n 1i a[ogrtr · 2. the gerade rydberg states of ne2 have been observed by laser excitation...

UNCLASSIFIED

:PORT DOCUMENTATION PAGEA D -A 247 293 - lb. RESTRICTIVE MARKINGS

1 1111 Ill 111111 11111 3.I T DISTRIBUTION/AVAILABILITY OF REPORT

* DECLASSIFICATIONIOOWNGRA EfR 0 6 1992 Distribution Unlimited4. PERFORMING ORGANIZATION RE % MBER(S) S.MONITORING ORGANIZATION REPORT NUMBER(S)

n 1i A[OgR*TR* 9 2 0 11461L NAME OF PERFORMING ORGANIZATION . FICE SYMBOL 7a. NAME OF MONITORING ORGANIZATION

University of Illinois (I a-6. ADDRESS (Ciy, State, and ZIP Code) . ADDRESS (Cily, State, and ZIP Code)

Department of Electrical and Computer Engineering1406 W. Green St.Urbana, IL 61801 ___ &_ _ _,__ _ __-

Sa. NAME OF FUNDING/SPONSORING '8b. OFFICE SYMBOL 9. PROCUREMENT INSTRUMENT IDENTIFICATION NUMBERORGANIZATION (It oplmce) Purchase Request No. F08671-9000284

AFOSR/# J L(I. AFOSR-89-00388C. ADDRESS (Cty, State, and ZIP Code) 10. SOURCE OF FUNDING NUMBERS

R IOJ ITASK 'WORK UNITBuilding 410 Eu. iT~.ELEETNO. NO. ACCESSION NO.

Boiling AFB, DC 20332-6448 ile If ,J, 2301 Al11- TITLE (Incude Secwitly Classiflcanmon)

VUV AND UV SOURCES AND SPECTROSCOPIC APPLICATIONS

12. PERSONAL AUTHOR(S)

J. G. Eden13a. TYPE OF REPORT 113b. TIME COVERED 114. DATE OF REPORT (Year, Monti, Day) i. PAGE COUINT

Final Technical FROM 11-1-89 TO 10-31-91 December 1991 14916. SUPPLEMENTARY NOTATION

17. COSATI CODES 18. SUBJECT TERMS (Continue on reverse if necessary and identify by block number)

FIELD GROUP SUB-GROUP laser, shortwavelength, energy storage, compound semiconductors,photochemical deposition

19 ABSTRACT (Continue on reverse it necessary and identity by block number)Extensive experimental studies of small molecules as potential "hosts" for short wavelength ( X < 200 nm) energy-storage

lasers have been conducted. Laser spectroscopy of the rare gas dimers and Group IS dimers have yielded structuralconstants for a significant number of previously unobserved states. Also, Fano "windows" have been observed in Ne 2 in theenergy region lying between v+

- 0 and v+. 3 of the Ne;2ground state (X 2 .). Bound - free emission studies of Cd 2 and Zn2have been carried out and the structural constants for the 1 . upper states have been determined.

The growth of GaAs and GaN by photo-assisted MOCVD has also been demonstrated. Gallium arsenide has been grownepitaxially at temperatures as low as 450"C by illuminating the surface with ArF (193 nm) radiation. Also, polycrystalline GaN(preferentially oriented (100)) has been grown on GaAs and sapphire at temperatures as low as 700"C and at growth ratesexceeding 2 Amlhour.

20. DISTRIBUTIONAVAILABILITY OF ABSTRACT 21. ABSTRACT SECURITY CLASSIFICATION

3 UNCLASSFIEDONLIMITED 0 SAME ASIRP Q" OTC USERS Unclassified22a. NAME OF RESPONSII NDID ,1. . 22c. OFFICE TEPLHN C

DO FORM 1473, 84 MAR 83 PR editonuy be used untl eshhted. SFCURITY CLASSJFrCATION OF 'HiS PGFAll other edtions are obeloie. UNCLASS;FIED

This document hes been OppToved

for public teleose ond sole; its

di_tibution is unlimited.

FINAL TECHNICAL REPORT FOR

AFOSR GRANT 89-0038:

"VUV AND UV SOURCES AND SPECTROSCOPIC APPLICATIONS"

NSP.CT( Accesion For

NTIS CRA&fDTIC 1AB []U. .a,.'ou;.:ed U3

Prepared for JJtI catio.

Dr. Howard R. SchlossbergPhysics D irectorate By ..............................................

AFOSR Dit b:tio-i/Boiling Air Force Base

Washington, DC Av.ia. it, c,-:e'

Dist Alv", (*.a

by

J. G. EdenDepartment of Electrical and Computer Engineering

Everitt LaboratoryUniversity of IllinoisUrbana, IL 61801

December 1991

Distribution Unlimited 92-05467

92 3 03 199 l39UUUU

2

I. INTRODU%. IION1

This AFOSR-supported program has been directed over the past 2 - years primarily towards

identifying new sources of coherent radiation with emphasis on the vacuum UV, UV and low end

of the visible (. < 500 nm). The motivation for this effort is provided by the tantalizing

applications in materials processing, (such as in situ photochemical processing), holography and

surface analysis that require short wavelength lasers that do not exist. A good example is the VUV

spectral region where tunable, high peak power sources are simply not available.

The work carried out under this grant has been multi-faceted and the following accomplish-

ments realized in the past 1 j years alone:

1. Photoexcitation experiments of the 11+ states of Zn2 and Cd2 have been completed and the

resulting bound--free emission spectra have been analyzed. Those molecules were

identified in joint theoretical studies with a Soviet physicist, Dr. V. Kochelap, as good

candidates for an anti-Stokes Raman laser operating in the deep UV or VUV. The results

have been published in the Journal of Chemical Physics.

2. The gerade Rydberg states of Ne2 have been observed by laser excitation spectroscopy and

the results have yielded precise structural constants for the molecule. More importantly, we

have observed Fano windows in Ne2 which arise from a Rydberg series of super-excited

31-g states approaching the v+ = 1, 2 and 3 states of Ne4. This is an exciting result

because we have learned a great deal about coupling of the continuum to these embedded

states.

3. Experiments in which an AlGaAs diode laser has been pulse-amplified in a system of four

dye amplifiers have been nearly completed. After frequency doubling the near IR output

beam, -100 pJ of narrow linewidth radiation is obtained at 390 nm. We are presently

testing this source by optogalvanic spectroscopy on Fe. The results are described in the

attached thesis that was prepared by Ann Farkas for her Master's degree.

3

4. The photochemical vapor deposition of GaN at 700"C has been accomplished. Although

the films are presently polycrystalline, they are preferentially-oriented (100) and grow at

rates exceeding 2 pm/hour.

In addition to these results, a book chapter on photochemical vapor deposition has been

published as a portion of Werner Kern and John Vossen's book entitled Thin Film Processes II

(Academic Press, 1991). This chapter has recently been expanded upon and will be published as a

monograph under the title Photochemical Vapor Deposition. It will be published by John Wiley in

1992. Also, the U. S. Patent Office awarded a patent to the Air Force this year for work done

under this grant. A copy of that patent is attached. The next section describes several of these

results in more detail.

II. RESULTS

A major component of this research program has been the investigation of the excited state

structure of the rare gas dimers and, more recently, the Column IIB metal dimers. The rationale

for this effort stems basically from the fact that the well-known energy storage capability of the rare

gas dimers has never been fully exploited. All of the excited states of the rare gas dimers are

Rydberg and lie more than halfway from ground to the first ionization limit. The lowest excited

states of these homonuclear molecules are essentially metastable. For the lighter species (He2 and

Ne 2), Hund's case (b) provides a good description of the excited states and the lowest one is

denoted 37+ which is, of course, not optically connected to ground (17+) by dipole transitions.

As the mass of the dimers increases, however, spin-orbit coupling gradually reduces the lifetime of

this state as Hund's case (c) notation of the state-l u, 0j-becomes appropriate.

In any event, the ability of these metastable molecules to store energy makes them attractive

candidates as amplifiers or anti-Stokes Raman media. Also, since Te for these states lie at >8 eV,

one can immediately envision their potential utility for the VUV and, perhaps, XUV as well.

The motivation, then, for this program was to develop a data base for these molecules that

would allow one to subsequently design the types of devices cited above. The paucity of data

4

available regarding the Rydberg states of all the rare gas dimers (except He2) prompted us to

conduct the experiments discussed in the next few sections.

Laser Excitation of Spectroscopy of Ne 2 and Ar 2The long radiative lifetimes (> 1 pisec) of their ns3 + metastable levels (n - 2-4 for He2, Ne 2

and Ar2 , respectively), and the high energies ('a 9 eV) and large rotational constants ( e 0.15 cm - 1)

for their Rydberg states make Ar2 and Ne2 , in particular, excellent possibilities for the applications

mentioned earlier. In contrast to He2 , unfortunately, little was known of the Ar2 or Ne2 states

lying above the lowest levels (correlated with ns(3/212 + 1So) because virtually all of their Rydberg

states are strongly predissociated. To circumvent this difficulty, we developed a simple

spectroscopic approach that is actually dependen upon the predissociation of the higher Rydberg.

Figure 1 shows the dramatic improvement in resolution that is afforded by this experimental

approach as opposed to conventional absorption methods. The upper spectrum is one that was

acquired in the present experiments whereas the lower profile was that obtained by absorption in an

electron-beam-pumped plasma in 7600 Torr of Ar. Spectra such as that in Figure 1 have made it

possible to extract a tremendous amount of information concerning the structure and kinetics of the

excited states of Ar2 and Ne2. The details of the analysis of the Ar2 excitation spectra will not be

described here but Table I, which summarizes the characteristics of a number of Ne2 Rydberg

states, conveys some idea of the wealth of information to which one has access. The attached

reprints describe some of our results obtained to date. The significance of this is that, for the first

time, one can design (rather than speculate about) a non-linear optical scheme for extracting energy

from the lowest ns37+u metastable states of the rare gas dimers.

Also, in Ne 2, vibrational autoionization of np3Il g Rydberg states (n > 15) that converge to

v+ = 1 of the Ne4 ground state has been observed. As indicated in Figure 2, these states are

observed as "windows" in a predissociative continuum lying above the Ne4 (v+ = 0) limit. Similar

series converging to the Ne v+ = 2 and 3 limits are also observed. The intriguing aspect of this

spectrum is the sharpness of the autoionizing resonances which implies that the vibrational

12npir,,0, 4sa USg~ 150 Torr Ar

8 n=6

0

0.4

0.01330 .570 410 450

Wavelength (nm)

Fig. I Comparison of Ar2 spectrum obtained by the LJF method with that recorded by Kileenusing a classical absorption method: (a) the Ar* excitation spectrum determined in thiswork by an LIF method; (b) a low resolution version obtained by Killeen et al. (J. Chem.Phys. 3A, 6048 (1986)) in absorption.

TABLE I. Effective molecular parameters (cm-1 ) and quantum numbers for the experimentally

characterized np. 3Ag (v = 0) states of 20Ne2.

SCate E(N-0) n Bo x 10 Do x 105 H3 x 108

6pa 3z 2 3 5 10 b 3.175

4pir 3fl 23972.80±0.08V 3.2442 5.743±0.008c 0.9±0.5 c 1.2±0.8c

7pa 3E 2 8 0 8 0 b 4.167

5pX 311t 28324.5±0.1 c 4.250 5.758±0.010c 0.5±0.4 c 0.1±0.2 c

8pa 3Z+ 3 0 2 9 4 b 5.162

6pc 31T+5 30424.6±0.2c 5.255 5.82±0.02c 0.9±0.5c 0.5±0.3c9pc 3Z~g 31526b 6.180

7px 3r1 31603,3±0.20 6.256 5.89±0.02c 0.9±0.5c 0.2±0.2c

10pa 3Z+ 3 2 2 7 7 .8±0.2 d 7 .1 9 3d 5.4 0±0. 0 2d -2.3±2. 0 d 4 .0±4 .5d

8pr 311 32320.2±0.20 7.266 5.98±0.02c 2.2±0.70 1.1±0.5cllp 3E3 32747+1 8.15 5.0±.20 N.D. N.D.

9pM 311l 32796.1±0.2c 8.274 5.77±0.02c 5.2±0.02c 0.1±0.20

12pa 3Z 33101.0±0.2 9.195 4.90±0.08 -15.±12. -18.±17.

1OPr 3r1; 33125.7±0.2 9.283 6.10±0.02 5.7±0.7 2.0±1.5

13pa 3-r 33350.8±0.3 10.232 4.85±0.10 -16±13 -15.0±15.0

lipir 311 33363.9±0.3 10.296 6.32±0.10 7.6±2.0 3.3±3.0

14pa 3E+ 33520.0±0.3 11.173 5.30±0.10 -2.0±2.5 -0.5±1.0

12p7r 3n f f f f f15per 3zE -33667.7±0.4 12.250 4.90±0.10 -4.0±4.5 -6.0±5.0

13pr 3fl+ 33677.8±0.4 12.335 6.30±0.15 8.0±3.0 N.D.

16pa 3r 33772.8±0.4 13.238 4.90±0.15 -16.0±14.0 -16.±8.0

14 pxc 3 it 33785.0±0.3 13.369 6.35±0.15 6.8±4.0 N.D.

17pa 3, 33853.3±0.4 14.181 4.80±0.20 -23.0±22.0 -30.±20.0

1Sp r 3r 33867.2±0.39 14.3658 6.37±0.105 13.0±10.09 N.D.

18pa 31: h h h h h

16pm 3T8 33938.4±0.3 15.435 6.30±0.15 N.D. N.D.

19pa 3E 33987.0±0.3 16.320 4.50±0.20 -25.0±20.0 -20.0±15.0

17p7r 3T$ 33993.4±0.2 16.448 6.60±0.20 20.0±20 N.D.

20pa 3Z 34030.9±0.2 17.266 4.50±0.30 N.D. N.D.

21pa 3E 34075.0±0.4 18.404 4.20±.60 N.D. N.D.

22pa 3Z 34099.0±0.3 19.125 4.50±0.30 N.D. N.D.

23pa 3E 34133.1±0.3 20.315 4.60±0.30 N.D. N.D.

00N

4-C

o >o

(00

C-,

OCO

A t200rn .2)

.00

Z

N

044

N CO,

N~lid]OS9 13AVCI.

5

autoionization lifetimes of these states exceeds several picoseconds, a property which may be

useful in a nonlinear mixing or wavelength conversion scheme. A paper describing these results

will be published in Physical Review Letters in early 1992.

Bound -+ Free Emission in Cd 2 and Zn 2

Another set of diatomic molecules that are potential energy storage systems are the Group UB

dimers Cd2 , Zn2 and Hg2 . Although the use of, for example, Hg2 as a laser is certainly not new,

the almost complete absence of information concerning all but the lowest-lying excited states of the

molecule has thus far prevented the design of the type of energy extraction scheme described earlier

for the rare gas dimers. In an effort to explore the structure of the Group IB dimers in more

detail, a number of emission spectra from the BIT + state of the dimers have been acquired by

photoassociating pairs of ground state atoms at various wavelengths in the ultraviolet. The

attached reprint, which was published in a recent issue of the Journal of Chemical Physics,

describes the accuracy with which excited state potentials can be constructed in this way.

Photochemical Growth of Wide Bandgap Ill-V Semiconductor Films

During this three year grant period we have also been pursuing the photo-CVD growth of III-V

films and the attached reprints describe the first results obtained on GaAs through a collaborative

program with Professor J. Coleman. In the past year, we have concentrated on GaN and have

successfully grown thin films at rates exceeding 2 gm/hour at 700"C. These films were grown on

GaAs (100) substrates and are polycrystalline but are preferentially oriented (100). The primary

impurity in the film is oxygen which appears to be originating from the Ga precursor (TMG). The

ease with which these films can be grown is encouraging and our plans for the immediate future

center on obtaining epitaxial films on sapphire and then studying the control on the electrical

properties of GaN that photo-CVD offers by substitutional doping.

6

APPENDIX:

RECENT REPRINTS PUBLISHED UNDER

AFOSR GRANT 89-0038

THIN FILM PROCESSES II

"'1-3

Photochemical Vapor Deposition

J. G. EDENUniversy of litnoiEventt ioratoryDepartnww of Electrical and Computer EngineeringUrbana, Iffin "

I. Introduction 44411. Fundamental Aspects of Photochemical Vapor Deposition 445

A. Photothermal and Photochemical Processes 445B. Overview of Photochemical Processes 449C. Adlayer Characteristics 462

Ill. Reactors, Optical Sources, and Associated Equipment 463A. Reactor Characteristics 463B. Optical Beam Delivery 469C. Optical Sources 471

IV Metal Films 473A. Column JIB Metals (Cd. Zn) 473B. Column IliA Metals (Al, Ga. In. TI) 474C. Column IVA, VIA, and VIIA Transition Metals

(Cr, Mo. W. Ti, Mn) 475D. Column IVB Metals (Sn. Pb) 476E. Column VIII Transition Metals (Fe. Pt, Ir) 477F Group IB (Cu. Au) 478

V Semiconductors 479A. Elemental Films 479B. Il-VI Binary and Ternary Compounds 484C. III-V Compound Films 487

VI. Dielectrics 490A. Aluminum Oxide 490B. Silicon Dioxide 491C. Silicon Nitnde 493D. Other Oxides and Nitrides 494

VII. Conclusions 495Acknowledgments 496References 4%

443CopVnght (r I99I hy Acadcm.1 Press. Inc

All rilhits of reproduction in in form rcscrcd

ISBN 0- 12.728251-3

444 J. G. EDEN

I. INTRODUCTION

Although the deposition of films by gas phase photochemical reactionswas first reported in the scientific literature more than 50 years ago [1. 21.only since the late 1970s has the application of laser and lamp radiation tothe growth of high-quality elemental and compound films been pursuedvigorously. The expansion of the variety of materials that can be depositedby photochemical processes is truly remarkable; to date, photo-assistedchemical vapor deposition (photo-CVD) encompasses 24 elements of theperiodic chart, in addition to at least 20 insulator and semiconductorcompound films.

Interest in photo-CVD partially stems from the ability of optical radia-tion to induce specific chemical reactions in the gas phase or at a surface.The selective, optical production of atoms, molecular radicals, or evenexcited species in the vicinity of a surface, and the ability to do soindependently of the substrate temperature, effectively decouples temper-ature from the number density of the species of interest. Said another way.the introduction of photons into a thin film deposition reactor allows one todrive the chemical environment far from equilibrium by sclectively pro-ducing species that are not normally present in significant concentrations inconventional CVD, MBE. or MOCVD reactors. This flexibility inherentwith the photodeposition of films often permits operation at lower temper-atures, where impurity redistribution and thermally induced mechanicalstress are minimized. Other attractive aspects of photo-CVD include thecapability for selective area deposition and, in general, the absence of iondamage to the film. In short, the spatial resolution, chemical specificity.and reduced temperature capability of photodeposition are among thefactors driving the current efforts in this field.

Several excellent reviews of photo-assisted deposition of semiconduc-tor, metal, and insulator films have appeared in the literature. Early workin this field was summarized by Osgood [31 and, over the past five years, ahalf-dozen journal articles [4-8] and several chapters or entire books [9-131devoted to reviewing the fundamental chemical and physical aspects of lasermicrochemical processing have been published.

It is the goal of this chapter to provide a broad overview of thephotodeposition methods and materials that have been reported to date.Emphasis is placed on the practical aspects of both laser and ultraviolet(UV) lamp photodeposition techniques for films of interest to microelec-tronics; for more detailed discussions of reactor photochemistry. precursorchemical structure, or the metallurgical quality of the films, the reader isreferred to selected references or to one of the reviews mentioned above.Because of their similarity to conventional CVD techniques, those photo-deposition processes that are primarily thermal in nature are not treated in

ni-3. PHOTOCHEMICAL VAPOR DEPOSITION 445

this chapter. Such photothermal processes rely on the spatial properties oflaser radiation to locally heat a substrate, and the reaction zone is laterallyconfined by the substrate's thermal conductivity and the nonlinear depen-dencies of reaction rates on temperature. Deposition processes involvingthe multiphoton absorption of infrared radiation to thermally dissociatepolyatomic molecules are also not discussed here. Rather, the applicationof visible or shorter-wavelength radiation to the deposition of films fromthe gas phase is stressed.

Photo-CVD is only one member of an ever-growing class of processes,including those fundamentally photochemical or thermal in nature, thatare now accessible with optical radiation. Laser-assisted doping, for exam-ple, is only briefly discussed here in conjunction with specific epitaxialsemiconductor films and the rapidly expanding and fertile field of laser-induced etching of films is reviewed in detail in Chapter V.3 by CarolAshby. Another closely allied processing subfield is that of selective areaprocessing. Frequently combining photochemical and thermal mecha-nisms, and often drawing upon physical processes identical to those under-lying broad-area deposition techniques, this area of film processing isdescribed by Thomas Mayer and Susan Allen in Chapter IV.3.

II. FUNDAMENTAL ASPECTS OF PHOTOCHEMICAL VAPOR DEPOSITION

A. Photothermal and Photochemical Processes

Photo-assisted film deposition methods and laser-assisted depositiontechniques, in particular, may be broadly subdivided into one of severalcategories, depending upon which of the three media (or combinationthereof) present in a reactor absorbs strongly at the laser wavelength. Asillustrated in Fig. 1, the substrate is normally immersed in a gas that givesrise to the formation of an adlayer at the surface of the substrate [141. Atthe core of all photo-assisted deposition processes is the gas, vapor, ormixture of gases from which we wish to extract the element(s) of interest.Known as the precursor or "parent," the gas is generally molecular and ischosen for its combination of thermochemical and spectral properties. Allphotodeposition methods, then, can be qualitatively distinguished fromone another according to the manner in which the precursor is decomposedto yield the desired film.

Depending upon the wavelength and intensity of the optical source,either the gas, adlayer, or substrate will be the predominant absorber thathas a critical bearing on the physical characteristics and growth rate of theresulting film. The first question to be considered when attempting tocategorize photo-assisted deposition processes is: How is the absorbed

446 J. G. EDEN

Molecular adlayer

Subst

Laser beam

Gas

Fig. 1. Diagram illustrating the three media-gas, adlayer, and substrate-that aregenerally present in a photodeposition reactor. The optical and chemical properties of thesethree play a key role in determining the nature of the reactions that culminate in film growthand in the metallurgical and electrical properties of the deposit. Although the laser is shownirradiating the substrate, films are often grown with the optical beam directed parallel to thesurface (adapted from Ref. 14).

energy disposed of? If the optical source energy is immediately convertedprimarily into heat, then the film growth process is said to be photother-mal. Often referred to as laser chemical vapor deposition (LCVD), thisapproach is based on the laser providing a spatially localized source of heatat the substrate that decomposes (pyrolyzes) the precursor molecules oncethe surface temperature exceeds a critical value that varies from moleculeto molecule. In this case, the gas in the reactor is often transparent at thelaser wavelength and, except for the heat source, LCVD is otherwisesimilar to conventional chemical vapor deposition (CVD). As is discussedin more detail in Chapter III.1, LCVD places few constraints on thewavelength of the laser, and the lateral extent of film growth in the plane ofthe substrate's surface is determined not only by the cross-sectionalgeometry of the laser beam itself, but also by the thermal conductivity otthe substrate and the nonlinear variation of the pyrolytic reaction rateconstant with surface temperature. A second variation of LCVD dependson the gas of interest absorbing the laser radiation in such a way that theprecursor molecules are not dissociated optically but are heated and even-tually fragmented by thermal processes. Whether the substrate or gas (orboth) is heated by the optical source, the results are similar and arecategorized as photothermal.

Photochemical vapor deposition, on the other hand, owes its existenceto the fact that UV and visible photon energies (- 2-6 eV) are often

w-3. PHOTOCHEMICAL VAPOR DEPOSITION 447

r --------Laser evaporationL--------

Laser chemical vapor depoitin ------------

L j

Laser-enhanced electroplating L------ ---

Fig. 2. Schematic representations of the various methods available for the laser-assisteddeposition of films. The darkened rectangles represent substrates that absorb strongly at thelaser wavelength, whereas the open rectangles are transparent materials. Note that laser-enhanced electroplating occurs in the presence of a liquid. This chapter is directed towardsthe third area, laser (or lamp) photodeposition of thin films (adapted from Ref. 15).

sufficiently large that chemical bonds in polyatomic molecules can beruptured by the absorption of one or perhaps two photons. A wide range ofphoto-assisted deposition techniques have been demonstrated, and, bear-ing the above principles in mind, most are readily classified as shownschematically in Fig. 2. In this diagram, substrates that absorb strongly atthe laser wavelength are denoted by solid rectangles, while optically trans-parent materials are represented by open rectangles. Laser evaporationconsists of rapidly heating a target with a pulsed laser, which results in theablation of surface molecules and their subsequent deposition onto anearby substrate. In this case, the gas (if present) is nonreactive, andfrequently the process is carried out in vacuum. A reactive ambient gas isessential for laser CVD or photochemical deposition, however, and theprocesses that culminate in film growth can occur either at the substratesurface or in the gas phase. The remainder of this chapter concentrates onfilm growth processes based predominantly on photochemical interactionswith molecules in the gas phase or in the adlayer.

The reactor configurations generally used in photo-CVD can, as de-picted in Fig. 2, involve an absorbing or optically transparent substrate,and it is not necessary for the optical beam to actually irradiate the surface.Both the so-called perpendicular (or normal incidence) and parallel

448 J. G. EDEN

Gas

WindowReactor

Laser -

Active medium - "

Vacuum system

(a) Perpendicular

indow Gas Reactor

Laser --

Substrate -7

Vacuum system Window(b) Parallel

Fig. 3. (a) Perpendicular (or normal incidence) and (b) parallel geometry configurationstypically used in laser photodeposition. Because of their lower intensities, ultraviolet lampscan irradiate the substrate (configuration (a)) without inducing a significant rise in surfacetemperature.

geometries common to photochemical vapor deposition are illustrated inmore detail in Fig. 3. As its name implies, perpendicular geometry in-volves irradiating the substrate directly, and the critical photochemicalmechanisms in this configuration generally occur in the adlayer. WhileFig. 3 illustrates a laser as the optical source, a large fraction of photode-position processes yield excellent results with ultraviolet or vacuum ultra-violet (VUV) lamps, and because of their lower instantaneous output

11I-3. PHOTOCHEMICAL VAPOR DEPOSITION 449

powers, one is able to directly irradiate the substrate as in Fig. 3a withoutconcern for significantly heating the surface by the optical source. Inparallel geometry, the optical beam does not impinge on the substrate, butrather produces atomic or molecular species that migrate to the substrateand react with the surface and adlayer, resulting in the desired film. In bothcases (perpendicular or parallel geometry), the "active" medium of Fig. 3is, in fact, initially inactive in the sense that no film is grown until theoptical radiation photochemically alters the molecules (in the adlayer orgas phase) into a more chemically reactive form. One of the attractivefeatures of photochemical deposition, then, is the control that it offers-spatial and temporal-since, in the absence of the external radiationprovided by the lamp or laser, the film growth rate is negligible or zero.

B. Overview of Photochemical Processes

Photochemistry in the gas and liquid phases is a well-established sci-ence, and considerable effort has been devoted over the last half-centuryto determining the products formed when small molecules absorb avisible, ultraviolet or vacuum ultraviolet photon (photon energy, hv--=2- 10 eV; bond-breaking with radiation of longer wavelengths [infrared]typically requires the absorption of more than one photon). Most of thisinformation has been obtained from gas phase experiments, and consider-ably less is known of the interactions of optical radiation with molecularadlayers.

The absorption of optical radiation by a column of gas of numberdensity N (expressed in cm - 3) and path length L is, at low intensities,governed by the relation known as the Beer-Lambert law:

It=Iie- NGL, (2.1)

where 1i and 1h are the optical intensities (W-cm- 2) incident upon andtransmitted by the gaseous medium, respectively, and o- is the absorptioncross-section (generally expressed in cm2) for the gas in question. Equation(2.1) is valid only when the precursor molecule absorbs a single photonfrom the optical source beam; multiphoton absorption rates, which varynonlinearly with the source intensity I and are usually significant only forI -Z 105-106 W-cm- 2, will be discussed later. Consequently, the fraction ofthe laser or lamp beam energy absorbed by the gas hinges upon the valueof a, at the source wavelength. For example, given a precursor partialpressure of I torr and an absorption cross-section of 10 -18 cm2. 3% of theoptical source energy will be absorbed over a 1-cm path length. Ifa= 10- 17 cm2 at the source wavelength, the absorbed power rises to 27%.Since the absorption cross-section is a function of the source wavelength,

450 1. G. EDEN

! I I I I I I I I I

2

.2

0110 130 150 170 190 210

Wavelength (nm)

Fig. 4. Disilane (Si2H6 ) absorption spectrum in the VUV (110 Z Az 210 nm) (adaptedfrom Ref. 16).

significant effort has been directed over the past decade, in particular,toward measuring the visible UV and VUV absorption spectra for volatileprecursors containing metal or semiconductor atoms.

Essential properties for any molecule to be a viable precursor forphoto-CVD are that: (1) it must be volatile (i.e., vapor pressure exceedingtens of millitorr at temperatures <100*C) in order to transport the desiredspecies to the deposition chamber; and (2) the resulting polyatomic mole-cule must absorb optical radiation in a portion of the spectrum whereefficient optical sources (lamps or lasers) exist. While most of the work todate in photo-CVD has involved precursors that were previously de-veloped for thermal (chemical vapor) deposition techniques (MOCVD,CBE, etc.), future efforts will undoubtedly focus increasingly on thesynthesis and engineering of new precursors.

Figure 4 shows the absorption spectrum of disilane (Si2H6), one of themost widely used precursors for the photodeposition of silicon. Note thatabsorption peaks in the vacuum ultraviolet (wavelengths below 200 nm orphoton energies >6.2 eV), which presents difficult decisions when choos-ing an optical source from the lamps and lasers that are commerciallyavailable (cf. Section III.C). For the acetylacetonates (known as acacs) ofplatinum and iridium, however, absorption maxima (cf. Fig. 5) lie in thedeep UV, which are conveniently accessible with both lamps and lasersources. Absorption coefficients at specific wavelengths, as well as entire

11-3. PHOTOCHEMICAL VAPOR DEPOSITION 451

2.0 I I1.0

1.60.8

~120.6

20.8 <~< 0.4

0.4 0.2

0.01 1 0.0 I I I I

200 300 400 500 200 300 400 500 600Wavelength (nm) Wavelength (nm)

(a) (b)

Fig. 5. Vapor phase absorption spectra of the acetylacetonates (acacs) of (a) platinumand (b) iridium in the UV and visible (reprinted from Ref. 17, by permission).

absorption spectra for several other molecular precursors of interest tophotochemical vapor deposition, can be found in Refs. 18 and 19.

Although o, describes the degree to which power is absorbed by a givengas, it says nothing of the manner in which the energy is disposed. One ofthe factors contributing most to the complexity (and breadth!) of photo-CVD is that the atomic and molecular fragments from a particular precur-sor are not only dependent upon the choice of parent molecule itself, butare also a function of the wavelength of the source. Furthermore, althoughabsorption cross-sections are generally straightforward to measure, theproducts and their relative yields are difficult to obtain, and for many ofthe photo-assisted deposition processes reported to date, are not known.The remainder of this section briefly discusses the most common photo-chemical processes underlying photo-assisted deposition- photodisso-ciation, photoionization, optically driven secondary (collisional) reactions,and multiphoton processes.

1. Photodissociation

Far and away the most common photochemical process invoked inphotodeposition is photodissociation, which involves the absorption of oneor more photons by a molecule, which ultimately results in the scission of achemical bond. Photodissociation of a polyatomic molecule ABC can be

452 J. G. EDEN

expressed as

ABC+nhy- A+BC. (2.2)

where A is the desired atom or molecular radical, BC represents theremaining molecular ligands, and n is an integer (generally, n = 1). Asmentioned earlier, the identity of the photofragments and the distributionof energy (internal and translational) among the products is a functionof wavelength. Also, photodissociation often results in the rupture ofthe weakest bond in the molecule, which may yield, for example,

ABC + h-- AB + C (2.3)

rather than the desired products of reaction (2.2). Similarly, the fraction ofthe total absorbed photon energy that appears as translational energy(heat) of the fragments depends on the strength of the bond(s) broken, thenumber of bonds involved, and the laser wavelength.

While a wide variety of precursors have been explored for photo-assisted deposition, the most common are carbonyls, alkyls or hydrides,which are not only (of necessity) volatile but also widely available. Theseparent molecules are of the form AB, where m is also an integer andB = CO, CH 3 (or C2H5), or H. Halides have also been investigated, but toa lesser degree. Early photo-CVD experiments were largely empirical inthe sense that little was known of the photoproducts that one could expectat a given wavelength. Consequently, only the most generally features ofthe photochemistry could be inferred either from the known chemical andoptical characteristics of analogue molecules (such as C-I4 for SiH 4 ) orfrom the chemical composition of deposited films. Such a situation is notconducive to exploiting to the maximum the capabilities of gas phasephotochemistry for film deposition and its ability to produce specificproducts, in particular. Recently, however, several studies have been re-ported that have investigated in detail the products of photodissociation forseveral precursors of interest to photo-CVD. A few of these are reviewedhere.

a. Column IliA Alkyls

Not only are the Column liA alkyls such as trimethylaluminum,AI(CH) 3 : TMA, or triethylgallium, Ga(C 2H5 ) 3 : TEG, widely used in theepitaxial growth of III-V semiconductor films (such as AIGaAs) by metal-organic CVD (MOCVD), but they were also among the first precursorsinvestigated for use in photo-CVD processes. Since the photochemistry ofTMA has been most thoroughly studied and is representative of that forother members of the Column IliA alkyl family, trimethylgallium and

i-3. PHOTOCHEMICAL VAPOR DEPOSITION 453

10

(0.Torr) 0.8

8 e( To ff))U 0.

E

20.

cX4 . 4%

2 Pressure (Tof)

0- -

200 220 240 260

Wavelength (rim)

Fig. 6. Expanded -iiew of the TMA absorption spectrum between 200 and 260 nm. Asshown by the inset, the cross-section is dependent upon pressure because of the risingconcentration of the TMA dimer (AI 2(CH 3)6) at higher pressures. At room temperature, theTMA monomer (AI(CH 3 )3 ) relative number density falls from 33% at a pressure of 10-2 torrto 4% at I torr of total pressure. For wavelengths longer than -220 nm, only the monomerabsorbs significantly. (Reprinted with permission from Springer-Veriag, Th. Beuermann andM. Stuke, Appl. Phys. B 49, 145 (1989).

trimethylindium (TMI), the discussion will be limited to its properties.In the 190-255 nm spectral region, TMA is photodissociated by the

absorption of a single photon, giving rise to the absorption spectrum ofFig. 6, which shows the 200-260 nm region in detail. The primary productsare free Al atoms and AICH 3 radicals and, as illustrated in Figs. 7 and 8,the relative yields are strongly dependent upon laser wavelength [20]. Forwavelengths below 255 nm, Al atoms are produced by the two-stepprocess [20, 21].

Al(CH,)., + h,(:-4.9 eV)- AI(CH3)2 + CH.,, (2.4)

AI(CH3)' - Al + CH,. (2.5)

454 J. G. EDEN

4

Yields:

3 ---- AI

- - Al + AICH3

AICH3 - Absorption spectrum

Al

190 2 210 220 230 240 250 260 270Wavelength (nm)

Fig. 7. Relative yields of Al and AICH3 resulting from the photodissociation of TMA inthe 190-270 nm spectral region. The absorption spectrum for TMA is represented by the solidcurve. Note that for wavelengths greater than -220 nm, Al is the predominant product.(Reprinted with permission from Springer-Verlag, Th. Beuermann and M. Stuke, Appl.Phys. B 49, 145 (1989).

The threshold for AICH3 production was observed by Beuermann andStuke [20] to be approximately 230 nm. For this fragment, the productionprocess appears to be

AI(CH3 )3 + hv(z5.4 eV) - AICH3 + 2CH3 . (2.6)

The key point to be. made here is that to produce free aluminum atomsrequires an optical source wavelength between -230 and 255 nm. ForA > 220 nm, it was also found that only the TMA monomer, AI(CH 3)3 ,absorbs, and the dimer, A12(CH3)6 , contribution rises rapidly at shorterwavelengths. TMA, TMG, and TMI apparently are all photodissociated inthe same manner (201. The results of Refs. 20,22, and 23 explain the presenceof objectionable levels of carbon in Al films deposited by photodissociatingTMA at 193 nm and suggest that the best future course would be to use asource in the A -> 220-230 nm region-perhaps the KrCI laser at 222 nm.As will be discussed in Section IVB, experiments in which TMA was

iu-3. PHOTOCHEMICAL VAPOR DEPOSITION 455

0 (Al 4 AICH3

1.0 oL ' Al

0

0p 0

0.8Al

. 0.6-

M8 0.4- CH

0.2 -

0.0190 200 210 220 230 240 250 260 270

Wavelength (nm)

Fig. 8. Absolute quantum yields for Al and AICH 3 in the photodissociation of TMA(Reprinted with permission from Springer-Verlag, Th. Beuermann and M. Stuke. Appl.Phys. 8 49, 145. 1989). (0) 0,, + 0AICH3; (') OAICH3; (0) 0AI. The dominance of atomic

aluminum as a product at longer wavelengths is again clearly evident.

photodissociated at 248 nm (KrF laser) to deposit Al films having lowercarbon impurity concentrations support the conclusions of Beuermannet al. [20, 22, 23].

b. Column V Hydrides

The Column V hydrides NH 3 (ammonia). PH3 (phosphine), AsH 3(arsine), and SbH 3 (stibine) are also frequently employed in the MOCVDand chemical beam epitaxy (CBE) growth of III-V compound films. Allabsorb strongly in the VUV, with absorption maxima occurring in the180-200 nm region [18], [19]. Also, except for ammonia, the absorptionspectra for all of the Column V hydrides are continua. The NH 3 spectrumis oscillatory in the 180-200 nm region, and the 193 nm cross-section is(1.5 ±t 0.3) x 10-17 cm

2. Photodissociation of the hydrides at these photon

456 J. G. EDEN

energies results in th removal of a single hydrogen atom. For phosphine,for example [19],

PH2(A(2 A1 )) - HPH, +h--. hpHz(k(IB,))+H (2.7)

where only -1% of the PH2 radicals are produced in the excited (A)state-most are formed directly into the ground state. Similar statementscould be made for the other hydrides-for NH 3 , the yield into the NH 2(A) state is 2.5%, with the remainder produced once again in the X groundstate.

c. Metal Carbonyls

The carbonyls of Cr, W, Fe, Mo, Mn, and Ni absorb strongly below300 nm, and in most instances the dissociation of these polyatomics pro-ceeds by the sequential elimination of carbonyl ligands [191. Althoughthe energy required to remove one or two CO ligands is 1.6 and 3.3 eV,respectively, the photodissociation of chromium hexacarbonyl, for exam-ple, in the 351-355 nm region proceeds predominantly by the process:

Cr(CO)6 + Av(-3.5 eV) - Cr(CO), + CO- (2.8)

At 248 nm, while two photons have sufficient energy (10 eV) to remove allof the carbonyl ligands, a single 5 eV photon leads to the removal of twocarbonyls by the two-step sequence

Cr(CO)6 + h&(5 eV) - [Cr(CO)l" + CO -- Cr(CO) 4 + 2 CO. (2.9)

where the asterisk indicates that the pentacarbonyl species is unstable. Atlaser intensities exceeding several hundred kilowatts per square centi-meter, electronically excited metal atoms are produced as a result of twophoton absorption at 248 nm by the parent molecule.

2. Photoionization

Although the photodeposition processes developed to date almost uni-versally involve electrically neutral products, selective photoionization ofreactants presents several advantages for photo-CVD, insofar as processcontrol and impurity rejection and chemical specificity, in particular, areconcerned.

Upon absorbing a single photon of sufficiently high energy. the entiremolecule AB can be photoionized,

AB + h - (AB)' + e -. (2.10)


or the molecule may be dissociated in the process. yielding a single.positively charged ion and an electron,

AB + h,-A- + B + e-, (2.11)

or only a pair of ions,

AB + h-- A' + B-. (2.12)

The first of these processes (reaction (2.10)) is generally not useful inphotochemical vapor deposition, since the precursor ligands are usuallyrequired only as a vehicle for raising the vapor pressure of the parent. Oncethe precursor is present at the substrate, the ligands are no longer neces-sary (and are often a nuisance since they frequently contain undesirableatoms such as carbon), and the molecule must be dismantled so as to gainaccess to the desired atom. Occasionally, however, photoionization of theentire precursor molecule at short (i.e., VUV) wavelengths is an attractivedeposition process, particularly for hydrogenated films. Calloway et al.1241, for example, have shown that thin Al films can be deposited byphotoionizing TMA with rare-gas lamp resonance radiation. At these shortwavelengths, the entire molecule is apparently positively charged, leadingto incorporation of carbon (and H) into the film.

An excellent example of the latter reaction (Eq. (2.12)), however, is thedissociative photoionization of selected metal halides. In 1932, Terenin andPopov [25] found that certain metal-halide molecules, when irradiated withphotons lying within a narrow energy range, undergo dissociative photo-ionization, resulting in the production of only charged fragments-a metal cation and a halogen anion. For the diatomic metal halide thalliumiodide, for example,

TII+ - -n' + -. (2.13)

Ion pair production processes analogous to reaction (2.13) occur inother metal halides such as InI. ArF laser photons (h Y = 6.4 eV) havesufficient energy to directly produce ln*-I - ion pairs. and this processcompetes with the formation of electronically excited, neutral In atoms anda ground-state I atom. One of the attractive features of dissociative photo-ionization of metal halides insofar as photochemical vapor deposition isconcerned is that, by the careful choice of the precursor and laserwavelength, only one positively charged ion is produced (that of the metalatom), and the undesired species (halogen atom) is oppositely charged-consequently, with the aid of an electric field, one can minimize theincorporation of contaminants into the resulting film.

458 J. G. EDEN

a. Column V HydridesAlthough little is known of the details of the photochemistry of arsine

and the other Column V hydrides, Koplitz et al. [261 have demonstratedthat the photoproducts for AsH 3 can be readily switched from predomi-nantly neutral to ionic species with a "two-color" laser approach. As notedearlier, the absorption of a single 193-nm photon by AsH 3 predominantlyproduces AsH 2 radicals in the X (ground) of A (excited) electronic states.Upon adding k - 365 nm (3.4 eV) photons concurrently with the ArFradiation, however, strong ionization of the molecule and the formation ofAsH , specifically, is observed. Consequently, the AsH 3 fragments can bereadily manipulated with the aid of UV radiation, which acts in concertwith 6.4 eV photons to photoionize AsH 3 (A) molecules. This scheme canbe expressed as

AsH3 + hv 1(6.4 eV) - AsH2(X, ,A) + H (2.14)AsH 3 + hi1 + hv,(3.4 eV)-AsHZ + e- -' H (x 3) (2.15)

and is shown schematically in Fig. 9. Similar approaches are almost cer-tainly available for other precursors of interest to photochemical vapordeposition, but further fundamental studies of the photochemistry of thesemolecules will be necessary to assess the full scope of the chemical path-ways and subsequent process engineering that is possible.

3. Optically Driven Secondary Reactions

a. SensitizersOften, a mismatch exists between the wavelengths available from ex-

isting (and efficient) UV and VUV lasers (and lamps) and the spectralregions in which volatile, high-purity precursors absorb. A more subtledifficulty arises if, as discussed previously, the photodecomposition of agiven precursor at a particular optical source wavelength does occur rap-idly but yields undesirable products. In such situations, intermediatespecies. known as photosensitizers, may occasionally be employed to act asa bridge between the optical source and the precursor molecule in thesense that the photosensitizer species must absorb strongl'v at the opticalsource wavelength and efficiently transfer its internal energy to the poly-atomic precursor molecule.

Atomic mercury is the oldest and most thoroughly studied photosensi-tizer. as the decomposition of silane by Hg photosensitization was reportedmore than a half-century ago [21. Mercury is an effective photosensitizerfor the Column IV hydrides in that it couples Hg resonance lamps emittingat 253.7 nm (and 184.9 nm) with precursors such as SiH 4 and Si2 H 6 , whichare virtually transparent for wavelengths beyond 160 and 200 nm, respec-


16 -

-As*

14 ASH, +

12 AsH

1 0 - AsH3

83M5 nrn

4

w IIIIAsH,+ H

AsH3 (A )

193 nm

0 AsH 3(X)

Fig. 9. Generalized energy level diagram for AsH3 showing the switch between photodis-sociation (denoted pathway 1) and dissociative photoionization that is induced by simul-taneously irradiating the molecule with 6.4 and -3.4 eV photons (from Ref. 26).

tively. Consequently, in the absence of a photosensitizer, decomposingprecursors such as SiH 4 by single photon processes precludes the use of themost efficient, commercially available sources (such as the ArF and KrFexcimer lasers). Central to Hg photosensitization is the 6p 3P, electronicexcited state of the atom, which lies -4.9 eV above the ground state.Mercury 3P, atoms, produced by 254 nm radiation from a low-pressureresonance lamp, are well suited for collisionally exciting (and dissociating)various polyatomic molecules. A complete discussion of the photochemis-try of the Hg(6p3pt) species can be found in Okabe [181.

The collisional processes that are critical to the photochemical growthof amorphous hydrogenated Si films from the photosensitized decomposi-tion of SiH 4 are [27]

Hg + hv(4.9 eV) - Hg*. (2.16)Hg+ SiH4 - Hg + Sil -t H, (2.17)

460 J. G. EDEN

H + SiH4 - H2+ SiH 3, (2.18)

SiH 3 + SiH 3 - SiH 2 + Sii-I4 , (2.19)

SiH 3 + SiH 1 - SiH 3SiH + H., (2.20)

where the asterisk denotes the 6p 3P1 excited state of mercury. Althoughthe actual production of a-Si: H films appears to occur when SiH 2SiH 3 orSiH 3 SiH radicals impinge on the substrate, efforts to unravel the compli-cated gas phase chemistry of this photochemical system are continuing.Nevertheless, the chemical versatility of Hg photosensitization, combinedwith the economy and simplicity of Hg resonance lamps, makes this anattractive approach to the deposition of the Column IV elemental films, inparticular.

b. Bimolecular Collisional Formation of Compounds

Compound film growth by photochemical vapor deposition is, at pres-ent, a class of complex and poorly understood processes. The deposition ofthe compound material AxBj, (where x and y are often not integers) usu-ally requires molecular precursors for both elements A and B and is similarto photosensitized deposition in that generally only one of the two precur-sors absorbs strongly at the optical source wavelength. Upon absorbing aphoton, this precursor is either excited to a relatively stable excited state ordissociated. In either case, the products formed by irradiating the moleculeare more chemically reactive than the ground state species, and the ensuingcollisions of these photofragments with the remaining precursor eventuallyculminate in the production of the desired film.

An example of this process is the photochemical sequence responsiblefor the deposition of the insulator films Si3N4 and SiO 2. In both cases, thesilicon precursor is a silane (SiH 4, Si2H6, or Si 3H8 ), all of which, as notedearlier, are essentially transparent (nonabsorbing) in the UV. Interactionof the optical source occurs primarily with the N or 0 precursor, whosedissociation fragments chemically react with the silane to form the desiredfilm. In the photodeposition of SiO 2 , the oxygen precursor is frequentlyN20, which, when photodissociated at 193 nm, yields

N20 + hi, (-6.5 eV) -- N, + 0*('P). (2.21)

Following relaxation of the excited O('P) species to the ground (3p) state,compound formation apparently proceeds by the reaction:

SiH 4 + 40(3 p) - SiO 2 + 2H20. (2.22)

The photochemistry of Si 3N4 deposition is considerably more complicated,since it entails the photodissociation of SiH 4 (or Si2H 6)/NH3 mixtures inthe VUV and generally at 193 nm. Two of the early steps in the reaction

[n-3. PHOTOCHEMICAL VAPOR DEPOSITION 461

chain are expected to be

NH 3 + h-* NH 2(X, A) + H, (2.23)

H + SiH, - H, + SiH 3 , (2.18)

with the final deposit-producing reactions involving the interaction oflaser-generated radicals (such as amidogen, NH 2) with SiH 4 .

4. Multiphoton Processes

The photoreactions discussed to this point have generally assumed theabsorption of a single photon per precursor molecule as a necessary condi-tion for triggering the photodeposition process of interest. As we have seen,however, several parent molecules of interest (such as SiH 4) are "trans-parent" in spectral regions where efficient optical sources presently exist.While transparency is a notion that is a result of a small single photonabsorption cross-section, o, muliphoton processes often become impor-tant at higher optical intensities and can alter the situation entirely. Recall-ing, for example, the generalized photodissociation process, Eq. 2.2,

ABC +nhvy- products,

where n is again an integer, we denote the nth-order photoabsorptioncross-section as o&(n), and the rate R at which ABC molecules are photodis-sociated is given by perturbation theory as

R = o.n)( , (2.24)

where I is the optical intensity (ex ressed in W-cm -2 ) and a " ) is expressedin units of cm 2"-(s) - 1 [i.e., cmy for n = 1, cm'-s for n = 2, cm 6-s2 forn = 3... ]. Note the strong nonlinearity of the process for n > 1 (i.e., theabsorption of more than one photon by the precursor), which favors largelaser intensities.

As an example, silane is essentially transparent at 193 nm (h Vy= 6.4 eV)owing to a single photon absorption cross-section at that wavelength ofcr( ) = 1.1 x 10-21 cm 2 . However, the two-photon cross-section, o-(2), atthe same wavelength is (6 ±t 2) x 10-

4 cm 4-s [28], and since the ratio of thesingle photon absorption rate to that for two photons is (oa1 )/o1 ' 2 ))h1'/l,then the rates for these two processes are equal for laser intensities of only-19 kW-cm - 2 (or 0.2 mJ-cm- 2 for 10 ns laser pulses). This intensity iseasily exceeded with commercially available excimer lasers, which gener-ally produce unfocused intensities >4 MW-cm 2 per pulse. Consequently,two (or more) photon processes can dominate single photon dissociation ofa molecule and, in such cases, conventional absorption spectra are of little

462 J. G. EDEN

value. Because of the large optical intensities required, multiphoton pro-cesses are usually accessible only with pulsed laser sources.

In summary, whether the pivotal photochemical process is photodis-sociation, photoionization, or photosensitization, the role of the opticalsource in photochemical vapor deposition is to transform the chemicalenvironment in the vicinity of the substrate from a quiescent to a reactiveone, but to do so in a manner that discriminates against undesiredproducts.

C. Adlayer Characteristics

Photochemical interactions in adlayers of molecules at a surface under-lie many photochemical deposition processes and are particularly useful inthose applications for which spatial resolution is an important consider-ation. In the gas phase, the lateral diffusion of photoproducts imposesa limit on the smallest features sizes obtainable by photo-assisted filmdeposition, and while this can be suppressed to some extent by controllingthe identity and pressure of the background gas, the restricted mobility ofatomic and molecular species in an adlayer is much more conducive todepositing patterned films. Because of the proximity of a surface, theoptical and chemical properties of adsorbed layers of polyatomic moleculescan differ considerably from those in the gas phase. Qualitatively, adlayerscan be categorized as chemisorbed or physisorbed. For the former, theinteraction of the adlayer with the surface is comparable to the precursor'sintramolecular bond strengths. Physisorbed adlayers, in contrast, interactweakly with the surface, and their absorption spectra are generally similarto those for the gas phase species.

The characteristics of adsorbed layers of a variety of molecules havebeen studied extensively and, for a broad review of the subject, the readeris referred to the excellent discussion by Rothschild [19]. Considerably lessis known regarding adlayers of the polyatomic molecules that are suitableprecursors for photochemical vapor deposition. Of the few precursors thathave been studied, the Column IIIB alkyls are the most thoroughly char-acterized and are briefly discussed here. Measurements made by Sasakiet al. [29] of the gas phase and adlayer absorption spectra of trimethyl-gallium show that the chemisorbed spectrum is broader than its gas phasecounterpart, and that the physisorbed spectrum, relative to the gas phaseprofile, peaks at a shorter wavelength. Their results were also in accordwith previous studies that revealed an elongation of the chemisorbedlayer's absorption spectrum towards the red. One other aspect of chemi-sorbed layers that has a direct bearing on photochemical deposition is thatthe absorption cross-section for the adlayer is generally larger than that forthe vapor.

inI-3. PHOTOCHEMICAL VAPOR DEPOSITION 463

Not only does the adlayer-surface interaction manifest itself in abroadened absorption spectrum and increased absorption strengths, but italso alters the photochemistry of the precursor relative to that for the gasphase. Zhang and Stuke [301 investigated the photochemistry of metalalkyl adlayers by laser time-of-flight mass spectroscopy and showed thatphotodissociation of chemisorbed adlayers of TMA (on Si or quartz) at248 nrm yields much higher abundances of AICH 3 radicals than are mea-sured for the gas phase process. Furthermore, photodissociation ofadsorbed aluminum-bearing alkyls at 308 nm occurs quite readily, in con-trast with the gas phase species, which is transparent at wavelength [30].

One practical ramification of adlayer photolysis is prenucleation [311 orphotonucleation [321, a process in which laser photodeposition from achemisorbed adlayer controls the nucleation of metal films. That is, sur-faces that have small sticking coefficients for gas phase metal atoms, forexample, can be "seeded" by photolyzing metal-alkyl adlayers with afocused UV laser beam. The resulting metal atoms on the surface act asnucleation sites on which further deposition can occur. This process, whichovercomes nucleation barriers that often exist at surfaces, has been ex-ploited to deposit patterned metal films with excellent spatial resolution(<4&m feature widths).

III. REACTORS, OPTICAL SOURCES, AND ASSOCIATED EQUIPMENT

All photochemical vapor deposition systems include these elements:

(1) Gas flow equipment for delivering the desired precursor(s) to thesubstrate at a specified pressure and mass flow rate;

(2) A substrate and, in most cases, heated susceptor;(3) An optical source (lamp or laser) and beam delivery scheme; and(4) An exhaust system.

Aside from item (3), photochemical vapor deposition reactors are quitesimilar to conventional MOCVD or CBE systems. Consequently, theremainder of this section concentrates on the characteristics of commer-cially available optical sources and the trade-offs associated with differentapproaches for delivering the optical beam to the gas or adlayer im-mediately adjacent to the substrate surface.

A. Reactor Characteristics

Even though all reactors have several common components, Figs. 10,12, and 13 illustrate the wealth of experimental configurations that have

464 J. G. EDEN

Modifiedarclamp

H 2 + Cd(CH 3)2 >

H2 ,-. -- Substrate

~~~'Thermocouple

Graphitesusceptor

ExhaustH2 + (C2H)Te Quartz-halogen

lamp

Fig. 10. Schematic diagram of a photochemical vapor deposition reactor for the growth ofCdTe and HgTe epitaxial films. In this arrangement, the optical source is a 1,000 W Hg-Xearc lamp whose spectrum is modified by reflecting the lamp radiation off a dichroic mirror thatis highly reflecting in the 200-250 nm region (from Ref. 33).

been employed thus far for photo-assisted CVD. A simplified schematicdiagram of a reactor for the photo-CVD growth of CdTe (33] is given inFig. 10. The precursors (diethyl tellurium (DET) and dimethyl cadmium(DMC), in this case) are entrained in hydrogen and introduced upstreamof the deposition region. The substrate is mounted on a graphite susceptorheated by a quartz halogen lamp, and the unreacted gases (and gaseousbyproducts of the deposition process) are exhausted at the rear of thereactor. A 1,000 W Hg-Xe arc lamp provides the necessary UV radiationfor photodissociating the alkyl precursors, but Ahigren et al. [33] haveintroduced a clever modification to the system. By reflecting the arc lampbeam off of a dichroic mirror, radiation outside the 200-250 nm regionis rejected and only photons in the -5-6.2 eV range reach the surface.

In order to introduce the UV or VUV radiation from the source intothe reactor, an entrance window must be provided, and the materialrequired is dependent upon the wavelengths involved. Since the reactor ofFig. 10 utilizes wavelengths in the 200-250 nm region, the optical beam canbe propagated efficiently through air, and quartz is an acceptable windowmaterial. For wavelengths below -180-190 nm, air absorbs strongly (ow-ing to the Schumann-Runge bands of 02), and different crystalline mate-rials are required for reactor windows. The optical transmission curves for

[in-3. PHOTOCHEMICAL VAPOR DEPOSITION 465

100

.80" Li MgF2

.60

.2a 40 Sapphire

u CaF2 prasil Quartz

0 20 -

0 I100 120 140 160 180 200

Wavelength (nm)

Fig. 11. Transmission of readily available window materials in the VUV: LiF-1.95 mmthick; MgF2-1 mm thick; CaFz-0.25 mm thick; sapphire-0.32 mm thick; BaF2 -1.95 mm thick; Suprasil (enhanced UV transmission quartz)- 10 mm thick; and quartz-2 mm thick. Copyright 1965 by The American Institute of Physics.

readily available window materials in the VUV are shown in Fig. 11. Mostare fluoride crystals, and the material offering maximum transmission atthe shortest wavelengths, lithium fluoride, "cuts off" below -100 nm (50%transmission at -112 nm). One of the materials in Fig. 11 is generallymore than adequate for most photochemical deposition processes. Notethat not only are new materials required in this spectral range for transmit-ting and focusing the incoming radiation, but it is necessary that the entireoptical path be evacuated or purged with N2 gas. The latter is only effectivefor wavelengths down to 150-160 nm. Below -105 nm, no windows areavailable, and it is necessary to resort to a windowless configuration [34].Although this allows one to utilize short-wavelength optical sources, thedisadvantage of the windowless approach is that the plasma generating theVUV radiation is not physically isolated from the thin film precursor.

A detailed diagram of a commercially available, low-pressure photo-CVD reactor (Tystar PVD 1000) designed for the deposition of SiO 2,Si3N4 , and silicon oxynitride (SiON,) dielectric films is given in Fig. 12

[351. Based on the decomposition of mixtures of SiH 4 and N20 or NH 3 byHg photosensitization, this reactor includes two growth chambers that arecapable of independently processing nine 75-mm diameter wafers or five100-mm diameter wafers. The optical source is an array of low-pressure Hglamps or a single grid lamp, and the substrate is heated by four quartz

466 J. G. EDEN

4AMIiN.E GUAM .W

A fVM W.G•

Fig. 12. Layout of a commercial photo-CVD reactor designed for the deposition of silicondioxide and silicon nitride dielectric films by Hg photosensitization (reprinted by permissionof Tystar Corp., licensee of Hughes Aircraft Co.-1989).

lamps. Depending on the grade of fused quartz chosen for the window,transmission at 2,54 nrn (resonance line of Hg) ranges from -50 to 90%.Tables I and II give details of the reactor's operating characteristics andfilm growth process parameters for both SiO 2 and Si3N, films. AlthoughHg is incorporated into the resulting films, concentrations are in theparts-per-billion range.

Broad-area deposition of dielectric films with a laser is generally accom-plished with a parallel configuration such as that illustrated in Fig. 13.Avoiding direct irradiation of the substrate eliminates transient heating ofthe growing film by the laser. Because of the -1 cm2 cross-sectional areaand large divergence angle of most commercially available excimer laserbeams, it is generally necessary to compress the rectangular beam in onedimension and collimate it with a simple optical telescope prior to directingthe radiation into the reactor [361. Note also in Fig. 13 that provision is

mTm lmm m mmWlmmmm

II-3. PHOTOCHEMICAL VAPOR DEPOSITION 467

TABLE I

GENERAL CHARACTERISTICS OF A COMMERCIALLY AVAILABLE PHOTO-CVD

REACTOR (TYSTAR PVD 1000) BASED ON Ho PHOTOSENSITIZATION(AFTER REF. 35)

1. Recommended gas flow ratesSiO 2 : Silane (SiH): 2 sccm

Nitrous oxide: 60 sccmNitrogen (pump injection): 42 sccmTotal gas flow rate: 104 sccrnReactor chamber pressure: I torrSubstrate temperature: 150*C

Si3N,: Silane: 2 sccmAmmonia (NH3): 50 sccmNitrogen (pump injection): 42 sccmTotal gas flow rate: 94 sccmReactor chamber pressure: 1 torrSubstrate temperature: 15(0C

2. Mercury: Triple distilled, reagent grade: 500 g3. Substrate heater: Infrared quartz lamps,

4 x 375 W = 1500 W totalHeat-up time to 150*C: 10 min

TABLE IITYPICAL PROCESS PARAMETERS FOR THE DEPOSITION OF S102 AND

SI3N4 FILMS By HG PHOTOSENSITIZATION IN A COMMERCIAL REACTOR(AFTER REF. 35)

Film

Parameter Silicon dioxide Silicon nitride

Thickness uniformity:within wafer ±3% ± 10%across heater substrate ± 10% = 15%repeatability :s±3% S ±5%

Deposition temperature 50 to 200*C 100 to 200CDeposition pressure 0.3-1.0 torr 0.3-1.0 torr

(40-133 Pa) (40-133 Pa)Cycle time (150 nm) 45 min 55 minMaximum deposition rate 15 nm-min- 6.5 nm-min-'

468 J. G. EDEN

Surface reactionphotons

Lens

Window purge Ratn ae

O-ring seal

Volume reactionphotons meterHeaterL

Cylindrical-len~tlsce

& smu-bstrae To vacuum pump(normal or parallelphoton flux)

Fig. 13. Parallel geometry reactor for :he laser photochemical deposition of ZnO films(from Ref. 36).

made for a low-intensity beam, directed onto the substrate, to stimulatesurface reactions.

As discussed in Section II.B.2, greater selectivity in the photodeposi-tion process is available by photoionizing only the atom or molecularradical of interest. Such an approach was reported by Geohegan and Eden[371 to deposit thin films of aluminum, thallium, or indium by dissociativelyphotoionizing metal halide precursors to generate metal-halogen ion pairs(i.e., M+-X - , where M and X represent metal and halogen atoms, respec-tively). In their apparatus (cf. Fig. 14), the substrates were arranged on anegatively biased plate, and the metal cations produced in the gas phaseabove the substrate were drawn to the substrates under the influence of theelectric field.

Regardless of the specifics of reactor design, one problem to be dealtwith in all photo-CVD processes is the tendency for films to grow on theoptical window in addition to the substrate. A flow of "purge" gas (typi-cally H2) across the interior surface of the window will minimize or elimi-nate this problem-cf.Figs. 10 and 13. An oil coating on the windowor in situ etching of the window during photodeposition with XeF 2 ordischarge-generated F atoms has also been demonstrated successfully.

m11-3. PHOTOCHEMICAL VAPOR DEPOSITION 469

A rF laser B e am splitter e l supply

or Slt - Collection

Photodiode circuit

Oscilloscope Oscilloscope~f

Fig. 14. Schematic diagram of apparatus employed in Ref. 37 to deposit thin thallium,aluminum, or indium films by the photoproduction of metal (M) and halogen (X) ion pairs(M -X-).

B. Optical Beam DeliveryThe optimal manner for delivering the optical beam to the reactor is

dependent upon the desired application. Figure 15 qualitatively illustratesthe four most common beam delivery configurations [15]. The first twoparts of Fig. 15 (shown as parts (a) and (b)) are intended for situationsrequiring spatially selective deposition. The simplest approach, the spotfocus, more commonly known as the direct write configuration, entailsscanning a focused laser beam across the substrate. Deposition processesbased on this approach generally involve photochemical interactions withinthe adlayer. Assuming the laser beam to be gaussian and in the fun-damental mode (TEMO), then the diffraction-limited spot size is approx-imately given by

W=--f 0"64 ' (3.1)

where X is the laser wavelength, and f and d are the focal length anddiameter of the focusing lens, respectively. Given a high-quality UVlaser beam such as the second harmonic of the Ar laser 514-nm line(257 nm), for example, feature sizes below I gm can readily be deposited.The major drawback of direct writing is its low scanning speed(typically <K 100 Arm-s-'). Patterned deposition (illustrated in Fig. 15b)can be obtained by projection, an approach in which a mask is opticallyimaged onto the substrate.

470 J. G. EDEN

(a) Spot focus (b) Projection

(c) LUne focus (d) Parallel

Fig. i. Four most common modes for optical beam delivery in photochemical vapordeposition: (a) spot focus, (b) projection, (c) line focus, and (d) parallel. The first two aredesigned for the deposition of small features, whereas (c) and (d) are intended for large-areadeposition techniques (adapted from Ref. 15).

Techniques (c) and (d) in Fig. 15, the line focus and parallel geonie-tries, are designed for large-area deposition. For the line focus, the opticalbeam (or substrate) is scanned, but for the parallel configuration, theoptical beam does not irradiate the surface. The primary advantage of thelatter is that the laser intensity can be increased at will without the possibil-ity of thermally damaging the substrate. Of course, diffusion of the gaseousphotoproducts makes this approach less attractive for spatially delineatedapplications.

m1[-3. PHOTOCHEMICAL VAPOR DEPOSITION 471

C. Optical Sources

A variety of continuous (cw) and pulsed lamps and lasers is availablefor photochemical vapor deposition. Several of the most widely usedsources-and their output wavelengths and power capabilities- arediscussed.

1. Lamps

The relatively low cost, high duty cycle, and reliability of dischargelamps have already made them attractive for commercial applications. Lowintensity is the only major drawback of lamps for broad area deposition,which relegates them to driving single-photon absorption processes. Sincethe poor beam quality of lamps is also an issue for applications involvingspatially delineated deposition, this area has thus far been the sole domainof lasers. One of the first lamps to be used in photochemical vapor deposi-tion is the low-pressure mercury discharge lamp. Its UV spectrum isdominated by two lines, the most intense lying at 253.7 nm and the other inthe VUV at 184.9 nm. As discussed earlier, the 253.7 nm radiation iseffective in triggering the deposition of elemental or compound films byphotosensitization processes. Because of their utility in commercial ap-plications other than photo-CVD (such as curing and sterilization), theselamps are commercially available in both linear and grid configurationswith output powers (at 254 nm) as high as 8 W.

Several other commercially available UV lamps, such as xenon andhigh-pressure Hg lamps, produce broad continua that extend from theinfrared to below 250 nm. The quartz halogen lamp also generates ablackbody spectrum, but its low color temperature results in weak outputbelow 350 nm, the region where the majority of precursors of interest tophoto-CVD absorb.

Shorter-wavelength (VUV) lamps include deuterium (D2 ), hydrogen,and the rare gases. Deuterium lamps produce negligible radiation in thevisible but generate continua that peak near A = 180 nm. Also, D2lamps having average powers up to 60 W are readily available. Hydrogenemits a highly structured spectrum consisting of two bands (Lyman andWerner) peaking near 120 and 160 nm. In contrast, microwave dischargesin the rare gases Ar, Kr, and Xe at high pressure (-200 torr) emit continuain the VUV that arise from electronic transitions of the transient moleculesAr2 , Kr2, and Xe 2. The spectral width of each of the continua exceeds200 nm.

A recent and important entry into VUV lamp family is the microwave-excited rare gas-halide lamp. Kumagai and Obara [381 have generated 29 Wof average power at 193 nm from ArF molecules produced in 2% F2 , 2%

472 J. G. EDEN

At, 48% He, and 48% Ne mixture discharges. The average depositedmicrowave power was 655 W for an intrinsic efficiency exceeding 4%.Because of their efficiency and potential for operating at several discretewavelengths in the UV and VUV, these newly developed lamps willundoubtedly play an increasing role in future photo-assisted depositionprocesses. To summarize, owing to their low intensities, UV and VUVlamps are most useful for those processes in which reliability, duty cycle,and low cost, rather than peak power, are the driving considerations.

2. Lasers

Despite the fact that the energies required for the rupture of chemicalbonds within a polyatomic molecule (-3-6 eV) generally dictate the needfor UV photons, prior to 1975 few efficient lasers existed in this spectralregion. However, the discovery and commercial development of the raregas-halide lasers has dramatically transformed that situation by makingavailable photons ranging in energy from 3.5 to 6.4 eV (193 :S X s 351 nm)in beams having peak intensities (unfocused) exceeding 10 MW-cm - 2 . Infact, much of the early work in photo-CVD was driven by the availabilityof these laser sources. It is not surprising, then, that a large fraction of thephotochemical processes to be discussed in Sections IV-VI rely on anexcimer laser as the optical source.

Over the last five years, the major improvements that have been madein excimer laser performance, such as gas fill lifetime and laser pulserepetition frequency, have permitted the introduction of these lasers intothe industrial environment. At the present time, commercial excimer lasersare capable of producing in excess of 100 W of average power at 248 nm(>30 W at 193 nm) at pulse repetition frequencies > 100 Hz for more than10a pulses (roughly 12 days of continuous operation at 100 Hz).

Another workhorse of photo-assisted deposition has been the argon ion(Ar ) laser. Although generally operated on the visible lines lying between454.5 and 528.7 nm (with the strongest line at 514.5 nm), this laser alsoproduces significant output power on several discrete lines in the ultra-violet (275-364 nm). Recent advances in the materials used in contact withthe high-current discharge have permitted the power deposited into theplasma to be increased, with a resulting improvement in output power. Atpresent, commercial models are available that will generate 1.5 and 7.0 Wof cw power on two separate groups of lines in the 275.4-305.5 nm and333.6-363.8 nm regions, respectively. In addition, a number of photochem-ical vapor deposition experiments and direct write studies, in particular,have employed the second harmonic of the 514.5 nm line (intracavitydoubled) at 257 nm.

III-3. PHOTOCHEMICAL VAPOR DEPOSITION 473

These UV lasers are well suited for those photodeposition applicationsin which the source intensity and beam quality are key considerations.Despite their relatively high cost and (for pulsed lasers) low duty cycleoperation, these characteristics, as well as the capability to quickly changethe output wavelength, make them attractive in photo-assisted processesfor which lamps are not suitable.

3. Other Sources

Although conventional lamps and pulsed lasers are most widely usedfor photo-CVD, several new optical sources have been reported. Window-less lamps, based on high-current, low-voltage (-0.5 A, 200-600 V) dcdischarges in gases such as He and N2, have been shown to efficientlygenerate VUV and deep UV radiation (He: 121.5 nm, N2 : 120-400 nm)[34]. Also, the increasing availability of synchrotrons, coupled with theirability to produce tunable UV and VUV radiation, has prompted severalreports of synchrotron radiation-assisted deposition of elemental andcompound films.

IV. METAL FILMS

Because of their varied applications to microelectronics, metal filmshave been the focus of much of the effort in photochemical vapor deposi-tion. This chapter reviews the elements that have been deposited to date,with emphasis given to the growth conditions. For a detailed discussion ofthe composition and structure of the films, the reader is referred to thereviews of Houle [61, Haigh and Aylett [7], Bauerle [9], and Jackson et al.[39].

A. Column 118 Metals (Cd, Zn)

Photodeposition of the Column IIB metals cadmium and zinc fromtheir respective alkyls has been the subject of intense scrutiny (mercurywill be discussed in Chap. V, Sect. B in conjunction with the II-VIcompound semiconductors). In 1966, Jones et al. [40] demonstrated thatthin Cd films (<100 A thick) could be patterned on Si0 2 substrates byphotodissociating dimethylcadmium (DMC) with 254 nm radiation froman Hg resonance lamp. Subsequent work has centered on the photodis-sociation of DMC in the gas phase or in an adlayer with excimer or At'(257 nm: second harmonic of 514.5 nm) radiation. In a series of papers,Deutsch, Ehrlich and Osgood [14], [31], [41] reported the broadareadeposition of Cd films, as well as he writing of lines with widths as small as

474 J. G. EDEN

0.7 ;m. DMC pressures of 1-4 torr in the reactor allowed deposition ratesof 7-13 A-sec- ' to be obtained by photodissociating chemisorbed andphysisorbed alkyl molecules on glass and quartz substrates. At higher in-tensities, deposition rates up to 1,000 A-sec-' have been achieved, and anHg arc lamp (200 W), ArF laser (30-300 mJ-cm - 2) or cw argon or kryptonion lasers have successfully served as optical sources.

Similar studies have been carried out for zinc where dimethylzinc,Zn(CH 3)2, or diethylzinc, Zn(C2H5)2, acted as the gas phase precursor.Several square centimeters of Zn films have been deposited on quartz orsapphire substrates by illuminating Zn(CH3)2/Ar gas mixtures with deepUV radiation from a mercury-xenon arc lamp. The threshold wavelengthfor dissociating the zinc alkyl is -245 nm, and film thicknesses up to0.6 pm have been obtained 1421. Also, Zn lines with widths as small as2 pAm and larger area films (400 pm 2) have been deposited with an ArF orKrF laser operated in an unstable resonator so as to minimize diffraction ofthe beam.

B. Column liA Metals (Al, Ga, In, TI)

The Column I1A metals aluminum, gallium, and indium have alsofound widespread use in microelectronic and optoelectronic devices asinterconnects and as elements in 111-V semiconductor compounds. Of thethree, aluminum has received the most attention in photodepositionstudies because of its excellent bulk conductivity. Both direct writingof Al lines and broad-area deposition of metal films have been demon-strated. With dimethylaluminum hydride (DMAH) as the precursor and60-100 mW of 257 nm laser power, for example, Al lines of -2-3 pm inwidth have been written on a variety of substrates at speeds of 1 pm-sec-11431. At this scanning speed and a laser intensity of -115 kW-cm - 2,deposition rates of 0.09 /m-sec - 1 were obtained. Broad-area depositionon polysilicon was demonstrated by Solanki et al. [44] by photodissociat-ing TMA in parallel geometry with a KrF laser. Deposition rates up to1,000 A-min - ' and film thicknesses > 0.5 pm were recorded.

Aluminum films having resistivities as low as 6.2 pfl-cm(2.3 times thebulk value) were deposited from TMA with a deuterium lamp by Hanabusaet al. [45]. Deposition rates up to 200 A-min- were realized at 2000C,and, in contrast to the photochemically deposited films, Al films depositedby pyrolysis at 260*C were highly resistive (140 p4-cm). With essentiallythe same experimental arrangement but studying DMAH as the Al precur-sor, Hanabusa et al. [46] compared the results obtained with a D2 lamp andthe ArF laser. At 200*C, deposition rates of 190 A-min- 1 and 380 A-min-'were obtained with lamp (140 mW-cm- 2) and laser (230 mW-cm - 2) radia-

III-3. PHOTOCHEMICAL VAPOR DEPOSITION 475

tion, respectively. Not only did the presence of VUV illumination at thesubstrate permit Al film deposition to occur at considerably lower tempera-tures than those accessible by pyrolysis, but the VUV once again dramati-cally improved the film conductivity. Because of the low concentrations ofcarbon in the photolytically deposited films (<6%), DMAH is a particu-larly attractive new precursor for optical source wavelengths below 200 nm.

Considerably less is known of the photodeposition of Ga and In.Gallium films have been deposited from Ga(CH 3)3 with both arc lamps[471 and a frequency-doubled Ar + laser [48]. Deposition rates up to70 A-s-1 have been achieved with the latter, while the large-area studies[471 yielded gallium metal droplets. Aylett and Haigh [471 carried outsimilar experiments with various indium precursors (indium trimethyl andcyclopentadienyl indium) and GaAs and Pyrex substrates. They once againobserved the formation of droplets on glass but obtained films withsmoother morphology on GaAs. Thin In films have also been deposited bydissociatively photoionizing InI at 193 nm to yield In+-I - pairs f37]. Foran Inl number density of 1015 cm - 3 , deposition rates were -400 A-h- 1 .Similar results were obtained for thallium films deposited on silver bydissociatively photoionizing TII with a Xe arc lamp [371.

C. Column IVA, VIA and VIIA Transition Metals (Cr, Mo, W, TI, Mn)

Because of their applications as interconnects and in gate metallization,the transition metals are of interest for photodeposition, and specifically inthose situations where processing temperature is a critical factor (such asin Ill-V device fabrication). Extensive studies of the photo-CVD of Cr,Mo, and W have been carried out, with most prior work involving thephotodissociation of their respective hexacarbonyls with an excimer laser.Large-area (>6 cm 2) Cr films have been deposited with XeCl (308 nm),KrF (248 nm), and ArF (193 nm) lasers in perpendicular and parallelgeometries [491, but most efforts have concentrated on the perpendicularconfiguration because of the poor adhesion of deposited films if the sub-strate is not irradiated. The highest deposition rates, 2,000 A-min - ', areobserved at 248 nm, but the films contain only -50% Cr, with the remain-der consisting of C and 0 because of the fact that all CO ligands are notremoved when adsorbed carbonyls are photodissociated. Direct writing oflines (and dots) has also been demonstrated [50] at 257 nm.

Similar results have been reported for tungsten. Lines (<3 jsm width)and ohmic contacts have been deposited from the carbonyl precursor, andbroad-area deposition rates up to 1,700 A-min- obtained at 248 nm [491.If one resorts to WE (in the presence of H2) as the precursor, deposition

rates in excess of 1.000 A-min - ' are obtained in parallel geometry at

476 1. G. EDEN

TABLE IIISUMMARY OF THE PHYSICAL PROPERTIES OF MO, W, AND CR FILMS PHOTODEPOSITED WITH

A KRF LASER (248 NM) FROM METAL HEXACARBONYL PRECURSORS (AFTER REF. 49)

Deposition rate Resistivity Percent carbon Adherence Tensile stress(A/min) (l/0) in film (PSI) (dyn-cm-')

Mo 2,500 0.9 0.9 >8,000 <3 x 10'W 1,700 4.5 0.7 >9,400 <2 x 109Cr 2,000 6 0.8 >7,800 <7 x 109

193 nm, with film resistivities within a factor of two of the bulk value [5i].Less work has been reported for the photodeposition of molybdenum, butthe results are comparable to those obtained for Cr and W. Gilgen et al. [52]have deposited molybdenum and tungsten lines on glass, GaAs, and sap-phire substrates by scanning the focused UV lines (350-360 nm) from anAr laser. For Mo(CO) 6 and W(CO)6 pressures of 0.17 and 0.035 torr,respectively, and a writing speed of 1 /.m-s- t , deposition rates up to2,700 A-s - t for Mo lines and W film thicknesses up to 1,100 A wereobserved. Several properties of Cr, W, and Mo films photodeposited inparallel geometry from their respective hexacarbonyls with a KrF laser(248 nm) are summarized in Table III.

Two other transition metals that have been deposited photochemicallyare titanium and manganese. Direct writing of 4-/Am-width Ti lines onLiNbO3 has been exploited to fabricate single mode waveguides [53] withonly 5 mW of power on the second harmonic of the Ar' laser (257.2 nm).Manganese films have also been photodeposited from a methylcyclopenta-dienyl precursor with a 2.5 kW Hg-Xe lamp [54].

0. Column IB Metals (Sn, Pb)

Although tin and lead films have been photodeposited from alkylprecursors, characterization of the gas and surface processes and resultingfilms has been limited. In both cases, adlayer photolysis appears to triggerfilm deposition and thin (<100 A thick) films have been deposited onGaAs, glass, and sapphire substrates. Chiu et al. [551 have deposited Pbfilms from tetraethyl lead with frequency-doubled Ar + laser radiation andobtained deposition rates up to 30 A-s' for a precursor pressure of 0.7torr. Mingxin et al. [56] deposited tin films on both quartz and carbon sub-strates by photodissociating Sn(CH3)4 with the second harmonic of theAr' 514 nm laser line or an Hg resonance lamp. Depending on the sourceintensity, both amorphous and polycrystalline films resulted. With Hg

iii-3. PHOTOCHEMICAL VAPOR DEPOSITION 477

resonance lamp (3 mW, 254 nm) dissociation of the metal alkyl, depositionrates were much smaller than those obtained with the laser- typically,0.06 A-min - '. Interference structures appearing in the photodepositedfilms indicate that patterned tin films with features as small as 0.2 Jim canbe deposited by this approach.

E. Column VIII Transition Metals (Fe, Pt, Ir)

Conducting films of iron were deposited onto various substrates byGeorge and Beauchamp [541 by photodissociating the pentacarbonyl withan Hg-Xe arc lamp. Carried out at Fe(CO) 5 partial pressures of 10-2 torr,these studies yielded -300 A thick films after an exposure period of 104 s.The authors attributed deposition to photoelectrons (ejected from thesubstrate by the optical source) that decomposed the carbonyl precursor.Thin iron films have also been deposited under UHV conditions by photo-lyzing Fe(CO) 5 adsorbed onto GaAs or Si substrates. Both excimer lasersand arc lamps have been used in perpendicular geometry, and the filmsstudied to date contain significant concentrations (> 10 at %) of carbon andoxygen. Iron has also been deposited onto Pyrex, quartz, and GaAs atrates up to 30 A-s- ' with 257 nm radiation from a frequency-doubled Ar4

laser [501.Recently, Armstrong et al. [571 deposited Fe/Ni composite films by

photodissociating mixtures of ferrocene and nickelocene vapor at roomtemperature. The ferrocene absorption spectrum peaks near 190 nm, whilethat for nickelocene reaches its maximum value at -285 nm. When theferrocene/nickelocene mixture was irradiated at 193 nm (ArF laser), thedeposit was found to consist of 92% Fe and 8% Ni. Similar measurementsmade at 337 nm with a pulsed N2 laser (X = 337 nm) yielded 65% Fe-35%Ni deposits. Film deposition rates at 193 and 337 nm were determined tobe 20 A-s- and >200 A-s- 1 , respectively.

Platinum films have been deposited from a variety of precursors ontoquartz, graphite, glass, sapphire, and GaAs substrates. Utilizing bothpulsed and cw ultraviolet sources, Koplitz et al. [58] demonstrated that-1,000 A thick Pt films can be deposited from CpPt(CH 3)3 whereCp = n5-CH s . For 26 mW of average power at 308 nm (10 Hz), filmswere observed to grow at a rate of -100 A-in- 1 over an area of -3 mm 2 .Similar results were obtained for Pt films photodeposited with a cw Ar + ionlaser operating simultaneously on the 351 and 364 nm lines. Platinum filmshave also been deposited on quartz and graphite with KrF excimer radia-tion and the frequency-doubled 514 nm line of Ar + [56]. Gilgen et al. [52]deposited thin Pt lines (from Pt(Hfacac) 2) on glass and GaAs at a scanspeed of 0.4jim-s - 1 with the 350-360 nm UV lines of an Ar4 laser. For

478 J. G. EDEN

glass substrates, the film resistivities were -2 to 10 times the bulk value.Both Pt and Ir films have also been deposited from their respective acetyl-acetonates by photoionizing the precursors with frequency-quadrupledradiation (266 nm) from an Nd:YAG laser [171. The beam was passed0.5 cm from the Si(111) substrate, resulting in the deposition of -330 pLmwide metal stripes consisting of >95 at % Pt.

F. Group 1B (Cu, Au)

Despite the importance of the Column IB metals copper, silver, andgold to an array of fundamental microelectronics processes (doping, metal-lization line and mask repair, etc.), photodeposition of these metals fromthe vapor phase is complicated by the scarcity of precursors. Simple alkylmolecules (i.e., of the form MX. where M = Cu, Ag, or Au; X = CH 3 ,C2H5 , C4H9.... and n is an integer) are not volatile and the carbonyls[M(CO),] are unavailable or do not exist. It was not until 1985 that Joneset al. [591 demonstrated the viability of volatile coordination complexes asprecursors by depositing Cu films upon photodissociating a fluorinatedacetylacetonate complex, bis(1,1,1, 5,5,5-hexafluoropentanedionate) cop-per (II)-(Cu(Hfac) 2), with both pulsed and cw UV optical sources. Afrequency-doubled Ar + ion laser (257 nm), ArF and KrF excimer lasers,and a low pressure Hg lamp were all investigated.

With the frequency-doubled Ar + ion laser, deposition was observedwith laser intensities of 0.2-2 x 10' W-cm- 2 at a Si substrate. Films-1 jAm in thickness were deposited at rates between 1 and 200 A-min - .ArF or KrF laser irradiation of quartz substrates at peak intensitiesZI MW-cm- 2 yielded thin Cu films (thicknesses s 500 A), and thethreshold intensity for film deposition was found to be -1 MW-cm- 2.Regardless of the optical source, the resulting Cu films also contained10-90% C depending on whether ethanol was added to the Cu(Hfac)2vapor.

Similar comments could be made for the photodeposition of gold fromthe vapor phase. Baum, et al. [601 demonstrated that Au can be pat-terned onto quartz substrates with a simple optical projection system toproduce metal line widths as small as 2 Am. In their experiments, di-methylgold (III) acetylacetonate, Me 2Au(acac) , was photodissociated bypulsed excimer radiation at 193 nm (ArF), 248 nm(KrF) or 308 nm(XeCI)or by an Hg resonance lamp. Films, 2,000-3,000 A thick were produced.and the highest Au content was achieved in films deposited with KrF laserradiation. No deposition was observed if A = 351 nm, which demonstratedthat the film growth processes are photochemical in nature. Deposition wasalso observed with a Hg resonance lamp. Aylett and Haigh [611 also photo-

1II-3. PHOTOCHEMICAL VAPOR DEPOSITION 479

deposited gold films by photodissociating trimethylgold(III)-trimethyl-phosphine [(CH 3)3 Au-P(CH 3 )3] at 248 nm. Both broad-area and patterneddeposition were demonstrated, the latter by projecting a mask ontothe substrate. Considerably more work has been reported in which Cu or Aufilms were deposited from liquids, and, for details, the reader is referred tothe excellent review by Haigh and Aylett [621.

V. SEMICONDUCTORS

The potential economic impact of photochemically deposited semicon-ductor films has driven the tremendous activity in the growth of bothelemental and compound materials that this field has witnessed over thepast decade. While efforts to deposit amorphous and epitaxial films of Sihave accounted for most of the work in this area, considerable success hasalso been realized in the deposition of II-VI and III-V compound, films.

A. Elemental Films

1. Carbon

The deposition of amorphous carbon films has been accomplished byphotodissociating halomethanes or vinylchloride (CC14 , C2H3C1) [63] oracetylene with an ArF excimer laser. For C2H3CI as a precursor, filmdeposition rates up to 150 A-min - 1 were measured, and the films con-tained up to -1 at % CI. Graphite films can also be deposited by thephotodissociation of C2H2 in parallel geometry or when the substrate isirradiated at a glancing angle. Recently, Ohashi et al. [64] deposited amor-phous hydrogenated carbon films by irradiating n-butane vapor with un-filtered radiation from a 750 MeV synchrotron. Films having thicknesses>0.4 /An were deposited with n-butane pressures under 1 torr, and the filmdeposition rate was found to increase linearly with negative bias on thesubstrate. Accordingly, the production of positive ions by the synchrotronradiation was identified as the dominant contributor to film deposition.

2. Silicon

All of the techniques that have been developed thus far for photode-positing Si may be classified into one of three categories: (1) Hg-photosensitized deposition; (2) direct decomposition of higher silanes; and(3) laser photodeposition, where the first two approaches rely on lamps asthe optical source.

Since the pioneering work of Emelus and Stewart [2] and Niki andMains [65] exploring the Hg photosensitized decomposition of monosilane

480 J. G. EDEN

(SiIHl) and germane (GeH 4), both amorphous hydrogenated and micro-crystalline Si films have been deposited by a number of groups [66,67].These studies showed photo-CVD deposited a-Si films to be of comparablequality to plasma CVD deposited films, with the added benefit that ion-induced damage is not present in photochemical processing. Early workreported film deposition rates of 20-35 A-in-, which was subsequentlyincreased to 600 A-min- 1 by improvements in reactor and Hg lamp design.In addition, Mizukawa et al. [68] deposited p-type amorphous Si films byadding small concentrations of diborane (B2H6) to the silane vapor in thereactor (200-250*C substrate temperatures).

Studies of the Hg photosensitization kinetics chain have identified theproduction of the higher silanes (Si 2H6 , Si3H8) from the collisional dis-sociation of Sill, as a critical step. Consequently, in an effort to bypass theunnecessary in situ conversion of SiH 4 to Si 2H6 , most recent Si photode-position by Hg sensitization has utilized Si2H6 directly. Inoue et al. [69]reported the deposition of amorphous Si films having thicknesses up to-1 jsm by irradiating a Si21H/Hg mixture for 10 min. Without the pre-sence of Hg in the reactor, the film deposition rate was observed to fall byone to two orders of magnitude. Both n- and p-type amorphous hydroge-nated and microcrystalline Si films have also been deposited [70] by addingphosphine (PH 3) or B2H6 , respectively, to the reactor. The ratio of PH3 orB2H6 to Si2H6 was 6,000 or 10,000 ppm, respectively, and the total gaspressure in the reactor was 2 torr. Solar cells (SiC/Si p-i-n structures)have also been fabricated [71] from photodeposited amorphous films andhave yielded efficiencies at AMI.5 of up to 8.5% [72].

A major development in the application of Hg photosensitization tofilm deposition was the report of Nishida et al. [72] in 1986 that epitaxialSi films could be grown at substrate temperatures as low as 200C by thisprocess. The addition of SiH 2F2 to the gas stream was found to be critical,not only in the early stages of film growth but throughout the depositionprocess. It was concluded that halogen-bearing radicals or free F atomsmay be responsible for removing the native oxide. Mass flow rates for theSiH 2F2 2xceeding -20 sccm at a total reactor pressure of 2 torr werenecessary to grow epitaxial films, and deposition rates up to 0.7 A-sec'were observed.

The underlying motivation behind Hg photosensitized deposition ofSi from monosilane is the weak absorption of the precursor in the UV andat the 185 and 254 nm lines of the Hg resonance lamp, in particular. Be-cause of the increasing absorption at longer wavelengths that is associatedwith resorting to progressively higher silanes (i.e., Sill4- Si 2H6-* Si3HS),however, one can avoid the presence of Hg in the reactor entirely byphotodissociating disilane (or trisilane, Si3Hs) directly. Early attempts at

ii-3. PHOTOCHEMICAL VAPOR DEPOSITION 481

I I I I I I I I I | I I I I

0 : Thermal

Photochemical

2~ o\

1100(2

0 lo - -4

, 1 0 0 s t l 1 1 -)0 s

if~o I I I I I I I I I I I

1.0 1.5 2.0 2.5IO0/ s (K - 1)

Fig. 16. Growth rate as a function of temperature for amorphous Si films deposited byphoto-CVD or pyrolysis of SiH 6. The data show a characteristic of photo-CVD: its lowinherent activation energy. (Adapted with permission from the Japanese Journal of AppliedPhysics, 22, L46, Y. Mishima. M. Hirose, Y. Osaka, K. Nagamine, Y. Ashida, N. Kitagawa,and K. Isogaya, 1983).

photodissociating Si2H6 with Hg lamp radiation resulted in deposition ratesthat were low compared to those obtained by Hg photosensitization.Mishima et al.73 observed deposition rates for amorphous Si films as high as15 A-min - ' when Si2H6 was directly photodissociated with the 185 and254 nm lines from an Hg lamp. The lamp intensity was -80 mW-cm - 2 , thequartz or Si substrate temperature was varied between 200 and 400*C,the gas feedstock mixture consisted of 1% Si2H 6 in He, and N2 acted asthe carrier gas. As shown in Fig. 16. the photochemical film growth rateis independent of substrate temperature in the 200-400'C range. which isconsistent with the low activation energy (E,) also reported in Ref. 72(0.18 eV).

Increased deposition rates and the growth of epitaxial films were re-ported by Gonohe et al.74, who illuminated a Si2 H6/N2 gas mixture and

482 J. G. EDEN

Si(100) substrates with the VUV radiation from a D2 lamp. Epitaxialfilms were obtained at 650"C, and at that substrate temperature, the de-position rate was measured to be -900 A-min - 1 . Over the 550-650'Ctemperature range, the activation energy for the deposition process is23 kcal/mol - eV, which is roughly 66% of the value for thermal deposi-tion (conventional CVD). Consequently, at the low end of the tempera-ture region studied (550"C), the photo-CVD deposition rate was found tobe more than double the thermal rate. The assistance of the incidentVUV photons in cleaning the Si surface was also clearly demonstrated.

Fuyuki et al. [75] also explored the use of a D2 lamp in depositing Sifilms but focused on a-Si:H films deposited in the 200-300"C temperaturerange. No dependence of the 15 A-in- ' deposition rate in this tempera-ture interval was observed, but substituting a microwave-excited Xe lampfor the D2 source dramatically improved the rate. For -120 and 50 W ofinput power to the lamp, the a-Si:H film deposition rate was observed to be75 A-in- 1 and 10 A-min- 1, respectively. By changing the optical windowon the reactor from CaF2 or MgF2 to quartz, it was shown that film growthrequires radiation wavelengths below 180 nm. After photolyzing Si2 H6 andSi3H8 with an Hg lamp producing 3.5 mW-cm - 2 at 185 nm, Kumataet al. [76] found that the deposition rate for a-Si:H films grown from Si3H8was five times that for Si-2H6 as a precursor. The conductivity of filmsdeposited from Si 3H8 at 300"C (10- 5(fl-cm) - ') makes them well-suited forincorporation into solar cells. Recently, Kim et al. [77] prepared p-i-na-Si:H solar cells from films photodeposited from Si2H6 and measuredefficiencies >11%.

Amorphous, polycrystalline, and epitaxial Si films have been depositedwith lasers from several different precursors and in both parallel and per-pendicular configurations. Andreatta et al. [78] deposited polycrystallineSi films on quartz substrates at rates up to 9.8 A-sec - 1 with a weaklyfocused KrF laser and peak intensities of -20 MW-cm- 2 . The high intensi-ties required are a consequence of the dominance of two-photon dissocia-tion of silane at 193 nm. Film thicknesses up to -2.000 A were obtainedwith Z400 s exposure times and 10% SiH,4 in N2 gas (total pressure ofseveral hundred torr). Several films were also deposited with Si(CH 3)4 asthe precursor.

Several groups [79, 80] have reported the deposition of hydrogenatedamorphous films from the photodissociation of Si2H6 and Si 3H8 with anArF excimer laser. Deposition rates typically range from 100 to 300 A-min - 1, and the film properties are comparable to those for films depositedby plasma processes.

Epitaxial Si films were deposited by excimer laser-assisted CVD byYamada et al. [81] in the 600-650"C temperature range. Not only did the

i-3. PHOTOCHEMICAL VAPOR DEPOSITION 483

presence of UV photons at the substrate triple the film growth rate, butcrystallinity and hole mobilities were both markedly improved with the UVlaser beam, particularly at the lowest substrate temperature studied(600C). At lower Si2H6 pressures (0.01-0.1 torr), thin Si films have alsooeen deposited on quartz or sapphire substrates. Frieser [82] found thatthe deposition of epitaxial Si films from the pyrolysis of Si2CI6/H2 gasmixtures was accelerated when the substrate was irradiated by an Hg arclamp. Since the lamp's 5 eV (A - 254 nm) photons are not sufficientlyenergetic to rupture the Cl3Si-SiCI3 bond, the observed enhancement wasattributed to surface interactions. Similar results were obtained by Ishitaniet al. [83], who measured significant increases in the deposition rate ofepitaxial Si (by CVD from SiH 2C12, dichlorosilane) when Hg-Xe UVradiation was directed onto the substrate. Specifically, the growth rate at9000C rose from 1.8 to 2.8 .m-min', which was attributed to thephotodissociation of the SiCI2 radical at the substrate surface.

3. Germanium

Fewer studies have been concerned with the photodeposition of ele-mental germanium. Polycrystalline films up to 1,500 A in thickness weredeposited by Andreatta et al. [78] by photodissociating germane (GeH 4)with 193 or 248 nm excimer radiation. The perpendicular geometry yieldeddeposition rates of 4-5 A-sec - for ArF laser intensities of 2 MW-cm - 2.Higher 248 nm intensities (-8 MW-cm - 2) were required to reach themaximum film deposition rate (5-6 A-sec - ) for that source wavelength.No film growth was observed when the laser wavelength was 308 nm (XeCIlaser), which demonstrates the wavelength threshold and, therefore, thephotochemical nature of the process.

Epitaxial Ge films were later grown by Eden et al. [84] on NaCl (100)substrates at substrate temperatures as low as 120C. Polycrystalline filmswere deposited on quartz substrates (Ts s 250"C) with grain sizes rangingfrom 0.3 to 0.7 m. In parallel geometry, thin epitaxial films have beendeposited on GaAs (100) substrates at temperatures as low as 285°C byphotodissociating GeH 4 with an ArF laser [85]. The position of the laserbeam was maintained at a2 mm above the substrate, and the dominantrole played by the laser appears to be the conversion of Ge H to Ge 2H6,which migrates more than 10 mean free paths to the substrate, where itsubsequently pyrolyzes. Etching the substrate in situ by flowing 5% HCI(in He) through the reactor for 5 min was found to be necessary for obtain-ing epitaxial films. After - 400-700 A of epitaxial growth, the growingfilm switches abruptly to amorphous material, which is apparently causedby the buildup of carbon (from the vacuum system) at the solid/vapor

484 J. G. EDEN

interface. More than an order of magnitude increase in the Ge film growthrate is observed if small amounts of ammonia are added to the GeH 4/Hegas flow stream. Kiely et al. [861 demonstrated that, despite the accelera-tion in the deposition rate, the Ge films are nevertheless epitaxial. Uponabsorbing a 193-nm photon, NH3 produces an H atom, which subsequentlydecomposes GeH 4 by hydrogen abstraction. Furthermore, the absorptioncross-section of NH 3 at 193 nm is almost three orders of magnitude largerthan that for GeH 4 at the same wavelength (O'NH,: HI:7 cm2; rGCH,:2-3 x 10- 20 cm2).

Structures consisting of alternating layers of amorphous Ge and Si weredeposited entirely by photochemical processes by Lowndes et al. [87]. Filmdeposition was carried out in parallel geometry (laser beam-to-substratedistance z 1 mm) by photodissociating Si2 H 6 or GeH 4 in the reactor.Ge and Si layer thicknesses of 5.4 ± 0.2 and 10.7 ± 0.4 nm, respectively,were deposited in fabricating a nine-layer structure. Germanium/siliconalloy films have also been photodeposited from GeH 4/Si 2H 6 mixtures,using an ArF laser [881. For substrate temperatures below 350*C, deposi-tion occurred entirely by photochemical processes, and deposition ratesranged from 150 to 300 A-min - '. The ratio of Si to Ge concentrations inthe films was found to be a factor of three higher than the Si 2H6 :GeH 4partial pressure ratio, which suggests the importance of cross-reactionsbetween Ge and Si-containing species in the film deposition process. Al-ternative precursors for the photo-CVD of Ge were explored by Stanleyet al. [89], who deposited polycrystalline Ge films from ethylgermane anddiethylgermanium. Although an infrared (CO 2 ), rather than UV, laserwas used, the photochemical nature of the deposition process was demon-strated by the need for the laser to be tuned to an IR absorption band ofthe precursor.

B. II-V Binary and Ternary Compounds

Photo-assisted CVD is a particularly attractive deposition technique forthose materials and devices that are sensitive to processing temperature.As Irvine (90] has pointed out, interdiffusion in the I-VI ternary com-pound CdxHg1 -. ,Te, for example, is typically a 100 jm at a growth tem-perature of 600C but is <1 Atm for temperatures at or below 400C. Effortsto fabricate multilayer structures with abrupt interfaces in the 11-VI com-pounds, therefore, hinge on an ability to deposit high-quality films atreduced temperatures. While the temperatures at which the binaryI1-VI semiconductors ZnSe, ZnS, CdS, CdSe, CdTe, and HgTe are depo-sited by MOCVD can be reduced to below 250C by resorting to lessthermally stable precursors, photo-CVD of these compounds at tempera-tures as low as 150-200C has yielded epitaxial films of high quality.

I1m-3. PHOTOCHEMICAL VAPOR DEPOSITION 485

1. Zinc Compounds

The earliest work in 1I-VI compound photodeposition centered on thezinc-containing binary compounds ZnSe and ZnO. Johnson and Schlie[421 were the first to demonstrate the deposition of zinc selenide films byirradiating mixtures of Zn(CH3 )2 and Se(CH3 )2 vapor with an Hg arclamp. Although the films were not stoichiometric, film thicknesses up to-0.6 gm over more than 3 cm 2 of surface were obtained. Exposureperiods of 1-2 h were necessary to deposit these films due to the lowfluence of the Hg lamp for wavelengths below -245 nm. which was foundto be the wavelength threshold for the photodissociation process. Polycrys-talline zinc monoxide films were subsequently reported by Solanki andCollins [361, who employed an ArF or KrF excimer laser to photodissociateZn(CH 3)2 and NO 2 (or N 20). Both parallel and perpendicular geometrieswere explored, but the highest optical quality (defined with regard to filmtransmission) and lowest resistivity films were obtained by irradiating theSi substrate. Deposition rates up to 18 L m-h-' were measured, and theuniformity in film thickness over 10 cm 2 of surface area was ±5%.

In 1985, Ando et al. [91] photodeposited ZnSe epitaxial films attemperatures as low as 200*C. Diethylzinc and dimethylselenide were thealkyl precursors and the UV radiation was provided by a low-pressure Hglamp. At 400*C, the presence of the UV improved the film deposition rateby a factor of two. Epitaxial ZnS and ZnSe films have also been depositedon GaAs with excimer laser radiation (193 and 248 nm) by Fujita et al. [921and Shinn et al. [93], respectively. In both cases, the UV laser beam passedparallel to the substrate and epitaxial films were obtained in the 100- 150*Crange fo: ZnS and at temperatures as low as 200*C for ZnSe. Althoughsingle pulse laser energies were varied from 7.5 to 50 mJ, deposition ratesdid not exceed 1 pim-h- 1.

2. Cadmium and Mercury Binary and Ternary Compounds

Considerably more effort has been devoted to the Cd and Hg-based[I-VI films. Virtually all of the photodeposition studies reported to datehave relied upon the alkyls as sources for one or both of the Column IIBand VIA atoms. The first photodeposited epitaxial 1i-VI film, HgTe, wasgrown by Irvine and co-workers [941 from Te (C2H5)2 and mercury vaporby Hg photosensitization. Films -1 Atm thick were deposited at rates upto -2 jsm-h - 1 in the 240-310*C interval, and epitaxial films could begrown at temperatures as low as -180*C. In the 180-250'C range, no filmdeposition occurs in the absence of UV radiation. Increasing the precur-sor partial pressures and/or Hg lamp intensity will offset the decline in filmgrowth rate at low temperatures. and it has been estimated that a deposi-tion rate of 1 /m-h- could be obtained at 180'C for UV intensities at the

486 J. G. EDEN

substrate of 100 W-cm - . Utilizing the radiation from an Hg/Xe arc lamp.Ahlgren er al. [33,951 also photochemically deposited HgTe films attemperatures as low as 182°C and successfully fabricated HgTe-CdTesuperlattices (>50 periods) by alternately introducing Te (CZHs)2 :adCd(CH3 )2 into the reactor (in the presence of Hg vapor). HgTe filmdeposition rates were <0.1 m-h-1 at 182°C when diethyltellurium was theTe precursor but, at 240*C, growth rates rose to roughly 1 Mm-h-1 whenusing the diisopropyl as the precursor. Laser photochemical vapor deposi-tion of HgTe was recently reported by Fujita et al. [921. With KrF or ArFlasers and a parallel configuration, epitaxial films were grown on CdTe(111) substrates at deposition rates up to -1 Mm-h- 1.

Kisker and Feldman [96, 97 grew epitaxial CdTe films on GaAs (100)at substrate temperatures down to 2500C with a low-pressure Hg lamp. Fordimethylcadmium and diethyltellurium as the precursors, Gbtaining epita-xial films required that the Cd(CH 3)2 :Te(C 2H5 )2 partial pressure ratio be1.5:1. Long-term film growth rates up to 2 Mm-h - were recorded at350°C. Cody et al. [98, 99] have also grown epitaxial CdTe films on GaAswith an Hg lamp. For only 100 mW-cm- 2 of radiation incident on thesurface (200 :s A r 250 nm), maximum deposition rates were found to be ashigh as 13 and 9 Mm-h- 1 for CdTe and GaAs (100) substrates, respectively[98]. For (CH 3 )2Cd and (C2H5)2Te partial pressures of 5.4 x 10- and1.6 x 10-3 atm, respectively, photodeposition rates were typically 6Am-h-' at 250'C [99].

Excimer laser-assisted photoepitaxy of CdTe was demonstrated byZinck et al. [100, 101]. Growth rates as high as 2 Mm-h- 1 were recorded ata deposition temperature of 165°C. The precursors for these studies werealso Cd(CH 3)2 and Te(C2 H5 )2 , but partial pressures for both reactantswere low: 10-3_10 - 2 torr. Also, while both 193- and 248-nm radiationwere investigated, most of the effort focused on the KrF laser.

The ternary alloy. CdxHg_ Te, was deposited by Irvine et al. [102,1031, who grew thin epitaxial films by Hg photosensitization. With diethyl-telluride, dimethylcadmium, and Hg vapor as the precursors and a3 kW Hg arc lamp, epitaxial films 1.3 ;m thick were grown at tempera-tures down to 250"C. The use of He rather than H2 as the carrier gas wasfound to be necessary for the growth of epitaxial films. Morris [1041subsequently grew epitaxial Cdfi) 8Hgo.2Te films with an ArF laser in par-allel geometry and, rather than Hg vapor, employed Hg(CH 3)2 , dimethyl-mercury, as the Hg donor molecule. With CdTe substrates, growth rates of-4 -h-' and >2 Am film thicknesses were obtained at a temperature of150'C. Since the laser did not irradiate the substrate, transient heatingof the growing film by the optical source was avoided. Recently, Cody et at.[991 obtained similar growth rates (4 Mm-h - ') for 20% CdTe/80% HgTe

iI-3. PHOTOCHEMICAL VAPOR DEPOSITION 487

alloy films that were deposited at 250C with the aid of a 2 kW Hg lamp.The best film quality resulted from using methylallyltelluride as the Teprecursor, and the variation in film composition, x (i.e., Cd1 Hgl-,.Te), wasonly 4% over 1 cm 2 of surface area.

C. ll-V Compound Films

Although the first efforts to photodeposit 111-V films were reportedearly in the last decade, the application of photo-CVD to this family ofmaterials initially developed more slowly than metal or 11-VI film deposi-tion, but progress has recently accelerated significantly. The presence of anoptical beam at the substrate during the growth of III-V films has beendemonstrated to yield several beneficial results, such as improvements insurface morphology, modification of carrier concentrations, and, in severalcases, increases in deposition rates. This section will provide an overviewof the work reported to date in this area with emphasis on the photo-assisted deposition of GaAs.

1. Gallium Arsenide

Most of the effort in photodepositing 111-V films has concentrated onthe binary compound GaAs. Putz et al. [105] examined the growth of GaAsin the 510-650C range and found that the presence of UV photons froman Hg lamp improved the maximum (saturated) growth rate from 5 to8 jum-h - ' at 510C. Also, they observed that "... in all epitaxial runsimproved surfaces were obtained upon illumination ... " [105] and thatmirrorlike surfaces were produced, even at temperatures as low as 510C.Similar results were obtained subsequently by Balk et al. [106], who em-ployed ArF and XeF excimer lasers in addition to the Hg lamp. While thefilm growth rate was observed to double for 193 nm illumination of the sub-strate, no effect was detected for A = 351 nm, which points to the photo-chemical nature of the effect.

Nishizawa et al. [107] explored the impact of excimer laser radiation atthe substrate on the growth of GaAs by chloride-transport vapor phase

epitaxy (VPE) and MOCVD. Not only were film growth rates enhanced,but the Hall mobilities of films grown in the presence of UV photons wereimproved over those deposited without substrate illumination. At a sub-strate temperature of 600°C, for example, the growth of epitaxial GaAsfrom AsC13 and GaCI3 by VPE was roughly quadrupled (from -1.2 to4.5 A.m-h-') by irradiating the GaAs(100) substrate with 2.7 W of averagelaser power at 248 nm (KrF laser). Similar improvements were observedover the 480-700C range. Also, epitaxial films were grown at tempera-tures as low as 350°C.

488 J. G. EDEN

Chu et al. [1081 grew epitaxial GaAs films on n-type GaAs(100)substrates at temperatures in the 425-500*C range by irradiating the sub-strate during film growth with an ArF laser fluence of 19-38 mJ-cm - 2 . Inthe absence of the UV irradiation, no deposition was observed at tempera-tures up to 600(C under conditions that were otherwise identical. Deposi-tion rates ranging from 0.02 Am-min -' at 425°C to 0.1 Am-min-1 at 500°Cwere measured and the electrical and chemical characteristics of the filmswere determined to be similar to those for films grown by conventionalMOCVD. Hole mobilities, for example, for films grown in the 425-500WCinterval varied from 150 to 200 cm 2/V-sec, with carrier concentrationsranging from 5 x 1016 to -3 x 1017 cm- 3. The carbon number density wasmeasured by SIMS to be 4 X 1017-2 x 1018 cm- 3 for films deposited at500*C. York et al. [1091 carried out similar experiments but with lowerlaser fluences (<13 mJ-cm- 2) and at wavelengths of 248 and 351 nm ratherthan 193 nm. They observed growth rate enhancements of 5-15% at 450°Cwhen KrF photons were present, but no measurable effect was observed at351 nm, which supports the photochemical nature of the enhancement.The effect was attributed to photolysis of adsorbed Ga(CH 3)3 molecules;KrF was chosen rather than ArF to intentionally avoid photodissociationof AsH3 .

Several groups have also successfully revisited the growth of GaAs filmswith UV lamps. In a clever series of experiments, Kachi er al. [1101combined a cw CO 2 laser with a high-pressure Hg lamp to grow epitaxialGaAs films at temperatures down to 500°C. The role played by the CO 2laser was to vibrationally excite AsH3 at the surface, while the Hg lampserved to improve layer quality and surface morphology. Confirmation ofthe CO2 laser's involvement in the growth process was demonstrated bytuning the laser to the 10.531 Am absorption line of AsH 3, whereupon thegrowth rate was enhanced by more than a factor of two at 500C. Tuningthe CO2 laser wavelength to a region where AsH3 does not absorb(A = 9.260 Am), or directing 10.531 Am radiation parallel to the substrate,yielded no effect.

Norton and Ajmera [111, 112] deposited polycrystalline GaAs onquartz by photodissociating Ga(C2Hs) 3 in the presence of AsH 3 with anHg-Xe arc lamp or by Hg photosensitization with a 500-W low-pressureHg resonance grid lamp. All of the films were deposited at 240°C, a tem-perature at which growth will not normally occur without lamp radiation.

2. Al-containing Compounds

Recently, Tokumitsu et al. [1131 combined an ArF laser and a metal-organic MBE (MOMBE) reactor to grow epitaxial AlAs and polycrystal-line GaAIAs films at 350°C. .rystalline AlAs films up to 1.25 Am in

ut-3. PHOTOCHEMICAL VAPOR DEPOSITION 489

thickness were grown by irradiating the substrate, whereas, in the absenceof UV photons, films only -500 A thick were obtained. Similarly, 0.2-Am-thick polycrystalline GaAIAs films were deposited with surface illumi-nation. Without the UV laser beam, film thicknesses were limited to200-300 A. For these studies, the gallium precursor was triethylgallium(Ga(C 2H5 )3), and heating of elemental arsenic provided a beam of As4 atthe substrate. AlAs, GaAs, and fAl, Ga] As were deposited by Zincket al. [114] from Lewis acid-base adducts such as (CH 3)3AI: As(CH 3)3 .The vapor pressure of this adduct at 25C is -0.5 torr, and both the Ga-Asand Al-As adducts investigated in Ref. 114 absorb strongly at 193 nm butonly weakly at 248 nm. Best film morphology was obtained when theexcimer laser beam was directed parallel to the Ge substrate, but themaximum growth rates were observed to be 0.03 A per pulse. For the GaAsfilms, the As/Ga ratio was found to increase for KrF laser fluences above140 mJ-cm- 2 and, for AlAs and [Al, Ga] As films as well, the film stoi-chiometry was observed to be a function of the laser fluence. The Ga/Alratio in [Al, Gal As films, for example, was determined to be 20 for A =193 nm, a laser fluence of 320 mJ-cm- 2 and a Ga/Al ratio in the gasphase of 17. At 248 nm, however, a Ga/Al ratio (gas phase) of 14, and100 mJ-cm- 2 of fluence, the Ga/Al ratio in the deposit fell to 0.2.

3. Gallium Phosphide

Epitaxial films of the wider-bandgap binary GaP have been depositedby Sudarsan et al. [115, 116] onto GaP (100) at 500C by photodissociatinga mixture of trimethylgallium and tertiarybutylphosphine (TBP) with anArF laser. In the absence of the excimer laser radiation, a thermal filmgrowth rate of 60 A-min-1 is observed, and the material is polycrystalline.With 110 mJ-cm- 2 of 193 nm laser fluence at the substrate, however,epitaxial films are grown at a deposition rate six times larger (i.e., -360 A-min ) than that of the thermal background. Although pulsed heating ofthe substrate undoubtedly plays a role in film growth for this fluence,identical experiments carried out at 248 nm yielded only polycrystallinefilms. Since both TMG and TBP absorb less strongly at 248 nm than at193 nm, it is apparent that photodissociation of the precursors also plays arole in determining the structure of the film. Epitaxial films were obtainedonly in a narrow range of laser fluences. Below 100 mJ-cm -2 , the filmswere polycrystalline, but above 120 mJ-cm - 2 , surface ablation dominated.

4. Indium Phosphide and Indium Antimonide

Indium phosphide (along with HgTe) was one of the first semiconduc-tor compounds to be deposited by photochemical processes in the gas

490 J. G. EDEN

phase. Donnelly et al. [117-119] deposited amorphous, polycrystalline, andepitaxial InP films on quartz, GaAs or InP substrates by photodissociatingmixtures of (CH 3)3InP(CH 3)3 and P(CH3)3 with an ArF laser beam atnormal incidence to the surface. Film thicknesses typically were < 1,000 A,and deposition rates were measured to be -0.2 A per pulse. Epitaxial filmswere obtained only for laser fluences exceeding -100 mJ-cm- 2, whichsuggests the contribution of thermal effects to the deposition process. Thiswas confirmed by the fact that film growth also occurred for an excimerlaser wavelength of 351 nm (XeF), in a spectral region where both precur-sors are transparent. However, at 351 nm, the film growth rates wereroughly a factor of four smaller than those at 248 and 193 nm, whichindicates that film growth is controlled by both pyrolytic and photochemi-cal mechanisms. The highest-quality epitaxial films were grown at 320°C.

Zuhoski et al. [120] have recently reported the photodeposition ofpolycrystalline InSb films on GaAs (100) at room temperature. The pre-cursors, In(CH 3)3 and Sb(CH 3)3 , both absorb strongly at 193 nm, butthe absorption cross-section falls by an order of magnitude in goingto 248 nm. Nevertheless, o at 248 nm for In(CH 3)3 and Sb(CH 3)3 is1.2 x 10-18 and 1.7 x 10-18 cm 2 , respectively [120], and it was found thatthe highest-quality InSb films were deposited with KrF radiation(50 mJ-cm- 2) and for an Sb(CH3)3/In(CH 3)3 partial pressure ratio of 17: 1.

Vl. DIELECTRICS

Among the materials that have been photodeposited, the dielectricswere the earliest to have significant commercial impact. Photo-CVD SiO 2 ,for example, has been introduced into the processing of 4-megabitdevices [121, 122], and reactors are now commercially available for thebroad-area deposition of both the oxides and nitrides of Si by Hg photosen-sitization [351 and ArF laser photodissociation. The primary factors drivingthe introduction of photodeposited dielectrics into commercial processingare its associated step coverage and low defect density. Aside from SiO 2and Si3N4 , a variety of other dielectric films, such as the oxides ofaluminum, tin, and zinc, have been deposited by photochemical processes.

A. Aluminum Oxide

Solanki et al. [44] demonstrated in 1983 that stoichiometric A12 0 3films can be deposited uniformly over 75-mm diameter wafers at 200C byphotodissociating mixtures of trimethylalumir, m and N20 with an ArF orKrF laser beam. At a source wavelength of 248 nm and an average laserpower of 10 W, film deposition rates up to 2,000 A-min' were measured

1I-3. PHOTOCHEMICAL VAPOR DEPOSITION 491

at the substrate temperature of 350'C. While higher deposition rates weremeasured at 193 nm, it was necessary to reduce the TMA partial pressurefrom 80-120 mtorr to 30 mtorr to obtain films of uniform thickness. Thecarbon content was <1 at %, and the pinhole defect density was <0.2-cm - 2. All experiments were carried out in parallel geometry, and Deutschet al. [123] subsequently demonstrated that irradiating the growing A120 3film with 193 or 248 nm fluences as low as 1 mJ-cm- 2 measurably increasedthe film's refractive index. An increase in the index from -1.6 to 1.7, forexample, was observed when the surface was illuminated with 1.6 mJ-cm - 2

of 193-rnm radiation.Hudson et al. [124] synthesized several novel precursors from the

aluminum (-diketonate family specifically for A120 3 photodeposition.Containing both the necessary Al and oxygen atoms, these moleculesabsorb strongly over a broad spectral region in the UV, making themsuitable for photodissociation by a lamp. One of the diisopropoxy alumi-num diketonates, Al(OPr )2Acac(OMe), which contains CH3 and OCH3ligands, was studied as a precursor. Although its vapor pressure at 120°C isonly -0.03 torr, stoichiometric A120 3 films having a carbon content of <1at % were deposited with a growth rate of 0.25 jm-h- .

B. Silicon Dioxide

Silicon dioxide is the most widely deposited and thoroughly character-ized of the photo-CVD dielectrics, owing to its fundamental importance inVLSI technology. The low processing temperatures (<2000 C), reduceddefect densities, and absence of heavy ion damage characteristic of photo-deposited SiO 2 , combined with good step coverage and capacity for broad-area deposition, make this dielectric of increasing interest for incorpora-tion into commercial semiconductor processing.

The deposition of SiO 2 by Hg photosensitization was demonstrated inthe early 1980s [35, 125]. Irradiating mixtures of SiH 4 , N20, and Hg vaporyielded deposition rates up to 150 A-min- at a total pressure of 0.3 to1.0 torr and substrate temperatures as low as 50°C. Tables II and IVsummarize the reactor conditions and electrical and structural characteris-tics of the deposited dielectric films [35]. More recent studies by Padma-nabban et al. [126] with a commercially available reactor examined siliconoxide deposition at 100 and 200*C. Deposition rates in excess of 150 A-min-' were again observed, but pinhole densities ranging from -200-cm 2

to 600-cm -2 and refractive indices between -1.5 and 2.0 were measuredfor 0.3-Mm-thick films grown at 200*C. Depending on the growth condi-tions, intermediate oxide phases such as Si 20 3 , SiO, and Si20 were alsopresent.

492 J. G. EDEN

TABLE IVCHARACTERISTICS OF SILICON OXIDE AND SILICON NITRIDE FILMS DEPOSITED BY

PHOTO-CVD IN A COMMERCIAL REACTOR (AFTER REF. 35)

Film

Characteristic Silicon dioxide Silicon nitride

Density 2.10 g/cm3 1.8 to 2.4 g/cm-"

Index of refraction 1.45-1.490 1.8-2.4"Dielectric constant 3.9 5.5'Dielectric strength 4.0-6.0 x 106 V/cm 4 x 10 V/cm"Defect density < 10/cm2 <10/cm 2

Film adhesion (tensile) >5.2 x 10 torr (10 lb/in2 ) 3.1-5.2 x 10' torr(0.6 to 1 x i03 lb/in 2)

Etch rate 9 nm-s - 6 to 10 nm-s -1Film composition SiO 2 Si OWN

'Film composition characteristics can be varied by reactant gas ratio.'Buffered oxide etch, film densified (15 min at 450rC in N2).'1: 10 HI, as deposited.

Mishima et al. [127 reported the direct photodeposition of SiO 2 frommixtures of SiH 4 and 02 with a low-pressure Hg lamp. For each tempera-ture studied in the 150-350C range, photo-CVD growth rates exceededthose obtained by pyrolysis. At 300C, the photochemical and thermaldeposition rates were >500 and -350 A-min- , respectively. Also, sincethe activation energy for the photo-CVD process in the 150 to -280°Crange (0.53 eV) is roughly half that for thermal deposition (0.96 eV),the difference between the photo and thermal deposition rates becomesmore pronounced at lower substrate temperatures. At 150C, the rates dif-fer by more than an order of magnitude. Inushima et al. [122] used a low-pressure Hg lamp to obtain a deposition rate of 200 A-min-' from SiH 4and N20 at a total pressure of 30 torr.

Scoles et al. [128] carried out similar experiments with SiH 4 and 02 withan Hg-Xe arc lamp acting as the optical source. For substrate tempera-tures between 150 and 350 0C, deposition rates ranging from 230 to 1140 A-min-' were recorded, and the electrical and structural properties of thefilms (such as the dielectric constant, breakdown voltage and capacitance)were found to be comparable to those for thermally deposited films.

Several groups have also explored the D2 lamp as an excitation source.Silicon dioxide films were deposited by Okuyama et al. [129] fromSi 3H 8/0 2 mixtures at a rate of 150 A-min- at 25°C. The deposition ratedeclined with increasing temperature, which was attributed to desorptionof an adlayer of lamp-produced radical species. Nonaka et al. [1301 havereported significant increases (i.e., factors of 2-4) in the deposition rate ofamorphous SiO 2 films when Si2F6 is added to the feedstock gas stream. At

m-3. PHOTOCHEMICAL VAPOR DEPOSITION 493

200C and an Si2H6 mass flow rate of 2.4 sccm, for example, the SiO 2deposition rate rose linearly from -140 A-min- 1 to almost 600 A-min- 1

when -3.5 sccm of Si2F6 was introduced to the gas stream. Similar im-provements in deposition rate (factors of two to three) were also observedat other disilane mass flow rates, and the presence of Si2F6 dramaticallyreduced the number of Si-OH and Si-H defects in the dielectric film.

Efforts to deposit Si0 2 with shorter-wavelength optical sources includethe work of Baker et al. [131] with a windowless N2 lamp. At 275C, thedeposition rate was 180 A-min- and the refractive index of the films was1.46. Although the deposition process is initiated by lamp radiation, ato-mic nitrogen produced in the lamp also appears to play a significant role.Thin film MOS field effect transistors fabricated from such photo-CVD,amorphous SiO 2 films yielded a gate leakage current of 1 pA. Si0 2 filmshave also been deposited with VUV resonance radiation at 106.6 nm froma microwave-excited Ar lamp. Marks and Robertson [1321 photodissoci-ated mixtures of NO 2 and SiH 4 at 100°C (and up to 2500C) with the lamp toobtain deposition rates of 100 A-min - on Si; the resulting films exhibitedrefractive indices between 1.45 and 1.47.

While most SiO 2 deposition studies have relied on lamps as the opticalsource, Boyer et al. [133, 134] resorted to an ArF (193 nm) laser in parallelgeometry to deposit films at rates up to 3,000 A-min- 1. The deposition ratewas insensitive to substrate temperature in the 20-600'C temperatureinterval but was linear in gas pressure and laser intensity. Transparent,scratch-resistant films were obtained at temperatures above 200'C, and therefractive indices ranged from 1.476 (250°C) down to 1.457 (350°C).

Oxynitrides of Si have been deposited by adding ammonia to a disilaneand nitrogen dioxide (NO 2) gas stream. Watanabe and Hanabusa [1351showed that the SiONy composition is quite sensitive to the NO 2 mass flowrate, which is the smallest of those for the three precursors. Depositionrates up to 120 A-min-' were reported.

C. Silicon Nitride

Although photodeposited Si 3N4 has not been as thoroughly character-ized as its oxide counterpart, the growth conditions for the two dielectricsare similar. Several groups [t5, 136] recognized that Hg photosensitizationof SiH 4 in the presence of NH 3 would yield silicon nitride films. Utilizing aprocess developed by Hughes Aircraft, Schuegraf [35] obtained depositionrates up to 65 A-min - in the 100-200°C temperature range with Hgimpurity concentrations in the resulting films of several parts per billion.Also, the film index of refraction is adjustable between 1.8 and 2.4 byvarying the gas mixture ratio. Further details regarding the deposition pro-cess parameters and film characteristics can be found in Tables II and IV.

494 J. G. EDEN

Inushima et al. [122] photodeposited SiN films by photodissociatingNH 3 and Si2H6 (in the absence of Hg vapor) with the 185 nm line from anHg resonance lamp. For substrate temperatures between 300 and 350°C,the deposition rate was 40 A-min - ', the hydrogen content was 20 at %,and the refraction index for the films was 2.0. Also, the deposition ratewas independent of substrate temperature, and the film thickness wasuniform to ±5%.

Early demonstrations of the ability of excimer laser radiation to depositsilicon nitride were carried out by Boyer et al. [1371. With the ArF laserbeam parallel to the substrate, they obtained deposition rates up to 700 A-min - , and the physical properties of the photodeposited films were foundto be comparable to those of plasma CVD films, but of higher quality(particularly insofar as pinhole densities are concerned) than those of filmsdeposited by Hg photosensitization. Weak illumination of the substratewith 254 or 193 nm photons ("surface photons") reduced the etch rate ofthe laser-deposited nitride films by more than a factor of five.

Obtaining stoichiometric silicon nitride (Si3N4) is hindered by the pres-ence of oxygen in most of the films reported to date. Sugii et al. [1381have explored this by photodissociating mixtures of Si2H6 and NH3 withan ArF laser in perpendicular geometry. Below 400°C, film growth isprimarily photochemical, and the deposition rate is roughly constant at50 A-min - . Thermal effects become dominant at higher temperatures,and at 550*C, the deposition rate has risen to -200 A-min-1. Auger elec-tron spectroscopy reveals the composition of the photodeposited films tobe Si3O N , where the nitrogen content gradually declines for substratetemperatures above 2000C, with y falling from -4 at 200*C to -3.5 at400C. Recently, Urisu et al. [139] have deposited hydrogenated siliconnitride films at 190*C with radiation from a synchrotron. Since VUV andextreme ultraviolet (XUV, A ! 100 nm) photons are available with thisdevice, precursors having stronger chemical bonds (i.e., more stable) areaccessible, and N2, rather than ammonia, can be used as the nitrogenprecursor. With a mixture composed of 0.02 tort SiH 4 and 0.1 tort N2.deposition rates of 4 A-min- 1 were obtained for a ring current of 100 mA.

D. Other Oxides and Nltrides

Several other oxides have been deposited by photo-CVD. Matsui andco-workers [140] successfully deposited TaO,, films from mixtures of TaCl5and 02. At 500C, the growth rate was >30 A-min- for a total reactorpressure of 1 tort. Also, TaO1, deposition rates are inversely proportionalto growth temperature but, without the external UV radiation. the deposi-tion rate falls to 2 A-min - ' at 4000C. At 350C. Yamagishi and Tarui


[141] deposited amorphous TaO (x 0.2) films at a rate of 120 A-in- ' byphotodissociating pentamethoxy tantalum, Ta(OCH 3)5 , in the presence of02. X-ray analysis revealed that the nonstoichiometric films were com-posed of TaO2 and Ta2 05 .

Tin dioxide films 1,000 A thick have been deposited from mixtures ofSnC 4 vapor and N20. Kunz et al. ff421 irradiated fused silica substratesat room temperature and normal incidence in the presence of such a mix-ture, and for laser energy fluences of 10-20 mJ-cm- 2, highly conductiveSnO 2 films resulted. Resistivities as low as 4 x 10-2 fl-cm were measured.At lower fluences, SnOCI2 is formed and the film resistivity rises rapidly.For fluences beyond 70 mJ-cm - 2, in contrast, ablation becomes dominantand the net deposition rate vanishes. Lines of 10 jzm width were also de-posited by proximity printing through a mask.

Reduction of the growth temperature of TiN films and 35-300% in-creases in the film deposition rate were realized by Motojima andMizutani [143] by irradiating TiC14 , NH 3 (or N2) and H2 gas mixtures witha deuterium lamp. In the absence of the UV radiation, and using N2 as thenitrogen donor, substrate temperatures above 800"C were required inorder to observe film growth. With the lamp, however, film depositionoccurred at 700(C and, at 900"C, the deposition rate, 2.1 /m-h-1, was afactor of three larger than that observed without the UV source. Substitut-ing NH 3 for N2 as the N precursor enables film deposition to occur attemperatures as low as 450"C, and at 8000C. the TiN deposition rate(55 Mm-h-') is -35% higher than that measured without the lamp radia-tion. Consequently, the addition of the D2 lamp to the TiN reactor permit-ted deposition temperatures to be reduced 50-100"C below those normallyrequired.

VII. CONCLUSIONS

Although photochemical vapor deposition is still in its infancy, thequality of the metal, semiconductor, and dielectric films obtainable by thistechnique has progressed rapidly. This is especially true of silicon dioxide,which has already been incorporated into the processing of VLSI devices.As increasing effort is invested in synthesizing new precursors that aretailored for the deposition of specific elements or compounds, it can beexpected that the broad-area deposition of other dielectrics and severalmetals will be the next to become economically attractive. Further en-gineering development of new, high duty cycle lasers and lamps operatingin the deep UV and near VUV. such as the new rare gas ion lasertransitions and the microwave-excited rare gas halide lamp discussed inSection III.C, will undoubtedly accelerate this trend.

496 J. G. EDEN

ACKNOWLEDGMENTS

The technical assistance of K. Voyles, D. Terven. and G. Woodcock and discussions withF. A. Houle, C. 1. Kiely, D. B. Geohegan, I. P. Herman. and S. D. Allen are gratefullyacknowledged. This work was supported by the Air Force Office of Scientific Research (H. R.Schlosberg) and the National Science Foundation through the University of Illinois MaterialsResearch Laboratory.

REFERENCES

1. H. Romeyn, Jr., and W. A. Noyes, Jr., J. Am. Chem. Soc. 54, 4143 (1932).2. H. J. Emeltus and K. Stewart, Trans. Faraday Soc. 32, 1577 (1936).3. R. M. Osgood. Jr., Ann. Rev. Phys. Chem. 34, 77 (1983).4. R. M. Osgood. Jr., and T. F. Deutsch, Science 227, 709 (1985).5. Y. Rytz-Froidevaux, R. P. SalathE, and H. H. Gilgen. Appi. Phys. A 37, 121 (1985).6. F. A. Houle. App. Phys. A 41, 315 (1986).7. J. Haigh and M. R. Aylett. Prog. Quant. Electr. 12, 1 (1988).8. 1. P. Herman. Chem. Rev. 89, 1323 (1989)9. D. Bliuerle, "Chemical Processing With Lasers," Vol. 1 of Springer Series in Materials

Science Springer-Verlag, Berlin, 1986.10. 1. W. Boyd, "Laser Processing of Thin Films and Microstructures," Vol. 3 in Springer

Series in Materials Science. Springer-Verlag, Berlin. 1987.11. "Laser Microfabrication: Thin Film Processes and Lithography," D. J. Ehrlich and

1. Y. Tsao, eds. Academic Press, New York, 1989.12. "Laser Chemical Processing for Microelectronics," K. G. Ibbs and R. M. Osgood, Jr.,

eds., Cambridge, 1989.13. V. R. McCrary and V. M. Donnelly, in "Chemical Vapor Deposition," K. F. Jensen.

ed., Academic Press, New York, in press.14. D. . Ehrlich, R. M. Osgood, Jr., and T. F. Deutsch, J. Vac. Sci. Technol. 21, 23

(1982).15. F. A. Houle and S. D. Allen. "Laser Deposition and Etching," American Vacuum

Society Education Committee-course notes.16. U. Itoh. Y. Toyoshima, H. Onuki, N. Washida, and T. Ibuki, J. Chem. Phys. 85, 4867

(1986).17 J. Heidberg, R. Daghighi-Ruhi. H. von Weyssenhoff. and A. Habekost, Mat. Res. Soc.

Symp. Proc. 101. 221 (1988).18. H. Okabe, "Photochemistry of Small Molecules," John Wiley and Sons, New York,

1978.19. M. Rothschild in Ref. 11, pp. 163-230, and refs. cited therein.20. Th. Beuermann and M. Stuke, Appl. Phys. 8 49, 145 (1989).21. G. S. Higashi and M. L. Steigerwald, Appl. Phys. Let. 54, 81 (1989).22. Y. Zhang and M. Stuke, Jpn. J. App. Phys. 27, L1349 (1988).23. Y. Zhang, Th. Beuermann, and M. Stuke. Appi. Phys. B 48, 97 (1989).24. A. R. Calloway, T. A. Galantowicz. and W. R. Fenner. J. Vac. Sci. Technol. A I, 534

(1983); A. R. Calloway and W R. Fenner, Aerospace Corp. Report No. ATR.84(8503)-1(1984).

25. A. Terenin and B. Popov, Phys. Z. Sowjetunion 2, 299 (1932).26. B. Koplitz, Z. Xu, and C. Wittig, Appl. Phys. Let. 52, 860 (1988).27. K. Kamisako, T. lmai. and Y Tarui, Jpn. J. Appl. Phys. 27, 1092 (1988).28. C. Fuchs. E. Boch, E. Fogarassy. B. Aka. and P. Siffert, Mat. Res. Soc. Symp. Proc.

101. 361 (1988).

i1-3. PHOTOCHEMICAL VAPOR DEPOSITION 497

29. M. Sasaki, Y. Kawakyu, and M. Mashita, Jpn. J. Appl. Phys. 28, L131 (1989).30. Y. Zhang and M. Stuke, J. Cryst. Growth 93, 143 (1988).31. D. J. Ehrlich, R. M. Osgood, Jr., and T. F. Deutsch. Appl. Phys. Lett. 38, 946 (1981).32. J. Y. Tsao and D. J. Ehrlich. App. Phys. Lett. 45, 617 (1984).33. W. L. Ahlgren, E. J. Smith, J. B. James, T. W. James, R. P. Ruth. and E. A. Patten,

J. Cryst. Growth 86, 198 (1988).34. H. Zarnani, Z. Q. Yu, G. J. Collins, E. Bhattacharya, and J. 1. Pankove, Appl. Phys.

Lett. 53, 1314 (1988).35. K. K. Sch',egraf, in "Microelectronic Manufacturing and Testing," pp. 1-4. Lake,

Libertyville. March 1983.36. R. Solanki and G. J. Collins, Appl. Phys. Lett. 42, 662 (1983).37. D. B. Geohegan and J. G. Eden, App. Phys. Len. 45, 1146 (1984).38. H. Kumagai and M. Obara, App. Phys. Lett. 55, 1583 (1989).39. R. L. Jackson, T. H. Baum, T. T. Kodas, D. J. Ehrlich, G. W. Tyndall, and P. B.

Comita, in Ref. 11.40. M. W. Jones, L. I. Rigby, and D. Ryan, Nature 212, 177 (1966).41. T. F. Deutsch, D. J. Ehrlich, and R. M. Osgood, Jr., App. Phys. Let. 35, 175 (1979).42. W. E. Johnson and L. A. Schlie, App. Phys. Len. 40, 798 (1982).43. T. Cacouris, G. Scelsi, P. Shaw, R. Scarmozzino, R. M. Osgood, and R. R. Krchnavek,

Appl. Phys. Lett. 52, 1865 (1988).44. R. Solanki, W. H. Ritchie, and G. J. Collins. Appl. Phys. Lett. 43, 454 (1983).45. M. Hanabusa, K. Hayakawa, A. Oikawa, and K. Maeda, Jpn. J. Appl. Phys. 27, L1392

(1988).46. M. Hanabusa. A. Oikawa. and P. Y. Cai. J. App!. Phys. 66, 3268 (1989).47. M. R. Aylett and J. Haigh, in "Laser Diagnostics and Photochemical Processing for

Semiconductor Devices," R. M. Osgood, Jr.. S. R. J. Brueck, and H. R. Schlossberg,eds., Mat. Res. Soc. Symp. Proc., 17, p. 177. Elsevier, 1983.

48. Y. Rytz-Froidevaux, R. P. Salatht. and H. H. Gilgen, Mat. Res. Soc. Symp. Proc. 17,29 (1983).

49. R. Solanki, P. K. Boyer, and G. J. Collins, App. Phys. Ler. 41, 1048 (1982) and refs.cited therein.

50. D. J. Ehrlich, R. M. Osgood, Jr., and T. F. Deutsch, J. Electrochem. Soc. 128, 2039(1981).

51. T. F. Deutsch and D. D. Rathman, App!. Phys. Lett. 45, 623 (1984).52. H. H. Gilgen. T. Cacouris, P. S. Shaw, R. R. Krchnavek. and R. M. Osgood, Jr..

App. Phys. B 42, 55 (1987).53. J. Y. Tsao, R. A. Becker, D. J. Ehrlich. and F. J. Leonberger, App. Phys. Ler. 42,

559 (1983).54. P. M. George and J .L. Beauchamp, Thin Solid Films 67, L25 (1980).55. M. S. Chiu, K. P. Shen, and Y. K. Ku, Appl. Phys. B 37. 63 (1985).56. Q. Mingxin. R. Monot. and H. van den Bergh, ScientUa Sinica A27, 531 (1984).57. J. V. Armstrong, A. A. Burk, Jr., J. M. D. Coey, and K. Moorjani. Appl. Phys. Le.

50, 1231 (1987).58. L. V. Koplitz, D. K. Shuh, Y. -J. Chen. R. S. Williams, and J. 1. Zink, Appl. Phys.

Lett. 53, 1705 (1988).59. C. R. Jones. F. A. Houle, C. A. Kovac. and T. H. Baum. Appl. Phys. Let. 46, 97

(1985).60. T. H. Baum. E. E. Marinero, and C. R. Jones. Appl. Phys. Lett. 49, 1213 (1986).61. M. R. Aylett and J. Haigh, Extended Abstracts of Mat. Res. Soc. Symp., Boston.

Massachusetts. 1985.

498 Ji. G. EDEN

62. J. Haighs and M. R. Aylett, in Ref. 11.63. H. Tachibana, A. Nakaue, and Y. Kawate. Mat. Res. Soc. Symp. Proc. 101, 367 (1988).64. H. Obashi. K. Ioue. Y. Saito, A. Yoshida, H. Ogawa. and K. Shobatake. App!. Plays.

Lea. 55, 1644 (1989).65. H. Niki and G. J. Mains. L. Plays. Chem. 68, 304 (1964).66. T. Saitoh. T. Shimada, M. Migitaka, and Y. Tarui. J. Non-Cryst. Solids 59, 715 (1983).67. Y. Tarw., K. Sonimachi, K. Fujii. K. Aota, and T. Saito, J. Non-Cryst. Solids 59/60, 711

(1963).68. S. Mizukawa. K. Sato, K. Yasuhiro, M. Isawa, K. Kuroiwa, and Y. Tarui. Jpn. J. App!.

Phys. 23, 961 (1989).69. T. Inoue, M. Konagai, and K. Takahashi, App!. Plays. Lett. 43, 774 (1983).70. S. Nishida. H. Tasaki, M. Konagai. and K. Takahashi, J. App!. Phys. 58, 1427 (1985).71. T. Tatsuya, W. Y. Kim, M. Konagai, and K. Takahashi. App!. Phys. Lett. 45, 865

(1984).72. S. Nishida, T. Shuimoto, A. Yamada, S. Karasawa. M. Konagai, and K. Takahashi.

App!. Plays. Lett. 49, 79 (1986).73. Y. NMishima, M. Hirose, Y. Osaka, K. Nagamine. Y. Ashida, N. Kitagawa, and

K. Isogaya. Jpn. J. App!. Plays. 22, L46 (1983).74. N. Gonohe. S. Shimizu, K. Tamagawa, T. Hayashi. and H. Yamakawa. Jpn. J. App!.

Phys. 26. L1189 (1987).75. T. Fuyuki. K. -Y. Du, S. Okamoto, S. Yasuda. T. Kimoto. M. Yoshimoto, and

H. Matsunami, J. App!. Plays. 64,.2380 (1988).76. K. Kumata, U. Itoh, Y. Toyoshima, N. Tanaka. H. Anzai, and A. Matsuda, Appi.

Phys. Let. 48, 1380 (1986).77. W. -Y. Kim. A. Shibata, Y. Kazama, M. Konagai, and K. Takahashi. Jpn. 1. App.

Plays. 23, 311 (1989).78. R. W. Andreatta. C. C. Abele. J. F. Osmundsen. J. G. Eden, D. Lubben, and J. E.

Greene. App!. Plays. Len. 40, 183 (1982).79. A. Yoshikawa and S. Yamaga. Jpn. J. App!. Plays. 24, 1585 (1985).80. H. Zarnan,. H. Demiryont. and G. J. Collins, J. App!. Plays. 60, 2523 (1986).81. A. Yamada, A. Satoh. M. Konagai. and K. Takahashi, J. App!. Plays. 65, 4268 (1989).82. R. G. Frieser. 1. Electrochem. Soc. 115S, 401 (1968).83. A. Ishitani, Y. Oshita. K. Tanigaki. and K. Takada. J. App!. Phys. 61, 2224 (1987).84. . G. Eden, J. E. Greene, J. F. Qsmundsen. D. Lubben, C. C. Abele. S. Gorbatkin.

and H. D. Desai. M~at. Res. Soc. Symp. Proc. 17, 185 (1983).85. C. J. Kiely, V. Tavitian, and J. G. Eden, J. App!. Phys. 65, 3883 (1989).86. C. J. Kiely, V. Tavitian, C. Jones. and J. G. Eden. App!. Plays. Lett. 55, 65 (1989).87. D. H. Lowndes, D. B. Geohegan, D. Eres, S. J. Pennycook. D. N. Mashburn. and

G. E. Jelison, Jr., App!. Plays. Lett. 52, 1868 (1988).88. H. H. Burke. 1. P. Herman, V. Tavitian, and J. G. Eden. App!. Plays. Lett. 55, 253

(1989).89. A. E. Stanley, R. A. Johnson, J. B. Turner. and A. H. Roberts. App!. Spectroscopy

40, 374 (1986).90. S. I. C. Irvine, in Ref. 11, pp. 503-538.91. H. Ando. H-. Inuzuka. M. Konagai. and K. Takahashi, J. App!. Phys. 58, 802 (1985).92. Y. Fujita. S. Fujii. and T. luchi. J. Vac. Sca. Trechnol. A 7, 276 (1989).93. 0. B. Shinn, P. M. Gillespie. W. L. Wilson. Jr.. and W. IM. Duncan. App!. Plays. Let.

54, 2440 (1989).94. S. J. C. Irvine. J. B. Mullin. and J. Tunnicliffe. I. Crvst. Growth 66, 188 (1984).95. W. 1. Ahigren. J. B. James. R. P Ruth. E. A. Patten. and 3. L. Staudenmann. Mat.

Res. Soc. Symp. Proc. 90, 405 f 1497).

mi-3. PHOTOCHEMICAL VAPOR DEPOSITION 499

96. D W. Kisker and R. D. Feldman. Mat. Let. 3, 485 (1985).97. D. W. Kisker and R. D. Feldman, J. Cryst. Growth 72, 102 (1985).98. N. W. Cody, U. Sudarsan. and R. Solanki. J. Mater. Res. 3, 1144 (1988).99. N. W. Cody, U. Sudarsan. and R. Solanki. J. AppI. Phys. 66, 449 (1989).

100. J. J. Zinck. P. D. Brewer, J. E. Jensen, G. L. Olson, and L. W. Tut. Appi. Phys. Let.52. 1434 (1988).

101. J. J. Zinck, P. D. Brewer, J.- E. Jensen, G.- L. Olson. and L. W . Tutt. Mat. Res. Soc.Symp. Proc. 101, 319 (1988).

102. S. J. C. Irvine. J. B. Mullin, G. W. Blackmore, and O.D. Dosser, J. Vac. Sci. Technol.8 3, 1450 (1985).

103. S. J. C. Irvine, J. B. Mullin, H. Hill. G. T. Brown, and S. J. Barnett, J. Cryst. GrowthSG, 188 (1988).

104. B. J. Morrs, Appi. Phys. Lett. .48, 867 (1986).105. N. Piltz, H. Heinecke. E. Veuhoff, G. Arens, M. Heyen. H. Lulth. and P. Balk. J. Cryst.

Growth 68, 194 (1984).106. p. Balk. M. Fischer, D. Grundmann. R.. Luickerath, H. Ldth, and W. Richter. J. Vac.

Sci. Technol. 8 5, 1453 (1987).107. 1. Nishizawa. H. Shimawaki, and Y. Sakuma. 1. Electrochem. Soc. 133, 2567 (1986),

and refs. cited therein.108. S. S. Chu, T. L. Chu, C. L. Chang, and H. Firouzi. Appi. Phys. Lett. 52, 1243 (1988).109. P. K. York, J. 0. Eden, J. J. Coleman. G. E. Ferna1ndez. and K. J. Beernink, Appi.

Phys. Lett. 54, 1866 (1989).110. T. Kachi. H. Ito, and S. Terada. Jpn. J. Appi. Phvs. 27, L1556 (1988).Ill. D. P. Norton and P. K. Ajmera. App. Ph 've. Lett. 53, 595 (1988).112. D. P. Norton and P. K. Ajmera. J. Electrochem. Soc. 136, 2371 (1989).113. E. Tokumitsu. T. Yamada. M. Konagai, and K. Takahash, J. Vac. Sdi. Tech.-to. A 7,

706 (1989); Mat. Res. Soc. Symp. Proc. 101, 307 (1988).114. J. J. Zinek, P. D. Brewer, J. E. Jensen. G. L. Olson. and L. W. Tutt. Mat. Res. Soc.

Symp. Proc. 75, 233 (1987).115. U. Sudarsan. N. W. Cody, T. Dosluoglu. and R. Solanki. Appi. Phys. Lett. 55, 738

(1989).116. U. Sudarsan, N. W. Cody. T. Dosluoglu. and R.. Solanki. App!. Phys. A. in press.117. V. M. Donnelly. M. Geva. J. Long, and R. F. Karlicek. Appi. Phys. Lett. 44, 951

(1984).118. V. M. Donnelly, D. Brasen. A. Appelbaum, and M. Geva. J. App. Ph vs. 58, 2022

(1985).119. V. M. Donnelly. D. Brasen. A. Appelbaum. and M. Geva. 1. Vac. Sci. Technol. A 4,

716(1986).120. S. P. Zuhoski. K. P. Killeen. and R. M. Biefeld. Mat. Res. Soc. S~ymp. Proc. 101, 313

(1988).121. K. Yano. S. Tanimura. T. Ueda. and T. Fujita. 1987 Symposium on VLSI Technology.

Business Center for Academic Society of Japan. Karuizawa. Japan (1987). p. 71.122. T. Inushima. N. Hirose. K. Urata. K. Ito, and S. Yamazaki. Appi. Phys. A 47, 229

(1988).123. T. F. Deutsch. D. J1. Silversmith, and R. W. Mountain. Mat. Res. Soc. Sjymp. Proc. 29,

67 (1984).124. M. D. Hudson. C. Trundle, and C. J. Brierley. J. Mater. Res. 3, 1151 (1988).125 R. F. Sarkozy. in "Technical Digest of 1981 Symposium on VLSI Technology." p. 68.

Japan Society of Applied Physics. Hawaii. 1981.126. R. Padmanabhan. B. J. Miller. and N. C. Saba. Mat. Res. Soc. Svmp. Proc. 101. 385

(1988).

500 J. G. EDEN

127. Y. Mishima. M. Hirose, Y. Osaka, and Y, Ashida. J. Appl. Phys. 55, 1234 (1984).128. K. J. Scoles. A. H. Kim. M. -H. Jiang, and B. C. Lee, J. Vac. Sci. Technol. B 6, 470

(1988).129. M. Okuyama, N. Fujiki, K. Inoue, and Y. Hamakawa, Jpn. 1. Appl. Phys. 26, L908

(1987).130. H. Nonaka, K. Arai, Y. Fujino, and S. Ichimura, J. Appl. Phys. 64, 4168 (1988).131. S. D Baker, W. I. Milne, and P. A. Robertson, Appl. Phys. A 46, 243 (1988).132. J. Marks and R. E. Robertson, Appl. Phys. Lett. 52, 810 (1988).133. P. K. Boyer. A. Roche, W. H. Ritchie, and G. J. Collins, Appl. Phys. Lett. 40, 716

(1982).134. P. K. Boyer, K. A. Emery, H. Zarnani, and G. J. Collins, Appl. Phys. Let. 4S, 979

(1984).135. J. Watanabe and M. Hanabusa, J. Mater. Res. 4, 882 (1989).136. 1. W. Peters, in -Technical Digest of the International Electron Devices Meeting 1981,"

p. 240. Institute of Electrical and Electronics Engineers, New York, 1981.137. P. K. Boyer. C. A. Moore, R. Solanki, W. K. Ritchie, G. A. Roche, and G. J. Collins,

Mat. Res. Soc. Symp. Proc. 17, 119 (1983).138. T. Sugii. T. Ito, and H. Ishikawa, Appl. Phys. A 46, 249 (1988).139. T Urisu, H. Kyuragi, Y. Utsumi, J. Takahashi, and M. Kitamura, Rev. Sci. Instrum.

60, 2157 (1989).140. M. Matsi. S. Oka. K. Yamagishi, K. Kuroiwa, and Y. Tarui, Jpn. J. Appl. Phys. 27,

506 (1988).141. K. Yamagishi and Y. Tarui, Jpn. J. Appl. Phys. 25, L306 (1986).142. R. R. Kunz. M. Rothschild, and D. J. Ehrlich. Appl. Phys. Lett. 54, 1631 (1989).143. S. Motojima and H. Mizutani. Appl. Phys. Lett. 54, 1104 (1989).

United States Patent o1 (II) Patent Number: 5,023,877Eden et al. (45] Date of Patent: Jun. 11, 1991

[54) MINIATURE. OPTICALLY PUMPED Attorney. Agent or Firm- Irwin P. Garfinkle, Donald J.NARROW LINE SOLID STATE LASER Singer

[75) Inventors: James G. Eden, Mahomet; Mark J. [57] ABSTRACTKushner. Urbana. both of Ill. A miniature solid state laser is optically pumped by

[73] Assignee: The United States of America as ultraviolet radiation produced by a surface or coronarepresented by the Secretary of the discharge. The device is monolithic in that both theAir Force, Washington. D.C. optical pump and the active medium are contiguous to

a base material which is optically transparent in the[21] AppI. No.: 546,462 ultraviolet range. The disclosed device consists of a[221 Filed: Jun. 29, 1990 substrate, a thin metal halide or crystalline or amor-

phous film deposited on the bottom of the substrate and[31] [at. C1.5 ................................................ H01S 3/30 a pair of spaced electrodes on the top of the substrate.[521 U.S. CL ........................................... 372/7; 372/39 When a high voltage pulse is applied across the spaced(58) Field of'Search ...................................... 372/7, 39 electrodes, discharge occurs between the electrodes

and produces strong ultraviolet emissions which propa-[561 References Cited gate through the substrate and dissociates the metal

U.S. PATENT DOCUMENTS halide molecules in the film. This causes laser emission3.573.653 4/1971 Smiley ............................. 372/7 to occur on an atomic transition of the metal atom. The3.573.654 4/1971 Smiley . ..... 372/7 active laser medium may also be a crystalline platelet or3.579.130 5/1971 Smiley ......................... 372/7 amorphous thin fidm containing an impurity atom or3.579.142 5/1971 Smiley .................................... 372/7 molecule (an example would be YAG doped with Nd3.747.021 7/1973 Smiley ........ ............ 372/7 or LiYFi(YLF) doped with cerium). In this case, the3.967.213 6/197 Yariv ............................ 331/94.5 C active medium may provide its own mechanical sup-4.002.998 1/1977 Conwedl et al ............ 372/7 port, provides a surface on which the discharge may4.087,764 i;7i Young ..................................... 372/7 occur and thus eliminates the need for a separate optical4.395.769 7/1983 Damen et a ............. 37Z/7 substrate.4,827.479 5/1989 Campbell et al ...................... 372/5

Primary Examiner-I.4on Scott, Jr. 27 Claims, I Drawing Sheet

PULSED ELEC. 20

k"6 -22 '18

14*12/

U.S. Patent June 11, 1991 5,023,877

PULSED ELEC. 20 PULSED ELEC. 20

POWER SUPPLY POWER SUPPLY

12 FIG. 1 FIG. 2

PULSED ELEC. 2040 214 "40 0aS M -1

FIG. 3 FIG. 4

FIG. 5 FIG. 5

FIG.E 5LFIG.26

5,023,877 2through the substrate and photodisssociates the metal

MINIATURE. OPTICALLY PUMPED NARROW halide molecules on the film which causes laser emissionLINE SOLID STATE LASER to occur on a metal transition of the metal atom. For

thallium. for example. this occurs at a wavelength ofSTATEMENT OF GOVERNMENT INTEREST 5 535 nm (green) and in atomic indium at a wavelength of

The invention described herein may be manufactured 451 nm (blue). The active laser medium may also be a

and used by or for the Government for governmental crystalline platelet or film of a material, such as YAG

purposes without the payment of any royalty thereon. doped with Nd. which is directly pumped by the optical

This invention relates to miniature solid state lasers emission from the discharge. In the case of a platelet.

that are optically pumped by the ultraviolet (UV) radia- 10 the substrate may no longer be necessary since the laser

tion emitted by a surface or corona discharge. The laser medium (platelet) now provides its own mechanicalis monolithic in that both the optical pump and the . support.active laser medium are contiguous to a *'base" or sub-strate material that is optically transparent in the visible. OBJECTS OF THE INVENTION

ultraviolet (UV) and in some cases, at least, a portion of 1S It is an object of this invention to provide a solid statethe vacuum ultraviolet (VUV) region. miniature laser that is optically pumped by the ultravio-

let radiation emitted by a surface or corona discharge.BACKGROUND OF THE INVENTION Another object of this invention is to provide a laser

The most relevant known background art is de- in which the optical pump and the active medium arescribed in a paper by Schmiele. Luthy. Henchoz and 20 both contiguous to a base material that is opticallyWeber. entitled Miniaturization of a Thallium Iodide transparent.Photodissociation Laser, published in Applied Physics. Still another object of this invention is to provide aVolume 29. pp. 201-3 in 1982. Schmiele et al. describe monolithic, miniature laser comprising a metal halidea miniature Thallium Iodide photodissociation gas laser laser material deposited as a film on one side of a trans-in which the active medium was in the form of a thin 25 parent base material, and a pair of electrodes on thelayer on a quartz wall. An ArF excimer laser was used other side for producing radiation that passes throughto pump the active medium in the vapor phase. The the base material to photodissociate the molecules of theSchmiele et al. laser differs from the present invention laser material to produce lasing. The active laser me-in that it was pumped by another laser, (an ArF laser). dium may also be a thin film or platelet of solid stateand it used an active medium in the vapor phase. 30 laser material.

A search of the patented background art revealedU.S. Pat. No. 3.573.653 issued to Smiley. U.S. Pat. No. BRIEF DESCRIPTION OF THE DRAWINGS3.967,213 issued to Yariv, and U.S. Pat. No. 4,827,479 For a more complete understanding of the objects.issued to Campbell cc al.

Smiley discloses a tunable thin film laser consisting of 35 nature and advantages of this invention, reference

laser material that is excited by a source which is de- should now be made to the accompanying drawings. in

scribed as comprising "an appropriate noncoherent which:

light source such as a gas discharge device or alterna- FIG. I is a diagrammatic representation of one em-

tively a source of coherent excitation energy such as bodiment of this invention.

another laser." Conductive means are positioned on . FIG. 2 is a diagrammatic representation of a second

either side of the laser material and insulated therefrom, modification of this invention;

and a variable electric field is generated between the FIG. 3 is a diagrammatic representation of another

electrodes to provide tuning for the laser output. Yariv modification of this invention:

discloses an X-ray laser comprising a single crystal in FIG. 4 is a diagrammatic representation of yet an-

the form of a thin film with an oriented set of prominent 45 other modification of this invention: and

atomic planes so that when the crystal is excited by an FIGS. 5 and 6 are diagrammatic representations of

appropriate pumping source, such as electrons, X-rays, another modification of this invention in which the

or intense laser radiation, X-ray photons, which are substrate is formed as a lens.

emitted from the crystal, experience internal feedback DESCRIPTION OF THE PREFERREDfrom the atomic planes. thereby eliminating the need for 50 EMBODIMENTSexternal feedback. In addition, the crystal functions as athin planar waveguide confining the X-ray waves Referring to FIG. 1, a miniaturized laser 10 corn-therein. thereby reducing the necessary pumping power prises, a thin film 12 of a metal monohalide, such asand increasing overall efficiency. Campbell et al. dis- indium monoiodide or thallium moniodide or a solidclose an operational X-ray laser comprising a free thin 55 state laser material such as Nd:YAG. deposited onto afoil which is optically pumped by a high power laser. substrate 14. The substrate 14 is transparent in the visi-

ble and UV regions, down to below 200 nm. On theSUMMARY OF TtE INVENTION other side of the substrate 14 are mounted-two elec-

The invention claimed herein is for a miniature solid trodes 16 and 18 across which is connected a pulsedstate laser comprising i thin film of a metal monohalide. 60 electrical power supply 20.such as indium monoiodide or thallium monotodide. When a pulsed voltage is applied to the electrodes 16%%hich is deposited onto a substrate. The device is opti- and 18. a discharge occurs through the air or preparedcall, pumped by the optical emission from an electrical gas adjacent to the substrate in the space bet,een thedischarge sustained between two electrodes in contact electrodes. This discharge tracks along the surface 22 ofwith the substrate When a pulsed .oltage is applied to 65 the substrate 14 and produces strong emission in the UVthe electrodes, a corona discharge results %.hich tracks and VUV ranges. The short %savelength radiationalo'g -he iurface adn produces strong emission in the passes through the substrate and photodissoctates theLV and VLV The short ,a,.elength radiation passes metal halide molecules in the film This causes la"r

5,023,877

emission to occur on an atomic transition of the metal approximately 1 mm in length. and thus. pulsed voltagesatom or in the solid state laser material. For thallium as low as I to 3 KV applied across the electrodes willthis occurs at a wavelength of 535 nm. and in atomic yield electric field strengths of 10 to 30 KV per cM.indium. at a wavelength of 451 nm. These field strengths are more than enough to generate

In the embodiment of FIG. 2. a laser 10 having a S the necessary corona or surface discharge. These volt-circular cross-section comprises a substrate 24 which is ages may be generated from compact. low voltage bat.flared at its ends 26 and 28. A monohalide film material teries using step up coils.30 is deposited on the inner surface of a central capillary It is pointed out that the layer of active material isportion 32. and band electrodes 34 and 36 are mounted very thin, and therefore, the devices are compatibleon the exterior of the flared ends. A source 20 of voltage 10 with integrated optics techniques. For example, thepulses is connected across the electrodes. With this surface of the substrate can be corrugated prior to ap-embodiment, the surface current flux can be small at the plying the metal-halide or other crystalline film, andelectrode but large at the center where the active laser thus, this laser can be operated in a distributed feedbackmedium is located. It would appear that it may also be mode without the need for mirrors. Also, this systemadvantageous for the active medium to be in a crystal- 15 can be coupled to optical fibers or planar waveguides.line state, such as Nd:YAG or Ti-doped sapphire which Another advantage is that waveguiding can be ob-.would be optically pumped in the same fashion as the tained in the active region, thereby reducing thresholdmetal-halides described here. pumping requirements.

In the FIG. 3 embodiment of the invention, the appa- The wavelengths obtainable using these lasers areratus is similar to that shown in FIG. I. and the same Z. much shorter than those realizable with IIl-IV semicon-reference characters are used to denote the same com- ductor lasers. Through the proper choice of a metalponents. The difference is that the embodiment of FIG. halide obtaining laser wavelengths in the UV range are

J dsafursetcnetr40 either between h aldotiiglsrwvleghenteU ag r3 adds a fluorescent converter 0quite possible. The green and violet wavelengths men-substrate and the active laser material 12 or mounted onthestexteorsurfacfthve laser material, as hown n te 2 tioned earlier are considerably smaller than the currentthe exterior surface of the laser material, as shown in the 25lasers (yellowred).FIG. 3. This fluorescent converter 40 improves laser oreover the radiation emitted by this laser is narrowefficiency by absorbing pump light that is not efficiently Moreoe.th a dto emitted byabsorbed by the film 12. but converts the pump radia- in linevidth and locked to a wavelength determined bytion wavelengths to a spectral region at which the laser the metal atom (or impurity ion in the crystal such asmedium does strongly absorb passes through it. That is, 30 Nd in YAG or Ce in YLF). This feature is extremelythe absorbed energy in the converter 40 is re-emitted at valuable for communications system Which require awavelengths that are absorbed by the active layer. reproducible frequency standard as a master oscillator.

The embodiment of FIG. 4 is also similar to the em- Because this invention is subject to numerous varia-bodiment of FIG. I. except that the lower surface of the tions. adaptations and improvements, it is intended thatsubstrate 14a is corrugated prior to applying the very 35 the scope of the invention be limited only by the follow.thin layer of active material 12a. It is pointed out that ing claims as interpreted in the light of the prior art.because the layer of active material is very thin, the What is claimed is:devices are compatible with integrated optics tech- 1. A miniature, optically pumped. narrow line widthniques. For example, with the surface of the substrate solid state laser comprising:corrugated as in this embodiment, the laser can be oper- 40 an optically transparent substrate;ated in a distributed feedback mode without the need an active laser medium comprising a film of lasingfor mirrors. Therefore, it can be coupled to optical material on said substrate, said laser material com-fibers or planar waveguides. prising a thin film of a metal monohalide:

Although not illustrated in the embodiment of FIGS. a pair of spaced electrodes on said substrate:1-4. the surface on the discharge side of the substrate ,5 a source of short duration high voltage pulses con-can be grooved to improve the reproducibility of the nected across said electrodes, said pulses arcingdischarge and. hence. the operation of the device, between said electrodes and producing short wave.

The embodiment illustrated in FIGS. 5 and 6 shows a length light energy, said light energy passingmodification of the basic idea, but in which the laser through said transparent substrate to photodissoci.medium 50 is deposited on a substrate 52 formed as a 50 ate molecules in said film, thereby causing laserlens. A pair of band electrodes 54 and 56 is attached to emission to occur on a transition of the metal atom,the substrate 52 and a pulsed electrical power supply thereby producing a narrow line width laser.(not shown in FIGS. 5 and 6) is connected across the 2. The invention as defined in claim 1 wherein saidbands. The curved surface of the substrate 52 is pro- substrate is transparent in the vacuum ultraviolet range.vided with a plurality of parallel grooves 58 to allow ss 3. The invention as defined in claim I wherein saidthe discharge to be reproducible. The light produced by metal monohalide is indium monoiodide.the various discharge grooves or "channels' is focused 4. The invention as defined in claim 1 wherein saidby the curved substrate to a line inside or at th; surface metal monohalide is thallium moniodide.of the active medium. This concentrated pump radia- 5. The invention as defined in claim I wherein saidton will allow laser oscillations to be realized with less 60 substrate is rectangular in cross section and whereinpower delivered to the corona discharge. said active medium is on one side of said substrate and

The disclosed embodiments of this laser have a num- said electrodes are on the opposite side of said substrate.ber of advantages. For one. the laser is in the solid state. 6. The invention as defined in claim 5 wherein saidIt uses a metal halide in a solid form. Any amorphous substrate is transparent in the v acuum ultraviolet range.crystalline or polycrystalline laser material is also ac- 65 7 The invention as defined in claim 4 wherein saidceptable The disclosed laser system is compact and. active material is indium monoiodideexcept for the power supply. is se!f contained The 8 The invention as defined in claim 6 %herein saidentire s .stem can be consiructed on a substrate only acrtie material is thallium monoiodice

5,023,877$ 6

9. The invention as defined in claim, I wherein said 16. The invention as defined in claim 15 wherein saidsubstrate is a capillary having flared ends and a central active material is indium moniodide.capillary section. and wherein said active material is 17. The invention as defined in claim I wherein saiddeposited inside the central section. and said spaced active material is thallium moniodide.electrodes are mounted on said flared ends. 5 18. The invention as defined in claim 14 wherein said

10. The invention as defined in claim 9 wherein said substrate is rectangular in cross section and whereinsubstrate is transparent in the ultraviolet range. said active medium is on one side of said substrate and

It. The invention as defined in claim 9 wherein said said electrodes are on the opposite side of said substrate.substrate is transparent in the vacuum ultraviolet range. 19. The invention as defined in claim 14 wherein said

12. The invention as defined in claim 11 wherein said 10 substrate is a capillary having flared ends and a centralactive material is indium monoiodide, capillary section. and wherein said active material is

13. The invention as defined in claim 11 wherein said deposited inside the central section. and said spacedactive material is thallium monoiodide. electrodes are mounted on said flared ends.

20. The invention as defined in claim 19 wherein said14. A miniature solid st rate comprising: 1 substrate is transparent in the vacuum ultraviolet range.an optically tnmsparen substrate; 21. The invention as defined in claim 20 wherein saidan active medium comprising a fim of lasing material active material is indium moniodide.

on said sbstratds,, 22. The invention as defined in claim 20 wherein saidan optical pump for driving said lasing material. said active material is thallium moniodide.

active material being a thin film of a metal mono- 20 23. The invention as defined in claim 14 wherein thehalide, said pump comprising a pair of spaced elec- said substrate is corrugated on said one side, andtrodes on said substrate and a source of short dura- wherein said film is deposited onto such corrugations.tion high voltage pulses connected across said elec- 24. The invention as defined in claim 14 wherein saidtrodes. said pulses arcing between said electrodes, substrate is shaped as a lens for focusing light onto saidthe resulting arc producing short wavelength light 25 active laser medium.energy. said short wavelength light energy passing 25. The invention as defined in claim 24 wherein saidthrough said transparent substrate to photodissoci- active material is indium moniodide.ate molecules in said film thereby causing laser 26. The invention as defined in claim 24 wherein saidemission to occur on a transition of the metal atom, active material is thallium moniodide.thereby producing a narrow line width laser. 30 27. The invention as defined in claim 24 wherein the

15. The invention as defined in claim 14 wherein said surface of said lens shaped substrate is grooved.substrate is transparent in the vacuum ultraviolet range. °

35

40

45

50

55

60

65

Reprinted from

THE JOURNALOF

CHEMICAL PHYSICS

VOLUME 94 NUMBER I

1 JANUARY 1991

Spectroscopic characterizations of the 9 and 1 componentsof nf complexes of Ne 2 with n = 4-6

S. B. Kim, D. J. Kane, and J. G. EdenEveritt Laboratory, University of llinoim Urbana, Illinois 61801

Marshall L GinterInstitute for Physical Science and Technology. University of Maryland. College Park Maryland 20742

pD. 145-152

Published by the

AMERICAN INSTITUTE OF PHYSICS

AlP

Spectroscopic characterizations of the 3+ and 3 -g componentsof nf complexes of Ne2 with n = 4-6

S. B. Kim, D. J. Kanea' and J. G. EdenEveritt Laborato. University of Illinois. Urbana. Illinois 61801

Marshall L. GinterInstitute for Physical Science and Technology. University of Marvliand. College Park. .aryland 20742

(Received 11 June 1990; accepted 24 August 1990)

Rotationally resolved, laser-induced fluorescence spectra for the transitions nf( ' ,' I)-a '!,," (n = 4-6) in Ne. are presented and analyzed. The rotational structures of the upperstates are found to be strongly 1-uncoupled. Effective rotational constants for the anf " , and nfr, '1 ! states with it = 4-6 are reported and the structures of the nfcomplexesdiscussed.

I. INTRODUCTION II. EXPERIMENTAL TECHNIQUE AND DATA

Until recently,' ' relatively little has been published ACQUISITION

concerning the spectroscopy of the excited electronic states The experimental approach invoked here is similar toof the Ne. molecule. Presumably, earlier experimental ef- that of Ediger and Eden " ' in that production of electronical-forts to directly observe spectra attributable to transitions ly excited rare gas dimers in their lowest metastable state,involving more than the lowest few excited states of Ne, a 1,, . is followed by dye laser excitation of the molecule.were unsuccessful because of the difficulties encountered in Subsequent predissociation of the highly excited Rydbergdetecting bound states which are coupled to open channels. diatomic state results in np-3s(n > 3) atomic fluorescenceIn the energy region below the X ,,, ground state of Ne, . which is a convenient measure of the relative excimer popu-the intrinsically bound excited electronic states of Ne, are lation in the dissociating state. Consequently, high-lyingRydberg states4 while the open channels lead to dissocia- rare gas dimer Rydberg states, which, because of predisso-tion. ciation. are inaccessible for study by conventional discharge

Well developed vibronic spectra involving electronic emission spectroscopy, can now be examined by an opticaltransitions between a number of Rydberg levels in Ne, have pumping technique in which the spectral resolution is deter-been reported previously.' '" However, the only published mined by the gas pressure. dye laser linewidth. and lifetimerotational analysis' for Ne, is for the transition tentatively broadening of the upper molecular state.assigned as 4p- 111, - 3.5o '1,, (0-0). which was observed Because ofthe energies of the Ne and Ne. excited states.in both the "Ne. and '-Ne. molecules. These observations however, the present experiments differ from those of Ref.confirmed the identity of the molecular carrier of an exten- 16 primarily in that Ne, metastable molecules are more con-sive body of spectra.' the vibrational numbering of the veniently formed in the afterglow of a pulsed corona or capa-observed bands, and the AA character of the transition. citive discharge as opposed to optical excitation. While ade-Comparison of these results with theoretical calculation%- ' quate Kr(6p) and, ultimately. Kr. number densities areand with the expected properties of similar Rydberg states in readily produced by the two photon excitation of Kr( 'S, )other molecules' implied the specific electronic state assign- with an untuned excimer laser. similar experiments withments indicated. Ne. are hampered by the lack of intense. tunable coherent

The present work deals with the analysis of rotationally sources in the VUV.resolved spectra for several hands associated with transitionsfrom the lowest stable I,, state of Ne,. denoted here as theu , state, to the/1-uncoupled' ... and !1. compo- A. Discharge cell design and operationnents of the 4j: 5j. and b'complexes. Such fcomplexes arethe collection of interacting electronic states of a single mul- A diagram of the corona discharge cell de~cloped fortiplicity from the set of four Rydberg united atom molecular these experiments is given in Fig. I To minimie the intri-orhitals nqfo, nf-1,. n./3, and n)'6 for a specific wil with I 3. duction of impurities ito the discharge. the metal cathodeBeside the Ne, spectra described below. rotationallv re- (either a section of wire or a threaded rod ) ,as encased In asolved spectra involing) complexes have been reported for glas tube which protruded into the a.,. A stainless steelHe -' H.- N ." and NO) screcn held in contact with the ilas,-shcathed calhode aidcd

i the production ofa uniform. corona-like discharge h% sup-pressing treamer formitlon. The ;iirde-io-cahodc eparal-

,(I -4 t)! i I .... \i.n,' \.,'il I i .,'< . t tion was set al ( im mnd all electrical fecdihrniuhs cni-\,, \\t .' phl\cd standard glass-to-ncial scals. .\1m). the anode \wi

94 a--, 9?0 C I30 '

146 Kim eta. Comrplexes ofNe,

FIG.as W iagrmofih corona Sicaiettb aotdi hs ealntn

tn!ht2hten.dicageslag itiotilsal:2M 2Wdi an, tclsistsi

kV/disungte ) oe nI th wowakfuirscne use hihars fo

reIiG. 1.g Diara the dcoroiara Thehrg ruse timete i thhs demnanesielmhtnulssiThe~~~~~~~~~~~ ~ ~ ~ ~ ~ ~ ~ pilo tlistilsste cerwa treuecagn ltesr FI 2. Waso m h e r eseti ta tive tro t he imp ed for: (ina)the htweeti theface~~~~~~~~~~~~~~~~~~~oe ofh ls uee-eog(i ahd n upesrae om- sdlgene -ror anicds hare ichel hoisel 214 ideri i ) an hh. Allhaen

mionedfo oydnm urai urzwnoswr dscrge liraphed tiiriNe prsurle 2() Todi ndrr ia eslIattached Noo each end ofe the discarg cellC viase qhich arsto-Py-resriin graded dieals.g After being etacate toth doi5n % h puls - srwit a tuboolcua valuum syste andon baked fro th mpdne00Cchbtc hforwe geeao to4d das eac discharge celll was ekihe with isoto- Al a~t

paical d efrche molydenu. Ise) i quert ind.os ofete were pobtogaped for time drelays ranging Torom4)t r()/satachwereoacuied wfthe 200cTore cel dichares bt- .twsfxdatby o ot f pcr rpre ee

pex rmeseas dAfterv beng caried to t prsue as low as Dy ae uleeeges pcly hs tb ewe

The 3 No4dyec discharge cl was ba rounded-rdtya toanmitiedo withi ± 0%osaseclidvauIron ~ ~ ~ ~ ~ ~ ~ ~ ~~wr pusrhvn nitra aaiac f5n n pr hogotae scan biedy a gn g thertm dela tor the/s

atdweat a ciren voltage r betw e 7adishre u ex- kA. For thee Nd ~ ater s forimohtf terdoubspegthareportedhen-.eperiments daeie haeren cared uta rssueslwayae pulsed neaie tnenaptssu iyrgen phphcat cstalto he bar-ee

with respect to the grounded anode for - 150 ns (full width monics were separated from the fundamental by a sequenceat half maximum. FWHM). Because little effort was ex- of four Pellin-Brocca prism%. Survey spectra were recordedpended in matching the driver impedance to that for the with a dye laser linewidth of -0.2 cm 'whereas the instal-discharge, a small fraction ( - 5%7c) of the power reflected by lation of an intracavirv etalon (FSR =I cm ') in the dyethe low impedance discharge reignited the plasma - 1. 1 Its laser reduced the linewidth to ,0.04 cm 'for higher reso-after the initial power pulse. Both the side-light fluorescence lution scans. The linewidth of the laser radiatiotn was miont-and discharge voltage waveforms I presented in Figs. 2(a) tored on a continuous basis with a solid etalon and a Si Pittand 2(h). respectively I show this clearly. For a charging photodiode. Wavelength calibration tof the excitationi spec-voltage of I5 kV, the peak voltage and discharge current for tra was accomplished by optogalvanic s pectroscopy usinigthe first pulse were IS1 kV and 10 A, respectively: for the U/He or Ne hollow cathode lamps. Line posittons were de-second, time-delayed pulse. the peak voltage was 5 kV and termined to within _0. 1 cm f'or the data presented he-the maximum current 1.5 A. Therefore, the total energy de- low.posited it) the gas was -5 mJ. which ctorresponds to an in-crease tn the tnstantaneous temperature in the discharge re-gion of - ) K.

8. Apparatus and data acquisitionA~ sChemnic dtagrairi of the experimentl :irratngcnitt _______ _____

Is iicn In F ig. 3. At adjUstable delaiv rime ..( Al ollov.1tiLet he producitIt111 iOt Ne metwathle intIlcules IIIt he pulseddis.harLe. ilie heatm troin a t'requ~elC\-doubl~ed. Nd YA( i-pumpeird (4~e latser I Lambda Phi 'tsik %%,;i ws pri ipatcat-Cdl alotlL' the axis )I' the discltarut: cdl. Fit dittxtctecc.

At 5 O i..tfitiied Is the. (timem at . tch peak ckrri icutti III .. . . * ,

thie N'condr pulsec throtueu, the (ltschanree ccli .\lthitiLi data 1.01 - Alim! f.""' 'I'Ail'. '1, . L 'I V

Kim etal. Complexes of Ne, 147

Ill. RESULTS

A low resolution (A '-2 cm ') scan of the 352-369nm region of the Ne, excitation spectrum. acquired in 20020W ns0 FIG. 4. Oscillogram of laser-induced Torr of-'Ne. is presented in Fig. b. Although the entire spec-

........... . .. Ne fluorescence at 703.24 nm trum of the molecule from - 282 to 450 nm will be discussed(Ne3p{ 1/21, -3s13/21, ) produced in detail elsewhere, it should be noted that the dominantby exciting Ne. from a Y, bandhead in Fig. 6 (A -356 nm) appears to be associated',- = 291.3 nm. The weak pulse at theleft of the photograph (indicated by with transitions to the 'II component of the third member ofthe vertical arrow) is an artifact of the the npi" Rydberg series. ' The second member of this series,photomultiplier pulsing network and which lies at - 425 11m, was reported previously.' Efforts toyet srves as a convenient tiduciA rotationally resolve the peaks near 356 nm have thus farmarker.

proven unsuccessful, presumably due to lifetime broaden-- - - ing. However, several other weaker bands in Fig. 6 clearly

exhibit resolved rotational structure.The spectroscopic data reported in this section were ob-

tained by analyzing rotationally resolved bands for transi-tions from vibrational levels v" in the a I,, state to vibra-tional levels tC in excited electronic states believed (seebelow) to be the 7, and 'I I, components of the manifolds

Dye laser-induced Ne atomic fluorescence was collected of states associated with nf Rydberg orbitals. The (0-0) andat 900 to the discharge tube axis with three piano-convex (1-0) bands [where (' - v") I of the 4f" member of thislenses and detected by a cooled photomultiplier. Spatial fil- series appear in Fig. 0. Figures 7 and 8 provide expanded andtering further defined the region of the discharge under ob- higher resolution ( - 0. 1 cm I) views of the 27 260-27 520servation and a dielectric interference filter selected a partic- cm I region which contains the 4( !, ,'11,)ular 3p or 4p-3s transition for study. Most of the -a ',, (0-0) bands.experiments to date have monitored the To d,,te. triplet splitting has not been resolved in Ne.3p 1/21, - 3s[3/21, transition at 703.24 nm. In order to spectra, a situation similar to electronic spectra observed inavoid saturating the photomultiplier with the intense visible He. where such splittings were observed'- ' to be negligi-fluoresLence produced by the active discharge. the photo- ble in most cases. Because the nuclear spins are zero inmultiplier bias was kept at a low value during the two dis- '"Ne-, only alternate branch lines can occur in Y, -. ,, andcharge pulses and pulsed "on" -7 pis prior to the arrival ifthe de laser beam. Ten microseconds later. the bias voltage type transitions. Thus. each hand of Ne. associated

with transitions of '", - ., type will consist of one R(.V)was returned to its lower ("'of) value. A representative and one P(N) branch only with odd N values. .V being thelaser-induced fluorescence waveform for Ne.. obtained for= 29.3 n, i shon i Fig 4.quantum number for total angular momentum excluding

2,,=291.3 nm. is shown in Fig. 4.

Data acquisition essentially consisted ofsampling the spin. Transitionsof 1l . - lv,, type will have three brancheswhich can he divided into two groups: one R(N) and one

photomultiplier signal immediately prior to and during the wi ch vidd o w a. - R,, and one

dye laser pulse with two hoxcar integrators having gate branch with N odd for Al, !,, T and 11

widths of 20 ns (cf. Fig. 5). The difference in signal ampli- he " " a 'tude was stored by a DEC LSI 11/73 computer which also states are reflection symmetry cigenstates which are called Adoublets when their energy splittings are small. When A

controlled the frequency doubling crystal. the scanning eta- doublingbecomes large, thesame phenomenon is referred toIon, and the dye laser grating. A minimum oftten pulses were d ouplingmeaccumulated at each dye laser wavelength. For higher reso- as/uncoupling.lution studies, the laser wavelength was stepped in 8 • It)

cm ' increments (i.e., 20-,, of the dye laser's linewtdth)while survey spectra were recorded in steps 4 times larger. 'Software written for these experiments controlled all phases I

Otf the data acquisition process. Fluorescence data obtained ...... ....

with probe (dye) laser energies lying more than -2 lO_from a preset value, for example. were automaticallv reject-ed. Also. the relative intensity data were averaged and I

checked for consistency b, comparing each datum to 3.5 . [tires the standard de%.iton ot Ithe Mt point ,et acquired ateach dye laser waselengih. Poti,, flliing, more than 3 5f" ), ___t__ _trom the avera.e \,alue were deemed to he tunacceptahle and .,,..,,. .....

the process was repeated uniit lhe criterion %% as satisfied for ....

each daium at each dye laser %%aielciwgth. Further det:iilsrciuatr"intg the experimentl techniqtes aind dalt ita li ,l, I, m IiIi, iq11iirii 0, .' i lh ,pr''~cdires aLlrl he. 1oitid i Rcf . I, 1\ '

•" ' ,

148 ,(im era Conpexes otNe

28400 28000 27600 27200

1 0 -12

e

8-

Cr)Z 6pifl,-a j3+ 3E

U 6- (0-0) 4f 9f .- a f 3 , g) 4. Eu

z (1-0) (0-0)

LU

6p1T 1 ng4 Q

2- (0-0)

0 - L--- i I

352 356 360 364 368

WAVELENGTH(nm)f- Ri f Porition oiI he Nc a 1 1aser ex 5 itation spectrumi ohser~ed I'm 32Q /. "b") 11tn. [it acqut11rmtg 1IO .til t. O le kiSer IirieC.dth t..;ts 1) 2 cii

P, 2mN rwr 3ME K and .11 (0 q, THe spcci ral regioni enclosed hN the dashed recialitlle. 1% hicit comialits tile 4/. tt -01 Iraisiloits. is I~li racd in

,-te~iter detail Ii FiVs ,lind M ote 111.11 the IS 7 om rettoit has het ttagtiid h,,a tactiir It 3 I'm ciiitmeniicc

12-

Ki, etai . Compiexes ot N. 149

Line positions. in cm vacuum, of the (0-0) bands of TABLE II. Line pIition% it cm xac) ofthe 5IAt I. . II -a Z

the nf( '1 '[1, ) -a '," transitions for n = 4-6 appear in 1 traitions it Ne..

Tables I-Ill. respectively, while typical high resolution sfrU -spectra of these bands appear in Figs. 7 through 10. Also

included in Table I are the line positions of the \ RN) PIN) R(N) QN)" P.A)

4fi'Z; .'llH,)-a',, (1-0) bands. Portions of (1-0)bands associated with the 5fand 6f-a transitions and the 1 24437.58 (29 1 35.13)3 2'o)35.13) 2')'432.2 2'4,463.76+{,051(0-0) bands associated with the 7f'-a transitions also were 5 (2) 432.21) 24925.53 ) 2') WO 45 (29 454.734 24o 5341

observed, but the current data on these bands were consid- 7 29929.02 29917.60 29')74.83 (29 95341) (29455.02)

ered too fragmentary for inclusion here. '4 (2025.53) 2')10.20* 299179.55 (29,-52.57) (29455.67)Band origins and effective rotational constants (i.e., ro- 1 29'922.05 29902.02 24983.71 (294'50.58) (24955.67)

13 (20918.38) (29)843.85) 29987.01 (29948.32) (29')55.67)tational constants not corrected for / uncoupling or other 15 291)[4.80 (29 885.78) 2)897+ (2X45.+8) '(20454,73)

perturbations) were determined from the data in Tables I- 17 29'11 14 29 877 80 29N2.30 2442.76 (2'452.57)Ill using least-squares-fitting procedures and the power se- it) 24 ')7.47 2') 869.54 29 '43.92 24 939'4 I2N 950.584

ries relations: 21 21)X)3.80 2)1)1.2) 29 496. 18 2'4937.11" (2448.32)

23 2) 9(X).43 20 853 24 29 998.4h 24 445.08)25 29) 896.57 29 845. IO

TABLE 1. Line posltions (In cm vac) of the 27 (29 893.85) 2' 837.054jf4 Z. Il,) - a M, tr' - v') trantions In "'Ne. 2t 2q 88.53 24 .82'.

31 12') 885.78) 24 S21.054/ou',-a 1, (0-0) 4/,-1 1-a ,, 10-)) 33 29883.05 2') 813.58

35 2' 147',77 2) 815.74N R( N) F) R S) Q N)" P(.%)

Parcnlhewss and an astcrisk detleoi a hlended hn and a shoulder I ature.I 27 398 73 27 39t 85 respectively3 27 397 85 27 343.07 'The .N nu.nhernoi htis hralch has hecn closen 1it make it ong i match5 27 396.02 t27 387.74) (27 426).01) 4the origin ,hcrvcd for the II com'ticni.7 27393.53 27 381 35 27 441 " 27425.37 427 426.0),) 27 340.71 27 374 43 7 452.54 274244') 27427.43

II 427 387 74) 27 367 1(0 27 458.53 27423.41 27428')13 27 184.56 27 35' )62 27 464 18 27422.)5 274313115 27381.35 127352.X)) 27469,54 274210.49 127431,4')) 1 F+

17 27 37XI' 273442h 27474.54 27418.78 427432.51)IL) 27 3745 27336.45 27479.24 27410.82 127432.8h) 'V(21 27371.44 27328.6') 27483 t4 27414.') 1274432.8h) Q ,

23 27 3b8.11 27 120).8 27487 5h 27412.82 427 432.8h)25 27 4 76 27 313 )8 2744135 2741054 427432.51) and27 27 161 41 27 305 31) 21 4)8 3) 427431 44)2 27358)') 272'75x 2740h 11 R(.N) + P(N)= V cN(- 1)'. (3)3) 27 354 S' 27 2X'4 17 27 403 8634 1 4' 3524844 27 282.23 27 -4)4.135 27 348 38 27 27445

I \I1 III. lI c posltiosI, cm , ac) o he hl/ , .11 .

4fir . a Z, ( !) 4I- 1 II -0a _ 1- ) 4-) tr111" ll lo tll N l

"N, R4 V) P(.\I R4V' Q4".4 /'4 4 " a 4. 6/' II - , .

I\ R44 A / \ 4 R44 \ I (,) %'44 I'! '4

214 M) 6')l hroad127 56 2 2) )012L peak 27147 1 1 11 ;1 1 2 11 128 25 .)1 124 ,7

_'1 ) Q 29) 27 )5 84 2 ' '''')') 3 1 2- '))x I I I ') 04 "1 "Oh 1h 4I I4 " " !; 14 1,1 +44 52425 4 1 1_2 I4 )70 2'45 4 2"'142 ° +

21M237 '"'1') 41

i I I )5 5I 1 ")x l M 'I,1 2347 1 1 12 147h2 627 29 6) 07 04044 0 4) 44)t 4) 1'2 4,4 1 '41 ) 116 I 22 i 11 4 ..

14 2 4; 2 8IM "'42-08 7 2S ) "4 ' 1 )')) ' 1 2) 2' ,x I ) 42,4 1 141 X4 1'l X.l 7) 31 2 I15 2- '4' 1, 2" 'fl') h 28 0U) 4 r 27 0'' 1 1 1 ) " 1144M4 1 '4 27- l) 11 '4 945 1 ') 1 "1 1 1 i3)1 11') x7,1 2"

'445 844 2" 19 2 4) 2444h I I 11 '404f ) 1) 1 5 '; , 1) 14 40), 4) ',I ;I ,p

) " 1142 144 2 144 4,) 2415) 2 15 1 2X8645 f1 25" 44 4) 4)) Ih

21 '" 'll".)) 2' I 5 , 1 282 46, 41 244 14 1 10 4h 4

7"I' 4," 14' 14 '44 -4 ;044o -

''4, ' '+' ""qu |. -- 't't| 5 247 " 4t') 24 21.' ,

"- 2- 2 I - til 4 "

I ll: \ mmitrhTne ,IIh s Ir -jC lia, Na. hc df'n h-c l i,, . , tlkc Itsl 4l4 0144 4"4,44,. r

,I v r 'ui ,ri . 4.,r ,ih, 11 mu4444,,mum4 Rll 111h4,", tid l , I, :, 'k ,4,.,,,4, 4 4,,, 4,, 4,1,., F, , , i,

I il Q \ ,n,.)'" *ll Th , '., ld4, "h h I,.i44 f.ik % hich4 hlWN f.

' M h4 1-4, , ) ), . " 4 . ,, , It .41214 h

'4).I1 '.s PI I ' '4 T., I,' .,f,.j ' . ,., t 't 4).11 ',i tl.llxl~, ll p, ,rh rc l'cttTit<iF lt'ldXlt'[+It [.t ' \

150 Kim at 3/ Complexes of Ne

-i IV

-. . .'" I) '

29)100 29900 30000 " .. "C;(c"- -)

FIG L) Laer-induced iuire 'l.'e -peciruni i the near ullra, obl illus- FIG tO Spectral %egion he ee,' 31 Il0 and 31 381) cm showing branchirating the rotational hraiche oif he 5J( - . II (1-)4 (rl|isi- asiglnmCent' r tf A( Z .11 I u 1 , (WI))) fransfiti N nOf ''Ne , Exp.ri-tion% in Ne. menial condition

, are the %ame as those fir Figs, 7,)

The coefficients offlN) in Eqs. (I )-(3) contain the effec- mined by averaging data from Refs. 3andbandTables 1-I11.tive rotational constants, and several orders (several values These energy levels were combined with the branch data inof i) were equally satisfactory in reproducing the data being the Tables to calculate upper state energies for each spectralfitted. Through i = 2 one has:' a,, = 4B transition. These spectrall, determined values were aver-- D . a, = - 8D +- 34H, + ". a, = 1211 - . aged to produce best sets of energy levels for the

b, =v '- t. b,=B, -B -6D + b, =D nft , .'11 ) states withn .=4-6. For each state. the result--D. c,,= 2v,, - B - 2 D ", , c, 2(B - B ing energy levels. F, (N). were fitted to the expression-6D, ,- ).andc.= 2(D -D - .- ). with higher-

order constants available '" when needed. A.F'(N) = R(N) F, (N) = C [N(N + 1)1'. (4)- P(N) values are used in (I) to obtain the upper state

constants B . D . etc.. while I.F'(.V) = R(N- I) where C,, represents the energy of the V = 0level (real or- P(.N I ) values arc used to determine the lower state hypothetical) above the zero chosen for the F energies (the(double primed) constants. The Ov(,' - I,") value in Eqs. N = Ox v 0 level of a I,, in the present case). C, = B,.(2) and (3) isthe band origin. which istheenergydifference C, = - D.and C, =H,.between the N' = Ox,' and .V" = Ox" levels. For the t11I and The results obtained from Eqs. ( 1 )-(4) were averageda TY,, states. the .N = 0 level does not exist and the (see below) to produce the effective molecular parameters

'v' - v") 'values are extrapolations. reported in Table IV. Energy levels (term energies) for theThe rotational energies for the r = 0 level of the a , a '11. 4f. .. ,). 5"( t,..A11). and 6'( .11)

state were calculated from a best set ofA, F" values deter- states are summarized in Table V.

I.)I I- I' I I .Isc m;loleculaIr cmisianis in tn

Six': ' )? I)Cj I) 1 1 * M

-/ ' fl l l i ;"-l,' 1 l - *

i

d i oI W , . 111 0. Si (11 o.14 29 o' 42C )+41 r" -

1) t Itso + I 2110 ZI- ]Il ' ) 5f i

4/- I1i 1 l-to I 2lis 140I 'I0I 40o 'i I i?

4/4 ~ S 4 1 ) 11) 0( 1III I 1,l 41) 12o 1 I 2,,r7 -1 1rt 4 90lf :lii !to .4(X 14 2,1

i I ) (I N)

¢ T ~ ~ ~ ~ ~ ~ ~ 9 II4 4! 1n4il:t ilJ+i 2)u€ ;t

' I- I 'I ilill II

"~ll, .l I " \ l t,( I i lt ( 1 k I< -

t - -i ll W 1 - , + 1 ; w , i) 1 -i 1 ' i { l l ; l l :i 4 1 i

i - I 1i1) ,i'0 1 - 1 1 ! 2 1 +

_T, c~ ~ ll k ll~ h ,l .% 1!1!Ilt.~ ltlk I lillt ~ lI t "-d C It ( H lll ,ilIl, lh tl

,.w tji Cor'piexes olN. J

TABLE V Level energies (er ') ol'ihea I and iM ( 1 .II ) s,.als The large uncertainties listed for the parameters in Ta-wuth n = 4. 5. and h.' ble IV represent the range of values obtained for these pa-

rameters from fitting the data to within experimental uncer-a I 4j- II 5frr II 6 II tainties using the different polynomial expressions presented

' C, = 0) (, = o) (, =0) (1. 0) above. When applying the method of least squares to polyno-

1I 12 31 325.79) mials such as Eqs. ( 1 )-(4). one finds that the values of the3 6.74 29962.41 31 330.919 molecular "'constants" depend on: (a) the total number of5 16.85 27 442.86 29 471.58 31 340.32 data points used in their determination (i.e., the number of N7 31.49 27456.86 2 9984.1X) 31 35379 50.b3 27475.12 30003.20 31 37.50 values included). (b) the order of the polynomial employed,

II 74.23 27497.64 30 024, 81 31 391,56 and (c) the specific equation used to obtain the molecular13 102.38 27 524.43 30050.70 31 416.68 constant. In most cases, the number of data points was var-15 135.01 27 555.50 30080.10 31 446.17 ied by - 10-25 % from the maximum possible number. all17 172.07 27 540 85 30 114.83 31 47(98719 213.64 2763051 30 153.h3 31 518 34 polynomial orders between 2 and 4 were considered, and the21 25.78 27607469 (1 9639 data were fitted to all of the equations ( 1 )-(4) to which they

23 310.3b 27723 1 could be applied. The spread in the set of values for a con-25 itri.3i 27 775 IM stant obtained from the range of procedures summarized in27 424.,0 27 833. 102) 488.63 27 N94 81 the preceding sentence is usually larger than predicted from31 556.85 27 9)60 71 the standard deviation analyses for individual fittings. This33 62 .49L 28 o3o. 5q spread will increase markedly in the presence of strong chan-

nel mixings (U uncoupling accompanied by configuration in-teractions).

1 The ranges of empirically determined molecular con-

0,' c, 1)) 0) 0) (, 1l) stants reproducing energy level data for electronic stateswhich are approximately case (a) or case (b) states are rela-

2 2" 347 7 2') ,)3s.25 tively small, because the case (a) and case (b) energy rela-' 2

7 3t4

3 S 2t)

108 70)4 27404 5 2) 3 21170 54) tions4 are functionally consistent with Eqs. (I )-(4). The

27 412.3h 27457 50 29 441, 07 29 9h 3) constants and "uncertainties" ( range of possible parametric274250o4 27478.11 2,) O).20 30) (1.32 values) for the a ',, state in Table IV provide a typical

I0 27 441 34 27 503 I1 24) 7h,20 1030 Is example. However. states which I uncouple approach case12 21 42.(X) 27 512 73 2) Qt14.2, I 115 )S7 '9414 274.569 4 2756.56, 31)02))077 3))fl59 3)) (d)' or its multichannel equivalent and are poorly repre-

I) 27 51 34 27 W04 55 10o049)4 )) 124 77 sented by Eqs. ( 1)-(4). which only slowly converge withIS 27 551) 15 27 646 h1 30(83 22 .(1 . 4 37 increasing order. This slow convergence leads to strong de-2)) 27 5X8847 2 h42931 30 121 1t 31 ' 7622 2' 631.23 2' "4 2 )3 1 1.5 ) 2()7 1 pendences of the parameter sets on the specific procedure22 2' 0)1.3 -" '43 42 10 1 h, 58 1)255 O,24 267 87 50 27 797 )2 ;( 2) i)2 1() 4)o 82 employed and. collaterally, to values of the effective con-26 2' '310 II 2' 85f '1 10 26 1. 9 stants which, although not interpretable in the same way as2, 27862) 12) 117 X I the constants obtained for states not affect ,d by I uncou-30 27 468 2 1)) ()2

'2 2 ,) o 443 )7 piing, are characteristic of multichannel coupled states.14 2' ')i 1 25 10 512 54

2s1)18 IXh 1 7 IV. DISCUSSION AND CONCLUSIONS

The assignments of the transitions in Tables -ill to

h/7 116 I 41,, 1 4 1 upper states associated with nJfi 11, and n!r 411, are based

If I If ) on a number of observational facts. Rotational analysesprove that: ( I ) all of the transitions have a common lowerstate which, from the experimental conditions, must he

1 I 7 " " 1- metastable to radiative decay to the X ' Z state. (2) the4~ ~ '1I~ , 41)2 "'*7'r,47 2X{1 ( M " 1S2' lo5 I~ I) 14) X 'N ,I transition types are of 1-- and I I-- type, and ( 3 ) the effec-

1I 12 1 7 I 2",), ) 2S 14') 1 tive molecular constants of he upper states are anonalous inI I I 4X 5 1 10o44 2 )s t) "I 's )l- 4 a manner similar to those for /-uncoupled states in otherI 1 1,.Ns 14 11 42) 14 _100 45 2x II, I molecules. General considerations' combined \W Ith14 1 92 I 2) l5 If It 2514h ' . 2" ,. 2" ) )I45 -4 observations on transitions involving the lowest lying excit-

Is Il 454 is III , 's _ ed states and X ' , i the heavier rare gases." ' lead to1, : 1 , 4 ' , 2t) m-x _ quite firm assignmelts of the lower state of t liese I rallsll (lls

to the metastabie ,j 1 , late. in this regard. It should he21 : I2')2 4.4 21

-,,' , 2 _ ,) noted thai the u 1 -,iate effective rotatlonal co sista1 it re-- I ,,, 42 _ 4', -"" ported here (B f ) 5"26 -+(.()(X)6 en r) t. lth Ii ' of

S lithe ,.alue in en in Ret S 10. 64 - ).0)4 cn

M s I,,() . 8' IC 'O1 L'ClClt It il h)\ 1 1IIshit'r iI(s , 1IIl

}g~~t tl I! , ' i t , X J ld -x" - - . .\ lh IId]LI l'l lcl I HI ,I I{ lS il )\.t' Dl-

152 Kim e a Comnpleyes atiN

by strong sp~in--orbit coupling. States of 11, and 1.' type are of reasons, this rule should be used with caution hut it isproduced by the addition of n/A orbitals with odd / v'alues to expected ito be reasonably applicable to the ni complexes ofthe X ', state of Ne;' which means that transitions from the lighter rare gas dimiers. For example. the origins oft he 4fa are allowed only to states associated with niA ,nA. complex in He. are knowtn experimentally2 to be 27 398.3and high n/i orbitals with I odd. cm ', 27 407.0 cm ', 27 432.7 cm I. and 27 473.5 cm 1

The a I! state is associated with a or type Rydherg tbr thle A = 0. I, 2. and 3 components. respecti~.elv. Theseorbital which often is designated as the lowest member of the values can be reproduced if .4 = 8.4 cm 'andnsoa manifold. Thus, with weak or no spin--orbit effects, one E,= E,-r- .AA ,. We have used this rule wvith .1 values nearexpects transitions from a 'I r, to npi ( " 11,.) to be the energy separations ofnija) 1, - u from itfit '11L - a tostrong with the next most intense (but much weaker) transi- predict the regions for the (0--0) hands of thetion being to nfA('7, 111, ). As spin-orbit effects increase. n/iA a .,and nfib t(P - a tY, segments. butsinglet-triplet transitions become weakly allowed with tran- found too few spectral features in these regzions to make con-sitions to npA( 'Y., .'11,) being much more intense than firmable spectral assignments. This is not unexpected sincethose to n~i!;., .'II,). The relative intensities of close ly- such .1-1 and ii- transitions are much more highly forbid-ing npir( 'Il1,,'l It ) -a and Pj- a transitions, such as seen in den than the observed 1- and 11-1 transitions. WithoutFig. 6. have led us to assign the upper states of the nj-a data on the A and (P components. it is not possible to deter-transitions to triplets rather than singlets. mine empirical multichannel parameters for these complex-

As noted above, all of the spectra give clear evidence of es. We will continue effoirts to obtain data which will permitstrong multichannel coupling while the upper states have experimental characterizations of complete nifcomplexes insmall quantum defects (6 = n* - n = -0.04). Both of Ne. and hope tc he able to produce such data in the future.these features are expected behavior for states associatedwith nonpenetrating orbitals without ancestor%' in their ACKNOWLEDGMENTScore. This, and the observations above, leave little doubt thatthe transitions reported in the Tables and illustrations of The Illinois work was supported by t he Air Force OfficeFigs. 7-1() involve the lowest three members of the of Scientitic Research ( AFOSR) (H. Schlossberg) under

n'CY_( A1.4 I ) series. Figure I I depicts the A doubling (meii- Grant No. 89-(X)38 and the National Science Foundationtioned earlier) that is present within the AII ', component of ( ~)(W. Grosshandler. K. Gillen) .Under Grant Nos.the 4fircomplex. Note the rapid rise in the II1, -1Islt CB'i 87-15691 and CTS 89-15795. The Marlyland work %%as

ting with increasing N as / uncoupling becomes prominent, supported by the same AFOSR office under Grant No. 88-Attempts w~ere made to find transition,, to the remaining 001l8 and by the NSF under Grant No. PHY-8814722. with

I. ft). Iportions of the nf complexes. To a first approi compter time supported in part through the I'lcilieges, of themat ion, the origis of the components of an ] complex are Cmue cec etro h n~est~0'Mr adcxpected to obey the 4A rule.' -. where.- .4 i constant andA is the , alue of the component A ,tate. For a vtariety K PIs ilicoiait J ii I dii. I ( Ii I'll%, 84. Nh(,1 0\,h

H \jilclhi,~ I I i1/cIIlII'l'ci I t 5lc.,Ilcr .I Ii I .11w0iolt, Ii'll%, It i9.

Il~, 1~, t S \ui~ici,) 1ii 1 hciii 1~ 9i4,l I o . nI%!R. \ii-lc 39 .441-

R S -l lio I \III li si 1 4ffI1Wiik

Z Z 27wlC R , ",111111k,11. I ( 1111 I'l 52. ;1-( 0'1

- n ~ ~~~ I S( c Itl It, Siticihi. 45 'Ii P)N ,,6. i

0~~~~~~~~~ %1ni 1 Ficinti lilt] 37- T.Il lil l 1 I)

0 A I s t.w-mi Kei )t -h I~ I, 'I 1 Pl

d. I , Ind I Wt. 69' 1- \A 'k" ;I'! - 61 '- 1 , l-.

F1 P Vt 11 Ph- I ti 134 F~. ~

Bound -. free emission spectra and photoassociation of 11"4 Cd2 and "Zn 2G. Rodnguez and J. G. EdenEveritt Laboratory, University of Illinois, Urbana, Illinois 61801

(Received 16 January 1991; accepted 12 July 199i)

Structured bound-free emission spectra of 'Zn, and "4 Cd, have been observed byphotoassociating pairs of ground state atoms (ns 'So, n = 4 and 5 for Zn and Cd,respectively) at several discrete wavelengths in the 222-277 nm region. The observedfluorescence is attributed to B IT.: -X 'I,' transitions of the molecules where the boundexcited state is correlated with np 'P, + ns2 'S, in the separated atom limit. The pressuredependence of the emission intensity demonstrates that the dominant '. productionmechanism is photoassociation. Numerical, quantum mechanical simulations of theexperimental spectra yield I.' spectroscopic constants that are generally - 10% larger thanab initio values in the literature.

I. INTRODUCTION the radiating state structure is well characterized and one isable to concentrate solely on the nature of the dissociative

Since bound-free transitions of diatomic molecules ter- portion of the ground state.inmate in the vibrational continuum of the lower (dissocia- The chemical and optical characteristics of the Columntive) electronic state, the interference of the free particle lib dimers are also interesting since the (n - I )d and ns(continuum) and bound excited state wave functions gives orbitals of Zn and Cd (n = 4 and 5, respectively), for exam-rise to the familiar Franck-Condon oscillations. Such spec- ple, are closed. Consequently, aside from a weak van dertra are often termed diffuse and do not as readily yield the Waals minimum at large internuclear separation, the dimerdetailed molecular structural information that can be ex- ground state interatomic potential is expected to be repul-tracted from bound-bound band intensity profiles. Never- sive. In 1978, Ehrlich and Osgood7 observed the Franck-theless. as Tellinghuisen' has noted in his excellent review of Condon modulated fluorescence from vibrationally excitedbound-free transitions, "...the [bound-free] spectrum may W'-5 7 ) Hg2 (0 ) excimers produced by pumping Hgstill display considerable structure, but that structure is en- ('S, ) ground state collision pairs with an untuned ArF lasertirely Franck-Condon structure...the intensity information (193 nm). Forming this highly excited molecule in the vicin-is no longer a redundancy of the analysis; rather it now repre- ity of the upper state's classical outer turning point results insents the primary data and hence is the only source of infor- bound-free emission spectra extending over 30-40 nmmation about the potentials..." Consequently. bound-free which corresponds to a large range (several Bohr) in inter-emission spectra are an intriguing spectroscopic tool in giv- nuclear separation, R. The spectra of Ref. 7 show no sign ofing insight into the electronic and vibrational structure of the discrete structure but rather display the well-modulated in-radiating molecular state but also in providing details re- terference oscillations described earlier.garding the ground state interatomic potential as well. Table I of Ref. I summarizes those molecules for which

To date, these oscillatory continua have been observed bound-free spectra had been observed and characterizedin a wide range of diatomic molecules including the homo- prior to )984. With the notable exceptions of Refs. 2-6, mostnuclear halogens, interhalogens and the alkali. Column I IA prior work entailed exciting the molecule at one wavelengthand IIB dimers.' Particularly extensive observations of and analyzing the resulting spectrum by the semiclassicalbound-free emission have been reported for the Column IIA stationary phase approximation or quantum calculationsdiatomicsMg 2 . Ca., and Sr. (Refs. 2-6). The A 'I. excit- (Franck-Condon integral).' ' Otherdiatomic molecules fored states of each of the three dimers were photopumped at which rcent studies have observed interference structure inone of several wavelengths with different visible or uv lines emission spectra include Hg, ( Ref. ') and lBr ( Ref. 10)from an At or cw dye laser. Since their X 'E; ground Experimental studies of bound-free emission from Zn.statesarestable (e.g., D, = M95 and 10f5 cm 'forCa, and and Cd. in the gas phase and numerical analysis of the ob-Sr,. respectively ).' both discrete and continuum spectra served spectra are reported here. For both molecules. a rangewere observed for the Column IIA dimers, the latter of of vibrational levels of the "_ j (0 .' ) state, derived fromwhich resulted from transitions to the repulsive portion of ,p 'P, n 2 S., 1n = 4 and 5 for Zn and Cd, respectively).the ground state. The presence of discrete line spectra in the were produced by photoexciting ( photoassociating) pairs of.4 - X emisson permitted Rydberg-Klein-Rees (RKR) po- ground state atoms with excimer or frequency-quadrupledtentials to be determined for the.4 and X states. As pointed Nd:YAG laser radiation. Analsisf the modulated hound-out by Gerber eral.. this is a distinct advantage when turn- free continua by numerical quantum calculation% has ield-ing to an analysis of the bound-free continua by calculating ed disscidtion energies and vibrational constants for thethe continuous Franck-Condon factors (FCFI. since now emitting -tates. Despite several jb ih ttI' tiulah ions, f ol-

j Chem P" 95 (8) '5 October 991 0021-9606M/91205539- 4S3 00 199, Amerncain irsttuie ot Ppvscs 5539

5540 G. Rodnguez and J. G. Eden: Photoassocation of "'C4 and 64Zn 2

umn IIB interatomic potentials,"-" little is known of the Consequently, as illustrated by the schematic diagram of thehigher excited states of Zn 2 and Cd2 as previous experimen- experimental apparatus in Fig. 1, excited Zn2 or Cd2 dimerstal studies have generally focused on the visible and uv con- were produced by irradiating pairs of ground state atomstinua emitted by the lowest lying ungerade excited states in with one of several wavelengths available from a rare gas-transitions to ground.'6 halide excimer laser (ArF: 193 nm; KrCG: 222 nm, KrF: 248

n.m and XeCl: 308 nm). The Stokzs Lnes of ArF and KrF1I. EXPERIMENTAL DETAILS Raman shifted in H2 and the fourth harmonic of the

Nd:YAG laser (266 nm) provided other useful excitationSeveral cylindrical optical cells, 5.5 cm in length and 2.2 wavelengths as well. Unless stated otherwise, all spectra

cm in diameter (o.d.), were fabricated from Suprasil grade were acquired for pump laser pulse energies of 20 mJ, al-quartz and evacuated to < 10 -' Torr by a turbomolecular though considerably larger energies were available at severalpumping station. After degassing the tube walls with a hy- wavelengths. For the excimer laser experiments, the pumpdrogen torch, several milligrams of isotopically enriched beam was rectangular in cross section and the quartz cell was(98%) "'Zn or "'Cd were introduced into the cells. Stan- irradiated transversely at a pulse repetition frequency of 10dard distillation techniques were followed to further purify Hz. A cylindrical lens focused the incoming radiation to athe metal after which each quartz tube was sealed off under line that was coincident with the axis of the optical cell.vacuum. The cells were placed into a firebrick lined oven and When the Nd:YAG 266 nm laser wavelength was studied,temperature was monitored by chromel-alumel thermocou- the beam was propagated along the cell's axis.ples located at the cell side arm. Metal atom number densi- Subsequent fluorescence from electronically excitedties were determined from the vapor pressure curves. In zinc or cadmium molecules was collected through a port atthese studies, the Column IIB operating pressures ranged the top of the oven by a large cylindrical collection lens madefrom 20 to 650 Tort which corresponds to the temperature of Suprasil quartz. An aluminized turning mirror and a finalinterval of 789-1013 K for Cd and 905-1173 K for Zn. At focusing lens ( 15 cm focal length) directed the emission to400 Tort, the Zn and Cd number densities are 3.4 X 10 and the entrance slit of a 1/4 m spectrograph equipped with an3.9 X 10' cm -', respectively, optical multichannel diode array (OSMA) which consisted

Photoassociation is an optically assisted collision in of a 660 element array enhanced for the uv by a phosphor onwhich a colliding pair of atoms in the ground state's thermal the multichannel plate. The spectrograph was equipped withcontinuum absorbs a photon, resulting in the production of 600 and 1200 lines mm -' gratings, giving an overall detec-an electronically excited diatomic molecule. In zinc vapor, tion system resolution of Ai, = 3.5, 7.5, or 15.0 cm -', de-for example, Zn2 ('!, 7) molecules can be produced by the pending upon the specific grating and spectrometer order inbound -free process use. All spectra were wavelength calibrated with known

lines from low pressure Ar, Ne, Kr, and Hg atomic lamps. InZn(4s'So) + Zn(4s'So) + 4A-Zn, ('7. ,u). (1) order to supprebb scattered pump radiation, one of several

LS1 f1/73Computer

__ Timing Circuitry- SMAMaster Oscillator

Oven .<;ui 1tpr FIG I. Partial schematic diagram ofthe experimental apparatus. For thoseexcitation wavelengths not requirng

45' Mirror Lens Lens ! Raman shifting, the 45' mirror and;_'ci e • se" tight angle prsm in the excimer laser

beam path were removed. The illustra-

tion depicts the selection of the second--, ,. .. ol --I Stokes of ArF Raman shifted in H, for

excitation of Zn vapor.Coec'roqroom

),ode Array

oar;

j Chem -",ys Vol 9S. No 9 '5 October 1991

G. Rodriguez and J. G. Eden: Photoassociaton of "'Cd2 and 64Zn2 5541

liquid filters (n-butyl acetate: ACUTOFF = 254 nm; chloro- As reviewed in detail by Tellinghuisen,' the semiclassi-form: 245 nm; dichloromethane: 233 nm) " was generally cal stationary phase approximation has also proven to beplaced in the optical path. valuable in describing bound - free transitions. For a two

Data were acquired and stored by a DEC LSI 11/73 state molecular system having a single point of stationarycomputer. The diode array controller was operated synch- phase,"S the quantum mechanical FCF (overlap integral)ronously at 10 Hz in a gated mode. That is, 50 ns after the accumulates primarily in the region of internuclear separa-pump laser was fired, the OSMA was activated (gated "on") tion in the vicinity of this point of stationary phase which, inand spectra were acquired by integrating the emission signal turn, lies close to the classical outer (or inner) turning point.for 5 js. Both background and scattered pump radiation Considering photoassociative (bound- free) transitions,were accounted for in the data reduction procedure. for example, previous work has demonstrated that the sta-

tionat 1 phase approximation is a reliable guide in visualizingIII. THEORETICAL CONSIDERATIONS such situations. One concludes that the photoexcitation of

Both exact quantum calculations and the semiclassical colliding ground state Zn (or Cd) atoms to produce an elec-

stationary phase approximation" t have been remarkably tronic excited state of Zn 2 [i.e., Znr* (v',J') ] is assumed to

successful in describing bound-free and free-bound elec- occur predominantly in the vicinity of the classical outer

tronic transitions in diatomics. The Einstein A coefficient for turning point for the bound excited state, v'.

spontaneous emission from a bound state (v',.I') into thecontinuum of a (dissociative) lower state is given by theexpression' IV. EXPERIMENTAL RESULTS AND SPECTRAL

ANALYSISA~v'J',e) = (64-r/3h)s,. .. (2J' + 1 A. Zn,

x g'Vj (e',(R) v',d') 2, (2) Irradiating 400 Torr of zinc vapor

(["Zn] =3.4x10"'cm -,T=lll8K) with the beamwhereg' is the upper (bound) state degeneracy, s.., is the from an untuned XeCl (308 nm) laser yields the unstruc-rotational line strength, p, (R) is the electronic dipole mo- tured blue continuum illustrated in Fig. 2. The spectralment for the transition, and v' and e represent the bound width ( - 120 nm FWHM) and position of the spectrum arestate and free particle (associated with an atomic pair of in accord with previous gas phase studies of the mole-energy e') wave functions, respectively. Consequently, the culeI

6 ,a and the assumption that this visible emission is dueintensity profile of the bound-free emission spectrum can be to transitions from a low-lying metastable state of the dimercalculated by evaluating the square of the matrix element. (37+)I.i3 to ground ('Ii ). The lowest electronic-excitedThat is: states of both Zn2 and Cd, are the 3F1g and '7 levels that

A W (R)! '3) are derived from np 3p+ nSi S."1.t 3 Of course, neither() ( (R)v', (3 triplet state is strongly connected to ground by dipole transi-

where I(A) is the experimentally observed emission intensi- tions and parity considerations preclude emission from the

ty spectrum. If the dipole moment is assumed to vary slowly 3Ji state.

over the internuclear separation region of interest, then Eq. The bound-free emission continua that were recorded

(3) can be rewritten as following the photoexcitation of 'Zn vapor at shorter wave-lengths-222, 230, and 248 nm-are shown in Figs. 3-5,

1(A) =#M1(Elu',J)I'), (4) respectively. For each of the spectra, strongly modulated

where ( W IV',j') 'is the continuous Franck-Condon fac- Franck-Condon oscillations are observed which extend 60-tor. The almost sinusoidal Franck-Condon oscillations that 100 nm to the red of the pump wavelength. Continua exhibit-are frequently characteristic of bound-free transitions arise ing such structure were predicted theoretically by Condon "from the interference between the wave functions for the who coined the descriptive term "internal diffraction bands"bound excited state and the free atoms associated with the to portray the interference between the bound and free parti-ground state's dissociation continuum. cle wave functions. Although atomic Zn lines are also fre-

Exact quantum calculations of bound-free spectra, quently present, the bound-free undulations dominate thetherefore, require knowledge of (or an assumed functional spectra and the S/N ratio and reproducibility are excellent.form for) the upper and lower state interatomic potentials The weak, high frequency structure superimposed onto thefrom which one calculates the continuous FCFs. As will be oscillations of Figs. 3-5 is an artifact of the detection system.discussed later, the Zn, and Cd, potentials of interest here It should be noted that the Zn, spectrum produced with aare not presently well characterized. Consequently, in this 248 nm excitation wavelength is distorted at the shorterstudy we adopt the approach taken by Bergeman and Liao' wavelengths by the n-butyl acetate liquid filter which has aby assuming both molecular potentials to be well described rapidly changing transmission between 240 and 260 nm.by the Morse function. For convenience, the constant dipole Previously published Zn, emission spectra"6 have shown noapproximation (Eq. (4) 1, invoked by Scheingraber and Vi- evidence of FC interference. Note, however, that the oscilla-dal for the Mg, A-X system,' is assumed to be valid for the tions of Figs. 3-5 do not exhibit complete modulation whichZn, and Cd, 'Y., .' transitions. The limitations im- are evidence for emission from several bound or quasiboundposed by this assumption will be dealt with later. vibrational levels. Because of the depth of the B '1,* state.

, *ha PVolI 9rqN 8. 1 ,, .q4~r 9)

5542 G. Rodnguez and J. G. Eden: Photoassociation of '"Cd2 and "Zn 2

64Zn2 Emission Absorption

X'p4mp= 3 0 8 nm Spectrum

["Zn]=3.4x10"e cmr-,

,A

to-n

FIG. 2. Unstructured continuum pro-S , duced by irradiating "Zn vapor at

A =308 am (XeC laser). TheC: 5s S, -4p 1P, triplet in the 468-481 nm

region is also present. The inset shows the200 240 280 320- ultraviolet absorption spectrum of Zn ob-

- served under identical experimental con-O ditions (Zn vapor pressure = 400 Torr).

A xenon arc lamp provided the probingcontinuum for this measurement. Neitherspectrum has been corrected for the spec-tral response of the detection system.

320 400 480 560 640Wavelength (nm)

free-free emission is expected to be negligible. However, if Stokes of ArF in H2 ), 266 (fourth harmonic of Nd:YAG),one were to form electronically excited Zn2 (or Cd 2 ) states or 277 nm (first Stokes of the KrF laser in H2 ). This result isindirectly by photoexciting the atomic 'P, - 'So transition consistent with the location of the well-known +).+ -- +

and relying on three body collisions to produce the excimer, absorption band of Zn2 that has been observed in the gasa strong free-f ree component to the fluorescence spectrum phase and in rare gas matrices.20 As displayed in the insetwould then be expected and in the vicinity of the resonance to Fig. 2, this broad continuum also shows no structure,transition wavelength, in particular. peaks near 230 nm and has a FWHM of - 22 nm.

No molecular emission was detected in the 200-300 nm The emission spectra of Figs. 3-5 are, therefore, attrib-region when the vapor was irradiated at 193, 210 (first uted primarily to bound-free transitions originating from a

64Zn2Xpump= 2 2 2 nm

0 FIG 3. Emission spectrum observedwhen 400 Tort of Zn vapor((Znl = 3.4X 10"cm -) is irradiated at

.4) 222 nm (KrCl laser). Although the emis-.- sion profile has not been corrected for theSspectral response of the detection system,

the decline in intensity observed at A - 298CX nm is real. No liquid filter was used for

these measurements.

220 240 260 280 300 320Wavelength (nm)

J Chem Phys Vo 95. No 8, 15 October 1991

G. ,drOiguz and J. G. Eden: Phooas3oatiaon of "'C4 and "Zn 2 5543

64_n"Zn2

X,,.. 230 nm .- Z.(4 ' 'P)

C > -

0C In .

"(1p --.is.)02

0 Z,(4' S)zN(4. ' S)

R (.)

230 29o .. 270 290 310 FIG. 6. Partial energy level diagram of Zn2 developed by Hay. Dunning,Wavelength (nm) and Raffenettifromabinidocalculations (adaptedfromRef. 1i).Thepro-

duction of a specific ':1 vibrational level by the photoassociation (bound.FIG. 4. Spectrum similar to that of Fig. 2 except ,Aup, -230 nm (S2 of - free absorption) of a pair ofZn (4 'So) atoms is also illustrated. Bound-ArF lae in H2 ). -bound absorption also occurs at larger R.

limited number of vibrational levels of the lowest ' + state (C) '11, states both lie close to one another (and presum-of the dimer. In accord with international convention, we ably intersect at R - 2.5 A)," are correlated with the samerecommend that this electronic state be denoted B 'I +. The separated atom limit and are connected to ground by dipole-nonselective nature of the pumping mechanism utilized here allowed transitions, the calculated dipole moment for thegenerally precludes exciting a specific v state. A partial ener- B 'T' -X' -' transition (- 9 Dat 5.25 Bohr) is consider-gy level diagram of Zn2 , adapted from Ref. 13 and given in ably larger than that for C 'I, -X 1z . transitions ( 1.9Fig. 6, illustrates the fact that this 11 + state is derived from D). " One anticipates, therefore, that the observed bound-4p 'P, and a ground state atom (3d"t 4s2 'So ). Hence the free emission will primarily arise from the B 'Y,7 state.molecular fluorescence is correlated with a resonance transi- Previous studies22 "2 and the results to be presented lat-tion ('P, - ,) of the Zn atom having an upper state life- er indicate that the ground state of Zn, is bound by severaltime of - 1.6 ns (Ref. 21). Finally, although the B I1,+ and hundred cm -' and R, - 4-5 A. Consequently, as noted in

t64

6Zn 2XPUMP= 2 48 n m

C -FIG. 5. Bound-free spectrum of Zn2 ob-0served for an excitation wavelength of 248> rim. As in Figs. 2-4. [Zn I 3.4X 10"6

cm- The rapid falloff in the emission in-__ tensity as oneC approaches the KPF latser7ij pump wavelength is a consequence of a

I KrF butyl acetate liquid filter placed in front ofh Loser the spectrograph.Pump

240 260 280 300 320Wavelength (nm)

J Chen. Phy.,V0t 95,No 8, 15Oletobor I"gI

5544 G. Rodriguez and J. G. Eden: Photoassociation of "'Cd and "Zn2

Sec. 11, both bound - free and bound - bound processes con- fluorescence dictates that bound - free transitions will domi-tribute to the production of particular vibrational levels of nate the spectrum, and particularly at the longer wave-the B 1Y'. state. While photoexcitation of bound Zn 2 lengths which correspond to transitions occurring at smallground state molecules to specific upper state vibrational R.levels (i.e., bound-bound transitions) will occur, the Zn2 The computational results discussed in Sec. IV C willvan der Waals bound ground state is shallow and so the demonstrate that the oscillatory Zn2 (and Cd2 ) spectra arenumber of atoms in the thermal continuum of the ground well represented by assuming that the emission emanatesstate greatly exceeds the number of Zn2 X 'I + molecules from three or fewer vibrational levels. This is a plausibleavailable. Said another way, the equilibrium constant that result if one can assume that the B '7.+ -X IM

+ dipole mo-describes the balance between the ground state atomic and ment is a slowly varying function of R. In that case, themolecular populations favors translational degrees of free- B 'M7 radiative lifetime (several ns) discourages thoroughdom. collisional mixing of the pumped vibrational level with adja-

From a semiclassical point of view, the spectra of Figs. cent states, even though the background Zn number densi-3-5 are textbook examples of reflection emission spectra in ties used in these experiments are as high as 5 X l0"l cm - 3.which the Franck-Condon "diffraction"" 8 of one (or a few) However, because of the spectral breadth of the excimer la-upper state vibrational level (v') onto the dissociative ser's untuned output (typically - 100cm - '), it is not possi-ground state is observed. In terms of their overall structure, ble to specify a particular B 'M X vibrational level.then, these spectra indicate that the radiating and ground The Zn number density ( [ Zn] ) dependence of the Zn2states are characterized by a single point of stationary phase bound-free emission spectrum is shown in Fig. 7 which pre-which lies close to the outer turning point for a given B 'Y,7 sents data acquired at four different vapor pressures. Notevibrational level (Ref. 1). Consequently, the overlap inte- that the overall spectral envelope and the oscillatory struc-gral describing the absorption process acquires the bulk of its ture itself remain unchanged but, for convenience, the lowervalue in the vicinity of the classical outer turning point. It is pressure data have been magnified. The figure also indicatesnot surprising, therefore, that excitation wavelengths to the the oven temperature corresponding to a chosen pressure.blue side of the Zn 4p 'P, - 4s 'So resonance line (213.9 nm) Although the oven temperature was varied by less than 20%or to the red of the B ' -X ' molecular absorption to obtain these spectra, the continuum intensity drops dra-band (inset, Fig. 2) will not produce Zn2 reflection fluores- matically as [Zn] decreases. Specifically, data plotted in Fig.cence spectra. 8 for three specific wavelengths in the spectra demonstrate

As expected, the Zn2 spectra show no evidence of a that the time-integrated intensity varies quadratically withbound -bound emission component. Since the B 'I,7 equi- the zinc pressure. This result is to be expected since the pho-librium internuclear separation R, is almost 2 A smaller toassociative process which produces the Zn2 B '..' mole-than that for the 'Z,+ ground state," .'3.24 the FC region for cules requires two Zn atoms and the third "body" is a pho-

- 64 'Zn 2+~ +

- ,,,,=248 nmu g 4 To"i

LKr FIG. 7 Dependence of 'Zn bound-free" urnoser . 200 To" (x2) emission intensity on the metal vaporW um (1060 ) i pressure. For convenience, the lower pres-

r sure spectra have been magnified by differ-

' ent factors (shown in parentheses). The0 1! pump laser wavelength is 248 nm and the"1V 00 Tor. (x4) temperature corresponding to each Zn va-

- A 1008 K~por pressure is also indicated.L

50 To"i (.5)(961 it)

2A0 260 280 300 320Wavelength (rrm)

J Chem Phys Vol 95. No 8. 15 October 1991

(3. Rodrguz and J. G. Eden: Photoassociabao of "'04 and "Z 2 5545

Zn Pressure (Tort) Specifically, when photoexciting Zn, at 222 inm, B '1. mol-

ecules are produced in the v 85 state which, when radiat-S//ing, is "reflected" onto a portion of the ground state poten-

tial that is flatter than is the ground state segment availablea.lo X =260 rim///.- X=26 ~to lower-lying 'I ' vibrational states. Assuming the B 'I

z23 n,, potential to be described by the Morse function (discussed in.6 & 3o .Sec. IV C), then (B) forv' = 25 and 85, for example, are 3.9

and 5.2 A, respectively. The ApLum = 248 nm data of Fig. 9

demonstrate that the ground state's slope is changing rapidlyin this region of internuclear separation.

, B. 114Cd20 100

[ZnI (10" Cm"3) Figures 10-12 present similar spectra that were ob-tained for "4Cd2 . For this molecule, oscillatory emission

FIG. 8. Data at three wavelengths ia. the Znz bound-free emission spectrum spectra were observed only when the vapor was irradiated at(260. 283. and 300 nm), portraying the quadratic dependence of the time 248 or 266 run. Pumping at 222 nm did not produce anyintegrated intensity [fl(A,)dtl on the zinc number density, [Zn ]. The observable emission which is consistent with the fact that thedata demonstrate that the 'Z. emitting state is populated by photoasso-ia- Cd 5p 'P, .- 'So transition lies at 228.8 rnm. Comments simi-tion (involving the absorption of a photon by a pair of colliding Zn atoms). lar to those earlier regarding the electronic states of Zn2

could be made for Cd2 . As a result, a comparison of Figs. 5and 11 shows that, other than the overall intensity envelopes

ton. Spectra similar to those of Fig. 7 also display a linear and the smaller vibrational frequency for the heavier Cd,dependence of bound-free intensity on the pulse energy of molecule, the two spectra are quite similar. The longestthe pump laser (KrF, in this case). wavelength maximum in the Cd2 spectra lies only - 5 nm to

Before leaving this section, one other interesting quali- the red of the analogous peak in Fig. 5.tative feature of these spectra should be noted. Since the The dependence of the emission on the Cd vapor pres-

anharmonicity of the Zn2 upper state ('T. *) potential be- sure is illustrated in Fig. 13 for a laser wavelength of 248 nm.comes more pronounced as the vibrational quantum number Once again, the interference structure is more prominent atincreases, (R) also gradually rises. Therefore, as shown in the higher Cd number densities. It is also expected that, asFig. 9, the spacing between adjacent maxima in the bound- was the case with Zn2 , the emission spectra of Figs. 10-13free spectra are sensitive to the slope of the ground state. result from the non-selective excitation of several ' levels.

1000- . .. . . . '. . . . . .

- 64

E n

["Zn]=3.4x 10' cm -

FIG. 9. Spacing between adjacent maxi-ai. ma (denoted by the integer N) in the

z- bound-freeeission spectra ofZn 2. The

& "difference in slope between the 222 nmc (triangles) and 248 nm (circles) data is

a consequence of the fact that the groundX,,=222 m state potential ('I,* ) is less steep in the

4 U'-85 emission FC corridor than it is) AN,,=248 nm/ (v'- 8 i) for lower-lying vibrational states (since0 "'

(' -25) (R) increases with v' ). The approximateE -25)vibrational quantum number for the

emitting state is also indicated for eachE set of data and the solid lines represent2.. linear least-squares hits to the highest N-data.C

1000 10 20 30 40N

J Chom. Phvs. Vol. 95, No, 8, 15 Octob e 1991

5546 G. Rodriguez and J. G. Eden: Photossociation of "'Cd, and "tZn,

114 Cd2[ 1

14Cd]=5.7x10"7 cm 5,

T=842 K (50 Torr)

"3 XPUImP= 2 4 8 nm

C

o FIG. 10. Ud1 bund-free emission spec-trum recorded upon irradiating SO Torr ofCd vapor with a KrF laser (248 nm). A

"* liquid filter was not used in this case toblock scattered pump radiation.

i I , i I

240 260 280 300 320Wavelength (nm)

C. Spectral simulations dure which calculates the eigenvalues for a given potential

with a finite difference technique using a predictor-correc-

Numerical calculations carried out to simulate the ex- tor formula. Convergence is approached iteratively with a

perimental bound-free spectra incorporated only theX 'Z+ second order variational procedure. For the bound ('Z. )ground and B 'I -* excited states mentioned earlier. The ap- state, the solution was obtained in two segments by integrat-proach entailed integrating Schrddinger's equation to deter- ing inward and outward and the two regions were matchedmine the radial wave functions for both potentials. The wave by requiring the wavefunction and its first derivative to befunctions were calculated by the Numerov-Cooley proce- continuous when the two solutions were merged.

114

Cd2XPUMp= 2 4 8 nm

.['Cd]3.9x10'a cm - 3 (400 Torr)

C E=20 m J/pulse

C FIG. II. Expanded view of the KrF laser-I produced Cd. spectrum. For them data.

> however, the Cd vapor pressure was 400Torr and the !xcimer laser pump energy

I was held com tant at 20 mJ/pulse. Also, a" butyl acetate liquid was inserted in the op-

M2 tical beam path.

265 275 285 295 305 315Wavelength (nm'

J. Chem. Phys.. Vol. 95, No. 8.15 October 1991

G. Rodr*Az and J. G. Eden: Photoassociation of " CA, and "Zn 2 5547

114 Cd 2

,pup=2 6 6 nm

["1 Cd]=6.1xlO" cr (650 Tort)

C E=13.3 mJ/pulse

C* FIG. 12. Q 2 spectrum obtained for

Z Arus~r = 266 nm (Nd:YAG fourth har-monic) and pcd = 650 Torr. The single

* pulse laser energy was fixed at 13.3 uJ.

260 270 280 290 300 310Wavelength (nm)

1. Zn, ground state separated atom limit). Owing to the spectralbreadth of the excimer laser radiation and the thermal distri-

The wave functions were evaluated over the 2.50 to 7.50 bution of ground state metal atoms, several v' levels areA internuclear separation region with a radial grid size of accessible for any one pump laser wavelength. Consequent-2X 10' A. Also, spatial normalization was accomplished ly, the FCF contributions from two or more B 'I.' vibra-with the trapezoidal integration rule. tional levels were often summed to produce the final simula-

After calculating the vibrational wave function for the tion.bound state, the continuous FCFs, I (e' v',J') 1', were evalu- The interaction potentials necessary to simulate Zn,ated for 500 points over the - 274.0 to 15 000 cm - ' range in emission or absorption spectra are not well characterized. Inlower state energies (where zero energy represents the fact, until recently, the molecular constants for neither state

1 4Cd 2 4o To-

(982 K)

Xp,.m= 2 4 8 nm

C KrF

Laser/Pump 20o Tor FIG. 13. Vapor pressure dependence of

(9 0 ) Cd bound-free emission. The laser exci-

1 tation wavelength was 248 nm for all of- J the data shown.ZS

100 Tort(584, K)

50 Twr(842 K)

240 260 280 300 320Wavelength (nm)

I 00 'I,, or ", 0 r~ - -ln

5848 G. Rodriguez and J G. Eden: Photoassociation of '"Cd2 and 1N

were known. Fortunately, Czajkowski and co-workers"' re- rected for by allowing for emission from several adjacent

ported in 1990 the ground state vibrational constants that vibrational levels and bearing in mind the approximationwere measured in a molecular beam experiment in which the represented by Eq. (4).

0 ' (H,) van der Waals excited state was examined by la- Similar results are obtained when simulating theset spectroscopy. Given this starting point and assuming APUMP = 222 and 230 nm spectra. As an example, the theo-

that both states can be adequately represented by Morse po- retical and experimental spectra for PUMP = 222 nm aretentials, the task of simulating the spectra of Figs. 3-5 was compared in Fig. 15. As was the case for 248 nm excitation,reduced to iterating D,, Ro, and w, for the upper ('I*) the positions of the maxima and minima in the spectrum are

state. As initial values for the iterative process, the B '2. predicted reasonably well and perhaps 20% of the back-constants determined from the ab initio calculations of Ref. ground continuum intensity can be reproduced by introduc-

13 were chosen. In an effort to guarantee a unique and con- ing two adjacent vibrational states (v' = 84 and 86) as emit-

sistent solution to this process, a final set of constants was ters. The simulated spectrum is, however, unable to

not accepted unless it was capable of reproducing all of the reproduce the modulation depth of the intensity variations

Zn2 (or Cd2 ) experimental spectra, regardless of the partic, at the "blue" end of the spectrum. Stated another way, the

ular pump wavelength, simulated spectra consistently displayed significant oscilla-

Taking the X 'I constants to be those of Ref. 24 tory structure down to the pump wavelength whereas the

(D." = 274 cm -',w = 25.7cm- , -,x, = 0.60cm -'and experimental spectra did not. Only in a few limited cases can

R - = 4.80 A). Figs. 14 and 15 show the best simulations this difficulty be attributed to the spectral resolution of the

obtained for the 248 and 222 nm pumped spectra, respective- detection system. Rather, this discrepancy at the shorter

ly. The process of fitting the data was initiated with the 248 wavelengths stems from having, to this point, ignored rota-

nm spectrum since v' for that pump wavelength is the small- tional effects in the simulation. Finally, the obvious differ-

est of those for the three excitation wavelengths investigated ences between the envelopes of the experimental and calcu-

and the emission structure is simplest. Examination of the lated spectra in Figs. 14 and 15 are directly attributable to

undulations in the experimental spectrum (cf. Figs. 5 and 7) the constant dipole approximation [ Eq. (4) ]. Trial and er-

indicate that v'> 18 and early simulations showed that ror iteration of the electronic moment for the B-X transi-

'= 25. Later work bore this out. In fact, assuming that the tion with the goal of reproducing the experimental spec-Zn2 emission of Fig. 5 emanates from a single vibrational trum's envelope appears to be a feasible approach to estimat-

state (Fig. 14) yields a simulated spectrum that is nearly ing the variation of the relatue value of p over a limitedidentical to the experimental version except for: (1) the de- internuclear separation region.gree of modulation of the FC oscillations, and (2) the overall A summary of the Zn 2 B '7.-' spectroscopic constantsintensity envelope which degrades too slowly towards short resulting from this work as well as values already in the liter-wavelengths. Both shortcomings are, to a large extent, cot- ature is given in Table I. The uncertainty associated with

64 rZn 2

pup.=248 nm

V)C

4 FIG. 14. Comparison of experimental(top) and simulated Zn spectra for an cx-

>2, citation wavelength of 248 nm. The calcu-.OO lated spectrum assumes radiation only(D | , fro. v' =25.

240 260 280 300 320

Wavelength (rim)

J Chem. Phys., Vol 95. No 8, 15 Octo er 1991

G. Rodriguez aN4 J. G. Eden: Photoassociabon of "Cd2 and "Zn 2 5549

64 Zn 2

.

APUMP=222 rim

C

C FIG. 15. Spectra similar to those of Fig. 14except the experimental spectrum (top)

0 was acquired for an excitation wavelengthof 222 nm (KrCI laser). The calculatedspectrum assumes that only the Zn2 '1,*,

u' --= 85 vibrational level is emitting.ir

22 240 260 280 300 320Wavelength (nm)

each constant was estimated from the observed sensitivity of vibrational state. The significant differences (cf. Fig. 1 ) aresimulated spectra to the constant in question. the intensity envelope at the short wavelength extremum of

the spectrum and the weak continuous background underly-2. "4Cd, ing the FC oscillations in the multivibrational level case that

A similar procedure was followed in simulating the Cd, results from broadening of the individual undulations. Theemission data. The wavefunctions for the upper and lower fact that the undulations in the experimental spectra arestate potentials were calculated over the 3.10-6.20 A inter- weaker and their spectral widths larger than their counter-nuclear separation interval with a radial grid size of 10- ' A. parts in the simulated spectra testifies to the conclusion thatAlso, the continuous FCFs were once again evaluated at 500 a rigorous simulation will require the inclusion of more radi-separate points for lower state energies in the range of - 323 ating. vibrational levels. This conclusion is not surprisingto 8000 cm -'. The potential barrier height in Cd2 is lower considering the spectral breadth of the laser excitation pulse,than that in the Zn2 case since the difference between the the metal vapor densities ( - 10" cm - 3) and the vibration-upper and lower state internuclear separations is smaller for al spacings (6G values) in this energy region. It must beCd2 than for Zn2 [i.e., AiR, (Zn 2 ) > &R, (Cd 2 )]' emphasized, however, that adding several vibrational levels

In similar laser spectroscopic experiments to those men- has only a minor effect on the spectroscopic constants ex-tioned earlier for Zn2, Czajkowski, Bobkowski, and tracted from the simulation.Krause" measured the vibrational constants for the van der The CdB a1' structural constants derived fromWaals ground state of Cd2 . From a Birge-Sponer extrapola- matching the simulations to the experimental spectra aretion of o* (I'n.) -XO , (Ii,) vibrational bandheads, also listed in Table I along with the ab initio values reportedthe X *0 £ constants were determined to be: D,- = 323.0 in Refs. II and 13.cm - 1, 23.Ocm -',w"x' =0.40cm - ',andR -5.10A. Again, it is assumed that these constants are valid at I Rotational effectsshorter internuclear separations as well as in the vicinity of The impact of rotational effects on the calculated Zn2the van der Waals minimum. Therefore, assuming a Morse and Cd2 spectra was assessed by incorporating a centrifugalfunctional form for both the X '1, and B ','7 interaction term into the expression for the rotationless potential. As-potentials and taking the '1*'* ab initio constants of Ref. 13 suming the B '1,+ (W) rotational level manifold to be ther-as the starting values in the iterative process, the solutions malized, the first iteration took into account only an averagethat best fit the emission spectra produced by the 248 and value forJ (Jo, = 142 for Cd 2 ). A more realistic approxima-266 nm excitation wavelengths are shown in Figs. 16 and 17, tion was subsequently tested by sampling the rotational dis-respectively. For Cd2 , we present the simulations that result tribution at seven points (every tenth level in the J' = 112-when assuming that emission occurs from an ensemble of 172 interval) and weighting the individual contributions ap-vibrational levels and, as stated earlier, they are similar to propriately. Figure 18 illustrates the spectra calculated forthe results obtained when considering only a single radiating emission from the v' = 22 level of "'Cd2 (B) assuming no

5550 G. Rodrguez and J. G. Eden: Photoassociation of "'Cd. and "Zn2

1 Cd 2'E,,+(V =57-59)""'Z9+

.pump 2 4 8 nm

C

FIG. 16. Experimental (Pc = 400 Torr)

0 and simulated spectra for Cd2 '1.+ as-- suming that the V' = 57-59 levels are radi-O ating. The calculated spectrum is almost

identical to the v' = 58 simulation exceptfor the region below - 260 nm.

240 260 280 300 320Wavelength (rm)

; rotational effects (top), an average value for J (middle) and11 4 ) the multilevel approximation (bottom). Although the fit ofCd 2 the simulated spectrum improves somewhat as rotation is

' I (v'=22)"=.!"1 " added, the impact is slight. Comparison of the B state con-\,,,,=266 rrn stants derived from the simulations also revealed only small

differences attributable to the rotational disifibution. For/v' = 22, for example, D; is determined to be 8350 cm - ' if. .rotational effects are completely ignored and rises to 8400

1 1 cm - ' if the average J value or the weighted distribution areZ_ usd imilarly, R; and Pi' remain essentially unchangedfand the quoted uncertainties for the constants in Table I

\more than adequately reflect the error introduced by ap-proximating the rotational distribution.

v'=20-24 V. DISCUSSION AND CONCLUSIONSFew molecular constants for Cd2 and Zn, with which

C our results can be compared exist in the literature. However,0 the ab inito calculations reported by Hay et al." and Bender

C and co-workers3 yield Zn2 constants that are in reasonable0 agreement with the values determined from our experiments

= (cf. Table 1). Specifically, our constants are - 10% largeraA ~ \~ \! than those given in Refs. I I and 13. However, the constants

Jl reported here must be viewed with some caution because of, 'the assumption adopted earlier that Morse potential param-

2. . .. eters determined for the X I' state at R, are valid at much

260 280 290 300 310 smaller R. Consequently, we perceive the differencesWavelength (nm) between the constants resulting from this work and the val-

ues of Refs. I I and 13 to be negligible. Ault and Andrews'FIG 17 Comparison of Cd, '1:, -'1; simulated spectra showing the experiments26 with the Column 1iB dimers in rare gas ma-diferences between assuming that one Wo' = 22, top spectrum) or several trices, on the other hand, yield estimated values for D, and( = 20-24. bottom spectrum) vibrational levels are emitting. The excita-tiomwavetengthis266nmandtheexpenmentalspectrum salsogivenatthe a, that are the largest yet reported but are still consistenttopof the figure. with both the theoretical and experimental constants of Ta-

J Chem Phys.. Vol 95, No 8. 15 October 1991

G. Rodrnguz and J. G. Eden: Pliotoassocation of "C4 and "Zn 5551

TABLE I. Spectrocopic constants for the B'1: states of Znz and Cd2 .

Molecule D.(cm-') 6, (cm-) aA,(cm-') .R(A). Ref(s). Comments

Rare gas matrix12 065' 132 b

...... 20, 26 experiments8145 115 0.41 '" 13 ab initio

calculationsZn2 8065 107 0.36 ... 1 ab initio

9 010 ± 200 122 ± 10 0.40 ±0.04 1.19 ±0.02 This work

Rare gas matrix10650" 14" ...... 20. 26 experiments

Cd2 7740 78 0.2 13 ab initio8 250 ± 200 92:± 10 0.26 ±0.04 0.92 ±0.02 This work

'Estimated from the value published for Do (Ref. 26).'The value cited her is not strictly a vibrational constant but is rather the average vibrational spacing mea-

sured in experiments in which Zn2 and Cd2 were observed in absorption in argon matrices.° 4R, •R,, - R~1.

ble I. The spread in values is partially due to the fact that quently, the resulting value for D is also excessive. 26

previous studies did not have access to a well-known ground The simulation procedure described in Sec. IV C is ex-state potential. For example, Ault and Andrews' (Ref. 26) tremely sensitive to the value assumed for the difference be-determination of D was based on an expression which as- tween the equilibrium internuclear separations for thesumed a value for D ' that is now known to be too large [i.e., X '1' andB 'I,7 states (-AR,). Combining the value forDJ(Zn, )=2000cm- 'and D (Cd, ) =600cm -']. Conse- AR, resulting from the spectral simulations ( 1.19 A) with

Re ( CI-) given in Ref. 24 (R , = 4.80 A.) yields R ; = 3.61A which was used to calculate the I..' vibrational wave

. .. functions. Although this excited state equilibrium internu-

v'=22 , Y /jRefs. 11 and 13 (2.9 and 2.97 A, respectively), our value is

Lim only - 10% larger than the R, determined by Su et al. fromgas phase absorption measurements (3.30 .

Similar comments can be made regarding a comparison/ \ of the Cd2 constants. The values from Ref. 13 for ca, and D,

- _ _ __ in Table I agree with the constants derived from the present• r experiments to within - 12% whereas Ault and Andrews'C - v=22 ' I value (10 650 cm-') is 13%-14% larger than the otherSJ',= 142 two.

Our value for AR, for Cd , (0.92 A) yieldsa)1 R;('Y) =4.18A ifthe ground state equilibrium internu-

,,1 1A I clear separation is taken to be 5.10 .A (Ref. 25). Once again,0 jJ' ' I this value for R ' is larger than that predicted by the ab initio_@ J J, calculations (3.24 A) " but agrees quite closely with the

/\ I value reported by Suet al.23 (R, = 4.10.A). Although rota-- =22 '

-'= 1 •2- 72 tionally resolved laser spectroscopic experiments will be re-quired to accurately determine molecular bond lengths for

A ithe Zn, and Cd., excited states, it is reasonable to conclude,1,' that: ( I ) the spectroscopic constants derived from compar-

ll ing the simulated spectra with the experimental spectra are,60'plausible within an uncertainty of 10%-20% and 2) the de-

20 280 290 300 310 scription of both molecular states with Morse potentials ap-Wavelength (nm) pears to be a valid representation over a limited range in R.

Several comments, however, must be made regarding

FIG. IS Effect of varnous approximatmns regarding the upper state rota- the simplifications made in simulating the emission spectra.tional distribution on the calculated spectra for the v' = 22 level of 'Cd. Rotational effects affect the positions and widths of undula-( ~ ~ ~ ~ ~ ~ Rttoa efet affect th poiin an width of uoannadfecsnre;(mide) naeag auef

1. : (top) rotational effectsignored; (middle) an average value forJ: tions in bound-free spectra and. as discussed earlier, areand (bottom) a multilevel distribution in which the contribution from eachofseven rotationalstates nthe 112-1 7 2interval (i e. I12. 122. 132 . 172) largely responsible for the uncertainties in he constants ofis weighted appropriately Table I. Assessing the impact on the results of assuming a

J Chem Phys. Vol 95. No 8. 15 October 1991

5552 G. Rodriguez arnd J. G. Edem Pttotoassociatioi, of "Td. and "Zn,

dipole moment that is insensitive to R is more difficult be- len, W. Grosshandler) under Grant No. CTS 89-15795 andcause of the competing effect of collisional relaxation of the Air Force Office of Scientific Research (H. Schlossberg)B 1:; vibrational states on the emission spectrum. Evalua- under Grant No. 89-0038.tion of simulated spectra for a wide range in molecular pa-rameters indicates that the adoption of Eq. (4) affects pri-manly the spectrum's envelope and the constants D,, w., 'J Tellinghuisen. in Photodissociation and Phottorzation. edited by K. Pand iaR, of Table I are, therefore, expected to be valid to Lawicy. Advances in Chemical Physics, Vol. LX (Wiley, New York,within the stated uncertainties. Because of the present ex- 1985), pp.299-369

perientl cndiion an, fr a ixe laer aveengh, he 2C. R. Vidal and H. Scheingraber. J. Mot. Spectrosc. 65. 46 (1977); H.perientl coditonsandfora fxed ase waelenththe Scheingraber and C. R. Vidal, J. Chem. Phys. 66, 3694 (1977).overlap of partial waves for the photoassociation of Cd-Cd 'K. Sakurai and H. P Broida. J. Chem. Phys. 65.,1138 (1976).or Zn-Zn pairs to adjacent, excited state vibrational levels, it 'C. R. Vidal, J. Chem. Phys. 12. 1864 (1980).is not possible to excite a specific B 'X,7 vibrational level. 'T Bergeman and P F. Liao, J. Chem- Phys. 72. 886 (1980).

'G. Gerber, R. Moiler. and H. Schneider, 1. Chem. Phys. 81, 1538 (1984).Indeed, the data and simulations confirm that the spectra 'D. 1. Ehrlich and R Ni. Osgood. Jr., Phys. Rev. Lett. 41, 547 (1978);reported here are the result of contributions from an ensem- Chem. Phys. Lett. 61, 150 (1979).ble of radiating v' levels. 'S. M. Adler and 1. R Wiesenfeld. J. Mol. Spectrosc. 66, 357 (1977);

Chem. Phys. 43, 21 (1979).'J. Supronowicz. R. J. Niefer, J. B. Atkinson, and L. Krause, 1. Phys. B 19,

VI. SUMMARY L717 (1986).'0K. P. Lawley. D. Austin. J. Tellinghuisen. and R. J. Donovan. Mol. Phys.

Bound-free emission spectra of 'Zn, and 'Cd, have 62.,1195 (1987).been observed in the 200-300 nm spectral region by irradiat- P. J7. Hay. T H. Dunning, Jr.. and R. C. Raffenetti. J. Chem. Phys. 65.

igthe respective metal vapors with several discrete wave- 2679 ( 1976).ing "F H. M~ies, W 1. Stevens, and M. Krauss, J. Mot. Spectrosc. 72, 303

lengths from an ex .imer or Nd:YAG laser. The observed (1978)fluorescence is attributed to B '1: -X'1; transitions of 'C. F Bender. T. N Rescigno. H. F. Schaefer 111. and A. E. Orel. J Chem.

themolcue wer th bunduper tae i deivd fom Phy 7, 122(199)the~~M molecul wher th boun upper stat ise derived from Phyd7, 1122e (1979)K

np 'P, + ni' 'So (n = 4 and 5 for Zn and Cd, respectively). Rhodes (Springer. Berlin, 1984). pp. 28-32.For the higher metal vapor densities ( t 5 X 10" cm - '), the 'H. TatewakiMN. Tomonari, and T Nakumura, J. Chem. Phys. 82. 5608spectra are dominated by Franck-Condon interference os- (1985).cillations that are broadened by rotational effects and the '(a) In the gas phase see, for example. H. Hamada. Philos. Mag. 12. 50

(193 111. G. Winans. Phys. Rev. 37. 902 (193 1): R. E. Druflinger andcontributions from adjacent V' levels that are also radiating. Mi Stock. 1. Chem. Phys. 68. 5299 (1978); A. Kowalski, Z. Naturforsch.Also, the dependence of the bound-free emission intensity TetiA 39. 1134 (1984); (b) The emission spectrum of Zn, ina rare-gason vapor pressure (number density) demonstrates that the matrix was reported by W Schroeder, H. Wiggenhauser, W Schritten-'1$ excited molecules are produced primarily by photoas- ,lacher. aind D) M4 Kolb, J. Chem. Phys 86, 1147 (1987)

"The cutoff wavelengths for these filters are those for which the absorbancesociative processes involving two free ground state atoms is 1 0 for a path length of 1.0 cm.and an ultraviolet photon. Numerical simulations of the ex- 'E. U Condon. Phys Rev 32. 858 (19281.perimental spectra yield constants that, in general, are with- 'For a recent review, see Mi D Morse, Chem. Rev. 86, 1049 ( 1986).

in - 10%of b inuiavalus.2'J. C. Miller and L. Andrews. Appi. Spectrosc. Rev. 16,1 (1980)C H. Corliss and W R Bozman. Experimental Transition Probabilitiesfor Snect 'al Lines of Seventy Elements, NBS Monograph 53 (U.S. Gov-erment Printing Office, Washington. DC. 1962). p 540.

ACKNOWLEDGMENTS "K D Carlson and K. R Kuschnir. Phys. Chem, 69, 1566 (1964).'C .H Su. P K. Liao. Y Huang, 5.5S. Liosi, and R. F Brebrick. J Chem.

The authors wish to acknowledge the technical assis- Phys 81, 11)(1984)tance of K. Kuehl and K. Voyles. Also, one of us (G.R.) -M. Czajkowski. R. Bobkowski. and L. Krause, Phys. Rev. A 41. 277

exprsseshisapprciaton o NAA fo th awad ofa gad- 9 I)Oexprsse hi aprecatin t NAS fo th awrd f agra- M Czajkowski, R. Bobkowski. and L. Krause. Phys. Rev. A 40, 4338

uate fellowship under Grant No. NGT-70 107. This work (1989)was supported by the National Science Foundation (K. Gil- 2B& S. Ault and L. Andrews, 3. Mol. Specirosc 65, 102 ( 1977).

iChem Piys Vol. 95. No 8. 15 October 1991

Laser excitation spectroscopy of Ne2 npir 11g Rydberg states observedIn the afterglow of a corona discharge

0. J. Kane,) S. B. Kim, and J. G. EdenEveritt Laboratory, University of Jllinom Urbana, Illinois 61801

M. L GinterInstitute for Physical Science and Technology, University of Maryland. College Park. Maryland 20742

(Received 4 March 1991; accepted 4 June 1991 )

Several Rydberg series attributable to transitions originating from the a '.' metastable stateof the neon dimer have been observed by laser excitation spectroscopy in the afterglow of apulsed corona discharge. The n = 4-9 members of an npmr H, .- a(0-0) series have beenrotationally resolved and the upper states characterized. The probable multiplicity of the [I,states is discussed. An unusual loss of intensity in the Q branches for the higher n membertransitions and a weak perturbation between the v = I vibrational level of the 5fa '1' stateand the 6pir 'Is, v = 0 level are also reported.

I. INTRODUCTION yielded rough values for the first ionization limit of each of

The rare gas dimers are members of a larger class of the four dimers. Laser pump-probe techniques introducedinteresting molecules having dissociative ground states and by Ediger and Eden', for Kr2 involved: (1) multiphotonRydberg excited states. Herzberg' has proposed the term and collisional production of the initial (metastable) dimerRydberg molecules to describe these species which represent state; (2) production of higher Rydberg levels by probinga subset of the excimer molecules. Such diatomic molecules the metastable diatomic species with a second laser pulse;afford an excellent opportunity to investigate Rydberg- and (3) detection of the Rydberg levels by monitoring theRydberg interactions without the added complexities intro- atomic photofragment emission resulting from the predisso-duced by interactions with valence or ion pair states such as ciation of these molecular states. Experimental advantagesare present in N2, NO, the group II B dimers, or the diatom- associated with a pump-probe approach include improvedic hydrogen halides, spectral resolution (arising from the narrow probe laser

Several experimental difficulties have, however, hin- linewidth and operation at reduced gas pressures) and tem-dered study of the heavier rare gas dimers. For most of the poral information. However, while multiphoton excitationexcited states of the Ne,, Ar,, Kr2 , and Xe2 dimers, predis- is an effective technique for preparing the lowest excitedsociation dominates radiative decay. Consequently, the states of the Kr and Xe dimers, the - 15-22 eV ionizationspontaneous emission studies that proved so effective in elu- potentials of the lighter rare gases make the procedure lesscidating in detail the electronic structure of He2 are of less attractive for producing Ne2 or Ar 2 excited states. To cir-utility for the other members of the rare gas dimer family. cumvent this difficulty, Ne2 metastables were produced inFurthermore, the equilibrium internuclear separation for the afterglow of a pulsed discharge (and optically probed asRydberg levels built on the diatomic ion ground state described above) in the studies reported in Ref. 6, with the[ Rg2' (X -11 ) I and the van der Waals (vdW) minimum result that several bands of the neon dimer were resolved andof the neutral species' ground state typically differ by more a rotational constant was determined for the lowest '7than I A, which complicates investigation of the lower vibra- metastable state which agreed well with theory. More re-tional levels of dimer excited levels by photoabsorption. cently, Kim et al." extended the approach with - I GHzThus, past experiments generally have examined the high v spectral resolution to studies of the rotationally resolvedportions of selected excited states by either: ( 1) single pho- '2 + and [IF components of the 4f, 5f, and 6f complexes ofton absorption (or photoionization) experiments involving Ne2 .ungerade Rydberg states excited by incoherent lamp, tuna- In this paper, the results of similar experiments involv-ble VUV laser, or synchrotron radiation; or (2) multiphoton ing npr I5 Rydberg states of Ne2 are presented. The (0-0)laser excitation combined with ion mass or photoelectron bands of transitions originating from the lowest 'I. excitedenergy spectroscopies. state of Ne2 (denoted here as a '1+ ) and terminating at

In 1985, Killeen and Eden' reported low resolution ab- npir 11, states (n = 4-9) have been rotationally resolvedsorption spectra of electron beam-produced rare gas plasmas and characterized. Several additional Rydberg series alsowhich revealed clear Rydberg series believed to originate are discussed briefly and a perturbation between the 6pr 1lfrom the lowest excited states of Kr 2 and Ar2 . When com- (v = 0) and 5fo, '; (v = ) levels is identified.bined with subsequent' work on Ne2 and Xe2 , these data

II. Ne ORBITAL AND STATE LABELS

P"w-h s aIdrm: CLS-4 Divisi, Lw Alama National L tory. Los The majority of the excited states reported here appearAla,,r "!M 87545. to be associated with the addition of a Rydberg electron to

J. Chem. Phys. " (6), 1 Seotewnber 1991 0021-9606/91/183877-11$0300 Z 1991 Amencan institute of Physics 3877

3878 Kane etal.: Spectroscopy of Ne2

the X 1Z1 ground state configuration89 of Net, sistent with Eqs. (1 )-(4), we note that caution should be

sitn wit E qsIr lr3oX. (1)-4,wnoetacuinshldbla 22a23a 2 1 r I 3 X (1) employed in interpreting the principal quantum number (n)O8 J 8 value as anything other than a relative index.

to produce the Rydberg configurations

to, 2 to .2a 2 .3o 2 li l' 3o,. nM. (2) Ill. EXPERIMENTAL APPROACH

TheX 1 ground state of the neutral dimer can be associat- Details regarding the experimental apparatus and dataed with the configuration acquisition procedure have been described previously."' °

I .2a 2 2 2 I Briefly, production of the a 37 metastable dimers is ac-eo 9 I9 .2 2 3 . Ile a' , (3a).complished with a corona discharge (having an E/N= 390which, when expressed in terms of the cm united atom Td; I Td- 10 - " V cm') in research grade 2'Ne or 22Ne gaslecular orbital (UAMO) designations, becomes at room temperature pressures of 120-200 Torr. At an adjus-

Isa 22pa "2w 3pa 23d7 22pr43dir44pa 2. (3b) table time delay (At) into the plasma afterglow, the dimer

Excitation of a 4po electron to the next available nsa orbital, metastables are photoexcited by a pulsed dye laser having a3s, results in the formation of the a 3 + and A '., excited linewidth of either -0.04 cm- ( - 1.3 GHz) or 2 cm-states, the lowest excited singlet and triplet A = 0 states, (for survey spectra). Atomic Ne fluorescence resulting fromfrom the configuration the predissociation of the photoexcited Ne states is collect-

(ls'2po 2sa'3p'3dcr'2p,3dr,4pa)3sa' 1.3y+ ed at 90' to the laser beam path (and the discharge axis) by aSI a4lens telescope and is detected by a cooled photomultiplier.

(4a) Dielectric bandpass filters were used to isolate the particular

which can also be expressed as Ne transition of interest which, for most of the experiments+ to be described here, was the 3p( 1/2), - 3s(3/2)o transition2 U I ,at 703.2 nm. Because of collisional mixing of the fine struc-

where 3sa denotes the Rydberg orbital and the square brack- ture levels of the ns and np atomic manifolds, few differenceseted term identifies the ground state configuration of the are observed between excitation spectra acquired by moni-molecular ion core. The usual practice in labeling Rydberg toting different np - as transitions. Throughout each spec-states is to specify an appropriate Rydberg UAMO and term tral scan, the relative dye laser wavelength was monitored bysymbol such as 3sa 'I Addition of a standard alphabeti- an etalon [free spectral range (FSR) = 2 cm ], while ab-cal label yields 3sa a 37 which we will usually shorten for solute wavelength calibration was carried out by optogal-convenience to a X. From Eqs. (3b) and (4b), it follows vanic spectroscopy in a U/He discharge. Also, the dye laserthat the first niA UAMOs outside the core with I = I or 3 are pulse energy was maintained by computer control to within5pa, 3pir, 4fa, 4fir, 4f3, and 4fo. If the 21r. orbital were to lie ± 2% across the entire scan region for a given dye.below Ir, as it does in He2 , then the lowest npr orbital Figure 1 shows the contrast between the temporal his-would be 4pir. While we will employ a labeling scheme con- tories of the Ne 3s(3/2)0 atomic and Ne a '7.' molecular

10

3 6po3+ - a

3 4p[ 3/2] 3s[ 3/2] 2 FIG. 1. Temporal variation of the excited4 -A atomic (open squares) and diatomic

0/ (open crosses) neon populations moni-

tored by laser-induced fluorescence in theO LJ• afterglow of a pulsed Ne discharge. The

z 0 Ne pressure was 200 Tort and the relativeZ 5 Ne 3s(3/2), metastable and Ne, a'l+

,g, number densities were recorded by excit.ing the 4p(3/2), - 3s(3/2)0 (A = 345.08

I- nm) and 6par37; -a'l.* transitions< (A = 425.25 nm), respectively, and moni-

*L toring the relative intensity of the Ne• o * 3p( 1/2), -3s(3/2)0 transition. Note

* that the peak atomic signal is off scale.

'0

0 o , 0o M MO!0 50 100 150 200

TIME DELAY, At(/4s)

J Chem Phys., Vol. 95. No. 6, 15 September 1991

Kane etal.: Spectroscopy of Ne2 3879

10

0Ne 2 A

X=425.25nm

CUZ FIG. 2. Pressure dependence of the rela-Wt tive No2 a '1: number density (moni-I-Z tored at 425.25 nm) for At =

20 ps. The1solid line drawn through the data repre-

',J sents a quadratic variation of the Ne6po.-a transition intensity with the

r\ neon pressure p.

0.1 . .......100 1000PRESSURE,p(Torr)

populations following the discharge. The atomic metastable from a number of experimental trials, are evident. The mostnumber density [ monitored by pumping the obvious of these is the virtual absence of 3p' emission in the4p(3/2), -3s(3/2)' transition at 345.08 nm] decays laser-induced fluorescence spectrum of Fig. 3(b)-notemonotonically following the termination of the discharge, especially the decline of the 3p'( 1/2),-3s'(1/2) 0 andwhereas the molecular population [observed by atomic la- 3p'( 1/2), -- 3s'( 1/2)0 transitions at 585.25 and 659.9 nm,ser-induced fluorescence upon pumping the 6pa-a(0-0) respectively. One is unable to invoke collisional processes totransition in the blue-425 nm] peaks - 20/As into the after- explain the differences in the spectra of Fig. 3 since both wereglow. Experience has shown that the long effective lifetime acquired under virtually identical conditions. Rather, theofthe Ne2 a ' metastable state results in excellent experi- dominance of emission from 3p states [and themental signal-to-noise ratios for At as large as 100/Ms. In 3p( 3 / 2 ), - 3s(3/2)0 transition lying at 638.3 nm, in partic-addition, as illustrated in Fig. 2, the relative Ne2 a 'I,+ ular] is likely attributable to: (1) the higher energies of thenumber density varies quadratically with the Ne pressure for Ne(3p') states relative to the 3p manifold and (2) the posi-

pN. 5 600 Torr. One concludes that the temporal and pres- tion and slope of the repulsive interaction potentials, corre-sure dependencies presented in Figs. I and 2 are consistent lated with Ne(3p') + Ne( So) in the vicinity of the pho-with those expected for a transient molecular species formed toexcited Ne2 state (6pe r7).by a three body collision.

Not surprisingly, the atomic fluorescence resulting IV. RESULTS AND DISCUSSIONupon photoexciting a Ne 2 inter-Rydberg transition differs The energy levels E( for relatively unperturbed Ryd-markedly from that observed during the active discharge. A berg series follow the expression'"comparison of the spectra recorded with a 1/4 m spectro-graph (30 A/mm reciprocal dispersion in first order) E(n) - E(n = ) = - _ R (5)equipped with a multichannel array detector (OSMA) is (n) (n- 6 )Zgiven in Fig. 3. In acquiring these data, the intensified detec- where E(n ao), R, n*, n, and 6 are the ionization energytor array was gated "on" for 60 ns. Final spectra are the for the series, the Rydberg constant, the effective quantumresult of averaging 2400 shots, obtained at a repetition fre- number, the principal quantum number, and the quantumquency of 2 Hz, and accounting for both scattered dye laser defect, respectively. For 2°Ne2, R = 109 735.81 cm -'whileand ambient radiation. The top spectrum [Fig. 3(a)] was the values of E(n = w) and E(n) depend on the referencerecorded just prior to terminating the corona discharge, zero of energy chosen. In the Ne, data presented below, wewhile the data of Fig. 3(b) were acquired 20 As into the have chosen to use the N = 0, v = 0 rotational-vibrationalafterglow by pumping the 6pti JI, -a(0-4) transition at level of the a '2,I state as the zero of energy, and have em-425 nm. In the absence of a dye laser pulse, no fluorescence ployed the level energies reported in Table V of Ref. 7was observed in this spectral region for At = 20ps. throughout this work.

Several reproducibl, differences between the two spec- As discussed in Ref. 7, triplet splitting has not, to date,tra of Fig. 3, which are representative of observations made been resolved in Ne, spectra which implies that excited state

J. Chem. Phys.. Vol 95, No. 6,15 September 1991

38M Kane et al.: Spectroscopy of No2

4 (a)

Discharge3 Fluorescence

*-3p'[ 1/2]. 3s'[ 1/2] 0FIG. 3. Comparison of (a) the atomic Ne

2 spontaneous emission spectrum observedz) from the active discharge (i.e., no probe

LLJ 1laser pulse) and (b the laser-inducedz atomic fluorescence (LIF) detected in theZ L d afterglow for a dye laser wavelength of

LJ 0 - 425.25 nm (corresponding to the peak of

> 4 (b) the unresolved 6pa '1, -a 'I band)and At = 2 0 ps. Without a dye laser pulse,

.- 3p[ 3/2] , -3s[ 3/2] of no emission was detectable in this spectral

~j3 LlF(A',,I425.25 nm) region at 20 jis into the afterglow. Two of

At =20 js the most prominent lines are identified.

2

0LA535 585 635 685 735

WAVELENGTH (nm)

structures are described by case b 'coupling (singlet or unre- of one R MN and one P( N) branch having only odd N val-solved spin multiplets). This result is analogous to the situa- ues, where N is the quantum number for total angular mo-tion observed for He, where such splittings are generally mentum, excluding spin. Case b' transitions of flq-YI,7 typenegligible. In addition, because the nuclear spins are zero in will have three branches which can be divided into two'N only alternate branch lines will appear in I1* -2, groups: one R(N) and one P(N) branch with N odd forandnI,-I,7 type transitions. Consequently, each band of 11 + -Z I. transitions and one Q(N) branch with N odd forNe 2 associated with transitions of I + -1, type will consist Ii, -1, transitions. The n1 + and fI states are reflection

200 rorr 2 Ne2

FIG. 4. Excitation spectrum of2ONe2 (a '2:) in the 300-370 nm region.In addition to the npr I'l, -a '1:7 (0-0)series discussed in this paper, the domi-

Z nant series, attributed to Wv~ 'I; -a(0-Lii 0) transitions, and several members of the

S300 305 .310 315 320 325 330 335 rn-aoo ndnuZ .aOOz __________-a_0_0__and__________-a_0_0_

Rydberg series ( Ref. 7) are also indicated.Lii nf)k (tZ.'T,)-o at: 4 The n =4 member of the npnr H._a(0.> nprn CE 0) series lies at A. 417 nm and is not in-

-- n; 'E -at: cluded here (see Fig. 5). Also, the "Nepressure was 200 Torr and the dye laser

Lii linewidth for this survey was 0.2 cm

335 340 345 350 355 360 365 370

WAVELENGTH(nm)

J. Chem. Phys.. Vol. 95. No. 6.1 lSeptember 1991

Kane etal.: Spectroscopy of Ne 3881

4pir' Ig- a ,(O-O) 20Ne2Q(N)

R(N) 25= N

F/ FIG. 5. Laser excitation spectrum of "Ne,

Z~. in the 23950<iS24010 cm-1 region

showing the resolved P and R branches ofz P(N ) the 4pr l, -a '1, (0-0) transition. While

the strong Q branch is collapsed and not re-. N =7 1 solved, theN I linesofboththePandA

f__ I I I I1 Ibranches are evident. The pressure of the99.5% enriched Ne was 200 Torr. The

LaJ ,lower trace shows the scan of the etalon'sLjJ . relative transmission (FSR = 2 cm" ')over

a portion of the frequency range.

ETALONTRANSMISSION

23950 23970 23990 24010

;(Cm-,)

symmetry eigenstates which are referred to as A doublets prominent bandheads in the figure (and another not pic-when their energy splittings are small. As A doubling in- tured at 425 nm) constitute a Rydberg series that, for quali-creases, the effect is known as I uncoupling. tative reasons, was tentatively attributed previously3' , toA. Experimental spectra, effective rotational constants npirl " -a 'I+ (0-0) transitions with n = 5-10. Further

experiments at higher resolution have shown no evidence ofA panoramic view of the 300-370 nm region of the Ne2 a Q branch. Consequently, while the details of rotationally

a 'Z excitation spectrum is displayed in Fig. 4. The most resolved spectra of these features will be discussed else-

5p7T rIT +- 0 o,-(O-o)

(J) 28330 28350 28370zLiJ P(N) FIG. 6. Laser excitation spectrum of the

5p[r 1, -a(0-0) band obtained at a resolutionZ N rT27 TT of 0.2 cm -'. Note that the Q branch is not re-

'III""" solved. The inset gives a higher resolution--> (Aie0.04 cm scan of the 28 320-28 370

cm 'region.

-J

LULSI 111111 II i i 1 I iI i

29= N

28290 28330 28370

J. Chem. Phtys.. Vol. 95, No. 6, 15 September 1991

3882 Kane et al.: Spectroscopy of '2

P(N)35 S5 5 6p7F n, -3EZ. (0.0)|5fX (3E"3,) - o'E:(l-0)

W N=s R() ssR 0 (N)N=5 R(N) 35 u Rn) NFIG. 7. Overview spectrum of the 30 390-

N=3 , 23 30 590 cm ' region showing 2"Ne, rota-

() RZ(N) .. . .. tional structure arising from 6p r fi, -a(O-

Z 39 29 23 17 11 5 P(N) 0) and 5.i( 31,,'I,)-a(I-0) transitions.W 25 13 Note the perturbation occurring between

-- rr-ro the 6pr fl, (u= 0) and 5fo '1,, (u I)states near 30 465 cm I For convenience,

LLJ the P and R branches associated with the> '1* or II components of the 5fcomplex

Q(N) are denoted by I or "subscripts.

I I 1 I kk /

17 P(N) 5

.I I I I .l

30390 30430 30470 30510 30550 .30590

z(cm- )

where, 2 suffice it to say that the strongest bands of Fig. 4 are remainder of this section deals with the analysis of a thirdassigned as pc 'I -a 'Z (0-0) transitions. Ne2 molecular series displaying rotational structure. One

As can be seen in Fig. 4, the increased sensitivity of the member ofthis series (near.I - 417 nm), previously labeled"laser-induced fluorescence technique, coupled with pulsing as the (0-0) band of the 4pir 3 Hg -a transition, is now iden-the photomultiplier, reveals several Rydberg series which tified as the n = 4 member of the npir g -a series. Sixare considerably weaker than the npa 31 .- a(0-0) transi- bands of this new Rydberg series have been rotationally-tions, but still well above the noise floor. Rotational analyses resolved and analyzed. Figures 5-10 present representativeof the strongly i-uncoupled 3 Y+ and 3nl t components of the spectra for the (0-0) bands of the n = 4-9 members of the4f, 5f, and 6fcomplexes of Ne2 were reported recently.7 The nplr i, -a 'I+ (0-0) series that were recorded in 200 Torr

N -25 7

N 1 5 7 ., (- -\

rn-n SpT rl -0i-0)

P(N) 7 p? flq -o(0-0) P(N)

Z R(N)UiUj ZSN,.

35

lI r I I II I 1 1 1 1N-SLL R(N)> 5 23

Q(N) - R(N)

L~A_ 32300 32320 32340 32360

31580 31600 31620 31640 31660 31680

FIG. q. 32 300-32 370 cm region showing the n = 8 member of the

FIG. S. 7pr 11, -a 1 (0-0) excitation spectrum of Ne, in the 31 580- n-pir 11, -a Rydberg series. Only those members of the P branch that are31 680cm I region. consistently reproducible are indicated.

J. Chem Phys., Vol. 95. No. 6. 15 September g91

Kane etal.: Spectroscopy of Ne2 3883

Q branch is still not resolved, but the lowest terms of both the9p (0-0) P and R branches are now observed. In comparing Figs. 5-

P(N) 9'fl -o(0O a10, it is obvious that the Hg ,-a transitions become weakerS 2- N R(N) and more difficult to resolve as the principal quantum num-

2 rrn N - ber increases.F- Rotational line positions for the n = 4-9 members of theZ 2°Ne2 npnr Hg -a(0-0) Rydberg series in Figs. 5-10 are giv-

en in Table I. The data of Table I demonstrate that all of theztransitions share the same lower state. The

t> A2F-(N) = R(N- 1) - P(N+ 1) lower state energy sep-Q(N) arations determined from the bands of Table I agree, to with-

in experimental uncertainties, with the published' energyspacings of the a 'I' (v = 0) etate, thus confirming ourlower state assignments." The upper state assignments giv-en in Table I will be discussed in the next section.

Energies (term values) of the upper state levels were0 , 2L90 , 2, , 23 calculated from the data in Table I and the a 'I,7 (v = 0).32770 32790 32810 32830

T(cm- 1) rotational level energies reported in Ref. 7. The results aresummarized in Table II in which all entries are referenced to

FIG. 10. Ne, excitation spectrum in the 32 770.32 840cm - 'region show- N = 0, v = 0 of the a state. The data of Table II areing the partially resolved 9p r [Is- a bands, estimated to have an overall uncertainty of + 0.15 cm- '

Band origins and effective rotational constants determinedfrom the data of Tables I and II using the power series rela-tions and least-squares fitting procedure described previous-

of 5'Ne (99.5% enriched). Similar spectra were also record- ly7 are summarized in Table IIl.ed for the stable isotope 22Ne. Although the 4pr 11, - a .Un(0-0) band was published previously, the resolution avail- B. 0 branch intensity anomalyable at that time (0.2 cm- I) was considerably lower than One intriguing trend noticeable in Figs. 5-10 is thethat for the spectrum of Fig. 5 (-0.04cm -'). The apparent weakening of the apparent Q branch in the npir Hg -a +

TABLE 1. Energies (cm - ') of the rotational lines observed for the npr H, -a 1.+ (0-0) transitions of :°Ne,, n = 4-9.

4pr 11, - a(G-0)' 5/ir n, -a(0-0), 6ptr 1, -a(0-0)' 7pr n, -a (0--0)f 8pVr fl, -a (0-0)= 9p € n, -a (O-0)

N R(N) P(N) R(N) P(N) R(N) P(N) R(N) P(N) R(N) P(N) R(N) P(N)

I 23975.08 2 3 97 1.6 8b 28326.83 23 977.49 23 969.54 28 329.405 23980.02 23967.43 28331.93 30432.63 30419.1 31 610.88 32328.47 23982.53 2395.35 28 334.80 30435.15 3u417.79 31 614.24 31 596.41 32 331.64 32313.6 32 806.039 23985.31 23963.48 28337.22 28315.4 30438.07 30416.04 31 617.50 31 595.15 32335.02 32312.5 32 809.02 32786.87

i 32988.01 23961.61 28340.20 28313.54 30441.27 30414.43 31620.55 31593.79 32338.64 32311.6 32811.85 32785.4113 23990.83 23959.92 28342.73 28311.86 30444.19 30413.13 31624.01 31592.53 32342.46 32310.55 32814.76 32783.7415 23993.77 23958.22 28345.94 28310.17 30447.56 30411.52 31628.01 31591.17 32346.26 32309.63 32817.67 32782.0417 23 996.82 23956.72 28 348.98 28308.66 30450.63 30410.37 31 631.48 32350.16 32 309.29 32820.6 32780.7819 23 999.74 28 351.86 28307.14 30453.84 30408.75 31635.36 32353.81 (32308.36) 32824.1 32779.421 24002.99 28355.40 28305.94 30457.44 30407.87 31639.57 32357.8 (32308.36) 32827.2 32778.223 24006.74 28358.62 28304.78 30460.37 30406.86 31643.78 32361.6 32307.2 32776.525 24010.62 28 362.00 28 303.60 30463.40 30405.41 31 647.9827 28365.38 28302.3 * 30403.95 31 651.9829 28368.60 30469.19 30402.06" 31656.5131 28371.3 30472.25 30401.03 31 660.7233 28374.9 30475.63 30 399.58 31665.1435 28377.9 30478.69 30398.56 31670.6

* Parentheses denote a blended line.

bTbeintense peak near 23 973cm - 'is likely an unresolved Q branch (see the text and Fig. 5). Ifthe feature at 23 971.68 cm 'isP( I) asitappearstobe. then

the upper state cannot be a I n state.'The moderately intense peak near 28 325cm - is probably an unresolved Q branch (see the text and Fig. 6).'Higher N-value levels perturbed by v = I ofthe 5fo '1, state. A very weak peak near 30 425 cm - ' is likely an unresolved Q branch (see the text and Fig. 7).Perturbed assignment sull questionable. The corresponding A(27) line would appear as a poorly defined shoulder near 30 466 cm - '+

'Higher P(,) lines unresolved with a peak near 31 589 cm - '. Weak head near 3 604 cm -' probably is an unresolved Q branch (see the text and Fig. 8).I The very weak feature near 32 320 cm - ' possibly is an unresolved Q branch (see the text and Fig. 9).'Weak features near 32 798 cm 'are likely an unresolved Q branch (see the text and Fig. 10).

J Chem. Phys., Vol. 95. No. 6, 15 September 1991

3864 Kane atal.: Spectroscopy of Ne.

TABLE II. Level (term) energies (cm - ') of the npir 11, states of "Ne, for m = 4-9.

N 4p "iri (u=0) 5pITfit (u=0) 6pirll, (u=0) 7pir [ t (v=0) 8pr I11 (u=0) 9pr l1 (v=0)

0 23 972.8Wb2 23976.24 28327.94 23 984.26 28 336.19 30436.06 23 996.95 28 348.76 30 449.1 31 627.76 32 345.38 24014.06 28 366.25 30466.65 31 645.75 32 363.13 32 837.51

10 24035.89 28 387.82 30488.68 31 668.10 32385.70 32859.6512 24062.27 28414.38 30515.50 31694.81 32412.90 32886.1014 24093.22 28445.14 30546.55 31 726.34 32444.74 32917.1016 24128.78 28480.86 30582.52 31763.02 32481.25 32952.76i8 24 168.89 28520.95 30622.65 31 803.55 32522.23 32 993.020 24 213.43 28565.62 30667.58 31849.05 32 567.50 33037.922 24262.77 28615.16 30717.22 31899.35 32617.6 33087.024 24317.10 28668.97 30770.75 31954.14 32671.926 24375.98 26 727.30 30828.75 32 013.3428 28 790.18 30 890.69' 32 076.7830 28 857.23 30957.85 32 145.14

32 28928.2 31029.09 32217.5734 29004.4 31 105.12 32294.6336 29084.5 31 185.25 32377.1

Ba ed on N = 0, u = 0 of the a '1: state.'See footnote b to Table 1.

"Value may be questionable (see footnote e to Table 1).

transitions as n increases. While the Q branch is reasonably obtained, the S/N ratio was too low to yield the detailedstrong for the n = 4 member (Fig. 5), the analogous transi- information required here. Subsequently, a two dye lasertions for n > 5 are barely discernible. As discussed below, it is pump-probe technique was developed which has proven tothe existence of the strong Q branchlike features in then = 4 be quite fruitful in observing weak inter-Rydberg abscrptionand 5 members which leads one to label the upper states for spectra.this series as 119 type. With the two laser pump-probe approach, one dye laser

One possible explanation for the Q branch intensity be- scans the spectral region of interest as before. However,havior outlined above revolves around the relative rates at small changes in the lower state population (in this casewhich different Rydberg states predissociate. In the experi- a '2, ) can be monitored indirectly by tuning a second dyemeats which produced Figs. 5-10, if a particular state is laser onto the strong 7po' ,-a(0-0) transition at 356stable against predissociation, its presence in the excitation nm' 2 which terminates on a predissociating Ne2 state. Withspectrum might be insignificant or go undetected altogether. this laser locked onto the predissociative transition, one thenSpecifically, it is conceivable that the 11 + and fl - states monitors the atomic 3p- 3s emission resulting from decay ofbehave entirely differently with respect to dissociation. One the dissociation products. If the pulse energy of the secondapproach to investigating this possibility is to compare the dye laser is kept constant, the peak atomic fluorescence in-excitation and absorption spectra for the same transition. tensity is a convenient measure of the instantaneous a 'T.1Toward that end, attempts were made to record Ne2 a 'I ' (v = 0) population. Consequently, even slight reductions inabsorption spectra by dye laser probe, multipass techniques the a 31+ state population that result from tuning the firstwith a UV-preionized transverse discharge. Although the dye laser through various transitions are readily detectable.experiments were successful in the sense that spectra were In other words, this technique effectively converts weak ab-

TABLE III. Effective molecular parameters (cm ') for the v = 0 levels of the npr I (n = 4-9) states of20Ne2 ,

State E(N=-0) jS6 0,.x 10 D.x 101 H.x 10'

4p. ( ' 23 972.80 ± 0.08 3.2442 5.743 t 0.008 0.9 ± 0.5 1.2 ± 0.8

5pr 1 28 324.5 +0.1 4.250 5.758 ± 0.010 0.5 ±0.4 0.1 ±0.26pr r' 30 424.6 ± 0.21 5.255 5.82 ± 0.02' 0.9 -0.5, 0.5 ± 0.3c7pir II 31603.3+0.2 6.265 5.89+0.02 0.9±0.5 0.2±028pi" fI 32 320.2 ± 0.2 7.266 5.98 0.02 2.2 ±0.7 1.1 ±0.59pir! 1 32796.1 + 0.2d 8.274 5.77 t 0.02 5.2 ±0.7" 0.1 0.2d

Energy of the extrapolated N = 0 level above the N- 0, v = 0 state of a 17,'

'Calculated from Eq. (5) assuming an ionization energy of 34 399 cm (Ref. 10).'Weakly perturbed by v = I of 5fo e'l,'.

Perturbed band.

J Chem. Phys.. Vol. 95, No. 6,15 September 1991

Kane eta.: Spectroscopy of Ne2 3885

The bottom trace in Fig. 11 is a segment of theTWO LASER 5pir 11, -a(0-0) excitation spectrum illustrated in Fig. 6,

P(N) obtained with a single scanning pulsed dye laser by monitor-ing 3p( 1/2), -. 3s(3/2)' fluorescence at 703 nm. The upperspectrum in Fig. 11 was acquired by driving the 356 nmtransition of Ne2 with a fixe-i wavelength dye laser, while the

z tunable laser was scanned over the 352.7-353.5 nm region.WR(N) Q(N) The fixed wavelength laser was delayed in time with respect

5pir rig -a (0-O) to the scanning laser pulse by 100 ns. Consequently, if the Ne"'. '3p-3s fluorescence is monitored, two pulses are detected

and the peak intensity of the second pulse is proportional tothe relative a . (v = 0) population at the moment the

.- second pulse arrives. When the wavelength of the first laserpulse is resonant with an Ne, inter-Rydberg transition, the a

state population is momentarily depleted which is reflectedONE LASER in a temporary suppression of the Ne ( 3p) population pro-

.duced by the second laser pulse and, hence, the 3p- 3s spon-352.7 35..1 353.5WAVELENGTH (m) taneous emission intensity falls. The upper spectrum of Fig.11, therefore, is a recording of the suppression of the 3p - 3s

FIG. I I. Comparison of (bottom) laser excitation spectrum of 5pir I'l -a fluorescence that directly results from a reduction in the a 'Y(0-0) transition in the 352.7-353.5 nm region acquired with one (scan- population induced by the first laser pulse. As a result, in thening) dye laser; and (top) two laser spectrum over the same spectral region two laser spectrum, absorption increases downward. Theobtained by scanning one dye laser with the wavelength of the second fixed two laser spectrum is noisier than the one laser excitationat 356 nm (see the text). The position of the transition's Q branch is indicat-ed by the dashed vertical line. The spectrum acquired by the two laser tech- spectrum because of shot-to-shot amplitude jitter in the sec-nique (where absorption increases downwards) is somewhat noisier than ond Nd:YAG laser which is driving the fixed wavelengththe single laser excitation spectrum due to amplitude jitter in the Nd:YAG dye laser. Finally, one attractive aspect of this technique islaser pumping the fixed wavelength dye laser, that both the one and two laser spectra can be acquired si-

multaneously using two boxcar integrators.Figure 12 shows experimental results analogous to those

sorption by a molecular transition into a suppression of tran- of Fig. II for transitions involving the 6pir Hig state. Fromsient atomic fluorescence. Although further details of the Figs. 11 and 12, it is obvious that all of the prominent fea-experimental technique will be published elsewhere,'" sever- tures in the excitation spectrum (lower trace) are repro-al examples of the spectra obtained which are important for duced in the two laser experiments. Most important, how-our purposes here are presented in Figs. 11 and 12. ever, is the fact that the Q branch (denoted by dashed

vertical lines in both figures) is anomalously weak for transi-tions to npir H1 states with n > 5 in both experiments. Thisdemonstrates that differences in the stability of the H1 " and

7WO LASER 1 - states against predissociation are not the cause of theweakening of the Q branchlike features in ournpr -a 'I, spectra that is observed with increasing n.

This interpretation of the experimental data of Figs. I Iand 12 would be invalid if the photopumped states do not

R(N) predissociate effectively, but rather reradiate to the a stateWN with a spontaneous lifetime that is short compared to thez 6pir r.- a3 E.(O-O) P(N) delay time of 100 ns. If such were true, however, the dye laser

I t pumping process would produce strong molecular emission> which has not been observed by us nor reported in the litera-

lure.R(N) As mentioned earlier, it is the existence of Q branch

features in the transitions of Figs. 5-10 that leads us to be-lieve that the upper states have 11 character. These Q featuresare exactly where predicted if one assumes the molecular

ON_ _ __SER constants for the I ' (Table l11) and H - sublevels to be32C2 328.6 329.0 similar (i.e., small A doubling). In addition, the relative in-

WAVELENGTH( nm) tensities of the branches in Fig. 5 are what one would predict

FIG. 12. Spectra similar to those ofFig. II for the 6 pi n,-a transition of fora H-1 type transition between reasonably isolated case brNe,. The top spectrum was. again, acqured by the two laser technique states, while the relative intensities within the band in Fig. 6and the solid vertical lines show the correlation between prominent lines in are also close to the values predicted on the basis of the sameboth spectra model. The higher members of the series more closely resem-

J. Chem. Phys., Vol. 95, No. 6, 15 September 1991

3M86 Kane et a.: Spectroscopy of No.

TABLE IV. Line positions (cm- -) of the 5f( 1," ,'11,,) -a 'l: (1-0) While we are reasonably confident that the upper statestransitions in 0INez." for this Rydberg series are of npir 11, character, it is difficult

1to be as certain about the states' multiplicity. Two factorsSfu 'XI -at 1-0)b 5fir 'n, - a(1-.[) have prompted us to, at present, refrain from making a more

V R(N) P(N) R(N) P(N) definitive assignment. ( I ) a fourth. and much weaker series

lies in the same energy region as the npir series, but as yet hasI not yielded any rotational structure and, therefore, remains3 30 539.55 unassigned; and (2) there appears to be a P(I) line in the

7 30502.71 (30491.43) 30549.25 k3l 529.41) 4pm n,-a transition (see Fig. 5) which cannot exist in a9 30499.34 (30483.96) 30553.52 (3! 529.41) 'fI,2-3 transition. Studies of transitions associated withI 30495.97 (30475.62) 30 557.05 (31 529.41) higher n terms of npir series and careful examination of the13 (30491.43) 30467.500 30 559.99 (31 529.41) weak, unidentified series should remove this remaining am-15 30 487.7" 30458.91 30562.93 31 527.60017 (30483.96) (30 450.77) 30 564.92 31 525.89 biguity in the near future.19 30479.72 30566.45 31 523.2521 (30475.62) 30567.79 31 520.31 C. 7pa I, -5f 3123 (30473.25) 30568.52 31 517.24 . perturbation

25 (30469.19) 31513.71 A careful examination of the 30 390-30 590 cm re-27 30464.85 gion (shown in Fig. 7) reveals a weak perturbation occur-29 30461.83 bet 'i" (t--0) and 5 (v: 1)31 (30457.44) ring ween the 6p1, =) ' 1

33 30454.43 states near 30 465 cm -'. This perturbation is first evident in35 30451.51 the sudden drop in intensity of the 6pr 1ig -a(0-0) R37 30 41.58 branch at higher N. The observed spectral line positions for39 30445.57 the 5fa 31 +, 5fr 3 n. - a( 1-0) transitions and the 5fa '

Parentheses and an asensk denote a blended line and a shoulder feature, 5 n/ , 3 -i (v = I ) level energies levels appear in Tables IV andrespectively. Note that although two decimal places are quoted for these V, respectively. Plots of the rotational level energies E(N)tine positions, in most cases the uncertainties in the last quoted digit are for both the 5lons (u = I) and 6pr rli (v = 0) levels in:t 5-10.AloftheJP(N) lines for N> 17 are obscured by much stronger transitbons. Fig. 13 show the convergence of the E(N) values for both

states and the crossing of the two series near N = 26. Theinset gives the same data, but the ordinate represents thedifference between the level energies of the two interactingstates as a function ofN. Examination of the data shows that

ble AA = 0, A > 0 transitions, but such an observation is of the interaction results in the N = 26 levels being displacedlittle help in explaining our observations because the one from their unperturbed positions by - 1 cm - with muchstate whose character does not change throughout the ob- smaller shifts for the N = 28 levels. Thus, the perturbationserved series of transitions is the lower state a 'Z' for which involves both transition intensities and energy although theA = 0. effects are relatively weak.

V. SUMMARY AND CONCLUSIONS

Several Rydberg series of the neon dimer have been

TABLE V. Level (term) energies (cm -') of the v = 1, 5fo ' and studied at a resolution of - I GHz by laser excitation spec-

5fr fIl, states of 'Ne2,. troscopy in the afterglow of a pulsed discharge. By photoex-citing th, metastable a 'x species with a time-delayed dye

S5f, 3y 5fr 3l laser pulse, Rydberg states of the 20Ne, and 22Ne, moleculeswere examined in detail. Six members of the npr fi1 -a(0-

6 30523.4 30556.96 0) series (4<n<9) were rotationally resolved and character-

8 30534.2 30 580.74 ized. An anomalous decrease in the intensity of the Q10 30550.0 30604 15 branches for the higher n members of the series has been12 30569.7 30631.28 discussed and a weak perturbation between the 6pr lI14 30593.8 30662.3716 30622.9 30697,95 (V = 0) and 5fo 9 (u = 1) states has been identified. In18 30 656.0 30 73697 passing, it should be noted that it is the 5pir 1lg -a photoab-20 30693.4 30780.12 sorption band of Ne2 that interferes with the extraction of22 30735.4 30827.60 energy from the XeF laser at 353 nm. Apparently, it was this24 30 783.6 30 888.9626 30834.6 band that motivated Chang and Champagne t4 to search for

28 30889.6 the origin of absorption they had observed in electron beam-30 30950.5 pumped Ne/Xe/NF , gas mixtures.32 31 014.3 " While at present we are unable to unambiguously assign34 31 083.9

36 31 158.1 the multiplicity of the npir [n, Rydberg states presented,, , _ _ , _,_here, it is expected that subsequent investigation of the com-

Basedon N-0. v-Oofthea '1: state plex region lying between the 9pr 1 - a(0-O) band and

J Chem. Phys.. Vol. 95, No. 6, 15 Septemnber 1991

Kane t al.: Spectroscopy of Ne 2 3887

Z for both the 6pirfl1 (v=O) and

31150 50l- . . . . . . . ,

"E (v 5f"atesa(afu)c-Euj 31325-'

" ) .tion of N. The larger graph shows_J_____________,_____________________.. the absolute values for the term en-

~ergies, whereas the ordinate for theXJ.,I/ -inset gives the deviations of the en-

J 30er0ies for the two states from their

_J average value E(N)AV.< 303650Z

P-- 305250 n," 6p' r ", (v= )

3070010 20 .0 40

N

the first ionization limit will resolve the outstanding ques- 'i. Herzerg, Annu. Rev. Phys. Chem. 38. 27 (1987).tions that remain. Since the spectral region beyond - 700 eK. P. Killeen and 3. G. Eden. 3. Chem. Phys. 83, 6209 (1985).

nm contains transitions from a 'Z,7 to the first member of ' K. P. Killeen and 1. G. Eden. 3. Chem. Phys. 84, 6048 (1986).SM. N. Ediger and J. G. Eden. Chem. Phys. Left. 126, 158 (1986).all of the n, Rydberg series with 1<4, we also plan to inves- M. N. Ediger and . G. Eden, . Chie sf. Phys. w s7s7 ( 9o6t).

tigate transitions in the near infrared. "D. J. Kane, C. J. Zietkiewicz, and J1. G. Eden, Phys. Rev. A 39. 4906

(1989).' S. B. Kim, D. J. Kane. J. G. Eden, and M. L. Ginter, J. Chem. Phys. 94,

145 (1991).ACKNOWLEDGMENTS "Elsewhere in the literature, the lowest energy bound state of Nez is occa-

sionally denoted ,4 ,2 :. We adopt X as its designation for consistencyThe authors gratefully acknowledge the expert techni- with international conventions tor labeling the ground state of a molecule.cal assistance of K. Kuehl and J. Sexton. This work was 5s. lwata, Chem. Phys. 37, 251 (!979).supported by the National Science Foundation under Grant ,D. J. Kane. Ph.D. thesis. University of Illinois, Urbana, 1989.NO.. CBT 87-15681 (W. Groeshandler), CTS 89.!5795 (K. 'tSee eg".B Edlen,'in Handbuch Der Physik,'edited byS.'Flu~gge (Spring-c er. Berlin. 1964), Vol. 27, pp. 80-220.

Gilln),andPHY 8-1722(S. mit) ad th Ai Foce 25 $.. Kim. J. G. Eden, and M. L Ginter (unpublished).Office of Scientific Research (H. Schlossberg) under Grant ,3s. B. Kim, D. 3. Kane, and J. G. Eden (unpublished).No.. 89-0038 and 88-0018. "R. S. F. Chang and L. F. Champagne, Appl. Phys. Lett. 36, 879 (1980).

J CIhem Phys.. Vol 95. No 8. 15 September 1991

Detection of Trace Concentrations of Column III and VHydrides by Laser-Induced Breakdown Spectroscopy

E. A. P. CHENG,* R. D. FRASER, and J. G. EDENtveritt Laboratory, University of Illinois, Urbana, Illinois 61801

The meitrement of tram concentratiom (1-400 ppa) of polatomic Two further advantages of this technique that havemeecalar Impurities In helium was inveutdgted by laser-induced break- been pointed out previously" are that LIBS is nonin-dewn spectrescopy (LIES) For tn Columa Mii and V hydrides such a vasive and does not require that the laser be tuned to aDi. PH, and AsH, the detectability amifs with the present a eratms resonance of the species under study. The latter is re-were determined to be 1, 3, and -1 I m respectively. It is expected that flected in the fact that LIB spect - are relatively insen-mesrtigto synchroousdetectloa techalquswililmprove the sensitvity sitive to the wavelength of the ex. -ation laser, thus al-oef LIBS for this application by as much as an order of magnitude. lowing one to simultaneous' amine the relativeIndex Heading&: Fluorescem Lasem Nd:YAG; UV.visible speroa- loig neosmuteus, aneteraivInr, Lar-ils. bremmkdoan specroNscopy. concentrations of several impurities that may be present

in the sample. Cremers8 has clearly demonstrated the

utility of this feature of LIBS in recent experiments in

INTRODUCTION which minor constituents in steel were observed.Loree, Radziemski, and co-workers" have demon-

The application of laser excitation and ionization pro- strated the ability of LIBS to measure small concentra-ceases to the detection of impurities in gases and solids tions of impurities in a hostile environment such as theis well developed. In flames, for example, Grant et al. I corrosive effluent from a coal gasifier plant. Early ex-have observed 25 parts per million (ppm) of NO by mul- periments involved the detection of 1-3% Na or K intiphoton ionization (MPI) with a tunable dye laser. Also, the product stream. In subsequent time-resolved studies,trace concentrations of a variety of elements in both 690 ppm of phosphorus (from diisopropylmethyl phos-solids and liquids (2 10-8 at. % Na and 5 10-1 at. % Al, phonate, DIMP), 8 ppm of chlorine, and 38 ppm of flu-for example, in Ge crystals and seawater, respectively) orine were observed in air.9

have been detected by Bekov and Letokhov,2 again uti- In this paper, the preliminary results of several ex-lizing multi-step, resonantly enhanced laser photoioni- periments are described in which the detection of sub-zation. Akilov et al.3 have invoked similar techniques to 100-ppm concentrations of several Column III and Vobserve aluminum concentrations as low as 2 10- 1 at. % hydride polyatomic molecules in the gas phase by LIBSin ostensibly "pure" germanium. was explored. These studies were motivated by the grow-

An alternative, and simpler, approach to impurity de- ing demands being placed on the purity of gases of in-tection is laser-induced breakdown spectroscopy (LIBS),' dustrial importance and the concomitant requirementwhich involves optical breakdown in a gaseous sample for (preferably) noninvasive and rapid diagnostic tech-or at a surface that is induced by a laser pulse having a niques for measuring trace concentrations of contami-peak intensity typically in the range of 107-1010 W cm - 2. nants in such gases. As discussed earlier, the inherentSuch high optical fields result in the dissociation of mo- characteristics of the method allow LIBS to lend itselflecular species present into their atomic constituents. well to such an application.Consequently, LIB spectra are largely free of molecular The goal of these experiments was the identificationstructure, and the small volume sampled by the tech- of one or more atomic emission lines associated with anique makes it attractive for the micro- (and remote, if given molecular impurity that are reliable indicatorsnecessary) analysis of toxic or corrosive substances. Fur- ("fingerprints") 4 of the contaminant's relative concen-thermore, as noted by Cremers and Radziemski,9 LIBS tration, particularly at the ppm level. Our initial effortsavoids the need for electrodes, which are common to have centered on monitoring the concentrations of PH,spark spectrometry. AsH, and BH, diluted in He. Although the use of these

three -,ases in the semiconductor industry is widespread,the toxicity of each has impeded the development of

Received I December 1990; revision received 5 March 1991. diagnostic procedures. Nevertheless, the impact of sourcePresent address: Lightweve Electronics Corp., 897-5A Independence gas purity on electronic device quality, for example, isAvenue. Mountain View. CA 9404.3. illustrated by the observation' 0 -l' of the presence of Ge

t Author to whom correspondence should be sent. as a donor in GaAs epitaxial films grown by metal organic

Volume 45, Number 6, 1991 ,0,.-7.0,UNI -094952.0o/o APPLIED SPECTROSCOPY 949r 1991 So(iety fnr Apphied Spectrecopy

TO VACUUM ANO 30GAS HANDUNGS;YSTrMS 361 ppm PH,/He

EH41

SAMPLE R EXHAUST P 024 4CHAMBER 20 P 60- 103.2

- INd:YAG "CLASER (10 HZ)

10

E5

0.5" SE TO R P Ls[I 1/7s COMPUTER

600 602 604 606 60a 610H Wavelength (nm)

MULETCTONEI ft. 2. Low-resolution view of the 600-610 nm spectral region for a(O .,TEC OS(O$MA) COMA

.CONLRLER mixture of 361 ppm of PH, in He (top) and He only (bottom). Theupper and lower traces have been intentionally offset for clarity.

FiG. 1. Schematic diagram of the experimental arrangement. cussed here are measured in air. All spectral data werestored and processed by a DEC LSI 11/73 computer.

Although the gas handling system was equipped forchemical vapor deposition (MOCVD). It now appears either flowing or static fill experiments, no detectablethat the Ge impurity is introduced to the reactor in the differences were observed between the LIB spectra forform o, germane (GeH,) as a contaminant in AsH 3. the two cases. Consequeutly, all of the measurements

with these toxic gases were made under static fill con-

EXPERIMENTAL ditions and, after each experiment, the gas was exhausted

The experimental apparatus, shown schematically in through a KMnO, bubbler.

Fig. 1, is described in detail elsewhere 12 and will be briefly Known levels of the impurity gas under study were

reviewed here. Optical field intensities up to - 10" W prepared by diluting calibrated samples (provided by the

cm -2 were produced in a vacuum chamber by focusing Liquid Air Corp.) with high-purity He (UHP; >99.999%).e red proce X -a2 vu m fIn this manner, it was straightforward to reliably mixthe second harmonic beam (X = 532 nim) from a pulsed sapecoting<1pm fthdsrd"ota -

Nd:YAG laser [10 Hz, 8 ns full width at half-maximum samples containing <1 ppm of the desired "contami-(FWHM) pulses] with a 10-cm-focal-length (FL) lens. nant." Fresh gas mixtures were prepared for each data

run, which simply consisted of firing a fixed number ofThe chamber was evacuated by a turbomolecuar p ip laser shots (typically 700-800) while monitoring a - 150-to a base pressure of 5-10-8 Torr, and the entire gas nm-wide spectral region. Preparation for the next run

handling/vacuum chamber system was constructed of consisted of evacuating the cell, backflling with He, and

stainless steel Visible and ultraviolet radiation from the evacutin oefin te befor aditing a fe as

laser-induced spark was collected at 90 to the pump laser evacuating one final time before admitting a fresh gas

beam and dispersed by a simple lens teleacope/spectrom- mixture.

eter combination. A 3.1-cm-diameter, 7-cm-FL lens col- RSULTS AND DISCUSSIONlimated the radiation from the plasma, and a second lens(2.5-cm diameter, 10-cm FL) focused the beam into a Phosphine. Experiments were carried out with mixtures0.25-m spectrograph having an / number of 3.8. of 3, 10, 50, and 180 ppm of PH3 in He that were prepared

Spectra were acquired by mounting one of two inten- by diluting a calibrated 361 ppm in He mixture. Scanssified diode arrays at the exit plane of the spectrograph. of the 300-900 nm region revealed several phosphorusThe first, composed of 512 diodes, had a sensitivity of and hydrogen lines, but most (such as the 391.4-nm and-100 photons/count and a useful spectral range of - 350- 434.0-nm transitions of P and H, respectively) vanished830 nm. Equipped with this optical multichannel ana- for PH, concentrations below -90 ppm.lyzer (OSMA) detector, 25-sm entrance slits, and a 300- In the 600-660 nm spectral region, five lines-fourlines/mm diffraction grating, the entire detection system arising from the phosphorus singly charged ion and onehad a first-order spectral resolution of -0.28 nm/diode. from hydrogen (P*: 602.4, 603.4, 604.3, and 605.5 nm; H:When necessary, the resolution could be improved by a 656.3 nm)-were readily detected. A low-resolution viewfactor of two by interchanging gratings or operating the of the 600-610 nm interval is given in Fig. 2 for both 361spectrograph in second order. For the arsine and dibor- ppm PH,/He and He-only static fills. The four lines men-ane experiments, it was desirable to monitor atomic lines tioned earlier, of which the 604.3-nm transition is thelying below 300 nm, which required a second intensified strongest, are clearly noticeable. Figures 3 and 4 showarray having adequate response over the 200-300 nm the dependence of the P 604.3-nm and H 656.3-nm tran-region. Unless stated otherwise, all wavelengths dis- sition peak intensities, respectively, on th t PH3 concen-

9O Volume 45, Number 6, 1991

1000 So,

PH,/HePH/ 60€.3nm 10 ppm AsH 3/HeP': 604.3 nm 3H

10o0

-

C 100

"to t.. . oo ,000

: '00

P10 o- ( 250 275 300 .. . " 50 375

FIG. 3. Dependence of the P" 604.3-nm line intensity on the PH, Wavelength (mm)

concentration for impurity levels ranging from -3 to -400 ppm. Fi. 5. Ultraviolet LIB emission spectrum of 10-ppm arsine in H-e.The strong As lines lying between 225 and -300 nm are indicated. Thelower spectral trace is that for pure He.

tration. Note that, for PH, concentrations above 20-30ppm, the variation of the time-integrated 604.3-nm tran- 290 nm. Figure 5 displays the spectra that were recordedsition intensity (cf. Fig. 3) is logarithmic in PH3 number between -212 and 378 nm for both 10 ppm AsH3 in Hedensity. Below 10 ppm, the signal falls less slowly, but and He itself. Five lines of neutral As (228.8, 235.0, 278.0,the underlying blackbody continuum complicated mea- and 286.0 nm) are quite clear and, as shown in Fig. 6 forsurements at lower PH3 concentrations. Consequently, the 228.8-nm transition, monitoring the intensity of anywith the present apparatus, the minimum PH3 level de- of the five limited one to detecting AsH3 concentrationstectable by monitoring the phosphorus orange line was of roughly 1 ppm or higher. Again, it is expected thatlimited to -3 ppm. However, we estimate that, by uti- the incorporation of synchronous detection techniqueslizing synchronous detection techniques, one could im- or the use of a fiber-optics bundle introduced into theprove the PH, detectability limit to below 500 ppb. Fi- vacuum chamber to improve the fluorescence collectionnally, data from three separate experimental trials, given efficiency would lower the detectability limit by as muchin Fig. 3, illustrate the reproducibility of the measure- as an order of magnitude.ments. Diborane. Results similar to those obtained with AsH,

Arsine. A known mixture of 100 ppm AsH, in helium and PH, were also realized with BH,. Two lines of atomicwas used to prepare all diluted samples. Scans of the boron, 434.5 and 336.0 nm, were readily observed, andvisible and ultraviolet revealed the most reliable atomic the latter was reproducibly detected down to BAH, con-transitions to be those of arsenic lying between 220 and centrations as low as - 1 ppm (cf. Fig. 7).

PH3/He AsH 3/He

H 656 3 nm As: 228.8 nm

oo

I I

100

g '0 2

0

0 1 00 200 300 400 0.1 1 0 100

PMa Concentration (pr) ASln Concentrtion (rpm)

FIG. 4. Date similar to those of Fig. 3 except that the H line at 666.3 FIG. 6. variation of the peak intensity of the 228.8-nm line of arsenic

nm is bening monitored. with the arsine concentration.

APPLIED SPECTROSCOPY 951

10- here. Nevertheless, these preliminary results demon-

B H6/Hestrate that LIBS is an attractive technique for makingB2H6 H remote measurements of sub-100-ppm concentrations ofB. 336.0 nm /gases which are of critical importance to the semicon-

ductor industry. In general, the rapid acquisition andanalysis of spectral data that are characteristic of LIBS

./ make it amenable to the computerized, in situ analysis* / of small volumes of toxic and/or corrosive gases.

0.17 ACKNOWLEDGMENTS

The authors gratefully acknowledge helpful discussions with andassistance from J. Kim, C. Abele, K. King, J. Schloss, D. Shannon, P.Griffis, D. Terven and K. Voyles. The work was supported by the AirForce Office of Scientific Research (H. Schlossberg) under Contract

0.01 F49620-85-C.0141 and the Liquid Air Corporation (F. Mermoud and0.1 lo 1000 J.-P. Barbier).

92, Concentration (ppm)

Via.7. Data similar to those of Figs. 3, 4, and 6 for the 338.0-nm line 1. E. R. Grant, B. H. Rockney, and T. A. Cool, Chem. Phys. Lett. 87,of boron The B 434.5-nm transition intensity was found to have a 141 (1982).depeondence on [BHe that was nearly identical to that shown here. 2. G. 1. Bekov and V. S. Letokhov, App!. Phys. B 30, 161 (1983).

3. R. Akilov, G. 1. Bokov, V.5S. Letokhov, G. 1. Maksiznov. V. 1. Mishin.V. N. Radaev, and V. N. Shiahov, Kvan. Elektron. (Moscow) 9.

SUMMAR"Y AND CONCLUSIONS 1859 (1982) [Sov. J. Quant. Electron. 12, 1201 (1982)].4T. R. Loree and L. J. Radziemski, Plasma Chem. and Plas Proc.

The capability of detecting small concentrations of 1,7 (1981).seletedpolytomc: oleclesin H byLIB hasbee 5.L. J. Radziemski and T. R. Loree, Plasma Chem. and Plasma Proc.seletedpolatoic oleule inHe y LBS as een 1,281 (1981).

demonstrated. In accord with previous studies, the sen- 6. L J. Radzieinski, T. R. Loree, D. A. Cremers, and N. M. Hoffman,sitivity of the technique is system dependent, but our Anal. Chem. 55% 1246 (1983).studies yielded a - 1 -ppm concentration "floor" for AsH3, 7. D. A. Cremers and L J. Radzemski, in Laser Spectroscopy andPH,, and BA1{. Consequently, the detectability limits Its Applications, L. J. Radzicinaki, R. W. Soiarz, and J. A. Paisner.

herearecomprabe t thoe dterinedpre 8.Eds. (Marcel Dekker, New York, 1987), pp. 351-415.reported heeaecmaal otoedtrie r- D. A. Cremers, Appi. Spectroec. 41, 572 (1987).viouuly for atomic species such as Cl and F. As noted 9. D. A. Cremers and L. J. Raduiemaki Anal Chem. 55, 1252 (1983).earlier, it is estimated that the implementation of syn- 10. G. K Stillman, C. M. Wolfe, and J. 0. Dimmock, in Semiconductorschronous detection techniques would reduce these limits and Semiinetals. R. K. Willardson and A. C. Beer, Eds. (Academic

b asmuch as an order of magnitude. Of course, the 1Press, New York 1977), Vol. 12, pp. 169-290.1.S. S. Bose, B. Lee, M. H. Kim, and G. E. Stillman, Appl. Phys.

homogeneity of neat samples available in an industrial Lett. 51, 937 (1987).envirlonment might adversely affect the limits reported 12. E. A. P. Cheng, M.S. thesis, University of Ilinois, Urbana (1987).

952 Volume 45, Number 6. 1991

Detection of Trace Concentrations of Column III and VHydrides by Laser-Induced Breakdown Spectroscopy

E. A. P. CHENG,* R. D. FRASER, and J. G. EDENtEWreitt Laboratory. Uniweruity of llinois, Urbana, llinois 61801

TM mmuewment of tram coacesiradow (1-400 ppm) of pelyatomic Two further advantages of this technique that havemIecular impuides in helium was Inwusdpted by laer-Induced break- been pointed out previously' 4 are that LIBS is nonin-dewn sp cresew (L=). For the Column I and V hydrides sab as vasive and does not require that the laser be tuned to aBL PH, and AsH, the delectability limits with the present appmft resonance of the species under study. The latter is re-wer detemimni to be 1, 3, ad -1 ppm, rmpectively. It Is exp ted that flected in the fact that LIEB spectra are relatively insen-morus Utosyhlroed edetecloe techakiues will improve thesemitivity sitive to the wavelength of the excitation laser, thus al-of Li for this applatlon by a mu = g orer of maitu lowing one to simultaneously examine the relative

Index Headings: FLaorescece asers, Nd:YAG; Uvibl e spectres- concentrations of several impurities that may be presentcopy Lasr-lduced brealkdown spectroscopy. in the sample. Cremerss has clearly demonstrated the

utility of this feature of LIBS in recent experiments inINTRODUCTION which minor constituents in steel were observed.

Loree, Radziemski, and co-workers" have demon-The application of laser excitation and ionization pro- strated the ability of LIBS to measure small concentra-

cesses to the detection of impurities in gases and solids tions of impurities in a hostile environment such as theis well developed. In flames, for example, Grant et al.I corrosive effluent from a coal gasifier plant. Early ex-have observed 25 parts per million (ppm) of NO by mul- perimenrs involved the detection of 1-3% Na or K intiphoton ionization (MPI) with a tunable dye laser. Also, the product stream. In subsequent time-resolved studies,trace concentrations of a variety of elements in both 690 ppm of phosphorus (from diisopropylmethyl phos-solids and liquids (2.10-8 at. % Na and 5-10 - 7 at. % Al, phonate, DIMP), 8 ppm of chlorine, and 38 ppm of flu-for example, in Ge crystals and seawater, respectively) orine were observed in air.9

have been detected by Bekov and Letokhov,2 again uti- In this paper, the preliminary results of several ex-lizing multi-step, resonantly enhanced laser photoioni- periments are described in which the detection. of sub-zation. Akilov et al.3 have invoked similar techniques to 100-ppm concentrations of several Column III and Vobserve aluminum concentrations as low as 2 .10- at. % hydride polyatomic molecules in the gas phase by LIBSin ostensibly "pure" germanium, was explored. These studies were motivated by the grow-

An alternative, and simpler, approach to impurity de- ing demands being placed on the purity of gases of in-tection is laser-induced breakdown spectroscopy (LIBS)," dustrial importance and the concomitant requirementwhich involves optical breakdown in a gaseous sample for (preferably) noninvasive and rapid diagnostic tech-or at a surface that is induced by a laser pulse having a niques for measuring trace concentrations of contami-peak intensity typically in the range of 10 7-10 10 W cm - 2. nants in such gases. As discussed earlier, the inherentSuch high optical fields result in the dissociation of mo- characteristics of the method allow LIBS to lend itselflecular species present into their atomic constituents. well to such an application.Consequently, LIM spectra are largely free of molecular The goal of these experiments was the identificationstructure, and the small volume sampled by the tech- of one or more atomic emission lines associated with anique makes it attractive for the micro- (and remote, if given molecular impurity that are reliable indicatorsnecessary) analysis of toxic or corrosive substances. Fur- ("fingerprints")' of the contaminant's relative concen-thermore, as noted by Cremers and Radziemski,9 LIBS tration, particularly at the ppm level. Our initial effortsavoids the need for electrodes, which are common to have centered on monitoring the concentrations of PH.,spark spectrometry. AsH 3, and B2H6 diluted in He. Although the use of these

three gases in the semiconductor industry is widespread,the toxicity of each has impeded the development of

Received I December 1990; revision received 5 March 1991. diagnostic procedures. Nevertheless, the impact of source* Present address: Lightwave Electronics Corp., 897-5A Independence gas purity on electronic device quality, for example, is

Avenue. Mountain View, CA 94043. illustrated by the observation"'0 1 of the presence of Get Author to whom correspondence should be sent. as a donor in GaAs epitaxial films grown by metal organic

Volume 45, Number 6, 1991 ooo.7o0MI4z5-,949MS2oon APPLIED SPECTROSCOPY 9494 1991 S ciety for Applied Spectroscopy

TO VACUUM AND 30GAS HANDLINGSYSTEMS 361 ppm PH3/He

tKn04

EXHAUST 25-4

S AMdPL ER P: 602.4 -

CHAMBER 20 1603.4

LASER (10 H2)15 .

;i

P, 0, DT

6 500 602 604 606 608 610

MULTtCHANN. Wovelenqth (nm)

UL"TICTOR FIG. 2. Low-resolution view of the 600-610 nm spectral region for aL DETECTOR OSA(OS A) O AC0NTROLLER mixture of 361 ppm of PH3 in He (top) and He only (bottom). The

upper and lower traces have been intentionally offset for clarity.

FiG. 1. Schematic diagram of the experimental arrangement. cussed here are measured in air. All spectral data werestored and processed by a DEC LSI 11/73 computer.

Although the gas handling system was equipped forchemical vapor deposition (MOCVD). It now appears either flowing or static fill experiments, no detectablethat the Ge impurity is introduced to the reactor in the differences were observed between the LIB spectra forform of germane (GeH 4) as a contaminant in AsH3 . the two cases. Consequently, all of the measurements

with these toxic gases were made under static fill con-EXPERIMENTAL ditions and, after each experiment, the gas was exhausted

The experimental apparatus, chown schematically in through a KMnO, bubbler.

Fig. 1, is described in detail elsewhere 12 and will be briefly Known levels of the impurity gas under study werereviewed here. Optical field intensities up to -1 w prepared by diluting calibrated samples (provided by thecm-1 were produced in a vacuum chamber by focusing Liquid Air Corp.) with high-purity He (U-P; >99.999%).

the second harmonic beam (X = 532 nm) from a pulsed In this manner, it was straightforward to reliably mix

Nd:YAG laser [10 Hz, 8 ns full width at half-maximum samples containing < I ppm of the desired "contami-(FWHM) pulses] with a 10-cm-focal-length (FL) lens. nant." Fresh gas mixtures were prepared for each data

run, which simply consisted of firing a fixed number ofThe chamber was evacuated by a turbomolecular pump laser shots (typically 700-800) while monitoring a - 150-to a base pressure of 5.10-' Torn, and the entire gas nm-wide spectral region. Preparation for the next runhandling/vacuum chamber system was constructed of mwdspcrlego.Ppatinfrheexruhandingvacum hamer yste wa costrcte of consisted of evacuating the cell, backfilling with He, and

stainless steel. Visible and ultraviolet radiation from the evacutin oeuin te cefor aditing a e as

laser-induced spark was collected at 90 to the pump laser mixture.

beam and dispersed by a simple lens telescope/spectrom-

eter combination. A 3.1-cm-diameter, 7-cm-FL lens col- RESULTS AND DISCUSSIONlimated the radiation from the plasma, and a second lens(2.5-cm diameter, 10-cm FL) focused the beam into a Phosphine.Experiments were carried out with mixtures0.25-m spectrograph having an f number of 3.8. of 3, 10, 50, and 180 ppm of PH, in He that were prepared

Spectra were acquired by mounting one of two inten- by diluting a calibrated 361 ppm in He mixture. Scanssifled diode arrays at the exit plane of the spectrograph. of the 300-900 nm region revealed several phosphorusThe first, composed of 512 diodes, had a sensitivity of and hydrogen lines, but most (such as the 391.4-nm and- 100 photons/count and a useful spectral range of - 350- 434.0-nm transitions of P and H, respectively) vanished830 nm. Equipped with this optical multichannel ana- for PH3 concentrations below -90 ppm.lyzer (OSMA) detector, 25-ism entrance slits, and a 300- In the 600-660 rm spectral region, five lines-fourlines/mm diffraction grating, the entire detection system arising from the phosphorus singly charged ion and onehad a first-order spectral resolution of -0.28 nm/diode, from hydrogen (P : 602.4, 603.4, 604.3, and 605.5 nm; H:When necessary, the resolution could be improved by a 656.3 nm)-were readily detected. A low-resolution viewfactor of two by interchanging gratings or operating the of the 600-610 nm interval is given in Fig. 2 for both 361spectrograph in second order. For the arsine and dibor- ppm PH.He and He-only static fills. The four lines men-ane experiments, it was desirable to monitor atomic lines tioned earlier, of which the 604.3-nm transition is thelying below 300 nm, which required a second intensified strongest, are clearly noticeable. Figures 3 and 4 showarray having adequate response over the 200-300 nm the dependence of the P 604.3-nm and H 656.3-nm tran-region. Unless stated otherwise, all wavelengths dis- sition peak intensities, respectively, on the PH, concen-

950 Volume 45, Number 6, 1991

1000

PH 3/HeP: 50.3 nm10 ppm AsH 3/HeP*: 604.3 nm

0

10000

C

ag ac50 As s

IA,

10 100 1000PH, Consceiirstion (ppm) 0 225 250 275 300 325 350 375

FIG. 3. Dependence of the P* 604.3-nm lUne intensity on the PH, Wavelength (nm)

concentration for impurity levels ranging from -3 to -400 ppm. FIG. 5. Ultraviolet LIB emission spectrum of 10-ppm arsine in He.The strong As lines lying between 225 and -300 nm are indicated. Thelower spectral trace is that for pure He.

tration. Note that, for PH 3 concentrations above 20-30

ppm, the variation of the time-integrated 604.3-nm tran- 290 nm. Figure 5 displays the spectra that were recordedsition intensity (cf. Fig. 3) is logarithmic in PH, number between -212 and 378 nm for both 10 ppm AsH3 in Hedensity. Below 10 ppm, the signal falls less slowly, but and He itself. Five lines of neutral As (228.8, 235.0, 278.0,the underlying blackbody continuum complicated mea- and 286.0 nm) are quite clear and, as shown in Fig. 6 forsurements at lower PH concentrations. Consequently, the 228.8-nm transition, monitoring the intensity of anywith the present apparatus, the minimum PH 3 level de- of the five limited one to detecting AsH3 concentrationstectable by monitoring the phosphorus orange line was of roughly 1 ppm or higher. Again, it is expected thatlimited to -3 ppm. However, we estimate that, by uti- the incorporation of synchronous detection techniqueslizing synchronous detection techniques, one could im- or the use of a fiber-optics bundle introduced into theprove the PH detectability limit to below 500 ppb. Fi- vacuum chamber to improve the fluorescence collectionnally, data from three separate experimental trials, given efficiency would lower the detectability limit by as muchin Fig. 3, illustrate the reproducibility of the measure- as an order of magnitude.ments. Diborane. Results similar to those obtained with AsH 3

Araine. A known mixture of 100 ppm AsH in helium and PH 3 were also realized with B ,i. Two lines of atomicwas used to prepare all diluted samples. Scans of the boron, 434.5 and 336.0 nm, were readily observed, andvisible and ultraviolet revealed the most reliable atomic the latter was reproducibly detected down to B 2H, con-transitions to be those of arsenic lying between 220 and centrations as low as - 1 ppm (cf. Fig. 7).

100 ?00-

PH3/He AsH 3/HeH 656,3 nm As: 228.8 nm C3

10l

" 0o3 =

0 100 200 300 400 01 1 10 1 00PM, Concentrton (ppm) ASH3 Concentrotion (ppm)

FIo. 4. Data similar to those of Fig. 3 except that the H line at 656.3 FIG. 6. Variation of the peak intensity of the 228.8-nm line of arsenicrm is being monitored, with the arsine concentration.

APPLIED SPECTROSCOPY 951

IC /here. Nevertheless, these preliminary results demon-strate that LIBS is an attractive technique for making

8 2H/He remote measurements of sub-100-ppm concentrations ofB: 336.0 nm gases which are of critical importance to the semicon-

ductor industry. In general, the rapid acquisition andanalysis of spectral data that are characteristic of LIBSmake it amenable to the computerized, in situ analysisof small volumes of toxic and/or corrosive gases.

a i ACKNOWLEDGMENTS

The authors gratefully acknowledge helpful discussions with andassistance from J. Kim, C. Abele, K. King, J. Schloss, D. Shannon, P.Griffis, D. Terven, and K. Voyles. The work was supported by the AirForce Office of Scientific Research (H. Schlossberg) under Contract

0.0 F49620-85-C.0141 and the Liquid Air Corporation (F. Mermoud and0" .1 .. . 1 .. I0 10 " i0 J.-P. Barbier).

82H% Concentration (ppm)

io- 7. Data si-ila to those of Fig. 3,4, and 6 for the 336.0-nm line 1. &. R Grant, B. H. Rockney, and T. A. Cool, Chem. Phys. Lett. 87,of boron. The B 434.5-nm transition intensity was found to have a 141 (1982).dependence on [BJ] that was nearly identical to that shown here. 2. G. I. Bekov and V. S. Letokhov, AppL Phys. B 30, 161 (1983).

3. R. Akilov, G. I. Bekov, V. S. Letokhov, G. 1. Maksimov, V. .Mishin,V. N. Radaev, and V. N. Shishov, Kvan. Elektron. (Moscow) 9,

SUMMARY AND CONCLUSIONS 1859 (1982) [Soy. J. Quant. Electron. 12, 1201 (1982)].4. T. R. Loree and L. J. Radziemski, Plasma Chem. and Plasma Proc.

The capability of detecting small concentrations of 1, 271 (1981).

selected polyatomic molecules in He by LIBS has been 5. L J. Radziemski and T. R. Loree, Plasma Chem. and Plasma Proc.1, 281 (1981).

demonstrated. In accord with previous studies, the sen- 6. L. J. Radziemski, T. K. Loree, D. A. Cremers, and N. M. Hoffman,sitivity of the technique is system dependent, but our Anal. Chem. 55, 1246 (1983).studies yielded a - 1-ppm concentration "floor" for AsH3, 7. D. A. Cramers and L. J. Radziemnki, in Laser Spectroscopy andPH, and BHA. Consequently, the detectability limits Its Applications, L. J. Radziemaki, R. W. Solarz, and J. A. Paisner,

Eds. (Marcel Dekker, New York, 1987), pp. 351-415.reported here are comparable to thoe determined pre- 8. D. A. Cremers, AppL. Spectrosc. 41, 572 (1987).viously for atomic species such as Cl and F. As noted 9. D. A. Cremers and L J. Raduiemaki, Anal Chem. 55, 1252 (1983).earlier, it is estimated that the implementation of syn- 10. G.E. Stillman, C. M. Wolfe, and J. 0. Dimmock, in Semiconductorschronous detection techniques would reduce these limits and Semimetals, 1. K. Willardsn and A. C. Beer, Eds. (Academic

by as much as an order of magnitude. Of course, the ~Press, New York, 1977), Vol. 12, pp. 169-290.11. S. S. Bose, B. Lee, M. H. Kim, and G. K Stiliman, AppL Phys.

homogeneity of neat samples available in an industrial Lott. 51, 937 (1987).environment might adversely affect the limits reported 12. E. A. P. Cheng, MS. thesis, University of Illinois, Urbana (1987).

952 Volume 45, Number 6, 1991

Ultraviolet laser-assisted metalorganic chemical vapor deposition of GaAsP. K. York, J. G. Eden, J. J. Coleman, G. E. Fernandez, and K. J. BeeminkMaterials Research Laboratory and Everitt Laboratory, University of Illinois, 1406 West Green Street,Urbana, Illinois 61801

(Received 20 April 1989; accepted for publication 30 July 1989)

The growth of GaAs irradiated with ultraviolet laser light in a metalorganic chemical vapordeposition reactor has been investigated. Growth rate enhancements of up to 15% wereobserved at 450 °C by illuminating the substrate with no more than 13 mJ/cm 2 of KrF laser(248 nm) radiation. For 5-eV photons, arsine is virtually transparent, while thetrimethylgallium (TMG) photoabsorption cross section is approximately 10 1 cm2. Dataacquired with and without the optical beam impinging on the substrate are well described bythe Langmuir-Hinshelwood model, and the results point to photodissociation of adsorbedTMG as the origin of the growth rate enhancement. Nearly identical experiments carried outat 351 nm (XeF) corroborate this conclusion since no measurable increase in growth rate wasobserved at this wavelength, where both arsine and TMG photoabsorption is negligible.Conversely, significant improvement in surface morphology for samples grown below 700 °C isobserved with ultraviolet laser irradiation of the substrate at each of the wavelengthsinvestigated ( 193, 248, and 351 nm). Smooth and specular surfaces are obtained with substratetemperatures as low as 550 "C and at fluences well below those which induce a significant risein the surface temperature.

I. INTRODUCTION experiments to clearly identify and model the epitaxial crys-

Laser photochemical vapor deposition (LPVD) draws tal growth regimes with and without ultraviolet (UV) laser

on laser photochemistry to produce, independently of sub- irradiation have been carried out and are reported here. The

strate temperature, atoms, molecular radicals, or excited KrF laser was used in the bulk of these studies to simplify

species which are not normally present in significant concen- analysis since arsine is virtually transparent at this wave-trations in conventional metalorganic chemical vapor depo- length,9 while TMG has a gas phase photoabsorption cross

sition (MOCVD) or molecular-beam epitaxy. The large section o of - 10- 9 cm 2 . On the basis of the results, the

photon energies available from existing ultraviolet lasers mechanism responsible for the enhancement in the film

(and the excimer lasers, in particular) are generally more growth rate of GaAs for A = 248 nm is attributed to photoly-

than ample to rupture metal-ligand bonds in a precursor sis of trimethylgallium (TMG) in the adlayer.

molecule by the absorption of a single photon. Hence the Furthermore, the influence of UV radiation on the epi-

reactor chemistry can be altered in a controllable, although taxial layer morphology has been investigated in the 500-

unconventional manner. Examples of the potential utility of 650 C temperature range where the surface morphology has

this technique include the growth of SiN 4 by photodisso- deteriorated from the smooth and specular appearance typi-

ciating mixtures of SiH 4 and NH3 with an excimer laser cally observed at higher growth temperatures. Noticeablebeam (ArF: 193 nm)" 2 or the recently demonstrated growth improvements in layer morphology are observed with UVof epitaxial Ge on GaAs with the assistance of laser pro- illumination for each temperature studied, and under appro-duced hydrogen atoms.' priate growth conditions, specular surfaces are obtained for

Over the past 4 years, several groups 7 have demon- temperatures as low as 550 *C. In contrast to the film growth

strated the growth of l11-V films by irradiating the substrate rate, this improvement is independent of wavelength in the

with focused or collimated excimer laser beams. In each 193<A<351 nm range.

case, however, transient heating of the substrate and grow- 1. EXPERIMENTAL DESCRIPTIONing film was a significant or controlling factor in the growthprocess owing to the high laser fluences employed ( > 50 A schematic diagram of the MOCVD reaction chambermJ/cm2 ). and UV laser optical system is illustrated in Fig. 1. Experi-

Recently, we reported' a photoenhancement in the ments were carried out in a 2.5-cm-square horizontal quartzgrowth of epitaxial GaAs films on GaAs substrates at a tem- chamber at atmospheric pressure with hydrogen as the car-perature of 450 °C by illuminating the substrate in an rier gas and source vapors of TMG ( - 10 =C) and arsineMOCVD reactor with no more than 13 mJ/cm- of KrF radi- ( 10% in hydrogen). The untilted graphite susceptor is 2.5ation (248 nm). The absence of any enhancement at 351 nm cm in length with its leading edge 10 cm from the gas inlet.(XeF laser) demonstrated the photochemical (as opposed Quartz plates are placed between the inlet and the susceptorto thermal) nature of this effect. Isolating the observed pho- to eliminate any discontinuity in the gas flow to the sub-toenhancement mechanism is a difficult task if the operating strate. Heat is provided by rf induction, and the temperatureregimes for the reactor are not known. Therefore, a series of is monitored by a thermocouple at the center of the susceptor

5001 J Ap . Phys. 66 (10), 15 November 1989 0021-8979/89/225001-0802.40 c) 1989 American Institute of Physics 5001

_._..._f placed immediately on top of the chamber.Laser Optical Window Growth rates were determined by measuring layer

Therrocouple thicknesses with a high-resolution scanning electron micro-0scope (SEM). Samples were cleaved and stained using the

aluminum selective etch 1 HF: 5 CrO3 : 30 H.O. The effectsjet loansc otoof the GaAs buffer and AlAs spacer layer on crystallinity

Gs Iwere investigated by growing a series of samples with andTMG+AsH3+H2 0 0 0 0 without buffer and spacer layers. When no GaAs buffer lay-

RF Coil er was grown, the AlAs spacer was omitted, and growth was

FIG. 1. Schematic diagram of the horizontal MOCVD reaction chamber initiated directly onto the substrate. The buffer and spacerwith optical window purge configuration and perpendicular substrate irra- layers were both grown at 700 *C. The crystalline quality ofdiation geometry, these epitaxial layers was determined" from transmission

electron microscopy (TEM) electron diffraction patterns.TEM specimens were prepared with standard thinning and

directly below the substrate. Although most experiments mounting techniques, and examined with a Phillips EM 430were performed between 450 and 650 *C, growth tempera- microscope operating at 300 kV.'2

tures ranged from 425 to 800 *C. Surface morphological studies were carried out atThe substrate was irradiated at normal incidence by a growth temperatures ranging from 500 to 650 *C in 50 *C

collimated excimer laser beam (ArF, KrF, or XeF) which increments withaTMG flow rateof2.5 or 10sccm, anarsinewas spatially filtered to a cross-sectional area of 1.4 cm 2, and flow rate of 100-500 sccm, a total hydrogen flow rate of 5100entered the chamber directly above the substrate through an sccm, and subsequent epitaxial layer thicknesses of 0.1-0.75optical window equipped with a hydrogen purge, as depicted Mum. For these studies, the 700 'C preheat and the GaAs buff-in Fig. 1. Pure hydrogen flowing through this purge path er/AlAs spacer layers were eliminated. Illuminated andprevents reactants in the main gas stream from reaching and nonilluminated samples were grown separately to spatiallythus depositing on the optical window and thereby attenuat- characterize morphology features. The substrate was illumi-ing the laser beam. The hydrogen flows through the purge nated by the UV laser beam at 193, 248, or 351 nm withwindow and the chamber are balanced for equal gas veloc- fluences up to 19 mJ/cm 2 for KrF and 8 mJ/cm2 for ArFities (60 cm/s) with a total flow rate of - 10 slpm. and XeF and pulse repetition frequencies between 4 and 30

Prior to each growth, the (100) GaAs:Si substrates Hz. Morphologies of epitaxial layers were examined with awere cleaned with organic solvents and hot 1:1 HCI:H,O. Nomarsky interference phase contrast microscope at magni-After the substrate was heated to 700 *C in an arsine-hydro- fications of up to 1000 x.gen ambient for 10 min to remove any residual oxide, a 0.25-pum GaAs buffer layer and a 150 . AlAs spacer layer were Ill. RESULTSgrown. The purpose of the AlAs layer is to serve as a marker A. Reactor characterizationto aid in delineation of the top GaAs layer. For the samples An Arrhenius plot for the growth rate of GaAs is showngrown below 700 *C, the temperature was lowered after the in Fig. 2, where two of the three regimes' ' in conventionalgrowth of the AlAs layer and stabilized for 5 min. Finally,0.1-0.2 Mm of GaAs was grown with the desired growth andoptical excitation parameters. For each sample, a totalgrowth time of 30 min was the maximum length of time over 5.Ea15 kcarnolwhich the excimer laser power output was stable relative to a athe initial output power at the start of the growth run.

Several experiments were performed to characterize , 5.0reactor growth temperature regimes and film growth rateswith and without laser irradiation. First, the temperature -<variation of the GaAs growth rates was ascertained by keep- M 45ing the TMG and arsine mass flow rates fixed at 3 and 500sccm. respectively, while varying the temperature in incre-ments from 425 to 750 'C. Next. to verify diffusion limited . 4.0growth at 700 °C, the TMG flow rate was varied between 2and 10 sccm for a constant arsine flow rate of 100 sccm. 3 mGFinally, the growth rates at 450 °C were measured for arsine 500 37 Asn3.5 . . . .flow rates between 30 and 500 sccm (0.23-3.7 Torr) with 0.9 1.1 1.3 1.5the TMG flow rate held constant at 3 sccm (0.23 Torr). In 1/T 'K (x1000)this last set of experiments, half of the substrate was maskedfrom the optical beam and the remaining half was irradiated FIG 2. Arrhenus plot of the growth rate for GaAs with fixed TMG (3m ) and arsine mass flow rates 500 ,ccm) to illustrate two of the three(A = 248 or 351 nm) with 7-13 mJ/cm at a pulse repetition growth regimes dictated by temperature in the MOCVD growth of GaAs.frequency (PRF) of 30 Hz. In order to minimize beam dif- The apparent activation energy E, in the kinetically controlled growth re-fraction effects at the substrate, the laser beam mask was gime is 15 kcal/mol.

5002 J. Appl. Phys., Vol. 66, No. 10, 15 November 1989 York t al. 5002

atmospheric pressure MOCVD are readily identified. These 80three regimes are i) low-temperature surface kinetics limit-ed growth, (ii) midtemperature diffusion limited growth, .and (iii) high-temperature desorption and diffusion limited 70 ""growth. In the kinetic regime (typically below 525-575 *C),the surface reaction rates are slow compared to the diffusion 60 J

of the source molecules to the surface, and the growth rate iscontrolled by the adsorption of both arsine and TMG mole- I-/

cules onto the surface. 5 For this regime, the apparent acti- 50

vation energy E. was found to be 15 kcal/mol. Between- 600 and 800 *C, diffusion of the source molecules to the 4

surface becomes slow compared to the surface reaction rates, T-4 0"Cand the film growth rate is then source diffusion limited. 0.3 Torr 1MGArsine adsorption is strong in this temperature range, and 32 4the growth rate is independent of arsine partial pressure as Arsine Pa Pressue (Ton')long as the concentration of arsenic-containing species islarger than the concentration of the gallium precursor.' 5 Un- FIG. 4. Data obtained for the growth rate variation with arsine partial pres-der these conditions, the growth rate varies linearly with the sure for both masked substrates (solid circles) and substrates illuminatedwith - I I mi/cm2 of 248-nm radiation (open circles) for fixed TMG par-TMG mass flow rate. At temperatures above - 800 *C, the tial pressure (0.23 Torr and temperature (450 °C). The solid curve repre-growth rate continues to be TMG diffusion limited, but de- sents the best fit of the Langmuir-Hinshelwood model to the nonilluminat-creases in magnitude, most probably caused by depletion of ed data, while the dashed curve corresponds to the best fit for data measured

reactants from gas-phase prenucleation and increased de- in the presence of UV radiation.

sorption of adsorbed molecules. t4 For the reactor configura-tion of these experiments, the growth rate appears, from Fig.2, to be diffusion limited above - 550 *C. The dependence of B. Effect of UV laser radiation at the substratethe GaAs growth rate on the TMG mass flow rate for a Data acquired for the GaAs growth rate variation withconstant arsine flow rate of 100 sccm that is observed at arsine partial pressure for both the nonilluminated (solid700 'C is shown in Fig. 3. The linear variation of the growth circles) and illuminated substrates (open circles) are plot-rate confirms TMG diffusion limited growth at this tem- ted in Fig. 4 for A = 248 nm, - 10 mJ/cm2 (30 Hz),perature, which is in a region which will not be of further T = 450 C, and the TMG partial pressure fixed at 0.23interest in this paper. Torr. The estimated relative error between illuminated and

All samples grown with a GaAs buffer layer were single masked samples, arising primarily from variations in thecrystal, even at 425 'C, and for those grown without the growth rate with position on the wafer, is 2.5% and is repre-AlAs spacer, the interface between the low-temperature lay- sented by the error bar in Fig. 2. To account for small fluctu.er and the buffer layer could not be discerned in cross-sec- ations in the laser intensity and to obtain a realistic compari-tional specimens. Those grown at 500 'C without the higher son between data points, the as-measured growth rates fortemperature buffer layer were polycrystalline, and there was the irradiated samples were normalized to an averageno measurable growth below 500 C without the buffer layer. fluence over the entire run as well as to an average initialThe presence of the AlAs marker layers resulted in no ob- fluence.8 Note the increase in growth rate for the illuminatedservable change in the crystallinity of the samples. regions for each arsine flow investigated, giving an enhance-

ment over the masked regions of between 5% and 15%. A0,12 T.700C sublinear dependence on arsine partial pressure is observed100 snefor both masked and irradiated regions of the substrate, and0.10 the growth rates eventually saturate with increasing concen-

trations of arsine. An increase in the pusle repetition fre-0.0e quency of the laser would be expected to increase the en-

hancement in the growth rate in a linear fashion, provided00 laser-induced temperature effects do not become important.0Surface-kinetics-controlled growth in MOCVD can be0.04 described by the Langmuir-Hinshelwood model,'4 - '6

which, indeed, predicts a sublinear variation with both0.02 TMG and arsine partial pressure. The growth rate R for this

0.00 . . . . ,model is given by

TMGMa FlowRate(sccrn) R= k, 0 k (, P ,, A I (I)

FIG 3. Dependence of the GaAs growth rate on TMG mass flow rate for where k is a rate constant, . is the adsorption rate constantthe substrate temperature and arsine mass flow rate held constant at 700 °Cand 100sccm. respectively. The linear variation demonstratesdiffusion lim- divided by the desorption rate constant for TMG (G) orited growth at this temperature, arsine (A). p is the partial pressure, and 0 is the molecular

5003 J. Appi. Phys., Vol. 66, No. 10. 1 5 November 1989 York et at 5003

surface coverage. This sublinear dependence is caused by a 90

saturation in surface coverage for large partial pressures ofthe source molecules.

Identifying the cause of the increase in growth rate with so

laser irradiation requires an examination of the low-tem-perature growth mechanisms. The Langmuir-Hinshelwood 11 70model assumes noncompetitive adsorption of TMG and ar-sine followed by surface reaction,' as opposed to gas-phasereaction. Furthermore, the growth rate is not hindered by 60the desorption of reaction by-products such as methane. ' It A 2.has been suggested'5 that arsine adsorption occurs prior to so -K G 79AI ndecomposition, with rapid loss of the first hydrogen atom ./Q-67 minafter surface adsorption, probably followed by the first bi- ICOG 9 Aiminmolecular methane elimination. At higher temperatures 40where the growth is TMG diffusion limited, arsine adsorp- 0 1 2 3 4

tion is very strong (indicating unity surface coverage), while Arne Partial Pressure (Tot)

desorption of TMG remains important' 4 so that p,,6,, ).Iand PGA5G 4 1. Under these conditions, Eq. ( ) predicts that 90

the growth is proportional to the TMG and independent ofarsine partial pressure, as expected for diffusion limited 80growth.

A least-squares fit of the Langmuir-Hinshelwood mod-el [Eq. (1) to our experimental data is represented by the ., 70solid (no laser) and dashed (laser) curves in Fig. 4. The i ,"'"

adjustable parameters in obtaining the fit arefl', and the ,product kO,. Without further empirical data on TMG sur- 60face adsorption and the measurement of the surface reaction 0 2rate constant k, in particular, fo and k cannot be deter- / A3.4

50 1 A 3mined separately. From the analysis./34 is found to decline sO 2.9by - 15% from 3.4 for masked samples to 2.9 when the A

surface was illuminated. However, the product kOG in- 4 . . ....creased by - 13% from 70 to 79 A/min with laser irradia- 0 1 2 3 4

tion. Figure 5 (a) depicts the sensitivity of the fit of Eq. ( 1) Arsine Partial Pressure (Tor)to the data for a change of ± 15% in kOG. Here the best fit isrepresented by the solid line, and clearly, the agreement with FIG. 5. Curve fits of the Langmuir-Hinshelwood model to the data ob-the data deteriorates rapidly with small changes in kO,. tained for the GaAs growth rate in the presence of UV radiation (I I mJ/

cm2. 248 nm) for different values of the parameters used in the analysis toConversely, the insensitivity of the curve fit to f., is demon- ascertain the sensitivity of the curve fit to these parameters. The solid curvesstrated in Fig. 5 (b) where a + 15% change in/3., produces represent the best fit using the values$, = 2.9 and kO, = 79 A/min; thea much smaller perturbation in the fit to the data, indicating dashed and dotted curves are computed for (a) kOG values of 67 and 91 A/that arsine surface adsorption is virtually unaffected by opti- min. and (b) 6 values of 3.4 and 2.6. respectively.

cal irradiation at this wavelength. These results are consis-tent with the fact that arsine is nearly transparent, whileTMG has a gas-phase photoabsorption cross section of- iO ' cm2 at 248 nm. We conclude, therefore, that the pendent of laser radiation, and there were few or no distin-GaAs growth rates measured in these experiments at 450 "C guishable morphological characteristics caused by opticalare governed by the Langmuir-Hinshelwood model and irradiation between these samples. For these reasons, we to-confirm the premise that the UV optical beam interacts pri- cused our attention on samples grown without buffers in themarily with adsorbed TMG at this wavelength. As an aside, temperature range 500-650 *C. Undoubtedly. incompleteit should be noted that Tanaka, Deguchi. and Hirose" have removal of the native oxide from the GaAs surface prior tointerpreted their recent experiments involving the laser pho- growth degrades the quality of the epitaxial layer, " particu-tochemical vapor deposition of Si with an ArF laser in an larly at temperatures near or below the desorption tempera-analogous manner. ture of the oxide ( - 580 °C"'). For this reason, the 700 °C

preheat and GaAs buffer layer were eliminated from the setC. Surface morphological studies of experiments carried out between 500 and 650 'C to ascer-

In the course of these experiments. it was discovered tain the effects of residual oxide, since, as a practical matter.that epitaxial layers grown at and below 500 'C had smooth it is desirable to do away with rregrowth thermal treat-and specular morphology if a GaAs buffer layer was first ments.grown at 700 *C. Additionally, no measurable growth was The epitaxial layer morphology of nonilluminated sam-initiated below 500 "C unless a buffer layer was present, inde- pies grown within this temperature range ( 500-650 'C) with

5004 J. Apo. Phys., Vol. 66, No. 10. 15 November 1989 Yorket al 5004

a TMG flow rate of 10 sccm and arsine mass flow rates > 100 samples grown at a TMG flow rate of 2.5 sccm. For fluencessccm were rough with irregular features. Reducing the below 8 mJ/cm2 , the resulting morphologies for the threegrowth rate by lowering the TMG flow rate to 2.5 sccm different wavelengths were very similar-mostly specular(0.23 Torr) significantly reduced the density of the irregular and smooth between small clusters of irregular features. Byfeatures, but did not produce specular surfaces at any of raising the fluence to 19 mJ/cm2 at 248 nm, it was possible tothese temperatures. Increasing the arsine flow rate from 100 obtain featureless surfaces at a TMG mass flow rate of 2.5to 500 sccm while keeping the TMG mass flow rate constant sccm even for temperatures as iuw as 550 *C. Micrographs ofhad no observable effect on morphology. Growth at 700 °C morphological features taken in a Nomarsky-interferencewas determined to be the lowest temperature for which fea- phase-contrast microscope at 625 X are shown for a sampletureless surfaces could be obtained for each of these TMG grown at 550 "C both (b) with (KrF, 19 mJ/cm2 ) and (a)flow rates. without illumination of the substrate in Fig. 6. The elongated

Optical irradiation at all wavelengths (193, 248, and region of poor morphology in Fig. 6(b) corresponds to an351 nm), fluences (5-19 mJ/cm2 ), and PRF (4-30 Hz) imperfection in the quartz window where the beam was at-investigated promoted smooth surface morphology for the tenuated. ArF and XeF laser fluences above 8 mJ/cm2 werechosen temperature range 500-650 °C, most notably for not available with our laser, and consequently, completely

featureless morphologies have not been observed for thesewavelengths. In general, increasing the PRF causes the clus-

ters of irregular growth to spread out, leading to a smoothen-ing in the surface morphology. Featureless surfaces were

achieved at each temperature only for the combined highestPRF (30 Hz) and fluence (19 mJ/cm2 ) available.

IV. ANALYSIS

A. Growth ratesThe conclusion that the growth rate enhancement of

Figs. 5 and 6 stems from the interaction of 5 -eV photons withadsorbed (as opposed to gas phase) TMG molecules is based

alkyl molecules in an adlayer or in the gas phase. Rytz-Froi-

devaux, Salathe. and Gilgen,2 ° Haigh,2' and, more recertly.Itoh et al.22 have shown that 248 nm lies in the red wing ofthe gas-phase photoabsorption spectrum of TMG. At that

4 Iwavelength, the single-photon (Beers law) absorption cross(ml 40 j.Lm section is ar- 1 10- " cm 2 which, when combined with the

low peak powers used in these experiments (7<(< 13 mJ/cm 2, 0.35<1<0.65 MW/cm2), yields low gas-phase photo-dissociation rates for the metal-alkyl. That is, sinceorlAt(hcw) -'- 1.5 x 10- 1 (where At is the KrF laser pulsewidth of 20 ns FWHM), the optical transition is* far fromsaturation and, for a TMG partial pressure at 300 K of 0.23Torr, the volumetric rate at which (CH,)_,Ga-(CH,) bondsare ruptured,22-' is -2.4x 10" cm s (assuming the

* scission of onc bond for every absorbed photon). This resultis consistent with the observation that < 1% of the beamenergy is absorbed by TMG in the gas phase for a window-to-substrate distance of - 1.2 cm. It appears, therefore, thatthe 248-nm laser radiation produces predominantly dimeth-

- , ylgallium radicals in the gas phase, and multiphoton effects

are negligible.While it is likely that partially photodecomposed TMG

molecules [such as (CH 3 ). Ga] in the gas phase have an

(b) 40im increased surface chemisorption probability owing to the re-moval of one or more methyl ligands. photodissociation

FIG 0 Surface morphological features observed 'i~th a Nomarskv-inter- products lying outside the boundary layer will have littleference phase-contrast microscope at 625 - (a) without laser irradiation impact on film growth. This constraint alone makes it im-and h with laser irradiation ( KrF. 19mJ/cm-2

of the suhstrate during possible to attribute the growth rate enhancement of Fig. 4 togrowth at a temperature of 550 *C. and TMG and arsine mass flow rates of2.5 and 10) sccm. respectively. The elongated region of rough morphology the gas phase photodissociation of TMG.in i) corresponds to an imperfection in the quartz window where the laser Several groups "24 -" have clearly demonstrated the im-heam has been attenuated. portance of the adlayer in determining the growth rate of

5005 J. AppI. Phys., Vol. 66, No 10, 15 November 198g York of al 5005

metal films by photodissociating column TIB or liA or no- sional, time-dependent heat-flow equation32

ble metal-alkyl molecular precursors. W 1ch regard to the / 21,(1 - R))present experiments, in particular, the TMG adlayer has T(t) J -'(2)several characteristics which indicates that it is the domi- \ K /

nant factor in film growth. Foremost is the fact that, as noted where I,, is the peak laser intensity. D is the thermal diffusiv-previously,3" the number density of reactive species at the ity, K is the thermal conductivity, and R is the reflectivity ofsurface that is produced by adlayer photolysis will typically the substrate. Although this relation assumes that the pa-be orders of magnitude larger than that available from gas- rameters defined above are temperature independent, K isphase photodissociation since ( 1) the TMG number density actually a strong function of temperature33 ; therefore, anin the adlayer is generally six to eight orders of magnitude average value in the temperature range of interest was used.larger than it is in the gas phase; and (2) as noted by Brai- For the values 33 K = 0.2 W/cm K at 450 "C,chotte and van den Bergh,3° the limited diffusion rates of Cp = 0.327 J/g K, p = 5.315 g/cm 3, D( K'pCp ) = 0. 11molecules in the adlayer are conducive to film growth if the cm 2/s, R = 0.6 at 248 nm, and a peak laser intensity of 0.55removal of more than one ligand of the alkyI is necessary to MW/cm 2 (which corresponds to a fluence of - 11 mJ/initiate growth. That is, more than one photon can be readily cm 2 ), the peak instantaneous temperature rise after oneabsorbed by an adsorbed TMG molecule since its mobility is pulse is - 55 *C. At a repetitior rate of 30 Hz, there is - 33low. At least two experimental studies 25 .3 ' have provided ms between pulses so that the average temperature rise isstrong evidence that the photoabsorption spectrum for an negligible, as expected.adsorbed metal-alkyl species is elongated to the red as com- From Fig. 2, the growth rate at 500 °C (substrate tem-pared to the same spectrum for its gas-phase counterpart. perature of 450 "C plus laser transient heating of - 50 *C) isThis has been most convincingly demonstrated by the re- nearly twice that at 450 'C, which is substantially larger thancent, detailed experiments of Zhang and Stuke"' for several the growth rate enhancement of 5%-15% observed here.alkyls of aluminum. Furthermore, the exposure of column However, the duty cycle of the laser (allowing for heatingIIB metal-alkyl adlayers to UV radiation has been shown to and cooling transients comparable to the laser pulse width)convert physisorbed molecules into chemisorbed species,27 a is still 10- .Extrapolation of the gas phase thermal crackingprocess which manifests itself in further broadening of the rates -" of TMG between 450 and 500 °C yield dissociationadlayer's absorption spectrum. times on the order of milliseconds. Consequently, the instan-

The relevance of these factors to our experiments is ob- taneous temperature rise of - 55 °C in a time frame on thevious. The ratio of TMG densities in the adlayer to that in order of tens of nanoseconds is unlikely to be responsible forthe gas phase under typical experimental conditions is esti- the observed film growth rate enhancement.mated to be 10'. Also, as noted earlier, 248 nm does lie in thered wing of the TMG gas phase absorption spectrum.'- 22 B. Surface morphologyHence it is plausible to expect the absorption spectra for the Lowering the substrate temperature in conventionalphysisorbed and chemisorbed species to shift progressively MOCVD correspondingly brings about a degradation of thetowards the red with the result that. as noted by Chen and surface morphology, which is caused primarily by a reduc-Ogood2' for (CH,) ,Cd, the absorption strength at 248 nm tion in surface mobility of the source species on the growingof a chemisorbed TMG layer can be expected to be consider- surface."' More specifically, the consequence of reduced mo-ably larger "...than for an equivalent quantity of molecules bility is that molecular or atomic species are not always ablein the gas phase." Rough calculations relating the observed to migrate to the energetically most favorable lattice sites,increase in film growth rate to the photon flux available at resulting in a roughening of the growth surface. This will bethe substrate demonstrate the plausibility of the last point, of particular importance at low growth temperaturesnamely, that the effective cross section for TMG adsorbed ( t 500 °C) where an independent source of energy must beonto GaAs is one to two orders of magnitude larger than that supplied to break up source molecules as well as providefor the gas phase species. ' 5 ample surface mobility to obtain homoepitaxial layers.

The most conclusive support for the claim of a photo- In the present experiments, it became increasingly diffi-chemically induced enhancement in growth rate at 248 nm cult to affect improvements in the surface morphology as thewas provided by measuring samples which were grown un- temperature was dropped. Presumably, this was due to theder otherwise identical reactor conditions, but irradiating dramatic decline in surface mobility of the source speciesthe substrate with a XeF laser beam (351 nm). For laser and perhaps because of the decrease in decomposition effi-fluences and PRF similar to those used at 248 nm, no change ciency of the source molecules, particularly arsine. Addi-in growth rate in the illuminated portion of the substrate as tionally, the native oxide on GaAs desorbs at - 580 °C, andcompared to the masked portion of the substrate was ob- may. in part. account for the rough surface observed atserved. These findings are consistent with the fact that gas 500 °C even at the highest available fluences, and the factphase TMG and arsine are optically transparent at 351 nm, that no growth was initiated without a buffer at tempera-despite broadening of the alkyl absorption spectrum towards tures below 500 'C. In fact, preliminary studies indicate thatthe red in the adlaver. preheating the substrate above the oxide desorption tem-

Furthermore, calculations of the instantaneous and perature without growing a buffer layer improves the subse-average temperature rise of the substrate surface due to laser quent epitaxial morphology. Source decomposition effi-transient heating were carried out by solving the one dimen- ciency and residual oxide, however, are not likely to be

5006 J ApoI Phys.. Vol 66, No. 10. 15 November 1989 York etat 5006

limiting factors in the quality of the morphology in the 600- particularly strong (as compared to that for 'he gas-phase650 *C range. Hence the attainment of featureless surfaces at species) for chemisorbed TMG.these higher temperatures by illumination with optical Finally, experiments carried out with 351-nm (XeF)fluences sufficiently low that the rise in the substrate tern- irradiztion of the substrate revealed no measurable enhance-perature during growth is negligible indicates that an enhan- ment of the film growth rate, which clearly demonstratesdement is occurring in the surface mobility of molecular or that the effect observed at 248 nm is photochemical in origin.atomic species at the growth interface. All of the available evidence suggests that the role played by

These results are similar to those obtained in previous the laser is to photodissociate physisorbed or chemisorbedstudies,36"' 37 where it was demonstrated that UV irradiation (rather than gas phase) 3" TMG. Since Zhang and Stuke3'of the substrate during growth at temperatures lower than have demonstrated that the photolysis of adsorbed trimethy-optimal improves the morphology of the epilayer. In the lat- laluminum at 248 nm generates considerably more A1CH.er study, Kukimoto er aL37 observed the same trends as those than does the corresponding reaction in the gas phase; it isreported here except that they were unable to generate fea- likely that a similar conclusion holds true for TMG. as well,tureless surface morphology below 700 *C, even at a higher and that, as suggested in Ref. 31, a logical course for thefluenceof30mJ/cm 2 (1.5W/cm 2 , 50Hz). Webelieve, how- future of laser-assisted MOCVD experiments would ne toever, that this is due to the high growth rates ( 1000 ,/min) explore halogen containing precursors such as dimethylgal-characteristic of their studies as opposed to those (000 ,/ lium chloride.min) that we found necessary to obtain specular morpholo- Contrary to the wavelength dependence observed forgies in our experiments. Although completely specular sur- the growth rate enhancement, significant improvement infaces were achieved in the present investigation only when surface morphology is observed at temperatures as low asusing KrF (248 nm) at the highest fluence of 19 mJ/cm2 , it 500 °C with UV laser irradiation at each of the three wave-is believed that similar results would be observed with ArF lengths investigated ( 193, 248, and 351 nm). Mirrorlike su-(193 nm) and perhaps even XeF (351 nm) if sufficiently faces are obtained for temperatures > 550 °C with KrF athigh fluences were available. The important point to be fluences well below those which will generate a significantmade here is that the improvement in surface morphology average substrate temperature rise. Residual surface oxidemust be attributed to a photochemical mechanism since the contributes deleteriously to the resulting layer morphologylowest temperature at which mirrorlike surfaces can be ob- at growth temperatures below that at which the native oxidetained without UV irradiation of the substrate is - 700 *C. desorbs, and at the lowest growth temperature (500 °C), la-The apparent lack of dependence of the effect on laser wave- ser irradiation does not render completely smooth morphol-length in the UV suggests that the photogeneration of free ogy. Between 425 and 500 °C, oxide or an otherwise imper-carriers by the optical source ma" be the dominant process. fect substrate surface may be the limiting factor in initiatingExperiments conducted at still longer wavelengths (i.e., growth. Other groups" ' have reported success with in situA > 350 nm) will be required to verify this possibility in pregrowth cleaning by illuminating the substrate withGaAs. Similar observations have been reported " for the fluences of - 100 mJ/cm2 (ArF). owing probably to thephotochemical growth of ZnSe and ZnS, where the growth nonnegligible transient temperature rise expected at thisrate exhibited a threshold for photoinduced enhancement at fluence. Clearly. it is desirable to remove all unnecessarya wavelength corresponding to the band gap of the material. thermal treatments from the growth process, and we intend

to investigate photochemical means of in situ pregrowthV. SUMMARY preparation of the substrate surface.

Significant and reproducible enhancements in thegrowth rate of GaAs have been observed in an MOCVD ACKNOWLEDGMENTSreactor by illuminating the substrate with 7-13 mJ/cm of The authors are grateful to S. Piette. V. Tavitian, D.248-nm radiation. Several factors demonstrate that this ef- Kane, K. King, C. Abele, E. Cheng, K. Kuehl. and C. Kielyfect is photochemical in nature. Data acquired with and for helpful discussions and technical assistance. This workwithout laser radiation are well described by the Langinuir- was supported by the NationAl Science Foundation (DMRHinshelwood model for which the curve fit to the raw data is 86-12860). the Air Force Office of Scientific Research (H.%irtually insensitive to 3 . regardless of whether the sub- Schlossberg, F49620-85-C-0141), and the Charles Starkstrate is irradiated or not. Therefore. photothermal effects Draper Laboratory (DLH 285419). G. E. Fernandezl.is up-( i.e., laser-induced substrate heating) may be discounted ported, in part. by a fellowship from G.T.E.;ince significant heating of the substrate rapid!y acceleratesarsint adsorption, '" which would, in turn, create a signifi-.ant lependence onfl ,. Calculation ofihe average tempera- 'P K. Boyer. G A. Roche. W H. Ritchie. tnd G. J. Collins. Appi Physture rise of the substrate surface support, this conclusion. In Lett. 40. 716 1 (992). R. Sohnki. W H. Ritchie. and G 1. Collin%. Appl.Phys. Lett. 43.454

I 53): P K Bover. K. A. Emery. H Zarnant. and K.contrmst. kN,, increases from -0 A/min without laser irra- I Collins. Appil Ph , Left 45. 70, JOQ, 4

diation to 79 A; mm in the presence of 5-eV photons. which T F Deutsch. D I SdverNmnth. and R. W Mountain. Mater Res Soc-is also consistent with the known gas-phase absorF:ton spec- Syrup. Proc. 17. 129 , 1043C. J Kjely. V Tahlllian, C. Jone,.. iiJ (3. Eden. Appi Phys. let 55. b5trum for TMG and the expected hroadening of that spec- C. iltttrum toward the red for the adsorbed species. An increase in 'v \ Donnelly. D Braen. A. Appelhaum. and M Geva, 1. Appi. Phys.

the absorption cro.s section at 248 nm is expected to become 58 2022 1 198)

5007 J Aopi Phys., VOl 6& No 10, 15 Novenber 1989 York etal, 5007

5C. W. Tu. V. M. Donnelly, J1. C Beggy. F. A. Baiocchi. V. R. McCrary, T. Appi. Phys. Lett. 50, 77 (1987),D. Harris. and M. G. Lamont, Appi. Phys. Lett. 52. 966 (1988) "Y. Rytz-Froidevaux. R. P. Salathi. and I. H. Gilgen, Maier. Res. Soc.

'V.MN. Donnelly, C. W. Tu. J. C. Begg, V R. McCrary, M. G. Lamont. T. Symp. Proc. 17, 29 (1983).D Harrs. F. A. Baiocchi. and R. C. Farrow. AppI. Phys.Lett. 52. 1065 2'1. Haigh, I. Mater. Sci. 18, 1072 (1983).(1988). 2 H. Itoh, M. Watanabe, S. Mcukai, and H. Yajima, J1. Cryst. Growth 93, 165

'S. S. Chu. T L. Chu. C. L. Chang, and H. Firouzi. Appi. Phys. Lett. 52. (1988).1243 (1988). 23S. Oikawa. M. Tsuda, M. Morishita. M. Mashita. and Y. Kuniya. J. Cryst.

T. K. York. J. G. Eden. J. J. Coleman. G. E. Fernandez, and K. I. Beer- Growth 91, 471 (1988).nink. AppI. Phys. Lett. 54. 1866 (1989). "'D. J. Ehrlich, R. M. Osgood, Jr., and T. F. Deutsch, Appi. Phys. Lett. 38,

'See, for example, J. Nishizawa and T. Kurabavashi. Maier. Res. Symp. 946 (1981).Proc. 101. 275 (1988) -,C. M. Humphries,.A. D. Walsh, and P. A. Warsop, D. J. Ehrlich and Rt. M. Osgood, Jr., Chemn. Phys. Lett. 79, 381 (198 1).Discuss. Farady Soc. 35, 148 (1963); J. H. Clark and R. G. Anderson, 'Y. R) zFroidevaux, R. P. Salathi, H. H. Gilgen, and H. P. Weber, AppI.Appi. Phys. Lett. 32.46 (1978); B. Koplitz. Z. Xu. and C. Wittig,. 4ppl. Phys. A 27. 133 (1982).Phys. Lett. 52. 860 (1988); H. Okabe. Photo-. hemistry of Small Molecules 2"C. J. Chen and R. M. Osgood, Jr., Mater. Res. Soc. Symp. Proc. 17, 169(Wiley. New York, 1978). p. 269. (10983),Y. Rytz-Froidevaux. Rt. P. Salathi, and H. H. Gilgen. Mawer.'Res. Soc. 'J. Y. Tsao and D. I. Ehrlich, J. Chemt. Phys. 81, 4620 (1984).Symp. Proc. 17,29 (1983). 2'G. S. Higashi and L. J. Rothberg, AppI. Phys. Lett. 47, 1288 (1985); G. S.

"C. J. Kiely, P. K. York. J. G. Eden, J. 1. Coleman, G. E. Fernandez. and Higashi, and C. G. Fleming, Appi. Phys. Lett. 48, 1051 (1986).K.)J. Beernink (unpublished dapa) - 30a Braichotte and H. van den Bergh, AppI. Phys. Lett. A 45, 337 (1988).

Z2S. J. Jeng. C. Ni. Wayman. G. Costrini. and J.3J. Coleman, Maier. Lett. 2, 31y. Zhang and M. Stuke, J. Cryst. Growth 90, 143 (1988).3590(984). 32P. Bacri and S. Campisano, in Laser Annealing of Semiconductors, edited

'"D. Shaw in CrystalGrowth T/reory and Techniques, edited by C. Goodman by J. Pdate and . Mayer (Academic, New York, 1932), pp. I5- 109.(Plenum. New York, 1974). Vol. 1, p. 1. 13J. S. Blakernore, J. Appi. Phys. 53, R123 (1982).

"D. H. Reep and S. K. Ghandi, J. Electrochem. Soc. 130, 675 (1981). ' 4M. Suzuki and M. Sato, J. Electrochern. Soc. 132, 1684 (1988)."D. 1. Schlyer and M. A. Ring. J Electrochem. Soc. 114. 9 (1976). "Surface morphology."'S Thomson and G. Webb, Heterogeneouss Catalysis (Wiley, New Yfork, 'N. Puitz, H. Heinecke, E. Veuhoff, G. Arens, Mi. Heyen, H. Lujth, and P.

Xi6), p. 87. Balk,)J. Cryst. Growth 68, 194 (1984)."'T. Tanaka, K. Deguchi, ,and M. Hirose, Jpn. 3. Appl. Phys. 26, 2057 "H. Kukimoto, Y. Ban, H. Komatsu, M. Takechi, and Mi. Ishizaki, I

(1987). P NCryst. Growth 77; 223 (196).'V. M. Donnelly, V. R. McCrary, A. Appelbaum. D. Brasen, and W. P'. "S. Fujita, A. Tanabe, T. Sakamoto, M. Isemura. and S. Fujita, J. Cryst.Lowe. J. Appl. Phys. 61. 1410 (1987). , Growth 93, 259 (1988).

"'A. J. SpringThorpe, S. J. Ingrey, B. Em'~erstorfer, and P. Mandeville,

16

500 J. AppI. P"'y.. Vol. 66, No. 10. 15 November 1989 York et a 5008

Low-temperature laser photochemical vapor deposition of GaAsP. K. York, J. G. Eden, J. J. Coleman, G. E. FernAndez, and K. J. BeeminkMaterials Research Laboratory and Everitt Laboratory. University of Illinois. 1406 West Green StrntUrbana. Illinois 61801

(Received 12 December 1988; accepted for publication 17 February 1989)

The growth of epitaxial GaAs at temperatures below 500 C by ultraviolet laser-assistedmetalorganic chemical vapor deposition has been investigated. Experiments were conducted at248 am (KrF excimer laser) and 351 un (XeF) in normal incidence with laser fluencesmaintained below 13 mJ/cm2. While the growth rate was enhanced by 5-15% at 450 C uponirradiating the substrate with 248 rin photons, no measurable effect was observed at 351 nm.This strong wavelength dependence at low fluence demonstrates that the film growthenhancement mechanism is photochemical in nature.

Laser-assisted chemical vapor deposition is often broad- 248nm (KrF) and at 351 nm (XeF) irradiated the substrately subdivided into laser photochemical vapor deposition at normal incidence. Prior to each growth, the(LPVD) or laser-induced chemical vapor deposition (100) GaAs:Si substrates were cleaned with organic sol-(LCVD), depending on whether the role played by the laser vents and hot 1:1 HCI:H 20. The substrate was then heatedin the growth process is primarily a photochemical or ther- to 700 C in an arsine-hydrogen ambient for 10 min to re-mal one. Both spatially delineated (selective area) and move any residual oxide, after which a 0.25 #im GaAs bufferbroad-area epitaxy have been demonstrated with focused vi- layer was grown. The temperature was then lowered and thesible' or ultraviolet - (typically, frequency-doubled Ar ) final GaAs layer grown with the desired growth and opticallasers and collimated excimer laser beams,' 4 respectively, excitation parameters. Laser irradiation was present onlyAlthough several studies have been reported in which III-V during the growth of the final GaAs layer.films were grown in metalorganic chemical vapor depo- Extensive experiments were performed to characterizesition3"-' (MOCVD) or molecular beam epitaxy7 systems the reactor growth regimes 2 and growth rates with andaccompanied by illumination of the substrate with ultravio- without laser irradiation. Specifically, the growth rates atlet (UV) radiation, the energy fluences and wavelengths 450 C were measured for arsine mass flow rates between 30used have made it difficult to resolve the relative contribu- and 500 sccm (0.23-3.7 Torr) with the TMG mass flow ratetions of thermal and photochemical processes to the overall fixed at 3 sccm (0.23 Torr). Half of the substrate wasgrowth rate. Furthermore, prior studies have typically in- masked from the UV beam while the remainder was irradiat-volved laser energy fluences at the substrate of 50 to > 100 ed (4k 13 mJ/cm2 ) at a repetition rate of 30 Hz. Measure-mJ/cm2 . Under such conditions, thermal effects dominate ments of the average power were taken immediately beforesince, for example, instantaneous surface temperatures ex- and after each growth run to monitor power fluctuations forceeding 1200 K are attained for an incident energy fluence of a given excimer laser gas mixture during any given run as100 mJ/cm2 and 10-20 us pulses. In short, whether UV a- well as changes between different runs. Growth rates wereser-auuisted film growth is driven photochemically or pho- determined by measuring layer thicknesses of cleaved andtothermally critically depends on the laser fluence and wave- stained cross sections with a high-resolution scanning elec-length. tron microscope (SEM). The crystalline quality ofthese epi-

The photochemically enhanced growth of epitaxial taxial layers was determined from transmission electron mi-GaAs by UV radiation in a MOCVD reactor is reported croscopy electron diffraction patterns. 3 ",here. Experiments were conducted at 248 and 351 am in Figure 2 illustrates the sublinear dependence of thenormal incidence with the laser fluence maintained below 13 growth rate on arsine partial pressure that was measuredmi/cm2 at all times in order to prevent transient heating ofthe substrate and adlayer. At 248 nm, arsine is tranapar-entO' while the absorption cross section" for trimethylpl- o wlium (TMG) in the gas phase is - 10 "cm 2.Neither arsine ~inor TMO absorbs significantly at 351 nm"' which provides H2xua convenient experimental means for differentiating between 0 0phtothermal and photochemical enhancement mecha-

A schematic diagram of the reaction chamber and laser TMG*A*4+,2 0 0 0 0apparatus is shown in Fig. 1. Experiments were carried out ,W Cal

in a square (2.5X2.5 cm2 ) horizontal quartz chamber atatmospheric pressure using hydrogen carrier gas and source I . Schemaic diagram of the horizontal MOCVD reaction chambervapors of TMG ( - 10 C) and 10% arsine in hydrogen. with opt"I wbdow purpe coniguration and perpendicular substrate ir,-The rectangular beam from an excuser laser operating at diation geometry.

t00 Ao. Phy&. LL 54 (19), 8 May 1980 0003-M1/89/191666-03801.00 IO Amero klmtd of Phylcs low

so Table I, where it can be seen that the fluence was not con-stant over the course of an entire run and between runs. To

~-.. account for this small linear variation in laser intensity and70 ., to obtain a realistic comparison between data points, the as-

=i.measured growth rate was normalized to an average fluenceso I over the entire run as well as to an average initial fluence.

/0 I These values are given in the last column of Table I. TheA Langmuir-Hinshelwood fit to the normalized data points

50 obtained with UV irradiation is represented by the dashedcurve in Fig. 2 for which kOG = 77k /min and 69A = 3.4.

Clearly, both masked and illuminated data are well de-40 scribed by the Langmuir-Hinshelwood model under the

Ta4S0Oc specified growth conditions.TMG-0.23 Torr Although the film growth enhancements obtained in

30 ... ... these experiments were no greater than 15%, these data are0 1 2 3T 4 representative of a larger set of data generated under differ-

Ans Poolal Pruasust (Tort) ent growth conditions, which consistently demonstrate en-

FIG. . Data illutrating the enhiancement of the GaAs growth rate as a hancement with UV illumination of the substrate at 248 nn.fiaction of arine partial pressure that is observed upon illuminating the However, experiments carried out at 351 nm with similarMbstrate With -10.7 mJ/CM2 Of248 m radiation with fiXed TMG panial fluences, repetition rates, and otherwise identical growthpressure (0.23 Torr) and temperature (450 *C). The solid circles representthe results obtained in the absence of U irradiation and the solid curve conditions show no measurable change in growth rate. Thisindicates the bat at of the Langmuir-Himneiwood model to these dam wavelength-dependent growth rate is consistent with the ab-while the open circles and the dashed curve correspond to growth rates and sorption cross sections for TMG and arsine at each wave-thebest fit fordatameasured in thepresenceofUV irradiation, respectively, length and demonstrates that the growth enhancement ob-

served at 248 nm is photochemical in origin.This result is also in agreement with estimates of the

with a fixed TMG partial pressure for the substrates blocked peak (instantaneous) and average temperature rise at thefrom the laser beam. An absolute uncertainty of approxi- substrate surface which are induced by the incident lasermately 10% in the ordinate of the graph was introduced in beam. Integration of the time-dependent, one-dimensionalthe measurement by procedures such as sample staining and heat flow equationIsSEM micrograph measurement. At 450 "C, the growth rateR is surface kinetically controiled and governed by the Lang- T(t) = (2o(1 -R)/K f ] 41 i, (2)muir-Hinshewood Model'" - 7 which, indeed, predicts a where! 0 is the peak laser pulse intensity (-7 X 0 W/cm2 ),sublinear growth rate dependence on both TMG and arsine D is the thermal diffusivity, K is the thermal conductivity,prta pressure: and R is the reflectivity of the substrate, yields a negligible

- P4, 1 PR.A4 average temperature rise.RI + kOpO, = (1) Previous investigations"' of the photochemistry of

l + P "'0, 1+P AUV-irradiated column IIIA, IIB, or noble metal-alkyl ad-where k is a rate constant, P is the adsorption rate constant layers have clearly demonstrated the enhanced efficiency ofdivided by the desorption rate constant for TMG (G) or photodisocating alkyl molecules in the adlayer as opposedarine (A), p is the partial pressure, and 6 is the molecular to the gas phase. As discussed in detail by Wood et al.' andsurface coverage. This sublinear dependence is caused by a Braichotte and van den Bergh,' the total number of alkylsaturation in surface coverage for large partial pressures of molecules which are able to contribute to film growth is con-the source molecules. The best fit of Eq. (1) to this data, siderably higher in the adlayer than in the gas phase. This is awhich is represented by the solid curve in the figure, yieldski =70 A/in and b = 3.4. The details of this model consequence of the large ratio of adlayer to gas phase num-

ber densities and the fact that only those gas phase moleculesand the sensitivity of the fit to assumed values for the con- that are photodissociated near the substrate actually con-stants k, P, and 0 will be discussed elsewhere. ' tribute to film growth. Not surprisingly, then, simple calcu-

The dependence of the normalized growth rate on arsminepartial pressure for substrates irradiated at a fluence of- 10.7 mj/cm2 (A= 248 nm) and otherwise identical TABLE 1. As-measured GaAs growth raes for illuminated -mp (241growth conditions is shown by the open circles in Fig. 2. The nm) nortalized to changes in fluence during and between growth runestimated relative error between illuminated and masked Growth Initial Final Aveage Normalizedsamples was 2.5% arising primarily from variations in the P G rote inuae Fuence Auene Norated

growth rate with position on the wafer and is represented by (Torr) (A/mm) (mi/cmz) (mi/cm') (mJ/cm2) (A/min)

the error bar in Fig. 2. As was observed for the masked sam-pies, the growth rate dependence is sublinear but the magni- 0.44 49.3 10.3 13.2 12.0 43.7

0.73 50.7 10.6 9.1 9.9 54.7tudeofthegrowthrateisreproduciblyincreasedbyapproxi- 2.26 69.3 10.6 11.1 10.7 61.7mately 5 to 15%, depending on anine partial pressure. The 3.74 71.3 10.7 10.5 10.6 72.0as-measured growth rates and laser fluences are shown in

167 W Phys. L.. Vl. 54, No. 19, 8 May1960 York at. 1667

lations show that the 5-15% enhancement in the growth 2S M. Bedsir, J. K. Whisnant, N. H. Karm, D. Grif N. A. EJ-Mavy,rate cannot be accounted for by photodissociation of TMG and H. H. Stadelmaier, J. Cryst. Growth 77, 229 (1986).rD. J. Ehrlich, R. M. Osgood, Jr., and T. F. Deutsch, J. Vac. Sci. TechnoLin the gas phase. Physisorbed TMG molecules, on the other 21, 23 (1982).hand, are prone to chemisorption once the scission of one or *T. H. Wood. J. C. White, and B. A. Thacker, Appl. Phys. Lett 42, 408more Ga--CH3 bonds by photodissociation occurs. Fur- (1983).theriore, Rytz-Froidevaux et alY and Zhang and Stuke2 ' 'G. S. Higashi and C. 0. Fleming. Appl. Phys. Lett. 48,1051 (1986).

eJ. Nishizawa and T. Kurabayashi, Mater. Res. Symp. Proc. 101, 275have investigated in detail the visible and UV photochemis- (1988).try of cadmium and aluminum alkyl adlayers, respectively, 'V. M. Donnelly, C. W. Tu, J. C. Beggy, V. P. McCrary, M. G. LaMont, T.and, in the latter study, the fragments resulting from the D. Harris, F. A. Baiocchi, and . Farrow, Appi. Phys. Ltt. 52. 1065

photodissociation of both the gas phase and adsorbed i (19).weree c a'S. S. Chu, T. L Chu, C. L Chang. and H. Ftrouzi, AppL Phys. Lett. 52,were compared at 193, 248, and 351 nm. These results,21 in 1243 (1988).

particular, demonstrate that the highly nonequilibrium dis- 9H. Kukimoto, Y. Ban, H. Komatsu, M. Takechi, and M. shizki, J.tribution of products emanating from the adlayer is deter- Cryst. Growth 77, 223 (1986).

mined by photochemical considerations and photothermal 'See, for example. C. M. Humphries, A. D. Walsh, and P. A. Wanop,Discuss. Faraday Soc. 35,148 (1963); J. H. Clark and R. G. Anderson,

effects can be ignored. We conclude therefore, that photoly- Appi. Phys. Lett. 32, 46 (1978); B. Koplitz, Z. Xu, and C. Wittig. Appl.

sis of adsorbed TMG at 248 nm is the dominant mechanism Phys. Lett. 52,S860(1988); H. Okabe, PiwtchemistryofSmallMoleules

responsible for the photochemical growth enhancement re- (Wiley, New York, 1978), p. 269.p Y. Rytz-Froidevaun, R. P. Salathe, and H. H. Gilgen. Mater. Res. Soc.

Syrup. Proc. 17,29 (1983).

In summary, an enhancement in the growth rate of '2P. K. York, J. G. Eden, J. 1. Coleman, 0. E. Feruindez, and K. J. Beer.GaAs has been observed by irradiating the substrate in a nink (unpublished data).

MOCVD reactor at 248 nm in normal incidence. No mea- "C. J. Kiely, P. K. York, J. G. Eden, J. J. Coleman, G. E. Fernindez, andK. J. Beernink (unpublished data).

surable increase in growth rate is observed for similar exper- "S. J. Jeag C. M. Wayman, 0. Costrini, and J. J. Coleman. Mater. Lett. 2,ments conducted at 351 nm. This strong wavelength depen- 359 094).dence demonstrates that the enhancement mechanism at 0D. H. Reep and S. K. Ghandi, J. Electrochem. Soc. US, 675 (1983).

248 am is photochemical in nature and is attributed to the "D. J. Schlyer and M. A. Ring. J. Electrochem. Soc. 114,9 (1976).of8 ishtoem in nurf e adar i "S. Thomson and G. Webb, Heterogeneous Catalysis (Wiley, New York,

photolysis of TMG in the surface adlayer. 1968), p. 87.The authors are grateful to S. Piette, V. Tavitian, D. "P. aeri and S. Campisano, in LaserAnnealing of Semkouctors, edited

Kane, K. King, C. Abele, E. Cheng, K. Kuehl, and C. Kiely by J. Poate and J. Mayer (Academic, New York, 1982), pp. 75-1409.

for helpful discussions andechnical assistance. This work D. J. Ehrlich and It. M. Osgood Jr., Chem. Phys. Let. 79, 381 (1981).feC. J. Chen and R. M. Osgood, Jr., Mater. Res. Soc. Symp. Proc. 17,169

was supported by the National Science Foundation (DMR (1983).86-12860), the Air Force Office of Scientific Research (H. 1J. Y. Tsao and D. J. Ehrlich, J. Chem Phys. SI, 4620 (1984).

Schlossberg, F49620-85-C-0141 and 89-0038), and the 2G S Higashi and L-. Rohberg, App. Phys. Lett. 47,1288 (1985).

Charles Stark Draper Laboratory (DLH 285419). G. E. "D. Braichotte and H. van den Bergh. Appl. Phys. A 45,337 (1988)"C. D. Stinespnng and A. Freeman, Appl. Phys. Lett. 52,1959 (1988).

Fernindez is supported, in part, by a fellowship from GTE. "'Y. Rytz-Froidevaa, R. P. SalaM H. H. Gilgean, and H. P. Weber. Appl.Phys. A27, 133 (1982).

'Y. Aoyagi. M. Kana=aw, A. Dot. S. lwai, and & Namba, . Appi. Phys. '. Zhang and M. Stuke, J. Cryst. Growth n, 143 (198).6, 3131 (1986).

loss A40. PhyS. Let, Vol. 54, No. 19,6 May INO York oSta. IM

Laser photochemical deposition of germanium-silicon alloy thin filmsHubert H. Burke and Irving P. HermanDepartment of Applied Physics, Columbia University. New York, New York 10027

Viken Tavitian and J. Gary EdenEve rit Laboratory, University of Illinois. Urbana. Illinois 61801

(Received 14 February 1989; accepted for publication 10 May 1989)

Thin films of Ge-Si alloys were deposited by 193 nm photolysis of GeH4/Si2H6 gas mixturesusing an ArF laser. For substrate temperatures below 350 C, deposition occurred only withthe laser present, while for temperatures above 400 T, film growth was little influenced bylaser photolysis and resembled conventional chemical vapor deposition (CVD). The Si/Geratio in the films was about three times the Ps, /Po, H ratio of reactant partial pressures fordeposition in either the laser photolysis or the CVD regime. This result indicates that there isstrong cross chemistry between silicon and germanium-bearing species in the gas phase. Filmstoichiometry was measured by Auger analysis and Raman spectroscopy, with both methodsleading to the same film composition.

The intriguing structural, electrical, and optical proper- tant pressure was fixed: PsnH. + POW = 0.5 Torr. This disi-ties of germanium-silicon alloys can be tuned by varying the lane fraction x, was varied in different experiments to growrelative Ge/Si composition.I This fact, along with the com- films over the full compositional range 0<x 1, where xfpatibility of these compounds with existing silicon-based = [Si /( [Si] + [Gel ) is the atomic fraction of Si in thetechnology, has led to the growing number of applications of film.thin-film Ge-Si alloys in band-gap engineering of silicon- The thickness of Ge-Si films on quartz was determinedbased microelectronic devices, solar cells, infrared detectors, by etching a trench down the center of film, and subsequent-and optical fiber communications.' Although several differ- ly measuring the step height by profilometry. Film stoichi-ent techniques of depositing Ge-Si alloy films have been in- ometry (silicon substrates) was measured by Auger electronvestigated, including pyrolytic chemical vapor deposition spectroscopy (AES). Depth profiles were obtained by prob-(CVD), and laser and plasma-assisted CVD,2 no studies of ing the 1147 eV Ge and 1619 eV Si peaks using 3 keV elec-laser photochemical vapor deposition (LPVD) of these trons directed to the center of the region that was sputteredcompounds have been performed to date. The advantages of by argon ions. Quantitative analysis was accomplished byLPVD include high deposition rates at substrate tempera- comparing the amplitude (peak-to-peak height) of the Au-tures lower than those characteristic of conventional CVD, ger signal from the component in the alloy to the corre-decreased atomic diffusion due to low-temperature oper- sponding amplitude from a pure elemental standard ana-ation, and monolayer control. These latter two points are lyzed under identical experimental conditions. Si-rich filmsparticularly important for multilayer fabrication. Investiga- and films grown on quartz were analyzed by x-ray photo-tions of excimer laser assisted LPVD of Ge-Si alloy thin electron spectroscopy (XPS) using the Al K, line.films are reported here for the first time.34 Several AES studies of alloys have shown that back-

Details concerning the experimental apparatus for the scattering of primary electrons and the escape depth of Au-deposition experiments can be found elsewhere .6 and will ger electrons depend on alloy composition, 7 and that, duringonly be briefly reviewed here. An ArF laser beam (193 nm, sputtering, the surface can be enriched by the alloy compo-40 Hz, 15 mJ/cm2 ), filtered by a rectangular slit (0.5 X 2.0 nent with the lower relative sputtering yield." Since thesecm2), entered the quartz reactor through one wall and pro- effects can influence the Auger determination of xf, theirpagated parallel to and 2.2 mm above quartz and Si sub- importance in Ge-Si alloys was tested by determining thestrates. Both substrates were resistively heated (", = 250- film composition of thick - I1-pm-thick CVD-grown films500 C) by an inconel heater containing a I-mm-diam hole, by both electron beam induced x-ray energy-dispersive anal-which permitted the passage of a He-Ne laser beam through ysis (EDAX) and AES. EDAX gave the same film composi-the quartz substrate, enabling in situ monitoring of the film tion as did the AES analysis, to within ±t 0.02.growth by transmission. An infrared pyrcmeter measured The stoichiometry of these thicker films was also deter-the quartz surface temperature at the point where the He-Ne mined by Raman scattering, using the shift of the Si-Si likelaser passed through the substrate. peak near 500 cm -'. Of the three peaks in the Ge-Si alloy

During deposition, a disilane/germane gas mixture (di- Raman spectrum, the variation of the frequency of this peakluted in a helium buffer-5% reactant gases/95% He) flowed with xf is the least sensitive to the method of prepara-over the substrate at a mass flow rate of 100 sccm, and a total tion.2 9., Measurements were performed with a microprobereactor pressure of 10 Torr. The mole fraction of disilane in system (described previously)," and using the 5145 A linethe gas, defined by x, P34,/(P.H + PeaH, ), where ofanAr4 laser operated at low power ( S2mW,0.6/mspotP3,. and P(,4, are the partial pressures of disilane and ger- size) to minimize substrate heating. After initial laser prob-mane, respectively, is taken to be a measure of the relative ing of the sample, the laser power was increased to about 40concentration of reactive gases. Furthermore, the total reac- mW for approximately 5-10 min, and then the sample was

253 A4*. Ptys. Let SS (3). 17 July INS8 0003-6951 /89/290253-03S01 00 © 1989 Amn ns Ittute of Physics 253

reexamined at low power to check for the effect of annealing. mixture for several minutes prior to initiating growth of theFor T, >400 C, the ArF laser was found to have only a alloy.

minor impact on film growth, which occurred primarily by The film composition (xf) determined by Auger andconventional CVD. However, below 350 C, no film growth XPS was in reasonable agreement with the Raman measure-was observed in the absence of UV laser radiation, indicating ments. The Raman spectrum given in Fig. I (a) depicts thecomplete laser control of growth in this experimental regime three-peak spectrum for CVD films grown on c-Si at 500 "Cof high flow velocity (20 cm/s) and low reactant gas partial with gas composition x, = 0. 10, and film composition xfpressure (0.5 Torr). Probe transmission, monitored during = 0.45, and indicates that the film is polycrystalline. Figurethe growth of photolytic films, showed that incubation times 1 (b) shows the Raman profile of a film deposited under sim-immediately preceding film growth were negligible for x, ilar conditions except that x, = 0.90 and x1 = 0.91. TheZ0.005 ( <1 min), but were significant for x. = 0 (pure broad peak shows that the as-deposited film is amorphous;GeH4 , 2 8 min). Deposition rates ranged from 150 to 300 the narrow peak is the result of Raman analysis after the filmA/min for both 250 and 350 "C substrate temperatures, and was converted to polycrystalline material by laser annealing.were insensitive to the disilane and germane mole fraction, Under these conditions, the CVD films were shown to bewith the disilane fraction x, >0.l and with constant total amorphous for x1> 0.78 and polycrystalline for x1 <0.45reactant gas partial pressure. (and were not examined for intermediate x).

Depth profiles of the photolytic films and several CVD If the dissociation (and film incorporation) probabilityfilms grown on silicon showed the composition to be inde- of Si2H6 were8 times that of GeH4, then it can be shown thatpendent of depth, and revealed the presence of an oxide layer the silicon fraction in the film (x,,) and the disilane fractionat the film-substrate interface for the photolytic films despite in the reactant mixture (xa) are related by the expressionin situ cleaning of the substrate with a 5% HCI/95% He gas xx s

=00 (a)600 (a) In Fig. 2, the measured silicon fraction in the film (x-) is

As deposited plotted versus the right-hand side of Eq. (1) for fl = 3,600 Ge-Ge showing very good agreement for the photolytically grown

G,-Si films and relatively good agreement for the CVD films, forwhich# 05:t 2. This means that the [Sil/[Gel ratio in the

400 s I-s photolytically grown film is enriched by a factor of 3 relativeto Pu.M/PO.H, (the gas phase partial pressure ratio) and a

- .00 factor of 1.5 relative to the atomic [Si]/(Ge] ratio in thereactant mixture.

At 193 nm, germane is weakly absorbing with an ab-f 0. . ..... . . . sorption cross section of approximately 3X 1020 cn 2 , 12

sooo Mhi while Si2H4 is strongly absorbing with an absorption crosssi-sl section -. two orders of magnitude larger, -2X 10- S

cm2. 3 If, after laser photolysis of the reactants, there werec tLaser annealed

2000 no "cross chemistry" between Ge and Si-containing species,then 8 would expected to be roughly 70 or 140 based on theAs deposited

1000- 1.0

04- a

0 .,.....;, .. . ....... .,,,. , , , .........

250 300 350 400 450 50 S 0.6Reman Shift (i

-m)

FIG. 1. Raman microprobe analysis of Ge-Si alloy thin films grown by 0.,4CVDon c-Si with T, =5 W0Cand I mWof 5145,A Ar Iaser power. TheSi

fraction in the film (x,) is determined by the frequency shift of the Si-Si

peak. (a) Spectrum acquired for a disilane fraction in the gas mixture of x 0.2

=0. 10 during deposition. Auger and EDAX measurements give x. = 0.45and 0.46, respectively, while Raman analysis gives 0.34 (using Ref, 9 forcaibration) or 0.42 (using Ref 10). depending upon the calibration curve. 0 0.2 OA 0.6 0.8 1.0

and shows that the as-deposited film is polycrystalline. (b) Analogous spec- X,' (0- 3)trum acquired for x, = 0.90. with x, determined to be 0.91 by XPS. The as-

deposited film is amorphous, while the laser-annealed film (40 mW, 5 min) FIG. 2. Measured Si fraction of the film (x,) as a function ofx," ' (0 = 3)

is polycrystalline, giving x1 = 0.92. For x, = 0.50, x, was measured to be for various x, under different experimental conditions. The solid symbols

0.78 from AEM 0.80 from EDAX, and 0.76 from the Raman spectrum of depict films grown by 193 nm photolysis at lower substrate temperaturesthe laser-annealed film (spectrum not shown). Each of the Si-Si Raman where the laser is necessary fordeposition: (circles) T, = 250"C; (squares)shift vs silicon fraction calibrations in Refs. 2. 9, and 10 gave the same x, for T, = 350 C. The open symbols depict films grown by conventional CVD at

the x, = 0.90 experiment. This was also true for the x, = 0.50 run, higher temperatures: (circles) T, = 400 "C; (squares) T, = 50 C.

254 AppI.Py Left., Vol. 55. No. 3.17 July 1989 Burke et al. 254

relative absorption cross sections, depending on whether between Ge and Si species have been found to be importantonly one or both of the disilane Si atoms contribute to film during film deposition and Raman spectroscopy has beengrowth. Since the value of 0 for LPVD is considerably shown to be a useful diagnostic of Ge-Si alloy film composi-smaller, collisions between GeH. and Si-bearing radicals or tion.H atoms from disilane photolysis, which form Ge-bearing This work was supported at Columbia by the Office ofradicals, undoubtedly play a pivotal role in deposition. Simi- Naval Research and at Illinois by the Air Force Office oflar cross chemistry between Ge and Si-containing species is Scientific Research (AFOSR F49620-85-C-0141) and byalso likely in conventional CVI) since fi-3 also describes the NSF (NSF DMR86-12860) through the Materials Re-the CVD-grown films fairly well. One further argument in search Laboratory. The authors would like to thank Dr. C.support of the impact of cross chemistry in forming the laser- F. Yu for assistance in making the Auger and XPS measure-deposited film stems from the measured rate of deposition ments.of Ge atoms, which exceeds the calculated rate at whichgermane molecules are photolyzed when disilane is present 'For example, see J. C. Bean, Science 230, 127 (1985); 0. Abstreiter, H.in the reactant mixture (specifically, for x, >0.01 ). For ex- B uge, T. Wolf, H. jorke and H. jL Hero Phys.Rev. Lemt K4 2441ample, the Ge atom deposition rate is 8.5 x lO"4/cm 2 s dur- (1985).ing photolytic growth of Ge-Si alloy films when the disilane 21. P. Herman and F. Magnotta, J. App!. Phys. 61, 5118 (1987), and refer-

fraction xe = 0. 1 and T, = 350 C, while the rate of photo- ence ctedthereinl.H..Buk,.PHemnS.A'These results were first discusaed by H .Bre .P emn .A

lyzing GeH4 molecules is only - 1.4X 10"Icme s. Pif V. Tavitian and J. G. Eden at the Conference on Lasers and Elec-It appears, therefore, that deposition of the alloy is initi- tro-Optics (CLEO'88), April 25-29, 1988, Anaheim, CA, paper W7.2,

ated by laser photodissociation of disilane, and photolysis of Technical Digest p. 264.

GeH4 is negligible except when x. <0.01. Recent studies 'After the completon ofths wok,the auhors beameaware ofan eper-suggst hat n te laer luece rngeof teseexpeimets, mental study entailing the photochemical vapor deposition of Ge, _isuggst hat n te laer luece rngeof teseexpeimets, alloys, in which the optical source was a Hg lamp: A. Yamada, Y. ha. M.

Si2H, absorbs only one 193 nm photon,'4 and that the pri- Konagai, and K. Takahashi. Jpn. J. AppI. Phys. 27, L2164 (1988).mary dissociation quantum yield is -0.7.' 'At least 20 exit JV. Tavitian. C. J. Kiely, D. B. Geohegan, and). G. Eden, Appl. Phys. Let

channels are c nergetically accessible in the photolysis of disi- 52, 1710 (1988).nM. 1-1. 16'C. J. Kiely, V. Tavitian, and J.0G. Eden, J. AppL. Phys. 65,3883 (1989).lane at 193 ti. Many of these channels may well be 'L. C. Feldman and J. W. Mayer, Fundamentals of Surface and Thin Film

important, producing H, H2, Siff, Siff, and other Si-bear- Ancaj'sis, 1st ed. (North-Holland, New York, 1986).ing species." Once formed, the silylene radical (SiH 2), for 1H. J. Mathieu and D. Landolt, Surf. Sci. 53, 228 (1975).

exmpe, reacts with GeH, via the process IT. Ishidate, S. Katagiri, K. Inoue, M. Sbibuya, K. Tauji. and S. Mino-exampl e, +MG iH+M(weeMianthr mura, J. Phys. Soc. Jpn. 53, 2584 (1984).

SiH +Ge4 MGeiH +M whreM s nythrd 'OMARenucci J.B.Renucci ndMCadoa, incfndi~pofthCs-body) which is analogous to the insertion reaction ference on Light Scattering in SoUi4 edited by M. Balkaki (Flammar-SiH2 + SiH, + M- Si2H, + M that is known to be critical ion, Paris, 1971). p. 326.to the pyrolysis of silane. ' In addition, H atoms produced "0. D. Pazionis, H. Tans, and!1. P. Herman. IEEE . Quantum Electron.

25, 976 (1989).from the photodissociation of Si2H6 Will form GeH3 and H2 '2M. J. Mitchell, X. Wang, C. T. Chin, M. Suto, and L C. Lee, 1. Phys. B 20,by hydrogen abstraction from GeH4. Since a given species 5451 (1987).formed by the laser within the photolysis zone suffers rough- "1U. Itoh, Y. Toyoshima, H. Onuki, N. Washida, and T. Ibuki, L. Chem

ly 5000 collisions with background reactant molecules be- Phs 85,4867 (1986)."E.. Boch. C. Fuchs, E. Fogarassy. and P. Sitfert, Mater. Res. Soc. Symp.

fore reaching the surface, the nascent products of disilane Pro. 129 (in press).photolysis are collisionally transformed by reactive colli- "53.0. Chu. M. H. Beenan, J. S. NicKillop, and). M. Jasinai, Chemsions into the final deposition precursors, perhaps Phys. Lett. 155, 576 (1989); J. M. Jasinski, J. 0. Chu, and M. H. Bege-

Ge, S. 2(-+ ),while en route to the surface. mann, Mater. Res. Soc. Symp. Proc. 131,487 (1989).,~ 16H. Stalaat, Appl. Phys. A 45,93 (1988).

In summary, low-temperature growth of Ge-Si alloy "t For example, see B. A. Scott, W. L. Olbricht, B. A. Meyerson, J. A.films has been demonstrated by laser LPVD. Cross reactions Reimer, and D. 3. Wolford, J. Vac. Sci. Technol. A 2,450 (1984).

255 AppI. Phys. Lett., Vol. 55, No, 3. 17 July 1969 &irks or*/. 255

44

Reprintd frorr

Journalica ofpr dpostio

7..

*A.L*.. %*x.

(inwersay of of.inhesAmUrbcaa Illtnouse61801Physics

pp 388 3895

Microstructural studies of epitaxial Ge films grown on [100] GaAs by laserphotochemical vapor deposition

C. J. Kiely,a) V. Tavitian, and J. G. EdenUniversity of Illinois, Urbana, Illinois 61801

(Received 11 July 1988; accepted for publication 10 January 1989)

Ge films grown on [ 100] GaAs by laser photochemical vapor deposition (LPVD) in parallelgeometry at temperatures ( T ) ranging from -240 to 415 °C have been examined bytransmission electron microscopy. For 285 < T, : 330 'C, a thin (250-700 A,) epitaxial film isgrown initially but a switch to amorphous material is subsequently observed. At highersubstrate temperatures ( T, Z 400 *C), thicker ( Z 800 k) epitaxial Ge films are grown beforethe transition to polycrystalline material takes place. In the absence of external 193-nm laserradiation (i.e., growing by conventional low-pressure chemical vapor deposition), the Ge filmsare completely amorphous (285 4 T <330 *C) or heavily defected polycrystalline(T, -400 C). The > 100 °C temperature reduction for the growth of epi Ge films madepossible by LPVD is attributed to the direct production of a species (GeH 3 ) by the laser whichis collisionally converted to GeH 6 en route to the substrate. Upon reaching the [ 1001 GaAssurface, the digermane is pyrolyzed. Experiments with [ 100] substrates tilted Y toward f 1101yielded thinner ( - 120 A), but smooth, epitaxial films which is attributed to the higher densityof available nucleation sites. Films grown at 280-330 "C on [ 111 oriented GaAs werecompletely amorphous which appears to arise from reduced adatom mobilities on [ 111 ]surfaces. These results clearly demonstrate the feasibility of photochemically generating aspecies which migrates to the surface and alters the chemistry at a substrate (which is itself notilluminated by the optical source) so as to permit the growth of epitaxial semiconductor films.

L INTRODUCTION traviolet (VUV) laser [ArF (Refs. 14 and 15) or F2 (Ref.

The field of laser photochemical vapor deposition 15) 1. At much higher temperatures (730 5 T, :< 925 C),

(LPVD) was born in the late 1970s amid considerable opti- Ishitani and co-workers"' 8 have reported volatilization of

mism regarding its potential for growing films at tempera- impurities and enhanced film growth rates when illuminat-

tures lower than those accessible to conventional growth ing Si substrates with 1-3 W cm-I (average power) of ultra-

techniques. Since that time, most of the effort in this area has violet (UV) radiation (from a Hg/Xe lamp or KrF or ArF

focused on the deposition of metals by photodissociating a excimer lasers) during the pyrolysis of SiHC12/H, mix-

precursor polyatomic molecule.' Much less is known of the tures.

feasibility of growing device quality, epitaxial semiconduc- Among the compound semiconductors, crystalline InP

tor films.; and polycrystalline GaAs have been grown by photodisso-

Early work with the optically assisted growth of elemen- ciating mixtures of columns III and V alkyls with excimer

tal semiconductors involved polycrystalline Ge (and to a laser radiation. 2 2 In the past several months, the results ofleser extent, Si) films that were grown on [100 ] NaCi, SiO 2, coupling an excimer laser to MOCVD (metalorganic chemi-

and [ 1T021 sapphire 4 at substrate temperatures (T,) as ca vapor deposion),- MBE (molecular beam epitaxy) low as -25 °C (room temperature) by photodissociating or CBE (chemical beam epitaxy 25 reactors have also beenGeH4 (or SiH 4 ) with an excimer (ArF or KrF) laser. Al- reported. Upon irradiating the substrate with 193-nm (ArF

though mercury photosensitization has been known (for laser) energy fluences of -40-300 mJ cm - 2 , epitaxial

two decades)5 to be capable of depositing amorphous Si GaAs or AIGaAs films were obtained. For such high peak

films, '" Nishida et al." 9 have recently grown polycrystalline powers (5-20 MW in - 15 ns pulses), however, film crystal-

Si:H layers and epitaxial Si films at 200 "C by irradiating linity and enhancements in the growth rate are primarilymixtures of Si2H,, SiH 2F,, H, and Hg vapor with a low- attributable25 to substrate heating (i.e., pyrolysis) rather

pressure Hg lamp (A= 185 254 nm; fluence (0 -30 than a photochemical effect.

mW cm-2). Resorting to the higher silanes allows one to Irvine and co-workers2 "29 were the first to grow high-avoid the sensitizer entirely and several groups' 5 have re- quality 11-VI epitaxial films (HgTe, HgCdTe) by Hg photo-

ported the deposition of a-Si films at temperatures below sensitization. Similar results have been obtained recently by300 C by directly photodissociating SiH, or SiH, with a Ahlgren and co-workers." Morris" and, more recently,

lamp [Hg (Refs. 10-12) or D, (Ref. 13)] or a vacuum ul- Zinck etal.'2 demonstrated the epitaxial growth of HgCdTeon CdTe and CdTe on GaAs, respectively, by photodisso-ciating the associated column II and VI alkyls with an ArFor KrF (248 nm) laser.

Visiting Postdoctoral Associate: 1986-88. Ultraviolet and visible laser (or lamp) radiation has also

3883 J AppI Pl ys 65 (10). 15 May 1989 0021-8979189/103883.13$02.40 c; 1989 American institute of Physics 3883

been shown to be effective in improving the quality of III-V phous or polycrystalline films. Therefore, it is apparent thatand II-VI films grown by conventional techniques. Ni- the laser directly produces a species which migrates > 10-60shizawa and co-workers 3 reported photochemically in- mean free paths (MFPs) to reach the surface or is collision-duced changes in film growth rates for GaAs grown by vapor ally transformed en route into a more stable species. Uponphase epitaxy (chloride transport) for 480< T, < 700 *C. arriving at the substrate (which is itself not illuminated),Similar results (improved growth rates and film morpholo- this species dramatically alters the surface kinetics so as togy) were obtained by Pitz et al.34 for GaAs films and Fujita make the growth of epitaxial films possible. Specifically, it isand co-workers35 for ZnSe films grown by MOCVD while concluded that the key role played by the laser is to produceirradiating the substrate with a low pressure Hg or Xe lamp. GeH 3 radicals which subsequently form GeH, by threeSignificant experimental activity3 '4 2 has also been devoted body collisions.to modifying the structural and electrical characteristics of This paper describes in detail the microstructural char-MBE or MOCVD grown films with Ar4 ion (visible) laser acteristics of these Ge films as examined by transmissionillumination of the substrate. electron microscopy (TEM) and energy dispersive x-ray

Virtually all of these previous experiments involved illu- (EDX) analysis. The next section discusses the experimen-minating the surface with UV (or visible) radiation and, in tal apparatus and data-acquisition procedure. Section IIImost cases, transient (or cw) heating of the substrate and presents the experimental results and the key conclusions ofgrowing film played a significant or dominant role in the this study are summarized in Sec. IV.formation of crystalline material. Aside from the early stud-ies of Collins and co-workers,"'" and Deutsch et al.,4 6 in II. EXPERIMENTAL APPARATUS, DATA ACQUISITION,depositing compound insulators (A120 3, SiO2 , Si 3N4 ), few AND ANALYSISLPVD experiments have been carried out in parallel geome-try (i.e., substrate is not irradiated). Notable exceptions are A. Apparatus: Reactor design and film thickness

the recent II-VI experiments (mentioned earlier) of Mor- measurements

ris3 'and Zinck and co-workers"2 and the Si:H work of Tagu- A partial schematic diagram of the experimental ar-chi et al.," the latter of which yielded amorphous material. rangement is given in Fig. 1. The reactor is constructed of

In summary, despite successful demonstrations of the commercial quartz tubing having a square cross sectiongrowth of both elemental and compound semiconductor (2.5 X 2.5 cm 2 ) and a length of - 30 cm. Each end of thefilms by optically assisted processes, few experiments have reactor is gradually tapered down to a final diameter of 8appeared in the literature reporting the LPVD growth of mm. Gas flow and pressure are both monitored and con-epitaxial materials. Those instances in which crystalline trolled upstream and downstream of the reactor by massfilms were reported have generally involved laser (or lamp) flow controllers, transducers, and a capacitance manometer.heating of the substrate or a sensitizer to couple the optical The entire gas handling and flow system (except the reactorsource to the gaseous precursor molecule. Also, as Houle2 itself) is constructed from welded stainless-steel tubing andhas stressed in a 1986 review paper, systematic studies corre- VCR fittings.5 'lating reactor conditions and laser parameters with film The unfocused beam from an ArF excimer laser ( 193quality (microstructural characteristics) have been few. nm) was spatially filtered by a 0.5 X 2.0 cm 2 rectangular slit,

Recently, we reported4' - the LPVD growth of thin entered the reactor (transverse to gas flow) through oneepitaxial Ge films on [ 1001 GaAs for substrate temperatures quartz wall and, as illustrated in Fig. 1, passed -2.2 mmas low as 285 *C. By photodissociating GeH4 at 193 nm in above and parallel to the substrate. For the gas pressures ofparallel geometry, single-crystal Ge films 250-800 & thick these experiments, 2.2 mm is the equivalent of 10-60 meanwere grown at rates of typically 0.6-5 nm/min (- I ;, s free paths (MFPs) for a collision with a background GeH4max). One attractive aspect of these experiments is that molecule. Statements made throughout the paper relative totransient heating of the substrate and adlayer photolysis are the height of the laser beam above the substrate are refer-not involved in the growth process, thus making it possible enced to the bottom of the beam since care was taken toto isolate the gas phase photochemical processes. Normally, ensure that the ArF radiation did not illuminate the sub-resolving the contributions of bulk and adlayer heating, ad- strate. Also, the transmittance ofeach wall of the reactor wassorbate photochemistry and gas phase processes is problem- measured to be 62% ± 3% at 193 nm.atic. However, separating adlayer chemistry and other sur- Although focusing of the UV laser beam was not neces-face reactions from laser-induced gas phase reactions is sary, the 193 nm intensity in the reactor could readily beimportant for determining those conditions (T,, pressure, varied from -0. 1 to 1.3 MW cm - by diluting the gas mix-flow) under which epitaxial material can be grown by ture in the laser with He or by inserting quartz flats into theLPVD or of the potential impact of gas phase photochemical beam path. Unless stated otherwise, all of the films in theseprocessing on conventional growth technologies. Although experiments were grown at an initial (i.e., at the beginning ofthis approach does not involve a gaseous sensitizer and the the growth run) laser fluence of 15 mi cm (- 25 ns pulseexperiments are carried out in parallel geometry, epitaxial width (full width at half maximum-FWHM) ] at a pulsematerial is consistently obtained for T, in the range 285- repetition frequency of typically 40 Hz. It must be empha-410 *C and for low laser fluences ( 5 15 mJ cm -2 ). The most sized that this fluence is considerably smaller (by more thansignificant aspect of these results is that, in the absence of an order of magnitude) than those typical of parallel geome-external laser radiation, this reactor produces only amor- try LPVD growth experiments to date."' 5 52

3884 J. App. Phys.. Vol. 85. No. 10, 15 May 1989 Kiely, Tavitian, and Eden 3884

PIN Diode

[M 11/73Computer f-4.5 cm

FExcimer I IIn frared Ca "Pyrometer Cancitance

Quartz ReactorDownstream Flow

Suansducer endsdce andController

Ehust

M FlowTransducers and .

ControllersSu s r t

Heater BypassA S Ts <60*C)Mechanical

Pump

FIG. i. Partial schematic diagram of the experimental apparatus (not drawn to scale). Although only the quartz substrate is shown, quartz and GaAssubstrates were placed in the reactor for each growth run. The in situ measurement of the film thickness for the quartz substrate was generally (as shown) aHe-Ne laser transmission measurement while, for higher film growth rates, the interference produced by A 5 jm radiation propagating through thegrowing Ge film was recorded by the pyrometer. The outline of the excimer beam, propagating parallel to the substrate and coming out of the page, and itsposition above the substrate are also shown.

After exiting the reactor, the UV beam was deflected by Over the pressure range investigated in these experi-a prism onto a pyroelectric detector (Gen-Tec) or absorbing ments (5% GeH4 in He; PTOTAL --PG.H, +PH. =5 to 30calorimeter (Scientech) for measurements of the beam's Torr), the Ge film growth rate (Ro) was found to vary lin-pulse energy or average power. For a given growth run, two early with pressure. At T, = 300 "C, for example, R, is ob-substrates--one -0.5 X0.5 cm'SiO2 (quartz) and the other served to increase from -0.6 nm min - 't for PToAL = 5.60.5 X - 2.0cm2 Cr or Si-doped [ 100] GaAs-were placed in Torr to almost 5 nm/min-' at 30 Torr.the reactor and could be resistively heated (by an inconelheater held in a Macor glass ceramic block) to temperaturesas high as 870 K. Both substrates were situated at the bottomof the reactor with the GaAs substrate placed further fromthe laser entrance wall so as to minimize the possibility ofparticles produced by laser ablation of the wall (expected tobe minimal) from reaching the semiconductor surface. A G./Si,

- 1-mm-diam hole drilled in the heater allowed for a He-Ne T, -_ 410' Claser beam to pass through the quartz substrate (and subse- 55 C../.

quently detected by ap-i-n photodiode), thus allowing in situ z 30.5 tor,10scsm

measurements of film growth rates to be carried out in trans- 10-

mission. An infrared (IR) pyrometer (IRCON) monitored 'the substrate temperature by viewing either the quartz or EGaAs surfaces through a sapphire window sealed to the topof the reactor. When viewing the SiO substrate surface, the FIR pyrometer was optically aligned by the He-Ne laser beam Lin order to ensure that the pyrometer monitored the sameregion of the substrate that was sampled by the red laserprobe. The thickness of the Ge film on the SiO 2 substrate wasalso measured in real time in a second way; namely, by re- ......

cording the interference fringes produced by blackbody radi- 0 Soo 000 ,500 Z000 2500 3000

ation from the GaAs substrate propagating through the Time (seconds)

growing Ge film. A pyrometer output trace that is typical ofthose observed throughout these experiments is given in Fig. FIG. 2. Interference signal recorded by the IR pyrometer from a Ge film

2 for the CVD growth of Ge at T, = 410 *C. All data-He- grown by conventional pyrolysis at 7", = 410 *C. Each cycle in the pattern

Ne transmission, substrate temperatures, gas flow rate and corresponds to the growth of an additional 0.625 pm in film thickness. Inpressure, and ArF laser intensity-were stored by a DEC acquiring these data, the IRCON pyrometer monitored blackbody radi-

ation from the film/SiO. substrate in the A - 5/um wavelength region. Simi-[SI 11/73 computer. It should also be mentioned that the lar traces were observed for the epitaxial Ge films grown by LPVD at lower

in situ film thickness measurements were subsequently veri- temperatures ( 7", > 285 'C) but because of the much smaller film growth

fled by TEM observations, rates, the He-Ne transmission technique was more frequently used.

3665 J. Appl. Phys., Vol. 65, No. 10, 15 May 1989 Kiely, Tavitian, ar d Eden 3885

throughout the growth period which typically lasted 40-50min. Film growth rates varied from -6 to 50 A/min, de-pending upon T, and (as mentioned earlier) gas pressure.The gas pressure limits (5.6 and 30.4 Torr) correspond, re-spectively, to flow velocities of 40 and -6 cm s'.

C. Sample preparationPlan view samples of the LPVD grown Ge/GaAs inter-

faces were prepared for TEM studies by the method pre-viously described by Rackham.5' This involved chemicallybackthinning the sample (from the substrate side) with a0.5-mm-diam jet of chlorine in methanol until perforationoccurred. Large areas of electron transparent material couldthen be found around the periphery of the hole. Cross-sec-tional samples were prepared by the procedure outlined by

FIG. 3. Selected area electron diffraction pattern (SADP) of an LPVD Ge Bravman and Sinclair." The LPVD wafers were cut intofilm grown on [1001 GaAs at 305 C and a GeH, partial pressure of 1.5 Torr strips which were then glued together (with epoxy resin)(PrOTAL .30Torr). Thepattern ischaracteristicofGe ( 1001 single crystal. into blocks of four in such a way that the Ge face was ar-(Note the absence of g = {200} reflections.) The diffuse rings arise fromthe presence of an amorphous Ge overlayer. ranged to lie closest to the center of the block. Subsequently,

the blocks were cut perpendicular to the interfaces with adiamond saw and the resulting strips were mechanically pol-ished down to a thickness of - 40 pm. After being mountedonto Ni slot grids, the strips were ion milled to perforation by

B. Substrate cleaning procedure, gas flow cycle a 6 keV Ar' ion beam at a 16* angle of incidence. All samples

Immediately prior to being placed in the reactor, the were examined in a Phillips EM 430 transmission electron

GaAs substrates were chemically cleaned by the procedure microscope operating between 150 and 300 kV.

described by Anderson, Christou, and Davey." The quartzsubstrates were rinsed in spectrophotometric grade acetone Ill. RESULTS AND DISCUSSION

and methanol and blown dry with N, prior to being intro- More than 20 LPVD Ge/[ 1001 GaAs samples, grownduced into the reactor. In order to obtain epitaxial films, it at temperatures ranging from room temperature towas found to be necessary to dry etch the substrates at the -415 *C, have been examined to date by TEM. Moreover,growth temperature (240 5 T, 5 415 *C) by flowing a 5 % several films have been grown by conventional low-pressureHCI/He mixture (total pressure = 3.5 Torr, - 50 sccm flow chemical vapor deposition (LPCVD) or by LPVD but onrate) for 5 min. Subsequently, a 5% GeH4 in He mixture at a GaAs substrates of orientations other than [ 100 1. Severaltotal pressure of 5.6-30.4 Tort and 100 sccm mass flow rate specific growth regimes have been identified and this sectionpurged the reactor for -270s. The UV laser was then turned presents and discusses the results of TEM (and EDX) ob-on and pulsed at 40 Hz (peak intensity -0.75 MW cm- 2 ) servations from samples in each category. Although micro-

24000

Ge

Ge

FIG. 4. Energy dispersive x-ray (EDX)spectra of the Ge film (grown at T,= 305C) of Fig. 3.

j GoG~ As

0.0 13.0ENERSY (KV)

3N6 J. App1. Phys., Vol. 65, No. 10, 15 May 1989 Kiely, Tavitian. and Eden 3886

hence the reflection is kinematically forbidden. Consequent-ly, {200) reflections are absent from the pattern, Severaldiffuse rings centered around the undiffracted beam indicatethe existence of an amorphous layer in addition to the crys-talline film. As illustrated in Fig. 4, EDX analysis of the Geoverlayer region reveals characteristic Ge peaks with only aweak Ga and As background and, therefore, one suspectsthat the diffuse ring in the SADP in Fig. 3 also arises from Geand not an impurity.

An SADP of the Ge/GaAs bicrystal, presented in Fig.5, is a superposition of a Ge [ 100] diffraction pattern (iden-tical to that of Fig. 3), a GaAs [100] pattern (withg = {2001 reflections now present) and the weak diffusescattered rings from the amorphous Ge layer. The patternsof Figs. 3 and 5 demonstrate clearly that crystalline [1001Ge has been grown on GaAs.

FIG. 5. (1001 SADP of the Ge/GaAs bicrystal for the sample of Fig. 3. Cross-sectional TEM images of the Ge/GaAs interfacesNote that the g = {200} reflections characteristic of GaAs are now present. verify the assumption that the Ge film is initially crystalline

but subsequently switches to amorphous material. Figure 6is a cross-sectional micrograph of a - 600-,-thick amor-

graphs and diffraction patterns for a single film in a particu- phous layer atop a crystalline film whose thickness variesfrom - 250 to 3 50 k , It is also obvious that the GaAs sub-

lar category will generally be shown, the comments and datasummarize observations made over a number of samples. strate surface is somewhat wavy (-.20-A undulations) but

there is no evidence for a gallium oxide layer at the interface.

A. LPVD Go films (285< , < 330 *C) Therefore, although the in situ HCI dry etch is apparentlyeffective in removing oxides from the substrate, it by no

1. Mlcrostructure of ifms means leaves the surface atomically smooth. Several inclined

Figure 3 ,iiows a selected area diffraction pattern growth defects in the Ge layer, which appear to propagate(SADP), acquired along the substrate normal direction, of a from the Ge/GaAs interface, are also evident in Fig. 6.Ge overlayer that is representative of those observed for all Convergent beam electron diffraction patterns of thethe LPVD films grown in the 285-330 "C temperature range. crystalline Ge layer and the GaAs substrate (near the inter-The film of Fig. 3 was grown at 305 *C (PGH,/H, = 30.2 face) are shown in Figs. 7 and 8, respectively. Both diffrac-Torr) and the SADP pattern is that for Ge [ 1001 single tion patterns were acquired by viewing the crystalline re-crystal. Since germanium has the diamond structure (FD 3m gions along the [0111 GaAs direction with a 150-A-diamspace group), the structure factor for g = (200} is zero and electron beam probe. As expected. the g = {200} disks are

FIG. 6. Cross- ectional micrograph of the Ge/GaAs interface (from the sar. -owth run as the %ample of Figs 3-5) showing the 250-350-A-thick crystal-line film and a 600-A amorphous layer. The rough topography of the crystalline film ts partially attributed to a switch from crystalline to amorphous materialfollowing the coalescence of epitaxial Ge islands. A few inclined growth defects i denoted d) are apparent in the Ge film as is an occasional soid (denoted v)but the interface is essentially free of gallium oxide.

3887 J. Apol. Phys.. Voi. 65, No. 10, IS May 1989 xiely Tavitian, and Eden 3887

is afforded by lattice images of the crystalline Ge/GaAs in-terface. Also taken along [011 ] GaAs at 300 kV, the imageof Fig. 9, for example, displays two sets of {I I l} latticefringes (with a characteristic spacing of 3.26 k) which are

inclined at 54.7* with respect to the interfacial plane. Thesefringes run continuously across the interface from the GaAs

substrate into the Ge layer, thus illustrating the epitaxy be-tween the two crystalline materials.

All of the Ge films grown on [ 100] GaAs and examinedto date by cro s-sectional TEM exhibit the rough morpholo-gy evident in Fig. 6 which suggests that the Ge initially nu-cleates on the substrate as epitaxial islands which grow insize and eventually coalesce to form a film. Soon after coales-cence, however, the material switches from crystalline toamorphous. As a result, the thickness of the epitaxial filmshas thus far been limited to - 250-800 A. Consequently, theLPVD growth of Ge films in this temperature range is char-acteristic of the Volmer-Weber process rather than theFrank-van de Merwe or Stranski-Krastanov mechanisms.5 6

Not only does the cluster (island) nucleation process ac-FIG. 7. (0ill Convergent beam electron diffraction pattern of a Ge film count for the observed rough morphology of the filmsgrown by LPVD at 305 *C. Acquired with a I,-A-diam electron beam by (owing to the differing growth histories of the individualviewing the crystalline layer along the (0111 GaAs direction, this diffrac- islands) but may also explain the small amorphous zonestion pattern is virtually identical with those for other photochemicallygrown films in the 285-330 T range. Note the absence of g = {200) disks, that are occasionally seen between Ge islands at the Ge/

GaAs interface (one such zone or void, denoted v, is indicat-ed in Fig. 6). A high magnification plan view micrograph ofthe Ge layer (, = 305 "C), taken under strongly diffracting

absent for Ge (cf. Fig. 7) and the fact that no symmetry conditions, is shown in Fig. 10. The mottled appearance ofinformation is visible within the disks is a consequence of the the film arises primarily from the variations in crystalline Gesample being only a few hundred A thick. These patterns (as overlayer film thickness that are apparent in Fig. 6.well as those taken on plan view specimens) demonstrate Figures 3-9 are quite representative of the characteris-that the thin Ge films are not only crystalline but have the tics of all of the LPVD films studied to date by TEM. Epitax-following epitaxial relationship with the substrate: ial Ge films can be reproducibly grown on GaAs [ 100] for

GeCl1001//GaAs( 100]; Ge[010]//GaAs(010]. T, in the range of 285-330*C. Upon increasing T, from

films -300 to 330 "C, the thickness of the epi Ge layer increasesslightly ( - 20%). Also, no obvious improvement in the filmmorphology or structure was observed if the film was slowlyetched during the growth run (by adding HCI to the GeH 4/He flow). Rather, the morphology remained unchanged and

Anm

FIG. 9. Lattice image of the Ge/GaAs interface (sample of Figs. 3-8) takenFIG. 8. [0111 convergent beam electron diffraction pattern of the GaAs along 10111 GaAs. The continuity of the two sets of { I l I}I lattice fnngesregion near the Ge/GaAs interface for the sample of Fig. 7. across the interface demonstrates the epitaxy between the two crystas.

3666 J. Appl. Phys., Vol. 65, No. 10. 15 May 1989 Kiely, Tavitian, and Eden 3888

where M is any background species (such as He)." Regard-less of whether GeH3 or the germylene radical GeH, areproduced initially, several arguments point to GeH, as thespecies most likely responsible for film growth."5 Althoughlaser spectroscopic experiments will be necessary to fullyresolve this issue, it is interesting to note that Taguchi andco-workers47 have inferred the photochemical production ofSiH, from SiH 4 (by an ArF laser) and its role in accelerat-ing the growth rate of hydrogenated amorphous Si films.

Recent studies of the CVD of GeH,/GeH, mixtures byGow et al.59 demonstrated that the introduction of eventrace amounts of digermane (0.2% GeH/4.8% GeH, inbuffer) into the reactor dramatically reduces the tempera-ture at which pyrolysis can occur. Experiments with pureGe,H 6 revealed that pyrolysis occurred on oxidized Si sub-strates at temperatures < 200 "C which is well below thelowest tempeature (285 *C) at which we have grown epitax-

FIG. 10. Plan view micrograph of a Ge surface grown at 7" = 305 C. The ial Ge films by LPVD. The pyrolytic temperature for mono-mottled appearance of the film arises primarily from the variations in Ge germane, on the other hand, is - 280 *C (Ref. 60).film thickness that are also apparent in Fig. 6. Further support for the pivotal role of GeH 6 in the

LPVD film growth process is suggested by the recent workof Gates6' who investigated the chemisorption of the (analo-gous) silanes on Si(1l)-(7X7) by temperature pro-

the major difference was that the thickness of the crystalline grammed desorption. The results showed a large enhance-layer was considerably thinner ( <200 A), owing tc the ment in the adsorption rate for Si2H6 as opposed to that forcompetition of growth and etching processes. Furthermore, monosilane. 62 Although this enhancement was attributed tothe quality of the epitaxial Ge film actually deteriorates the presence of the Si-Si bond in disilane, a second Si-Si

somewhat if HC is continuously flowed through the reactor bond (for Si3H8 ) leads to a much less dramatic improve-witl the GeH4/ri Ie mixture. This conclusion is based on the ment in the adsorption rate. It was concluded that the mono-observation of misoriented domains of Ge that are apparent silane "...is relatively unreactive due to its short residencein samples grown under these conditions. However, dry time and because adsorption requires Si-H bond scis-chemical etching of the substrate (in situ) prior to film sion." " Moreover, Perrin and Allain63 have also recentlygrowth was found to be essential for obtaining epitaxial lay- demonstrated that, in the presence of UV radiation, GeH 3ers. Failure to do this consistently yielded amorphous films. radicals will form GeH, at the surface.

All of the available evidence, therefore, points to Ge2 H6as the critical species in these LPVD experiments. Although

It is not likely that the crystalline-to-amorphous switch not directly produced by the excimer laser, the molecule isin the film structure arises from the decrease in excimer laser apparently formed in the gas phase (or, for low GeH4 partialbeam intensity that occurs as the growth run proceeds owing pressures, at the surface) by the recombination of germyleneto the formation of a thin Ge film on the quartz wall where radicals. The photochemical production of digermane notthe beam enters the reactor. After 600 s of growth time, the only accelerates the film growth rate (owing to a presumedlaser fluence entering the reactor was measured to have fal- higher adsorption rate for Ge,H6 as opposed to that forlen only 12% + 1% from its initial value. Rather, it appears GeH4 ) but also allows for the growth of epitaxial films atthat the transition from epitaxial to amorphous material can lower T, since digermane pyrolyzes at lower temperaturesbe attributed to one of three explanations which immediately than does monogermane.present themselves. Each argument implicitly assumes that Let us now turn our attention to the Ge film growththe species produced (indirectly) by the UV laser is diger- process itself and, in particular, to the epitaxial-to-amor-mane. phous transition that is observed. The first plausible view is

The peak laser intensities in these experiments have in- that, if the GaAs [ 1001 surface is more catalytically activetentionally been maintained below 1 MW cm in order to (in pyrolyzing GeH,) than is Ge, then once the substrate isminimize the impact of two photon absorption processes on covered by the coalescing islands, epitaxial film growth willthe gas phase photochemistry occurring in the reactor. stop. The second possibility revolves around the limited mo-Hence, extensive photofragmentation of GeH4 (beyond bility of Ge adatoms that is an inherent characteristic of theGeH, or. perhaps, GeH,) is considered to be unlikely. Single low surface temperatures of these experiments. If these ada-photon dissociation of GeH4 at 193 nm is expected to pre- toms have a higher mobility on GaAs [ 100] or Ge [ 100]dominantly yield a germyl radical (GeH,) which rapidly than they do on other surface orientations [such as GaAsforms Ge:H, by the reaction: ( I ll), for example], film thickness may be limited by the

fact that the growth of individual crystalline islands hingesGeH, + GeH, -4 M-Ge,H, + M, (I) on the migration of Ge adatoms over [ I ll] surfaces.

3889 J Appi Phys. Vol 65, No 10. 15 May 1989 Kiely. Tavitian. and Eden 3889

- - ,-r- --.' ."",- .. -- . .. -- -

.. ., FIG. 11. TEM lattice image ofa thin epitax-ial Ge film grown on a tilted GaAs 1100]substrate (offcut 3* toward [1101) atT = 320 C. Note the relatively smooth

morphology of the epilayer as compared tothe film of Fig. 6.

In order to explore these possibilities further, several Ge crystalline (rather than amorphous) material is observed.layers were grown by LPVD at 280< T, <320 "C on sub- The microstructure of a Ge/GaAs [ 100] sample grownstrates of various orientations. Films grown at 320 *C on at 400 *C can be seen in the cross-sectional micrograph of the[ 100] substrates which were offcut 3* towards [ 110] once interface (Fig. 12). Immediately adjacent to the interface isagain were initially epitaxial but the crystalline-to-amor- a layer of epitaxial Ge having a thickness of - 800 ,. On topphous transition occurred sooner (at tz 120 k.). Also, as is of this is a region, -0.9 ym thick, consisting of heavily mi-apparent in the cross-sectional TEM micrograph (lattice crotwinned, columnar-type polycrystals with typical grainimage) of Fig. 11, the epilayers were considerably smoother diameters of 0.25 1um and lengths of -0.5 .um.than those discussed earlier (on untilted [ 1001 substrates). Figures 13 and 14 are [O I] convergent beam patternsBoth of these results are a consequence of the higher density of the GaAs substrate and the single-crystal Ge layer (sam-of surface steps that are available with the tilted surface ori-entation. Since each of these steps is a potential nucleationsite, more islands are presumably formed initially and co-alescence occurs sooner due to the reduced lateral growthrequired of an island.

Several LPVD Ge layers were grown on [ 1111 orientedsubstrates and all were completely amorphous. Under noconditions (varying T, up to 350 *C, etc.) was a crystallinefilm observed. One tentatively concludes that adatom mobil-ity is, indeed, a crucial factor in determining the thickness ofthe LPVD grown Ge films. The development of nonthermalapproaches to increasing Ge atom surface mobilities willlikely remove the thickness limit encountered and allow forcontinuous epitaxial Ge films to be grown on GaAs at evenlower temperatures.

Recent analysis of these films by SIMS suggests that an V

additional (and perhaps critical) factor in the transition toamorphous material is the gradual buildup of carbon in theepitaxial film as time proceeds. It is quite possible that, sincethe solubility of C in c-Ge is low, the carbon concentration atthe Ge/gas phase growth front will rise as the front movesout from the substrate surface. Eventually, the C number . . "density will exceed a critical level beyond which only a-Gewill grow. Studies to explore this possibility further with animproved vacuum system are continuing.

B. LPVD grown films: Tr, Z 400 'CAll of the Ge films grown on GaAs [100) in the -240-

280 'C range by LPVD are amorphous in character. At high-er temperatures ( T Z 400 °C), on the other hand, epilayers FIG. 12. Cross-.ecttonal TEM micrograph of a LPVD Ge/GaAs 1001-800 A thick are grown. Subsequently. a transition to poly- interface grown at 400 'C

3890 J AopI PhyS., VOI 65, No, 10. 15 May 1989 Kiely, Tavitian, and Eden 3890

FIG. 15. Plan view SADP of the polycrystalline layer in the LPVD grownFIG. 13. [0aI convergent beam diffraction pattern of the GaAs substrate sample of Figs. 12-14.forthesampleof-Fig. 12 =400C).

pie of Fig. 12), respectively, and illustrate the epitaxy be- crystalline layer is strongly in contrast in this imaging condi-tween the two crystals. As expected, the Ge pattern again tion which provides further confirmation of the epitaxial na-reveals no f{200}-type reflections. An SADP of the polycrys- ture of the thin interfacial Ge film. It is worth noting, how-talline layer, given in Fig. 15, is characteristic of a textured ever, that the top surface of the epitaxial layer again has apolycrystalline morphology and the obvious streaking of the fairly rough topography and inclined defects are frequentlydiffraction pattern is caused by the microtwinning within observed within the film.individual columnar grains. Before leaving this section, it should be mentioned that

A g = {I Il I} dark field image of the Ge/GaAs interface Baert et aI." have recently observed thin ( -200 A) epitax-for the sample of Figs. 12-15, is shown in Fig. 16. The entire ial Si films at the interface between [ 100] Si substrates and

hydrogenated amorphous silicon. In their experiments,films were grown at 250 *C by dissociating SiH, in a radiofrequency discharge. Although the transition from epitaxialto amorphous films was not discussed, the presence of thecrystalline layer was attributed to cleaning of the substrate

(chemically with HF and in situ by species produced in thedischarge).

C. Ge flms grown by LPCVD (285< T, :410 *C)

In the absence of the external UV radiation (i.e., con-ventional low-pressure CVD), the Ge films grown in the

285-330 fC range are completely amorphous while thosegrown at higher temperatures 4 410 'C) are heavily defect-ed polycrystalline.

Figure 17 is a plan view micrograph of an amorphousGe overlaver grown on [ 100 1 GaAs by LPCVD at T,

= 305 *C. Recall that, at the edge of the hole in the TEMspecimen. the Ge film overhangs the GaAs substrate. Notethat in this region no band contours can be seen which is anearmark of amorphous material. Such contours are clearlyevident in the GaAs substrate but end abruptly at the a-GelGaAs interface. A SADP of this layer, displayed in Fig. 18.exhibits only the diffuse ings expected for amorphous mate-

FIG. 14, 10111 convergent beam electron diffraction pattern of the - rial. Across-sectional view of thissampleisshown in Fig. 19.-thick epitaxial Ge fim for the sample of Figs. 12 and 13. Although this film is - 1000 A thick, in contrast to the epi-

3691 .2. Appl. Prys., Vol, 65, No. 10. 15 May 1989 Kiely, Tavtian. and Eden 3891

FIG. 16. Dark field image (g= {litI) of theGe/GaAs interface ( , 400 =C).

taxial films discussed earlier, it has a smooth surface and {111} type and are being viewed edge-on. The lamellaeexhibits no diffraction contrast. thickness of the microtwins are, in some cases, as small as a

Ge films grown pyrolytically on GaAs in the same reac- few Ge unit cells (i.e., down to - 17 A). Dislocation struc-tor but at higher temperatures (T, Z 400 °C) are polycrys- ture is also frequently observed within individual Ge grainstalline. A plan view micrograph of a Ge layer grown at as well as at the grain boundaries. A SADP of a plan viewT, = 415 °C is given in Fig. 20. The grains are typically 0.4 specimen ( T. again = 415 °C) is shown in Fig. 22. Acquiredpm in diameter along this projection and exhibit no obvious with a 20-14m selected area aperture, this pattern verifies thattexturing (nor do the boundaries between individual grains the film is indeed polycrystalline and the diffraction ringsexhibit well-defined facetting). Furthermore, as illustrated can be indexed (working out from the center) as corre-in Fig. 21, most ofthe microcrystallites are heavily defected. sponding to {11},{220}, {311}, {400}, {331}, {442}, ... ,This micrograph shows several Ge grains from the same plane spacings, respectively. The obvious streaking of somesample imaged in Fig. 20 that are viewed along Ge [011]. reflections is attributable to the high density of microtwinsThe two equivalent types of twin planes that are observed are discussed earlier.

A cross-sectional view of this textured polycrystallinesample, taken along GaAs [011], is given in Fig. 23. Thetotal thickness of the Ge layer is - 1.9 pm and the film is seen

FIG. 17. Plan view micrograph of an amorphous Ge film grown at - 305 Cby conventional low pressure chemical vapor deposition (i.e., in the absenceof parallel UV radiation). FIG. 18. SADP of the Ge overlayer of Fig. 17 (7, = 305 T, I,,, = 0).

3892 J. Api. Phys., Vol. 85, No. 10, 15 May 1989 Kiely, Tavitian, and Eden 3892

e1.

FIG. 19. Cross-sectional view of the - 1000-A-thick amorphous Ge film ofFigs. 17 and I8 ( T, = 305 *C).

FIG. 20. Plan view micrograph of a polycrystalline Ge layer grown on[ 1001 GaAs at 7" = 415 *C in the absence of (parallel) excimer laser radi-ation.

to be composed of three distinct and separate regions. Adja-cent to the substrate is a layer, -950 A in thickness, whichconsists of fine-grained, randomly oriented polycrystals ofGe. The intermediate region is -0.7 um in thickness and iscomposed of columnar-shaped Ge grains. The outer layerconsists of large columnar grains, typically 0.4,um in diame- geometry for T, ranging from room temperature to 415 *C.ter and - 1.1/ im in length, that are heavily microtwinned. Resorting to parallel geometry prevents transient heating of

the substrate and adlayer photolytic processes.D. General comments For 285 < T, 5 330 *C, the Ge film initially grows epitax-

ially but, after reaching a thickness of 250-700 A (depend-As mentioned previously, epitaxial Ge films have been ing on gas pressure and flow, T, and the etching procedure),

reproducibly grown on [ 1001 GaAs for substrate tempera- switches to amorphous material. At higher substrate tem-tures as low as 285 *C. More than a half-dozen films have peratures ( T. Z 400 *C), the epitaxial Ge film grows thickerbeen grown in the 285-320 'C region and examined by TEM ( . 800 A) before a transition to apolycrystalline layer takesand all exhibit the same characteristics as those for the film place.of Figs. 3-1I. That is, the Ge film initially grows epitaxially pae

an, ftr eahig ticnes f250-700 Acnet o One of the most interesting aspects of these experimentsand, after reaching a thickness of A , converts to is that, over the temperature range investigated, film growthamorphous material. These results represent a decrease of does occur in the absence of UV laser radiation. Since themore than 100 'C over the lowest temperature in the litera- pyrolytic temperature for GeH4 is - 280 °C, Ge films grownture at which epitaxial Ge films have been grown on GaAs. by LPCVD (i.e., with the laser blocked) are amorphous forKrautle and co-workers6 " have reported the epitaxial 285<T, : 330°C but, at the higher temperaturesgrowth of Ge on GaAs by CVD at temperatures as low as (7', Z 400 'C), heavily defected polycrystalline material is400 'C (in contrast to our polycrystalline results) but, even obtained.at this temperature, the Ge film is heavily auto-doped. The reduction in the epitaxial growth temperature madeTherefore, the present experiments demonstrate that the possible by the UV laser is attributed to the photochemicalnonequilibrium chemical environment produced by the ul- podion of derVase attrbute in the eatoSeical

traviolet laser termits the growth of epitaxial Ge films at ly, GeH 3 radicals produced from monogermane rapidly

temperatures at least 100 'C lower than those required for form GeH (by three body collisio onnce outside the

conventional CVD. The lower growth temperatures should beam or at the substrate. Digermane has previously been

considerably reduce interdiffusion at the Ge/GaAs inter- identified as one of the few (collisionally) stable photoche-

face.mical products of the excimer laser photodissociation ofGeH4 and is known to pyrolyze at temperatures below200 'C (Refs. 57 and 59).

It is apparent, then, that the laser directly produces aIV. SUMMARY AND CONCLUSIONS species (radical) which is collisionally transformed en route

Ge films grown on 11001 GaAs by LPVD have been to the substrate into a more stable species. Upon reaching theexamined in detail by transmission electron microscopy. GaAs [ 1001 surface, this species (Ge,H,) is pyrolyzed, re-The LPVD growth process consisted of photodissociating suiting in the growth of a Ge epilayer. The ability to growGeH, at 193 nm (ArF laser, (b - 15 mJ cm -) in parallel epitaxial Ge layers at such low temperatures is critically de-

3893 J AppI. P ys., Vol. 65. No. 10, 15 May 1989 Kiely, Tavitian, and Eden 3893

FIG 2 1. Ge grains grown on GaAs 11001 atT, = 415 °C and viewed along [0111, show-ing distinct microtwinning.

pendent upon both the surface orientation and cleaning pro- The results presented are interesting in that we arecedure. For the present apparatus, in situ etching of the aware of no other example in which photons have altered theGaAs and the use of [ 1001 (or slightly oficut [ 100]) sub- gas phase (and subsequently, the surface) chemistry of astrates are essential for obtaining crystalline films. The data reactor which, in the absence of external radiation, producesobtained to date indicate that the mobility of Ge adatoms on amorphous or polycrystalline material. The laser initiatedGaAs[ 1I11 is smaller than that for [ 100] surfaces. From chemistry, however, culminates in the growth of epitaxialthe results reported in Ref. 57, one would expect that, if films.adatom mobility were not a problem, higher ArF laserfluences (and, perhaps, lower pressures) would allow for epiGe films to be grown at temperatures below 200 C. That is,at the higher 193-nm intensities, the peak Ge number density(produced directly from GeH4 by multiphoton absorption)rises to sufficiently high levels to allow monolayer s-'growth rates without the need to pyrolyze polyatomics(higher germanes) such as GeH,.

L1.

FIG. 22. Plan view SADP of a polycrystalline Ge film (T =415'C) The FIG. 23. Cross-sectional micrograph. taken along GaAs I OT 11. of the poly-

streaking of the diffraction pattern is attributable to the high mtcrotwin crystallinesampleofFigs. 20-22. The total thicknessoftheGelayeris -1.9

density in the sample. 'Um.

3894 J. Appl. PhyS., Vol, 65. No. 10. 15 May 1989 Kiely, Tavitian, and Eden 3894

ACKNOWLEDGMENTS 2"S. J. C. Irvine and J. B. Mullin, J. Vac. Sci. Technol. A 5, 2100 (1987).29IS. 3. C. Irvine, J. B. Mullin. H. Hill. G. T. Brown, and S.)J. Barnet, ..

Discussions with and suggestions from S. D. Allen, S. Cryst. Growth 86. 188 (1988).M. Bedair, D. B. Geohegan, 1. P. Herman, R. J. von Gutfeld, "VW. L. Ahigren, E. J1. Smith, J. B. James, T. W. James, R. P. Ruth, E. A.R. 1. Masel, and S. A. Piette are gratefully acknowledged. Patten, R. D. Knox, and J.-L. Staudenmann. 1. Cryst. Growth 86, 198

Also, the technical assistance of E. J. M. Colbert, T. R. Dill, (1988).11B. J. Morris, App!. Phys. Lett. 48. 867 (1986).

M. N. Ediger, K. K. King, M. Ross, P. Griffis, D. Terven, '2J. l. Zinck, P. D. Brewer,). E. Jensen. G. L. Olson, and L. W. Tutt, App.and K. Voyles is appreciated. This work was supported by Phys. Lett. 52. 1434 (1988); Mater. Res. Soc. Symp. Proc. 101, 319the Air Force Office of Scientific Research (H. Schlossberg) (1988).

under Contract No. F49620-85-C-O 141, the National "J. Nishizawa, Y. Kokubun, H. Shirmawaki, and M. Koike.)J. Electrochem.Soc. 132.,1939 (1985).

Science Foundation (through the University ao' Illinois Ma- 'N. Piitz. H. Heinecke, E. Veuhoff. G. Arens, M. Heyen, H. Liith, and P.terials Research Laboratory) under grant DMR 86-12860, Balk,)1. Cryst. Growth 68, 194 (1984).

and the Eaton Corporation. The electron microscopy was "S. Fujita. A. Tanabe. T. Sakamoto, M. Isemura, and S. Ftzjita, Jpn. J.carried out in the Center for Microanalysis of Materials at A~pl. Phys. 26, L2000 ((987).

'.A. Doi, Y. Aoyag, and S. Namba, Mater. Res. Soc. Symp. Proc. 75, 217the University of Illinois which is supported by the U. S. (1987); App!. Phys. Lett. 49, 785 (1986).Department of Energy under Contract No. DE-AC-02- "N. H. Karam, N. A. EI-Masry, and S. M. Bedair, App!. Phys. Lett. 49, 880

76ER 0 1198. (1986); S. M. Bedair, J. K. Whisnant, N. H. Karam, D. Griffis, N. A. El-Masry, and H. H. Stadelmaitr, J. Cryst. Growth 77, 229 (1986).

"'Y. Aoyagi, M. Kanazawa, A. Doi, S. Iwai, and S. Namba, J. App. Phys.60,3131 (1986).t 'For a review of early work in the field, see R. MI. Osgood. Jr., Ann. Rev. "H.j Kukimoto, Y. Ban, H. Komatsu, MI. Takechi, and M. tshizaki. J.

Phys. Chem. 34, 77 (1983). Cryst. Growth, 223 (1986).IF. A. Houle, App. Phys. A 41, 315 (1986). 40Y. Ban. MI. Takechi, M. Ishizaki, H. Murata, and H. Kukimoto, Jpn. J.3R. W. Andreatta, C. C. Abele, 3. F. Osmundsen, J. G3. Eden. D. Lubben. App!. Phys. 25. L967 (1986).and)J. E. Greene, App!. Phys. Lett. 40. 183 (1982). "R. N. Bicknell, N. C. Giles, and)J. F. Schetzina. Appl. Phys. Lett, 49, 1735"I. G. Eden, J. &. Greene, J. F. Osmundsen, D. Lubben. C. C. Abele, S. (1986).Gorbackin, and H. D. Desa,, Mater. Res. Soc. Symp. Proc. 17, 185 (1983). 42R. Solanki, U. Sudarsan, and J. C. Johnson, Appl. Phys. Lett. 52, 919'H. Niki and 0. 1. Mains,)1. Phys. Chem. 68. 304 (1964). (1988)."T. Saitob, S. Murmactsu, T. Shimada. and M. Migitaka, App!. Phys. Lett. 4'P. K. Boyer, (3. A. Roche, W. H. Ritchie, and G3. J. Collins. App!. Phys.42, 678 (1983). Lett. 40.,716 (1982); R. Solanki. W. H . Ritchie, and G.)J. Collins, App!.

'T. Inoue, MI. Konagai, and K. Takahashi, App!. Phys. Lett. 43, 774 Phys. Lett. 43, 454 (1983).(1983). "'P. K. Bayer. C. A. Moore, R. Solanki. W. K. Ritchie, G3. A. Roche, and G3.'S. Nishida. H. Tasaki. M. Konagai, and K. Takahashi.)J. Appi. Phys. 58, J. Collins. Mater. Res. Soc. Symp. Proc. 17, 119 (1983).1427 (1985). 3P. KC. Boyer, K. A. Emery, H. Zarnani, and (3. J . Collins, App!. Phys.

'S. Nishida, T. Shjimoto, A. Yamada, S. Karasawa. MI. Konagai, and K. Lett. 45, 979 (1984).Takahashi, App!. Phys. Lett. 49. 79 (1986). "T'. F. Deutsch, D. 1. Silversmith, and R. W. Mountain, Mater. Res. Soc.

"'1Y. Mishima, MI. Hirose. Y. Osaka, K. Nagamine, Y. Ashida, N. Ki- Symp. Proc. 17. 129 (1983).tagawa. and K. (sogaya. Jpn. 3. App!. Phy3. 22, L46 (1983). 47T. Taguchi, M,. Morikawa, Y. Hiratsuka, and K. Toyoda, App!. Phys.

"Y. Mishinia, Y. Ashida, and M. Hirose.)1. Non-Cryst. Solids 59&60, 70 Lett. 49, 971 (1986).(1983). "V. Tavitian. C.)J. Kiely, and)J. 0. Eden, Mater. Res, Soc. Synip. Proc. 101,

"K. Kumata, U. Itoh. Y. Toyoshima. N. Tanaka, H. Anzai, and A. Mat- 349 (1988).suda, Appl. Phys. Lett. 48. 1380 (1986). ~"J. G3. Eden, V. Tavitian, and C,.J. Kiely, Proc. SPIE 894, 26 (1988).

"N. Gonohe, S. Shimizu, K. Tamagawa. T. Hayashi, and H. Yamakawa 5"V. Tavitian. C. J. Kiely, D. B. Geohegan, and)J. G3. Eden, App!. Phys.Jpn. J. App!. Phys. 26, L 1189 (1987). Lett. 52. 1710 (1988).

"'A. Yoshikawa and S. Yamnaga. Jpn. 1. AppI. Phys. 23, L473 (1984). "VCR isa trademark of the Cajon Company, 9760Shepard Road, Macedo-"Y. Toyashima. K. Kumata. U. Itoh, and A. Matsuda. App!. Phys. Lett. nia, OH 44056.

51.,1925 (1987). 34 . Zinck, P. D. Brewer,.). E. Jensen. G3. L. Olson. and L. W. Tutt, Mater."A. Ishitani, M. Kanamon. and H. Tsuya. J. App!. Phys. 57,2956 (1985). Res. Soc. Symp. Proc. 75,.233 (1987)."A. Ishitani and S. Itoh, Paper D 3.6, Materials Research Society Fall "AW. T. Anderson, Jr.. A. Christou. and J. E. Davey, J1. Appi. Phys. 49,2998

Meeting. Boston. MA (1985). (1978)."'A. (shitani, Y. Ohshita, K. Tanigaki. K. Takada. and S. Itoh, J. App!. "G. M. Rackham, Ph.D. thesis. University of Bristol (1976).

Phys. 61. 2224 (1987). "3. C. Bravmnan and R. Sinclair.)J. Electron Microsc. Tech. 1, 53 (1984)."V. M. Donnelly. XI Geva. J. Long, and R. F. Karlicek, App!. Phys. Lett. "See, for example, 1. W. Boyd. Laser Processing of Thin Films and Micro-4.951 (1984). structures (Springer. Berlin. 1987), pp. 85-88 and references cited there-

'V. MI. Donnelly, D. Brasen, A. Appelbaum, and NI Geva.)1. App!. Phys. in.So. 2022 (1985). "K. K. King, V. Tavitian, 0. B. Geohegan. E. A. P. Cheng, S. A. Piette, F.V. MI. Donnelly. V. R. MlcCrary, A. Appelbaum, D. Brasen. and W. P. J. Scheltens, and)J. G3. Eden. Mater. Res. Soc. Symp. Proc. 75, 19(1987).Lowe.)3. App!. Phys. 61,1(410 (1987). "J. F. Osniundsen. C. C. Abele, and J. (3. Eden.)J. App!. Phys. 57, 2921

2'V. R. McCrary, V MI. Donnelly. D. Brasen. A. Appelbaum, and R. C. (1985).Farrow, Mater. Res. Soc. Symp. Proc. 75. 223 (1987). "T. R. Gow. D. G3. Coronetl. and R. [.Masel, 1. Mater. Res. (to be pub-

"S. S. Chu, T. L. Chu, C. L. Chang, and H. Firouzi. Appl. Phys. Lett. 52. lished).1243 (1988). "L. H. Hall,.J. Electrochem. Soc. 119.1(593 (1972).

2"C. W. Ta, V. MI. Donnelly,)J. C. Beggy, F. A. Baiocchi. V. R. McCrary, T. "S. MI. Gates, Surf. Sci. (in press).D. Harris. and MI. G3. Lamont, App!. Phys. Lett. 52. 966 (1988). 'The Si,H. adsorption rate was measured to be roughly an order of magni-V NI. Donnelly. C. W. Tu. J. C. Beggy. V. R. McCrary,. MG. Lamont. T. tude larger than that for SiH. at an exposure of - 10" molecues cm'-.D. Harris. F. A. Buiocchi. and R. C. Farrow. App!. Phys. Lett. 52, 1065 The diff'erence appears to be even greater for smaller surface exposures.(1988). "J3. Perrin and B. Attain, J. Non-Cryst. Solids 97&98, 261 (1987).

2'S. 1. C. Irvine,)J. B. Mullin and)1. Tunniclifre. J. Cryst. Growth 68, 188 "'K. Bert, J. Symons, W. Vandervorsti . Vanhellemont. I. Caymax. J.(1984). Poortmans. 1. Nijs. and R. Mertens, App!. Phys. Lett. SI, 1922 (1987).

"~S. 1. C. Irvine.). Giess.). B. Mullin. (3. W. Black more, and 0. D. Dosser, 'H. Kraiutle, P. Roentgen. and H. Beneking. J. Cryst. Growth 65. 4393. Vat. Sci. Technol. 8B3. 1450 (1985). (1983).

3896 J. Aopl. Ptiys.. Vol. 85. No. 10, IS May 1989 Kiely, Tavitian. and Eden

n 1i a[ogr*tr* · 2. the gerade rydberg states of ne2 have been observed by laser excitation...

Documents

n 1i a[ogrtr · 2. the gerade rydberg states of ne2 have been observed by laser excitation...