AD-A257 361

Defense Nuclear AgencyAlexandria, VA 22310-3398

DNA-TR-92-12

Effects of Processing on MOS Radiation Hardening

Dr. Ralph J. JaccodineDr. Donald R. YoungLehigh UniversityOffice of Research and Sponsored Programs DT(C526 Brodhead Avenue DBethlehem, PA 18015 $ L C TE 1• EP 3 0.1991I •

September 1992

Technical Report

CONTRACT No. DNA 001-88-C-0122

Approved for public release;distribution Is unlimited.

92-26120i2iqi0i17fl111U'

Destroy this report when it is no longer needed. Do notreturn to sender.

PLEASE NOTIFY THE DEFENSE NUCLEAR AGENCY,ATrN: CSTI, 6801 TELEGRAPH ROAD. ALEXANDRIA, VA22310-3398. IF YOUR ADDRESS IS INCORRECT, IF YOUWISH IT DELETED FROM THE DISTRIBUTION LIST, ORIFTHEADORESSEE IS NO LONGER EMPLOYED BY YOURORGANIZATION.

0u

Form ApprovedREPORT DOCUMENTATION PAGE OMB No. 0704-0188

Pulic r f burdlen for this COileesue ofi nfornisheon is setnam to ave0I oge I or beerM rSemesS. ciuOimg tris time for rnivtniri i rnsructions. searcritig existlng own sources.gapmow" a m intsmgmno the dean needed and conoeg and r mevieinM iMe Colc"if1 of infOrneton. Send coninenet regavitng win burden *etmatie or any cover s oect at mscolteon ot infontmln. naudg suue•oeaons for recingn mits burden. to Weashgton HeeA& uaefl e Services. Orectortss for into , entin Ooersrions am Reports. i215 jeffersonOawis Hqhnfwv. Suite 1204. Allngto . VA 22202.4302. am to me Office of Management and Budget. Paseriwork Reduction Prmec 10704-01681. Wasnington. OC 20503

1. AGENCY USE ONLY (Leeve Wlonc) 2. REPORT DATE 3. REPORT TYPE AND DATES COVERED920901 Technical 880827 - 920817

4. TITLE AND SUBTITLE 5. FUNDING NUMBERSC - DNA 001-88-C-0122

Effects of Processing on MOS Radiation Hardening PE - 62715HPR - RV

6. AUTHOR(S) TA - RA

Dr. Ralph J. Jaccodine and Dr. Donald R. Young WU - DH880122

7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES). 8. PERFORMING ORGANIZATIONREPORT NUMBER

Lehigh UniversityOffice of Research and Sponsored Programs526 Brodhead AvenueBethlehem, PA 18015

9. SPONSORING/MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSORING/MONITORING

Defense Nuclear Agency AGENCY REPORT NUMBER

6801 Telegraph Road DNA-TR-92-12Alexandria, VA 22310-3398RAEE/Cohn

11. SUPPLEMENTARY NOTESThis work was sponsored by the Defense Nuclear Agency under RDT&E RMCCode B4662D RV RA 00226 RAKE 3400A 25904D.

12a. DISTRIBUTION/AVAILABIUTY STATEMENT 12b. DISTRIBUTION CODE

Approved for public release; distribution is unlimited.

13. ABSTRACT (Maximum 200 words) This study substantiated the beneficial effects of usingfluorine as an additive during the oxidation of silicon for application to InsulatedGate-Field Effect transistor technology. The modern trend towards smaller devicesmakes it imperative to reduce the time and/or temperature required for the oxidationprocess. The dramatic increase in oxidation rates resulting from very small fluorineadditions suggests the desirability of using fluorine.The results of this study have clearly shown a significant reduction in the strain atthe Si-SiO2 interface, and in addition, a reduction in the oxidation enhanced stackingfaults in the silicon. Electrical characterization confirmed the conclusion that therise of fluorine is beneficial. The results are discussed.

In the course of this work, a relatively new technique was developed to use Q-V meas-

urements to evaluate the interface and results of these measurements are included.In summary, we are excited about the significeat technological improvements that

result from the use of fluorine during the oxidation process.

14. SUBJECT TERMS Oxidation 15. NUMBER OF PAGES

Fluorine 42Interface Measurements 16. PRICE CODEPoint Defects

17 SECURITY CLASSiFICATION 18. SECURITY CLASSIFICATION 19. SECURITY CLASSIFICATION 20. LIMITATION OF

OF REPORT OF THIS PAGE OF ABSTRACT ABSTRACT

UNCLASSIFIED UNCLASSIFIED UNCLASSIFIED SAR

NSN 7540-01-290-S600 Standard Form 298 (Rev. 2-89)iPrse ied bpAMf SU&. 22 -251- I02

UNCLASSIFIEDSECURITY CLASSIFICATION OF THIS PAGE

CLASSIFIED BY:

N/A since Unclassified.

DECLASSIFY ON:

N/A since Unclassified.

SECURITY CLASSIFICATION OF THIS PAGE

UNCLASSIFIED

SUMMARY

The objective for this study was aimed at the kinetics of growth ofoxides using low concentration additions of Fluorine to the oxidant stream,the study of related cxidation phenomena such as strain and oxidation induceddefects (stacking faults), and, importantly, the significance of fluorine,either grown in or implanted on, MOS properties and hot carrier trapping.

The experimental growth and measurements work carried out under thisproject has shown that this is a technologically successful, easilyimplemented process which has a major influence on interfacial strain and hotcarrier trapping. During the course of these investigations, a new method ofmeasuring interfaces was implemented which allows one to measure accuratelyand conveniently interface distribution without the use of C-V measurement.

The report will be organized along the following sections:

1) Oxidation kinetics with fluorine additions (NF 3 anddichlorofluorethane)

2) Influence of this process on stacking faults and straincharacterization techniques

3) MOS characterization techniques and hot carrier trapping

Each of these tools was successfully carried out and objective met.Additionally, a great deal of insight and important information was uncoveredduring these studies. The investigators believe that this process is superiorto present state-of-the-art practice.

In addition to the reduced (by two orders of magnitude) impurityinclusion for Fluorine sources vs. Chlorine sources, many of the otherbeneficial effects on point defects, traps, MOS quality, etc., are included inthis study and make the above statement valid.

iii

PREFACE

Permission has been granted by The Electrochemical Society, Inc. and theAmerican Institute of Physics to use copyrighted information.

iv

TABLE OF CONTENTS

Section Page

SUMMARY iii

PREFACE iv

FIGURES vi

1 OXIDATION KINETICS 1

2 INFLUENCE OF PROCESS ON DEFECTS AND "OTHER EFFECTS" 3

3 MOS CHARACTERIZATION TECHNIQUES AND HOT CARRIER TRAPPING 5

4 SUMMARY OF EXPERIMENTAL RESULTS 7

5 REFERENCES 19

Appendices

A THE EFFECT OF FLUORINE ADDITIONS TO THE OXIDATION A-1OF SILICON

B ELECTRON INJECTION STUDIES ON FLUORINE-IMPLANTED B-iOXIDES

Accession For J

NTIS GRA&I 5DTIC TAB ElUnrmnnounced ]

=fC QUALM TUgp D 3 JuStIfI!atk,_ n__

By -

Distrlbutloz!

Availability Codes

D Avall and/cr

Di ,s

FIGURES

Figure Page

Oxide thickness vs. oxidation time for the 9oxidation of lightly doped silicon in various gasambients at 1000°C. The points represent theexperimental data. Solid line is the least squaresfit with the power of time

2 OSF shrinkage rates in various thermal heat 10treatments

3 Equilibrium partial pressure in N-F-O-H system vs. 11temperature. (Total pressure - 1 atm, NF3 - 0.044vol% and H2 - 0.005 vol%)

4 Fluorine and oxygen profiles of an oxide grown at 12800°C with 0.011 v/o C2H 3Cl 2F in 02. A, oxygenprofile; B, 8h and C, 2h

5 Fluorine and oxygen profiles of oxygen grown in 13different amounts of NF3 addition at 900°C. A,oxygen profiles; B, 2h 0.011% NF3 ; C, 2h 0.033% NF3 ;D, 4h 0.022% NF3

6 Oxide stress against NF3 concentration with oxidation 14temperature as parameter

7 Oxide stress against thickness for dry and 15fluorinated oxides, oxidation at 1000°C

8 High freqygncy C-• curves before and after injection 16(doesi 10" 5+/cm ; dox - 825A; current density:2x10 F+/cm ). The C-V curve after room temperatureinjection shows a stretch-out while the curve afterhigh temperature injection shows a parallel shift

9 Decrease of minimum capacitance lfter r om 17temperature injection (Iose: . 10 F+/cm ; dox - 600A;current Iensity: 2.2x10 A/cm ; injection fluence:1.9 C/cm ). A decrease of effective dopingconcentration in the silicon substrate

10 Decrease of minimum capacitance after room 18temperature injection (co trol pxide; dox - 600 A;current dInsity: 2.35x10" A/cm ; injection fluence:2.06 C/cm ). A decrease of minimum capacitanceresults from a decrease of effective dopingconcentration in the silicon substrate

vi

SECTION 1

OXIDATION KINETICS

The first phase of our studies was to adapt "standard equipment" used inindustry, i.e. horizontal thermal oxidation furnace to be used in the newprocess. The practical implementation of this process was to use standardhigh temperature (Mini Brute) furnaces modified with an appropriate gashandling and temperature controlled "bubbler" system capable of mixing theproper flows of impurity to oxidant gases. With this system, we establishedthe experimental source, gas delivery system, apparatus and conditions of thestudy. The work established the growth characteristics of the Si02 , as wellas some of the influence of the amount of fluorine additions on the variousgrowth parameters.

The early phases of this work were established in an M.S. Thesis by C.Wolowodiuk (Wolowodiuk, 1985) and later in a Ph.D. Thesis by U.S. Kim (Kim,1990). Some of the most recent work has been summarized by Kim, Wolowodiuk,Jaccodine, et al. (Jaccodine, et al., 1990).

We have established first, that small amounts of a fluorine source (NF3or C2H3Cl 2 F in the region of .0x/100% - hundredths of a percent) greatlyenhances the oxidation rate.

As can be seen in Figure 1 (Jaccodine, et al., 1990), there is a largeenhancement of the dry oxide thickness with fluorine added compared to theother state-of-the-art sources, other industrial processes, even over that fora ten percent volume addition of HCl with as low an addition of 400 ppm. (byvolume) of the Fluorine compound. In this same reference (Jaccodine, et al.,1990) is also summarized the effective B and B/A parameters for the Deal-Grovelinear-parabolic model as well as the constant for the power of time model(Deal, Grove, 1965; Nicollian, Reisman, et al., 1987). These parametersdefine the relevant growth laws for this process.

One of the key findings of this earlier work, which was anticipated butwas even more dramatically demonstrated, was the need for good control overthe active Fluorine concentration. "Too much of a good thing" was found to bedeleterious. This is a clear indication of the competition between oxideenhanced growth and the tendency for Fluorine species to act as an etchant ofthe growing oxide.

For the analyses of the growth kinetics previously mentioned, it wasshown that Fluorine plays a key role in the structure of the oxide formed,namely the oxide structure was opened sufficiently for the parabolic rateconstant (related to oxidant diffusion) to show a marked enhancement. Theinterface kinetics were also modified indicating that Fluorine acts almostcatalytically with a resultant increase in the linear growth rate regime.Another measure of these effects was seen by the effect that F additions haveon the dielectric properties and optical index.

The work in this part of the research was also aimed at specific kineticinformation on the new processes. We have explored this issue and have

1

discussed the kinetic data in terms of both the more traditional Deal-Grove(linear-parabolic) model (Deal, Grove, 1965) and also using the power of timemodel favored by E. Nicollian and A. Reisman (Nicollian, Reisman, et al.,1987).

This information is found in tabulation of temperature and fluorineimpurity and concentration in Table II and Table III as noted in (Jaccodine,et al., 1990).

In addition, the resultant inclusion of Fluorine in the SiO2 was studiedas a function of the two "generic" sources used in the kinetic study by SIMS.It was a finding of this work that the shape of the (F) profile depended onwhen it was introduced into the growth process, as well as the type of thesource impurity used. Specifically, these profiles were highly dependent onthe "dryness" of the oxidation. It was postulated (to explain this date)that, with hydroxides present, F will move by a replacement reaction similarto that which occurs for Cl. In contrast with our NF3 source and dryoxidation, the F profiles are generally quite flat and "saturaced" through theoxide.

2

SECTION 2

INFLUENCE OF PROCESS ON DEFECTS AND "OTHER EFFECTS"

This area covers important "other effects" that are a consequence of orbear on the technology of oxidation. Oxides grown by the new process wereexamined for any anomalous breakdown or ionic migration behavior. Theseoxides were shown to be as good as those of the best present practice, andevery indication is that F acts as an efficient materials getter forimpurities.

More pertinent, because it represents new findings, is the role F playsin suppressing oxidation induced stacking faults. It was conjectured in ourproposed work that fluorine may be able to suppress stacking fault nucleationand growth; in fact, it was extremely effective for this purpose. Weinvestigated the behavior of oxidation induced stacking faults (OSF's) duringfluorine-oxidation in temperature range 850° - 1100 * C. It was observed thatpregrown OSF's shrink very rapidly and that no OSF's are generated even whendamage is purposely done to the wafe-s prior to oxidation. This work(Jaccodine, Kim, 1986) also showed that the activation energy for shrinkage ismuch lower than for other comparable processes. This is shown in Figure 2.The more complete findings of this work are published in Lehigh Ph.D. Thesisby U.S. Kim (Kim, 1990).

One of the unexpected results of this work is that the new F oxidationprocess completely reverses the normal point defects balance at the growthinterface such that vacancies are induced over the equilibrium concentrationCv > Cv eq. and that these vacancies (instead of interstitial silicon) areinjected back into the substrate. The important implication of this is thatoxidation enhanced/retarded diffusion effe,•.s are reversed over those normallyexpected (Kim, 1990).

As a part of our general investigation, we anticipated that a few otherfluorine containing additives may be also readily useful sources. We alsoexpected that a great deal of trial and error substitution of additive gaseswould be required. It was quite clear that the original source(Dichlorofluorethane - C2H3Cl 2 F) we selected and used had very convenienthandling and vapor pressure properties for easy replacement into a state-of-the-art oxidation facility. This is where the work started. It became alsoobvious by considering the SIMS profiles of included F in the oxides that thepresence of hydrogen in this source resulted in an effective "wet" oxide.This was also seen in growth rate vs. temperature results. Our next step wasto investigate a source with neither H or the Cl. This source NF3 turned outto be the simplest and most flexible and useful of the ones investigated. Allthe work by Wolowodiuk, Kim, McCluskey and Huang (theses and publications)were carried out using one or other of these two sources.

In order to come to grips with the potential multitude of availableFreon-like sources (Freon 11, 12, 22, etc.), we not only explored theliterature for current use but importantly we used the SOLGAS program(Eriksson, 1971) to calibrate the partial pressure of possible oxidizingspecies under the actual high temoerature oxidation conditions in the furnace.

3

These calculations could be accomulished for all the reasonable sourcefamilies that were considered provided sufficient thermodynamic informationwas available. In that way families of like compounds could be compared togive insight into what species were responsible for enhancing rates, etc. Forexample, the C2 H3 Cl 2 F source that gzve strong indications as to the importantactual species partial pressures that were different and thus effective forunderstanding the process.

This simulation and calculation techniqu-t were also employed to explore(theoretically) the role of small additions ot H2 (and, of course, H20) to theambient. This approach also allowed us to recognize the advantages of the useof the NF3 source. This can also be seen in Figure 3 (Kim, 1990) which is aplot of partial pressure in N-F-O-H system vs. a function of temperature.

By means of these calculations and simulations, many experimentalconditions and source compositions could be compared without the need for thelarge expenditure of effort, materials and cost entailed in one for onecomparison experimental work.

Guided by these programs, 7e then took the most attractive resources andcommitted these to the experimental studies.

The analytic work involved the determination of the F concentration indepth in the oxide and its role on mechanical properties was included as partof all the experimental studies. Data from SIMS was obtained from manysamples. A sample of the different resulting F concentration from bothsources can be seen from the Figures 4 and 5 (Jaccodine, et al., 1990).

Figure 4 is illustrative of the "peaked" concentration when the presenceof H2 in the source allows H2 0-F replacement reaction to occur. Figure 5shows an example of the "saturated" level concentration using the NF3 sourceand very dry conditions. Further details of this work and particularly thework in which the Fluorine was included as pulses during oxidation processesare now part of the Lehigh Thesis of Dimitrios Kouvatsos (Kouvatsos, 1991).

Study of Interfacial Strain as a function of F concentration was carriedout and reported in Lehigh thesis of D. Kouvatsos (Kouvatsos, 1991) in Chapter3. As shown in this work, oxide stress as measured optically is markedlyreduced when the concentration of additive NF3 exceeds 100 ppm in the oxidesgrown from 800 - i00° C (Figure 6). Additional work was done to profile thestress throughout the thickness of the oxide both dry and with NF3 addition.It was shown in this study (Figure 7) that again the total stress was relievedwith NF3 addition but also a variation of stress as a function of depth foroxides grown at 10000 C. This effect is less dramatic than 900° C and moredramatic than 10000 C for 11000 C.

4

SECTION 3

MOS CHARACTERIZATION TECHNIQUES AND HOT CARRIER TRAPPING

The electrical characteristics of Si-SiO2 interface:

The beneficial effects of using fluorine in the oxidation process havebeen described elsewhere in this report. The objective for this part of thework has been to investigate the effect of fluorine on the electricalcharacteristics of SiO2 and its interface with silicon. The equipmentavailable for this investigation consists of the routine equipment normallyused such as high frequency C-V measurements, quasi static measurements andhigh field injection current measurements.

In addition to this type of equipment we have an automatic avalancheinjection apparatus that was originally designed by one of the authors (D.Young) while he was at IBM Yorktown Heights Research Laboratory andsubsequently donated to Lehigh by IBM. The results of these studies, usingthis equipment, enable us to determine the cross section and density of thetraps involved. The injected carriers can be either electrons or holes asdesired. This apparatus makes it possible to pass a constant current throughthe sample using a silicon avalanche injection process. this current can beset as desired over a wide range. For the study of large cross section trapsa large current is used. The injection process is periodically halted (forexample, every 100 sec.) and the resultant trapped charge measured byautomatic high frequency C-V measurements. The results are stored in apersonal computer and can be subsequently analyzed to determine the trapcharacteristics.

Since this apparatus runs automatically, a large number of measurements(200-500, for example) can be made, and in some cases it has been possible tomeasure three different traps for one particular run. This large crosssection traps are filled first and then the smaller cross section traps arefilled subsequently. The means for analyzing the data has been described innumerous previous publications (Young, 1981) and will not be repeated here.

The trend in modern semiconductor technology to use thinner insulators(down to 100 A, for example) has placed increasing emphasis on the propertiesof the interface as compared to bulk effects. To enable us to learn moreabout the interface we have designed a special apparatus to measure in arelatively direct way the interface potential as a function of voltage appliedto the gate electrode. The technique for making this measurement that we haveused is called the Q-V technique. By comparison of the results of thismeasurement of interface potential as a function of the applied gate voltage,we are able to determine the interface state distribution without the use ofC-V measurements. This technique has many advantages that include thepossibility of making the measurements over the entire band gap of siliconusing only one sample, the interface potential measurements are independent ofthe interface state density. However, the results of the interface potentialvs. the applied gate voltage do depend on the interface state densities and acomparison of our measured results with the ideal curves of interfacepotential vs. applied voltage enables us to determine the interface state

5

density. This compression is made without the use of capacity measurements orcapacity calculations which is unique for our work. We have also used thisapparatus to evgluate the effect of fluorine.

The results of measurements of trapped charge as a function of timeresulting from the passage of electron current are frequently complicated bythe competing mechanisms of negative charge build up in the bulk of the oxideand positive charge build up due to the generation of slow states at theinterface. In addition, we also have observed the build up of fast states atthe interface. We have been able to separate these effects by means oftechniques developed in our laboratory. One significant means to do this isthe use of an elevated temperature for the injection (120°C). Thisessentially eliminates the charge build up due to the slow states. There aretwo reasons for this: this temperature anneals out some of the these statesas they are generated and, in addition, under conditions of our measurementsthe states tend to be discharged at these temperatures and hence do notcontribute to the results. As a result, we are able to evaluate bulk chargetrapping effects without this complication. During the course of this work wehave also observed a significant decrease in the generation of fast interfacestates at 120°C (Figure 8).

Most of the results to be briefly described are based on the work ofDunxian Xie and are described in detail in his Ph.D. dissertation. This workhas been published in the Journal of Applied Physics (Xie, Young, 1991). Acopy of this paper is enclosed. The equipment for the Q-V measurements wasdesigned by Ta-Cheng Lin and his work is described by a paper that has beenaccepted for publication in the Journal of Applied Physics. New means areused to determine the unknown constant in the Q-V measurement and, inaddition, the technique for determining the interface state densitydistribution uses a comparison of the measured interface potential as afunction of applied gate voltage with the calculated potential vs. gatevoltage curve. Using this method we do not involve capacity measurements asis usually done.

6

SECTION 4

SUMMARY OF EXPERIMENTAL RESULTS

A brief summary of the results is given below. For a more completediscussion, see the paper included with this report.

Initially the fluorine was applied during the oxidation of the wafers.We were concerned about the purity of the gases we were using for this purposeand to be certain that our results were not due to impurities we shifted tothe use of implantation as a source of fluorine. The use of implantation alsomade it easy to conduct comparison experiments since the fluorine can beimplanted over part of the wafer and the remaining part of the wafer was usedfor comparison with the non-implanted case. There is always concern with theuse of implantation that the results might be due to the implantation damageand not due to the chemical impurity. We have annealed all of our samples at1000C after the implantation to hopefully eliminate the radiation damage. Tocheck on this we have implanted an inert species Ne+ and the results aresimilar to the non-implanted case so we conclude that, for the purpose of ourexperiments, implantation damage is not playing a significant role (Figure 5of the enclosed publication).

One of the complications of electron injection-trapping experiments isthat the effective trapped charge does not build up monitonically but there isa turn around in the characteristics wizich is known to be due to the build upof positive charge at the interface due to the formation of slow donor statesas a result of the passage of electron current (Dekeersmaecker, et al., 1979).Our results with fluorinated samples also show this effect as shown in Figure2 of the publication en ose• . However, we also observe that for a particularimplantation dose of 10 /cm this effect disappears.

The increase in electron trapping rate as a result of fluorine is to beexpected since it is generally observed that the addition of chemically activeimpurities increases the electron trapping rate of SiO2 . Many of the othermodifications of SiO2 that are being considered have the same effect such asthe use of oxinitrides. It is interesting to note that for the case offl rin , the cross section of the traps involved is in the range of 10-17-1OY?/cml. These relatively small cross sections are not of concern for deviceoperation since they would not be filled under normal conditions of deviceoperation. It would require a very long time to fill these traps with theinjected current density present with operating devices. The traps of concernhave much larger cross sections than these.

It has also been observed that there is a larger stretch out in the C-Vcurves for the fluorinated oxides as a result of electron injection suggestingthe generation of fast interface states. To investigate the effect this hason our measurements, we have followed the suggestion originally made by Laiand Young and compared the flat band voltage shift measured at flat band withthe shift measured at mid-gap. The suggestioxi was that the interface stateswould not be charged at mid-gap. The observed shift is much larger at mid-gap(Figure 6 of enclosed publication). It appears that the relatively smallershift observed at flat band is due to the compensation of the bulk negative

7

charge due to trapped electrons by the positive interface charge due ot donor-like interface states. Measurements made at an elevated temperature (1200 C)also show that these differences tend to disappear indicating that thesestates are being annealed out during the electron injection process.

As a result of the application of a fluence of injected electrons, thecapacity in the depletion region decreases. This has been investigatedextensively by C.T. Sah and his students (Pan, Sah, 1987). They haveexplained this effect as being due to the release of hydrogen by the electrontrapping process which diffuses to the interface and into the silicon. In thesilicon, the hydrogen forms complexes with the boron dopant impurity andrenders it ineffective as a dopant. We have observed that implanted fluorineenhances this effect. This enhancement might be due to the increased hydrogendiffusion due to implanted fluorine. These results are shown in Figures 9 and10 where the results for the unimplanted sample are compared with theimplanted sample.

8

i0i-<' 104

W 1o3Co)LUzY_ 102 0 O.o0Sr vol% C2.H,3clF

U 0.044 vol% IW.,A 10 vol% HC- ief.15)

13 3 vol% C22 F ef.16)

0 Dry 0 2

101

102 103

OXIDATION TIME (MIN)

Figure 1. Oxide thickness vs. oxidation time for the oxidation of lightlydoped silicon in various gas ambients at 1000°C. The pointsrepresent the experimental data. Solid line is the leastsquares fit with the power of time.

9

Tempera ture(C°)

1200 liIl 1000 900

G~hift work 61-1.7 eV&Fast shrinkaqe iLn 3 9,-2.3 @V

- w±Ltridatioa A -3.6 OeIE0 5 o O s~ a h ,• ,in ] k a e i n, x. A Z - 4 .3 *

O@.3% NC1. oxidatim 2 A1,4.3 *

"0!~U) A~w4 -- 4.

- C1 0

Ii 7 :? I

0.7 -O 0.9

Figure 2. OSF shrinkage rates in various thermal heat treatments.

10

10.2

F

F2 N2LU

,.n HFLU 4S10

-IA. NO

= F02. SiF4

10 "5 2

0

500 700 900 1100 1300

TEMPERATURE (CC)

Figure 3. Equilibrium partial pressure in N-F-O-H system vs. temperature.(Total pressure - I atm, NF3 - 0.044 vol% and H2 - 0.005 vol%).

11

1 022 Si-SiO 2 inter-aceI '

01 O.

I

10<0 o2

< 10 19

0 101J

0 0.1 0.2 0.3DEP1H (W.)

Figure 4. Fluorine and oxygen profiles of an oxide grown at 800*G with0.011 v/o C 2H 3Cl2F in 02 . A, oxygen profile; B, 8h and C, 2h.

12

10

Si-SiO 2 interfaceSI ', J

o on

21 -

U 10' !.02

20 O

z 1

131

U A

1018 D

10 17 1

0 0.1 0.2 0.3,

Figure 5. Fluorine and oxygen profiles of oxygen grown in differentamounts of NF3 addition at 900*C. A, oxygen profiles; B, 2h0.011% NF3; C, 2h 0.033% NF3; D, 4h 0.022% NF3.

13

4

, 0 800C

SM 900C

A 1000C

3- \

'U

02. . -

1I. I ..

0 100 200 300 400 500

NF3 concentration (ppm)

Figure 6. Oxide stress against NF3 concentration with oxidationtemperature as parameter.

14

4,C44

2

O'

0

0 1000 2000 3000oxide thickness (A•)

Figure 7. Oxide stress against thickness for dry and fluorinatedoxides, oxidation at 10000C.

15

I*LS

1.1-

_- 10" Fr/cm' ('I 1 5A, H)

N -- after injection at 1200C. after injection at 220'

C.9 . - before injection

-V M

0 o \

S• .... iI°°.

0 .5 I i i I I I i i t i I i i i I I 1 i i i 1 i , t i I i i i i i i • , i i 1 i i i i i i i i

-10 -5 0 5 10 15VG (V)

Figure 8. High freq~ncy C-y curves before and after injection(dose& l0O' r/cm ; dox,-825A; current density:

2x10"• F /cm ). The C-V curve after room temperatureinjection shows a stretch-out while the curve after hightemperature injection shows a parallel shift.

16

1.1-

mN

0.9

0

0.7 .

S 10' F*/cm2 (:i5C.N)

- before injec:ion

-- cfter injecticr- ct 22°C

0.5-10 -5 0 5 10 15

VG (v)

Figure 9. Decrease of minimum gapacitince after room temperatureinjectfgn (dole: 101 F+/cm ; dox-60OA; curient density:2.2x10 A/cm ; injection fluence- 1.9 C/cm ). A decreaseof minimum capacitance results from a decrease of effectivedoping concentration in the silicon substrate.

17

1.1-

Control (R0D)

- before injection

0.9 -- cfter injection ct 22°C

0 .7 - -

0 . 5 - I A I A I A I A A A I A I I I I I I I A A A I a A A i A I A I A i

-10 -5 0 5 10 15VG (v)

Figure 10. Decrease of minimum capacitance after room temperatureinjec~iou (con rol oxide; d~x - 600 A; currený density:2.35x10"J A/cm ; injection Hluence: 2.06 C/cm ). Adecrease of minimum capacitance results from a decrease ofeffective doping concentration in the silicon substrate.

18

SECTION 5

REFERENCES

Deal, B. and Grove, H., J. Appl. Phys. 36, 3770 (1965).

Dekeersmaecker, R.F., DiMaria, D.J., Irene, E.A., Massoud, H.Z., Young,D.R., J. Appl. Phys. 50, 6366 (1979).

Eriksson, G., Acta Chem. Scand., Vol. 26, pg. 2651 (1971).

Jaccodine, R. and Kim, U.S., Appl. Phys. Lett. 49 (18), pg. 1201, (1986).

Jaccodine, R.J., Kim, U.S., Wolowodiuk, C., et al., Electrochemical Society,Vol. 137, pgs. 2291-2294, July, (1990).

Kim, U.S., Ph.D. Thesis, Lehigh University, (1990).

Kouvatsos, D., Ph.D. Thesis, Lehigh University, (1991).

Nicollian, E., Reisman, A., et al., J. Electron. Material, Vol. 16, pg. 45,(1987).

Pan, S.Cheng-Seng and Sah, Chih-Tang, Appl. Phys. Lett. 51, 334 (1987).

Wolowodiuk, C., M.S. Thesis, Lehigh University, (1985).

Xie, D., Young, D.R., J. Appl. Phys. 70, 2755 (1991).

Young, D.R., J. Appl. Phys. 52, 4090 (1981).

19

APPENDIX A

The Effect of Fluorine Additions to the Oxidation of Silicon

U. S. Kim, C. H. Wolowodiuk,1 and R. J. Joccodineo

Sherman Fairchild Center for SoLid State Studies. Lehigh Universtty, Bethlehem. Pennsylvania 18015

F. Stevie and P. Kahora

AT&T Bell Laboratories. AllentomWu, PennsyLvania 18013

ABSTRACT

Experiments were carned out to study the effects of fluorine additions to a dry oxidation ambient. Two distinctclasses of fluorine sources. liquid dichlorofluoroethane (C2H3Cl2F), and gaseous nitrogen trifluonde (NF 3). were investi-gated. We experimentally found that small fluorine additions (up to 0.11% by volume) caused large enhancements m oxi-dation kinetics. The oxidation kinetics data were analyzed by both the power of time and linear-parabolic models as afunction of fluorine addition, temperature, and the type of fluorine additive. Thermodynanuc calculations for these classesof fluorine sources were extensively carried out to determine the active oxidizing species that cause the significant en-hancement of the oxidation. According to these calculations, the enhancement of oxidation could be explained by thepresence of hydrogen fluoride (HF) and atomic fluorine (F). Secondary ion mass spectrometry (SIMS) was performed tostudy the incorporation behavior of fluorine into the oxide layer. C2H-3C 2F oxides displayed peaks at the silicon-oxide in-terface. while NF3 oxides exhibited flat fluorine profiles.

The addition of percent concentrations of a chlorine rates from 1 to 10 ml/nmr. This corresponded to fluorinebearing compound to the silicon oxidation process has additions from 0.011 to 0.11 vio. N• 3 is a gaseous fluorinebeen widely practiced in silicon integrated circut process- source, and its gas flow was monitored directly by micro-ing technology (1. 2). A new process has been reported (3-5) flowmeter before introduction into the oxidation furnace.showing that small additions of fluorine in the parts per Again, the oxidizing gas flow was controlled at 1 liter/mmmillion range can result in a greatly enhanced oxidation 02 plus the addition of NFj . The NF2 additions were fromrate as well as improved dielectric strength. Additional 0.011 to 0.044 vlo. The oxidations were carried out in a con-benefits of fluorine additions include the elimination of ox- ventional quartz furnace after a standard RCA cleaning foridation induced stacking faults (6), the retardation of various times at 90(r and 1OWC. Oxide thickness and re.boron diffusion (7). and, as reported by Ma and his co- fractive index were determined by a Rudolph Researchworkers (8). improved radiation resistance of oxide. Auto ELIl ellipsometer using a helium-neon laser. Physi-

In this paper. experimental data showing +he effect of cal integrity of the oxides was observed by optical andfluorine additives on the growth kinetics of dry oxide as a scanning electron microscopy (SEM).function of the volume percent (v/o) addition of a fluorine The calculated concentrations of the various chemicalsource compound is presented. Two different classes of species existing in the oxidation furnace were determinedfluorine compounds were selected, namely, dichlorofluo- u4ng the SOLGAS program, which computes the partialroethane and gaseous nitrogen trifluoride. The alphatic liq- pres-ures of all the possible species in equilibrium in theuid hydrocarbon, dichlorofluoroethane (C2H1C12F), has an system (9). JANAF thermodynamic data (entropy, en-appropriate vapor pressure at room temperature for deliv- thalpy) (13) for all the species in the system and the num-ery by a gas bubbler system. Nitrogen fluoride (NF?) is a ber of moles of each element with stoichiometric coeffi-gas that can be easily diluted and delivered directly into an cients were an input into the program, along withoxidation gas system. The former source is easy to handle temperature. The total pressure, temperature, and initialand combines both fluorine and chlorine species whereas input gas mixture were assumed to be constant through.the NF3 has only fluorine and contains none of the addi- out the process. To determine the effect of fluorine addi-tional carbon and hydrogen species as does the liquid tions on an active species present at high temperature, twosource. different systems containing fluorine were initially con-

The SOLGAS program (9) was used to calculate the par- sidered. For the O,/C2H3CI2F oxidation, as a model for thetial pressure of possible oxidizing species under oxidation general class of hydrocarbons containing hydrogen, fluo.conditions to ascertain which of the many possible sources rine, and chlorine, the input data consisted of 0.11 vlowould be suitable. This program was also used for assess- CIHsCItF in an O, ambient, for different temperatures. Foring which specific species correlate with the enhanced od- 2OOM3 oxidation, as a model for the general class of somedation rate. Fluorine profiles in the oxide and silicon were Freons not containing hydrogen and chlorine, the programdetermined using secondary ion mass spectroscopy was initially run at 0.11 v/o NF3 as a function of tempera-(SIMS). Oxidation kinetics data were used as guides to ture, and later rerun under the same conditions but withinfer which physical processes are being influenced and the addition (ppm) of hydrogen (HO) to simulate the hydro-enhanced, and thus lead to the marked enhancement in carbon impurities in the O and the in-diffusion of mois-oxidation kinetics. Data were fit to the traditional linear- ture. The latter condition was thought to be the actualparabolic treatment (10) and also to the power of time composition within the furnace.model, the latter providing a better flt (UI, 12). Secondary ion mass spectrometry (SIMS) was per-

Expetrimentel formed to determine the fluorine concentration vs. depthprofile in the oxides. The oxide films were rastered using

The silicon wafers used were chem-mechanically pol- Cs" ions across a 500 x 500 Iun square at a rate ofished p-type, Czochralski-grown crystals of (100) orienta- 12-16 A/min. The fluorine profile was determined by moni.tion. with a resistivity of 2-10 fl/cm. Oxides were grown by toting "SiF', which is equivalent to directly monitoringinserting wafers into the furnace tube in flowing N2 and 'I'F. These rates have been verified by comparing the orig-then switching over to the oxidant supply. The oxidizing inal oxide thickness measurement with the position of thegas flow for O/CIH3Cl2F oxidations was composed of 1 li run *SiF' and "0' profiles. Quantitative SIMS analysis re.ter/min of oxygen (03) along with a much smaller flow of quires standards of similar composition to the unknown.C2,H2ClF. The bubbler filled with the C2H3Cl 2F was kept at Since no fluorinated thermal silicon dioxide standardsroom temperature, while oxygen was bubbled through at exist, we used a F-implanted standard typically with

Electiochemcal Society Active Member. <<1% F. In particular, it was observed that the film den.'Presnt addrms: Shipley Corpoat Newton, Masschus sity decreases with fluorine incorporation. This matrix ef-

02181 fect complicates the analyses and thus the fluorine concen-

J. Electrochem. Soc., Vol. 137, No. 7, July 1990 © The Electrochemical Society, Inc. 2291A-I

2292 J. Electrochem. Soc., Vol. 137, No. 7, July 1990 © The Electrochiemical Society, Inc.

Table I. Composition of gas phase at 9000C with 07/0. II vi

C2HICl2F and 0,/0. 11 wo NF, system. (Ten ppm of 1H2 added to each104system considenng the hydrocarbon impurities in the oxygen.)

0/CHHClzF system 02 NF1 systemChemical Parual Chemical Partia,

C 103 species pressure species pressureU)

L0 O, 0.99 x 100 02 099 10'"HCI 0.43 10-' F 0.10"lHF 0.22 x 10' ONF 0.76, 10--'

U CO? 0.22 x I0- N2 0.71 , 10-2

102 0 0.05 vo C,HCxF H 20 0.28 x l0- F, 0.29 , 10-"a 0.0" va% W, Cl, 0.22 x 10-1 HIF 0.99 X 10-S10 v1i% HO af.15) CI 0.12 x 10' F0 2 0.11 10-40 3 vai% 0, etF.1e)* ry 0'.

1 _0___ _,_, _,_ ,_ __ amounts of moisture and hydrocarbons will provide someH2 which will quickly give rise to HF in the furnace am-

102 10 bient. Thus, it is postulated that both F and HF readilyform in either case and are the active species responsiblefor the increase in the oxidation rate. Similar simulated

OXIDATION TIME (MIN) runs with other related compounds (various Freons. etc.)

Fg. .Oxide thidiss. axiduin timiefoethexiastion of lightly lead to essentially the same conclusions.

doped silicon in vas gas ambients at I000*C. The points rprese Oxidation resuLts.-The fluorine oxidations were carriedthe expenmental date Solid line is the least squars fit with rte power out by varying four parameters--oxidation temperature.of time. time. fluorine concentration, and the type of fluorine addi-

tive. In Fig. 1. we plot oxide thickness vs. oxidation time

tration was determined on a relative rather than on an ab- for oxidations at 1000*C with 0.055 Wo C2H3CIF and

solute basis. 0.044 v/o NF, addition. These data were chosen so that a di-rect comparison with the published data for films grown in

Results 10 v/o HC1 and 3 v/o ClI (15, 16) could be made. It is ob-

Gas phase calculation results.-Table I contains results served that the resultant oxide is thicker for films grownof the calculated composition of the gas phases at 900*C for using the fluorine compounds despite the two orders ofthe 02/0.11 vlo C2H 3CI2F and the 0/0.11 wo NF3 systems. magnitude difference in the volume percent addition.The partial pressures of the various species are propor- Figures 2a and b present oxide thickness vs. oxidationtional to their concentrations in the system. At elevated time data for increasing C2H2Cl2F concentration at 900*oxidation temperatures. C2H3CIj F dissociates to form and 1000C, respectively. It is seen that increasing themany of the reaction products formed upon the disso- amount of CIH1ClIF acts to enhance the oxidation rate. Inciation of the commonly used chlorine additives. [e.g., Fig. 2b, for the case of 1000*C at the high concentrations (attrichloroethylene (C2H 3C13) and trichloroethane (C2HCI3)]. a0.011 v/o) and longer times, represented by data shownSome of these reaction products, such as HCI. H2 0, or C12, with the dashed line, poorer quality oxides are observed,are known to enhance the growth rate (14). However, there i.e., the oxides have pinholes. This observation is consist-is one compound that was unique to the 02C,2HIC12F sys- ent with the role of carbon as a nucleating agent for pin-tem. which was present in relatively large amounts, holes in the presence of HF.namely, hydrogen fluoride (HF). Since the chlorine com- Figures 3a and b present the comparable data of oxidepounds cannot account for the large observed increase in thickness us. oxidation time for increasing volume concen-the oxidation rate, it was postulated that HF was the pri- tration of NF3. It is found that the NF 3 species is also effec-mary active species. tive in enhancing the oxidation rate as can be expected

For the case of OViN 3 system. it can be seen from the from its chemistry. It should also be noted that contrary toTable I that a significant amount of atomic fluorine (F) is the result of the above hydrocarbon source, optically clearpresent. However, it is also expected that residual pinhole free oxides result when using NF 3 over the entire

10'

10'

10= ~~103 • ••I Q1 u I % a m 0.m 1 V01%

0 0.011 vov 0 0.0 11 MIo%

101 0 Coy Amilent 102 Dry Aflribiart

102 102 10" 103

OXIDATI0N TIME MN) OXIDATION TME GMNOFig. 2. Oxide t vsictno v. ouidiien fime fo the oxidetwen of (100) silicon in eariou OWCiH 1CF oxidsien (e( left) at WOa end (W right) at

IO , The peinis srpsnt the exp imenl dIae. SOid line is the least squern fit with the power of tme.

A-2

J. Electrochem. Soc., Vol. 137, No. 7, July 1990 C The Electrochemical Society, Inc. 2293

10'

10'

Cn(n 101 10)

6 9 OO"vcA 0.044 Vol% oono" VOI-- 00.022 voe% oC.O1 Vo%

102 00.011 vol% 1 D rbit0ocry Ambient1

102 103 10a 10o

OXIDATICN TvE (MuN OX[DATION TIVE MEFig. 3. Oxide thickues vs. oxidation time far the oxidation of (100) silicoa in various 0I/NF, oxidation (a, left) at 90OC and (b, rigt) at 100(0C.

The points repn t he expAimentp a doa Solid line is the lean squom fit wmt *a pow of tim.

concentration, oxidation time-temperature range of these stant increases over the full HCl concentration range. Instudies. the case of oxidation data at 900( and 1000*C. an almost

The experimental data was treated in terms of the linear- order of magnitude increase occurs in these parameters inparabolic model (10) and the more recently proposed spite of the much lower fluorine concentration. As increas-power of time model (11, 12). The following two relation- ing amounts of NF3 are added to the oxidation, the opticalships were used to extract the appropriate parameters refractive index was found to decrease from 1.46 to 1.451.

This leaves one to infer that the fluorine incorporation de-X - a(t. + tb [I] creases the density of the oxide and also has an influence

on the viscoelastic properties of the resultant oxide muchX - A[I + 4B(t + T)/Ar - 11/2 [2] as water vapor does. In fact, this correlates with some of

where X is measured oxide thickness. For the power of the experience of workers in the area of silica optical fibers

time model G and b are coefficient and exponent, resW- (18).tively, while t, and to are the oxidation times required to SIMS results.- Typical fluorine concentration vs. depthgrow the final measured oxide and the pre-existing oxide. profiles for oxide films grown using O,/C2H3C12F source areThe constants a, b, and t4 were determined by least shown in Fig. 4 and Fig. 5, for the same W/o addition but atsquares analysis. Since these oxidations were all per- the extremes of the investigated temperature range. Theformed at long oxidation time intervals, a simple leastsquares fit the data very well. The A and B in Eq. (21 are theconstants in the linear-paraboac model, t is the oxidation Table il. Effect of fluorine additive on the consat from wt powertime, and r is a fitting parameter which accounts for initial of time model oxidation at 900( and 1000*Crapid growth regime. The constants A and B were deter-mined from X vs. (t + r)/X plots. The fitting parameter - M Ambient a b t, (mi) Errorwas determined by the intersection of the extrapolated lin-ear portion of the linear-parabolic model with the time 900" Dry 02 9.30 0.72 6.5 0.0008axis. This extrapolation provided a constant value of the 0.011 NMF 46.77 0.66 1.6 0.0003initial rapid growth of 180 + 20A for all the data of these 0.022 NF3 46.77 0.70 1.6 0.0097fluorine oxidations. 0.044 NF3 63.09 0.64 5.0 0.01410.011 CtH3CIF 52.48 0.65 5.0 0.0022

The resultant calculated constants for both these models o.o55 CtHsCIF 107.15 0.56 5.0 0.0015are given in Table IH and Ill. The solid line in Fig. 2 and 0.11 C2HC1F 147.91 0.53 5.0 0.0013Fig. 3 are the least squares fit with the power of time 1000' Dry 02 128.82 0.45 0.3 0.0034

0.011 NF3 131.82 0.50 35.0 0.0001model. This fit was superior to the inear-parabolic model 0.022 Ni'F 169.82 0.47 30.0 0.0239and therefore was used. As pointed out by Blanc (17) and 0.044 NF3 338.84 0.36 30.0 0.0107others, reasonable physical models can lead to equivalent 0.011 C2HCIF 208.92 0.45 30.0 0.0029mathematical expression for the thickness as a function of 0.055 C2HIC12F 213.79 0.51 24.0 0.0016time and that a large class of models' exhibit similar math-ematical structure. The value of carrying forward analysisof the above two models is that the power of time model Table Ill. Effect of Nho- additive am Ow constants froan ft linear-provides the best fit to the data while the linear-parabolic pamhe€ models dring oidatio at 900" and 10O'Ctreatment ties back to the large body of data already pub-lished. 7rM Ambient B(rm'/h) A/B(Pm/h) Tih)

The enhanced oxidation rate for the oxides grown withincreasing concentrations of fluorine compounds is 900" Dry 02 0.0021 0.0122 1.4readily evident from Fig. 2 and Fig. 3. When the data are 0.011 NF3 0.0047 0.0127 1.2

0.022 NF, 0.0059 0.0152 1.0compared to previously published results of the oxidation 0.044 NF 0.001 0.0194 0.8of 1-10 vlo HCI (15). several interesting differences are evi- 0.011 C2930CIF 0.0258 0.0280 1.0dent apart from the nature of the impurity. For all case of 0.055 C2HCItF 0.0438 0.0629 0.5fluorine oxidation both the linear and parabolic constants 0.11 CHsCIF 0.0724 0.0946 0.25

1000" Dry Ot 0.0073 0.0709 0.4increase with increasing fluorine concentration, whereas 0.011 NFi 0.0192 0.1053 0.4in the comparable HCI data only the parabolic rate con- 0.022 NF3 0.0203 0.2203 0.4

0044 NIP 0.0278 0.2530 0.42We acknowledge the mueon made by the reviewer for this 0.011 CAWscisF 0.0471 0.1162 0.2

approach. 0.055 CIH3C1,F 0.1105 0.1162 0.1A-3

2294 J. Electrochem. Soc., Vol. 137, No. 7. July 1990 © The Electrochemical Society. Inc.lO, 10"M1

Si-SiC 2 interface

10 SI-S12 titaf'1ce 10

S11201

1 O 101

C10"t I I 1 1017 I

0 0.1 0.2 0.3 0 O,1 0.2 0.3

Fig. 4. I .hi Oy1m6 Po4 ile of doxide gro1w at 8OCCwit Fig. 6. Fluorine and oxygen profiles of oxide grown in difeit0.011 iao CI•CIt, in O2 . A. oxygen profile; 8, 8 arid C, 2h. anaonts of NF3 addition at 900C. A, oxygen pruoes; B. 2b 0.011%

NF,; C, 2A 0.033% NFI; 0, 4. 0.022% NF3.

oxygen signal is presented on an arbitrary scale to showthe position of the interface as well as to calibrate the sput- the oxide as if the fluorine were tightly bonded into thetering depth. Some of the data for the O,/NF 3 oxidation are oxide network and not mobile.shown in Fig. 6 for NF3 concentrations of 0.011, 0.022, and0.033 vio. Discaian

A comparison of the fluorine profiles show a consistent Gas-phase compontiow Active fluormine species.-Thedifference between the two sources. The fluorine profiles SOLGAS calculation for fluorine compound additionsusing C2H 3C12F are characterized by a fluorine peak which provided a convenient guide in selecting potential sourcesmoves with the silicon-oxide interface. The fluorine con- and assessing the partial pressure of the resulting activecentration is low at the surface, and increasing to a peak species. Table I lists the species with the highest concen-concentration at almost three-quarters of the oxide thick- trations that are present during the oxidation at 900"C forness where it remains more or less constant to the inter- each of the sources used in this study. These calculationsface. In contrast to this, the oxides, grown using NF 3 are showed that commonly used compounds such as CrHClstypified by the profiles shown in Fig. 6, namely, they ex. or C2HsC1,, when heated, will give simila active specieshibit a high uniform concentration of fluorine throughout (except for the fluorine-related species) as the C2H2Cl2F

type source. Those findings are similar to those observedfor HCl or chlorine oxidation (15) leading one to concludethat the essential difference in the enhanced oxidationrates is the presence of the high concentration of HF. How-

Si-SO 2 nteftce ever, when the concentration of this active species (IF) in-creases, especially in the presence of carbonaceous inpur-ities, pinholes result

M + When the NF2-type source is used, it dissociates rapidlyat elevated temperatures. Table I shows that an apprecia-ble amount of F is present in the furnace which results inthe enhanced oxidation rate. The presence of even smallamounts of H, (between 0.000005 and 0.005 moles) affects

10 2o the concentration of active species and forms HF. The re-10= suits of such a simulation are shown in Fig. 7. where the

active species are plotted as a function of temperature. Itcan be seen that for 0.005 vlo of H2 an appreciable concen-

10O9 tration of HF is formed throughout the entire temperaturerange. The HF concentration changes directly with vlo NF 3

A added to the O% ambient ifH, is present.

Thermal oxidation in the presence offluormne.t-The sig-U Onificant enhancement of the oxidation rate was achieved

sg by small fluorine additions to the oxidation stream. In thecase of a hydrogen bearing compound, some of the en-hancement is due to the in sttu formation of HI0. This ef-fect has been observed for the enhancement of oxidationwhich takes place using only hydrogen-chlorine corn-

0 0.1 0.2 0.3 pounds (such as HC), but in no case is it of the same mag-nitude, as can be seen from the comparison of the growthcurves in Fig. 1. From the compisison of the 0,'C2HCIF

Fi. S. Pluvine aed zyget peila of oeis witi 1000"C system with the OINF3 systm Mi the oxide thickness V.wi10 0.011 W/e C2H1C in (h. A, oxygen profiies; B. 2h; C, Ik. time data, one can see that the pIwese of a significant

A-4

J. Electrochem. Soc., Vol. 137, No. 7, July 1990 © The Electrochemical Society, Inc. 2295

10-2 Q A mechanistic model for the power of time relationshiphas also been recently proposed by Nicollian and Reisman(23). In this model, they propose a surface reaction con-trolled mechanism driven by the generation of free volume

" •through the v6lume expansion and the subsequent vis--5 10-1- O . NF Jcous flow in the oxidation process. The average viscosity

of the oxide is time, temperature, and pressure dependentWJ N2 because the oxide undergoes oxide ring reconfiguration

NF during growth. i.e.. the time dependent viscosity is thecause of a time dependent forward specific reaction coef-

(0 i -F ficient. This model shows some of the characteristics ofLU •other models (24) but differs in its treatment of the tume de-m NO pendence of the viscosity, i.e., it does not make the as-0.. sumption that the oxide is a Maxwellian (viscoelastic)

Sfluid. More to the point this model involves the conversion

<1-- e of the Si lattice into Si--O-Si rings, and as such involvesmicroscopic mechanisms through the power of time

H- model exponent b. It is found that b varies from about 0.85to 0.57 depending on the oxidation conditions at oneatmosphere (11). In general, for a diffusion-limited oxida-

-10-e tion. the value of b is expected to be 0.5. On the other hand.if strictly reaction-limited b should be equal to 1. Our data300 700 900 1100 1300 gives values varying between 0.7 and 0.5 with the b valuegenerally decreasing with increasing concentrations of the

"TE PERATURE () fluorine additive. This indicates that even at low concen-trations of fluorine b values are affected and are generally

Fig. 7. Etuilibrium parti pms me. in syste vs. tow lower than those at 900*C for 1 atm dry oxidation. Atpietwe. (Total pImew - I am, NFI - 0.044 vi. m H2 1000"C, all the b values determined in this study are at or0.005 v/0.) below 0.5.

Looking at the parameters extracted from the linear-parabolic treatment, one can see an increase in both the

partial pressure of HC1 and H20 along with the HF aids in parabolic and linear constants as a function of temperaturethe enhancing the oxidation rate. The observation of pin- and additive concentration. It is also shown that both theholes at the higher temperatures and higher concentra- parameters associated with diffusion (B) as well as the li-tions of C2HsClF indicates that local etching occurs in ear rate constant (BIA) increase with fluorine additive con-competition with growth. In the same temperature and centration. As stated previously, the power of time gives asimilar concentration ranges, with care, the NF, source re- superior fit to the data and one measure of this is the largersuits in optically clear and pinhole-free oxides. In this re- values of i needed in the linear-parabolic fitgard, it was reported that etching and growth occurs in the Several posibilities can be proposed for the effective-lower concentration range used in this work (4). However. ness of fluorine as an oxidation enhancer. Flourine can actMorita's data has been recently corrected (19) for underis- catalytically at the interface and competes with oxygen totimating the actual amounts of NF3 by 9.7 times and now form Si-F bonds (25) (based on electronegativity differ-their results agree generally with our data. The remaining ences). This increases the available number of Si danglingdifferences can be accounted for by recognizing that bonds (analogous to the mechanism proposed for F etch-Morita et aL. used a cold-wall quartz reactor with a halogen ing of Si). This enhanced activity increases the reactivity atlamp heater vs. the standard hot-wall quartz oxidation sys- the interface and subsequently the growth rate of the oxi-tern in this study. The presence of a large area of hot quartz dation. Fluorine can also effectively substitute for oxygenhas a strong influence on the attack of active species and as a oxide network modifier and thus modify the structure.their effective participation in the growth process. ie., open the structure to more readily allow ingress of the

The effect of fluorine additions on the model parameters oxidant This also directly relates to the change in the localis shown in Tables; U and MIf for the Linear-parabolic and stress and viscosity at the interface, and correlates with thepower of time models. These parameters will be used to lower index of refraction observed. Analysis of our dataguide the understanding of the detailed processes that does not enable us to eliminate either of these mecha-may be occurring. The assumptions and factors that relate nisms. A more complete mechanistic theory, especially,interface reaction, diffusion, and solubility of oxidant in a one that considers the chemical effect and oxide history oncomprehensive and complete theory are still under mtense oxidation is still lacking. Such a theory would lead to spec-investigation (20. 21). The addition of chemical species to ific experiments testing these mechanistic models.enhance the oxidation rate further complicates the devel- Fluorine profiles in the oxide.-The fluorine concentra-opment of a theoretical model. Several recently published tion profiles determined by SIMS were found to have dis-articles have dealt with the relation of mechanism to the tinct differences depending on which chemical source wasmathematical formulations used in this study. Following used, ie., whether large amounts of moisture forming im-hovew (of= have a(17), Wter aon beg e en e purities were present during the growth. The profiles that

result from the oxides grown with C2iH3ClF yield profilespower of time and the linear-parabolic expressions. They peaked near the silicon-oxide interface. As the oxidationconcluded that the Deal and Grove's linear-parabolic ex- temperature increases the fluorine continues to peak andpression is an approximation to the power of tite expres- saturate at a high conctntiont broadening from this po.sim as suggested by Blanc and can be useful asa guide for sition at an almost constant concentrat, on before decrets-technology. From their analysis, it can be seen that the ina namstcmsatcnetato eoedces"linear" and "parabolic" constants. uid in the Deal and ing to the outside interface. This shaped profile is directly"a aanalogous to the profiles found for chlorinated oxides (26).Grove model are in reality only two terms of a series ex- This behavior can be explained by a replacement reactionpression and the r in their theory is added to compensate proposed for chlorine and the critical role of hydroxylfor the missing higher order termn. The larger the values of (OH) species in this mechanism.the parabolic and linear constants, the less important are As the oxidation of CIHIC12F proceeds, a significantthe higher order terms and the small er value added (22). amount of water is generated and the bound fluorine canWolter and Zegers-van Duynhoven suggest an ionic spice be weakened and replaced by OH groups. Continung bycharme-limited growth mechanism for their growth model. the analogy to chlorinated oxide, one needs to postulateHowever, the expression for oxide growth rate given by that a replacement reaction similar totheir Eq. [1] makes a correlation with the extracted param-etsfs fom this study diffMcult vSi-.F + HtO0- Si--OH + HF [31

A-5

2296 J. Electroch em. Soc., Vol. 137, No. 7. July 1990 (0 The Electrochemical Society, Inc.

occurs during the growth. This replacement reaction de- 0094 and by the Sherman Fairchild Foundauon Fellow-creases the bonded fluorine concentration in the bulk ships.oxide from the initial high concentration of bonded fluo-nne at the silicon-oxide interface. The mechanism out- Manuscript submitted June 23. 1988: revised manuscriptlined in Ref. (26) needs to be pursued as a possible detailed received ca. Jan. 15. 1990.description of these fluorine profiles.

In contrast, the typical profiles obtained from the 02/NF3 Lehigh Universlty assisted in meetng the publicatonoxidation behave in a markedly different way as can be costs of this article.seen from Fig. 6. Here, in the absence of a large amount ofmoisture, the fluorine is bonded into the oxide and re- REFERENCESmains at a high saturating concentration throughout the 1. R. J. Kriegler, Y. G. Cheng, and D. R. Colton, Thisbulk. Insights into the behavior of NF 3 in oxides can be Journal. 119, 388 (1979): R. J. Kriegler. Y. G. Cheng,gleaned from work on optical fibers (27). In this case. fluo- and D. R. Colton. U.S. Pat. 3.692.571 (1972).rme can compete and isomorphously replace OH in the 2. J. Monkowski, Solid State Technol., 58 (July 1979);structure. It has been found that fluorine is thus randomly ibid., 113 (Aug. 1979).incorporated into the silica network in the form of silicon 3. P. F. Schmidt, R. J. Jaccodine. C. H. Wolowodiuk. andmonofluonde sites [Si(01•F1 and silicon difluoride sites T. Kook. Mat. Lett., 3,235 (1985).[Si(O-)2F (27). The Si---O--F complex is thermodynami- 4. M. Morita, T. Kub. T. Ishihara. and M. Hirose. IEDMTech. Dig., 144 (1984).cally more stable than SiOl. Thus the fluorine profiles in 5. Y. Nishioka. Y. Ohji, K. Mukai. T. Sugano. Y. Wang,this case are expected to be more uniform throughout the and T. P. Ma. AppL Pkys. Lett., 54, 1127 (1989).bulk and limited by the solubility of fluorine in the oxide. 6. U. S. Kim and R. J. Jaccodine, ibid.. 48. 1201 (1986).Note that in the Ref (27). no evidence was found for the 7. U. S. Kim. T. Kook, and R. J. Jaccodine, This Journal.formation of silicon trifluoride or silicon tetrafluoride 135.270(1988).(SiF4), which would lead to a higher fluorine concentration 8. E. F. da Silva, Jr., Y. Nishioka, and T. P. Ma. IEDMat the interface. Tech. Dig., 848 (1987).

9. G. Erikssoon. Acta Chem. Scand., 26, 2651 (SOLGASSummary Program) (1971).

It was shown that small additions of compounds con- 10. B. E. Deal and A. S. Grove, J. AppL. Phys., 36, 3770taining fluorine such as C2H3CIF and NF3 to a dry oxida- (1965).tion ambient enhanced the oxidation kinetics consider- 11. A. Reisman. E. H. Nicollian, C. K. Williams. and C. J.

Me=z J. Electron. Mater., 16. 45 (1987).ably. The enhanced kinetics indicate that diffusion 12. R. Doremus and A. Szewczyk, in "Silicon Nitride andthrough the already grown oxide as well as the reaction at Silicon Dioxide Thin Insulating Films," (PV 87-10)the silicon-oxide interface was increased by the fluorine V. J. Kapoor and K. T. Hankins, Editors, p. 350. Theadditives. Electrochemical Society Softbound Proceedings

The crucial distinction between NF3 and C2H3C12F in the Series. Pennington. NJ (1987).chemistry is that C2H3C12F provides reactive fluorine and 13. D. R. Stull and H. Prophet. JANAF Thermochemicalchlorine as well as water vapor. From the thermodynamic Tables. 2nd ed., NSRDS-NBS37 (1971).analysis of the oxidation ambient, it was determined that 14. R. E. Tressller, J. Stach, and D. M. Metz. This Journal,HF and F were the major active species for OI/C2H 3Cl2F ox- 124,607 (1977).idation and for OI/NF% oxidation, respectively. Thus. dur- 15. D. W. Hess and B. E. Deal. ibid., 12A, 735 (1977).ing these oxidations, a local etching process competes IF B. E. Deal, D. W. Hess. J. D. Plummer. and C. P. Ho,

ibid., 12M 339 (1978).with the growth process because of the presence of the HF. 17. J. Blanc, Philos. Mag., 55, 65 (1987).If this concentration exceeds a crucial amount, poor qual- 18. P. D. Lazay and A. D. Pearson. IEEE J. Quantum Elec-ity oxide results, in which pinholes and weakened struc- tron.. QE-18, 504 (1982),ture have been observed. 19. S. Aritoine, M. Morita. T. Tanaka. and M. Hirose, Appl.

It is also found that CHCI2F oxides yield fluorine pro- Phys Lett., 981 (1987).files which are peaked at the silicon-oxide interface. The 20. Oxidation Workshop, Philos. Mug., 55 (1987).peaks are sharp for the thinner oxides, and broader for 21. "The Physics and Chemistry of SiO5 and Si-SiO2 Inter-thick oxides. For the NF 3 oxides, the fluorine profile is face," C. Rt. Helms and B. E. Deal. Editors, Plenumconstant at a high concentration throughout the oxides. Press. New York (1988).The incorporation behavior of fluorine into the growing 22. D. R. Wolters and A. T. A. Zegers-van Duynhoven. J.oxide is based on the availability and competition between AppL Phvs., 65, 5126 (1989); D. R. Wolters andthe hydroxyl group and the bonded fluorine for bridging A. T. A. Zegers-van-Duynhoven. ibid.. 65. 5133sites in the oxides. (1989).

23. E. H. Nicollian and A. Reisman. J. Electron. Mater., 17,Acknowledgments 263(1988).24. E. A. Irene. J. AppL Phys., 4, 5416 (1983).The authors gratefully acknowledge Dr. T. Kook for use. 25. M. Seel and P. S. Bagus. Phus. Rev. B., 2L 2023 (1983).

ful discussions in the course of this experiment. In addi- 26. Y. D. Sheu, S. R. Butler, F. J. FeigL and C. W. Magee,tion. they would like to thank Dr. J. Parks for his contribu- This Journal, 13, 2137 (1986).tion to this manuscript. This work was supported in part 27 T. ML Duncan, D. C. Douglass, R. Csencsits. and K. L.by the Army Research Office under grant DAAL 0388-K- Walker, J. AppL PhV&, 4, 130 (1986).

Reprinted h~un JMJfkftALJVTh Eizcr1HmWCE~AL~:1CrTTWV . 137. N.. 7. Jil, I1")

Printed in U.S.A.C A-oyn6ht 1W)

A-6

APPENDIX B

Electron injection studies on fluorine-implanted oxidesDunxian 0. Xie and Donald R. YoungSherman Fairchild Center. Lehigh Univenity. Bethlehem. Pennsylvania 18015

(Received 21 December 1990; accepted for publication 23 May 1991)

The effect of fluorine on electron trapping in SiO2 films has been studied by avalancheelectron injection. Samples are prepared by 25-keV fluorine implantation into dry oxidesfollowed by a 1000 'C N 2 ambient anneal. Bulk electron traps with capture crosssections on the order of 10- 1_i0 - 19 cm 2. which are not due to implantation damage, arefilled by avalanche electron injection. An optimum dosage of fluorine implantation tosuppress the so-called turnaround effect during avalanche injection exists. This suggests thatfluorine might passivate slow interface donor states or reduce bulk hydrogen diffusion.The observation that high-temperature (120 °C) injection can eliminate most fast and slowinterface states for conventional oxides is also true for fluorinated oxides. Our resultsindicate an enhanced generation of fast donor states for oxides containing fluorine. These statescontribute a positive charge when the Fermi level is in the lower portion of the band gap.

I. INTRODUCTION keV) on half of each wafer. The other half is the controlsample used for comparison. The fluorine doses are se-As metal-oxide-semiconductor (MOS) devices ar lected in the range from 10'3 to 103cm - 2* After implan-

widely used in modern very large scale integration (VLSI) tatin the sampe are anle fo 30 M i a Nt (n-

technology, many investigators have been trying to im- gat)oambientaatl1000e'Cntolremove theiphysical damageprove the qualities of oxides by various approaches. The gen) ambient at 10300 °C to remove the physical damage

resulting from the implantation. Aluminum dots (area:use of fluorine in the SiO 2 films, although it is still in re- 7 x 10- 3 cm2 ) are deposited in a vacuum chamber withsearch stages, is one of the possibilities to make good qual- low sodium contamination to define the gate of MOS ca-ity devices. Fluorine reduces the generation of interface pacitors for electrical measurements. Finally, the devicesstates when it is properly introduced into SiO,. Nishioka et receive a 400 *C. 30-m post-metal anneal in a forming gasaL.-3 observed the densities of radiation-induced oxide (20% H2/80 N, mixture.charge and interface states were reduced when small quan- (20% H/80% N2) mixture.tities of fluorine were added before or during dry oxidation. The fluorine distribution in the oxides is profiled using

secondary-ion-mass spectrometry (SIMS). Electron injec-They explained their results by the oxide bond strain d tion current is induced in the MOS devices via the ava-dlient model which suggests that the fluorinated oxides, lanche injection technique.7 During the measurement, the

with a smaller strain gradient, reduce the migration of ra-

diation-induced defects to the Si-SiO2 interface. Wright et average avalanche current density, monitored by a Kei-

al."' studied theeffect of fluorine on SiO2 via F *" (fluorine thley 616 electrometer, is set at a value of 2-2.3 x 10-:

ion) implantation using different doses, ranging from 1012 A/cm2 . The measuring equipment has been previously de-to 1016 cm - 2 at 40 and 90 keV. Physical ion damage was scribed by Young er al.8 '9 The constant current injection is

achieved by a I150-kHz saw-tooth wave form. Automaticobserved after high-field current injection for their acqi on ena usato mase the AtoadanF '-implanted oxides even after annealing at 900- 1000 *C. data acquisition enables us to measure the flathand andTheir -implated aloxindesee aftecreanneaing ath 900- 0 t. midgap voltage shifts at room and elevated temperatures.Their results also indicated a decrease in the charge to Bt -~ ihfeunyadqaittcCVdt

breakdown for the highest dose (1016 cM - 2) F- implan- Both I-MHz a igh-frequency and quasistatic C-A V data

tation used.4 Charge to breakdown means the integrated search) and Keithley 602 electrometer, respectively. By

injected charge per unit area from a virgin sample to a cmaring c and after a valac iesample showing a resistive characteristic. comparing C-V curves before and after avalanche injec-

In this paper, we will discuss the electron trapping tion, we are able to obtain the information containing thephenomna thi pluoiaper wwidiscu the ealahetonjctrping bulk and interface characteristics due to the effects of av-

phenomena of fluorinated oxides by the avalanche injection alanche electron injection.

technique. Both bulk and interface traps were studied

mainly for low-energy (25 keV) F * -implanted MOS ca-pacitors. The results of fluorinated oxides were compared Ill. RESULTSwith those of nonimplanted and Ne ÷ (neon ion) im-planted samples. The results of SIMS measurements of F - -implanted

oxides after a 30-min N, anneal at 1000 C are shown in

II. SAMPLE PREPARATION AND EXPERIMENTS Fig. 1. These samples were implanted with different doses,ranging from 1013 to 101- F - /cm 2 at 25 keV as mentioned

Silicon wafers, (100) B-doped p-type, 0.1-0.2 fl cm above. The peak positions are in the oxide and only 40-50are used for our experiments. The oxide is grown in a dry A. away from the Si-SiO, interface. When the dose is in-oxidation furnace at 1000"C (dry oxygen, 2 I/min). The creased to 10 i5 F/cm 2, an additional shoulder appearssamples are implanted with F - or Ne ý at low energy (25 towards the bulk of the oxide. This effect is likely due to

2755 J. Appi. Phy&. 70 (5). 1 September 1991 0021-8979/91/052755-0550.00 Oc 1991 American Institute of Physics 2755

B-1

7 22 _______________

c., 0 . /, - -'r',m

- o" 1 ,~i / i i'\ /cn 2 2

10

0 00 0.05 0'0 0.15 2.20 0.0 C.5 1.0 •.5 2.0

Depth (microns) Fluence(C/cm 2 )

FIG. I. SIMS profiles of fluorine implantation (25 keV) in oxides(•=700A), after aN2 annal at IO00"T for 30rain. FIG. 3. Avalanche ,electron injection at 120"T for fluorinated oxide(dose- 1015 F+/cnm-; current density: 2 x 10- A/cm-: d_, = 825 k)

showing no turnaround effect.

the enhanced diffusion of fluorine species at high concen-trations above 1010 cm - 3.

The electron injection characteristic is studied by mon- Another way of eliminating the turnaround effect is toitoring the flatband voltage shifts versus injection time or perform the avalanche injection at 120 °C as reported influence for different devices. Figure 2 shows the compari- the literature.8 As we can see in Fig. 3 for the sample withson of nonimplanted (control) with F ÷-implanted (fluor- a dose of 101 F÷ /cm 2, the flatband voltage shift does nothuated) samples with the injecting current density of 2 show the turnaround effect. The corresponding high-fre-X 10-' A/cm2. Fluorinated samples have larger flatband quency and quasistatic C- V curves before and after thevoltage shifts than those of the control sample. Almost all injection at both room temperature and 120 "C with a flu-the curves show a turnaround effect when the fluence be- ence of 2 C/cm 2 are shown in Fig. 4. In Fig. 4(a) we cancomes 0.2-0.8 C/cm 2 except for the sample with an ima- see that the high-frequency C-V curve shows a stretch outplant dose of 1014 F + /cm 2. The positive flatband voltage after room-temperature injection and the bulk trapping ef-shift results from the negative charge trapped in the oxide. fect is not easily estimated because of the buildup of theThe turnaround effect is due to a positive charge that anomalous positive charge near the Si-SiO2 interface. How-slowly builds up and compensates the negative charge in ever, the high-frequency C-V curve becomes nearly parallelthe oxides after a certain period of time of avalanche elec- to the initial one after the injection at 120 *C. This enablestron injection. This effect of the anomalous positive charge us to determine C- V shift due to bulk electron trapping. Inhas been discussed in previous papers"' and explained as Fig. 4(b) the quasistatic C-V curve after the injection atresulting from donorlike slow states near the Si-SiO2 inter- 120 "C is significantly deeper than the one after room-tern-face. The optimum implant dose of 1014 F + /cm 2 elimi- perature injection. This means that the generation of inter-nates the buildup of the anomalous positive charge near the face states is obviously reduced for elevated temperatureSi-SiO2 interface. injection.

In order to compare the physical damage due to im-plantation, we have implanted neon instead of fluorinewith the results shown in Fig. 5. All of these samples re-

4l-4e ceived a N2 anneal at 1000 "C for 30 min after implantation1 10"r /c "m2 or after oxidation for the nonimplanted sample. The flat-

10F/n band voltage shifts of the Ne ÷-implanted sample afterelectron injection are comparable to the nonimplanted con-trol sample. These results indicate that physical damage

21013F,/cm 2 does not play an important role in trapping electrons as aresult of the annealing treatment we have used. We there-fore conclude the additional flatband voltage shift of fluor-inated samples during electron injection is not caused by

Control physical damage resulting from the implantation. Our re-

0 ........ sults are not the same as the observation of Wright and3. . 1.0 1.5 2.0 2.5 co-workers.*'5 They attributed the negative flatband volt-

Fluence(C/cmz) age shift during Fowler-Nordheim injection to the physi-

FIG. 2. Avalanche electron injection curves of control and fluorinated cal damage of F + and Ne I implantation; however, our

ouides at rooma temperature (d,. - 60 A. current density: 2 x 10 results did not indicate damage for the Ne- implantedA/zcn) showing Minaroumni effect. sample after annealing. The flatband voltage shifts of

2756 J. Api. Pl•ys., Vol. 70, No. 5, 1 Sepanier 1991 D. 0. XK and .L R. Young 2756

B-2

11 106

S--- after injection at 1 20,C

1 --- • .after injection at 22°C

0:91 b efore injection0.9 ". \V

• l1

0.7 -¶ \...4

0.5;0

0.5 i ......... .......... Fluence(C/cm2 )-I 5 0 5 10 15

(a) V; (v)

1.1 FIG. 6. Avalanche electron injection data showing the difference betweenSmidgap (V,,) and flatband (Vrb) voltage shifts of the fluonnated oxide

(dose: 101'F /cm-; d., = 600A•).

0.9 face state generation. Following the suggestion of Lai and1 eiYoung,10 that the measurements be made at midgap tot.~ . " /eliminate the effect of the interface state charge, we have

0.7 obtained the results shown in Fig. 6 for samples with 1010.7 before injection F /cm2. As the midgap is usually the point that is less

after injection at 22"C sensitive to the interface states, this difference characterizesa-- fter injection at 1 200C interface state generation in these fluorine-implanted ox-

0.5 ........................ ,,,,,, , ,,, ides after avalanche electron injection. We will discuss this-10 -5 0 5 10 15 phenomenon in the following section.

(bI VG (v) In analyzing the data of avalanche electron injection.we have obtained capture cross sections and trap densities

FIG. 4. (a) High-frequency C-V curves before and after avalanche injec- of various electron trapping centers via measuring midgapnon (current density: 2 x 10- A/cml: do&e lO' F' -/cm: d,, =- 825 voltage shifts for both control and F+-implanted samples.A). M Quaistatic C-(V curv before and after avalanche injection (cur- as shown in Table I. The calculation method we used hererent density: 2 x 10-' A/cm2: dose. 10" F+/cm'; d. = 825 ). is the one described in Young's previous paper.9 The initial

experimental curve of voltage shift versus fluence is fittedF ÷- and Ne +-implanted samples we observed were POsi- exponentially for different fluence regions. The magnitudetive during avalanche injection. of the exponentials is related to the trap density and the

We have also observed a larger stretch out in the C- V time constant is related to the trap cross section. Thecharacteristic of our fluorinated oxides than that of control smallest cross-section trap is extracted first using the re-samples. This suggests the possibility of a relatively large sults for the largest fluences. The exponential due to theeffect on our flatband voltage measurements due to inter- above trap is subtracted from the original data and the

resultant used to characterize the next larger trap. Thisprocedure continues until all the traps are extracted fromthe initial experimental curve. These exponentials are sub-

i - 2tracted from the original data and the results are plotted tosee if a good overall fit has been obtained. We measured

TABLE 1. Capture crosa section and trap density of nonimplanted and> "fluorine-implanted oxides.

0'Ne/c rn Cros section Trap density

Cont Sample Trap No. a0O- s cm-) ,VT(10l" cm-)

0.01 Nonimplaned 1 1.83 2.31Fluence(C/cm 2 ) (midpp) 2 0.11 3.63

10"F/cm" (FIg) 1 33.20 0.91FIG. 5. The difference of flatbend voltage shifts of control, neon- and (midgap) 2 2.78 4.04hwrine-implanted oxides (d., - 600 A). The implantation was followed 3 0.24 15.10

by a N, annal at 1000 C* for 30 min.

2757 J. ApI. Phiys., Vol. 70. No. 5, 1 Septembtr 1991 D. D. Xis aind 0. R. Young 2757

B-3

three different dots on each wafer. The variations involved avalanche injection is analogous to what da Silva and co-in cross sections were around 10%. workers found in the radiation-induced effect. ' They also

We can see from the table that the trap with a cross showed that an optimum amount of fluorine exists to mm-section of 3.32x 10- 17 cm" does not appear in control imize the radiation-induced interface trap density.samples and fluorine does not create any larger traps than The process of interface state generation during elec-the above one. This might be an additional electron trap tron injection is somewhat complicated. Both fast and slowdue to fluorine. Finally, in midgap measurements, two states are generated at and in the vicinity of the Si-SiO.trapping centers with capture cross sections of 2.44 interface along with the bulk charge trapping in the oxide.x 10- 0 9 and 2.78 x 10- 19 cm 2 become dominant as seen The high-frequency C-V curve after injection will show ain the table for their higher trap densities. The reason we stretch out mainly due to the fast interface states since slowuse midgap results instead of flatband results will be dis- states are not sensitive to the stretch out but only result incussed in the following section. a rigid shift in C-V ramp.10 Thus, measuring the flatband

voltage shift gives us the total trapping effect both near theIV. DISCUSSION interface and in the bulk of the oxide, whereas the midgap

Based on above results, we have observed that fluorine voltage shift primarily characterizes the bulk trapping ef-

implantation introduces new electron traps. fect since the interface states are not charged at this point.

First, in examining current injection data of nonim- Therefore the data of capture cross sections of bulk trap-

planted. Ne -implanted, and F '-implanted samples, we ping should be extracted from the midgap measurement

found that physical damage resulting from the implanta- rather than the flatband measurement. The difference be-

tion is largely annealed out by our 1000 "C N2 heat treat- tween midgap and flatband shifts characterizes the gener-

ment. However, fluorine can react with silicon atoms in the ation of interface donor states that have a positive charge

oxide when it is implanted. It may cleave Si-O bonds to when the Fermi level is close to the valence band of theform Si-F bonds or gaseous compounds SiF 3 (Ref. 13) and silicon but become discharged as the Fermi level moves toleave the oxide with a more open structure. When electrons midgap.are injected into this oxide, the open structure may act as In addition to the interface effect, fluorinated oxideselectron traps to cause the additional flatband voltage shift. also contain bulk electron traps with cross sections on the

Second, the turnaround phenomenon during avalanche order of 10 - 17 cm 2. The neutral trap with the cross sectionelectron injection disappears for the sample with a of 2.78 X 10- I cm 2 is larger than the usual water trapF -implanted dose of 10i1 cm - 2. From calibrated SIMS ( -2 X 10 - is cm 2) in the conventional oxides. Anothermeasurements, this dose corresponds to the fluorine con- electron trap with a cross section of 2.44 X 10 - '9 cm 2 iscentration with a peak close to 1020 cma- 3 near the Si-SiO, twice as large as the one on the same order ( 1.11interface. Excess or smaller amount of the F implant X 10- 19 cm2 ) in the control sample. These traps might bedose fails to eliminate the turnaround phenomenon. created by the open structure of fluorinated oxides. How-

The turnaround effect during avalanche electron injec- ever, fluorine does not seem to create any Coulombic trapstion for conventional oxides was discussed by Feigl et al.14 or other neutral traps with cross sections larger than theyears ago. The model they proposed was that atomic hy- order of 10- 17 cm 2. It would be expected that only thedrogen released from water traps in the oxide diffuses to larger capture cross-section trap' would be of concern tothe Si-SiO2 interface. This neutral species could further device designers.release an electron at the interface to build up the anoma-lous positive charge. It was also considered to be a donor-like slow interface state10 because it only built up withlong-time constants during current injection and could becharged and discharged by the application of a moderate V. CONCLUSIONelectric field but does not respond during the usual C-Vmeasurements. For the F -implanted sample with a dose We have studied fluorinated samples by low-energyof 10I1 cm - 2, this turnaround effect does not exist. Sup- (25 keV) ion implantation. SIMS profiles show that thepose the atomic hydrogen diffuses to the Si-SiO, interface, fluorine peaks are in the same spatial position after postox-It might form H-F bonds if there are enough fluorine ions idation anneal in N, ambient at 1000 *C and are close tonear the interface. Thus, less positive charge builds up and the Si-SiO2 interface. The high-dose ( 10-5 cm - -) fluorinethe flatband voltage shift is increased. A dose of 1013 profile has a shoulder going into the bulk oxide. This indi-F 1/cm 2 does not seem to have enough fluorine to passi- cates a concentration-enhanced diffusion. Our electricalvate the hydrogen based on the above argument. Whether measurements indicate that, after avalanche injection, flu-the high-concentration ( > 1020 F ,/cm

3 ) diffusion ob- orinated oxides have additional bulk electron traps that areserver' in SIMS profile accounts for the turnaround effect not due to implantation dam-4e. The generation of inter-of the high-dose (10"1 F , /CM2) fluorine-implanted oxides face donor states in fluorinated oxides could be annealedremains to be understood. Although we do not know ex- out during avalanche injection at 120 'C. Finally. an opti-actly what chemical reaction is involved in eliminating the mum dosage ( 1014 cm - 2) of fluorine implantation existsturnaround effect by fluorine implantation, it is worth no- to suppress the turnaround effect during avalanche elec-ticing that an optimum dose does exist. This result from tron injection.

2758 J. AppIl. Phys., Vol. 70, No. 5, 1 September 1991 0. 0. Xie and 0. R. Young 2758

B-4

ACKNOWLEDGMENTS 3Y. Nishioka. Y. Ohji, K. Mukau. T. Sugano. Y. Wang. and T. P. Ma.Appl. Phys. Lett. S4. 1127 (1989).

The work is supported by a grant from Defence Nu- 4 P. J. Wright and K. C. Saraswat. IEEE Trans. Electron Devices ED-36.clear Agency (DNA). We would like to thank Fred Stevie 879 (1989).

of AT&T Bell Labs, Allentown, PA for valuable SIMS 5 P. J. Wright, M. Wong, and K. C. Saraswat. IEDM Tech. Dig., 574

measurements. We are indebted to Professor Ralph Jacco- (1987).

dine of Lehigh University whose long standing interest in e v Wright. N. Kasai, S. Inoue, and K. C. Sarswat. IEEE ElectronDevice Lett. EDL-1O, 347 (1989).

the effect of fluorine has stimulated our work in this field. 'E. H. Nicollian and C. N. Berglund, J. Appl. Phys. 41. 3052 (1970)We also wish to thank Mr. Saikumar Vivekanand and Dr. 'D. R. Young, E. A. Irene, D. J. DiMana. R. F. Dekeersmaecker. and

Jeffrey Parks of the Sherman Fairchild Lab of Lehigh Uni- H. Z. Massoud. J. Appl. Phys. 50. 6366 (1979).yfor the help in making samples for our measure- D. R. Young, J. AppL. Phys. 5Z 4090 (1981).versity 0 S. K. Lai and D. R. Young. J. Appl. Phys. 52. 6231 (1981).

ments, 'C. Sahi J. Y. Sun. and J. J. Tzou. J. Appl. Phys. 54. 944 ( 1983).

"2

C. Sah. J. Y. Sun, and J. J. Tzou, J. Appl. Phys. 54, 2547 (1983).'E. F. da Silva. Jr.. Y. Nishioka, and T. P. Ma, IEEE Trans. Nucl. Sci. '3 F. G. Kuper, 1. Th. M. De Hosson, and J. F. Verway, J. App!. Phvs. 60.NS.X 1190 (1987). 985 (1986).

:Y. Nishioka, E. F. da Silva, Jr., Y. Wang. and T. P. Ma, IEEE Electron "F. J. Feigl, D. R. Young, D. J. Di Maria, S. Lai. and J. Calise. J. Appi.Device LetL EDL-9, 38 (1988). Phys. 52, 5665 (1981).

2759 J. Appi. Pý is.. Vol. 70, No. 5, 1 September 1991 0. 0. Xie and 0. R. Young 2759

B-5

DISTRIBUTION LIST

DNA-TR-92-12

DEPARTMEf' OF DEFENSE USAISCATTN: ASOP-DO-TL

ASSISTANT TO THE SECRETARY OF DEFENSEATrN: EXECUTIVE ASSISTANT DEPARTMENT OF THE NAVY

DEFENSE ADVANCED RSCH PROJ AGENCY DEPARTMENT OF THE NAVYATTN: ASST DIR ELECTRONIC SCIENCES DIV ATTN: A NEWHOUSE NAVSEA 081ATTN: R REYNOLDS

NAVAL AIR SYSTEMS COMMANDDEFENSE ELECTRONIC SUPPLY CENTER ATTN: AIR-536T

ATTN: DESC-ENAVAL ELECTRONICS ENGRG ACTVY, PACIFIC

DEFENSE INTELLIGENCE AGENCY ATTN: CODE 250ATTN: DIW-4ATTN: DT-1B NAVAL POSTGRADUATE SCHOOL

A"TN: CODE 1424 LIBRARYDEFENSE NUCLEAR AGENCY

ATTN: RAEE TREE NAVAL RESEARCH LABORATORY2 CYS ATTN: TITL ATTN: CODE 4600 D NAGEL

ATTN: CODE 4613 A B CAMPBELLDEFENSE TECHNICAL INFORMATION CENTER2 CYS ATTN: DTIC/FDAB NAVAL SURFACE WARFARE CENTER

ATIN: CODE H21 F WARNOCKDNA PACOM LIAISON OFFICE ATTN: CODE H23 R SMITH

ATTN: DNALONAVAL SURFACE WARFARE CENTER

FIELD COMMAND DEFENSE NUCLEAR AGENCY ATTN: CODE H-21ATTN: FCPF R ROBINSON

NAVAL TECHNICAL INTELLIGENCE CTRFIELD COMMAND DEFENSE NUCLEAR AGENCY ATTN: LIBRARY

ATTN: FCNM2 CYS ATTN: FCTT W SUMMA NAVAL WEAPONS EVALUATION FACILITY

ATTN: CLASSIFIED LIBRARYJOINT DATA SYSTEM SUPPORT CTR

ATTN: C-330 NAVAL WEAPONS SUPPORT CENTERATrN: JNSV ATTN: CODE 6054 D PLATIETER

STRATEGIC AND THEATER NUCLEAR FORCES STRATEGIC SYSTEMS PROGRAMATTN: DR E SEVIN ATTN: JIM HOWARD

DEPARTMENT OF THE ARMY DEPARTMENT OF THE AIR FORCE

HARRY DIAMOND LABORATORIES AERONAUTICAL SYSTEMS DIVISIONATTN: SLCHD-NW-RP ATTN: ASD/ENSSATTN: SLCHD.NW-TNATTN: SLCHD-NW-TS AIR FORCE CTR FOR STUDIES & ANALYSIS

ATTN: AFSAA/SAKIU S ARMY MISSILE COMMAND

ATTN: AMCPM-HA-SE-MS AIR UNIVERSITY LIBRARYATTN: AUL-LSE

U S ARMY NUCLEAR & CHEMICAL AGENCY OGDEN AIR LOGISTICS CENTERATTN: MONA-NU DR D BASH AIR LOGSTCS ENE

ATT"N: OO-ALC/LMIM

U S ARMY RESEARCH OFFICE ATTN: O.O-ALCIMMDEC HARDNESS CONTROLATTN: R GRIFFITH PHILLIPS LABORATORY

U S ARMY STRATEGIC DEFENSE CMD ATTN: NTC M SCHNEIDERATTN: CSSD-SA-E ATTN: NTCTDATTN: CSSD-SD-A ATTN: PL/VTE

ATTN: PL'VTEE S SAMPSONU S ARMY STRATEGIC DEFENSE COMMAND

ATTN: CSSD-SL ROME LABORATORYATTN: RBR

U S MILITARY ACADEMYATTN: LTC AL COSTANTINE ROME LABORATORY

ATTN: ESR

Dist-1

DNA-TR-92-12 (DL CONTINUED)

UNITED STATES STRATEGIC COMMAND ALLIED-SIGNAL. INCATTN: J 63 ATTN: DOCUMENT CONTROL

WRIGHT RESEARCH & DEVELOPMENT CENTER AMPEX CORPATTN: AFWALIELE ATTN: B RICKARDATTN: WRDC/MTE ATTN: K WRIGHT

3416TH TECHNICAL TRAINING SQUADRON (ATC) ANALYTIC SERVICES. INC (ANSER)ATTN: TTV ATTN: A HERNDON

ATTN: A SHOSTAKDEPARTMENT OF ENERGY

BDM INTERNATIONAL INCDEPARTMENT OF ENERGY ATTN: D WUNSCHALBUQUERQUE OPERATIONS OFFICE

ATTN: NESD BOEING COATTN: A JOHNSTON

LAWRENCE LIVERMORE NATIONAL LAB ATTN: M ANAYAATTN: L-156 J YEE ATTN: ROS WOODATTN: L-84 G POMYKAL ATTN: D EGELKROUTATTN: W ORVIS ATTN: 0 MULKEY

LOS ALAMOS NATIONAL LABORATORY BOOZ-ALLEN & HAMILTON, INCATTN: E LEONARD ATTN: D VINCENT

SANDIA NATIONAL LABORATORIES ATTN: L ALBRIGHT

ATTN: DR T F WROBEL DIV 9341 CALIFORNIA INSTITUTE OF TECHNOLOGYATTN: L D POSEY DIV 9351 ATTN: C BARNESATTN: ORG 2146 T A DELLINATTN: ORG 2535 F SEXTON CALSPAN CORPATTN: P WINKUR DIV 1332 ATTN: R THOMPSONATTN: 1332 P WUNOKURATTN: 2140 J WOODARD CHARLES STARK DRAPER LAB, INC

ATTN: J BOYLEOTHER GOVERNMENT ATTN: N TIBBETTS

CENTRAL INTELLIGENCE AGENCY CLEMSON UNIVERSITYATTN: OSWR/NED ATTN: P J MCNULTYATTN: OSWR/STD/MTB

COMPUTER SCIENCES CORPDEPARTMENT OF TRANSPORTATION ATTN: A SCHIFF

ATTN: ARD-350DAVID SARNOFF RESEARCH CENTER, INC

NASA ATTN: R SMELTZERATTN: CODE 100.0 DR JAMES TRAINORATTN: CODE 313 V DANCHENKO E-SYSTEMS. INCATTN: CODE 724.0 M JHABVALA ATTN: MAIN LIBRARYATTN: CODE 900 E STASSINOPOULOS EATON CORP

NATIONAL INSTITUTE OF STANDARDS & TECHNOLOGY ATTN: R BRYANTATTN: P ROITMAN ELECTRONIC INDUSTRIES ASSOCIATION

DEPARTMENT OF DEFENSE CONTRACTORS ATTN: J KINN

ADVANCED RESEARCH & APPLICATIONS CORP GENERAL ELECTRIC COATTN: R ARMISTEAD ATTN: B FLAHERTY

ATTN: L HAUGEAEROSPACE CORP

ATTN: AAMRAM GENERAL ELECTRIC COATTN: C RICE ATTN: G GATI MDATTN: D S, HMUNKATTN: K G HOLDEN GENERAL ELECTRIC COATTN: M HOPKINS ATrN: DAREN NERADATTN: N SRAMEK GENERAL ELECTRIC COATTN: P BUCHMAN ATLN: J MILLERATTN: R KOGA

Dist-2

DNA-TR-92-12 (DL CONTINUED)

GENERAL RESEARCH CORP LEHIGH UNIVERSITYATTN: A HUNT 2 CYS ATTN: DR D R YOUNG

2 CYS ATTN: OR R J JACCODINEGEORGE WASHINGTON UNIVERSITY

ATTN: A FRIEDMAN LITTON SYSTEMS INCATTN: F MOTTER

H M WElL CONSULTANTS, INCATTN: H WElL LOCKHEED MISSILES & SPACE CO, INC

ATTN: F JUNGAHARRIS CORP ATTN: TECHNICAL INFORMATIION CENTER

ATTN: J C LEEATTN: J W SWONGER LOCKHEED MISSILES & SPACE CO, INC

ATTN: E HESSEEHARRIS CORPORATION ATTN: GLUM ORG 81-63

ATTN: E YOST ATTN: J CAYOT DEPT 81-63ATTN: W ABARE ATTN: L ROSSI

HONEYWELL INC ATTN: PBENE

ATTN: R JULKOWSKI LOCKHEED SANDERS, INCATTN: BRIAN G CARRIGG

HONEYWELL SYSTEMS & RESEARCH CENTER

ATTN: R BELT LOGICON R & D ASSOCIATESATTN: D CARLSON

HONEYWELL INCATTN: MS 725-5 LORAL AERONUTRONIC

HUGHES AIRCRAFT COMPANY ATTN: TECHNICAL LIBRARY

ATTN: E KUBO LTV AEROSPACE & DEFENSE COMPANYATTN: L DARDA 2 CYS ATTN: LIBRARY EM-08

IBM CORP MARTIN MARIETTA CORPATTN: DEPT L75 ATTN: J TANKE

IBM CORP ATTN: TICIMP-30

ATTN: J ZIEGLER MARTIN MARIETTA CORPIBM CORP ATTN: S BUCHNER

ATTN: N HADDAD MARTIN MARIETTA DENVER AEROSPACE

lIT RESEARCH INSTITUTE ATTN: P KASEATTN: I MINDEL ATTN: RESEARCH LIBRARY

MARYLAND, UNIVERSITY OFINSTITUTE FOR DEFENSE ANALYSES ATTN: H C LIN

ATTN: TECH INFO SERVICESMCDONNELL DOUGLAS CORP

JAYCOR ATATTN: A P MUNIE DEPT 4100ATTN: D WALTERS ATTN: R L KLOSTER

JAYCOR MCDONNELL DOUGLAS CORPORATIONA-rN: RSULLIVAN ATTN: P ALBRECHT

JAYCOR MESSENGER. GEORGE CATTN: R POLL ATTN: G MESSENGER

JOHNS HOPKINS UNIVERSITY MISSION RESEARCH CORPATTN: R MAAURER ATTN: R PEASE

KAMAN SCIENES CORP MISSION RESEARCH CORP

ATTN. DASIAC ATTN: J LUBELLATTN: E CONRAD ATTN: W WARE

KAMAN SCIENCES CORPORATION MITRE CORPORATIONATTN: DASIAC ATTN: J R SPURRIERATTN: R RUTHERFORD ATTN: M FITZGERALD

KEARFOTT GUIDANCE AND NAVIGATION CORP NATIONAL SEMICONDUCTOR CORPATTN: J D BRINKMAN ATTN: F C JONES

Dist-3

DNA-TR-92-12 (DL CONTINUED)

NORDEN SYSTEMS. INC SCIENCE APPLICATIONS INTL CORPATTN: N RIEDERMAN ATTN: P ZIELIEATTN: TECHNICAL LIBRARY

SCIENTIFIC RESEARCH ASSOC. INCNORTHROP CORPORATION ATTN: H GRUBIN

ATTN: J R SROURSUNDSTRAND CORP

PACIFIC-SIERRA RESEARCH CORP ATTN: C WHITEATTN: H BRODE

SYSTRON-DONNER.PHYSITRON INC ATTN: SECURITY OFFICER

ATTN: MARK CHRISTOPHERTECHNOLOGY DEVELOPMENT ASSOCIATES

PHYSITRON INC ATTN: R V BENEDICTATTN: MARION ROSE

TELEDYNE BROWN ENGINEERINGRAND CORP ATTN: G R EZELL

ATTN: C CRAIN ATTN: LEWIS T SMITH

RAND CORP TEXAS INSTRUMENTS, INCATTN: B BENNETT ATTN: J SALZMAN

ATTN: T CHEEKRAYTHEON CO

ATTN: DO LEE TRWATTN: JOSEPH SURRO ATTN: M J TAYLOR

RENSSELAER POLYTECHNIC INSTITUTE TRW SPACE & DEFENSE SECTORATTN: J HOWARD ATTN: C BLASNEK

ATTN: DR D R GIBSONRESEARCH TRIANGLE INSTITUTE

ATTN: M SIMONS TRW SPACE & DEFENSE SECTORATTN: D M LAYTON

ROCKWELL INTERNATIONAL CORP ATTN: HL DEPT LIBRARYATTN: V DE MARTINOATTN: YIN-BUTE YU UNISYS CORPORATION-DEFENSE SYSTEMS

ATTN: P MARROFFINOS-CUBED

ATTN: J M WILKENFELD VISIDYNE. INCATTN: C H HUMPHREY

SCIENCE APPLICATIONS INTL CORP ATTN: W P REIDYATTN: D MILLWARDATTN: DAVID LONG FOREIGN

SCIENCE APPLICATIONS INTL CORP FOA 2ATTN: J RETZLER ATTN: B SJOHOLM

SCIENCE APPLICATIONS INTL CORPATTN: W CHADSEY

Dist-4