cross-sectional scanning tunneling...

Cross-Sectional Scanning Tunneling Microscopy

Edward T. Yu

Department of Electrical and Computer Engineering, University of California at San Diego, La Jolla, California 92093-0407

Received September 23, 1996 (Revised Manuscript Received March 24, 1997)

ContentsI. Introduction 1017II. Sample Geometry and Surface Properties 1018III. Imaging and Spectroscopy 1020

a. Voltage-Dependent Imaging 1021b. Scanning Tunneling Spectroscopy 1021c. Spectroscopy on Unpinned SemiconductorSurfaces

1022

d. Current Imaging Tunneling Spectroscopy 1023IV. III−V Semiconductor Characterization 1024

a. Electronically Induced TopographicContrast

1024

b. GaAs/AlGaAs Heterostructures 1025c. GaAs/InGaAs Heterostructures 1027d. Arsenide/Antimonide Heterojunctions 1028e. Arsenide/Phosphide Heterostructures 1032f. Cross-Sectional Scanning TunnelingSpectroscopy

1034

g. Profiling of Dopant Distributions 1035h. Individual Dopant Atoms and Defects 1035i. Nanometer-Scale Optical Characterization 1036

V. Group IV Semiconductor Characterization 1037a. Group IV Epitaxial Structures 1037b. Si Device Structures 1039

VI. Summary and Conclusions 1041VII. Acknowledgments 1042VIII. References 1042

I. IntroductionSince its invention in the early 1980, scanning

tunneling microscopy1-4 (STM) and techniques suchas atomic force microscopy5 (AFM) that have evolvedfrom it have become tools of paramount importancein fundamental studies of surfaces. In addition,increasing effort has been directed towards the useof STM and related techniques to address problemsof technological interest, particularly in the area ofsemiconductor materials and devices. The extremelyhigh spatial resolution afforded by these techniqueshas led naturally to their application to a wide rangeof scientific and technological problems in semicon-ductor device physics. Indeed, their growing impor-tance in semiconductor science and technology is anatural consequence of the desire to design andfabricate ever smaller structures to improve deviceperformance or achieve new device functionality, andof the limitations of more traditional experimentaltechniques in performing detailed and comprehensivecharacterization at the nanometer to atomic scale.The unique and powerful advantage offered by

STM and other scanning probe techniques is the

ability to perform direct studies of structural, elec-tronic, and other properties of materials with ex-tremely high spatial resolution. Many other tech-niques, such as X-ray diffraction, electron diffraction,and transmission electron microscopy, typically pro-vide only indirect information about sample structureand, while offering the ability to probe certainstructural or compositional features at the atomicscale, inevitably average these properties over sub-stantially larger areas or volumes. In scanningtunneling microscopy, however, direct imaging offeatures corresponding to individual atoms on asurface has been demonstrated for a wide range ofmaterials. Other scanning probe techniques, whileusually providing somewhat lower spatial resolutionthan STM, allow greater flexibility in samples thatcan be studied, imaging conditions that can betolerated, and properties that can be characterized.Rapid progress in the development of new scanningprobe techniques and in the commercial availability

Edward Yu is Associate Professor in the Department of Electrical andComputer Engineering at the University of California, San Diego. Hereceived his A.B. (summa cum laude) and A.M. degrees in Physics fromHarvard University in 1986, and his Ph.D. degree in Applied Physics fromthe California Institute of Technology in 1991. While at Caltech he wasthe recipient of a National Science Foundation Graduate Fellowship andan AT&T Bell Laboratories Ph.D. Scholarship. He joined the faculty ofUCSD in September 1992 following a one-year postdoctoral appointmentat the IBM Thomas J. Watson Research Center in Yorktown Heights,NY. Professor Yu’s research interests are in the area of semiconductormaterials and devices, including atomic-scale characterization of semi-conductor heterostructures and devices; fundamental properties andapplications of novel heterojunction material systems; and techniques forthe fabrication and characterization of nanometer-scale semiconductordevice structures. Recent activities in his research group at UCSD haveincluded extensive studies using cross-sectional scanning tunnelingmicroscopy of atomic-scale properties, and their relation to epitaxial growthconditions and device performance, of a wide range of III−V compoundsemiconductor heterostructures. Professor Yu has been the recipient ofan Office of Naval Research Young Investigator Award, and Alfred P.Sloan Research Fellowship, and a National Science Foundation CAREERAward.

1017Chem. Rev. 1997, 97, 1017−1044

S0009-2665(96)00084-2 CCC: $28.00 © 1997 American Chemical Society

of scanning probe instrumentation has led to thewidespread application of these techniques to prob-lems of both scientific and technological importance.Cross-sectional scanning tunneling microscopy has

proven to be a particularly useful tool in the studyof atomic-scale properties in III-V compound semi-conductor heterostructures and of nanometer-scalestructure and electronic properties in Si microelec-tronic devices. As characteristic dimensions in semi-conductor devices continue to shrink and as advancedheterostructure devices increase in prominence, theability to characterize structure and electronic prop-erties at the atomic to nanometer scale has come tobe of greater and greater importance. Cross-sectionalSTM offers unique capabilities for characterizationthat, in conjunction with a variety of other, comple-mentary experimental methods, are providing newand important insights into material and deviceproperties at the atomic to nanometer scale. In thisreview, we present a description of the basic experi-mental techniques involved in cross-sectional STMand a detailed discussion of the application of thesetechniques to the characterization of III-V andGroup IV materials and device structures. Althoughnot completely exhaustive, our review includes refer-ences to and descriptions of a wide range of cross-sectional STM work reported in the literature; inaddition, detailed discussions of several results fromour own laboratory are presented to illustrate theutility of cross-sectional STM for obtaining specific,detailed information about material and device struc-ture at or near the atomic scale. We have focusedspecifically on cross-sectional STM as applied to thecharacterization of semiconductor epitaxial layers ordevice structures; studies in which the properties ofgeneric cross-sectional semiconductor surfaces pro-duced by cleaving are of primary interest are includedonly to the extent that issues of relevance to cross-sectional STM characterization of epitaxial layersand devices are addressed.

II. Sample Geometry and Surface PropertiesIn cross-sectional scanning tunneling microscopy,

semiconductor wafers are cleaved to expose a crosssection, typically a (110) or (111) plane, of epitaxiallayers grown or device structures fabricated on thewafer. Scanning tunneling microscopy experimentsperformed on the cross-sectional surface can thenreveal information about the atomic-scale morphol-ogy and electronic structure of these epitaxial layersor devices. The typical geometry of the sample andprobe tip in cross-sectional STM are shown in Figure1. Two factors are of critical importance in suchexperiments: first, the production of a cross-sectionalsurface that is atomically flat and as free as possibleof surface states or defects that can act to pin theFermi level at the surface; and second, the ability toposition the tunneling tip reliably and reproduciblyover specific regions of interest on the cross-sectionalsurface.Production of an electronically unpinned surface,

i.e., a surface at which the Fermi level is the sameas that in the bulk rather than being determinedexclusively by surface states, is a major considerationfor spectroscopic studies of electronic structure in

semiconductor heterostructures and devices. Figure2 shows schematic energy band diagrams of an STMtip and sample for (a) an unpinned surface and (b) apinned surface of a semiconductor sample. For theunpinned surface, no surface states are present inthe band gap; the surface Fermi level is therefore thesame as that in the bulk material. For the pinnedsurface, surface states in the band gap force theFermi level to assume a specific value in the bandgap, independent of its value in the bulk material.For most III-V semiconductors, an atomically flat(110) surface will not have surface states in the bandgap and the surface will therefore be unpinned;defects such as monoatomic steps or contaminationon the surface can produce dangling-bond states atenergies in the band gap, leading to partial orcomplete pinning of the surface Fermi level. Inscanning tunneling spectroscopy, described in detailin section III.b, the tunneling current is measuredas a function of the bias voltage applied betweensample and tip. For positive or negative bias voltageapplied to the sample, tunneling of electrons occurs

Figure 1. Sample and probe tip geometry in cross-sectional scanning tunneling microscopy. Typically, an(001) semiconductor wafer is cleaved to expose a (110)cross-sectional surface. Tunneling measurements are thenperformed with the STM tip positioned over the region ofinterestsgenerally consisting of exposed cross sections ofepitaxial layers grown or devices fabricated on the top ofthe waferson the cross-sectional surface.

Figure 2. Schematic energy band diagram of an STM tipand semiconductor sample for (a) an unpinned and (b) apinned semiconductor surface. For the unpinned surface,no surface states are present in the band gap and the Fermienergy at the surface assumes the same value as in thebulk material; for the pinned surface, surface states in theband gap force the Fermi energy to assume a specific valueat the surface.

1018 Chemical Reviews, 1997, Vol. 97, No. 4 Yu

between tip states and empty (conduction-band) orfilled (valence-band) states, respectively, in the sample,allowing the detailed electronic structure of thesample to be probed.A cross-sectional surface with the requisite struc-

tural quality can at present be obtained only bycleaving. III-V compound semiconductors cleaveeasily to yield a (110) surface that is atomically flat,with monolayer steps typically spaced tens to hun-dreds of nanometers apart. For Si and Si-basedstructures cleaving is more difficult, although flat(110) surfaces can be obtained. Figure 3 shows aschematic diagram of the (110), (1h10), and (001)surfaces of a zincblende semiconductor crystal; arepresents the cubic lattice constant of the semicon-ductor (e.g., 5.65 Å for GaAs). For a compoundsemiconductor, the empty and filled circles wouldcorrespond to the cation and anion species, respec-tively, e.g., the Group III (cation) and Group V (anion)species for a III-V compound semiconductor. For aGroup IV elemental semiconductor all the atoms are,obviously, of the same valence. It should be notedthat the definition of the (110) and (1h10) cross-sectional planes used in some STM studies of (110)surfaces of III-V semiconductors6 differs from thatconventionally used in STM and other studies ofepitaxial growth, in transport measurements, andother investigations. Some care must therefore betaken in comparison of STM characterization of (110)semiconductor surfaces with results of other studies.As seen in Figure 3, the unreconstructed (110) and

(1h10) surface planes contain zigzag chains of atoms,with the cation and anion species alternating alongeach chain in a compound semiconductor. It may beseen from the figure that, for a compound semicon-ductor, the (110) and (1h10) surface planes are in-equivalent and, indeed, differences are observedbetween the two in, for example, the ease with which

atomically flat cleaved surfaces can be produced. Ina Group IV semiconductor the (110) and (1h10) planesare of identical structure.It is also worth noting that a given (110) surface

plane contains atoms from only half of the (001)layers within the semiconductor crystal. As shownin Figure 4, the chains of atoms on the (110) surfaceare bonded to similar chains in a second layer locateda distance a/2x2 below the surface plane. The atomsin the second layer are not imaged directly in cross-sectional STM, but do influence the contrast observedin cross-sectional STM imaging via their effect on thelocal electronic structure of the surface layer. Somepossible effects of this aspect of the (110) surfacestructure are discussed in section IV.b.For compound semiconductors, the cation and

anion sites in the (110) and (1h10) surface planescontain unoccupied and occupied dangling-bond states,respectively. STM imaging of the GaAs (110) surfacehas shown that for positive bias voltage applied tothe sample, one obtains images of the unoccupiedorbitals corresponding to the Ga (cation) sites on the(110) surface, while for negative sample bias voltageone images the occupied orbitals corresponding to theAs (anion) sublattice on the (110) surface.6 Figure 5shows constant-current STM images from ref 6 of theGaAs (110) surface at positive and negative samplebias, clearly demonstrating bias-dependent, selectiveimaging of the Ga and As sublattices on the surface.Detailed STM studies of the GaAs (110) surface haveshown that the surface unit cell remains 1×1, withthe surface reconstruction consisting only of somebuckling and bond reorientation.6 Furthermore,scanning tunneling spectroscopy measurements haveshown that for atomically flat (110) surfaces of GaAsand most other III-V compound semiconductors,surface states are not present in the band gap or atthe band edges. These surfaces are therefore elec-tronically unpinned, allowing the bulklike electronicproperties of the sample at energies near the band

Figure 3. Schematic diagram of the zincblende crystalstructure for the (001), (110), and (1h10) lattice planes. a isthe cubic lattice constant of the crystal (5.65 Å for GaAs).Empty and filled circles represent the cation and anionspecies, respectively, of a compound semiconductor. ForIII-V compound semiconductors, there is buckling of thelattice in the (110) and (1h10) surface planes, but the surfaceunit cell remains 1×1. As-cleaved Si (110) surfaces aregenerally disordered and can undergo a variety of complexsurface reconstructions.

Figure 4. Structure of the top two atomic planes of the(110) surface for a zincblende crystal. In cross-sectionalSTM, atoms in the first layer are imaged directly, whilethose in the second layer are probed indirectly via theirinfluence on the local electronic structure of the surfaceplane.

Cross-Sectional Scanning Tunneling Microscopy Chemical Reviews, 1997, Vol. 97, No. 4 1019

edges to be probed by tunneling spectroscopy.7For Group IV elemental semiconductors, the as-

cleaved (110) surface has been found to be disor-dered,8 with complex surface reconstructions appear-ing upon annealing at high temperature.9,10 However,hydrogen passivation of the (110) surface can beaccomplished by dipping the cleaved wafer in hydro-fluoric acid or by exposure of a clean, cleaved (110)surface to atomic hydrogen in vacuo, and is requiredto obtain a (110) surface that is not strongly pinned,i.e., a surface for which the density of surface statesassociated with defects or contamination is suf-ficiently low that the electronic properties of the bulkmaterial can be probed using tunneling spectroscopy.Positioning of the STM tip over the region of

interest on the cross-sectional surface can be anontrivial matter, as one is often interested inlocating structures extending across ∼1 µm or lesson a cross-sectional surface that is a substantialfraction of a millimeter in width. One approach thathas been used to locate epitaxial layers in suchstudies is to mount the STM in a scanning electronmicroscope (SEM).11,12 By imaging the sample andtip using the SEM, it is possible to position the tipin the vicinity of epitaxially grown layers rapidly andreliably. An alternate method utilized by a numberof researchers is shown in Figure 6. The tip ispositioned as close as possible to the sample edge,and gradually moved toward the edge using thecoarse positioning mechanism. When the tip hascrossed over the sample edge, tunneling current canno longer be detected. The sample edge can then beused as a reference to locate the region of intereston the cross-sectional surface. This approach is mosteffective for samples consisting of epitaxial layers or

device structures on a conducting substrate (althoughits application is not precluded for samples on semi-insulating substrates13) and when a sharp, well-defined sample edge exists.

III. Imaging and SpectroscopyScanning tunneling microscopy provides informa-

tion about the electronic structure of the sampleunder investigation via the dependence of the tun-neling current on the tip-sample separation and onthe local electronic densities of states in the sampleand tip. A clear understanding of these dependencesis essential for interpretation of cross-sectional STMdata, and we therefore address this issue in somedetail. For studies of semiconductor surfaces, thedensity of states in the tip may usually be assumedto be relatively featureless compared to that of thesample in the energy range of interest. This allowsconstant-current images, obtained by rastering thetunneling tip across the sample surface and, at eachpoint, adjusting the tip-sample separation to yielda constant tunneling current, to be interpreted asprofiles of the local electronic density of states of thesample. Details of the probe tip shape and cleanli-ness, however, can affect the appearance of suchimages, which must therefore be interpreted withappropriate caution.Spectroscopic studies, in which the dependence of

the tunneling current on tip-sample voltage, tip-sample separation, or other parameters is measured,can be used to probe the electronic properties of thetip-sample system with spatial resolution at or nearthe atomic scale. Detailed reviews of scanning tun-neling spectroscopy can be found in refs 14-16. Inthis review we concentrate primarily on the aspectsof spectroscopy of greatest relevance to studies ofsemiconductor electronic structure. It should benoted that spectroscopic measurements, to an evengreater degree than constant-current imaging, areparticularly sensitive to probe tip structure andcleanliness.16 For example, relatively blunt tips,while not necessarily producing the highest spatialresolution in imaging, have been found to yield veryconsistent and reproducible spectroscopic results.17

Figure 5. Constant-current STM images of the GaAs (110)surface obtained at (a) +1.9 V and (b) -1.9 V sample biasvoltage. (c) Schematic diagram of atomic arrangement onthe GaAs (110) surface; solid and empty circles representGa and As atoms, respectively. Comparison of the contrastobserved in a and b within the rectangle representing theunit cell in the surface plane demonstrates that the Ga andAs sites are selectively imaged for positive and negativesample bias, respectively. The cubic lattice constant forGaAs is a ) 5.65 Å. (Reprinted from ref 6. Copyright 1987American Institute of Physics.)

Figure 6. One procedure for locating epitaxial or devicelayers in cross-sectional scanning tunneling microscopy.The initial tunneling approach is made over the substrateregion, which is easily visible in an optical microscope. Thetip is then gradually stepped toward the edge, the locationof which is signaled by a failure to detect tunneling currentwhen the tip is advanced.


a. Voltage-Dependent ImagingThe simplest extension of the constant-current

imaging measurement described above is voltage-dependent imaging, in which constant-current im-ages are obtained at different tip-sample bias volt-ages. Because of the dependence of the tunnelingcurrent on the tip and sample electronic densities ofstates, the tunneling current at different bias volt-ages will be dependent on the densities of statesintegrated over different ranges of energies. Contrastin the corresponding constant-current images willreflect these variations in electronic structure. Inparticular, positive bias applied to the sample willyield images of unoccupied electronic states on thesample surface, while negative sample bias will yieldfilled-state images. Some early examples of thisapproach include atom-selective imaging of the GaAs(110) surface6 as described in section II, and studiesof surface chemical bonding on clean and NH3-dosedSi (001) surfaces.18

b. Scanning Tunneling SpectroscopyFurther information about the local electronic

density of states on a sample surface can be obtainedusing techniques such as constant-separation spec-troscopy, in which the feedback loop that maintainsa constant tunneling current is momentarily discon-nected and the tunneling current measured as afunction of tip-sample bias voltage.19-21 A detailedtheoretical understanding of the dependence of thetunneling current on bias voltage and on the elec-tronic structure of the sample and tip then allowscurrent-voltage spectra obtained in this manner tobe interpreted in terms of the local electronic densityof states in the sample. In cross-sectional scanningtunneling microscopy performed on semiconductorheterostructures and devices, one often desires toobtain detailed, quantitative information about thestructure and electronic properties of the sampleunder investigation. Because the electronic structurein such samples is often complex and highly inho-mogeneous, especially at the nanometer scale, theo-retical modeling and simulation of tunneling current-voltage spectra are likely to play an essential role inthe detailed interpretation of both imaging andspectroscopic data.The tunneling current I(V) can be calculated using

a variety of theoretical methods, two of which wedescribe here. In one approach, developed by Tersoffand Hamann,22,23 the tunneling current is calculatedfrom the transition rate between tip and samplestates obtained using time-dependent perturbationtheory. The total current arising from electronstunneling both into and out of the sample is thengiven by

where fs and ft are the Fermi distribution functionsin the sample and tip, respectively, Mµν is thequantum-mechanical tunneling matrix element be-tween the tip state ψµ and the sample state ψν, Eµ

and Eν are the energies of the states ψµ and ψν,respectively, and δ(Eµ - Eν) is the Dirac δ function.The sum is taken over all unperturbed tip and samplestates. If the spectrum of states in the sample andthat in the tip are continuous rather than discrete,the sum in eq 1 can be converted to an integral overenergy

where Ft and Fs are the electronic densities of statesin the tip and sample, respectively. From eq 2 wesee explicitly the dependence of the tunneling currenton the densities of states in the tip and sample andon the tunneling matrix element. The tunnelingmatrix element Mµν was shown by Bardeen24 to begiven by

where the integral is computed over a surface con-tained entirely within the vacuum barrier region.

Selloni et al.25 and Lang26 have suggested that, fornonzero bias voltages applied between the sampleand tip, the tunneling current should be givenapproximately by an expression of the form

where T(E,V) is the transmission coefficient forelectron transport between the sample and tip. Thesimilarity between the expressions for I(V) in eqs 2and 4 is immediately evident, with T(E,V) and |Mµν|playing corresponding roles. If the electronic densityof states in the tip is assumed to be constant, theconductivity derived from eq 4 is then given by

The transmission coefficient in eqs 4 and 5 dependsvery strongly on the tip-sample separation and thetip-sample bias voltage; these dependences canobscure features in the current-voltage spectrumI(V) and the conductivity dI/dV arising from varia-tions in the electronic density of states. Because thesample electronic density of states, Fs(E), is mostoften the quantity of greatest interest, a procedurefor extracting Fs(E) from I(V) and dI/dV is highlydesirable.

It has been shown17,21 that a good approximationto the surface electronic density of states can beobtained from the normalized conductivity,dI/dV/(I/V). From eqs 4 and 5 we can obtain anexpression for the normalized conductivity

I(V) )2πe

p∑µ,ν[ft(Eµ) - fs(Eν + eV)] ×

|Mµν|2 δ(Eµ - Eν) (1)

I(V) ) 4πep∫Ft(E)Fs(E + eV) ×

[ft(E) - fs(E + eV)]|Mµν|2 dE (2)

Mµν ) p2

2m∫ dS (ψ*µ∆ψν - ψν∆ψ*µ) (3)

I(V) ∝ ∫0eVFs(E)Ft(E - eV)T(E, V) dE (4)

dIdV

∝ Fs(eV)Ft(0)T(eV,V) +

∫0eVFs(E)Ft(E - eV)dT(E,V)dV

dE (5)


Because the transmission coefficient appears in boththe numerator and the denominator of both inte-grands, the dependence of the normalized conductiv-ity on separation and voltage is much weaker thanthat of the current I(V) or the conductivity dI/dV, andthe normalized conductivity can be written in theform14

where A(V) and B(V) are slowly varying functions ofV. Experimental studies have confirmed that thenormalized conductivity has a much weaker depen-dence on tip-sample separation and applied voltagethan does the conductivity dI/dV, and can provide agood approximation to the local electronic density ofstates.17,21In another theoretical model, developed by Bono

and Good,27,28 the metal tip is modeled as a hemi-sphere of radius R projecting from a plane parallelto the sample surface. The probe tip is modeled as afree-electron metal, and the sample, in the case of asemiconductor, is assumed to be well described byan effective-mass model. The tunneling current fora voltage V applied to the tip is then given by28

where A ) πR2 is the effective cross-sectional areaof the tip, fs and ft are the Fermi distributionfunctions in the sample and tip, respectively, k| isthe wave vector parallel to the sample surface, E⊥ isthe energy corresponding to the wave vector compo-nent normal to the sample surface, and T(E⊥) is thetransmission coefficient. The integral is computedover all energies for which the electronic densities ofstates in the sample and tip are both nonzero. Thisapproach has been used to develop semiquantitativesimulations of current-voltage spectra for scanningtunneling spectroscopy performed on semiconductorsurfaces.7Theoretical models of the tunneling current such

as those described above provide the basis for devel-oping a detailed, quantitative understanding of thetunneling process. A comprehensive theoretical un-derstanding of tunneling processes in STM anddetailed simulations of tunneling currents for avariety of sample structures and experimental condi-tions will be essential for the most effective applica-tion of cross-sectional STM to the characterizationof semiconductor materials and devices, particularlyfor heterostructure materials and advanced devicestructures in which the composition and electronicstructure of the surface under investigation arespatially inhomogeneous and highly complex.

For studies of semiconductors, in which the pres-ence of a nonzero band gap can lead to tunnelingcurrents and conductivities ranging over severalorders of magnitude, an additional issue of impor-tance is dynamic range in the measurement oftunneling current. Wide dynamic range is especiallyimportant if features in the electronic density ofstates near or within the energy band gap are ofinterest. This issue has been addressed most ef-fectively by the development and application ofvariable-separation spectroscopic techniques, in whichthe tunneling current is measured while both thetip-sample bias voltage V and the tip-sample sepa-ration s are varied in a controlled manner.17,21,29 Forexample, a series of constant-separation current-voltage spectra can be measured at several differentseparations, or the separation may be varied con-tinuously as the voltage is ramped. The variationin tip-sample separation is then accounted for bymultiplication of the current by e2κ∆s, where κ is thetunneling decay constant and ∆s is the relativeseparation at which the tunneling current was mea-sured. This factor represents the approximate changein tunneling probability associated with the changein separation ∆s, and therefore allows tunnelingcurrents measured at different tip-sample separa-tions to be normalized, approximately, to a single,constant separation. Using this technique it ispossible to measure values of the tunneling currentover a limited dynamic range and, via the transfor-mation described above, obtain the equivalent con-stant-separation current-voltage spectrum over amuch larger range of current magnitude. A detaileddiscussion of variable-separation tunneling spectros-copy is presented in ref 16.

c. Spectroscopy on Unpinned SemiconductorSurfaces

Tunneling spectroscopy performed on unpinnedsurfaces of heavily doped semiconductors is of par-ticular relevance to the characterization of semicon-ductor heterostructures and devices by cross-sec-tional STM. Unpinned surfaces, or at least surfacesthat are not strongly pinned, are important for suchapplications because it is the electronic properties ofthe bulk material, rather than those arising predomi-nantly from surface effects, that are of primaryinterest. For such surfaces, three components havebeen found to contribute to the tunneling current:tunneling of electrons from the tip into empty con-duction-band states at large positive sample biasvoltages, tunneling of electrons from filled valence-band states into empty tip states at large negativevoltages, and tunneling involving dopant-inducedcarriers (electrons in the conduction band for n-typematerial or holes in the valence band for p-type) atvoltages for which the Fermi level in the tip is at anenergy aligned within the band gap of the sample.7Figure 7 shows current-voltage characteristics fortunneling on unpinned p-type and n-type Si surfacescalculated using the expression given in eq 4 for thetunneling current. The three components contribut-ing to the tunneling current can be seen clearly ineach case. Experimentally, it has been found thattunneling involving dopant-induced carriers can gen-

dI/dVI/V

∝

Fs(eV)Ft(0) + ∫0eVFs(E)Ft(E - eV)

eT(eV,V)dT(E,V)dV

dE

1eV∫0eVFs(E)Ft(E - eV)

T(E,V)T(eV,V)

dE(6)

dI/dVI/V

)Fs(eV)Ft(0) + A(V)

B(V)(7)

I(V) ) eA2π2h∫ dE[fs(E) - ft(E + eV)]∫d2k|T(E⊥)

(8)


erally be measured only for unpinned surfaces of veryheavily doped material and that observation of thedopant-induced current is a good indication that asurface is electronically unpinned. However, thedopant-induced tunneling current is much smallerin magnitude than the conduction- and valence-bandcomponents, particularly for material with a sub-stantial energy band gap; resolving this componentin spectroscopic measurements therefore requiresthat the tunneling current be measured over a largedynamic range.Feenstra30 has performed detailed scanning tun-

neling spectroscopy measurements on cleaved (110)surfaces of several III-V semiconductors. III-Vsemiconductors are particularly well suited for suchstudies, as they cleave easily to yield atomically flatand, for most III-V semiconductors, electronicallyunpinned (110) surfaces ideally suited for tunnelingmeasurements. Detailed tunneling current spectrawere obtained for GaAs, InP, GaSb, InAs, and InSbusing variable-separation spectroscopy. Figure 8shows normalized conductivities measured for cleaved(110) surfaces of several III-V semiconductors. Inthese spectra, the conduction-band and valence-bandedges, L-point band edges, and features associatedwith surface states can be distinguished; the energiesof these features can be determined to within (0.03eV, with shifts in energies arising from tip-inducedband bending estimated to be less than 0.1 eV fordopant concentrations higher than ∼1 × 1018 cm-3.For samples in which the dopant concentration issubstantially lower, tip-induced band bending canshift the apparent energies of the band edges by asmuch as several tenths of an electron volt, makingaccurate, quantitative interpretation of features ob-served in current-voltage spectra considerably moredifficult.

d. Current Imaging Tunneling SpectroscopySpatially resolved spectroscopic measurements can

be used to construct images that reveal atomic- tonanometer-scale variations in electronic structure onthe sample surface. In one such technique, referred

to as current imaging tunneling spectroscopy (CITS),a constant-current topographic scan with the currentstabilized at a fixed value I0 for a specific voltage V0is performed over the sample surface and, at eachpoint, a current-voltage spectrum is measured.Variations in electronic structure across the samplesurface will produce corresponding variations in thecurrent-voltage spectra; these spatial variations canthen be revealed by plotting the current measuredat specific bias voltages other than V0sso-called“current images”. Hamers et al.20 first used thistechnique to study atomic-scale variations in theelectronic structure of the 7×7 reconstruction of theSi (111) surface.Current images cannot, however, necessarily be

interpreted as direct images of particular features inthe local electronic density of states. Because thetip-sample separation at each point is determinedby the constant-current condition at a single biasvoltage V0, topographic and electronic informationmay be difficult to distinguish from each other in aCITS measurement. Contrast in current images canin fact be caused both by localized variations inelectronic structure and by changes in the back-ground of the spectrum that can arise from, forexample, the dependence of the tunneling probabilityon separation. Hamers et al.20 have argued that, forCITS performed with a constant-current bias voltageV0 that produces a topographic image reflectingprimarily the actual physical topography of the

Figure 7. Current-voltage spectra for tunneling performedon unpinned surfaces of heavily-doped (1 × 1019 cm-3)p-type and n-type Si, calculated using eq 4. Note thelogarithmic scale used for the current axis. The conduction-band, valence-band, and dopant-induced components of thetunneling current, labeled C, V, and D, respectively, canbe seen in each case.

Figure 8. Normalized conductivities measured by scan-ning tunneling spectroscopy for cleaved (110) surfaces ofn-GaAs, n-InP, p-GaSb, n-InAs, and n-InSb. Valence-band-edge and conduction-band-edge energies EV and EC, re-spectively, are indicated by dashed lines; L-valley conduc-tion-band energies are indicated by the arrows, and varioussurface states by solid lines. The energies of these featurescan be determined to within approximately (0.03 eV.(Reprinted from ref 30. Copyright 1995 American Instituteof Physics.)


sample surface, constrast in current images willreflect primarily differences in electronic structure.Satisfying this condition, however, requires a detailedprior knowledge of the physical topography of thesurface. Indeed, studies of the Si (111) 2×1 surfaceby Stroscio et al.31,32 showed explicitly that contrastin current images can arise from topographicallyinduced changes in the tunneling probability as wellas variations in the electronic density of states;however, it was also noted that these effects couldbe reduced by generating images of the normalizedconductivity (dI/dV)/(I/V) rather than simply thetunneling current I.From this discussion it should be apparent that

imaging and tunneling spectroscopy are extremelypowerful tools for probing the structure and electronicproperties of conducting or semiconducting samples,but also that considerable care is required in per-forming and interpreting such measurements. Con-tinued advances, particularly in the application oftunneling spectroscopy to advanced semiconductormaterials and device structures, are likely to requirefurther improvements in sample and probe tip prepa-ration, continued development and implementationof advanced techniques such as variable-separationspectroscopy, and perhaps most important, the de-velopment of a more complete understanding of therelationship between the measured tunneling currentand the detailed electronic structure of the sampleand tip. This is particularly true for complex, highlyinhomogeneous sample structures that are likely tobe encountered in technological applications.

IV. III−V Semiconductor CharacterizationThe development of epitaxial crystal growth tech-

niques such as molecular-beam epitaxy (MBE) com-bined with great flexibility in choice of materials hasallowed semiconductor heterostructure device con-cepts to be explored extensively in III-V compoundsemiconductor material systems. Advances in het-erostructure materials and device technology havedepended to a considerable degree upon studies of,and consequently improvements in, crystal growthtechniques and the resulting structure of epitaxiallygrown layers. As material quality has improved andas device structures have grown in sophistication, ithas become increasingly apparent that an under-standing of, and ultimately control over, growthprocesses and epitaxial layer structure at the atomicscale will be needed for realization and optimizationof advanced semiconductor heterostructure and nano-scale devices.III-V semiconductors are in many ways ideally

suited for study by cross-sectional STM. Cleaving of(001) wafers can produce (110) cross-sectional sur-faces with atomically flat planes extending over tensto hundreds of nanometers. To avoid contaminationof the cleaved surface, cleaving under ultra-highvacuum conditions is necessary. Furthermore, foratomically flat (110) surfaces of most III-Vmaterials33,34sGaP being a notable exception35-37stheFermi level at the surface is unpinned, and tunnelingspectroscopy performed on the surface can yieldinformation about the electronic properties of thebulk material.

The first cross-sectional STM experiments in whichactual epitaxial and device structures were investi-gated were performed by Muralt et al.38-40 In thesestudies, scanning tunneling microscopy and poten-tiometry were used to probe potential distributionsin a GaAs pn junction and a GaAs/Al0.40Ga0.60Asdouble-heterojunction laser structure. Salemink etal. performed the first studies in which a GaAs/Alx-Ga1-xAs heterojunction interface was resolved di-rectly by scanning tunneling spectroscopy,41 andsubsequently extended this work to obtain the firstatomically resolved images of GaAs/AlxGa1-xAsheterostructures.42-45 To date, a number of groupshave reported scanning tunneling microscopy studies,both in air and under ultrahigh-vacuum conditions,of epitaxially grown structures in a wide range ofmaterial systems.

a. Electronically Induced Topographic ContrastIn constant-current cross-sectional images of het-

erostructures, apparent “topographic” contrast be-tween the different materials present in the hetero-structure is typically observed, with corrugationamplitudes ranging from less than 1 Å to over 100Å. This contrast is generally ascribed to differencesin the electronic properties of the constituent materi-als, rather than to actual physical topography. Fac-tors such as differing energy band gaps, dopant andcarrier concentrations, and electron affinities cancontribute to the apparent topographic contrastbetween materials in a heterostructure, with varia-tions in band-edge energy from one layer to the nextgenerally being the dominant contributing factor tothe observed heterostructure contrast.To explain the electronically induced topographic

contrast observed in such images, we note that thetunneling current integral in eq 2 includes factorsfor the densities of states in the tip and sample andfor the difference between the Fermi distributionfunction in the sample and that in the tip. At biasvoltages typically applied for tunneling in semicon-ductors, the main contribution to the tunnelingcurrent will therefore involve an integral of thesample density of states over the energy rangebetween the Fermi energies in the sample and tip.This contribution to the tunneling current will varyfrom layer to layer in a heterostructure because ofthe changes in electronic structure from one materialto the next. In constant-current imaging, thesevariations are manifested as changes in the apparenttopographic height of the surface as the tip-sampleseparation is adjusted to yield a constant tunnelingcurrent at each point. In complex structures suchas superlattices, in which the electronic structurechanges considerably from that in bulk material,these effects can become more complex; however, thebasic principle remains valid.Figure 9 shows the band-edge energy structure and

a cross-sectional STM image, obtained at negativesample bias, of an InAs/InAs1-xSbx superlattice. Asshown in the band-edge energy diagram, tunnelingof valence-band electrons from the sample into emptytip states occurs over a wider range of energy in theInAs1-xSbx layers than in the InAs layers for negativebias voltage applied to the sample. Consequently,


maintaining a constant tunneling current requiresthat the tip-sample separation be increased whenthe tip is located over the InAs1-xSbx layers, leadingto an apparent difference in topographic height.Furthermore, an additional periodic structure isvisible within each InAs1-xSbx layer. This contrastarises from the periodic compositional structurewithin each InAs1-xSbx layer induced by the modu-lated MBE technique46,47 employed during the growthof the InAs1-xSbx layers in the superlattice. In thisgrowth technique, the average group V compositionin the InAs1-xSbx alloy layer is controlled by growingalternating layers of InAs and InSb rather than arandom alloy, resulting in the periodic structureevident in the cross-sectional STM image. Detailedstudies of this structure are described in section IV.d.The wide range of height differences from layer to

layer reported by different investigators for cross-sectional STM imaging of semiconductor heterostruc-tures is likely to be due at least in part to variationsin tip and sample cleanliness. If excessive contami-nation is present on the tip or sample, the apparentbarrier height will be anomalously low and thechange in tip-sample separation required to main-tain a constant tunneling current will be much larger

than that for an ideal vacuum barrier. Furthermore,different rates of oxidation in air for different materi-als in a heterostructure can lead to actual physicaltopography that is correlated with composition. Thelatter phenomenon has been observed in AFM studiesof GaAs/AlxGa1-xAs heterostructures.48 In addition,STM studies of InGaAs/InP multiple quantum wellsconducted in air revealed that observations of large(several nanometers or more) heterostructure cor-rugation amplitudes were correlated with low appar-ent barrier heights and with increases in ambienthumidity.49

Cross-sectional STM offers unique capabilities forcharacterization of atomic-scale structure in semi-conductor alloy layers and at heterojunction inter-faces. It has been noted45,50 that because STMimaging is sensitive primarily to the one or twouppermost layers on the surface, atomic-scale fea-tures in structure and electronic properties can beresolved with considerably greater detail than inother techniques. High-resolution X-ray diffractioncan in some circumstances yield information aboutatomic-scale interface structure in superlattices,51,52but this information is inevitably averaged over manylayers and also over a large area of the sample.Transmission electron microscopy (TEM), particu-larly the chemical lattice imaging technique,53-55 canprovide information about atomic-scale interfacestructure, but averaging that occurs through thethickness of the samplestypically on the order of 100Å or moreslimits the technique’s ability to resolvefeatures such as atomic-scale interface roughness andfluctuations in composition.Atomically resolved imaging of cross-sectional sur-

faces, which thus far has been achieved only forIII-V semiconductors and under ultra-high vacuumconditions, is therefore a uniquely powerful tool forexploring alloy layer and heterojunction interfacestructure at the atomic scale. Constant-currentimages of cross-sectional surfaces are sensitive pri-marily to composition within the atomic plane at thesurface, and to a limited degree to composition a fewto several layers below the surface. Compositionalfluctuations at the atomic scale, which would beundetectable using other characterization techniquesdue to averaging effects, are therefore clearly visiblein high-resolution cross-sectional images. By com-bining information obtained from cross-sectionalSTM with results obtained using complementarycharacterization techniques such as X-ray diffraction,electron diffraction, electron microscopy, and electri-cal and optical measurements, one is able to developa detailed understanding of atomic- to nanometer-scale properties of semiconductor materials anddevices, confirm that this information is representa-tive of properties at the larger length scales relevantfor most device structures, and ultimately establishcorrelations between atomic-scale material propertiesand the behavior of actual devices.

b. GaAs/AlGaAs HeterostructuresCharacterization of GaAs/AlxGa1-xAs heterojunc-

tions using cross-sectional STM performed underultrahigh vacuum conditions has proven to be apowerful probe of atomic-scale alloy layer and inter-

Figure 9. Current flow, and the resulting topographicimage for negative sample bias voltage, in cross-sectionalscanning tunneling microscopy performed on an InAs/InAs1-xSbx heterostructure. For negative bias voltage ap-plied to the sample, the tip Fermi level, EF

tip, will be belowthe sample Fermi level, EF

sample, and tunneling of electronswill occur from filled (valence-band) states in the sampleinto empty states in the tip. The arrows provide a sche-matic indication of the tunneling current arising fromcarriers in the sample at energies between the Fermi levelsin the sample and tip. Because the electronic structurechanges from one layer of the heterostructure to the next,the tunneling current measured for a fixed tip-sampleseparation would vary accordingly. In constant-currentimaging, the tip-sample separation is adjusted to yield aconstant tunneling current, resulting in electronicallyinduced topographic contrast as seen in the image.


face structure. Studies of GaAs/AlxGa1-xAs hetero-structures either cleaved and imaged in air56 orcleaved in air and imaged in vacuum57 have yieldedconstant-current images in which apparent contrastbetween the GaAs and AlxGa1-xAs layers was ob-served, but with substantially lower resolution thanthat achievable for cleaving and imaging performedentirely under ultrahigh vacuum conditions.The first atomically resolved images of semicon-

ductor heterostructures in cross-sectional scanningtunneling microscopy were obtained by Salemink etal. in studies of GaAs/AlxGa1-xAs heterostructuresunder ultrahigh-vacuum conditions.42-45 In theseand subsequent studies, Albrektsen et al.45,58 wereable to determine the locations of heterojunctioninterfaces with a precision of (1 unit cell, allowingmonolayer steps at the interfaces to be resolved.These studies yielded direct images of step bunchingand consequent terrace formation at GaAs/AlxGa1-x-As interfaces for heterostructures grown on GaAs(001) substrates with a 2° miscut in the [11h0] direc-tion.45 Cross-sectional STM has also revealed thepresence of atomic-scale fluctuations in the composi-tion of ternary III-V alloys. In studies of GaAs/Alx-Ga1-xAs heterojunctions, Albrektsen et al.58 observednanometer-scale “mottling” in constant-current im-ages of the AlxGa1-xAs layers, indicative of nonuni-formity in the electronic structure of the alloy thatwas tentatively attributed to fluctuations in composi-tion. Detailed imaging of a GaAs/Al0.35Ga0.65As/GaAsstructure by Salemink and Albrektsen59 revealedclustering of Al atoms in the ternary alloy with atypical length scale estimated to be two to three unitcells (10-15 Å). Further studies by Johnson et al.60confirmed these observations of interface roughnessand alloy fluctuations, and revealed Al-rich regionsrunning along the [1h12] and [11h2] directions (asobserved on the (110) cross-sectional surface); thelatter observation was interpreted as possible evi-dence for preferential growth of Al-rich regionsaligned along these directions in the AlxGa1-xAslayers. Figure 10 shows an atomically resolved filled-state cross-sectional image of a GaAs/AlxGa1-xAs/GaAs heterostructure, obtained at -2.1 V samplebias and 0.1 nA tunneling current. The AlxGa1-xAslayer is clearly lower topographically, on average,than the surrounding GaAs layers, consistent withthe valence-band-edge discontinuity at the hetero-junction interface. In addition, local atomic-scalecontrast within the AlxGa1-xAs layer can arise fromthe bonding of the surface As atoms to both Al andGa atoms on the surface and one layer below,although the atomic-scale features visible in theimage correspond to occupied orbitals on the GroupV (As) sublattice of the (110) cross-sectional surface.Mottling in the AlxGa1-xAs layer may therefore beinterpreted as reflecting local variations in the Alconcentration within the ternary alloy layer.A dependence of interface structure in GaAs/Alx-

Ga1-xAs heterojunctions on growth sequence andgrowth conditions in molecular-beam epitaxy has alsobeen observed by cross-sectional STM. Smith etal.61,62 have performed ultrahigh-vacuum cross-sectional STM studies of GaAs/Al0.3Ga0.7As superlat-tices and observed that interface roughness ampli-

tudes were greater by approximately 10 Å for theinverted (GaAs-on-Al0.3Ga0.7As) interface comparedto the normal (Al0.3Ga0.7As-on-GaAs) interface. Stud-ies of AlAs/GaAs short-period superlattices revealedthat incorporation of 30 s growth interrupts at eachinterface substantially improved interface abruptnesscompared to that achieved with 5 s growth inter-rupts.62,63 Also addressed in these studies was theinfluence of bonding configuration on the (110)surface on atomic-scale contrast observed by cross-sectional STM.63 Figure 11 shows the four possiblearrangements of Al, Ga, and As atomic layers for a(110) cross section of a GaAs/AlAs interface. Thenumbers to the left of each first-level As layerindicate the number of Al atoms to which the Asatoms in that layer are bonded. Since only the first-level As atoms are imaged directly for cross-sectionalSTM imaging performed at negative sample biassasis often the case for III-V semiconductorssfourpossible sequences of bonding configurations forsuccessive first-layer As rows may be present at agiven interface. Thus, even for interfaces that, asshown in Figure 11, are all perfectly abrupt, theapparent interface abruptness observed in a filled-state cross-sectional STM image may vary slightlywith the atomic configuration at the (110) surface.

Figure 10. Atomically resolved filled-state cross-sectionalSTM image of a GaAs/AlxGa1-xAs/GaAs heterostructure,obtained at -2.1 V sample bias and 0.1 nA tunnelingcurrent. Topographic line scans along the rows indicatedare shown below the image. The average topographicheight within the AlxGa1-xAs layer is clearly lower thanin the surrounding GaAs, reflecting the change in electronicstructure across the interface. Mottling in the AlxGa1-xAslayer reflects local variations in Al concentration. (Re-printed from ref 60. Copyright 1993 American Institute ofPhysics.)


c. GaAs/InGaAs HeterostructuresThe detailed structure of GaAs/InxGa1-xAs inter-

faces is of considerable importance for heterostruc-

ture devices such as pseudomorphic modulation-doped field effect transistors (MODFET’s) andquantum-well and quantum-wire lasers. For elec-tronic devices, such as MODFET’s, based on lateraltransport of carriers near a heterojunction interface,local composition variations and interface roughnesscan significantly affect carrier mobility and, conse-quently, device performance. For quantum-well- andquantum-wire-based optoelectronic devices, precisecontrol over layer thickness and alloy composition isessential to ensure optimum device performance atspecific wavelengths. Clustering of In and asym-metric interface broadening have been observed bySTM in cross-sectional studies of GaAs/InxGa1-xAsheterostructures performed by Zheng et al.64 Figure12 shows an empty-state cross-sectional STM imageof a pseudomorphic GaAs/InxGa1-xAs multiple-quan-tum-well structure grown by MBE. Chainlike clus-ters containing 2-3 In atoms aligned in the [001]direction are observed in coherently strained In0.20-Ga0.80As layers grown on GaAs (001) substrates. Inaddition, segregation of In on the growth surface wasfound to produce asymmetric broadening at theGaAs/In0.20Ga0.80As interfaces. For In0.20Ga0.80As grownon GaAs, the interfaces were found to be 2-4 atomic

Figure 11. Arrangement of atoms at the (110) surface foran AlAs/GaAs heterostructure. The precise location of eachAlAs/GaAs interface, indicated by the dashed lines, withrespect to the first-level As atoms influences the bondingconfigurations for the first-level As rows near the interface,and may consequently affect the apparent interface abrupt-ness observed in a cross-sectional STM image. Such dif-ferences would be artifacts of the cross-sectional STMtechnique, since all the interfaces shown are atomicallysharp.

Figure 12. Cross-sectional STM image, obtained at asample bias of +2.0 V and a tunneling current of 0.5 nA,of a GaAs/In0.2Ga0.8As multiple-quantum-well structure.Asymmetric broadening of the GaAs/In0.2Ga0.8As interfacesarising from In segregation during growth is observed, asis the formation of 2-3 atom In clusters (indicated byarrows) aligned along the [001] direction. (Reprinted fromref 64. Copyright 1994 American Institute of Physics.)


layers wide; for the opposite growth sequence, incor-poration of excess In segregated on the growthsurface was believed to be responsible for producinginterfaces extending over 5-10 atomic layers. Thisasymmetry in the structure of the GaAs/In0.20Ga0.80-As interfaces is qualitatively apparent in the imageshown in Figure 12, and has been confirmed byquantitative determination of In concentration as afunction of position in the [001] growth direction.64In cross-sectional STM studies performed by Pfister

et al.65 of InxGa1-xAs quantum wells and InxGa1-xAsquantum wires embedded in (AlAs)4(GaAs)4 super-lattice barriers grown on V-groove-patterned GaAs(001) substrates, substantial segregation of In fromthe InxGa1-xAs quantum well and incorporation of Ininto the subsequently grown (AlAs)4(GaAs)4 barrierwas also observed. However, no In clustering beyondthat expected statistically for a random alloy wasdetected, nor was evidence seen for preferential Inincorporation at the center of the quantum wire.Studies of InxGa1-xAs quantum-wire structures byPfister et al.65 have also demonstrated the possibilityof performing quantitative chemical identification ofatoms in the surface and first subsurface layers ofthe (110) cleaved cross-sectional plane for ternaryIII-V alloys. Figure 13 shows empty-state and filled-

state images of an InxGa1-xAs quantum-wire struc-ture surrounded by (AlAs)4(GaAs)4 superlattice bar-riers. In the empty-state image, bright spots areobserved that correspond to unoccupied In orbitalson the (110) surface; fainter features in the empty-state image appear to correspond to In atoms in thefirst subsurface layer. In the filled-state image,bright spots appear to correspond primarily to filledAs orbitals associated with As atoms back-bonded toIn atoms in the first subsurface layer which causethe surface As atoms to protrude from the surfaceplane, yielding the observed contrast. These inter-pretations are supported by comparisons of filled-state and empty-state images of identical areas: thebright As sites are correlated with faint double-sitefeatures in the empty-state image, and surface Inatoms visible in the empty-state image produce onlyweak contrast in the filled-state image for As atomsto which they are bonded.

d. Arsenide/Antimonide HeterojunctionsMixed-anion material systems, in which the Group

V as well as the Group III constituent changes acrossa heterojunction interface, are particularly rich inpossibilities for complex interfacial structure. Inthese material systems, two distinct bond configura-tionssboth of which correspond to a structurallyperfect heterojunctionscan be present at each inter-face.66 The InAs/AlSb heterojunction, for example,can contain either InSb-like or AlAs-like bonds at theinterface. Furthermore, it has been demonstratedthat interfacial composition can exert a considerableinfluence on electron transport66 and optical proper-ties67 in InAs/AlSb quantum wells and superlattices,and on the superlattice band gap and backgrounddoping characteristics of InAs/Ga1-xInxSb superlat-tices.68

Cross-sectional STM studies of InAs/GaSb69-71 andInAs/Ga1-xInxSb72,73 superlattices have begun to pro-vide insight into some of these issues. A cross-sectional constant-current STM image of an InAs/Ga0.75In0.25Sb superlattice is shown in Figure 14.Atomic-scale interface roughness is clearly visible inthe image, as is a dependence of the InAs/Ga1-xInx-Sb interface structure on growth sequence. Detailedstudies of InAs/GaSb and InAs/Ga1-xInxSb superlat-tices have revealed clear evidence, in both spectros-copy69,70 and imaging,72,73 of a dependence of inter-facial electronic structure on interface growthsequence; such an asymmetry could arise from anumber of sources. In studies of InAs/GaSb super-lattices, Feenstra et al.69-71 attributed the observedasymmetry to incorporation of Sb riding on thegrowth surface into InAs grown on GaSb. Lew etal.72,73 have suggested that an additional factor maybe the asymmetry in superlattice electronic structurethat can arise if different interfacial bond configura-tions are formed for different growth sequences.Quantitative information about atomic-scale inter-

face roughness can be obtained very effectively bycross-sectional STM. Feenstra et al.69-71 have per-formed quantitative studies of interface roughnessin InAs/GaSb superlattices as a function of MBEgrowth conditions. Fourier spectra computed forheterojunction interface profiles obtained from InAs/

Figure 13. (a) Empty-state and (b) filled-state images ofa 112 Å × 250 Å region of an InxGa1-xAs quantum-wirestructure surrounded by (AlAs)4(GaAs)4 superlattice bar-riers. The images are of the exact same cross-sectional areaof the quantum-wire structure. (c) Topographic profilesfrom the empty-state and filled-state images along the linesindicated by the arrows. (Reprinted from ref 65. Copyright1995 American Institute of Physics.)


GaSb superlattices were found to be consistent witha Lorentzian spectral distribution expected to becharacteristic of interfaces with random step distri-butions. A significant dependence of interface struc-ture on growth sequence was observed, with InAs-on-GaSb interfaces being significantly rougher thanthose in which GaSb was grown on InAs. Thisdependence of interface roughness on growth se-quence was attributed to incorporation of excess Sbfrom the GaSb growth surface into the subsequentlygrown InAs layer. Consistent with this interpreta-tion, growth by atomic-layer epitaxy and growth athigher substrate temperatures (460 °C vs 380 °C)were found to yield significant reductions in interfaceroughness.As an illustration of the quantitative analysis of

interface roughness using cross-sectional STM, wedescribe studies by Lew et al.74 of the dependence ofinterface roughness in InAs/Ga1-xInxSb superlatticeson orientation and growth sequence. Figure 15ashows a cross-sectional STM image of a 17 Å InAs/50Å Ga0.75In0.25Sb superlattice grown on a GaSb (001)substrate. Interfaces between the InAs (dark) andGa0.75In0.25Sb (bright) layers can be extracted fromimages of both (110) and (11h0) surfaces by enhancingthe grey-scale contrast in the image and employingan edge-detection algorithm to identify the interfaces.Figure 15, parts b and c, show representative inter-faces extracted in this manner from images of the(110) and (11h0) surfaces, respectively. Analysis ofimages in this manner allows detailed, quantitativeinformation about interface structure, and its depen-dence on orientation and growth sequence, to be

obtained. Considerable care must be taken, however,to ensure that interface profiles extracted from cross-sectional STM images are independent of the methodof interface definition and extraction. While themethod employed by Lew et al.74 was checked toensure that this criterion was satisfied, it is necessaryto note that, at the atomic scale, inherent ambiguityin the definition of heterojunction interfaces can arisefrom the fundamental structure of the interfaceregion, e.g., if atomic-scale intermixing or grading ispresent; this should be distinguished from the ex-perimental uncertainty that can arise from thetechnique used to extract the interface profile fromthe STM image.Interface roughness in these structures can be

quantified by performing a Fourier analysis of theinterface profiles extracted from the cross-sectionalSTM images. Given a real-space interface profilea(x), one may calculate a discrete Fourier transformto obtain the spectral components, Aq, of the rough-ness amplitude. The roughness amplitude compo-nents Aq are then given by

Figure 14. Cross-sectional STM image of an InAs/Ga0.75-In0.25Sb strained-layer superlattice, obtained at a samplebias of -1.5 V and a tunneling current of 0.1 nA. The brightand dark regions correspond to the Ga0.75In0.25Sb and InAslayers, respectively. Features corresponding to individualatomic orbitals on the Group V sublattice of the cleaved(110) surface are clearly visible, allowing phenomena suchas atomic-scale interface roughness and atomic cross-incorporation to be investigated.

Figure 15. (a) Cross-sectional STM image of a 17 Å InAs/50 Å Ga0.75In0.25Sb superlattice, obtained at a sample biasof -1.5 V and a tunneling current of 0.1 nA. (b) Profiles ofheterojunction interfaces in the superlattice extracted froma (110) cross-sectional STM image of the superlattice; (c)heterojunction interface profiles extracted from a (11h0)cross-sectional STM image.


where q ) 2πn/L, L is the length of the interfaceprofile, and d is the spacing between data pointsalong the interface. A random distribution of stepsat the interface is expected to yield an exponentialcorrelation in real space and a Lorentzian powerspectral distribution75

where ∆ is the roughness amplitude and Λ thecorrelation length. Power spectral distributions cal-culated for interface profiles extracted from cross-sectional STM images were found to be well describedby the Lorentzian function of eq 10.Figure 16 shows power spectral distributions cal-

culated for interface profiles extracted from (110) and(11h0) cross-sectional STM images of the InAs/Ga0.75-In0.25Sb superlattice structure. Also shown in thefigure are Lorentzian distributions fitted to theexperimental data. The results shown in the figureincorporate analyses of several dozen interface pro-files extracted from several images of both (110) and(11h0) cross-sectional surfaces. Separate distributions

were calculated for InAs-on-Ga0.75In0.25Sb and Ga0.75-In0.25Sb-on-InAs interfaces for each orientation, al-lowing the influence on interface structure of growthsequence as well as orientation to be determined. Asindicated in Figure 16c, interfaces imaged in the(11h0) cross-sectional plane were found to be signifi-cantly rougher, on average, than those imaged in the(110) plane. In addition, a substantial dependenceof interface roughness on growth sequence wasobserved. For imaging in both the (110) and the (11h0)cross-sectional planes, interfaces in which Ga0.75In0.25-Sb was grown on InAs were found to be significantlyrougher than those in which InAs was grown onGa0.75In0.25Sb.The observation that interface roughness is greater

in the (11h0) plane than in the (110) plane is consistentwith the formation on the epitaxially grown surfaceof islands elongated preferentially along the (11h0)direction during epitaxial growth, a phenomenon thathas been widely observed in planar studies of III-Vepitaxial growth.76-80 Islands or terraces elongatedalong the (11h0) direction present on the surfaceduring growth would be expected to produce corre-sponding features in epitaxially grown interfaces.Cross-sections of such interfaces would then exhibitgreater roughness along the [110] direction, i.e., inthe (11h0) plane, than along the [11h0] direction.The dependence of interface roughness on growth

sequence is consistent with the formation of inter-faces with mixed stoichiometry for Ga1-xInxSb grownon InAs. For infrared detectors based on InAs/Ga1-x-InxSb superlattices, it has been found that devicecharacteristics may be optimized by deliberate growthof superlattice structures in which the InSb-likecomposition is present at both interfaces.81 However,obtaining InSb-like composition at the Ga1-xInxSb-on-InAs interface requires that the As layer presenton the InAs surface during growth by MBE bereplaced by a layer of Sb prior to initiating depositionof Ga1-xInxSb. While studies by reflection highenergy electron diffraction and X-ray photoelectronspectroscopy indicate that substantial substitution ofSb for As can be achieved by exposure of the As-terminated InAs surface to a pure Sb flux,82,83 X-raydiffraction measurements indicate that only partialsubstitution of Sb for As occurs under typical growthconditions.84 As shown schematically in Figure 17,the presence of mixed stoichiometry at a Ga1-xInx-Sb/InAs interface will inevitably lead to interfaceroughness, with the spectral distribution of rough-ness observed dependent upon the spatial distribu-tion of Sb and As on the epitaxially grown surfaceand, consequently, at the heterojunction interface.Additional insight into cross-incorporation of As

and Sb in arsenide/antimonide heterostructures canbe gained by examination of InAs/InAs1-xSbx hetero-structures, in which the issue of interface abruptnessis not complicated by the possible presence of mul-tiple interface stoichiometries.85 A cross-sectionalSTM image of an InAs/InAs1-xSbx superlattice grownby modulated molecular-beam epitaxy was shown inFigure 9. As indicated in Figure 18, the InAs1-xSbxlayers grown by modulated MBE consisted of four-period 7.8 Å InAs/5.2 Å InSb superlattice structures.The resulting periodic structure within the InAs1-xSbx

Figure 16. Power spectral distributions for InAs-on-Ga0.75-In0.25Sb and Ga0.75In0.25Sb-on-InAs interfaces imaged in the(a) (110) and (b) (11h0) cross-sectional planes. Both experi-mental data (symbols) and Lorentizian fits (solid anddashed lines) to the experimentally observed spectra areshown. (c) Amplitudes (∆) and correlation lengths (Λ)corresponding to the fitted Lorentzian spectra plotted in aand b.

Aq )2

N∑n)0

N-1

a(nd)e-iqd (9)

|Aq|2 ) 1L

2∆2(Λ/2π)

1 + (qΛ/2π)2(10)


layer is clearly visible in Figure 9. The nominalaverage composition of the InAs1-xSbx layer is InAs0.60-Sb0.40, based on the ratio of InAs to InSb layerthickness; X-ray diffraction measurements, however,yielded an average composition of InAs0.76Sb0.24,indicating the presence of substantial incorporationof As into the InSb-like layers.Figure 19 shows a detailed, filled-state constant-

current image of a single InAs0.76Sb0.24 layer withinthe InAs/InAs0.76Sb0.24 superlattice, obtained at asample bias voltage of -2.4 V and a tunnelingcurrent of 0.1 nA. Several significant features maybe noted in the image. The modulation in contrastarising from the modulated MBE growth techniqueis again clearly visible. However, substantial na-nometer-scale lateral inhomogeneity is observed withinthe InSb-like layers, suggesting the presence of

compositional fluctuations arising from incorporationof As into the InSb-like layers. A dependence ofinterface abruptness on growth sequence is alsovisible in the image: the InAs0.76Sb0.24-on-InAs in-terface is seen to be atomically abrupt, while theInAs-on-InAs0.76Sb0.24 interface appears to suffer fromcompositional grading across one to two monolayers.Lew et al.85 interpreted this grading to be evidencefor Sb segregation at the growth surface duringgrowth of the InAs0.76Sb0.24 layer followed by incor-poration of Sb into the subsequently grown InAslayer, a phenomenon also found by Feenstra etal.69-71 to occur during growth of InAs/GaSb super-lattices by MBE. Finally, detailed examination ofcross-sectional STM images of the InAs/InAs0.76Sb0.24superlattice revealed that the average contrast, andby extension the average Sb content, increased fromthe first to third InSb-like layers within the InAs0.76-Sb0.24 alloy, then decreased in the fourth and finalInSb-like layer. This observation indicates that Asincorporation into the InSb-like layers is most severefor the initially grown InSb-like layer, then graduallydecreases. The reduced contrast observed in the finalInSb-like layer is consistent with Sb segregation atthe surface and incorporation into the subsequentlygrown InAs layer, and also suggests the occurrenceof a slight amount of As diffusion into the underlyingInSb-like layer.

Figure 17. Schematic diagram of atomic layer structurefor an InAs/Ga1-xInxSb superlattice in which the stoichi-ometry is InSb-like at the InAs-on-Ga1-xInxSb interface andhas mixed InSb-like and Ga1-xInxAs-like character at theGa1-xInxSb-on-InAs interface. The dashed lines delineatethe interfaces between the layers. The presence of mixedstoichiometry at the Ga1-xInxSb-on-InAs interface cancreate additional interface roughness arising from thelateral inhomogeneity in composition.

Figure 18. Schematic diagram of shutter sequence em-ployed during growth of an InAs/InAs1-xSbx superlatticeby modulated MBE. The lines indicate periods duringwhich each shutter is open. In the modulated MBE growthtechnique, an InAs/InSb short-period superlattice is grownrather than a random InAs1-xSbx alloy. This approach hasbeen demonstrated to yield higher structural quality andgreater control over average alloy layer composition thancan be achieved by solid-source MBE growth of a randomgroup V alloy.

Figure 19. Filled-state cross-sectional STM image, ob-tained at a sample bias of -2.4 V and a tunneling currentof 0.1 nA, of an InAs0.76Sb0.24 alloy layer grown by modu-lated MBE surrounded by InAs. An averaged topographicline scan is shown below the image. Modulation of thecomposition in the growth direction and nanoscale lateralinhomogeneity within the InAs0.76Sb0.24 alloy layer, alongwith a growth sequence dependence in interface abrupt-ness, are visible in the image and the topographic profile.


e. Arsenide/Phosphide HeterostructuresAtomic-scale cross-incorporation and variations in

interface composition are also significant issues forheterojunctions containing combinations of arsenideand phosphide compounds, e.g., InP/In1-xGaxAs, InP/In1-xGaxAsyP1-y, and GaAs/In1-xGaxP. In cross-sectional scanning tunneling microscopy of InGaAs/InP multiple quantum well structures performed inair, topographic contrast between the InGaAs andInP layers was observed in constant-current imageswith corrugation amplitudes of up to 50-150 Å.49,56,86Cross-sectional STM studies of In1-xGaxAs/InP mul-tiple-quantum well structures under ultrahigh-vacuum by Osaka and Kato87 yielded atomicallyresolved images of the In1-xGaxAs quantum-welllayers; furthermore, the detailed contrast observedin constant-current images was found to vary withthe dopant concentration in the sample: for positivebias voltages applied to the tip, the shape of thetopographic profile reflected the shape of the valence-band-edge profile in the sample at a given dopantconcentration. In addition, the corrugation ampli-tude measured as a function of bias voltage was used,in conjunction with theoretical modeling of the tun-neling current, to estimate a valence-band offset ∆Ev) 0.60∆Eg between In1-xGaxAs and InP.The detailed atomic-scale structure of alloy layers

and heterojunction interfaces in various arsenide/phosphide heterostructures has been studied quiteextensively by cross-sectional STM. In early studiesof an InP/In0.25Ga0.75As0.5P0.5 heterojunction, Sale-mink and Albrektsen88 obtained direct images ofatomic-scale fluctuations in Group V composition inthe quaternary alloy. In addition, recent studies ofvarious arsenide/phosphide heterostructures by cross-sectional STM have revealed a considerable depen-dence of atomic-scale interface and alloy layer struc-ture on growth conditions, with significant implica-tions for the realization and optimization of arsenide/phosphide heterostructure devices. In cross-sectionalSTM studies of In1-xGaxAs/InP double-barrier reso-nant-tunneling diodes grown by metalorganicMBE,89,90 a dependence of both interface roughnessand interface chemical composition on growth se-quence was observed. As in the studies of interfaceroughness in InAs/GaSb69-71 and InAs/Ga1-xInxSb74superlattices, interface profiles were extracted fromconstant-current images of the In1-xGaxAs/InP reso-nant-tunneling-diode structure, and spatial fre-quency spectra were computed from these profiles.Fitting of the resulting spectra to a Lorentzianspectral distribution function revealed that the in-terfaces in which InP was grown on In1-xGaxAs werecharacterized by significantly shorter correlationlengths and larger roughness amplitudes than theIn1-xGaxAs-on-InP interfaces. In addition, constant-current images of the double-barrier regions revealeda growth-sequence-dependent asymmetry in the in-terfacial electronic structure which was believed tobe a consequence of a difference in bonding types atthe In1-xGaxAs-on-InP and InP-on-In1-xGaxAs inter-faces resulting from the growth procedures used.89,90

Lew et al.91,92 have demonstrated that, undercertain growth conditions, considerable atomic cross-incorporation and interfacial grading can occur at

arsenide/phosphide heterojunction interfaces. Dif-fusion during growth interrupts was observed to bethe most significant contributing factor, particularlyas the growth temperature was increased. Figure 20shows a schematic diagram and a filled-state cross-sectional STM image, obtained at a sample bias of-2.5 V and a tunneling current of 0.1 nA, of astructure in which growth of GaAs by gas-sourceMBE at a substrate temperature of 550 °C has beeninterrupted periodically and the surface exposed for40 s to a pure P2 flux. Growth interrupts and avariety of anion-switching procedures are often em-ployed to optimize interface abruptness during growthof mixed arsenide/phosphide heterojunction inter-faces. The filled (valence-band) states associatedwith P in the GaAs1-xPx layers produced during the

Figure 20. (a) GaAs/GaAsP heterostructure created byperiodic exposure of a GaAs surface to a pure P2 flux duringgrowth by gas-source MBE. (b) Cross-sectional STM imageof the structure shown in a, obtained at a sample bias of-2.5 V and tunneling current of 0.1 nA. The samplestructure was grown at a substrate temperature of 550 °C.Topographic contrast indicated in the image reveals sub-stantial incorporation of phosphorus into the GaAs duringthe exposure to P2.


growth interrupts are expected to have a lowerenergy than those associated with As;93 the darkareas in the image therefore correspond to regionsof phosphorus incorporation into the GaAs layersduring the growth interrupts. Nanometer-scale lat-eral variations in phosphorus incorporation are clearlyvisible; furthermore, the contrast within the phos-phide interlayers, averaged in the lateral direction,gradually increases in the [001] growth direction atthe lower interface while decreasing more abruptlyat the upper interface, suggesting that diffusion ofphosphorus into the underlying GaAs is more sig-nificant than carryover of phosphorus into GaAsgrown after the interrupt. Figure 21 shows a filled-state cross-sectional STM image of a nominallyidentical structure grown at 450 °C, obtained at asample bias of -2.5 V and tunneling current of 0.1nA. Comparison of the STM images of structuresgrown at 450 and 550 °C shows that the contrastassociated with the GaAs1-xPx layers formed duringthe growth interrupts, and consequently the amountof phosphorus incorporated during the interrupts, isdramatically reduced by growth at lower substratetemperatures, consistent with the interpretation thatphosphorus incorporation occurs primarily via diffu-sion into the underlying GaAs.Figure 22 shows a filled-state cross-sectional STM

image, obtained at a sample bias of -2.5 V and atunneling current of 0.1 nA, of a 67 Å In1-xGaxAs/142 Å InP multiple-quantum-well structure grownby low-pressure organometallic vapor-phase epitaxy(OMVPE) at a substrate temperature of 650 °C.92 Thenominal composition of the ternary alloy layers, asdetermined by X-ray diffraction, is In0.53Ga0.47As;however, substantial variations in heterostructure

contrast, and by extension in composition and elec-tronic structure, appear to be present within theIn0.53Ga0.47As layers. The In1-xGaxAs and InP layersappear bright and dark, respectively, consistent withthe lower valence-band-edge energy in InP. Sub-stantial contrast is observed in the InP layer nearthe InP-on-In1-xGaxAs interface, extending approxi-mately 50-75 Å into the InP layer. Arsenic incor-poration into InP layers grown on In1-xGaxAs, arisingfrom the presence of substantial residual As in thegrowth chamber, is believed to be a significant factorcontributing to the contrast in the InP layer near theInP-on-In1-xGaxAs interface. Little such contrast isobserved in the InP layer near the In1-xGaxAs-on-InP interface. The heterojunction interface structureis thus clearly dependent on growth sequence, withthe In1-xGaxAs-on-InP interfaces appearing to befairly abrupt and the InP-on-In1-xGaxAs interfacesexhibiting substantial interfacial grading. In addi-tion, considerable inhomogeneity and a clear asym-metry in electronic structure in the direction ofgrowth are visible in the In1-xGaxAs layer: themaximum contrast within the In1-xGaxAs layer rela-tive to InP occurs near the In1-xGaxAs-on-InP inter-face, with the contrast gradually decreasing in thegrowth direction. These features could arise fromintermixing of either Group III or Group V constitu-ents; however, the reduced contrast near the InP-on-In1-xGaxAs interface combined with the observa-tion of phosphorus diffusion into GaAs during growthinterrupts in gas-source MBE91 suggests that phos-phorus diffusion into the In1-xGaxAs layers duringInP growth might contribute substantially to theobserved asymmetry in samples grown by MOVPE.This interpretation of the cross-sectional STM data

is consistent with calculations of electrostatic bandbending within an In1-xGaxAs/InP heterostructure inwhich atomic cross-incorporation as described abovehas occurred.94 Figure 23 shows the conduction-band-edge and valence-band-edge energies calculatedfor an In1-xGaxAs/InP quantum-well structure inwhich cross-incorporation of As and P has occurred.Specifically, a self-consistent solution to Poisson’sequation was obtained for a four-layer n-type (n ) 1

Figure 21. Cross-sectional STM image of a GaAs/GaAsPheterostructure nominally identical to that shown in Figure20, but grown at a substrate temperature of 450 °C. Theimage was obtained at a sample bias of -2.5 V and atunneling current of 0.1 nA. The contrast between the GaAsand GaAsP layers is extremely faint, indicating a muchlower level of phosphorus incorporation than that presentat higher growth temperatures.

Figure 22. Filled-state cross-sectional STM image, ob-tained at sample bias voltage of -2.5 V and 0.1 nAtunneling current, of a 67 Å InGaAs/142 Å InP multiple-quantum-well structure grown by organometallic vapor-phase epitaxy. As incorporation in the InP layer near theInP-on-InGaAs interface is clearly visible. Contrast ob-served within the InGaAs layer suggests the occurrence ofP diffusion into the InGaAs layer near the InP-on-InGaAsinterface. Arrows indicate the region from which thetopographic line profile shown in Figure 23 was extracted.


× 1017 cm-3) periodic structure with each periodconsisting of the following layers: (1) 82 Å InP; (2)60 Å InAsxP1-x, with x increasing quadratically from0 to 0.2; (3) 57 Å InGaAsxP1-x, with x increasinglinearly from 0.9 to 1; and (4) 10 Å InGaAs. Theindividual layer widths in this structure were chosento correspond approximately to the regions of differ-ent contrast observed in the cross-sectional STMimage. Band-gap bowing parameters from the lit-erature for the ternary and quaternary compoundswere included when available;95,96 otherwise valueswere obtained by linear interpolation between binarycompounds. Valence-band offsets were assumed tovary linearly with composition, and the full effectsof strain were included.97 As seen in the figure, theband-edge profile calculated for a structure incorpo-rating the effects of As carryover into InP and Pdiffusion into In1-xGaxAs at the InP-on-In1-xGaxAsinterface is qualitatively consistent with the elec-tronically induced contrast observed in the cross-sectional STM image of the InP/In1-xGaxAs multiplequantum-well structure grown by OMVPE.

f. Cross-Sectional Scanning TunnelingSpectroscopyDetailed information regarding nanometer-scale

electronic properties of semiconductor heterostruc-tures can be obtained from cross-sectional scanningtunneling spectroscopy. Using scanning tunnelingspectroscopy, Salemink et al.41 were able to detectthe two-dimensional electron gas formed at a GaAs/Al0.5Ga0.5As interface with spatial resolution ap-proaching 1 nm. In subsequent, detailed spectro-scopic measurements on GaAs/AlxGa1-xAs hetero-

junctions,42,43 the variation in valence-band-edgeenergy near the GaAs/AlxGa1-xAs interface was probedquantitatively with nanometer-scale resolution. Fig-ure 24 shows valence-band-edge energies near aGaAs/AlxGa1-xAs interface determined using cross-sectional scanning tunneling spectroscopy; also shownare the valence-band-edge energy profile calculatedin the vicinity of the interface in bulk material andat the (110) surface, the latter taking into accounttip-induced band-bending in the sample that occursduring the spectroscopic measurement. A cleartransition in the valence-band-edge energy, occurringover approximately 35-50 Å in the growth direction,is evident in the spectroscopic measurements. Thedistance over which the transition from the GaAs tothe AlxGa1-xAs valence-band-edge energy occurredwas taken to reflect the inherent width of theelectronic transition across the heterojunction inter-face. The observed shift in valence-band-edge energyat the GaAs/AlxGa1-xAs interface is seen to be quan-titatively consistent with the expected band offset forthe heterojunction, provided that tip-induced elec-trostatic band bending in the sample is taken intoaccount.Detailed, quantitative interpretation of spectro-

scopic measurements on unpinned (110) surfaces ofIII-V semiconductors can be complicated consider-ably by tip-induced band-bending in the semiconduc-tor, as described in section III. However, complica-tions in quantitative interpretation of spectroscopicdata arising from tip-induced band bending can to alimited extent be circumvented by performing spec-troscopic measurements on surfaces for which theFermi level is pinned. Such an approach is viableprovided that the pinning does not completely ob-

Figure 23. (a) Conduction-band-edge and valence-band-edge profiles calculated for an InGaAs/InP multiple quan-tum well structure in which As carryover into the InP layerand P diffusion into the InGaAs layer have occurred at theInP-on-InGaAs interface. (b) Topographic profile of theInGaAs/InP structure extracted from the filled-state imageof Figure 22, obtained at -2.5 V sample bias. Nonuniformcontrast within the quantum-well layer and evidence of Asincorporation in the barrier layer near the InP-on-InGaAsinterface can be seen in the topographic profile.

Figure 24. Valence-band-edge energies across a (110)crosssection of a GaAs/AlxGa1-xAs interface determined byscanning tunneling spectroscopy (symbols). The solid line(D) is a fit to the spectroscopic data. The dashed curve (C)is the calculated valence-band-edge profile for the hetero-junction in bulk material, and the solid curve (S) is thecalculated profile at the (110) surface, taking into accounttip-induced band bending that occurs during the spectro-scopic measurement. (Reprinted from ref 43. Copyright1992 American Institute of Physics.)


scure the electronic properties of the bulk material.Gwo et al.98 have explored this approach, performingimaging and spectroscopy on sulfide-passivated cross-sectional surfaces of GaAs/Al0.3Ga0.7As heterostruc-tures. Their experiments were interpreted as sug-gesting that the pinning induced by sulfide passivationdid not alter the shift in valence-band-edge energybetween the GaAs and Al0.3Ga0.7As layers as mea-sured by scanning tunneling spectroscopy, and thatspectroscopic measurements were in fact able toprobe the bulklike electronic properties of the sample.Pinnington et al.99 have also used sulfur passivationof the cross-sectional surface to perform imaging andspectroscopy, in air, of GaAs/AlxGa1-xAs heterostruc-tures and pn junctions. While these experimentsdemonstrate that sulfur passivation layers can fa-cilitate certain types of spectroscopic measurementson (110) surfaces of III-V semiconductors, the abilityto probe atomic-scale compositional structure andelectronic properties is considerably reduced. Con-trast between alternating layers in a heterostructurecan generally be observed provided the layers are ofsufficient thickness. However, atomic-scale detailsin the structure of alloy layers and heterojunctioninterfaces are generally obscured.

g. Profiling of Dopant Distributions

Characterization of dopant distributions and, in-deed, of individual dopant atoms in III-V semicon-ductors has also been accomplished using cross-sectional STM. Muralt38 performed the first cross-sectional studies of GaAs pn junctions, using scanningtunneling potentiometry to probe the potential dis-tribution across a junction. Feenstra et al.100 carriedout the first detailed imaging and spectroscopicstudies of GaAs pn junctions, observing electronicallyinduced topographic contrast in which the p-typeregions appeared to be ∼2 Å lower than the n-typeregions for a sample bias voltage of -2 V. Alsoobserved were unexpectedly large widths for then-type layers compared to those expected from thegrowth procedure, and correspondingly smaller widthsfor the p-type regions; this was interpreted as evi-dence for diffusion of Si from the n-type layers intothe surrounding p-type regions. Current-voltagespectroscopy allowed the n-type, p-type, and depletedregions to be identified unambiguously, confirmingthe interpretation of contrast in the constant-currentimages. Images and spectroscopic measurementsconsistent with these results were obtained in studiesof GaAs pn junctions by Gwo et al.101 Measurementsof tunneling conductivity have been used by Feenstraet al.102,103 to facilitate the study of pn junctions oncross-sectional surfaces containing relatively highdensities of monoatomic steps. In constant-currentSTM imaging of these surfaces, the electronicallyinduced topographic contrast between p-type andn-type regions is most likely diminished by partialpinning of the Fermi level,102 and further obscuredby actual topographic features on the stepped surface.Tunneling conductivity, however, was observed tovary with carrier type, allowing p-type, n-type, anddepleted regions to be distinguished. In measure-ments of tunneling conductivity, the distance overwhich the transition in conductivity from the p-type

to the n-type value occurred was found to be close tothe expected depletion layer width for the junction,thereby providing a direct measure of the depletionlength and, consequently, the dopant concentration.Pinnington et al.99 and Tseng et al.104 have per-

formed imaging and spectroscopy on GaAs pn junc-tions in air, using sulfide passivation to preventoxidation of the cross-sectional surface. In bothcases, electronically induced topographic contrast wasobserved between the n-type and p-type regions, withthe p-type regions appearing 4-8 Å lower than then-type regions for approximately -4 V bias appliedto the sample; this degree of topographic contrast iscomparable to that observed for sulfur-passivatedGaAs pn junctions studied by cross-sectional STMunder ultrahigh-vacuum conditions.98 In addition,Tseng et al. observed a discrepancy between theapparent widths of the p-type and n-type regions incross-sectional images and the layer widths expectedfrom the growth procedure, with the n-type regionsbeing substantially wider and the p-type regionsnarrower than intended. This was attributed to ashift in the apparent position of the pn junctionarising from tip-induced band bending effects, ratherthan diffusion of Si from the n-type material into theadjacent p-type regions as postulated by Feenstra etal.100 However, Gwo et al.101 observed that the Fermilevel on sulfur-passivated cross-sectional surfacesappeared to be pinned, which would tend to minimizetip-induced band-bending effects. While further workis required to understand fully the relative influenceof actual sample structure and factors such as surfacepreparation, tip quality, and tip-induced band bend-ing, it is clear that cross-sectional imaging andspectroscopic measurements with the STM offer thepossibility of performing dopant profiling with aspatial resolution approaching 1nm, far better thanthe resolution achievable with more traditional char-acterization techniques.

h. Individual Dopant Atoms and DefectsIn addition to nanometer-scale profiling of dopant

distributions, cross-sectional STM has been used tostudy the properties of individual dopant atoms,defects, and other atomic-scale structures in GaAs.It has been demonstrated in high-resolution imagingof cleaved (110) surfaces of GaAs that Be and Zndopant atoms in GaAs up to four layers below thecross-sectional surface can be detected, appearing assmall hillocks in filled-state images of the cross-sectional surface; from the size and shape of thefeatures, dopants at the surface and one layer belowthe surface could be distinguished from each otherand from dopants farther below the surface.105,106 Theappearance of these hillocks was consistent with thatexpected on the basis of theoretical modeling of theeffect of a Coulomb potential on the tunneling cur-rent. Figure 25 shows a constant-current cross-sectional STM image of a Be-doped GaAs sample.Individual dopant atoms, labeled 1-9, appear as“hillocks” in the image; the features labeled “a” and“b” were interpreted as As vacancies and adatoms,respectively. Detection of dopants in this mannerwas used to obtain dopant profiles in modulation-doped and δ-doped GaAs/AlGaAs epilayers.107,108


Results consistent with these were reported by Zhenget al., who obtained images of both Si donors109 andZn acceptors110 in GaAs using cross-sectional STM.In addition to dopants, features such as As antisite

defects111 and As precipitates112 in low-temperature-grown GaAs have been studied by cross-sectionalSTM. Figure 26 shows constant-current cross-sectional STM images of defects observed in low-temperature-grown GaAs, in which a high concen-tration of As-related point defects were expected tobe present. The four images shown in the figure wereinterpreted as each corresponding to an individualAs antisite defect located zero, one, two, or threeatomic planes below the (110) surface for images A,B, C, and D, respectively. The extended nature ofthe contrast observed in the images was believed tobe a consequence of the spatially extended tails ofthe As antisite wave function.

i. Nanometer-Scale Optical CharacterizationOptical measurements such as photoluminescence

can serve as highly sensitive probes of certain atomic-to nanometer-scale properties of semiconductor het-

erostructures. For example, photoluminescence mea-surements performed on quantum-well structures bySakaki et al.113,114 were able to provide informationabout atomic-scale interface roughness and layerwidth fluctuations in GaAs/AlxGa1-xAs heterostruc-tures. However, these photoluminescence measure-ments were sensitive primarily to interface roughnessand layer width fluctuations at lateral length scalescomparable to the typical exciton size (assumed tobe ∼250 Å for GaAs114). Furthermore, such mea-surements inevitably average such atomic-scale prop-erties over much larger length scales, and provideonly indirect, and typically incomplete, informationabout local optical properties and their dependenceon atomic-scale structural and electronic propertiesof materials and devices. Scanning probe techniqueshave made possible the direct characterization ofoptical properties at or near the nanometer scale.Detection of luminescence induced by recombina-

tion of carriers injected during imaging in a scanningtunneling microscope has allowed energy-band pro-files and carrier transport characteristics in GaAs/AlxGa1-xAs quantum wells to be probed at the na-nometer scale.115-117 In these experiments, p-typeGaAs/AlxGa1-xAs quantum-well structures werecleaved in situ in an ultrahigh-vacuum scanningtunneling microscope system. Light emission arisingfrom recombination of electrons injected during con-stant-current imaging with holes present in thep-type material was measured, allowing lumines-cence properties to be probed with nanometer-scalespatial resolution. Higher luminescence intensitywas observed while tunneling into the GaAs layerscompared to that observed when injecting electronsinto the AlxGa1-xAs layers, with the ratio of lumi-nescence intensity in the different layers increasingfor increasing magnitude of the bias voltage. Lumi-nescence contrast from GaAs quantum wells asnarrow as 20 Å was observed. For tunneling ofelectrons into both GaAs and AlxGa1-xAs layers,luminescence at photon energies corresponding to the

Figure 25. (a) Cross-sectional STM image of a Be-dopedGaAs sample, obtained at a sample voltage of -2.1 V anda tunneling current of 0.1 nA. Individual Be dopant atoms,labeled 1-9, appear as “hillocks” in the image. (b) Topo-graphic line scans through individual hillocks in a. (c)Distribution of areas under topographic line scans for eachhillock. Larger areas correspond to dopant atoms closer tothe surface; it was estimated that dopants up to fivemonolayers (∼10 Å) below the surface produced visiblefeatures in the image. (Reprinted from refs 105 and 106.Copyright 1993 American Institute of Physics.)

Figure 26. Constant-current STM images of defectsobserved in low-temperature-grown GaAs. Images A, B, C,and D were believed to correspond to As antisite defectslocated zero, one, two, or three atomic planes, respectively,below the (110) cross-sectional surface. (Reprinted from ref111. Copyright 1993 American Institute of Physics.)


GaAs band gap was observed, suggesting that mostof the injected electrons diffused to the GaAs quantum-well layers before recombining. Indeed, measure-ment of luminescence intensity as a function oflateral tip distance from a GaAs quantum well wasused to estimate the diffusion length in AlxGa1-xAs,and the value obtained was found to be in reasonableagreement with previously reported values. Detailedexamination of the dependence of luminescence in-tensity on bias voltage revealed that the onset ofluminescence occurred at bias voltages for which thetip Fermi level was just above the lowest energyconduction-band state in the sample; this contributedto a substantial difference at low to moderate biasvoltages between luminescence intensity for electroninjection into GaAs and that for electron injection intoAlxGa1-xAs. This difference was, as expected, con-sistent with contrast observed in luminescence imag-ing of the GaAs/AlxGa1-xAs quantum-well structures.A mapping of the threshold voltage for the onset ofluminescence as the STM tip was scanned across aGaAs/AlxGa1-xAs multiple-quantum-well structurewas found to correspond to the profile of the conduc-tion-band-edge energy in the barrier regions, and ofthe lowest confined-state energy in the quantumwells, with energy variations observable at the na-nometer scale.117Given the eminent suitability of III-V semiconduc-

tors for characterization by cross-sectional STM andthe increasing realization of the importance of atomic-scale material and device properties in heterostruc-ture and nanoscale device technology, it is likely thatcross-sectional STM studies of epitaxial layers anddevice structures will play a significant role in thedevelopment and application of advanced semicon-ductor heterostructures and nanostructures. Furtherdevelopment and effective utilization of this tech-nique will require effort in a number of areas. Adetailed understanding of the relationship betweenmeasured tunneling current (and, consequently, im-ages and spectroscopic data) and the actual atomic-scale structure of the sample, including aspects suchas local alloy-layer composition, ordering effects,interface stoichiometry, defects, and impurities, isessential; theoretical simulations of tunneling experi-ments are likely to be particularly useful. In addi-tion, systematic comparisons of cross-sectional STMstudies with data from various other characterizationtechniques such as transmission electron microscopy,X-ray diffraction, diagnostics such as RHEED per-formed during epitaxial growth, and optical andelectrical measurements on a wide range of actualdevice structures will aid in the interpretation ofSTM data and ensure that high-resolution measure-ments performed on relatively small areas of a devicedo indeed provide information representative of theentire device structure.

V. Group IV Semiconductor CharacterizationThe overwhelming importance of Si in microelec-

tronics technology provides a powerful motivation forapplying scanning tunneling microscopysin both theplanar and cross-sectional geometriessto the char-acterization of Si and Si-based materials. Extensivestudies using STM have been performed of funda-

mental properties of various Si surfaces and ofprocesses such as epitaxial growth, adsorption ofatoms and molecules, and etching on Si surfaces.Cross-sectional STM studies of epitaxial and devicestructures have been more limited, largely becausepreparation of a cross-sectional surface suitable forstudy by STM is much more difficult for Si than forIII-V semiconductors: first, Si (001) wafers are moredifficult to cleave than III-V wafers, although withsome care flat (110) cleaved surfaces can be obtained;and second, the as-cleaved (110) cross-sectional sur-face has been found to be atomically disordered withelectronic structure dominated by dangling-bondstates.118 Despite these complications, nanometer-scale cross-sectional characterization of Si-basedmaterials and devices has been achieved. Detailedatomic-scale studies of such structures will, however,require the ability to prepare cross-sectional surfacescomparable in quality to cleaved (110) surfaces ofIII-V compounds.

a. Group IV Epitaxial StructuresDetailed cross-sectional imaging and spectroscopy

of semiconductor materials and device structuresrequires, ideally, the production of an atomically flat,electronically unpinned surface. For III-V com-pounds, this is readily achieved by cleaving in vacuo.As mentioned above, however, cleaving of Si (001)wafers to obtain atomically flat (110) cross-sectionalsurfaces is highly problematic. Furthermore, the as-cleaved Si (110) surface tends to be disordered, andthe Si (110) surface can exhibit a highly complexsurface reconstruction. Nevertheless, Johnson andHalbout119 have shown that an ex situ cleave of a Si(001) wafer followed by immersion in dilute HF canyield a (110) cross-sectional surface that is nearlyatomically flat and is electronically unpinned. Usingthis sample preparation procedure, Yu et al. haveperformed electronic profiling of epitaxially grown Sipn junctions,120 imaging and spectroscopy of Si/Si1-x-Gex superlattices,121 and two-dimensional electronicprofiling of Si MOSFET structures.122 Spatiallyresolved spectroscopic measurements obtained oncross sections of pn junctions grown by MBE revealedsystematic variations in current-voltage spectra asthe junction region was traversed; an analysis ofthese variations allowed the potential distribution inthe junctions to be mapped out with spatial resolutionon the order of 10 nm. In addition, changes in thepotential distribution across the pn junction occurringas a reverse bias voltage was applied across thejunction could be detected using potentiometric tech-niques. Spectroscopic techniques applied by Yu etal.122 have been demonstrated to yield two-dimen-sional profiles of carrier type in Si MOSFET struc-tures with spatial resolution on the order of 10-20nm.Cross sectional STM imaging of Si-based hetero-

structures has yielded results qualitatively similarto those obtained in studies of III-V heterostruc-tures, but with reduced spatial resolution and sen-sitivity to local electronic structure because of thelower quality of the Si (110) cross-sectional surface.Electronically induced contrast between Si andSi1-xGex layers similar to that seen in studies of


III-V heterostructures has been observed in constant-current cross-sectional images obtained under ultra-high-vacuum conditions of Si/Si1-xGex superlattices121cleaved and passivated ex situ, and in images ofpassivated Si/Ge multilayers obtained in air.123 Fig-ure 27 shows a (110) cross-sectional STM image of a260 Å Si/75 Å Si0.76Ge0.24 (001) strained-layer super-lattice, obtained at a sample voltage of +1.5 V and atunneling current of 0.1 nA.121 The (110) cross-sectional surface was obtained using an ex situ cleavefollowed by immersion in a dilute HF solution asdescribed above. The Si0.76Ge0.24 layers appear asbright vertical stripes in the image; three Si0.76Ge0.24layers can be seen, with the underlying Si bufferlayer visible in the left part of the image. Severalmonoatomic steps on the cleaved surface, runningfrom the lower left to the upper right in the image,can also be seen. The presence of monolayer-levelroughness on the (110) surface reduces the visibilityin the gray-scale image of the electronically inducedcontrast between different materials in the super-lattice, although the Si and Si0.76Ge0.24 layers can stillbe distinguished.Figure 28 shows more detailed cross-sectional STM

images of the Si/Si0.76Ge0.24 superlattice obtained at0.1 nA tunneling current and sample bias voltagesof (a) -2.5 V (filled-state) and (b) +1.8 V (empty-state). The characteristic shape of the superlatticecorrugation is clearly different for images obtainedat positive and negative sample bias, confirming that

the contrast is indeed electronically induced.124 Alsovisible in the figure is monolayer-level roughness ofthe cross-sectional surface induced by the cleavingand HF etching procedure. Such roughness is likelyto induce partial pinning of the surface. However,the electronic properties of the bulk material are stilldiscernible in tunneling spectroscopy measurements,as evidenced by the spectroscopic studies of pnjunctions described in refs 120 and 122. Indeed,

Figure 27. (a) (110) cross-sectional STM image of a 260Å Si/75 Å Si0.76Ge0.24 strained-layer superlattice, near theinterface between the superlattice and the underlying Sibuffer layer. The growth direction is from left to right inthe image. The image was obtained at a sample bias of+1.5V and tunneling current of 0.1 nA. The bright regionscorrespond to the Si0.76Ge0.24 layers, and the dark to Si.Monolayer roughness of the surface is also visible in theimage. (b) Line scan across the superlattice; the Si0.76Ge0.24layers appear as peaks in the topography. (Reprinted fromref 121. Copyright 1992 American Institute of Physics.)

Figure 28. A (110) cross-sectional STM image of a 260 ÅSi/75 Å Si0.76Ge0.24 strained-layer superlattice at 0.1 nAtunneling current and sample bias voltages of (a) -2.5 Vand (b) +1.8 V. Below each image are single topographicline scans traversing the superlattice layers shown in eachimage. The growth direction in both images is from left toright.


current-voltage spectra obtained for sequences ofpoints traversing multiple periods of the Si/Si1-xGexsuperlattice appeared to reflect nanometer-scale spa-tial variations in the electronic structure of thesuperlattice.121

Alternate methods for obtaining high-quality Si(110) cross-sectional surfaces have also been ex-plored. Lutz et al.118 have performed detailed studiesof (110) surfaces cleaved and hydrogen passivatedunder ultrahigh-vacuum conditions, and have com-pared measurements from these surfaces to thoseobtained from samples cleaved ex situ and hydrogenpassived by dipping in HF. Imaging and spectros-copy of (110) surfaces obtained by in situ cleavingrevealed that the as-cleaved surface is disordered,with dangling-bond states in the band gap; substan-tial variations in local electronic structure wereevident from spectroscopic measurements performedon as-cleaved surfaces. Upon exposure to atomichydrogen, mostly passivated (110) surfaces wereobtained, although completely homogeneous passi-vation of the entire surface was not observed. Spec-troscopic measurements on epitaxial p-type andn-type Si layers indicated that surfaces prepared inthis manner were partially pinned; however, p-typeand n-type material could clearly be distinguished.Compared to the ex situ cleave and HF passivationprocedure, the in situ approach yielded somewhatless homogeneous passivation but offered the pos-sibility of reduced contamination and consequentlymore stable tunneling characteristics.A comparison of images and spectroscopic data

obtained from (110) cross-sectional surfaces of GroupIV and III-V semiconductor heterostructures makesevident the importance of surface quality in high-resolution cross-sectional scanning tunneling micros-copy. The atomically flat, electronically unpinned(110) surfaces obtainable by in situ cleaving for III-Vsemiconductors allow structure and electronic prop-erties to be probed with atomic resolution, as de-scribed in section IV. For Group IV semiconductors,resolution and sensitivity to bulk electronic proper-ties are currently limited to the nanometer scale byroughness, contamination, or inhomogeneous passi-vation on the (110) cross-sectional surface. Improve-ments in cross-sectional surface quality might beachievable either using improved passivation tech-niques for cleaved (110) surfaces or by cleaving toexpose a (111) cross-sectional surface, for whichetching in buffered HF has been shown to yieldatomically flat surfaces.125 Such improvements wouldbe particularly valuable for improving spatial resolu-tion and sensitivity to electronic properties in cross-sectional STM studies of Group IV materials anddevices.

b. Si Device StructuresAs feature sizes in advanced Si microelectronic

devices such as metal-oxide-silicon field-effect tran-sistors (MOSFET’s) enter the ultrasubmicron, andeventually the nanometer-scale, size regime, theability to characterize and ultimately to controlfabrication processes and the resulting structure ofa device with nanometer-scale spatial resolution hasbecome essential. Indeed, successful fabrication of

ultrasubmicron Si MOSFET’s with gate lengths ap-proaching 0.1 µm (100nm) and below will require theability to characterize dopant and carrier profiles inthe source and drain junction regions of a MOSFETwith vertical and lateral resolution approaching 10nm or better. Nanometer-scale characterization ofand control over the source and drain junctionprofiles in ultrasubmicron MOSFET structures areespecially critical for low-voltage, low-power applica-tions that are becoming of central importance inmicroelectronics. However, such information can beextremely difficult to obtain using traditional char-acterization techniques such as spreading resistanceprofiling and secondary ion mass spectrometry (SIMS);the vertical resolution required for effective junctionprofiling is approaching the limits of these tech-niques, and the lateral resolution that can be achievedis far from adequate for this application.A number of particularly promising approaches to

this problem based on various scanning probe mi-croscopies have begun to be explored. In this sectionwe describe studies in which scanning tunnelingmicroscopy and spectroscopy performed in the cross-sectional geometry have been used to delineate theelectronic structure of Si MOSFET’s with extremelyhigh spatial resolution. Related techniques such asscanning capacitance microscopy,126-130 scanning re-sistance microscopy,131-133 Kelvin probe forcemicroscopy,134-137 and other scanning probe tech-niques138-140 have also been used for extensive high-resolution characterization of Si pn junctions in boththe planar and cross-sectional geometries. As featuresizes in state-of-the-art commercial microelectronicdevices such as Si MOSFET’s continue to shrink,scanning probe techniques are likely to assumeincreased importance for performing the nanometer-scale electronic and structural characterization ofdevice structures and fabrication processes that willbe essential for reliable, reproducible, and uniformfabrication of large-scale circuits based on ultra-submicron devices.Kordic et al.141,142 performed the first cross-sec-

tional STM studies of Si pn junctions, using thescanning tunneling microscope to study cross sectionsof implanted junctions exposed by cleaving. Experi-ments were conducted in which sample cleaving andsubsequent tunneling measurements were performedboth in air141 and under ultrahigh vacuum condi-tions.142 In both cases tunneling measurements werepeformed directly on cleaved (110) surfaces ratherthan on surfaces that had undergone hydrogen pas-sivation by immersion in an HF solution. Althoughthe as-cleaved (110) surfaces were most likely to havebeen electronically pinned, the pn junctions could bedetected using potentiometric techniques. Sampleswere fabricated with separate electrical contacts tothe p-type and n-type regions of the pn junctions, andtunneling currents were measured while forward orreverse bias voltages were applied across the pnjunction (in addition to an overall bias voltage appliedto the sample relative to the tip). Differences intunneling current measured in p-type, depleted, andn-type regions for various bias voltages applied to thepn junctions then revealed the electronic structureof the biased junctions. For experiments performed


under ultrahigh-vacuum conditions, the locations ofthe junctions could be resolved to within ∼30 nmusing these techniques.Current imaging tunneling spectroscopy performed

by Yu et al.122 under ultrahigh-vacuum conditions oncleaved, hydrogen-passivated cross sections of Simetal-oxide-semiconductor (MOS) structures pro-vided a demonstration of the ability, using cross-sectional STM, to characterize realistic device struc-tures with spatial resolution in both the vertical andlateral directions on the order of ∼10-20 nm. In thiswork, MOS structures were fabricated with gate andsource/drain junction structure and topology charac-teristic of 0.1 µm Si MOSFET technology. Thesample and tip geometry employed in these experi-ments is shown in Figure 29a. The geometry of theimplanted junctions used in these studies was suchthat cleaving through regions containing multiplejunctions and subsequently locating and imaging thejunction regions was relatively straightforward.Current imaging tunneling spectroscopy measure-

ments were performed on the exposed cross sectionsof the MOS junctions as indicated in Figure 29b.Spatially resolved tunneling spectroscopy confirmedthat clear differences were observable among cur-rent-voltage spectra obtained from p-type, n-type,and depleted regions of the junctions. Because thecurrent-voltage spectra differed significantly for thep-type, n-type and depleted regions, current imagesgenerated from the spatially resolved tunneling cur-rent spectra were able to reveal the profiles of thepn junctions with extremely high spatial resolutionin both the lateral and vertical directions.Figure 30a shows a 160 nm × 160 nm constant-

current topographic image of a junction region,obtained at a sample bias voltage of +1.5 V and a

tunneling current of 1 nA. The location and orienta-tion of the image region with respect to the junctionstructure are as indicated in Figure 29b; it is evident

Figure 29. (a) Schematic diagram of the sample-tipgeometry used for characterization of Si MOSFET struc-tures by cross-sectional STM. (b) Location and orientationof topographic and current images.

Figure 30. (a) A 160 nm × 160 nm constant-currenttopographic image of the junction region of a Si MOSFET,obtained at a sample bias of +1.5 V and a tunneling currentof 1 nA. The gray-scale range is 0.5 nm. (b) Current imagefrom the same area obtained for a sample bias of 2.0 V.The dark area, representing low tunneling current, corre-sponds to the n+ source/drain region; the brighter arearepresents larger tunneling currents present in the p-typeregion. (c) Current image for a sample bias of -2.0 V. Thedark band, representing large negative tunneling current,clearly delineates the depletion region of the junction; thebrighter areas represent smaller negative tunneling cur-rents present in the n+- and p-type regions. (Reprinted fromref 122. Copyright 1996 American Institute of Physics.)


from the figure that there is no clear correspondencebetween the topographic image and the electronicstructure of the junction. Parts b and c of Figure 30show current images from the same area obtainedfor sample bias voltages of +2 and -2 V, respectively,with the tip-sample separation at each point fixedby requiring a tunneling current of 1 nA to flow for+1.5 V sample bias voltage. In Figure 30b, the darkregion corresponds to the small tunneling currentspresent in the n+ source/drain region, while thebrighter areas correspond to the larger positivetunneling currents measured in the p-type channelregion. Intermediate values of the tunneling currentare measured in the depletion layer, which thereforedoes not appear as a distinct region in the currentimage.In contrast, one sees in Figure 30c that, for nega-

tive sample bias voltage, a small negative tunnelingcurrent is present in the n+ source/drain region, anda somewhat larger negative current in the p-typeregion; most significantly, however, a dark band inthe image, representing large negative tunnelingcurrents present in the depletion layer, clearly de-lineates the shape of the junction. Figure 31 showsprofiles of the tunneling currents measured at posi-tive and negative sample bias along the line indicatedby the arrows in Figure 30. For negative bias appliedto the sample, a sharp increase in the magnitude ofthe tunneling current is seen at the edge of the n+

region, providing a precise indication of the locationof the pn junction to within ∼10-20 nm. Thedepletion layer width determined from tunnelingspectroscopy measurements was found to agree wellwith that calculated using dopant concentrationsobtained from SIMS measurements on the samesamples, and the junction shapes seen in the currentimages were consistent with junction profiles ob-tained from transmission electron microscopy imagesof chemically delineated junctions.122 The ability toperform two-dimensional carrier profiling of sourceand drain junctions in Si MOSFET’s with spatialresolution on the order of 10 nm, as demonstratedin these experiments, will be a critical capability forsuccessful realization and optimization of MOSFETcircuits as gate lengths approach 0.1 µm and below.

VI. Summary and ConclusionsAs the typical dimensions of advanced semiconduc-

tor devices approach the nanometer to atomic scale,the need to characterize, understand, and ultimatelyto control structure and electronic properties at ornear the atomic scale is increasing. Cross sectionalscanning tunneling microscopy has proven to be apowerful technique for probing the atomic to nanom-eter-scale properties of advanced semiconductor ma-terials and device structures, providing valuableinformation about advanced compound semiconduc-tor heterostructure materials and devices as well asultrasubmicron Si microelectronic devices of rela-tively near-term commercial interest.The utility of cross-sectional STM for performing

characterization of semiconductor structures withextremely high spatial resolution is strongly depend-ent upon the ability to obtain cross-sectional surfacesthat are uncontaminated, atomically flat, and elec-tronically unpinned. For III-V semiconductors, suchsurfaces are achievable by cleaving under ultrahigh-vacuum conditions to expose (110) cross-sectionalplanes. Studies of III-V semiconductors by cross-sectional STM have, consequently, been extremelyfruitful. Investigations of semiconductor heterojunc-tions in a variety of material systems have providednew information concerning atomic-scale structureof heterojunction interfacessparticularly for so-calledmixed-anion material systemssand its dependenceon epitaxial growth conditions. Detailed informationconcerning atomic-scale fluctuations in composition,clustering of atoms, and other such phenomena internary and quaternary III-V alloys has been re-vealed by cross-sectional STM; cross-sectional STMis particularly valuable in such studies by virtue ofits sensitivity predominantly to local, atomic-scalecomposition and electronic properties at or extremelyclose to the cross-sectional surface. The spatialresolution afforded by STM has also made possiblethe characterization of individual dopant atoms andpoint defects in III-V semiconductors.Studies of Group IV materials and device struc-

tures are of profound interest by virtue of thepreeminence of Si in microelectronics technology.Cross sectional STM studies of Group IV semicon-ductor structures have been quite limited, however,due to the difficulty of obtaining a high-quality cross-sectional surface. Shortcomings in cross-sectional

Figure 31. Tunneling currents measured along a linetraversing the MOSFET junction as indicated by thearrows in Figure 30. Currents measured at sample biasvoltages of (a) 2.0 V and (b) -2.0 V are shown. (c)Conduction- and valence-band-edge profiles calculatedusing dopant concentrations measured by SIMS. The regionof increased negative tunneling current at negative samplebias corresponds to the depletion region of the pn junction.


surface quality also limit the lateral spatial resolutionthat can typically be achieved to the nanometerrather than atomic scale. Nevertheless, cross-sectional imaging of Si/Si1-xGex heterostructures andcarrier-type profiling of Si MOSFET structures withnanometer-scale resolution have been successfullyachieved.The application of scanning probe techniques gen-

erally, and cross-sectional STM in particular, to thecharacterization of advanced semiconductor materi-als and device structures is still in its early stages.A wide range of possibilities exists for further inves-tigation of atomic-scale interface, alloy layer, anddefect structure in a variety of III-V compoundsemiconductor material systems. Extensive studiesof Group IV heterostructure materials and devicestructures would be enabled by improvements intechniques for preparing cross-sectional surfaces ofGroup IV semiconductors. In addition, STM andother scanning probe techniques have shown greatpromise and generated considerable interest fornanometer-scale characterization of ultrasubmicronSi electronic devices; continued advances in this areapromise to have a significant impact on progress inadvanced Si microelectronic technology.Future work on cross-sectional STM is likely to

benefit substantially not only from improvements inexperimental techniques, e.g., for performing spec-troscopy or for sample preparation, but also fromincreased use of detailed modeling and simulation toaid in the interpretation of images and spectroscopicdata. As discussed in the previous sections, inter-pretation of images and especially quantitative in-terpretation of spectroscopic data can be complicatedconsiderably by the fact that the atomic compositionof the cross-sectional surface plane is often highlyinhomogeneous and not completely known a priori.Theoretical modeling and simulation of tunnelingcurrents are likely to help in providing a sound basisfor interpretation of images and spectroscopic dataobtained by cross-sectional STM.In addition, it will be important to continue to

investigate and establish correlations between resultsof atomic- to nanometer-scale characterization bycross-sectional STM and information about materialand device properties obtained using other measure-ments that probe different material and deviceproperties, that are sensitive to properties at largerlength scales, or are otherwise complementary tocross-sectional STM. This is in fact of great impor-tance for any characterization technique that pro-vides information with extremely high spatial reso-lution but over a very small area or volume of thesample. For example, comparisons between atomic-to nanometer-scale characterization using cross-sectional STM and results of electrical transportmeasurements or optical measurements performedon actual device structures will provide informationabout the detailed relationship between atomic-scaleproperties and device behavior, and help confirm thatinformation obtained from measurements over anextremely small area or volume are indeed represen-tative of the whole.Fabrication and characterization of nanometer-

scale materials and devices have become, and prom-

ise to remain, pervasive themes of research in a widerange of fields. As scanning probe techniques gener-ally and cross-sectional STM in particular are appliedin an increasingly routine manner to a broad rangeof problems, they are likely to play a significant rolein much of the progress that will be made in nanom-eter-scale science and technology. Work in this areapromises to provide fertile ground for exploration andinnovation in the coming years.

VII. AcknowledgmentsPart of this work was supported by the National

Science Foundation under Award nos. ECS 93-07986and ECS 95-01469, and by an Alfred P. SloanResearch Fellowship.

VIII. References(1) Binnig, G.; Rohrer, H. Helv. Phys. Acta 1982, 55, 726.(2) Binnig, G.; Rohrer, H.; Gerber, Ch.; Weibel, E. Appl. Phys. Lett.

1982, 40, 178.(3) Binnig, G.; Rohrer, H.; Gerber, Ch.; Weibel, E. Phys. Rev. Lett.

1982, 49, 57.(4) Binnig, G.; Rohrer, H. Surf. Sci. 1983, 126, 236.(5) Binnig, G.; Quate, C. F.; Gerber, Ch. Phys. Rev. Lett. 1986, 56,

930.(6) Feenstra, R. M.; Stroscio, J. A.; Tersoff, J.; Fein, A. P. Phys. Rev.

Lett. 1987, 58, 1192.(7) Feenstra, R. M.; Stroscio, J. A. J. Vac. Sci. Technol. B 1987, 5,

923.(8) Lutz, M. A.; Feenstra, R. M.; Chu, J. O. Surf. Sci. 1995, 328,

215.(9) Van Loenen, E. J.; Dijkkamp, D.; Hoeven, A. J. J. Microscopy

1988, 152, 487.(10) Dijkkamp, D.; Van Loenen, E. J.; Hoeven, A. J.; Dieleman, J. J.

Vac. Sci. Technol. A 1990, 8, 218.(11) Salemink, H.; Albrektsen, O. J. Vac. Sci. Technol. B 1991, 9,

779.(12) Gomez-Rodriguez, J. M.; Baro, A. M.; Silveira, J. P.; Vazquez,

M.; Gonzalez, Y.; Briones, F. Appl. Phys. Lett. 1990, 56, 36.(13) Lew, A. Y.; Yu, E. T. Unpublished results.(14) Hamers, R. J. Annu. Rev. Phys. Chem. 1989, 40, 531.(15) Feenstra, R. M. Surf. Sci. 1994, 299/300, 965.(16) Stroscio, J. A.; Feenstra, R. M. In Scanning Tunneling Micros-

copy; Stroscio, J. A., Kaiser, W. J., Eds.; Academic Press: SanDiego, 1993; p 95.

(17) Feenstra, R. M.; Stroscio, J. A.; Fein, A. P. Surf. Sci. 1987, 181,295.

(18) Hamers, R. J.; Avouris, Ph.; Bozso, F. Phys. Rev. Lett. 1987, 59,2071.

(19) Feenstra, R. M.; Thompson, W. A.; Fein, A. P. Phys. Rev. Lett.1986, 56, 608.

(20) Hamers, R. J.; Tromp, R. M.; Demuth, J. E. Phys. Rev. Lett. 1986,56, 1972.

(21) Stroscio, J. A.; Feenstra, R. M.; Fein, A. P. Phys. Rev. Lett. 1986,57, 2579.

(22) Tersoff, J.; Hamann, D. R. Phys. Rev. Lett. 1983, 50, 1998.(23) Tersoff, J.; Hamann, D. R. Phys. Rev. B 1985, 31, 805.(24) Bardeen, J. Phys. Rev. Lett. 1961, 6, 57.(25) Selloni, A.; Carnevali, P.; Tosatti, E.; Chen, C. D. Phys. Rev. B

1985, 31, 2602.(26) Lang, N. D. Phys. Rev. B 1986, 34, 5947.(27) Bono, J.; Good, R. H., Jr. Surf. Sci. 1985, 151, 543.(28) Bono, J.; Good, R. H., Jr. Surf. Sci. 1986, 175, 415.(29) Martensson, P.; Feenstra, R. M. Phys. Rev. B 1989, 39, 7744.(30) Feenstra, R. M. Phys. Rev. B 1995, 50, 4561.(31) Stroscio, J. A.; Feenstra, R. M.; Fein, A. P. J. Vac. Sci. Technol.

A 1987, 5, 838.(32) Stroscio, J. A.; Feenstra, R. M.; Newns, D. M.; Fein, A. P. J.

Vac. Sci. Technol. A 1988, 6, 499.(33) Van Laar, J.; Huijser, A.; Van Rooy, T. L. J. Vac. Sci. Technol.

1977, 14, 894.(34) Bechstedt, F.; Enderlein, R. Semiconductor Surfaces and Inter-

faces; Academie-Verlag: Berlin, 1988; Chapter 3.(35) Gudat, W.; Eastman, D. E. J. Vac. Sci. Technol. 1976, 13, 831.(36) Norman, D.; McGovern, I. T.; Norris, C. Phys. Lett. A 1977, 63,

384.(37) Straub, D.; Skibowski, M.; Himpsel, F. J. J. Vac. Sci. Technol.

A 1985, 3, 1484.(38) Muralt, P. Appl. Phys. Lett. 1986, 49, 1441.(39) Muralt, P.; Meier, H.; Pohl, D. W.; Salemink, H. W. M. Appl.

Phys. Lett. 1987, 50, 1352.(40) Muralt, P.; Pohl, D. W. Appl. Phys. Lett. 1986, 48, 514.


(41) Salemink, H. W. M.; Meier, H. P.; Ellialtioglu, R.; Gerritsen, J.W.; Muralt, P. R. M. Appl. Phys. Lett. 1989, 54, 1112.

(42) Albrektsen, O.; Arent, D. J.; Meier, H. P.; Salemink, H. W. M.Appl. Phys. Lett. 1990, 57, 31.

(43) Salemink, H. W. M.; Albrektsen, O.; Koenraad, P. Phys. Rev. B1992, 45, 6946.

(44) Salemink, H. W. M.; Albrektsen, O. J. Vac. Sci. Technol. B 1992,10, 1799.

(45) Albrektsen, O.; Meier, H. P.; Arent, D. J.; Salemink, H. W. M.Appl. Phys. Lett. 1993, 62, 2105.

(46) Zhang, Y.-H.; Chow, D. H. Appl. Phys. Lett. 1994, 65, 3239.(47) Zhang, Y.-H. Appl. Phys. Lett. 1995, 66, 118.(48) Reinhardt, F.; Dwir, B.; Kapon, E. Appl. Phys. Lett. 1996, 68,

3168.(49) Kato, T.; Osaka, F.; Tanaka, I. Jpn. J. Appl. Phys. 1989, 28,

1050.(50) Salemink, H.; Albrektsen, O.Microelectron. Eng. 1991, 15, 101.(51) Vandenberg, J. M.; Panish, M. B.; Temkin, H.; Hamm, R. A.

Appl. Phys. Lett. 1988, 53, 1920.(52) Chang, J. C. P.; Chin, T. P.; Kavanagh, K. L.; Tu, C. W. Appl.

Phys. Lett. 1991, 58, 1530.(53) Ourmazd, A.; Tsang, W. T.; Rentschler, J. A.; Taylor, D. W. Appl.

Phys. Lett. 1987, 50, 1417.(54) Ourmazd, A.; Taylor, D. W.; Cunningham, J.; Tu, C. W. Phys.

Rev. Lett. 1989, 62, 933.(55) Ourmazd, A.; Baumann, F. H.; Bode, M.; Kim, Y. Ultramicros-

copy 1990, 34, 237.(56) Tanaka, I.; Kato, T.; Ohkouchi, S.; Osaka, F. J. Vac. Sci. Technol.

A 1990, 8, 567.(57) Gomez-Rodriguez, J. M.; Baro, A. M.; Silveira, J. P.; Vazquez,

M.; Gonzalez, Y.; Briones, F. Appl. Phys. Lett. 1990, 56, 36.(58) Albrektsen, O.; Arent, D. J.; Meier, H. P.; Salemink, H. W. Appl.

Phys. Lett. 1990, 57, 31.(59) Salemink, H. W. M.; Albrektsen, O. J. Vac. Sci. Technol. B 1992,

10, 1799.(60) Johnson, M. B.; Maier, U.; Meier, H.-P.; Salemink, H. W. M.

Appl. Phys. Lett. 1993, 63, 1273.(61) Smith, A. R.; Gwo, S.; Sadra, K.; Shih, Y. C.; Streetman, B. G.;

Shih, C. K. J. Vac. Sci. Technol. B 1994, 12, 2610.(62) Smith, A. R.; Chao, K.-J.; Shih, C. K.; Shih, Y. C.; Streetman,

B. G. Appl. Phys. Lett. 1995, 66, 478.(63) Smith, A. R.; Chao, K.-J.; Shih, C. K.; Shih, Y. C.; Anselm, K.

A.; Streetman, B. G. J. Vac. Sci. Technol. B 1995, 13, 1824.(64) Zheng, J. F.; Walker, J. D.; Salmeron, M. B.; Weber, E. R. Phys.

Rev. Lett. 1994, 72, 2414. Zheng, J. F.; Walker, J. D.; Salmeron,M. B.; Weber, E. R. Phys. Rev. Lett. 1994, 73, 368.

(65) Pfister, M.; Johnson, M. B.; Alvarado, S. F.; Salemink, H. W.M.; Marti, U.; Martin, D.; Morier-Genoud, F.; Reinhart, F. K.Appl. Phys. Lett. 1995, 67, 1459.

(66) Tuttle, G.; Kroemer, H.; English, J. H. J. Appl. Phys. 1990, 67,3032.

(67) Brar, B.; Ibbetson, J.; Kroemer, H.; English, J. H. Appl. Phys.Lett. 1994, 64, 3392.

(68) Chow, D. H.; Miles, R. H.; Hunter, A. T. J. Vac. Sci. Technol. B1992, 10, 888.

(69) Feenstra, R. M.; Collins, D. A.; Ting, D. Z.-Y.; Wang, M. W.;McGill, T. C. Phys. Rev. Lett. 1994, 72, 2749.

(70) Feenstra, R. M.; Collins, D. A.; Ting, D. Z.-Y.; Wang, M. W.;McGill, T. C. J. Vac. Sci. Technol. B 1994, 12, 2592.

(71) Feenstra, R. M.; Collins, D. A.; McGill, T. C. SuperlatticesMicrostruct. 1994, 15, 215.

(72) Lew, A. Y.; Yu, E. T.; Chow, D. H.; Miles, R. H. Appl. Phys. Lett.1994, 65, 201.

(73) Lew, A. Y.; Yu, E. T.; Chow, D. H.; Miles, R. H. Mat. Res. Soc.Symp. Proc. 1994, 340, 237.

(74) Lew, A. Y.; Zuo, S. L.; Yu, E. T.; Miles, R. H. Appl. Phys. Lett.1997, 70, 75.

(75) Goodnick, S. M.; Ferry, D. K.; Wilmsen, C. W.; Liliental, Z.;Fathy, D.; Krivanek, O. Phys. Rev. B 1985, 32, 8171.

(76) Pashley, M. D.; Haberern, K. W.; Gaines, J. M. Appl. Phys. Lett.1991, 58, 406.

(77) Sudijono, J.; Johnson, M. D.; Snyder, C. W.; Elowitz, M. B.; Orr,B. G. Phys. Rev. Lett. 1992, 69, 2811.

(78) Heller, E. J.; Lagally, M. G. Appl. Phys. Lett. 1992, 60, 2675.(79) Pond, K.; Maboudian, R.; Bressler-Hill, V.; Leonard, D.; Wang,

X.-S.; Self, K.; Weinberg, W. H.; Petroff, P. M. J. Vac. Sci.Technol. B 1993, 11, 1374.

(80) Tanimoto, M.; Osaka, J.; Takigami, T.; Hirono, S.; Kanisawa,K. Ultramicroscopy 1992, 42, 1275.

(81) Miles, R. H.; Schulman, J. N.; Chow, D. H.; McGill, T. C.Semicond. Sci. Technol. 1993, 8, S102.

(82) Wang, M. W.; Collins, D. A.; McGill, T. C.; Grant, R. W. J. Vac.Sci. Technol. B 1993, 11, 1418.

(83) Collins, D. A.; Wang, M. W.; Grant, R. W.; McGill, T. C. J. Appl.Phys. 1994, 75, 259.

(84) Miles, R. H.; Chow, D. H.; Hamilton, W. J. J. Appl. Phys. 1992,71, 211.

(85) Lew, A. Y.; Yu, E. T.; Zhang, Y.-H. J. Vac. Sci. Technol. B 1996,14, 2940.

(86) Osaka, F.; Tanaka, I.; Kato, T.; Katayama, Y. Jpn. J. Appl. Phys.1988, 27, L1193.

(87) Osaka, F.; Kato, T. J. Vac. Sci. Technol. B 1993, 11, 1697.(88) Salemink, H. W. M.; Albrektsen, O. Phys. Rev. B 1993, 47, 16044.(89) Skala, S. L.; Wu, W.; Tucker, J. R.; Lyding, J. W.; Seabaugh, A.;

Beam, E. A., III; Jovanovic, D. J. Vac. Sci. Technol. B 1995, 13,660.

(90) Wu, W.; Skala, S. L.; Tucker, J. R.; Lyding, J. W.; Seabaugh, A.;Beam, E. A., III; Jovanovic, D. J. Vac. Sci. Technol. A 1995, 13,602.

(91) Lew, A. Y.; Yan, C. H.; Tu, C. W.; Yu, E. T. Appl. Phys. Lett.1995, 67, 932.

(92) Lew, A. Y.; Yan, C. H.; Welstand, R. B.; Zhu, J. T.; Tu, C. W.;Yu, E. T.; Yu, P. K. L. J. Electron. Mater. 1997, in press.

(93) Harrison, W. A. J. Vac. Sci. Technol. 1977, 14, 1016.(94) Lew, A. Y.; Yu, E. T. Unpublished results.(95) Adachi, S. Physical Properties of III-V Semiconductor Com-

pounds; J. Wiley and Sons: New York, 1992.(96) Madelung, O., Ed. Landolt-Bornstein: Numerical Data and

Functional Relationships in Science and Technology; Springer:New York, 1982; New Series III, Vol. 17. O. Madelung, ed.,Landolt-Bornstein: Numerical Data and Functional Relation-ships in Science and Technology; Springer: New York, 1987;New Series III, Vol 22.

(97) Van de Walle, C. G. Phys. Rev. B 1989, 39, 1871.(98) Gwo, S.; Chao, K. J.; Shih, C. K.; Sadra, K.; Streetman, B. G.

Phys. Rev. Lett. 1993, 71, 1883.(99) Pinnington, T.; Patitsas, S. N.; Lavoie, C.; Sanderson, A.; Tiedje,

T. J. Vac. Sci. Technol. B 1993, 11, 908.(100) Feenstra, R. M.; Yu, E. T.; Woodall, J. M.; Kirchner, P. D.; Lin,

C. L.; Pettit, G. D. Appl. Phys. Lett. 1992, 61, 795.(101) Gwo, S.; Smith, A. R.; Shih, C. K.; Sadra, K.; Streetman, B. G.

Appl. Phys. Lett. 1992, 61, 1104.(102) Feenstra, R. M.; Vaterlaus, A.; Yu, E. T.; Kirchner, P. D.; Lin,

C. L.; Woodall, J. M.; Pettit, G. D. In Salemink, H. W. M.,Pashley, M. D., Eds.; Semiconductor Interfaces at the Sub-Nanometer Scale; Kluwer Academic Publishers: Dordrecht,1993; p. 127.

(103) Vaterlaus, A.; Feenstra, R. M.; Kirchner, P. D.; Woodall, J. M.;Pettit, G. D. J. Vac. Sci. Technol. B 1993, 11, 1502.

(104) Tseng, W. F.; Dagata, J. A.; Silver, R. M.; Fu, J.; Lowney, J. R.J. Vac. Sci. Technol. B 1994, 12, 373.

(105) Johnson, M. B.; Albrektsen, O.; Feenstra, R. M.; Salemink, H.W. M. Appl. Phys. Lett. 1993, 63, 2923.

(106) Johnson, M. B.; Albrektsen, O.; Feenstra, R. M.; Salemink, H.W. Appl. Phys. Lett. 1994, 64, 1454.

(107) Johnson, M. B.; Meier, H. P.; Salemink, H. W. M. Appl. Phys.Lett. 1993, 63, 3636.

(108) Johnson, M. B.; Koenraad, P. M.; van der Vleuten, W. C.;Salemink, H. W. M.; Wolter, J. H. Phys. Rev. Lett. 1995, 75, 1606.

(109) Zheng, J. F.; Liu, X.; Weber, E. R.; Ogletree, D. F.; Salmeron,M. J. Vac. Sci. Technol. B 1994, 12, 2104.

(110) Zheng, Z. F.; Salmeron, M. B.; Weber, E. R. Appl. Phys. Lett.1994, 64, 1836.

(111) Feenstra, R. M.; Woodall, J. M.; Pettit, G. D. Phys. Rev. Lett.1993, 71, 1176.

(112) Feenstra, R. M.; Vaterlaus, A.; Woodall, J. M.; Pettit, G. D. Appl.Phys. Lett. 1993, 63, 2528.

(113) Sakaki, H.; Tanaka, M.; Yoshino, J. Jpn. J. Appl. Phys. 1985,24, L417.

(114) Tanaka, M.; Sakaki, H. J. Cryst. Growth 1987, 81, 153.(115) Abraham, D. L.; Veider, A.; Schonenberger, Ch.; Meier, H. P.;

Arent, D. J.; Alvarado, S. F. Appl. Phys. Lett. 1990, 56, 1564.(116) Alvarado, S. F.; Renaud, Ph.; Abraham, D. L.; Schonenberger,

Ch.; Arent, D. J.; Meier, H. P. J. Vac. Sci. Technol. B 1991, 9,409.

(117) Renaud, Ph.; Alvarado, S. F. Phys. Rev. B 1991, 44, 6340.(118) Lutz, M. A.; Feenstra, R. M.; Chu, J. O. Surf. Sci. 1995, 328,

215.(119) Johnson, M. B.; Halbout, J.-M. J. Vac. Sci. Technol. B 1992, 10,

508.(120) Yu, E. T.; Johnson, M. B.; Halbout, J.-M. Appl. Phys. Lett. 1992,

61, 201.(121) Yu, E. T.; Halbout, J.-M.; Powell, A. R.; Iyer, S. S. Appl. Phys.

Lett. 1992, 61, 3166.(122) Yu, E. T.; Barmak, K.; Ronsheim, P.; Johnson, M. B.; McFarland,

P.; Halbout, J.-M. J. Appl. Phys. 1996, 79, 2115.(123) Pinnington, T.; Sanderson, A.; Tiedje, T.; Pearsall, T. P.; Kasper,

E.; Presting, H. Thin Solid Films 1992, 222, 259.(124) Yu, E. T.; Johnson, M. B.; Powell, A. R.; Halbout, J.-M.; Iyer, S.

S. J. Vac. Sci. Technol. B 1993, 11, 1149.(125) Higashi, G. S.; Becker, R. S.; Chabal, Y. J.; Becker, A. J. Appl.

Phys. Lett. 1991, 58, 1656.(126) Williams, C. C.; Slinkman, J.; Hough, W. P.; Wickramasinghe,

H. K. Appl. Phys. Lett. 1989, 55, 1662.(127) Williams, C. C.; Slinkman, J.; Hough, W. P.; Wickramasinghe,

H. K. J. Vac. Sci. Technol. A 1990, 8, 895.(128) Huang, Y.; Williams, C. C.; Slinkman, J. Appl. Phys. Lett. 1995,

66, 344.


(129) Huang, Y.; Williams, C. C.; Smith, H. J. Vac. Sci. Technol. B1996, 14, 433.

(130) Neubauer, G.; Erickson, A.; Williams, C. C.; Kopanski, J. J.;Rodgers, M.; Adderton, D. J. Vac. Sci. Technol. B 1996, 14, 426.

(131) Shafai, C.; Thomson, D. J.; Simard-Normandin, M.; Mattiussi,G.; Scanlon, P. J. Appl. Phys. Lett. 1994, 64, 342.

(132) Shafai, C.; Thomson, D. J.; Simard-Normandin, M. J. Vac. Sci.Technol. B 1994, 12, 378.

(133) DeWolf, P.; Clarysse, T.; Vandervorst, W.; Snauwaert, J.; Helle-mans, L. J. Vac. Sci. Technol. B 1996, 14, 380.

(134) Nonnenmacher, M.; O’Boyle, M.; Wickramasinghe, H. K. Ultra-microscopy 1992, 42-44, 268.

(135) Henning, A. K.; Hochwitz, T.; Slinkman, J.; Never, J.; Hoffman,S.; Kaszuba, P.; Daghlian, C. J. Appl. Phys. 1995, 77, 1888.

(136) Kikukawa, A.; Hosaka, S.; Imura, R. Appl. Phys. Lett. 1995, 66,3510.

(137) Hochwitz, T.; Henning, A. K.; Levey, C.; Daghlian, C.; Slinkman,J.; Never, J.; Kaszuba, P.; Gluck, R.; Wells, R.; Pekarik, J.; Finch,R. J. Vac. Sci. Technol. B 1996, 14, 440.

(138) Abraham, D. W.; Williams, C.; Slinkman, J.; Wickramasinghe,H. K. J. Vac. Sci. Technol. B 1991, 9, 703.

(139) Weaver, J. M. R.; Wickramasinghe, H. K. J. Vac. Sci. Technol.B 1991, 9, 1562.

(140) Bourgoin, J.-P.; Johnson, M. B.; Michel, B. Appl. Phys. Lett. 1994,65, 2045.

(141) Kordic, S.; Van Loenen, E. J.; Dijkkamp, D.; Hoeven A. J.;Moraal, H. K. J. Vac. Sci. Technol. A 1990, 8, 549.

(142) Kordic, S.; Van Loenen, E. J.; Walker, A. J. Appl. Phys. Lett.1991, 59, 3154.

CR960084N


cross-sectional scanning tunneling...

Documents