SURFACE AND INTERFACE ANALYSIS, VOL. 26, 105È108 (1998)

Electron-stimulated Desorption Total Cross-sectionDetermination for Digermane on Si(100)

A. F. Aguilera,1,2 J. H. Campbell,3 J. H. Craig, Jr.2,3,* and K. H. Pannell1,21 Department of Chemistry, 500 W. University Ave., University of Texas at El Paso, El Paso, TX 79968-0513, USA2 Materials Research Institute, 500 W. University Ave., University of Texas at El Paso, El Paso, TX 79968-6664, USA3 Department of Physics, 500 W. University Ave., University of Texas at El Paso, El Paso, TX 79968-0515, USA

We have studied digermane-covered Si(100) using electron-stimulated desorption (ESD). Estimates are presentedfor the total H(a) ESD removal cross-section for digermane-exposed Si(100) substrates at 85 K using electronsincident at 150 eV energy. It is found that electron-enhanced deposition of Ge occurs only when physisorbeddigermane is present. Auger electron spectroscopy provided the means for determining the relative amounts ofgermanium adsorbed on the Si(100) surface following digermane exposures, electron irradiation and surface recon-struction. It is found that two coverage regimes are important : initial dosing of digermane on Si(100) at 85 Kresults in overlayers consisting of both physisorbed digermane and chemisorbed (x = 1, 2 or 3) species ;GeH

x(a)

and short anneals to 200 K following exposure of the Si(100) surface at 85 K lead to the presence of only chemi-sorbed The two coverage regimes exhibit di†erent ESD behavior. Two kinetic energy distribution (KED)GeH

x(a).

peaks are seen when physisorbed digermane is present, and only one when it is absent. The ESD signal decay curvesobtained from the two surfaces are also di†erent : the presence of physisorbed digermane results in a two-component exponential signal decay ; the absence of the physisorbed species results in a single-exponential decay.The total H removal cross-section from the physisorbed digermane overlayer was determined to ber ¿ 1.4 Â 10—15 cm2, while that from Si(100) with only adsorbed present was found to be r ¿ 2.6 Â 10—16GeH

xcm2. Our results suggest that adsorbed species remain intact on the surface even when the Si(100) sub-GeHx(a)

strate is annealed to 200 K, indicating that hydrogen migration from surface to Si surface sites does notGeHx(a)

occur at 200 K. 1998 John Wiley & Sons, Ltd.(

Surf. Interface Anal. 26, 105È108 (1998)

KEYWORDS: ESD; Si(100) ; electron-stimulated desorption ; silicon

INTRODUCTION

The e†ects of an electron beam on digermane overlayerson Si(100) surfaces are interesting in part becauseenhanced Ge deposition has been observed.1 Theobservation of enhanced Ge deposition and e†orts tocharacterize this phenomenon may lead to a betterunderstanding of adsorption and decompositionGe2H6mechanisms on Si(100). Although we have examined

on Si(100) previously, and have demonstratedGe2H6that Ge deposition is enhanced under electron irradia-tion, further studies are needed to better characterizethe interaction of the electron beam with adsorbed

In this work, we report the H(a) removal cross-Ge2H6 .section from Si(100) surfaces using anGe2H6-coveredelectron beam incident upon the surface at 70¡ o† thesurface normal and 150 eV kinetic energy.

Samples of p-silicon(100) with a resistivity of D0.1 )cm were prepared and placed into an ultrahigh vacuumsystem operating at D2 ] 10~10 Torr. Auger electron

* Correspondence to : J. H. Craig, Materials Research Institute, 500W. University Ave., University of Texas at El Paso, El Paso, TX79968-6664, USA. E-mail : jcraig=utep.edu.

Contract grant sponsor : National Science Foundation ; grant no. :CHE8920120.

spectroscopy (AES) analysis conÐrmed contaminant-free initial surface conditions. First, electron-stimulateddesorption (ESD) kinetic energy distributions (KEDs)for positive hydrogen ions from digermane-dosedSi(100) surfaces held at 85 K were obtained by methodsdescribed in detail elsewhere.2 The H` ESD KEDsexhibited bimodal peak shapes that were Ðtted by theleast-squares method to a model based upon the workof Nishijima and Propst,3 as shown in Fig. 1. Thebimodal peak in Fig. 1(a) is suggestive of the presence oftwo distinct binding states from which the desorbingpositive hydrogen ions originate,2 and is consistent withthe fact that no annealing step was used (i.e. physi-sorbed digermane was present on surface). Previoustemperature-programmed desorption (TPD) resultsindicated the presence of physisorbed digermane evenfor low exposures on Si(100) at 85 K. Work doneelsewhere4 indicates that for adsorption at 85 KGe2H6there is at least one chemisorbed state (x \ 1, 2GeH

xor 3) on Si(100) that produces at 590 K during TPDH2experiments. Thus, we conclude that our bimodal H`ESD KEDs arise from hydrogen ions desorbing viaESD from molecularly adsorbed (physisorbed) Ge2H6and directly from adsorbed germyl fragments on(GeH

x)

the Si(100) surface. In a separate set of experiments,ESD KEDs from adsorbed fragments on Si(100)GeH

xwere obtained after Ñashing the Si(100) sample to 200 Kto remove molecularly adsorbed (physisorbed)

CCC 0142È2421/98/020105È04 $17.50 Received 20 May 1997( 1998 John Wiley & Sons, Ltd. Accepted 29 August 1997

106 A. F. AGUILERA, J. H. CAMPBELL, J. H. CRAIG, JR. AND K. H. PANNELL

Figure 1. (a) An ESD KED profile showing two distinct states. (b) An ESD KED obtained after a 200 K anneal to remove physisorbedwhich exhibits only one state. The insets in (a) and (b) show fits of our decay data using Eqn (1). The inset to (a) shows aGe

2H

6,

double-exponential fit to data, obtained from a Si(100) substrate on which and are adsorbed. The inset to (b) requiredGe2H

6(a) GeH

x(a)

only a single-exponential fit.

digermane. The H` ESD KEDs from Si(100) havingonly adsorbed exhibit only one ESD KEDGeH

x(a)

peak, as seen in Fig. 1(b). The TPD results conÐrmedthat no physisorbed digermane molecules were presentwhen performing this second set of ESD measurements.

In this work we directly measure the H` ESD signaldecays by setting the energy window (pass energy) ofour Bessel Box energy analyzer5 to a constant energy(the Bessel box energy resolution is D1 eV). A singleexponential decay, given by Eqn (1)

I(t) \ Ib] aC1 [ exp([bt)

btD

(1)

can be used to Ðt the experimentally obtained H` ESDsignal decays as described elsewhere.6 In Eqn (1), I(t) isthe ion current at time t, is the background signal, aIbis a Ðtted amplitude parameter and b is the decay con-

stant (b \ pJ/e). We note that Eqn (1) was derived in amanner that included the spatial dependence of the elec-tron beam current density proÐle within the beamspot.6 A Faraday cup was used to obtain electron beamcurrent density proÐles. The experimentally determinedelectron beam current density proÐle was found to bewell approximated by

J(r) \ J0 exp([r2/a2) (2)

where J(r) is the spatially dependent current density, J0is the current density at the center of the electron-beamspot, r is the independent variable and a is the Gaussianwidth of the spot. From the Ðts of the exponential decayin Eqn (1) to our data (see insets to Fig. 1), it is quitestraightforward to extract the decay constant. Providedthat the relation holds true, we may then ploteb \ J0 peb vs. to obtain a line of slope p, which in this case isJ0the total H(a) removal cross-section for the process.

SURFACE AND INTERFACE ANALYSIS, VOL. 26, 105È108 (1998) ( 1998 John Wiley & Sons, Ltd.

ESD OF DIGERMANE-EXPOSED Si(100) 107

We obtained experimental H` ESD signal decaycurves from Si(100) surfaces both with and without amolecularly adsorbed overlayer, as shown in theGe2H6insets to Fig. 1. From single-exponential decay Ðts tothe experimental decays at various plots of eb vs.J0 , J0indicate that the H(a) removal cross-section for surfaceswith physisorbed digermane present are a factor of twogreater than for surfaces with no physisorbed digermanepresent on the surface. However, it is known thatmolecularly adsorbed coexists with onGe2H6 GeH

x(a)

the Si(100) surface at 85 K even for relatively lowexposures. Thus, in the ESD experimentsGe2H6(g)

where no e†ort is made to remove physisorbed Ge2H6 ,our Si(100) surface is covered with a mixture of stronglybound (chemisorbed) species and more weaklyGeH

xbound (physisorbed) molecules. Intuition sug-Ge2H6gests that the ESD signal would have contributionsfrom both and and bimodal ESDGeH

x(a) Ge2H6(a),

KEDs such as that shown in Fig. 1(a) conÐrm thissuspicion. With two surface states contributing to theESD signal, it is appropriate to Ðt the experimentalESD signal decay curves obtained in the presence ofphysisorbed digermane with double-exponential decaycurves. Double-exponential decay Ðts were used for theexperimental decays obtained from surfaces having phy-sisorbed digermane, and an example of such a Ðt isshown in the inset to Fig. 1(a), where the dashed linesrepresent the component single-exponential decaycurves that are summed to obtain the resultant best-Ðtdouble-exponential decay. Plots of eb vs. are madeJ0using the decay constants obtained from each com-b

iponent decay curve, and are shown as the squares andtriangles in Fig. 2. Cross-sections deduced from thebest-Ðt straight lines to the points in Fig. 2 indicate that

cm2 andp1\ 1.4 ^ 0.45] 10~15 p2\ 2.3 ^ 1.5cm2 are the values of the component H(a)] 10~16

removal cross-sections. We note that the value for isp1an unusually large cross-section, of the order of the geo-metrical cross-section. However, we have previouslyobserved that physisorbed digermane is very susceptibleto electron beam-induced decomposition.1

Figure 2. Values of eb obtained from exponential decay fits tothe experimental signal decays are plotted vs. the value used toJ

0obtain the experimental decay. Single-exponential decays wereused to fit ESD signal decays obtained following a short anneal to200 K, and the resulting data are shown as circles. Squares andtriangles represent data obtained from double-exponential fits tothe ESD signal decays obtained with physisorbed digermanepresent on the surface.

We obtained several H` ESD signal decays followingÑashing of the Si(100) to 200 K to remove the weaklybound species, and found that a single-Ge2H6exponential decay Ðt the experimental data quite nicely.An example of such a Ðt to the data is shown in theinset to Fig. 1(b). Using the relation and theeb \ J0 psame analysis as was used above, we obtained the datarepresented by circles in Fig. 2, for which the best-Ðtleast-squares straight line gives a slope ofp \ 2.8^ 0.4] 10~16 cm2. The observation madeabove that ESD KEDs for exhibit onlyGeH

x/Si(100)

one ESD desorption state supports the use of the single-exponential Ðt to the decays in Fig. 1(b).

Our results indicate that there is an ESD processoccurring with high cross-section when physisorbeddigermane is present on the surface. The high cross-section process correlates well with the presence of phy-sisorbed digermane, so it is reasonable to attribute thisH` ESD signal to removal of H(a) from physisorbed

species. The high cross-section process onlyGe2H6occurs when the high-energy (D5 eV) KED peak ispresent, so we assign the high-energy KED peak asbeing due to H` desorbed from physisorbed digermane.Just as the low-energy KED peak at D3.1 eV is alwayspresent, so is the H` ESD process having a cross-section of p D 2.6] 10~16 cm2, where the two lowcross-section values are essentially the same withinexperimental error and have thus been averaged.

The origin of the H` ESD signal decay having across-section of p D 2.6] 10~16 cm2 is less certain.Possible surface states from which this H` ESD signalmight arise include as well as various surfaceGeH

x(a)

silicon hydride species.7 The primary question iswhether hydrogen atoms associated with surface germylspecies migrate to silicon surface sites during the 200 Kanneal used to remove physisorbed digermane. We esti-mate that the 200 K anneal supplies up to 12 kcalmol~1 towards the activation of surface processes suchas di†usion of hydrogen from species to neigh-GeH

x(a)

boring silicon surface sites. If the activation barrier is aslow as 9 kcal mol~1, as recently postulated by Russelland Eckerdt,8 such di†usion would be expected tooccur and the post-annealed samples would actually beexhibiting H(a) removal cross-sections for removal ofH(a) from surface silicon hydrides. Alternatively, if theactivation barrier is higher, or the efficiency of energytransfer from the surface to the adsorbate is lower thanwe estimate, the low-energy KED peak would originatefrom electron-stimulated removal of H(a) from surface

species.GeHx(a)

Our ESD data suggest that the speciesGeHx(a)

remains intact on the surface during the 200 K anneal.Support for this conclusion arises from several observ-ations. The low-energy ESD KED peaks in Fig. 1appear at the same energy within Ðtting errors bothbefore and after the 200 K anneal, and the low-energyKED peaks reported here appear at slightly higherenergies than similar peaks reported for ESD from

Additionally, the cross-section for theSi2H6/Si(100).7low-energy KED peak found in this work(p D 2.6] 10~16 cm2) is somewhat larger than thecross-section for the low-energy (dihydride) KED peakfound in the work done with disilane adsorbed onSi(100).7 Finally, the monohydride state would beexpected to form before the dihydride, and the ESD

( 1998 John Wiley & Sons, Ltd. SURFACE AND INTERFACE ANALYSIS, VOL. 26, 105È108 (1998)

108 A. F. AGUILERA, J. H. CAMPBELL, J. H. CRAIG, JR. AND K. H. PANNELL

KED peak from the monohydride occurs near 5 eV andis not present in our post-anneal ESD KEDs. Thus,ESD from surface silicon hydrides does not appear tobe occurring.

In summary, previous work in our laboratory hasshown that molecularly adsorbed (physisorbed) Ge2H6must be present on the Si(100) surface for electron irra-diation to signiÐcantly enhance Ge atom deposition.The work reported here shows that enhanced Ge depo-sition is the result of electron-induced dissociation ofphysisorbed digermane. Following exposure of theSi(100) surface at 85 K to digermane gas, our ESDKED data strongly suggest that two distinct Ge-Hstates exist on the surface, which we ascribe to GeH

x(a)

and Finally, after annealing of Si(100) sur-Ge2H6(a).faces covered with digermane to 200 K, our ESD datasuggest that the only adsorbate remaining on thesurface is GeH

x(a).

Acknowledgements

This work was supported in part by the Science and TechnologyProgram of the National Science Foundation, grant no. CHE8920120.The authors also wish to acknowledge support received from the NSFMaterials Research Center of Excellence at the university of Texas atEl Paso, cooperative agreement No. HRD-9353547. A.F.A. acknow-ledges support from the Fulbright Institute of International Educa-tion and Consejo Nacional de Ciencia y Tecnologia (Mexico).

REFERENCES

1. J. H. Campbell, J. Lozano, A. F. Aguilera, J. H. Craig, Jr. and K.H. Pannell, Appl . Surf . Sci . 108, 345 (1997).

2. J. H. Campbell, M. V. Ascherl and J. H. Craig, Jr., J . Vac. Sci .Technol . A 12, 2128 (1994).

3. M. Nishijima and F. M. Propst, Phys.Rev.B 2, 2368 (1970).4. B. M. H. Ning and J. E. Crowell, Surf . Sci . 295, 79 (1993).

5. J. H. Craig, Jr. and W. G. Durrer, J . Vac. Sci . Technol . A 7,3337 (1989).

6. B. Xia and S. C. Fain, Jr., Phys.Rev.B 50, 14565 (1994).7. J. Lozano, J. H. Craig, Jr., J. H. Campbell and M. V. Ascherl,

Nucl . Instrum.Methods Phys.Res.B 100, 407 (1995).8. N. M. Russell and J. G. Eckerdt, Surf . Sci . 369, 51 (1996).

SURFACE AND INTERFACE ANALYSIS, VOL. 26, 105È108 (1998) ( 1998 John Wiley & Sons, Ltd.