Surface Science 602 (2008) 2055–2060

Contents lists available at ScienceDirect

Surface Science

journal homepage: www.elsevier .com/locate /susc

On the existence of a stable, room temperature dihydride-terminatedGe(100) surface in ultrahigh vacuum

Grant Underwood a, Lynette Keller Ballast b, Alan Campion c,*

a Intel Corporation – SiTD MA Labs, 2501 NW 229 Street, Portland, OR 97124, USAb Cerium Labs, 5204 East Ben White Blvd., Austin, TX 78741, USAc Department of Chemistry and Biochemistry, The University of Texas at Austin, Austin, TX 78712, USA

a r t i c l e i n f o a b s t r a c t

Article history:Received 27 July 2007Accepted for publication 24 March 2008Available online 29 March 2008

Keywords:GermaniumLow index single crystal surfacesHydrogenAdsorptionLEEDThermal desorptionRaman scattering spectroscopy

0039-6028/$ - see front matter � 2008 Elsevier B.V. Adoi:10.1016/j.susc.2008.03.030

* Corresponding author. Tel.: +1 512 471 3012; faxE-mail address: campion@mail.utexas.edu (A. Cam

We present strong evidence, using low energy electron diffraction, temperature programmed desorptionand Raman spectroscopy, for the formation of a dihydride-terminated Ge(100) surface phase that is sta-ble in ultrahigh vacuum at room temperature. This phase was prepared by exposing the Ge(100) surfaceto a large fluence of atomic hydrogen generated by dissociation of molecular hydrogen on a tungsten fil-ament. We observed a strong uptake dependence on filament placement and temperature and proposethat the reason a number of previous reports did not observe this phase was an insufficient fluence ofatomic hydrogen. Isotopic substitution of deuterium for hydrogen produced identical results and furtherallowed us to demonstrate that facile hydrogen abstraction may play an important role in the observeddifferences for hydrogen adsorption on the (100) surfaces of silicon and germanium.

� 2008 Elsevier B.V. All rights reserved.

1. Introduction

Over the past decade or so there has been a resurgence of inter-est in germanium as a microelectronics material due in large partto the demonstration of atomic layer growth [1], and the produc-tion of device quality SiGe and Si1�xGex heterostructures [2]. TheseSiGe systems are of interest because of their unique electronic andoptoelectronic [3] characteristics and because they take advantageof existing silicon base technology. Although it is common to growthese structures on silicon substrates there is an increasing interestin film growth on germanium for the investigation of novel devicesthat take advantage of the smaller band gap, increased dopant sol-ubility and higher hole mobility of germanium [4]. Although thesemicroelectronics applications have provided a new practical moti-vation for understanding the chemistry of germanium surfaces, thesubject is also of fundamental interest.

The surface chemistry of the low index faces of silicon has beenextensively explored and the reactions of hydrogen with these sur-faces are well understood, especially on the most technologicallyrelevant Si(100) surface. For the purposes of the present contribu-tion it is sufficient to note that the clean Si(100) surface undergoesa 2 � 1 reconstruction in vacuum; exposure to atomic hydrogenproduces a monohydride phase that preserves the reconstruction.

ll rights reserved.

: +1 512 471 8696.pion).

Additional hydrogen adsorption produces mixed mono- and dihy-dride phases that eventually saturate to form the 1 � 1 dihydride-terminated surface. More extensive discussions of the adsorptionand reactions of hydrogen with silicon surfaces are presented inseveral reviews [5,6].

The situation for Ge(100) is much less clear, however, despite anumber of investigations spanning the past two decades. In 1986,Papagno and coworkers observed a (1 � 1) low energy electron dif-fraction (LEED) pattern and obtained a vibrational spectrum (usinghigh-resolution electron energy loss spectroscopy (HREELS)) thatidentified the germanium dihydride scissor vibration [7]. Cohenand coworkers observed a low temperature shoulder in their tem-perature programmed desorption (TPD) spectra that they attrib-uted to the formation of a germanium dihydride phase [8]. Onthe other hand, Chabal found no IR spectral evidence of dihydrideformation upon exposure to atomic hydrogen and concluded that itwas not possible to prepare this phase under ultrahigh vacuum(UHV) conditions [9]. A number of experiments using various com-binations of TPD, LEED and STM measurements also concluded thatno significant extended dihydride phase results from large expo-sures to atomic hydrogen under UHV conditions at 300 K [10–14]. There is still considerable current interest in this system asevidenced, for example, by the recent work of Chabal and cowork-ers, who reported a wet-chemical protocol for preparing a hydro-gen-passivated Ge(100) surface that was a mixture of bothmonohydride and dihydride phases [15]. In an effort to reconcile

2056 G. Underwood et al. / Surface Science 602 (2008) 2055–2060

the results of the previous UHV investigations, we have used Ra-man spectroscopy, LEED and TPD to study this system. We reportcompelling evidence for the existence of a stable dihydride-termi-nated Ge(100) surface at 300 K in ultrahigh vacuum and suggestan explanation for the discrepancies observed in the previousstudies.

Fig. 1. TPD spectra and uptake curve resulting from relatively small doses of atomicdeuterium.

2. Experimental

The experiments were performed using an ultrahigh vacuum(UHV) chamber, pumped by a Leybold Heraeus 360 l/s turbomolec-ular pump and Varian Multiple Vacion pump system, that routinelyachieved base pressures of 8 � 10�11 torr. A Stanford Research Sys-tems RGA200 quadrupole mass spectrometer was used for TPDexperiments, to monitor the purity of species delivered during dos-ing and to provide residual gas analysis.

The Ge(100) single crystal (undoped, Eagle–Picher, 1.1 �1.6 cm.) was mounted in tantalum clips and resistively heated bythe ‘sandwich’ method previously described by Gillis et al. [16].The crystal mount, which provides for resistive heating, and cool-ing to liquid nitrogen temperatures, consists of two electrically iso-lated OFHC blocks attached to an x,y,z,h manipulator arm.Although we initially tried to clean the sample using a CO2 snownozzle, we found that the resultant temperature gradient causedthe formation of visible surface microfractures. We opted, there-fore, to mount the sample as received, with no additional cleaningprior to loading into the UHV chamber. Although our as-loadedsamples showed a considerable degree of carbon and oxygen con-tamination, as detected by Auger spectroscopy (PHI cylindricalmirror analyzer), it was easily removed by just a few sputter(600 eV argon, 15–20 lA/cm2) and anneal (�650 �C, 10 min) cycles.In our laboratory, we use Raman spectroscopy to detect surfacecarbon at levels below the Auger sensitivity limit (which is 0.5%),having shown a number of years ago that the large Raman cross-section for graphitic carbon makes it readily detectable at verylow concentrations.

Atomic hydrogen (Scientific Gas Products, Research Grade) anddeuterium (Matheson, UHP) were dosed by backfilling the cham-ber to 10�7–10�6 torr of molecular hydrogen (deuterium) andmoving the sample to within a few centimeters of a hot, coiledtungsten filament. The flux of atomic hydrogen was a strong func-tion of both the filament temperature and the distance betweenthe sample and the filament; we discuss these issues later. Alldoses were performed at a sample temperature of 300–310 K andare expressed as partial pressures (uncorrected for gauge sensitiv-ity) of H2 or D2.

For the TPD experiments, the sample temperature and heatingrate were measured with both an infrared pyrometer and a K-typethermocouple attached to one of the tantalum clips near the sam-ple. A computer-controlled temperature ramp produced a con-stant, stable heating rate of approximately 1.5 K/s as verified byindependent pyrometer and thermocouple measurements.

Raman spectra were obtained using 514.5 nm radiation from aCoherent Innova-200 argon ion laser. Unless otherwise noted allspectra were obtained using 1000 s integration times with300 mW of power incident on the surface (higher powers causeda slow rise in sample temperature). The angle of incidence wasca. 70� with respect to the surface normal and the laser radiationwas polarized in the plane of incidence (p-polarized). A Navitar f/0.95 camera lens, positioned at 60� from the surface normal in aplane orthogonal to the plane of incidence, was used to collectthe scattered radiation. This configuration is the most efficientfor surface Raman experiments on highly reflective surfaces [17].The collected light was passed through a holographic notch filter(Kaiser Optical Systems) and dispersed by a Spex HR-320 spec-

trometer. The detector was a liquid nitrogen cooled PhotometricsCH210 charge coupled device and the system resolution was about1 cm�1 per channel.

3. Results

Fig. 1 shows the TPD spectra for D2 following exposure to vari-ous amounts of atomic deuterium. We used deuterium for theseexperiments to eliminate the interference from background hydro-gen in the chamber; the corresponding experiments using hydro-gen produced identical results except for a sloping background inthe TPD spectra due to the residual hydrogen. For these initialexperiments the filament-to-sample distance was set at 4.5–5.5 cm, with a filament temperature of 1700 �C as determined bya pyrometer. The inset in the upper right of Fig. 1 is a standard up-take curve showing TPD peak area as a function of deuterium expo-sure. The coverage increases rapidly with increasing exposurereaching a plateau at approximately 1000–1500 L. We assume thatthis plateau represents monolayer coverage of some surface spe-cies and set it equal to 1 ML; all other surface concentrations arereferenced to this value. At high coverage, a small shoulder beginsto appear on the low temperature side of the TPD peak; we origi-nally, and mistakenly, ascribed this feature to contamination oretching.

Fig. 2 shows a typical Raman spectrum of the bare Ge(100) sur-face. The intense peak located at 561 cm�1 is the second order Gephonon whereas the smaller peak at 840 cm�1 is the third orderphonon. The features at 653 and 748 cm�1 have not previouslybeen reported but appear to be combinations or overtones of otherbulk modes. In the upper trace the spectrum has been expandedapproximately five fold to emphasize the less intense peaks. Thetwo broad peaks near 1330 and 1560 cm�1 are due to diamond-like

Fig. 2. Raman spectrum of the clean Ge(100) surface.

Fig. 3. Raman spectra of monohydride and monodeuteride terminated Ge(1 00)surfaces.

G. Underwood et al. / Surface Science 602 (2008) 2055–2060 2057

and graphitic carbon contamination, respectively. As mentionedabove, Raman spectroscopy is particularly sensitive to graphiticsurface carbon and we estimate the residual coverage to be lessthan 0.1% of a monolayer. The carbon is undetectable by Augerspectroscopy. The tiny feature at 1976 cm�1 is also seen in Figs. 3and 6; it is due to the Ge–H stretching mode that results from asmall but persistent hydrogen background pressure in ourchamber.

Our TPD results suggest that exposures of ca. 1000 L hydrogenproduce a mono-hydride saturated surface. Raman spectra werecollected after exposures of 1000 L hydrogen and deuterium,respectively (Fig. 3). For these spectra, and the Raman spectra thatfollow, the bare germanium spectrum of Fig. 2 has been subtractedto produce a flat background. The features at 1444 and 1976 cm�1

correspond well with previously published HREELS and IR data forthe Ge–D and Ge–H monohydride stretching frequencies, respec-tively [7,9,18]. Based upon our experience with Raman signalintensities from molecules adsorbed on single crystal surfaces atmonolayer coverages, it appears that the Raman scattering fromthese adsorbed hydrides is slightly enhanced, relative to whatwould be expected from the gas phase Raman cross-section forGeH4, for example. Similar enhancements have been reported forthe Si–H vibrations on a Si(111) monohydride-terminated surface[19].

To determine the source of the low temperature shoulder visi-ble in the TPD spectra at higher exposures shown in Fig. 1, we thensystematically exposed the sample to much higher doses of atomichydrogen. To produce these large doses without drasticallyincreasing the exposure time, we decreased the sample-to-fila-ment distance to �2.5 cm and increased the filament temperatureto �1900 �C. We found that these conditions enable us to providethe required doses with an acceptably small temperature rise (less

Fig. 4. TPD spectra and uptake curve resulting from large doses of atomicdeuterium.

2058 G. Underwood et al. / Surface Science 602 (2008) 2055–2060

than 10 �C) for the longest exposures. The resultant TPD spectraand uptake curve are presented in Fig. 4. This uptake curve agreeswell with the one shown in Fig. 1 up to exposures of about 1000 L.The increased flux, however, reduced the time required to achievea given coverage by a factor of 2–4. For higher doses the low tem-perature shoulder becomes a well-resolved separate peak. Thisnew peak saturates at a total coverage of roughly two monolayersas can clearly be seen by the appearance of a second plateau in theuptake curve for exposures above approximately 5000 L. Our TPDspectra agree reasonably well with those of Shimokawa et al.[13]. The inset in this figure also shows the regimes in which verysharp and well-defined LEED patterns were observed. For expo-sures below about 1000 L, the LEED pattern was identical to thatof the bare (2 � 1) reconstructed surface. Higher exposures wereaccompanied by a concomitant reduction in the intensity of thehalf-order spots resulting in a clear (1 � 1) pattern by about4000 L. The total integrated area of the dihydride plus monohy-dride TPD peaks saturates at twice that of the monohydride peakarea and the LEED pattern changes from a 2 � 1 to a 1 � 1 patternat exposures where the dihydride TPD peak is observed. Theseobservations, taken together, strongly suggest the formation of astable, dihydride-terminated Ge(100) surface, analogous to thatformed on Si(100).

To corroborate the results of our TPD and LEED experiments, weattempted to obtain Raman spectroscopic evidence for a germa-nium dideuteride species. The unfortunate coincidence of theGeD2 scissor mode frequency with that of the intense second ordergermanium phonon made it impossible to detect using Ramanspectroscopy, however. Switching to hydrogen helps but sincethe GeH2 scissor mode is located on the shoulder of the third orderGe phonon, extreme care must be taken during background sub-traction to avoid the creation of artifacts. We employed the follow-ing procedure for reproducible background subtraction.

Fig. 5. Raman spectra of Ge(100) after various exposures of atomic hydrogen. TheRaman spectrum of the clean surface has been subtracted.

After dosing atomic hydrogen, the crystal was translated to theRaman position and a spectrum was collected. The crystal was thenflashed to 500 �C, while remaining in the Raman position, allowedto cool to room temperature and a spectrum of the bare surfacewas acquired. So a background spectrum was taken at each hydro-gen exposure and subtracted from the spectrum of the hydroge-nated surface. Raman spectra of hydrogenated Ge(100) surfacesat several hydrogen exposures are shown in Fig. 5. For exposuresgreater than 1000 L, where the shoulder first appears on our TPDdata, we observe a clear peak at 830 cm�1 in our Raman spectra,which grows in and saturates with increasing exposure. Not shownin Fig. 5 is a spectrum resulting from a 10,000 L dose that shows nosignificant difference from the 5000 L spectrum. The feature we ob-serve at 830 cm�1 corresponds closely to the 105 meV loss previ-ously reported in an HREELS study of Papagno et al. [7], whichthey assigned to the GeH2 scissors mode. Their assignment wasbased upon analogy to the HREELS results for the dihydride-termi-nated Si(100) surface. The frequency of this mode is in the rangeexpected (based upon the gas phase vibrational frequencies of ger-mane, for example) and the failure to observe this mode by IR spec-troscopy was one piece of evidence that led Chabal to concludethat he had not prepared a dihydride under the conditions of hisexperiment [8]. To ensure that these features were not due to con-tamination from dosing we used the same background subtractionprocedure described above to generate a spectrum from a sampleexposed to 5000 L of deuterium. The resulting spectrum is shownin Fig. 6 along with one resulting from exposure to 5000 L of hydro-gen, for comparison. Although both the scissor and stretchingmodes from background hydrogen adsorption can be discernedin the upper spectrum, they are much weaker than in the lowerspectrum. Only the Ge–D stretching mode is visible for the deuter-ated surface as the expected isotopic shift for the 830 cm�1 scissor

Fig. 6. Raman spectra of Ge(100) after exposures of 5000 L of atomic hydrogen andatomic deuterium. The Raman spectrum of the clean surface has been subtracted.

Fig. 7. Raman spectra of Ge(100) exposed to 1000 L of atomic hydrogen (a) follo-wed by exposure to 4000 L of atomic deuterium (b).

G. Underwood et al. / Surface Science 602 (2008) 2055–2060 2059

mode would put it at ca. 600 cm�1, a region where strong substratescattering renders it unobservable.

4. Discussion

As stated in the introduction, the existence of an extended, sta-ble GeH2 phase at room temperature in UHV remains controver-sial; Refs. [7,8] provide support for the existence of this phase,whereas Refs. [9–14] concluded that a dihydride-terminated ger-manium surface could not be prepared in UHV at room tempera-ture. The shape of the uptake curve in our study provides onepossible explanation for the discrepancy between the two previoussets of studies. Since very high exposures to atomic hydrogen arenecessary to form the dihydride, and since the monohydride up-take curve has a well-defined plateau, some of the previous exper-iments may not have exposed the surface to enough hydrogenatoms to create an extended dihydride phase. Comparison of expo-sures from different experiments is problematic because exposuresare generally reported in terms of molecular hydrogen although itis the flux of atomic hydrogen that matters. Small differences in thedosing arrangement, therefore, have significant consequences. It isnot uncommon to find the source of atomic hydrogen described asa ‘‘hot” tungsten filament or foil placed a ‘‘few” cm away from thesample. As we have shown, even small changes in filament temper-ature and position have significant effects on the initial uptakecurve, presumably by varying the atomic flux. These details mustbe important for in Ref. [14] the authors conclude that the reactionto form the dihydride ‘‘is very limited, even after a H(g) exposure of�67 times that for the 1 ML HH”.

Although the observation of a dihydride-terminated Ge(100)surface adds to the list of similarities between germanium and sil-icon, interesting questions can still be found in the differences.Why, for example, does the formation of the (100) germaniumdihydride surface require such high exposures, compared to theSi(100) surface? One contributing factor may be the competitiveabstraction of surface hydrogen by atomic hydrogen. Adsorptionof atomic hydrogen by a surface can be represented by the follow-ing equation:

HðgÞ þ ! HðaÞ ð1Þ

where H(g) represents gas-phase atomic hydrogen, ‘_’ represents abare, unreacted surface site, and H(a) is a hydrogen atom adsorbedon the surface. A process often ignored during exposures to atomichydrogen is that of hydrogen abstraction, which can be representedby the following equation:

HðgÞ þHðaÞ ! þH2ðgÞ ð2Þ

Koleske and coworkers [20] have shown that the abstraction ofhydrogen from the Si(100) surface is a competitive reaction that al-ways accompanies adsorption of atomic hydrogen with an observedreaction probability equal to 0.36 that of adsorption. Since the Ge–Hbond is approximately 20% weaker than the Si–H bond [21],abstraction is likely to be more efficient from a hydrogen (deute-rium) terminated Ge(100) surface than from the correspondingSi(100) surface. Hydrogen atom abstraction reactions have indeedbeen invoked to explain the results reported in Refs. [13,14]. Morerecent experiments using atomic beam sources have measuredthe relative efficiencies of deuterium abstraction by hydrogenatoms to be 0.40 for Si(100) and 0.45 for Ge(100) [22]. To the ex-tent that these results are relevant to our conditions, in whichatoms are produced by dissociation over a hot filament, the smalldifferences observed cannot completely account for our uptake re-sults. Nevertheless, a greater abstraction rate on the germaniumsurface would certainly reduce the net adsorption rate and at leastpartially help explain why it takes higher exposures to saturate thedihydride surface on Ge(100) than on Si(100).

Although we have not quantified the importance of hydrogenatom abstraction in this system we have definitively detected it.Fig. 7 shows the Raman spectra of a germanium sample exposedto (a) 1000 L of hydrogen followed by (b) a subsequent exposureto 4000 L of deuterium. Here we see that deuterium has almostcompletely displaced hydrogen on the monohydride-terminatedsurface, mostly likely through an abstraction reaction that formsHD(g), leaving behind an empty surface site for the adsorption ofa deuterium atom. Subsequent TPD experiments also showed thatthis process produced a surface terminated almost entirely withdeuterium.

5. Conclusions

We have presented strong evidence for the formation of a sta-ble, room temperature, dihydride-terminated Ge(100) surface.Ours is the first study of which we are aware which combinesTPD, LEED and vibrational spectroscopy to study the Ge(100) sur-face exposed to fluences of atomic hydrogen large enough to pre-pare a stable extended dihydride-terminated phase. We havespeculated that the previously reported discrepancies about thissurface may arise from subtle experimental differences, which af-fect the net atomic flux incident on the surface. We have also pro-posed that facile hydrogen abstraction may play an important rolein the observed differences in hydrogen adsorption on the (100)surfaces of silicon and germanium.

Acknowledgements

The work reported here was funded in part by a Grant from theWelch Foundation (F-751) whose support we gratefullyacknowledge.

2060 G. Underwood et al. / Surface Science 602 (2008) 2055–2060

References

[1] See, for example: S. Dey, S. Joshi, D. Garcia-Gutierrez, M. Chaumont, M. Jose-Yacaman, A. Campion, S.K. Banerjee, J. Electron. Mater. 35 (2006) 1607. andreferences cited therein.

[2] See, for example: J.L. Hoyt, in: Proceedings – Electrochemical Society 2004-07,2004, p. 15, and references cited therein.

[3] T.P. Pearsall, CRC Crit. Rev. Solid State Mater. Sci. 15 (1989) 551.[4] S.M. Sze, Physics of Semiconductor Devices, Wiley, New York, 1981.[5] D.J. Doren, Adv. Chem. Phys. 95 (1996) 1.[6] Y.J. Chabal, Mater. Res. Soc. Symp. Proc. 259 (1992) 349.[7] L. Papagno, X.Y. Shen, J. Anderson, G.S. Spagnolo, G.J. Lapeyre, Phys. Rev. B 34

(1986) 7188.[8] S.M. Cohen, T.I. Hukka, Y.L. Yang, M.P. D’Evelyn, Thin Solid Films 225 (1993)

155.[9] Y.J. Chabal, Surf. Sci. 168 (1986) 594.

[10] L. Surnev, M. Tikhov, Surf. Sci. 138 (1984) 40.[11] J.A. Appelbaum, G.A. Baraff, D.R. Hamann, H.D. Hagstrum, T. Sakurai, Surf. Sci.

70 (1978) 654.[12] L.B. Lewis, J. Segall, K.C. Janda, J. Chem. Phys. 102 (1995) 7222.[13] S. Shimokawa, A. Namiki, M.N. Gamo, T. Ando, J. Chem. Phys. 113 (2000) 6916.[14] J.Y. Maeng, J.Y. Lee, Y.E. Cho, S. Kim, Appl. Phys. Lett. 81 (2002).[15] S. Rivillon, Y.J. Chabal, F. Amy, A. Kahn, Appl. Phys. Lett. 87 (2005) 253101.[16] J.P. Chamberlain, J.L. Clemons, A.J. Pounds, H.P. Gillis, Surf. Sci. 301 (1994) 105.[17] D.R. Mullins, A. Campion, J. Phys. Chem. 88 (1984) 8.[18] Y.J. Chabal, K. Raghavachari, Phys. Rev. Lett. 53 (1984) 282.[19] H. Sano, S. Ushioda, Phys. Rev. B: Condens. Mat. 53 (4) (1996) 1958.[20] D.D. Koleske, S.M. Gates, J.A. Shultz, J. Chem. Phys. 99 (1993) 5619.[21] C. Mui, S. Bent, C.B. Musgrave, J. Phys. Chem. B 108 (2004) 6336.[22] S. Shimokawa, A. Namiki, M.N. Gamo, T. Ando, Diam. Relat. Mater. 10 (2001)

1659.