Japan Advanced Institute of Science and Technology

JAIST Repositoryhttps://dspace.jaist.ac.jp/

TitleGermanium nanostructures on silicon observed by

scanning probe microscopy

Author(s) Tomitori, Masahiko; Arai, Toyoko

Citation MRS Bulletin, 29(7): 484-487

Issue Date 2004-07

Type Journal Article

Text version publisher

URL http://hdl.handle.net/10119/4958

Rights

Copyright (C) 2004 Materials Research Society. It

is posted here by permission of the Materials

Research Society. Masahiko Tomitori and Toyoko

Arai, MRS Bulletin, 29(7), 2004, 484-487.

http://www.mrs.org/

Description

484 MRS BULLETIN/JULY 2004

IntroductionIt is widely accepted that scanning probe

microscopy (SPM)1 has greatly contributed tothe development of nanoscale materialsscience and technology. Using SPM, we canobtain information on surface topographyand simultaneously measure the physicalproperties of the surface with atomic reso-lution. In the early 1980s, the advent ofscanning tunneling microscopy (STM)2 asthe first member of the SPM family oftechniques was a landmark achievement inthe field of high-resolution microscopies: amechanically sharpened tip was able to pre-cisely depict a sample surface with atomicresolution. Moreover, the more recent tech-nique of noncontact atomic force micros-copy (nc-AFM)3,4 can also achieve atomicresolution, even on low-conductivity andinsulating materials;5 this noticeable fea-ture has not been achieved by any othermicroscopic technique to date. With SPM,

the development of a mechanical scanningmethod using an atomically sharp tip isleading us into a fascinating “nanoworld.”

In materials science and technology on the nanoscale, the fabrication and evalua-tion of heteroepitaxial nanostructures haveattracted much interest for realizing noveloptoelectronic devices, in particular, by di-rect observation and manipulation usingSPM. In general, nanostructures fabricatedon surfaces are surrounded by characteristiccrystallographic small planes (i.e., facets) insuch a way that the total energy around thenanostructures decreases under pseudo-equilibrium conditions. In principle, the optoelectronic properties concerning quan-tum states—for example, light emissionwith a blueshift—are governed by the shape,atomic arrangement, and composition ofnanostructures. Thus, three-dimensional(3D) topographic observation is indispen-

sable for characterizing the properties offaceted nanostructures and designing andcreating novel functional devices usingthem. It is not easy to observe nanostructureswith complicated facets, but SPM makes itpossible, as does low-energy electronmicroscopy (LEEM).6 Although LEEM hasadvantages in analyzing nanostructurefacets in real time without any tip-inducedartifacts,7 SPM is superior to LEEM in regard to the achievable spatial resolution,and the two techniques can be used complementarily.

Atypical example for heteroepitaxial crys-tal growth leading to the self-organizedformation of nanostructures is to be foundin the Si-Ge system. Extending the capa-bilities of Si as the most widely used semi-conducting material is of great importancefrom an industrial viewpoint. The integra-tion of other materials can provide a Si-containing material with light-emittingand fast-switching capabilities. Integrationhas been carried out by fabricating hetero-interfaces, quantum dots, and superlatticeson Si. Germanium deposited on a Si(001)surface exhibits pronounced features ingrowth mode: it was previously knownfrom electron microscopy and electron dif-fraction studies that a Ge film deposited onSi(001) grows layer-by-layer up to a criti-cal thickness of about three monolayers,and that sequentially, large Ge clusters formthree-dimensionally beyond that criticalthickness. This heteroepitaxial growth iscategorized as the so-called Stranski–Krastanov growth mode, according to asimple classification usually employed inthe field of heteroepitaxial crystal growth.8Since there is a lattice mismatch of 4.2%between Si and Ge, the interface betweenthe Si and Ge surfaces is significantlystrained. The strain energy is released byforming Ge clusters on the wetted layerswith a critical Ge thickness to prevent thestrain energy in the layers from piling up:the relaxation of the Ge atomic configura-tion in the growing cluster and on the sur-face of the cluster with various steps candecrease the total strain energy throughgradual, slight deformation. This morpho-logical change in the heteroepitaxial growthhas not only invoked practical issues inthe control of the Si-Ge interface, but alsohas initiated a debate in the materials sci-ence community concerning heteroepitax-ial cluster growth in preferential shapeswith respect to energetics. SPM providesextremely valuable images for analyzingthis type of heteroepitaxial growth withatomic resolution.9

Furthermore, a well-defined nanostruc-ture with specified facets and atomic re-construction is required for establishing astandard tip for SPM. For atomic-resolution

GermaniumNanostructures onSilicon Observed byScanning ProbeMicroscopyMasahiko Tomitori and Toyoko Arai

AbstractScanning tunneling microscopy and noncontact atomic force microscopy have been

used to observe germanium growth on Si(001) and Si(111). The atomically resolvedimages provide invaluable information on heteroepitaxial film growth from the viewpointsof both industrial application and basic science. We briefly review the history ofcharacterizing heteroepitaxial elemental semiconductor systems by means of scanningprobe microscopy (SPM), where the Stranski–Krastanov growth mode can be observedon the atomic scale: the detailed phase transition from layer-by-layer growth to three-dimensional cluster growth was elucidated by the use of SPM. In addition, wecomment on the potential of SPM for examining the spectroscopic aspects ofheteroepitaxial film growth, through the use of SPM tips with well-defined facets.

Keywords: AFM, atomic force microscopy, crystal growth, elemental semiconductors,germanium, nanostructures, scanning probe microscopy, scanning tunneling microscopy, silicon, SPM, STM.

Germanium Nanostructures on Silicon Observed by Scanning Probe Microscopy

MRS BULLETIN/JULY 2004 485

imaging by STM, an atomically sharp tipwith a stable structure is required. However,to measure the physical properties of asample (i.e., the surface electronic statesand the force interaction between the tipand the sample), the atomic species at theapex of the tip and the structure of theshank should be well defined; we alwaysmeasure the quantity convolutedly relatedto the electronic properties of both the tipand the sample. A well-defined tip havingan apex surrounded with specified facetscan promote theoretical studies of the in-teraction between the tip and the sample.In this review, we mainly describe Gecluster growth on Si substrates in variousmodes observed by SPM, and discuss theadvantages of SPM tips with well-definedfacets from the viewpoint of nanoscalematerials science.

STM Observation of theStranski–Krastanov Growth Mode: Ge Clusters on Si

In the 1980s, Si(111) and Si(001) surfacesplayed an important role in confirming anddemonstrating the atomic resolution of theSTM instruments developed in laborato-ries around the world.10 Afterward, manyresearchers started to observe molecularadsorption and thin-film growth on thesesurfaces, as well as Si growth on Si. Subse-quently, in the early 1990s many researchersstarted to observe the heteroepitaxial thin-film growth of semiconductors by STM, inparticular, Ge on a Si(001) substrate.11–18 Gecrystals have the same diamond structureas Si, in which each atom has four bondswith one electron per bond. Thus, in prin-ciple, initial growth of less than one mono-layer (ML) of Ge on Si exhibits a similartopography to Si overlayers on Si.

Figure 1a shows a simplified model of aSi(001) 2 � 1 reconstructed surface consistingof symmetrical Si dimers with monoatomicheight steps: this surface has a periodicityof two times the primitive cell on an idealSi(001) diamond surface. One dimer consistsof two Si atoms bound to each other; byforming a dimer, the number of danglingbonds at each Si atom on the surface is reduced by one-half, while increasing thestrain energy through tilted bonds, result-ing in a stable arrangement with the lowestsurface energy. To be precise, the dimerachieves this lower energy by asymmetri-cally buckling, with one Si atom up and theother Si atom down. The two atoms in anasymmetrical dimer can quickly flip-flopbetween buckling up and down at roomtemperature, and consequently, the dimerlooks symmetrical on a time-averaged basisusing STM, which has a slow response time19

(Figure 1b). The Si dimers form a row,leading to a 2 � 1 reconstruction with re-

spect to the primitive surface cell of the(001) plane. On the upper terrace, the dimerrows run perpendicular to those on thelower terrace. This arrangement conforms ina reasonable manner to the dimer formationon a stepped surface of a diamond structure:the dimer orientation is rotated 90� on theupper terrace with respect to the lower terrace, with a monoatomic step height.

Figure 2 shows an STM image of a Ge-deposited Si(001) surface at 300�C with aGe coverage of �0.1 ML. The Ge dimers aredepicted as bright protrusions, and someof them form a line: a buckled-up Ge atomin a dimer clearly protrudes in the image,and some isolated Ge dimers look sym-metrical with a depression at the middle ofthe dimer, corresponding to a node ofelectronic states.13,16 The Ge dimers in theline exhibit asymmetry as a zigzag pattern,and the asymmetry is also induced in thesurface regions of Si without Ge coverage.This means that the Ge deposition freezesthe flip-flop motion of Si dimers on thesubstrate by means of the induced strain. Byincreasing Ge deposition up to 1 ML, the

surface becomes covered with Ge dimers,while the Ge dimers form rows that meeteach other at many places over the substrate.Beyond 1 ML Ge coverage, missing-dimerrows are introduced to decrease the surfacestrain, as has been explained by Köhleret al.;15 generally speaking, a Ge atom, witha radius larger than that of Si, cannot enterthe narrow missing-dimer rows. Figures 3aand 3b show STM images of 2 ML and 3.5 ML Ge coverage, respectively, grownat 500�C. The missing-dimer rows appearas dark lines in Figure 3a and are regularlyintroduced in Ge layers with a periodicityof 8–10 dimers in a belt-like structure.17 InFigure 3b, the patch-like structures foundat about 3.5 ML Ge coverage are combina-tions of perpendicularly crossing missing-dimer rows. The Ge atoms deposited onthe belt-like structure cannot be adsorbedin the missing-dimer row due to an in-crease in strain energy. This leads to theformation of a Ge dimer colony in a “patch”area, surrounded by missing-dimer rows:missing-dimer rows in the layer underneathlook like trenches, and the dark rows run-ning perpendicular to them are newlyformed missing-dimer rows generated bythe Ge overgrowth.

Further Ge deposition initiates 3D Gecluster growth. Figure 4 shows an STMimage of 3 ML Ge coverage at 300�C, ex-hibiting precursor nuclei for Ge clustergrowth over a few patches. Proceeding withGe deposition, we observe the formation ofmany “hut” clusters, first found by Moet al.,11 as shown in Figure 5a. Each hutcluster has four {015} facets with ridge linesat the meeting of two facets and a narrow(001) top consisting of dimer rows. The{015} facet has perfectly regular atomic rows.Tomitori et al. found that Ge growth onSi(015) continues layer-by-layer beyond

Figure 2. Scanning tunneling microscopyimage of Ge dimers on a Si(001)surface at 300 �C with a Ge coverage of�0.1 ML.The Ge dimers are depictedas bright protrusions; the scanning areais approximately 13 nm � 10 nm.

Figure 1. (a) Simplified atomistic modelof a Si(001) 2 � 1 surface consisting ofsymmetrical Si dimers. Red circlescorrespond to atoms on the lowerterrace, and blue circles correspond toatoms on the upper terrace, at a heightdifference of a monoatomic step (�0.14 nm). (b) Typical scanningtunneling microscopy image of aSi(001) 2 � 1 surface at a sample biasvoltage of –1 V with respect to the tippotential. The scanning area isapproximately 20 nm �15 nm.

486 MRS BULLETIN/JULY 2004

Germanium Nanostructures on Silicon Observed by Scanning Probe Microscopy

10 ML with almost no defects.20 This sur-face is regarded as a small (001) terrace re-peating with a single-atom-height stepand a 45�-rotated two-dimer width. Thus,the strain in the heteroepitaxial growthcan be released through the step, whichhas a more flexible atomic configuration. Adetailed atomic configuration model hasrecently been proposed by Fujikawa et al.21

and confirmed using density functionaltheory calculations by Hashimoto et al.22

The rebonding configuration is introducedat the step of the small (001) terrace to sta-bilize the electronic states: back bonds ofGe atoms at the step edge are bound todangling bonds of Ge dimer atoms on thelower terrace. This configuration was alsodiscussed by Tomitori to explain defect-free Ge growth on Si(015).20

As shown in Figure 5b, larger Ge clusterswith fourfold symmetry were found usingSTM; they were termed “dome” clustersby Tomitori.17 The dome is surrounded by a(001) facet on the top, and by {113}, {015},and other high-index facets in fours, dueto the symmetry.

The surface phase transitions for Gecoverage of less than 10 ML at growth

temperatures of 300�C, 400�C, and 500�C aresummarized in Reference 17. In that study,several large clusters were observed byannealing the sample. At the base of theclusters, close to the substrate, {015} facetsalways appear, probably to effectively re-duce the strain energy at the interface be-tween Si and Ge by introducing {015} facetswith many (001) steps. Medeiros-Ribeiroet al.23 statistically analyzed the abruptvolume change from “pyramid” clusters(hut clusters with a square base) to domeclusters, with few intermediate volume sizes,as a thermodynamic phase transition withan energy barrier, as evaluated from STMimages of Ge on Si(001). On the otherhand, Tromp et al. clearly observed thetransition from pyramid to dome clustersin real time at high temperatures by LEEMand concluded that the shape change is dueto an anomalous coarsening process simi-lar to Ostwald ripening, with a slow trans-formation via several transition states.24

Although Voigtländer succeeded in ob-serving hut cluster growth at high tem-peratures by fast STM,9 it is not easy torecord the images at a video rate, particu-larly for rough surfaces [e.g., dome clusterson Si(001)]. Thus, complementary observa-tions by LEEM and STM are a promisingapproach for studying heteroepitaxialgrowth proceeding by means of complicatedmodes such as the Stranski–Krastanovmode; LEEM is best for fast observationsover wide areas, and STM for imagingwith atomic spatial resolution.

Over the past few years, noncontact-AFM(nc-AFM) has become one of the mostpowerful atomic-resolution microscopies.Recently, the nc-AFM technique has beenapplied to studies of Ge on Si(111). Atomic-resolution images were obtained,25 asshown in Figure 6. The reconstructions of7 � 7, 5 � 5, and 9 � 9 unit cells, explainedby the DAS model,26 and their domainboundaries, as well as islands and step-flowgrowth, are observed as clearly as by STM(for example, see Reference 9). Using theinteraction between the sample and a Sitip on an nc-AFM cantilever, we can observethe topography as well as simultaneouslymeasure tunneling current and dampingenergy. The damping energy is a nano-mechanical quantity that can be measuredas the power input to a modulation piezonecessary to maintain a constant oscillationamplitude of the AFM cantilever. These si-multaneously measured quantities provideus with surface spectroscopic information.

By changing the bias voltage between thetip and the sample, one can sometimesspontaneously modify the tip properties.When this occurs, the atomic contrast innc-AFM topography and damping energydrastically changes: three Si corner adatoms

Figure 3. Scanning tunneling microscopyimages of (a) �2 ML Ge coverage witha belt-like structure (scanning area,approximately 100 nm � 90 nm), and (b) 3.5 ML Ge coverage with a patch-like structure (scanning area,approximately 70 nm � 70 nm), bothgrown at 500 �C and observed at roomtemperature.

Figure 4. Scanning tunneling microscopyimage of 3 ML Ge coverage at 300 �C,exhibiting precursor nuclei for three-dimensional Ge cluster growth.Scanning area, approximately 45 nm � 45 nm.

Figure 5. Scanning tunneling microscopyimages of 6.5 ML Ge coverage grownat 500 �C. (a) Ge “hut” clusters. Scanningarea, approximately 70 nm � 70 nm.(b) Round “dome” clusters surroundedby hut clusters. Scanning area,approximately 200 nm � 150 nm.

Germanium Nanostructures on Silicon Observed by Scanning Probe Microscopy

MRS BULLETIN/JULY 2004 487

on the faulted half of a 7 � 7 unit cell aredistinctly depicted; moreover, in the damp-ing image, one observes the three adatomsat specified sites in Ge overlayers.25 Thismight indicate that nc-AFM can be used todistinguish Si and Ge at surfaces. It shouldalso be noted that tip modification some-times occurs accidentally when the tip isscanned over a sample.

To clarify the capabilities of nc-AFM, weshould discuss the effect of bias voltage onthe image contrast and the interactionchange induced by the characteristics of thetip apex. We have suggested that a changein bias voltage shifts the Fermi levels in thetip and the sample and subsequently en-hances a quantum mechanical resonance in-teraction between the energetically tunedsurface states of the tip and the sample.27

Furthermore, the background force changedby applying a bias voltage alters the imag-ing conditions, resulting in a contrast changeand sometimes a contrast inversion.28 Theimaging mechanism of damping energy isstill under debate in the field of nc-AFM.4,29

Recently, Morita et al. distinguished Si andGe atoms in a Si and Ge intermixing layerby nc-AFM: they reported that the differ-ence in electronic states between Si and Geatoms alters the interaction with a Si tip30

under their conditions, compensating thedifference in contact potential between thetip and the sample. On the other hand,Kawamura et al. reported a contrast changedue to Si and Ge on the surface of step-flowsuperstructures covered with Bi observedby STM, which may be ascribed to the dif-ferent electronic state of the Bi overlayer.31

Intermixing Ge into a Si layer plays an im-portant role in releasing the mismatch strain;recognition of Si and Ge atoms in the mix-ing layer by SPM is becoming increasinglyimportant for heteroepitaxial systems.

The tip shape and composition are muchmore critical in forming atomic contrast innc-AFM images than they are with STM: thechemical bonding states are governed by the electronic states of both the tip and thesample on the atomic scale. To elucidate theimaging mechanism for atomic contrast, the tip apex should be controlled in a well-defined manner. We usually use a com-mercially available [001]-oriented Si tip on apiezoresistive Si cantilever for nc-AFM. Thetip can be heated in ultrahigh vacuum bypassing a small current through the piezo-resistive cantilever.32,33 Nanoclusters withwell-defined facets, such as the Ge hut anddome clusters on Si(001), are suitable can-didates for an nc-AFM tip: Ge or a similarmaterial deposited on a [001]-oriented Si tipand subsequently heated can lead to the for-mation of a well-defined tip surrounded byatomically flat facets having sharp corners atthe meeting points of the facets. This proc-ess looks promising as an addition to tra-ditional surface science methods forfabricating well-defined tips.

SummaryScanning probe microscopy studies of a

heteroepitaxial elemental semiconductorsystem [Ge on Si(001) and Si(111)] have beenbriefly described in this article, as well as thepotential of SPM for studying hetero-epitaxial film growth. The atomically re-solved images obtained by SPM inspire usto understand nanostructures fabricated byheteroepitaxial growth; the complicatedgrowth stages, referred to collectively as theStranski–Krastanov growth mode, havebeen elucidated on an atomic scale by SPM.By finding initially grown facets on a wet-ting layer in the Stranski–Krastanov modeusing SPM, we can choose a substrate having the same plane indexes as the facet tocontrol overlayer morphology layer-by-layer—for example, the {015} facet of Ge onSi(001). In addition, an SPM tip with a repro-ducible and well-defined shape and compo-sition can be fabricated by utilizing strainedheteroepitaxial overlayers. These advanceswill open the way for the detailed analysisand fabrication of nanomaterials with newfunctional optoelectronic properties.

References1. For example, see R. Wiesendanger, ScanningProbe Microscopy and Spectroscopy: Methods andApplications (Cambridge University Press,Cambridge, UK, 1994).2. G. Binnig, H. Rohrer, Ch. Gerber, and E.Weibel, Phys. Rev. Lett. 49 (1982) p. 57.

3. F.J. Giessibl, Science 267 (1995) p. 68.4. S. Morita, R. Wiesendanger, and E. Meyer,eds., Noncontact Atomic Force Microscopy (Springer-Verlag, Berlin, 2002).5. M. Reichling and C. Barth, Phys. Rev. Lett. 83(1999) p. 768.6. E. Bauer, Rep. Prog. Phys. 57 (1994) p. 895.7. For example, see M. Tomitori, F. Iwawaki, N.Hirano, F. Katsuki, and O. Nishikawa, J. Vac.Sci. Technol., A 8 (1990) p. 222.8. For example, see C. Ratsch and A. Zangill,Surf. Sci. 293 (1993) p. 123.9. B. Voigtländer, Surf. Sci. Rep. 43 (2001) p. 127.10. R. Becker and R. Wolkow, in Scanning Tunnel-ing Microscopy, Chapter 5, edited by J.A. Stroscioand W.J. Kaiser (Academic Press, San Diego,1993) p. 149. 11. Y.-W. Mo, D.E. Savage, B.S. Swartzentruber,and M.G. Lagally, Phys. Rev. Lett. 65 (1990) p. 1020.12. F. Iwawaki, M. Tomitori, and O. Nishikawa,Surf. Sci. Lett. 253 (1991) p. L411.13. F. Iwawaki, M. Tomitori, and O. Nishikawa,Surf. Sci. 266 (1992) p. 285.14. F. Iwawaki, M. Tomitori, and O. Nishikawa,Ultramicroscopy 42-44 (1992) p. 902.15. U. Köhler, O. Jusko, B. Müller, M. Horn-vonHoegen, and M. Pook, Ultramicroscopy 42-44(1992) p. 832.16. M. Tomitori, K. Watanabe, M. Kobayashi, F.Iwawaki, and O. Nishikawa, J. Vac. Sci. Technol.,B 12 (1994) p. 2022.17. M. Tomitori, K. Watanabe, M. Kobayashi,and O. Nishikawa, Appl. Surf. Sci. 76/77 (1994)p. 322.18. J. Knall and J.B. Pethica, Surf. Sci. 265 (1992)p. 156.19. R.A. Wolkow, Phys. Rev. Lett. 68 (1992)p. 2636.20. M. Tomitori, K. Watanabe, M. Kobayashi,and O. Nishikawa, Surf. Sci. 301 (1994) p. 214.21. Y. Fujikawa, K. Akiyama, T. Nagao, T. Sakurai,M.G. Lagally, T. Hashimoto, Y. Morikawa, andK. Terakura, Phy. Rev. Lett. 88 176101-1 (2002).22. T. Hashimoto, Y. Morikawa, Y. Fujikawa, T.Sakurai, M.G. Lagally, and K. Terakura, Surf.Sci. 513 (2002) p. L445.23. G. Medeiros-Ribeiro, A.M. Bratkovski, T.I.Kamins, D.A.A. Ohlberg, and R.S. Williams,Science 279 (1998) p. 353.24. F.M. Ross, R.M. Tromp, and M.C. Reuter,Science 286 (1999) p. 1931.25. T. Arai and M. Tomitori, Appl. Surf. Sci. 188(2002) p. 292.26. K. Takayanagi, Y. Tanishiro, M. Takahashi,and S. Takahashi, J. Vac. Sci. Technol., A3 (1985)p. 1502.27. T. Arai and M. Tomitori, Jpn. J. Appl. Phys.,Part 1 39 (2000) p. 3753.28. T. Arai and M. Tomitori, Appl. Surf. Sci. 157(2000) p. 207.29. T. Arai and M. Tomitori, Appl. Phys. A 72(2001) p. S51.30. S. Morita and Y. Sugawara, in Nanotechnol-ogy and Nano-Interface Controlled Electronic Devices,Chapter 21, edited by M. Iwamoto, K. Kaneko andS. Mashiko (Elsevier, Amsterdam, 2002) p. 431.31. M. Kawamura, N. Paul, V. Cherepanov, andB. Voigtländer, Phys. Rev. Lett. 91 096102 (2003).32. T. Arai and M. Tomitori, Jpn. J. Appl. Phys.,Part 1 36 (1997) p. 3855.33. T. Arai and M. Tomitori, J. Vac. Sci. Technol.,B 18 (2000) p. 648. ■■

Figure 6. Noncontact atomic forcemicroscopy image of Ge on Si(111) withGe coverage of 0.7 ML grown at 450 �C.Ge islands and step flow with 5 � 5, 7 � 7, and 9 � 9 unit cell reconstructionsare observed. Darker areas correspondto the Si(111) 7 � 7 substrate withoutGe coverage.The scanning area isapproximately 75 nm � 75 nm.