Ultramlcroscopy 47 (1992) 212-222 North-Holland

Resolution in the scanning tunneling microscope

R J W i l s o n a, p H L l p p e l b, S C h l a n g a, D D C h a m b l l s s a a n d V M H a l l m a r k d

~ IBM Research Dwtston, Almaden Research Center, San Jose, CA 95120-6099, USA Department of Physws, Unwerslty of Texas, Arhngton, TX 60019, USA

Received at Editorial Office 11 May 1992

Scanning tunnehng microscopy (STM) has been used for many atomic-scale observations of metal and semiconductor surfaces Experience has shown that a variety of effects must be considered m interpreting h~gh-resolut~on data For example, images may be distorted by the mohon of tap or surface atoms which results from hp-surface interactions at small tunnel gaps Lower-resolution stu&es generally appear to be qmte reliable Empmcal hmlts on ~mage interpretation are &scussed for some of the metal, molecule and semiconductor systems whtch we have measured

1. Introduction

Scanning tunne l ing microscopy (STM) is a rel- atively new tool for s tudymg the s t ruc ture of surfaces [1] Expe r i ence ga ined over the pas t few years has shown that a tomic-sca le fea tu res can be imaged for a wide vartety of ma te rmls including semiconductors , meta ls and a tomic and molecu- lar over layers A c c u r a t e i n t e rp re t a t i on of such tmages requt res a good u n d e r s t a n d i n g of the ori- gins of image cont ras t and of the hmlts imposed by the r e so luhon of the S T M Exper tmen t s have shown that several p h e n o m e n a are of varying impor t ance for the success of h igh- reso lu t ion ~maglng Consequen t ly one must crtt tcally assess the da ta in an a t t emp t to es tabhsh the r ehabd i ty of i n t e rp re t a t i ons In th~s pape r , we review a range of expe r imen ta l resul ts r egard ing hlgh-reso- lut ton STM and men t ion a few theore t tca l mode l s of STM imaging which he lp to explain the obser - vat ions

2. Experiment

The S T M we use res ides in one c h a m b e r of an u l t r ah igh-vacuum system with several d i f ferent connec t ed chambers for sample p r e p a r a t i o n and

analysis [2] The STM itself is compr i sed of a p iezoe lec t r i c t r ipod, which carr tes a tungs ten tlp at small d is tance ( ~ 10 .~) f rom a single crystal surface of interest , a p iezoe lec t r i c walker for br inging the sample to the tip, and v~bratlon-lso- la t lon m e c h a m s m s The wldth of the vacuum gap be tween the sample and tip 1s ma in t a ine d by apply ing a small bias vol tage ( ~ 1 V) and maln- t a lnmg a cons tan t tunnel cu r ren t by using feed- back to cont ro l the t ip pos l t lon The tip is slowly scanned para l le l (x , y) to the surface, and the vol tage which ~s app l i ed to the z -p lezo is r eco rded and conver ted into an l inage using var ious da ta- process ing a lgor i thms The la te ra l and vert ical " r e s o l u t i o n " of the ins t rument a re in fe r red from the size of the smal les t de t ec t ab l e objects

STM differs f rom many mtcroscoples m that the r e so luhon d e p e n d s s t rongly on an uncharac- ter tzed var iable , namely the S T M tip The STM tips we use are m a d e of tungs ten wire, which is e lec t rochemica l ly s h a r p e n e d in a K O H etching solut ion A tip is checked for r ea sonab le sharp- ness m a 100x opt ica l mic roscope and then t rans- f e r red via an a l r lock into U H V The tip 1s f lashed r e pe a t e d ly to ~ 800 °C by e lec t ron b o m b a r d m e n t and l oaded into the STM W e bel ieve this proce- du re is effective in reduc ing tip lns tablht les which might be due to diffusing adso rba te s We do not

0304-3991/92/$05 00 © 1992 - Elsevier Science Pubhshers B V All rights reserved

R J Wdson et al / Resolutton m the scannmg tunnehng rmcroscope 213

imagine that tips thus formed are atomically clean tungsten or that they are always sharp on an atomic scale Some groups have invested consid- erably more effort m preparing and characteriz- ing tips using field ion microscopy, for example Our experience has shown that tip changes occur about hourly, due to, for example, accidental oscillation of the feedback loop when passing over a defect The performance of such an al- tered tip can be improved or degraded In the latter case, the tip can often be restored more quickly than a new one can be made and charac- terized Consequently, we have not pursued the ultimate luxury of using tips of known structure Because of the variability of tips, it is not un- usual, during a given STM session, to confront problems such as fluctuations of the apparent t~p height on the A scale, possibly from motion of tip adsorbates, bluntness on a 20 ,~ scale, or ghost images which result from multiple effective tips These difficulties require that deliberate tip changes be carried out by slightly increasing the bias, to ~ 5 V, or by reducing the bias to small values while monitoring changes in the surface profile from one line scan to the next Not infre- quently, this procedure falls and larger biases, 100-1000 V, are momentari ly applied In this case, surface damage is certain and the sample must be translated so that a new area can be explored Tips which cannot be made to work satisfactorily with a few hours' effort are dis- carded

During this "sharpening" procedure, we find that tunnel current noise levels are qmte variable and not necessarily correlated to the quality of images which result after servo filtering If the

o

vertical fluctuations are small enough, 0 1 A, we sometimes find apparent height fluctuations which oscillate at the vlbratlonqsolation reso- nance frequency (2 Hz) This behavior, which we attribute to the ringing following vibrational im- pulses, can also be quite tip dependent It is typically found that high vertical resolution is more easily obtained ff the lateral resolution is low, suggesting a blunt tip which somehow aver- ages out fluctuations The key point of this dls- cuss~on is that the limits on resolution are usually not those which might be ~mposed by the tunnel

current shot noise, but rather depend on the stablhty and electronic properties of the sample and tip Experience has taught us that many subtle artifacts must be eliminated by using only high-resolution images which can be repeatedly obtained over several regions of a carefully pre- pared sample using a reasonable range of bias voltages and tunnel currents

3. Theoretical models of STM imaging

STM involves the elastic tunneling of electrons from filled states of the sample or tip through the complex t ip -sample potential barrier, which in- cludes the applied tunnel gap bias potential, and into the empty states of the complementary elec- trode [1] For large planar tunnel gaps, the tunnel

current roughly follows the form l T ~ 10- ( ~ 4 ) z Here V is the tunnel barrier height in eV, which corresponds roughly to a workfunctlon, and z is in .~ For a hemispherical tip geometry, simple models have included explicit forms for an s-state tip wavefunctlon and calculations of the sample surface density of states [3], and further refine- ments of the t ip -sample potential [4] Experi- mentally, it became apparent that some atomic- resolution images could be obtained under condi- tions which revolved highly unusual values of V In particular, STM images of graphite could be obtained where a ten-fold decrease In t T caused a 100 ,~ change m the apoparent tunnel gap In- stead of the expected 1 A, implying unphyslcal values of a few meV for V The explanations proposed to account for these observations in- volved mechanlcal interactions between the tip and sample [5,6] In these models, the measured "apparen t" motion of the tip Induces a corre- sponding deformation of the sample so that the actual tunnel gap changes only slightly In an extreme hmlt, it was proposed that one could even obtain atomic-resolution images by Imbed- dlng the t~p in a portion of the sample which then shdes as a whole over the underlying bulk planes [7] Studies of semiconductor surfaces, on the other hand, had usually led to observations of relatively large corrugations where hlgh-resolu-

R J Wdson et al / Resolutton tn the scannmg tunnehng mtcroscope

tIon Images rehably showed distinctive Nas de- pendence which could be associated with the expected spectrum of electronic states within a few eV of the Fermi level [8,9] This was under- standable because of the strongly localized and directional bonding in semiconductors, but no corrugation was expected for delocahzed free electron states of metals It was therefore surpris- ing that atomic corrugation Images of close- packed metals were also obtained [10] with rea- sonable values of V of a few eV [11] It was suggested that this corrugation could result from smaller mechanical deformations which were at- tributable to soft metal-cluster tips [11] or to insulating species on the tip which were not ac- tive in tunneling but contributed to force effects [12] Although this is not an unlikely possibility, It has recently been shown that large metal corruga- tions can be obtained without force effects if the tip wavefunction is of particular p- or d-like char- acter [13] Further, It has been shown that for small t ip-sample gaps the electronic interactions between the tip and sample can result in local- lzed states which strongly perturb the surface electronic structure and alter the expected corru- gations [14] The result of this controversy is a deep concern over our ability to make detailed interpretations of STM images without support- ing information from other techmques a n d / o r electronic structure calculatmns Precise state- ments can be made if one wishes to refine exist- ing models of relatively simple, well studied sys- tems Alternatively, one can proceed into un- known systems If one is willing to make limited interpretations and draw speculative conclusions

4. Imaging close-packed metal corrugations

The first STM observations resolving individ- ual close-packed metal atoms on a surface were made on an epltaxlal A u ( l l l ) film on mica [10] These measurements were performed both in the constant-current mode in UHV and also in the constant-height mode in air Fig la is an image of where atomic corrugation is observed on two terraces adjacent to a monatomlc step The ob- servation of a corrugation of ~ 0 3 A IS still not

214

Fig 1 (a) Image showing atomic corrugation on a A u ( l l l ) surface The fine structure on the terraces matches the A u ( l l l ) lateral atomic spacing (b) A larger area image after a remote tip contact has created straight steps Normally steps are rounded during sputter-anneahng A constant has been subtracted from each terrace height to reveal the corrugation

of the Au(111) reconstruction

fully understood The image is somewhat unsta- ble in the VlClnlty of the step edge, although the atomic corrugation seems quite repeatable It ~s tempting to speculate that the atoms near the steps are somewhat mobile at room temperature Fig lb shows a much wider scan image of a A u ( l l l ) single-crystal surface which includes a number of straight steps which were created by striking the surface with the STM tip at a distant location [15] The tip used here does not resolve the atomic corrugation, but it does show a small corrugation (0 1 A) associated with a long-period reconstruction of Au(111) which is discussed later The straight steps are found to retain their char- acter for days, arguing against rapid diffusion of step atoms With these conflicting observations, ~t appears that one must use extreme caution m

R J Wilson et al / Resolution m the scanning tunnehng microscope 215

mterpre tmg the step mstabdlty because these m- stabthttes can have other origins, such as t~p-m- duced motion of step edge atoms or step-induced motion of tip atoms Similarly, the origin of atomic corrugation must be viewed with suspicion, and hmltattons on ~ts interpretation must be tested

5. Reconstruction of A u ( l l l )

To test the usefulness of atomic corrugation images, this work was extended to studies of the (23 x x/3-) stacking fault reconstruction on a sm- gle-crystal A u ( l l l ) sample [16] Previous trans- mission electron microscope (TEM) images led to a model for the surface atom rearrangement in- volving an ordered array of boundaries between surface regions with face centered cubic (fcc) type stackmg (ABC) and hexagonal close-packed (hcp) stacking (ABA) [17] Atomic corrugation STM images showed periodic lateral and vemcal undu- lations in the atomic rows which agreed well with previous models of the reconstruction and ruled out models in which the corrugation results from contact interactions which induce shppage of bulk A u ( l l l ) planes [16] It was found that, even with- out atomic resolution, the domain boundaries between regions with hcp- and fcc-type atomic stacking were clearly observed as 0 15 .~ high ridges which were assocmted with the vertical displacement of Au atoms not positioned over hollow sites The height of these ridges agrees well with the displacement expected for hard- sphere models In extending this STM work to larger fields of view, it became apparent that, in defect-free regions, the dislocation ridges are not linear but zigzag to form a herringbone superlat- tice which is comprised of alternating sections of two rotatlonally equivalent unlaxtal domains [15] This superlatt~ce is distract from, but related to, the Y fault structures observed with TEM on small Au platelets evaporatwely deposited on MoS 2 [17] We expect this difference results from differences in the sizes of crystals accessible with these two techniques Analysis of the surface crystallography showed that the boundaries be- tween rotational domains must contain a d~sloca- tlon at the bends, or elbows, where the dlsloca-

200,~

Fig 2 STM images of A u ( l l l ) surface with and without NI and Fe Islands (a) Herringbone pat tern of clean surface The spacing between unlaxial dislocations IS ~ 70 ,~ (b) Hexago- nal islands of NI decorating elbow dislocations at low cover- age (c) Wide-scan image showing NI island coherence as nucleated by herringbone domains (d) Atomic-resolution im- age of A u ( l l l ) surface surrounding a small, triangular Fe island Statistical differencing has been used to bring out

details within the island

216 R J Wtlson et a l / Resolutton m the wannmg tunnehng mtcro~cope

t lon ridges change direct ion [18] A large region of this superlat t ice s tructure is seen in fig l b and an image of a few super-cells, with 70 A × 140 ~, cell dimensions, is shown In fig 2a

intrinsic to the her r ingbone reconstruct ton of A u ( l l l ) Striking variat ions in the morphology for each of these e lements have been observed

6. Metals adsorbed on A u ( l l l )

Studies of metal film growth have been carried out for a n u m b e r of metals evaporatively de- posited on room- tempera tu re slngle-crystal sub- strates The goals of these exper iments involved under s t and ing the morphology of the deposits and de te rmin ing the ability of STM to discrimi- nate be tween different types of metals The nu- cleation p h e n o m e n a we observe are best thought of in terms of the aggregat ion of isolated atoms which ram upon the surface, thermaIlze after a short interval, and diffuse at room tempera tu re until they form or jom a stable nucleus Nucle- at ion can take place when a diffusing atom inter- acts with intrinsic defects such as steps and dislo- cations, creates an extrinsic defect as in alloy formation, or encounte rs a sufficient n u m b e r of other diffusing atoms so that homogeneous nucle- at ion creates islands in defect-free regxons of the surface

The simplest expertments involve observations of the lnlttal stages of film format ion for non-M- loymg matertals as exemplified by studies of evaporated films of Fe, NI, Au and Ag onto a room- tempera tu re A u ( l l l ) single crystal For the metals and condi t ions employed in these studies, all growth takes place at steps or at dislocations

Fig 3 (a) Hexagonal Au islands grown on a clean, annealed Au(lll) surface (b) The heights of the islands in (a) are reduced so that the distorted dislocations near islands and step edges are visible (c) A wide-scan tmage (3000 A) showing the finger growth mode of Ag on Au(lll) at steps (d) Higher-resolution image of Ag decorated edge Deformations of the reconstruction are seen on the lower terrace, while the upper terrace retains a herringbone structure The Ag deposit ~s the unreconstructed nm on the upper terrace A constant has been subtracted from points on the upper terrace to Improve image contrast (e) A 1000 A wide image of submono- layer him of Ag sputtered onto Au(111) for comparison with

(c) for evaporative deposition

(a)

(©)

R J Wtlson et al / Resolutton m the scanning tunnehng mtcroscope 217

and understood m terms of a model where the intrinsic dislocations have a variable probability m serving as nucleation sites In some cases, modification of the intrinsic dislocations by grow- mg deposits also strongly affects film growth

Nucleation at intrinsic defects on the Au(111) surface can involve steps, umaxlal dislocations of the 23 × v~ reconstruction, or elbow dislocations of the superlattlce The umaxaal dislocations are almost completely inactwe as nucleation s~tes, but the elbow d~slocatlons act as nucleation s~tes whose binding probability varies strongly for dif- ferent metals When the dislocations are strongly actwe m nucleation, as for N1 [18,19], then or- dered arrays of irregular hexagonal islands form for submonolayer deposits, as shown m fig 2b Large, coherent arrays of these islands cover the entire surface, as shown m fig 2c Similar arrays form for sub-monolayer deposits of Fe on Au(111) The nucleation at surface-lattice dislo- cations is proved by images in which the Au atomic lattice is Imaged clearly and the atomic rows are easily counted (fig 2d) The apparent topography of the Fe island, however, is comph- cated and irregular, with no atomic lattice wslble This may reflect the presence of adsorbed con- tamlnants on the Fe island A comparison of several islands Imaged usmg the same tip suggests also that these images involve tunnehng through several electronically different locations on the tip One cannot safely assume, therefore, that the high spots In the image of the A u ( l l l ) lattice can be identified with individual Au atoms, even in this case where the lattice could be imaged cleanly and reproduobly over several hours Similar am- blgmty results when one at tempts to interpret atomic-resolution images of the bare elbow dislo- cations, this precludes the verification of atomic models of the dislocation from existing data [19] Nonetheless, the lattice penod lo ty and the pres- ence of a dislocation can be reliably deduced from the ~mages

On a broader scale the mterpre tahon is much more rehable For example, ~sland s~zes and shapes are easily observed at 5 A lateral resolu- tion ff tips are sufficiently sharp One must still remain cautious about possible tip artifacts, how- ever An STM operating poorly, as when a t~p

change reduces feedback-loop oscillation, can change the sample In gross or subtle ways The t~p-crash-mduced steps of fig lb are an example of a gross change More subtly, tip effects have been observed to cause Nl islands on A u ( l l l ) to break up into smaller islands Similar problems for TEM of small particles are well known [20]

When Au or Ag is deposited, the growth struc- ture ~s very different from the island arrays seen for Nl and Fe on A u ( l l l ) For Au on Au, most growth takes place at the step edges at low cover- ages, but a few ~slands have been seen at the elbow sites, so the effect of the dislocations is lmtlally qmte weak [18] However, the growth of small Au Islands dramatically d~storts the recon- structxon, as seen m figs 3a and 3b, and elimi- nates the herringbone order The result is the appearance of relatwely large islands which do not form an ordered array The formation of multllayers ~s again evident at coverages near 0 5 ML The growth mode is very different for films of Ag on A u ( l l l ) as the elbow dislocations ap- pear to be completely mactwe so that all growth takes place at step edges [15] In this case, Ag grows from initial Au step edges m the form of monolayer-hlgh peninsulas enclosing small mono- layer voids, as shown m fig 3c These penmsular growths are explained by the instability of dlffu- slon-hmlted growth [15] Close scrutiny of the data shows that the growing Ag deposits do per- turb the reconstruction, as shown m Fig 3d, and that the reconstruction in turn affects the Ag aggregation

It ~s remarkable that each different metal we have studied is readily identifiable from its growth morphology on A u ( l l l ) A brief study of sputter-deposited films shows that the deposition technique can also produce readily dlStlngmsh- able growth morphology Fig 3e shows the struc- ture of a submonolayer Ag film deposited on A u ( l l l ) by sputtering with 500 eV Ar ions The island sizes are much smaller, as though nucle- ation sites were created during the sputter depo- smon This leads to a more homogenous island film More systematic work on sputter deposition appears possible but it would be difficult to ad- dress the wide variety of sputtering conditions employed to produce various devices

218 R J Wtlson et a l / Re~olutton m the scanning tunnehng mtc ros~ope

One important hm~tat~on of STM as an ~dentxfy- mg atomic types Interestingly, it has been found that the height of metal adsorbate islands ob- served m the STM agree, w~thm experimental errors, with the heights predicted from hard spheres wlth radxl taken from the metal lattice constants For Ag and Au, the lattice match to A u ( l l l ) is close and no height differences are observed, but substrate and room-temperature deposits can be distinguished from the behavior of the reconstruction For Fe and NI, the mis- match IS substantial and the Island heights are quite distinct from the A u ( l l l ) step helghts [18] In most of these cases, the images seem purely topographic since there Is no ewdence for signifi- cant electromc structure effects, and interpreta- tion seems straightforward Amb~gmtles, such as possible intermixing or the unlikely inversion of the deposit and substrate, cannot be absolutely ruled out with existing data Occasionally, certain t~ps g~ve rise to unusual height differences for the substrate and deposit [18] These may be due to adlayer-dependent t ip-surface interactions whlch m~ght be probed w~th atomic force microscopy to prowde insight into the tip dynamics

(c)

7. STM Imaging of molecules on metals

STM studies of molecules are particularly dif- ficult because molecules are typically weakly bound to surfaces and are not electrically conduc- tive One thus antxclpates tip contact and contam- ination problems when the layer thickness ex- ceeds the nominal 10 A, tunnel gap For small molecules, thermal diffusion and t ip-molecule forces are problematic Although many of our attempts to image molecules have not succeeded, we have obtained good results for flat-lying aro- matics on non-noble metal surfaces

The (3 × 3) ordered superlattlce of coadsorbed benzene (C~,H6) and CO on Rh(111)was studied first [21] The unit cell contains one flat-lying benzene molecule and two upright CO molecules, all chemlsorbed over hcp-type threefold hollow sites directly over second-layer Rh atoms At tip voltages V T s - - 0 0 1 0 V and tunnehng currents t v = 2 hA, 2 A high, threefold symmetric rlng-llke

Ftg 4 (a) A 60 ,~ wide image of benzene and carbon monox- ide molecules coadsorbed on a R h ( l l l ) surface Individual benzene molecules appear as distorted donut-shaped struc- tures (b) A z;00 A wide image of Cu-phthalocyanme molecule,, on Cu(100) Below this coverage, defects make It difficult to be certain that images arise from deposited molecules (c) An image where a mid-frame tip change eliminates atomic corru- gation and reveals distorted molecular images (d) A ~00 ,~

wide image ol fullerenes on Au(l 11)

R J Wdson et al /Resolutton m the scannmg tunnehng mtcroscope 219

s t ruc tures were obse rved which could be associ- a t ed wtth the benzene molecules , as shown in Fig 4a, a l though the CO molecu les were not resolved Signif icant d i s tor t ions m many of the htgh-resolu- t lon images of ten e l imina ted all expec ted symme- try None the less , these th ree fo ld s t ruc tures were found r epea t ed ly and a p p e a r e d to have a ftxed o r i en ta t ion r e l a twe to the subs t ra te Thts th ree- fold symmetry is p re sumab ly due to the mte rac - t lon of the ~- bonds of the b e n z e n e molecu les with the rhod ium subs t ra te a toms be low Here , the S T M was tmagtng empty s ta tes nea r E F whtch must be mostly metal l ic , with some con t r ibu t ton f rom hybr id iza t ion with mo lecu la r s ta tes Never- theless, It would not be surpr is ing if t i p - s a m p l e in te rac t ions were mvolved m the cont ras t mecha- msm, as in te rna l s t ruc ture was not observed at 1 V biases C o m p a r i s o n of STM images nea r a s tep edge t aken abou t ten m m u t e s apa r t showed evi- dence of b e n z e n e molecu la r diffusion be tween images [21] In some cases, molecu les which ap- p e a r e d diffuse in the tmage, p robab ly due to the i r diffusion dur ing the m e a s u r e m e n t of the S T M image, s e e m e d to shift into thei r favored (3 × 3) la t t tce pos i t ions This diffusive mot ion a p p e a r e d to not al low h igh- reso lu t ion m e a s u r e m e n t s in non-c lose -packed regtons of the sample These s tudies were ex tended to the c(2v~ × 4)rect over- layer whose pr lmt t lve umt cell conta ins one f lat- lying b e n z e n e molecu le and only one C O molecu le [22] In those studies, C O molecu les were resolved spat ia l ly for the first t tme, with the small p ro t rus ions assoc ta ted with the CO appea r - mg approx ima te ly one th i rd of the height of the t h ree fo ld r ing-l ike s t ruc tures resolved for the benzene molecu les

W e next chose to examine Cu-ph tha locyan lne molecu les on d i f fe ren t subs t ra tes [23] This molecu le has a dis t inct ive c lover- leaf shape and can be sub l imed to form e i the r submono laye r or mul t I layer depos i t s W e did not o b t a m useful Images for A u ( l l l ) or S I ( l l l ) subs t ra tes F o r a Cu(100) subs t ra te , we found tha t h igh- reso lu t ion images could be o b t a i n e d up to mono laye r cover- age, bu t tha t tmages of second- layer molecu les were m a d e q u a t e to ident i fy m o l e c u l a r s t ruc ture Fig 5b shows a wtde-scan image of a low-cover- age depos t t where a cons tde rab le n u m b e r of de-

Fig 5 (a) A 90 ,~ wide image of the transition region between clean (top) and Ag-coated (bottom) Sfflll) Data processing algorithms have been used to adjust the corrugation amph- tudes so that they can be simultaneously wsuahzed (b) A S~(lll) mesh gwmg the posmon of bulk S~ atoms superim- posed on the previous data (c) Image of Cu/Sffll l) "5 ×5" structure w~th superimposed mesh showing d~screte spacings

of triangular subunlts

220 R J VVtlson et a l / Resolunon m the scanning tunnehng mwroscope

fects can be seen in ad&tlon to stable, isolated molecules of two onentahons In efforts to obtain high resolution, it turned out to be useful to compare images of molecules w~th differing rota- honal orientations in order to ehmmate t~p asym- metries The fine structure in the resulting im- ages were in reasonable accord with simple Huckel calculations of the molecular electromc states which predicted valence charge densities locahzed on specific C atom p states [23] If the bins voltage was reduced so that the Cu(100) corrugation was perceptible, then the molecular images became more unstable, or no molecules were detected at all Fig 5c is a "bad" image where a t~p giwng a reasonably stable atomic corrugation image suddenly loses atomic resolu- tion and begins to generate noisy, multiple-tip images of molecules Has the tlp p~cked up a molecule to lose the atomic corrugation and form the multlple-zmages of molecules or has a non- conducting molecule fallen off so that reduced contact interactions account for the reduced atomic corrugatlon'~ Although the observation of four petals of the cloverleaf was common, the common observation of small &storhons and tip changes is of great concern in attempting to un- derstand internal molecular structure from STM data alone Near 1 ML coverage, two rotational domains, which correspond to the molecular ori- entations present at low coverages, were easily resolved Unusual molecular binding s~tes at step edges were also directly observed Good STM images of Cu-phth were not obtainable for cover- ages higher than 1 ML on Cu(100), or at sub- monolayer coverages on A u ( l l l ) or $1(111) Fad- ure in high-coverage ~mages may have been asso- ciated with a build-up of molecules on the tip, resulting m poorly conducting contacts For Au( l l l ) , low actlvat~on barriers allowing rota- tions or translations of molecules may contribute to smearing of the ~mages The ablhty to choose a substrate on which the molecules are bound by a h~ghly corrugated molecule-surface interaction potential appears to be essential for high-resolu- tion imaging of small, isolated molecules Tem- perature remains an important, but largely unex- plored, variable m these measurements

More recently, STM has been used to examine

ordered arrays of Buckminstertullerene (Bf) on a Au(111) sample [24] Small quantities of a mix- ture of C60 and C70 were sublimed onto the Au(111) crystal used in the metal-deposition stud- les discussed previously The images revealed an ordered hexagonal array of molecules wlth a 10 spacing, as in fig 4d This spacing was consistent with a 7 .~ diameter soccer ball We were inter- ested in the hmltatlons on the vertical height of molecules which can be studied w~thout excessive tip-sample contact This hmltatlon Is ~mportant here because molecular corrugations are smaller than the molecular diameter, which implies small t ip-molecule separations Indeed, the images be- came noisy at boundaries of the close-packed ~slands, or at elevated protrusions which might represent second-layer molecules Although ~t was straightforward to obtain these images, especially since the close-packed corrugation was a few A, tip changes were frequent, and 0 1 A stablhty was not obtained Internal structure was not ob- served, possibly due to molecular rotations which are now known to occur m bulk Bf [25]

These examples reveal a variety of phenomena in STM imaging of molecules, ranging from com- plete failure to the ablhty to observe internal features which correlate with crude calculations of internal electronic structure Tip-sample in- teractions are frequently apparent, hmltlng the value of more accurate quant~tatwe calculations It is still difficult tO pre&ct whether a given molecule on a given substrate will provide rea- sonable images, but there ~s a good chance for success if thin molecules form tightly packed ar- rays, or if aromatic molecules are deposited on metals hke Cu, Pt, or Rh

8. Semiconductors

STM studies of semiconductors have grown dramatically m recent years, and we have not kept pace w~th other groups These systems differ from metals and molecules in that high-resolution images are relatively easy to obtain and often show strong electromc structure variations This ~s in part due to the fact that corrugations are often near 1 .A so that 01 A effects do not

R J Wtlson et al / Resolutton m the scannmg tunnehng mtcroscope 221

dominate and operation at small t ip-sample spacing IS not necessary In many cases previous LEED and photoemisslon work has already re- suited in a good understanding of both geometric and electronic structure, so that much more de- tailed interpretations of the bias dependence of images is possible For unknown structures, how- ever, the situation is complicated because of the extensive rearrangements involved in many semi- conductor reconstructions and metal overlayers

Some of the strengths and hmltatlons of STM are apparent in our STM work on metals on the S1(111) surface [2hi Fig 5a shows an image oh- tamed on a SI(111) 7 × 7 crystal onto which 0 1 ML of Ag was evaporated at 600°C Large, clean areas of the 7 × 7 reconstruction are found at the top of this image, but near step edges a distinct v~ × v~ honeycomb structure appears on both upper and lower terraces STM images [8,28] of the 7 × 7 reconstruction offered strong support of formulating Takayanagl's d lmer-adatom-stack- lng fault model [27] which is now generally ac- cepted The honeycomb features arise from an- other well known A g / S I structure which had been previously studied by many techniques Two STM papers by different groups appeared with similar data but different interpretations, namely the protrusions were associated with Ag atoms in one paper [26] and Sl atoms in another paper [29] In an attempt to further establish the iden- tity of the protrusions, we examined boundaries between the clean and metalhzed surface and superimposed a mesh corresponding to the bulk SI atom positions as determined from the Sl adatom positions in the 7 × 7 region, as in fig 5b Although this method seems to adequately spec- ify the positions of the protrusions, which do not shift with bias voltage, one cannot unambiguously identify these protrusions as Ag or S1 atoms, or rule out the posslblhty that they represent charge density localized away from any atomic cores In fact, this STM work helped to stimulate interest m Ag/Sl(111) which has led to a number of more complex models which should again be tested by STM [30] However, with this lesson in mind, it would be foolish to simply repeat our measure- ments because it is clear that complex spatial structures cannot be determined from STM data

without a thorough knowledge of the resulting electronic structure It should be noted that both the interpretations of measurements using other techniques and the results of electronic structure calculations have also changed within the decade

Fig 5c shows a view of the C u / S I ( l l l ) "5 × 5" surface [31] Little was known about this surface beyond some Auger coverage measurements and LEED patterns The Image here shows many protrusions with a 1 × 1 spacing and an irregular pattern The corrugations are small, near 0 1 A, as might be expected for closely spaced features Triangular pockets are evident in this image and form an irregular quasi-5 × 5 array No attempt has been made to develop an atomic model for this complicated surface, but some insight into the type of lncommensuratlon was obtained This type of data is thus questionable if details of atomic positions are required but is of value if qualitatwe insights into packing and uniformity are desired Indeed, more detailed treatment of variations in unit cell structures have received considerable attention since this early work [32,33]

The frustrating complexities discussed above have resulted largely from examining large cell structures whose reconstructions can extend a few layers below the surface Numerous outstand- ing studies of simpler dImerlzed surfaces have been more successful and are known throughout the literature Recent studies of nucleation and diffusion phenomena during semiconductor epl- taxy provide tremendous Insight without requir- ing controversial Interpretations of STM data [34]

9. Conclusion

This brief review is Intended to convey a sense of the limitations of interpretation of STM which are associated with uncertainties regarding reso- lution and contrast in STM There are many problems, mysterious effects, and continual con- troversy Nonetheless, STM has provided many new insights into surface morphology and contin- ued improvements are expected Nearly ten years have passed but we are still in the early stages of exploring a wide range of phenomena which can

222 R J Wtlson et a l / Resolutton m the scanning tunnehng mtcroscope

b e v i e w e d , a n d pos s ib ly c o n t r o l l e d , o n t h e m i c r o -

s cop ic scale

Acknowledgement

T h i s w o r k was p a r t m l l y s u p p o r t e d by t h e O f -

r ice o f N a v a l R e s e a r c h ( N 0 0 0 1 4 - 8 9 - C - 0 0 9 9 )

References

[1] G Bmmg and H Rohrer, IBM J Res Dev 30(1986)355 [2] S Chlang, R J Wdson, Ch Gerber and V M Hallmark,

J Vac So Technol A 6 (1988)386 [3] J Tersoff and D R Hamann, Phys Rev Lett 50 (1983)

1998, E Stoll, Surf Scl 143 (1984) L411

[4] N D Lang, Phys Rev Lett 56(1986)1164 [5] J M Soler, A M Baro, N Garcla and H Rohrer, Phys

Rev Lett 57 91986) 444 [6] H J Mamm, E Ganz, D W Abraham, R E Thomson

and J Clarke, Phys Rev B 34 (1986) 9015 [7] J B Pethlca, Phys Rev Lett 57 (1986)3235 [8] R J Hamers, R M Tromp and J E Demuth, Phys Rev

Lett 56 (1986) 1164 [9] R M Feenstra. J A Strosoo, J Tersoff and A P Fern,

Phys Rev Lett 58 (1987)2575 [10] V M Hallmark, S Chlang, J F Rabolt, J D Swalen and

R J Wdson, Phys Rev Lett 59 (1987)2879 [11] J Wlntterhn, J Wlechers, H Burne, T Grltsch, H

Hofer and R J Behm, Phys Rev Lett 62 (1989) 59 [12] N J Zheng and I S T Tsong, Phys Rev B 41 (1990)

2671 [13] C J Chen, Phys Rev Lett 65 (1990)448 [14] S Clrao, A Baratoff and I P Batra, Phys Rev B 42

(1990) 7618

[15] D D Chambhss and R J Wilson, J Vac Scl Technol B 9 (199l) 928

[16] Ch Woll, S Chmng, R J Wdson and P H Llppel, Phys Rev 1339 (1989) 7988

[17] K Takayanagl, Y Tamshlro, K Yag,, K Kobayashl and G Honjo, Surf So 205 (1988) 205

[18] D D Chambhss, R J Wilson and S Chlang, J Vac Scl Technol B 9 (1991) 933

[•9] D D Chambhss, R 3 Wilson and S Chlang, Phys Rev Lett 66 (1991) 1721

[20] S hjlma and T lchlhashl, Phys Rev Lett 56 (1986)616 [21] H Ohtam, R J Wdson, S Chlang and C M Mate, Phys

Rev Lett 60 (1988) 2398 [22] S Chlang, R J Wilson, C M Mate and H Ohtam, J

Microscopy 152 (1988) 567 [23] P L~ppel, R J Wdson, M D Mdler, Ch Woll and S

Chmng, Phys Rev Lett 62 (1989)171 [24] R J Wilson, G Me0er, D S Bethune, R D Johnson,

D D Chambhss, M S DeVrles, H E Hunzlker and H R Wendt, Nature 348 (1990) 6302

[25] CS Yannonl, R D Johnson, G Me0er, D S Bethune and J Salem, J Phys Chem 95 (1991) 9

[26] R J Wdson and S Chlang, Phys Rev Lett 59 (1987) 2329

[27] K Takayanagl, Y Tamshlro, S Takahashl and M Taka- hash1, Surf So 164 (1985) 367

[28] R S Becker, J A Golovchenko, E G McRae and B S Swartzentruber, Phys Rev Lett 55 (1985) 2028

[29] E J van Loenen, J E Demuth, R M Tromp and R J Hamers, Phys Rev Lett 58 (1987) 373

[30] M Katayama, R S Wdhams, M Kato, E Nomura and M Aono, Phys Rev Lett 66 (1991)2762

[31] R J Wdson, S Chlang and F Salvan, Phys Rev 38 (1988) 12696

[32] J E Demuth, U K Koehler, R J Hamers and P Kaplan, Phys Rev Lett 62 (1989)641

[33] K Mortenson, Phys Rev Lett 66 (1991)461 [34] Y W Mo, J Klemer, M B Webb and M G Lagally, Phys

Rev Lett 66 (1991) 1998