Hawkes P.W., Spence J.C.H. (Eds.) Science of Microscopy. V.1 and 2

Подождите немного. Документ загружается.

Chapter 8 LEEM and SPLEEM 635

The ﬁ rst-order transition at about 1005 K from the high-temperature

“(1 × 1)” to the low-temperature “(16 × 2)” structure of the Si(110)

surface and the stress-stabilized coexistence of the two phases was the

subject of another study.

148

Similar to the work on Si(111) and Si(100),

various surface thermodynamic data have also been obtained for the

Si(110) surface from island decay measurements.

149

Finally, an LEEM

study of the (2 × 1) to (1 × 1) phase transition above 925 K on Ge(100)

led to the conclusion that the transition was due to dimer breakup and

roughening.

150

Because of its importance in the semiconductor industry Ge or more

precisely SiGe on Si(100) has been the most-studied system.

151–161

This

system initially forms a several monolayer thick layer that is highly

strained depending upon alloy composition. From this layer three-

dimensional islands develop with facets that depend upon size and

composition, leading to a rich variety of phenomena well suited for

LEEM studies. The transition from the initial layer to the three-

dimensional islands is strain driven and does not require three-

dimensional nucleation.

156,157,159

Surface steps play a critical role in the

alloying of Ge with the Si(100) surface.

160

The results up to 2000 are

summarized in an excellent review.

161

The growth of Ge on other Si

surfaces has been studied in much less detail. On the (111) surface the

inﬂ uence of the surfactant Sb on the growth was studied,

162

on the (311)

surface the transition from the two-dimensional layer to three-

dimensional islands.

163

These showed a much more complicated facet

structure than on Si(100). Finally, the growth of Ge on GaAs(100) has

also been studied brieﬂ y.

164

The growth of metals on semiconductor surfaces can also be studied

very well with LEEM. Most of the work was done on Si(111) surfaces,

some also on Si(100), using video recording of the growth, dif-

fusion, ordering, and disordering processes. Au on Si(111) is a good

example: after several two-dimensional superstructures have formed

with increasing coverage, three-dimensional particles grow that show

an interesting temperature dependence due to the formation of an

Au–Si eutectic.

165,166

On vicinal Si(100) surfaces with a miscut of 4°,

adsorption of a monolayer of Au causes pronoun ced faceting at ele-

vated temperatures.

167–170

The growth of Ag on Si(111) has also been

studied extensively,

162,171–174

in particular the growth shape,

172,174

and has

been used to demonstrate that buried interfaces can be imaged via

their strain ﬁ elds.

173

In the various studies of Ag on Si(100)

175–177

a strain-

induced shape transition of Ag crystals into “quantum wires”

175

and

bamboo-like growth

178

have been observed. Substrate material incor-

poration into the two-dimensional superstructure

177

and the inﬂ uence

of steps on the domain ordering in it

176

were found. On vicinal Si(100)

self-assembled Ag “quantum wires” form at high temperatures.

179

The

growth of Cu on Si(111) has been studied both on the clean and on the

hydrogen-passivated surface.

180–184

Hydrogen termination strongly

inﬂ uences the formation of the two-dimensional “(5 × 5)” superstruc-

6.5 Thin Films on Semiconductors

636 E. Bauer

ture and the three-dimensional nucleation of Cu silicide. At higher

temperatures at which hydrogen is desorbed large Cu

Si crystals in a

variety of shapes and facets form.

In Al ﬁ lms on Si(111) the two-dimensional phase diagram and the

phase transitions between the phases have been studied.

185

The growth

of Pb on Si(111) has been used to obtain an understanding of how

interfactants, that is substrate surface layers that remain at the interface

during growth, produce quasi-monolayer-by-monolayer growth, using

Au and Ag as interfactants.

186–188

Figure 8–32 illustrates the inﬂ uence

of an Au interfactant layer on the growth of Pb. On the Si(100) surface

Pb grows in <111>-oriented crystals on the two-dimensional initial

layer. The crystal, have frequently “ashtray” shape due to the large

supply of Pb atoms by diffusion on the initial layer.

189

On vicinal and

high index surfaces of Si, which have been faceted by Au adsorption,

Pb grows in mesoscopic wires.

190,191

In was found not to wet the Si(111)-

(7 × 7) surface but to grow at low temperature monolayer-by-monolayer

on the Si(111)-(

33×

)-R30° surface with a variety of superstruc-

tures. Three-dimensional crystals that grow at somewhat higher

temperatures have predominantly (100) orientation with a surface

reconstruction similar to that of the underlying two-dimensional

layer.

192,193

The Si(100) surface is etched at high temperatures by In

similar to the situation mentioned above for Ag, at lower temperatures

a superstructure forms.

194

Transition metals are highly reactive with Si and form silicides that

frequently are more stable at high temperatures than Si. An example

is Co. When deposited or annealed at high temperatures Co layers form

more or less hexagonal CoSi

crystals that show misﬁ t dislocation

contrast when thin enough. When heated to temperatures at which Si

sublimes CoSi

-topped hillocks form because of the lower vapor pres-

sure of CoSi

(Figure 8–33).

97,195

At lower temperatures the crystals

Figure 8–32. Images taken during the growth of Pb on Si(111) at 290 K with

(top) and without Au surfactant (bottom). Electron energy 8 eV.

186

Chapter 8 LEEM and SPLEEM 637

are triangular and act as Co scavengers cleaning the surrounding

surface from Co so that it develops the (7 × 7) structure of the clean

surface. At very low Co coverages an interesting “ring cluster” (RC)

phase forms, which has been studied in considerable detail.

196–198

Ni forms a similar RC phase.

199

Another transition metal, Ti, spon-

taneously forms Ti silicide “nanowires” when deposited at about

1120 K. Their formation process and stability have been studied in

detail.

200

Only a few nonmetal ﬁ lms on Si have been studied with LEEM: CaF

and Si nitride. In CaF

ﬁ lms both the complexities of the formation of

the ﬁ rst two layers have been studied

201

as well the formation mecha-

nism of the interfacial dislocation network with increasing ﬁ lm

thickness.

202

The Si nitride work was concerned mainly with the forma-

tion of the initial epitaxial layer by reaction with NH

at high tempera-

tures, which occurs via nucleation and growth of nitride domains

similar to the (7 × 7) domains on the clean surface.

203

Very little work has been with LEEM on these materials. There is a

cursory study of the temperature dependence of the structure of the

6H-SiC(0001) surface used as a substrate in the growth of GaN layers.

204

The structure of the GaN(0001) surface was studied with dark-ﬁ eld

imaging, which allowed determination of the surface termination,

205

similar to that of the SiC(0001) surface.

204

The importance of the Ga/N

ratio in homoepitaxial growth of GaN was demonstrated in several

papers,

204,206,207

in particular the necessity of a Ga double layer on top

of the GaN surface for the growth of ﬁ lms with (0001) surfaces.

204,207,208

The double layer shows an interesting phase transition.

204

Without an

adsorbed Ga layer the GaN ﬁ lms grow rough, with {10-11} and {101-2}

facets. GaN growth on 6H-Si(0001) is similar

209

and not as reported in

another paper,

210

except that initially three-dimensional crystals form

instead of the spiral and step ﬂ ow growth in homoepitaxial growth on

GaN.

Figure 8–33. Hillocks on an

Si(111) surface caused by CoSi

crystals during the sublimation

of Si.

6.6 Wide-Band Semiconductors

638 E. Bauer

Refractory metal surfaces have been the most popular subject in LEEM

because of their high melting point and low vapor pressure that allow

experiments over a wide temperature range. W and Mo can be cleaned

easily by heating in oxygen. The chemisorbed oxygen is then ﬂ ashed

off at high temperatures, a procedure that is not successful in Nb and

Ta, in which oxygen goes into solid solution and can be partially

removed only by lengthy sputter and annealing cycles. The imaging

of the step structure on the Mo(110) surface (Figure 8–23) was one of

the ﬁ rst demonstrations of the power of LEEM and step contrast has

been one of the mayor tools in the study of surface processes on clean

surfaces. While W(110) and Mo(110) have been used frequently as sub-

strates for thin ﬁ lms, and to a much lesser extent also W(100) and

W(111), little has usually been published about the clean surface, except

for a brief study of the Mo(110) surface.

211

Extensive work was done,

however, on epitaxial Mo(110)

212–217

and Nb(110)

215,218–220

layers grown

on sapphire (11–20) surfaces at high temperatures, which after proper

cleaning produces surfaces that are comparable in quality to single

crystal surfaces. A complication is the interfacial strain and the disloca-

tions introduced upon cooling and thermal cycling by the different

thermal expansion coefﬁ cients between ﬁ lm and substrate. Neverthe-

less, pure surface quantities such as the step stiffness could be extracted

from such ﬁ lms. In the case of the Nb(110) ﬁ lms the situation is

complicated by the residual oxygen. This causes reconstruction and

faceting,

215,218,219

which in themselves are interesting processes and are

useful for the study of extended defects.

220

Some work on Ta(110) ﬁ lms

was also done.

215

Noble metal surfaces have also been the subject of several LEEM

studies. For Pt(111) single crystal surfaces step stiffness, step–step

interactions, step free energy,

215,217,221

and bulk-surface vacancy

exchange

222,223

have been determined. For Pd(111) surfaces the step

stiffness

217

has been obtained and sputter erosion processes

224

have

been observed. Studies of the island decay on Rh(100) indicated a new

surface diffusion process.

225

Step ﬂ uctuation spectroscopy of Au(111)

yielded the surface mass diffusion coefﬁ cient and the orientation-

dependent step stiffness.

226

Dark-ﬁ eld imaging of the reconstructed

Au(100) surface was used to establish the connection between the

reconstruction domains and the step orientations (Figure 8–34).

80,227,228

On the Ag(111) surface a critical island size for layer-by-layer growth

was found.

229

From an investigation of the homoepitaxial growth of Cu

on Cu(100) the Ehrlich–Schwoebel barrier, the energy barrier for diffu-

sion across a step, was deduced.

230

Similar to other fcc(110) surfaces, the

Pb(110) surface reconstructs. A LEEM study of the various reconstruc-

tions, some of them alkali induced, revealed the topography of the

various phases and the inﬂ uence of surface defects on the transitions

between them.

231,232

Finally, studies of the surface morphology of the

NiAl(110) surface demonstrated the importance of bulk diffusion for

surface smoothing.

233

Most experiments on clean surfaces rely on the

6.7 Metal Surfaces

step contrast discussed in Section 5, wh ic h illustrates its usefulness.

Chapter 8 LEEM and SPLEEM 639

Although the growth of many metals on W(110) and Mo(110) has been

studied with other methods, LEEM has been used infrequently, often

only in a very cursory manner. Cu is the most investigated layer mate-

rial, both on Mo

86,123,166,234,235

and W. The growth on both substrates is

similar and has been studied in detail on W(110).

236

Layer spacings

have been determined using the quantum size effect discussed in

237

The more qualitative work on Mo revealed an interesting

striped phase upon annealing at high temperatures and details in the

structural phase transition in the double layer. Layer spacings have

also been obtained from the quantum size effect for Ag on W(110).

Au has been studied brieﬂ y on Mo(110) in the submonolayer range

where it forms needle-like crystals.

123,235

In Pd layers on W(110) the

transition from two- to three-dimensional growth was the subject of a

combined LEEM–XPEEM study.

238

The growth of Ag and Cu on the

Ru(0001) surface served as a demonstration of the inﬂ uence of

substrate steps on the growth of three-dimensional crystals.

239

In the

not modiﬁ ed or is modiﬁ ed only a little by the growing ﬁ lm. This is

not the case on less densely packed surfaces such as the W(111) surface

that facets upon deposition of certain metals at high temperatures. Pt

growth on W(111) is a case study for this process.

240,241

The growth of

with SPLEEM.

Figure 8–34. Au(100) surface. Bright-ﬁ eld image (a) and dark-ﬁ eld images

(c and d) taken with the “(5 × 1)” superstructure reﬂ ections indicated in the

LEED pattern (b). Electron energy in the images 16 eV.

227

6.8 Metal Layers on Metals

Section 5.

ferromagnetic layers will be discussed in Section 7 in connection

metal-on-metal systems discussed up to now the substrate surface is

640 E. Bauer

In many cases the deposited metal forms a surface alloy with the

substrate. Pd on Mo(100) is an example that was studied in detail by

LEEM up to several monolayers.

242

Clear alloying was observed up to

a monolayer followed by two-dimensional Pd growth without faceting.

Alloying of Sn with Cu(111) is strikingly different: at very low cover-

ages large two-dimensional islands form that travel across the surface,

leaving alloy behind and, thereby, decreasing in size.

243

Pb forms on

the Cu(100) surface initially several two-dimensional structures

followed by the growth of three-dimensional crystals. LEEM clearly

shows the correlation between the crystals and the initial layer,

244

evidence for the surface alloying in the initial layer at low coverages,

followed by dealloying.

245,246

The nature of the disordering transitions

of these phases was also determined.

246

One of the most impressive

results that LEEM has produced up to now is the self-assembly of stress

domain patterns in the two-dimensional Pb–Cu alloy on Cu(111),

which consist of domains of a Pb-rich and Pb-poor phase. This process

has been studied in great detail, which has produced a wealth of infor-

mation on stress-induced ordering phenomena

247–251

and has been well

reviewed.

252

Finally, the inﬂ uence of the interface between Pb droplets

and the Cu(111) surface on the shape and melting of the droplets has

been the subject of an LEEM study.

253,254

Although LEEM is well-suited for the study of reactions on surfaces

with gases or impurities from the bulk, very little work has been done

up to now. Segregation of impurities from the bulk and the formation

of precipitation products on the surface are routinely observed before

the crystal has been cleaned completely. Surface carbide formation on

W and Mo surfaces is an example. Images of such carbides and of car-

bides formed by CO dissociation have sometimes been published,

123,213,234

but the precipitation process was never studied in detail. As far as

oxidation is concerned, only three systems have been studied: the

initial oxidation of W(100),

255,256

the low oxygen coverage region on

Nb(100),

257

and the growth of oxide domains on NiAl(110).

258

All three

studies show interesting unexpected phenomena.

More work has been done on surface reactions in which the reaction

products desorb, that is in heterogeneous catalysis. The ﬁ rst studies

looked with very limited resolution at reactions of CO with O

259,260

with NO

261

and at the reaction of NO and H

262

mainly at the propaga-

tion of the reaction front and its pinning by defects. On Pt(110) pattern

formation in the CO oxidation has been studied

263,264

and on Rh(110)

the reaction of NO

265

and of O

with H

266,267

Most of this work has been

done with low resolution, but the possibilities of LEEM in this particu-

lar ﬁ eld are evident in some studies.

265,267

The contrast is due to the fact

that surface regions with different reactant composition have different

structures, which produce characteristic LEED patterns. These can be

used for dark-ﬁ eld imaging of these regions.

265

Such dark-ﬁ eld images

are shown in Figure 8–35 of the NO + H

reaction on Rh(110). The

propagation of a spiral wave reaction is imaged with the diffraction

6.9 Reactions on Metal Surfaces

Chapter 8 LEEM and SPLEEM 641

spots of pure N phases (a and b), a mixed N + O phase (c), and a pure

O phase (d), which occur during the oscillatory reaction. Because of

the high brightness, LEEM is preferable to PEEM because it allows fol-

lowing the kinetics of the reaction wave propagation and it is superior

to MEM because of its better resolution.

The only LEEM work on oxides published up to now is that on the

TiO

(110) surface.

268–271

TiO

is a particularly good example of the inﬂ u-

ence of the exchange of defects between the bulk and the surface on

its structure because its stoichiometry can vary considerably. On the

nonstoichiometric surface a (1 × 1) to (1 × 2) phase transition occurs as

a function of temperature due to vacancy exchange between the bulk

and the surface, which was studied thoroughly.

268–270

When exposed to

oxygen, the nonstoichiometric, surface grows via step ﬂ ow at high

temperatures but via two-dimensional nucleation at lower tempera-

tures. With suitable growth conditions the surface topography can be

271

The

only nitride studied to date—silicon nitride and GaN, which have been

surface of TiN layers. These layers form mounds with spiral steps or

stacks of two-dimensional islands. The mass transport at high tem-

peratures can be studied well by following the step motion and the

diffusion processes and constants can be derived from it.

272,273

The applications discussed in this subsection clearly show the many

possibilities of LEEM in the study of surfaces and thin ﬁ lms of a wide

cal properties of surfaces and thin ﬁ lms has been extracted from these

studies. These can be found in the references cited. Diffraction contrast,

step contrast, and quantum size contrast, together with the real time

capability, are the essential features that make LEEM so powerful in

these studies. It should also be emphasized that LEEM is primarily an

imaging method for in situ studies. Sample exchange is more time-

consuming than in standard transmission and scanning electron

Figure 8–35. Dark-ﬁ eld images taken during the reaction of NO with H

on Rh(110) with LEED spots

characteristic for the various phases in the oscillatory reaction.

265

6.10 Oxides and Nitrides

6.11 Concluding Remarks

variety of materials. Important information on the physical and chemi-

discussed in Sections 6.5 and 6.6, respectively, excepted—is the (111)

changed and the oxygen content of the bulk can be increased.

642 E. Bauer

microscopy because the sample has to be transferred into ultrahigh

vacuum, usually after cleaning in a preparation chamber, and then

aligned in the microscope. These steps are necessary because of the

high surface sensitivity of the method and because the sample is part

of the objective lens so that its surface has to be exactly perpendicular

to the optical axis.

An important aspect of LEEM is that it can be easily combined with

other, in part complementary surface imaging techniques. MEM is

useful whenever no strong diffracted beam is available for imaging, for

example, on ﬁ ne-grained polycrystalline or amorphous samples. In this

case the sample potential is chosen such that the electrons cannot pene-

trate into the sample. Contrast is then determined by surface topography,

surface potential, and work function differences. MEM has been used,

for example, in the study of chemical reactions.

259,260

Ultraviolet light-

excited photoemission electron microscopy (UVPEEM) is probably the

most popular auxiliary imaging method in LEEM instruments. It is used

when ﬁ elds of view larger than those possible in LEEM have to be imaged

or in samples in which it produces better contrast than LEEM. A good

example is the study of the growth of pentacene ﬁ lms on oxidized Si.

274,275

The application range of UVPEEM is the same as that of MEM, but the

resolution is in general much better. Synchrotron radiation-excited

XPEEM provides chemical information and chemically speciﬁ c mag-

netic information. It can be combined easily with LEEM, in particular in

instruments equipped with an energy ﬁ lter, the SPELEEM mentioned in

Other imaging methods that are possible in LEEM instruments are

thermionic emission electron microscopy (TEEM). TEEM is useful in

general only for samples with locally varying work functions and that

can be heated high enough for thermionic emission. Metastable impact

electron emission microscopy (MIEEM) is another imaging method

with limited application range. In this method

276

deexcitation of meta-

stable He* atoms at the surface causes electron emission up to energies

of 15–20 eV. Because of the chromatic aberration resolutions of 100 nm or

less can be achieved only with a band pass energy ﬁ lter.

277

The main

application of MIEEM is in the study of adsorbates, which consist of

regions with different deexcitation processes. In LEEM instruments that

allow higher energies in the illumination system than in the imaging

system SEEM and in particular AEEM are possible. AEEM is useful for

chemical analysis but inferior to XPEEM because of the larger character-

istic peak width and the larger background. As in XPEEM, a band pass

energy ﬁ lter is indispensable for selection of a narrow energy band at

the Auger electron peaks. The energy ﬁ lter is also useful for SEEM

whose application range is similar to that of UVPEEM and MEM.

SPLEEM is a version of LEEM that requires a separate treatment

because it does not give structural but magnetic information. It differs

from LEEM only in that the illumination system produces a partially

7 Spin-Polarized LEEM (SPLEEM)

Section 3. XPEEM is discussed elsewhere in the book.

Chapter 8 LEEM and SPLEEM 643

spin-polarized electron beam. The polarization is achieved by illumi-

nating a GaAs (100) surface with circular polarized light with the

wavelength corresponding to the band gap of GaAs. The surface is

activated by Cs and O

exposure to negative electron afﬁ nity so that

electrons that have been excited to the bottom of the conduction band

can escape the surface. Optical selection rules produce a spin selection

in the excitation process such that the spin of the emitted electrons

points normal to the surface either inward or outward, depending

upon the helicity (right or left) of the exciting light. With specially

designed photocathodes spin polarizations up to 80% have been

achieved, but the ordinary cathodes used in SPLEEM usually have a

polarization of only about 20–30%. Details of this kind of cathode can

be found in Pierce.

278

After extraction from the cathode, the electron beam is deﬂ ected 90°

by a combined electrostatic–magnetic sector ﬁ eld. In pure electrostatic

deﬂ ection the direction of the spin polarization P is unchanged, in pure

magnetic deﬂ ection P is deﬂ ected 90°, and if both ﬁ elds are used for 90°

deﬂ ection P can be rotated in any direction in the plane indicated in

Figure 8–36. In the early SPLEEM studies

279–284

only electrostatic deﬂ ec-

tion was available. Later the magnetic rotator lens indicated in this

ﬁ gure was added, which now allows P to rotate in any direction in

space.

285

Usually, however, only three directions are selected, one normal

to the surface of the crystal and the other two in preferred directions in

the surface plane (easy and hard magnetic axes). The “spin manipula-

tor” shown in Figure 8–36 i s i n cor pora te d i n t he or i g i n a l LEEM d e s c r ib e d

50,286

Figure 8–36. Schematic

of the spin manipulator.

For explanation see

text.

285

in Section 3 (Figure 8–9) and in a second instrument.

644 E. Bauer

The magnetic contrast is due to the fact that the 180° backscattering

from magnetic materials depends upon the relative orientation of the

spin of the incident electrons and of the electrons in the material, or in

other words upon the relative orientation of the polarization P and the

magnetization M. The intensity in the image is then given by I = I

str

c P ⋅ M where I

str

is determined by the structure and topography of the

surface, c is a small proportionality constant, and the dot indicates the

scalar product. Pure magnetic contrast is obtained by subtracting two

images taken with opposite P direction pixel by pixel. Maximum con-

magnetic domains with opposite magnetization vanishes and only the

domain walls produce contrast. Maximum contrast requires imaging

at very low energies, typically a few electronvolts. At these energies

the exchange splitting of the band structure and the difference between

the inelastic mean free paths of the spin-up and spin-down electrons,

between the backscattering for the two spin directions. Because the

magnetic signal is only a small fraction of the total signal, the signal-

to-noise ratio in the difference image is small and frequently limits the

resolution.

287

Addition of two images with opposite P direction

produces only structural contrast, which makes SPLEEM ideal for

the correlation between magnetism and structure. More information

on magnetic contrast formation may be found in recent SPLEEM

reviews.

288,289

In most of the SPLEEM studies the ferromagnetic samples are pre-

pared in situ while observing the magnetic structure together with the

crystal structure and topography via several contrast mechanisms.

However, ex situ prepared samples can also be studied after proper

surface cleaning as illustrated in Figure 8–37 for a Co(0001) surface that

had been sputter-cleaned and annealed.

279

Ex situ prepared samples

can also be studied when passivated with a thin layer that is sufﬁ -

ciently transparent for slow electrons such as noble metals.

290

In this

method it has to be taken into account that such overlayers can change

the magnetization in the ﬁ lm below.

291,292

The in situ studies cover single

Figure 8–37. SPLEEM image of

the closure domains on a

Co(0001) surface. Electron

energy 2 eV.

279

which were mentioned in Section 2, cause the strongest difference

trast occurs obviously for P || ± M. For P ⊥ M the contrast between