Adsorption and electric measurements on germanium (111) surfaces covered with potassium and cesium

Surface Science 85 (1979) 413-431 0 North-Holland Publishing Company

ADSORPTION

AND ELECTRIC MEASUREMENTS

ON GERMANIUM (111)

SURFACES COVERED WITH POTASSIUM AND CESIUM

L. SURNEV and M. TIKHOV Institute of General and Inorganic Chemistry, Bulgarian Academy garia

of Sciences, 1040 Sofia, Bul-

Received 26 April 1978; manuscript received in final form 15 March 1979

Potassium and cesium adsorption on a Ge(ll1) surface and its effect on the surface conductivity, Au, surface recombination velocity, S, and work function changes, A9, has been studied. The K and Cs Thermal desorption spectra consist of multiple peaks. To each peak corresponds a straight line segment of the A9 versus coverage curves. The surface conductivity versus coverage curves successively pass through a minimum and a maximum. The decreasing parts of these curves are attributed to the appearance of alkali-metal-induced extrinsic surface states, whereas the increasing one are ascribed to the disappearance of the intrinsic states of the clean Ge(ll1) surface. The surface recombination velocity has been found to be multi-valued function of the surface potential. At a K overlayer coverage close to one monolayer new high-temperature peaks appear in the TD spectra due to irreversible surface reconstruction. 1. Introduction

The study of the changes in the electric and adsorption properties of the germanium surface during alkali metal adsorption is mainly of fundamental interest. As is known, adsorption on covalent surfaces is much more localized than adsorption on metal surfaces. The redistribution of the directed covalent bonds during adsorption causes rather unusual changes in the electric properties of the Ge surface [l-4]. In a previous paper [l] the adsorption of sodium and its effect on some electric properties of the Ge(ll1) surface was studied. In the present paper the same investigation is extended to potassium and cesium. The larger atom radii of these elements in comparison with sodium are the reason for their less complex thermal desorption spectra which ensures a more definite interpretation. A comparison of the results obtained with the three metals proved to be very useful for understanding the processes occurring on the Ge(ll1) surface during adsorption. 2. Experimental The investigations were carried out using several experimental methods, namely thermal desorption (TD) and measurements of the surface conductivity changes, 413

414

L. Surnev, hf. Tikhov /Adsorption

and electric measurements

Au, the surface recombination velocity, S, and the work function changes, Aq. TD spectra recording, and Au and S measurements, were performed in a glass tube with cylindrical symmetry and which was connected with the Pyrex UHV system described previously [l] . Work function and repeated Au measurements were performed in a stainless steel UHV chamber being pumped out by a 100 l/s ion pump and a liquid-nitrogen-cooled sublimation pump. The system was equipped with a nude Bayard-Alpert gauge and a quadrupole mass spectrometer (QMS). After baking the chamber up to 2.50°C, the base pressure reaches 1 X lo-” Torr (mainly CO, Hz and traces of HaO, Ar and COz). Fig. 1 shows the experimental tube and part of the UHV chamber (top view). The sample was mounted on a movable car which cduld be driven by means of a long screw rotated by an external magnet. The car slid along the MO rod (diam. 0.2 cm) which, together with the srew, was used as a current lead. This car enabled us to place the sample successively in front of the Ar ion gun 1, the alkali metal ion guns 2, and the W whiskers 3 (0.01 cm, magnetically moved) for Au measurements, as well as in front of the reference electrode 4 for work function measurements. The latter was a MO ribbon (0.5 X 0.3 X 0.02 cm) welded in a thin glass envelope covered with a conducting SnO layer. By means of an isolator the reference electrode was mounted at the end of the MO rod (diam. 0.1 cm) provided with a Ni anchor for resonance vibration (-120 Hz). The screw used for moving the car was slightly excentric. This facilitated the precise adjustment of the sample reference electrode distance. This, together with the application of a lock-in amplifier enabled

I I OMS

Fig. 1. Schematic diagram of experimental arrangement (top view): (1) sputtering alkali metal ion guns; (3) tungsten whiskers for Au measurements; (4) reference WF measurements; (QMS) quadrupole mass spectrometer; (BAG) Bayard-Alpert sample.

chamber; (2) electrode for gauge; (S) Ge

L. Sumev, M. Tikhov /Adsorption and electric measurements

415

us to measure work function changes with an accuracy better than 1 mV. The impurities in the K and Cs flows, as determined by the QMS, were less than 0.05%. The methods used for Au and S measurements and TDS recording were described elsewhere [l]. For the TDS experiments a surface ionization gauge with an Ir filament was used [l]. The gauge calibration for desorbed atoms and ions was performed by the following procedure. The same amount of the alkali metal N, was deposited two times. After the first deposition, the integral Sy of the atom current was recorded by the surface ionization gauge (a retarding potential was applied to prevent the ions from reaching the Ir filament and the collector). After the seond deposition the Ir filament was switched off and the integral Si of the ion current was recorded alone (without using a retarding potential). The same procedure was repeated for another initial coverag&, NZ, and the new values, $ and S’, , were obtained. Using these measurements, the ions and the neutral components nr, n;, ny, and n! of the desorbed flux could be determined from the equations n:+n;=Nl,

n8 tn’,=N2,

n:Jn! = S,“/Sl ,

nt/ni = $/S;

.

(1)

The samples with n-type conductivity and a specific resistance ranging from 20 to 30 R cm were rectangular plates (2.5 X 0.5 X 0.03 cm in the glass tube and 1.5 X 0.4 X 0.02 cm in the metal tube). Their largest planes were oriented in the (111) plane with an accuracy of 0.5’. The cleaning procedure, which was similar to that described in ref. [ 11, consisted of several cycles of Ar bombardment (450 V, 20 PA/cm’) and annealing at 850 K. Since we were unable to check the surface cleanliness by AES, stringent requirements for Ar purity were imposed.

3. Results 3.1. Surface conductivity and surface recombination velocity ‘Fig. 2,shows the surface conductivity changes, Aa, as a function of the K and Cs coverages, 0. For the sake of comparison, the same dependence obtained for sodium adsorption is indicated with a dashed line. This curve is in good agreement with the similar one obtained in a previous study [l]. The changes in Au are relative to an arbitrary zero, i.e. the clean surface. A logarithmic scale is chosen for the coverage. We assume that 0 = 1 monolayer (ML) when one alkali atom corresponds to each Ge atom. It is evident that the curves for K and Cs show the same peculiarity as that obtained for Na. Some general features of the results in fig. 2 can be summarized as follows: (1) The curves minimum is observed at the same coverage (-0.03 ML) with the three metals and its depth increases from Na to Cs. (2) The curve maximum is shifted towards lower coverages and increases in height in the sequence NaCs.

L. Surnev, M. Tikhov /Adsorption

416

and electric measurements

US

120

-12 : 80 Q: co E LO lo 4

-11

0

-10 -LO

-9 2 N (atoms /cm-2

Fig.2. Changes in the surface conductivity, Au, as a function tion. On the right-hand side of the figure the surface potential

5

1

of Na, K, and Cs adatom (Us) scale is given.

popula-

(3) The final Au values reached after the maximum decrease in the same sequence. From the Au values the surface potential, U,, was calculated using a method described previously [I]. The U, value for a clean surface was determined by reducing the conductivity to its theoretical minimum following oxygen adsorption. The course of U, is almost the same as the Au course but has an opposite sign. Using the right-hand side scale in fig. 2, the Au values could easily be transformed into U, values. Recently Riach and Peria [4] also determined the surface potential changes following Cs adsorption using photoelectron-energy distribution measurements. For 0 < 0.2 ML, our data almost coincide with those obtained in ref. [4]. For higher coverages, however, Riach and Peria have found that the Us remains almost constant, whereas from fig. 2 it is seen that at 0 > 0.2 ML, the band bending strongly decreases, reaching values lower than those for a clean surface. It is difficult to understand the reason for this discrepancy. Figs. 3a and 3b show the changes in Au measured at 27°C after depositing alkali metals and heating the sample for 30 set at about 150°C. It is evident that heating causes an increase in Au. This effect is most strongly expressed at the curves maxima. Simultaneously with the surface conductivity the stationary photoconductivity was measured, from which, by means of a known method [5], the surface recombination velocity, S, was determined. Figs. 4a and 4b show the S versus Us curves obtained for K and Cs. The arrows indicate the direction of increase in 0. The S values for a clean surface varied by 10-l%% with the different samples after the surface cleaning procedures, but these variations were always associated with small changes

L. Surnev, M. Tikhov /Adsorption and electric measurements

417

b 250 200

150

100

2 50 Q Lo $0

-50

10”

2

5

lo’&

2

N,(atoms/cm‘2)

5

1o13

2

5

10“

2

5

NJ atoms /cm-z)

Fig. 3. Change in the surface conductivity, Au, measured at 27°C as a function of K (a) and Cs (b) adatom population, obtained after alternatively depositing alkali metals and heating the samples at 150°C for 30 sec. The vertical dashed lines show the variation of Au obtained by heat treatment and the sloping dashed lines show the variation of Au obtained by additional dosing of K and Cs.

in the initial values of Us (see curves 1 in fig. 4). For the sake of clarity, the S values were normalized in such a way that for a clean surface at Us = 10.7 kT, So = 500 cm/set. It is evident from fig. 4 that the experimental points obtained could be fitted to four curves. Curves 1 are plotted through the points obtained for S before the minimum in the Au versus 0 curves, and curves 2 through points after this minimum. Curves 3 are fitted to the experimental points obtained at 8 higher than those corresponding to the Au versus 0 maximum. Curves 4 are fitted to the points obtained for surfaces heated at 150°C for half a minute, following alkali deposition. These four curves are similar to those obtained previously [l] for Na. The main difference between present and previous results is that for Na, curve 1 and curve 2 coincide; for K, curve 2 is slightly shifted, whereas for Cs this shift is significant. It is impossible to determine the recombination center parameters owing to the narrow Us range within which Schanges and to the fact that one cannot be sure whether the system of recombination centers remains unchanged during deposition.

L. Surnev, M. Tikhov /Adsorption

418

and electric measurements

800 -

B c

E 2 cn

600 400 200 -

, -13

-12

L

-11

-10

-9

"s

-14

-13

-12

-11

-10

-9

-8

"S

Fig. 4. Typical plots of the surface recombination velocity, S, as a function of the,+rface potential, Us. Curves 1 are fitted to the experimental points of S obtained for coverages before the minimum in the Au versus 0 curves, and curves 2 are fitted to the points after this minimu. Curves 3 are fitted to the point of S obtained after the maximum of the Au versus 0 curves and curves 4 are obtained for heat-treated alkali-metal-covered surfaces.

From curve 1 alone we obtained the value of 10.5 kT for the reduced depth, W;, of the recombination centers, using the method described before [l]. This value is very close to the one obtained previously for Na adsorption and to that obtained in ref. [6]. 3.2. Thermal desorption measurements At adatom densities up to 5 X 10” atoms/cm2, 65% of K and 75% of Cs are desorbed as ions. With increasing 8, the ionic part of the desorbed alkali gradually decreases. At K concentrations of about (2-3) X 1013 atoms/cm2 this ionic part is saturated with about (7-8) X 1012 atoms/cm 2. With Cs the saturated value (-2 X 10r3 atoms/cm2) is attained at a total overlayer concentration of more than (7-8) X 1Or3 atoms/cm’. The ratio of ions to atoms in the gas phase is given by the Saha-Langmuir equation n+/n” = A’ exp [e(q, - I/ia)/kT]

9

(2)

where A’ is the ratio of the partition functions for ions and atoms, cpeis the work function of the substrate and Via is the alkali metal ionization potential. Since, for instance, the Via value for K (3.42 eV) is lower than the qe value for Ge (4.79 eV), 100% ionization should take place. This conclusion does not contradict the experimental data as equation 2 is valid in the case of thermal equilibrium.

L. Surnev, M. Tikhov /Adsorption and electric measurements

419

In the present paper we shall report only the thermal desorption of atoms. The results on ion desorption will be the subject of another communication. Fig. 5 shows a series of TD spectra (desorption rate r versus ZJ of the K atoms at gradually increasing initial coverages up to 0.5 ML. Four desorption peaks, labeled as PI to f14, according to the sequence of their appearance, can be distinguished in the spectrum. The first three peaks do not change their position, whereas peak fi4 is shifted to the higher temperatures with increasing 19. With further increase of 8., the TD spectrum drastically changes. Fig. 6 presents a series of TD spectra recorded at 8 > 0.6 ML. It is evident that two new desorption peaks appear on both sides of peak &. The lowest-temperature peak was attributed to the second physical layer. With further increase of the K population the & and & peaks disappear and two or three new unresolvable peaks are observed instead of them (labeled 0’ and 0” in fig. 6). Fig. 7 shows a series of Cs desorption spectra obtained at 0 < 0.40 ML. Evidently four desorption peaks, labeled PI, pz, &, and fi4, could be seen in the spectrum. Fig. 8 shows the cesium spectrum for f3> 0.4 ML. As in the case of K, two new peaks, flS and &, appear on both sides of peak p4 with increasing 8. Peak & appears

750

T( K)

850

Fig. 5. Desorption spectra of K from a (1ll)Ge surface at various initial coverages (ML), e, up to 0.5ML. (a) 0 = 0.02;(b)0 = 0.08;(c)e = 0.15;(d)e = 0.19(e) 0 =0.26;(f) e = 0.35;(g) e = 0.43.

L. Surnev, hf. Tikhov /Adsorption

420

1

and electric measurements

1

550

650

6

750

Fig. 6. Desorption spectra of K from a (111)Ge surface recorded 0.6 ML: (a) 0 = 0.65; (b) .9 = 0.77; (c) 0 = 1.1; (d) 0 = 1.3 ML.

550

650

750

T( IO

850

at initial coverages

850

higher

than

950

T( IO Fig. 7. Thermal desorption spectra of Cs from a (1ll)Ge surface at various initial coverages, 0, up to 0.4 ML: (a) e = 0.02 ML; (b) e = 0.07 ML; (c) e = 0.13 ML; (d) e = 0.2 ML; (e) e = 0.3 ML; (f) 0 = 0.4 ML.

L. Surnev, M. Tikhov /Adsorption

and electric measurements

1

550

650

750

T( K)

421

850

Fig. 8. Desorption spectra of Cs from a Ge(ll1) surface recorded at initial coverages higher than 0.5 ML: (a) -9 = 0.5 ML; (b) 0 = 0.61 ML; (c) 0 = 0.69 ML; (d) 0 = 0.8 ML.

in the spectrum after the saturation of the lower-temperature peak &. With increasing 0 this peak increases and, similarly to the K p4 peak, is shifted to the higher temperatures. At highest ~9values both fls and /I3 peaks overlap. Two methods were used for the interpretation of the desorption spectra. (I) Assuming that the alkali metals are adsorbed in a binding state whose parameters change with coverage, we carried out the analysis by the method of Bauer et al. [7]. According to this method the quantity n/r is plotted as a function of the instantaneous concentration, n, for spectra obtained with different initial 0. From the family of curves obtained we plotted the Arrhenius dependences for n/r at a constant n. When the slopes of these plots are constant one may determine the values of E and v for any value of IZ. The analysis of the spectra made by this method gave no satisfactory results. We restricted the interval of the n changes to different values beginning from zero. Thus different numbers of desorption peaks were analysed. It turned out that none of the chosen values of the final surface concentration yielded a straight-line Arrhenius plot. This result shows that at f3> 0.04 ML, more than one binding states contribute to the desorption process at a given moment. (II) Assuming that the peaks are due to alkali metal desorption from different binding states, their parameters can be determined using the Polanyi-Wigner equa-

L. Surnev, M. Tikhov /Adsorption

422

ri = -dni/dt = i?rivi exp(-Ei/RT)

and electric measurements

.

(3)

Here ri = -dni/dt is the desorption rate, ni the instantaneous surface concentration, Vi the frequency factor, Ei the activation energy of clesorption, Ki the order of clesorption of the ith binding state, R the Boltzmann constant and T the absolute temperature. The analysis of the peaks according to eq. (3) for first order clesorption kinetics is difficult because of the partial overlapping of the peaks. By preliminary desorption of all lower-temperature peaks than the /3, peaks, the latter were completely resolved. The resolution of the p2 peaks was not difficult also. By appropriate preliminary desorption we can obtain traces which include only p1 and pz peaks. The population of the fll peaks is more than three times smaller than the flZ peaks population. Thus the p1 peaks affected considerably only the high temperature tails of the fiZ peaks. For this reason we evaluated the up2 and ED2parameters from the Arrhenius plots of the r/n quantity as determined from the low-temperature tail of the pZ peaks. The instantaneous concentration, n, was determined by graphic integration. Following Redhead [8], the n value at the peak maximum was assumed to be l/2.72 of the whole amount corresponding to the peak. This method can be used for the rest of the desorption peaks too, which follow first order clesorption kinetics. The existance of repulsive lateral interactions between chemisorbed species, however, may result in complex spectra at medium and high coverages [9]. Thus the multistate model becomes inapplicable. For this reason the values of E and v are presented in table 1 for /3, and p2 peaks only since the former reaches saturation and the latter appears at coverages as low as 0.03 ML. It can be seen that the E and v values for the f12peaks exceed considerably those for the p1 peaks. It follows from thermodynamical considerations that for localized adsorption v is about 1016 set-‘, while for mobile adsorption v is -10’ 3 set-‘. The K p4 and Cs p5 clesorption peaks are moving towards the higher temperatures with increasing 0. This fact, and the coverage independence of the initial clesorption rate, indicate that K and Cs are desorbed with thse peaks according to the fractional order of desorption kinetics. The K p4 peak could be relatively well resolved. This facilitates its study. Fig. 9 shows the Arrhenius plot of r/nl’* for four

Table 1 Desorption energies and frequency factors of pr and pz binding states for K and Cs

K

Pl P2

Cs

Pl P2

E (kcal/mol)

u (set-l)

3 65 * 3 58 * 3 60 f 5

(3 (5 (3 (6

5.5 f

2 f * f

2) 4) 2) 3)

x x x x

10’ 3 10’7 10’3 10”

L. Surnev, M. Tikhov /Adsorption

and electric measurements

423

E z LO k.cal /mol

Fig. 9. The Arrhenius plot of r/n1 I2 for the K p4 peaks at various initial 04 state populations: (0) 0 = 0.03; (a) tI = 0.08; (0) 0 = 0.10; (x) 0 = 0.12 ML.

different initial coverages of the /I4 K state. It can be seen that at a given temperature rfn ‘I* is coverage-independent within experimental uncertainty, i.e. r is proportional to n ‘I* . This is indicative of a : order desorption kinetics. The slopes of the plots in fig. 9 are almost the same, showing an activation energy of 40 kcal/ mol, regardless of coverage. By analogous procedure we evaluate that the EpS value for Cs is about 50 kcal/ mol. As can be seen from fig. 8, at highest coverages the OS and pa peaks overlap and are difficult to resolve. For his reason the value obtained for Eps should be taken with caution. It is almost impossible to analyse the peaks which appear in the higher-temperature region of the K spectra at 0 > 0.8 ML, due to their overlapping. Using the desorption parameters of the different peaks one may choose their population to produce the best fit to the desorption spectra of Cs and K (for 0 < 0.8 ML). From the areas under the peaks one can determine the corresponding coverages 8,. Table 2 shows the Bi maximum values obtained. It is seen that each newly formed peak has a higher population than the preceding one. The amount of K desorbed with the new two or three highest temperature peaks (curve d in fig. 6) was found to be 0.40 ML. This amount was determined after preliminary desorption of all lower-temperature peaks. It is interesting to note that this amount is almost equal to the coverage at which Weber and Johnson [3] have observed a

424

L. Surnev, M, Tikhov /Adsorption

Table 2 Coverages (in ML) corresponding

to the K and Cs desorption peaks

Element K CS

and electric measurements

005

9p3

0.03 0.03

0.1 0.09

(1 X 6) LEED structure on .a (111) from a K-saturated surface.

0.2 0.12

0.4 0.2

0.2

Ge surface after partial thermal

desorption

3.3. Work function measurements The main purpose of these measurements was to find a correlation between the work function (WF) changes during adsorption and the behaviour of the TD peaks. For the sake of comparison with Ao and S data obtained for unheated and heated surfaces with an alkali metal overlayer, WF measurements were performed under the same conditions. Fig. 10 shows the WF changes, Aq, as a function of the K and Cs overlayer concentrations. For the sake of clarity, the Cs curves are translated along the ordinate by 1 eV. Above each curve, the curves obtained for heat-treated surfaces are plotted. It is evident that, similarly to Weber and Peria’s [lo] results, the experimental points can be fitted to several striaght-line segments. Since our results differ from those in ref. [lo] only for low and medium coverages, in fig. 10 the WF changes are shown in a more stretched 8 scale. The first difference is that for unheated surfaces we obtained two segments instead of the first one in ref. [lo]. Hence the first break points of our Aq versus 0 curves is already to 0.04 ML and not at 0.12-0.125 ML as in ref. [lo]. The second difference concerns only cesium. There is no straight-line segment in the Aq versus 0 curve [lo] corresponding to the fourth segment of the analogous curve for Cs in fig. 10. The experimental points which correspond to the fourth segment of the latter curve, however, are in agreement with those obtained by Weber and Peria [IO]. In fact some of their experimental points deviate from the plotted straight-line segments. Thus the experimental points at 8 = 0.35 ML is slightly deviated from the segments whereas with the points at 0 = 0.40 ML this deviation is considerable. If we connect the latter point and that corresponding to 0 = 0.25 ML with a straight line we shall obtain one more segment which will correspond to our fourth segment from the Cs curve in fig. 10 and will provide a much better fit to the experimental points. Table 3 shows the Bi value between the break points of the Aq versus B curves for K and Cs. A comparison with table 2 shows that the coverages for each straightline segment are in complete agreement with the alkali amount desorbed with the corresponding peak, i.e. to each TD peak corresponds a segment of the Aq versus 8

L. Surnev, M. Tikhov /Adsorption and electric measurements

425

&(E

Y aI

‘;: 0 3(4 t 3 IL Y f5 3 2(!

!)-_

(2

0

N

3.10“

2'10“

atoms/cm2

Fig. 10. Work function of Ge(ll1) against K and Cs coverages. The filled circles denote the unheated surfaces and the open circles correspond to the heated surfaces. The Cs curves are translated along the ordinate by 1 eV with respect to the K curves. In the right-hand side of the figure the three types of the binding sites are shown according to ref. [lo].

curves. The curves obtained for heated surfaces show that heating causes an increase in WF only at coverages of up to 0.3 ML. One should distinguish between this increase and the increase in WF by about 1 eV as a result of heating Na [2] and K [3] saturated Ge( 111) surfaces. Fig. 10 shows that the first segments in the Aq versus 8 curves for the heated surfaces have almost the same slope as the second segments for unheated surfaces. We shall calculate only the initial values of the dipole moment (in Debye units)

Table 3 Coverages (in ML) corresponding to different straight-line segments (between the break points) of the Aq versus 0 curves of K and Cs Element

*1

e2

83

94

05

K cs

0.04 0.04

0.105 0.08

0.22 0.125

0.38 0.21

0.18

L. Surnev, M. Tikhov /Adsorption

426

for heated and unheated p = A&n

and electric measurements

surfaces using the formula of Helmholtz:

X 300 An ,

(4)

where Ap is the work function change in eV, and An is the change in the surface concentration in atoms/cm’. The first column in table 4 contains the dipole moments of K and Cs as calculated from the slopes of the first segments in the Acp versus 0 curve. We assumed that on the (11 l)Ge surface there are three types of adsorption sites [IO] denoted as A, B, and C in fig. 10. For each of them the fraction of transferred charge, F, from the adsorbate to the substrate can be calculated. Columns 2, 3, and 4 in table 4 show the calculated F values for heated and unheated surfaces. The fifth column presents the Pauling ionic character (PIC) [lo] or the alkali-germanium bond. It is evident that the F values calculated for unheated surfaces with the three types of adsorption sites considerably exceed PIC. The values of F calculated from the initial dipole moments for heated surfaces are very close to PIC, the best agreement being found with adsorption at the B sites.

3.4. Comparison of the experimental data with the Gyftopoulos-Levine

theory

As was shown in the preceding sections, adsorption of alkali metals on the covalent Ge surface is accompanied by complex and unexpected changes in the kinetic parameters (E, v, and K). These changes cannot be predicted on the basis of the phenomenological theory of Gyftopoulos-Levine [ 111. For this reason we calculated only the initial values of the alkali adsorption energies for heated and unheated surfaces using the F values obtained in the previous section. We used the equation given in ref. [ 1 l] :

E = F~e[l + F(e’/d-

Via)/PeI

+ 2 [@sub)Ge@-sub)AM(1

where

-

F2h2)1

1’2/(SAM/SGe

+ &h/SAM)

>

(5)

is the valence charge of the alkali metals, e is the unit electron charge, and and the angular (&ut,)c& @sub)AM, SGe> and SAM are the heats of sublimation K

Table 4 Fraction of the transferred charge F calculated from the initial slopes of the & versus 0 curves for the three different binding sites A, B, and C shown in fig. 10 (upper right-hand side) Element

p (Debye)

FA

FB

Fc

K (unheated) K (heated) Cs (unheated) Cs (heated)

5.3 3.32 6.63 4.24

0.49 0.30 0.51 0.32

0.43 0.21 0.47 0.30

0.34 0.21 0.39 0.25

Pauling ionic character 0.26 0.30

421

L. Surnev, M. Tikhov /Adsorption and electric measurements Table 5 Binding energies (in kcal/mol) calculated from the Gyftopoulos-Levine values from table 4

formula (5) using the F

Element

EA

EB

EC

Ec(1.3)

EB(PIC)

K cs

96.5 93.6

86.5 86.9

12.2 14.1

62.6 64.0

66.0 68.1

strengths of the valencd orbitals of the substrate and the adsorbate, respectively. The thickness of the adlayer, d, is different for the three kinds of adsorption sites [lo]. The following data (in eV) were used for the calculation: qe= 4.78, viK= 4.32, I’tcs = 3.87, (Esub)Ge = 3.98, (EsurJK = 0.92, (EsutJCs = 0.80. The covalent radii of Ge, K, and Cs were 1.22,2.02, and 2.35 A, respectively, and SK = ScS = K = l,SGe= 1.87. Columns 1, 2 and 3 in table 5 show the adsorption energies aetermined using formula (5) with F values from columns 2, 3, and 4 of table 4 for unheated samples. It is evident that the E values calculated for the three types of adsorption sites (A, B, and C) considerably exceed the values found from TD data for the /3r states (table 1). When the calculations were performed with F equal to PJC and a d value which is 1.3 times as high as the sum of the covalent radii, we obtained the value given in column 4 of table 5. These values are closer to the experimentally obtained ones for the /3r binding states. Column 5 shows the values of E calculated for adsorption of alkali metals at the B sites when F is equal to PIC. It is evident that these values are in a good agreement with those obtained by TD for the K and Cs pa peaks.

4. Discussion 4.1. Thermal desoolption and work function measurements The presense of several straight-line segments each corresponding to a desorption peak, seems to be in accordance with the viewpoint of Weber and Peria [lo] that alkali metals are adsorbed at various sites on the Ge surface, a discrete dipole moment corresponding to each of them. On the other hand the Cs thermal desorption spectra are more complex in comparison with the K ones, although the Cs atoms are larger in size. It can be assumed that the multiple-peak desorption spectra may be due in part to the existence of direct or indirect [ 121 lateral interactions between adsorbed species [ 131. The complete interpretation of the desorption traces is difficult because no structural data are available. Indeed, no characteristic diffraction patterns were observed at the

428

L. Surnev, M. Tikhov /Adsorption

and electric measurements

break points in the Acp versus 0 curves [lo], i.e. when the new desorption peaks appeared. It is known [9,13] that the repulsive lateral interactions affect the desorption traces at moderate and high coverages. The appearance of the flz K and Cs peaks at 0 - 0.05 ML, however, cannot be explained by the lateral interaction model. The higher E and v values found for /Zz peaks compared to the f3r peaks (see table 1) cannot be understood on the basis of thermodynamic equilibrium considerations. Thus we assumed that the flz peaks are formed during adsorption, most probably as a result of surface covalent bond redistribution. As it can be seen in fig. 10 the appearance of Pz peaks is accompanied by the appearance of the second segments in the AV versus 0 curves. The smaller slope of these segments suggests that the newly formed binding states are characterized by lower values of the dipole moments, most probably due to a decrease in the dipole lenght. Assuming that the structure of the Ge(ll1) surface, covered with such small alkali metal quantities, is metastable and tends to reconstruct, it must be separated by a free energy barrier from the more stable structure. Hence, heating should accelerate this process. Indeed, fig. 10 shows that after heat treatment, the WF changes and the slope of the first segments in the Ap versus 0 curves decreases. The availability of this process is additionally supported by the fact that the first segments in the Acpversus 0 curves for the heated surfaces have almost the same slope as the second segments for unheated surfaces. Parallel structure measurements have to show the change in LEED patterns after heating. Indeed Goldstein [ 141 has found that similar heating of GaAs(ll1) surface with Cs coverage as low as 0.05 ML causes the disappearance of the : order spots in the (2 X 2) LEED patterns observed for clean surfaces. Unfortunately no analogical measurements have been available for Ge( 111) surfaces yet. Because of the above mentioned difficulties we will not consider all desorption peaks. Further we shall discuss only K /I4 and Cs /3s peaks which follow half an order desorption kinetics. This kinetics can be explained assuming that a twodimensional patch structure arises as a result of the lateral attractive interactions. It is interesting to note that a desorption peak with similar behaviour was observed with sodium as well [l]. The surface concentration at which these twodimensional patch structures appear increases in the sequence from sodium to cesiurn. Thus peak flz for Na appears at 0 = 0.15 ML, peak f14 for K is observed at 0 = 0.25 ML, and peak /Is for Cs at 0 = 0.45 ML. This behaviour is apprehensible when one takes into account that the repulsive forces between the posotively charged species hindering the appearance of two-dimensional patches increase in the sequence from Na to Cs. In the case of Cs, a higher concentration is needed for the attractive forces to overcome the action of the repulsive ones. Weber and Peria [IO], however, have shown that if two-dimensional patches appear on the Ge surface in some alkali metal coverage region, they are not characteristic of a nucleation type of overlayer formation. But from the available data it is impossible to draw any conclusion about the alkali overlayer structure.

L. Surnev, M. Tikhov /Adsorption and electric measurements

429

4.1.2. Heat-treated potassium-covered surface An interesting fact observed in the present work is the appearance of the new TD peaks in the high-temperature part of the K spectrum at 0 - 1 ML (curve d in fig. 6). Undoubtedly, this behaviour of the TD spectra can be associated with the husteresis in the Acp versus 0 curves obtained during deposition and during stepwise desorption of K from the Ge(ll1) surface [3]. It means that at the same e value the WF changes are smaller for heat-treated surfaces. At 8 - 0.5 ML the difference may reach the value of 1 eV. Weber and Johnson [3] established that after gradual desorption of K, when 0 reaches 0.41 ML, the (1 X 6) LEED pattern appears, which corresponds to a straight-line segment in the Aq versus 8 curve. We have found that about the same amount corrrsponds to the two or three highest-temperature peaks appearing in the K spectrum at t’3- 1 ML (curve d in fig. 6). Obviously these hightemperature peaks and the hysteresis in the Aq versus 0 curves are due to the surface reconstruction caused by the adsorbed K and probably stimulated by the heattreatment. As in the preceding section, we assume that the reconstruction results in the formation of new adsorption sites for which the K-Ge bond is stronger and therefore shorter. This explains the appearance of the new high-temperature peaks and the decrease in strength of the surface dipoles of which K is a part. To overcome the potential barrier of the surface reconstruction, the K coverage and the temperature have to rise above certain critical values. Fig. 6 shows that the critical value of 8 is about 1 ML. However, it was impossible to determine the critical temperature on the basis of our measurements. It should be pointed out that this reconstruction is irreversible even after the removal of half of the critical K coverage. Similar reconstruction is not achieved with Cs adsorption. Indeed, no new high-temperature peaks analogous to those for K appear in the Cs spectrum. In the literature there are also no data concerning the existence of hysteresis phenomena of the Aq versus 8 curves with Cs adsorption. The most probable reason for this behaviour is the relatively low surface concentration which can be reached by Cs at moderate temperatures. 4.2. Surface conductivity and surface recombination velocity Recently Rowe et al. [ 151 measured the electron energy loss spectra (ELS) of Si, Ge and GaAs surfaces with indium and gallium adsorption and found that partial metal overlayers remove the intrinsic surface states and replace them with extrinsic states localized near the metal adatoms. On the other hand, Housley et al. [16] established, in the Auger spectrum of a Si(ll1) surface covered with a silver monolayer, a new peak which, according to their opinion, had arisen from an Auger process involving an induced Ag-Si interface state. Although, to our knowledge, there are no ELS measurements for the alkali metal adsorption on the Ge surface, we believe that such processes take place in this system. Thus the changes in surface conductivity, Au, and the surface recombination velocity, S, with increasing alkali coverage can be ascribed to the competitive action of the removal of intrinsic sur-

430

L. Surnev, M. Tikhov f Adsorption

and electric measurements

face states of the clean Ge(ll1) surface and their replacement by alkali-induced extrinsic states. Hence, the initial decrease in Au can be associated with the alkaliinduced surface states as this process provokes an increase in the positive surface charge. In fact the depth of the initial Au minimum increases in the same sequence in which the ionization potential, Via, of the alkali decreases, i.e. from Na to Cs. The increase in Au is associated with an increase in band bending which means that the negative charge on the surface increases. Evidently, owing to the much lower Vi, values of the alkali compared to the WF of the Ge surface, this fact cannot be explained with the alkali-metal-induced surface states and we shall attribute it to the removal of the intrinsic states of the clean Ge(l1 I) surface. It is interesting that the increasing part of the Au versus 0 curves begins at coverages at which the second segments in the Au versus 0 curves (fig. 10) and the p2 binding states appear. Since we attributed the appearance of these adsorption phases to surface reconstruction, it follows that the beginning of reconstruction coincide with the beginning of the process of disappearance of the intrinsic states. If this is true, the increasing part of the Au versus 0 curves could be a measure of surface reconstruction in the process of alkali adsorption on Ge. From fig. 2 it can be seen that these parts of the Au versus 0 curves have slopes increasing in the sequence from Na to Cs. Therefore, with the same 0, the surface reconstruction degree is inversely proportional to the Vi, values of the adsorbed alkali. If cesium is most effective in removing the intrinsic surface states and probably in changing the system of recombination centers this explains why the hysteresis effect of the S versus U, curves in fig. 4 is most pronounced in this case. As heating facilitates surface reconstruction, according to the above model one can explain why Au increases after heating, whereas S changes according to another curve (curve 4 in fig. 4). The decrease in Au after the maximum will be ascribed to extrinsic surface states induced by alkali metals. Indeed alkali metal with lower Vi, values results in lower final Au values. 5. Summary It has been found that a thermal desorption peak corresponds to each straightline segment of a Aq versus 8 curve. The beginning of the increasing part of the Au versus 8 curves coincides with the appearance of new adsorption phases and new straight-line segments of the Acpversus 19curves. Hence, these phenomena are due to alkali-metal-induced redistribution of the covalent surface bonds accompanied by disappearance of the intrinsic surface states. At a given 0 value the surface reconstruction degree increases in the sequence from Na to Cs. At a certain critical surface overlayer concentration increasing from Na to Cs, peaks corresponding to f kinetic order appear in the desorption spectrum. This is explained with the appearance of two-dimensional patches on the surface. At a potassium concentration close to 1 ML, new, more tightly bound adsorption phases appear, probably due to additional irreversible surface reconstruction.

L. Surnev, M. Tikhov /Adsorption

and electric measurements

References [l] [2] [3] [4] [S] [6] [7] [8] [9]

L.N. Surnev, Surface Sci. 55 (1976) 625. P.W. Palmberg and W.T. Peria, Surface Sci. 6 (1967) 57.. R.E. Weber and A.L. Johnson, J. Appl. Phys. 40 (1969) 314. G.R. Riach and W.T. Peria, Surface Sci. 40 (1973) 479. A.V. Rzhanov, Fiz. Tverd. Tela 2 (1960) 2431. S. Wang and G. WaIlis, Phys. Rev. 105 (1957), 1459; 107 (1959) 947. E. Bauer, F.Bonczek, H. Poppa and G. Todd, Surface Sci. 53 (1975) 87. P.A. Redhead, Vacuum 12 (1962) 203. D. Adams, Surface Sci. 42 (1974) 12. [lo] R.E. Weber and W.T. Peria, Surface Sci. 14 (1969) 13. [ll] J.D. Levine and E.P. Gyftopoulos, Surface Sci. 1 (1964) 171. [ 121 T.B. Grimley, Proc. Phys. Sot. (London) 90 (1967) 751. [ 131 D.A. King, Surface Sci. 47 (1975) 384. [ 141 B. Goldstein, Surface Sci. 47 (1975) 143. [ 151 I.E. Rowe, S.B. Christman and G. Margaritondo, Phys. Rev. Letters 35 (1975) 1471. [ 161 M. Housley, R. Heckinbottom and C.J. Todd, Surface Sci. 68 (1977) 179.

431