RAYMOND F. BADDOUR AND CHARLES W. SELVIDGE
2536
Catalytic and Chemical Properties of Clean Germanium Surfaces
by Raymond F. Baddour and Charles W. Selvidge Department of Chemical Engineering, Masaachusetta Institute of Technology, Cambridge, Massachusetts OBI 69 (Received December 27, 1966)
The interaction between clean germanium powders (prepared by crushing under ultrahigh vacuum) and ethanol was studied in a high-vacuum catalytic reactor. Germanium powders catalyzed the steady-state dehydrogenation of ethanol, and the reaction rate was determined over the range 120-325" and 0.3-5 X torr. The catalytic activity decreased as a function of the maximum temperature to which the samples were heated in the presence of ethanol; e.g., the specific rate constant was decreased by more than 100 when a sample was heated from 120 to 325". Clean germanium surfaces extracted oxygen from ethanol producing ethylene, hydrogen, and a nonreducible surface oxide. The extent of oxide formation was also a function of the maximum temperature to which the surfaces were heated, and the decrease in catalytic activity is associated with oxygen poisoning of catalytic sites. The dehydrogenation activity of the germanium surfaces was independent of bulk doping over the entire range investigated (lozo n type, intrinsic, 1020 ~ m p- type). ~ The dehydrogenation activity of the germanium surfaces was attributed to the unsaturated bonding in a clean surface.
Introduction Elemental semiconductors are well suited for studying electronic factors in catalysis because their bulk electrical properties can be accurately varied over wide ranges with minute amounts of impurity dopants. Furthermore, the fundamental understanding of elemental semiconductor surfaces is highly advanced because they have been investigated in such great detail in connection with their use in semiconductor devices. Several investigations have been reported in which the relationship between catalytic activity and bulk electrical properties of germanium was studied. Contradictory results have frequently been obtained without apparent explanation and no fundamental interpretation has been possible. The probable reason for the ambiguities in these results is the presence of uncontrolled surface contamination. The present investigation was undertaken to study the catalytic and chemical processes occurring a t a clean germanium surface prepared by cleavage under ultrahigh vacuum. Particular emphasis has been given to determining the effect of changes in bulk doping upon catalytic properties. The ethanol decomposition reaction was chosen for study because The Journal of Physical Chemiatry
germanium catalyzes this reaction, and the activity has been reported to be a function of bulk doping.'
Experimental Section A high-vacuum reactor system (Figure 1) was developed for these studies. Purified ethanol vapor is metered into the high-vacuum reactor section through a variable leak. I n the reactor section, ethanol passes through a bed of the semiconductor sample which is supported on a fritted Pyrex disk. The gases resulting from the interaction of ethanol with germanium surfaces flow through a mass spectrometer and are pumped from the system by a 75-l./sec Varian Vac Ion pump. Product Analysis. A Machlett radiofrequency mass spectrometer described by Merril12was used to obtain mass spectra of the product mixtures flowing from the reactor. Two different techniques were employed to determine the composition of the product stream from these mass spectra. One method employs the absolute magnitude of the (1) V. M. Frolov, 0. V. Krylov, and S. Z. Roginskii, Dokl. Akad. Nauk S S S R , 126, 107 (1959). (2) R. P. Merrill, Sc.D. Thesis, Department of Chemical Engineering, Massachusetts Institute of Technology, 1964.
2537
CATALYTIC AND CHEMICAL PROPERTIES OF CLEAN GERMANIUM SURFACES
GAS HANDLING
UHV-CV-
Figure 1. Schematic diagram of the high-vacuum catalytic reactor.
peaks corresponding to difierent masses to determine compositions. The most important peak is that for mass 31 (CHzOH+), which is the principal fragment in ethanol ionization for the 104-v ionizing potential employed. For steady-state conditions, the ratio of the mass-31 peak with the reactor a t some temperature (T) to the 31 peak for ethanol with the reactor a t room temperature (RT), where no reaction occurs, gives the fraction of ethanol unreacted, Y.
This technique can also be used to detect compounds other than ethanol which are in the product spectra.* The other procedure relies on ratios of peak heights within a given spectrum to determine compositions. The mass-31 peak height was directly proportional to the ethanol partial pressure (numbers in parentheses refer to peak heights)
PE =
(YE
(11)
(31)
The mass-29 peak (CHO+) was the principal acetaldehyde peak and the acetaldehyde partial pressure was related to the 29 peak height by the relation
where (29/31)~is the ratio of the 29 peak to the 31 peak for pure ethanol. This ratio was a weak function of operating conditions and for a given run the value was constant within *0.01. When the only reaction products were acetaldehyde and hydrogen
"[(g) - (E)J
5 = (YE P E
where Y
=
PE/(PE
+ PA).
=
(1/Y) - 1 (IV)
The value for
((YA)/
((YE) was determined by calibration with mixtures of known composition to be nearly unity. The determination of ethylene concentrations was more difficult because the mass-28 peak (CzHb+) which is the principal fragment in ethylene ionization was a nonlinear function of the ethanol and acetaldehyde concentrations. Consequently, a detailed calibration curve for the ratios of peaks (25/31) and (28/29) us. ethanol and acetaldehyde concentrations was required. Use of the mass-27 peak (C2H3+) to determine ethylene concentrations gave results equivalent to those obtained from mass 28. Flow rates into the reactor could be determined accurately by measuring the rate of decay of the pressure in the gas-handling manifold at a constant variable leak setting. The flow rate out of the system could be estimated because the current reading on the pressure gauge of the Varian Vac Ion pump is proportional to the flow rate into the pump. Using the relation F = K,P, where F is the flow rate and P , is the Vac Ion pump current in milliamperes, it was possible to determine K , by calibration. The measured value of K , was 5.5 X 10-lO mole/sec ma, and this was reproducible to within 10% from run to run. Within a given run the value of K , remained constant to better than 5%. There was no significant drift in the value of K , over several months of pump operation. Mixtures of ethanol, acetaldehyde, and hydrogen in the proportions which were developed in the dehydrogenation reaction gave the same Vac Ion pump current for a constant flow rate of ethanol into the system independently of the degree 0' conversion. Mixtures containing ethylene were found to exhibit values for K , somewhat different from pure ethanol. The proper values for K , of mixtures with ethylene were found by calibration. Semiconductor Samples. High-purity single crystal specimens of germanium were used in the preparation of the samples investigated. These were supplied by Bell Telephone Laboratories, Murray Hill, N. J., Lincoln Labs, M.I.T., Lexington, IIass., and Semimetals, Inc., West Bury, N. Y. The clean germanium surfaces studied were prepared by crushing large chips (4 mm3) into fine powders under high vacuum (Figure 2 ) . Details of the crusher construction are reported elsewhere.* The entire crushing system was baked out at -300" and 10 at 200"). Because of the high-temperature activation procedure (outtorr) gassing powder crushed in air a t 650" and in the previous work, the doping effect reported is believed to arise from chemical changes in the surface caused by diffusion of dopant to the surface layer. The diffusion of n-type dopants becomes significant a t temperatures > 600°.6 Dillon and Farnsworth observed significant diffusion of antimony to the surface while annealing germanium samples a t 675" prior to work function studies.' The probable occurrence of chemical changes in the surface can account for the results in other studies in which correlations between activity and doping have been reported for germanium. In particular for the H2-D2 exchange reaction, studies on samples activated by heating a t -675" indicate large poisoning effects resulting from n-type dopants,8 whereas vacuumcrushed samples possessed activity which was independent of dopingeg Table II'O summarizes results showing the doping effect which has been found in studies on samples heated to high temperatures and the absence of a doping effect in studies using vacuumcleaved surfaces. Several recent investigations have been reported on the catalytic activity of doped germanium, in which
Table 111: Recent Studies in Which the Catalytic Properties of Germanium Have Been Found Independent of Bulk Doping Ref
Doping extremes
Reaction
11
3 X 1OI8 cm-3 As 3 X 1018 em+ Ga
H
12
1017 cm-s -n 1017 cm-3 -p
Hz '/zOz HsO (200-400")
13
1018 cm-3 Sb 1018 cm-3 Ga
Isopropyl alcohol vacuumcleaved Ge + H2 (few hundredths of monolayer) (85')
14
10l6cm-3 Ga 1018 cm-a Ga
Formic acid decomposition (325-375")
+
+
H - S + HZ S (Hydrogen atom recombination a t germanium surface) R.T.-100"
+
+
+
(6) W.C.Dunlap, Jr., Phya. Rev., 94, 1531 (1954). (7) J. A. Dillon and H. E. Farnsworth, J . A p p l . Phya., 2 8 , 174 (1957). (8)G.E. Moore, H. A. Smith, and E. H. Taylor, J . Phys. Chem., 66, 1241 (1962). (9) V. L. Kuchaev and G. K. Boreskov, Kinetika i Kataliz, 1, 356 (1960). (10) V. M.Frolov, ibid., 6, 149 (1965).
Volume 71 Number 8
J u l y 1967
2542
RAYMOND F. BADDOUR A N D CHARLES W. SELVIDGE
t
I
I
I
I
I
1I
0.3
0.6
0.9
1.20
1.50
1.80
%$!4
MOLECULES ETHYLENE EVOLVED TOTAL SURFACE SITES
Figure 6. Ethanol dehydrogenation activity of sample 6-In-Ge relative to activity of Level I vs. total quant'ity of ethylene produced.
activity was independent of bulk doping. The reactions and doping levels investigated are summarized in Table III.11-13 In a recent study by Bracken,'* the catalytic activity of germanium as a formic acid decomposition catalyst was independent of bulk doping but was greatly increased by high levels of surface doping (-5 X 1020 cm-a) with gallium and boron. It is probable that a t such high doping levels chemical alteration of the surface is significant and that the enhanced catalytic activity cannot be attributed directly t o an increase in the surface p-type carrier concentration. The appearance of ethylene in the product stream is attributed to the surface chemical reaction CH&H20H or CH3CH0
+ Ge(S) +
+ H2 + GeO(S)
(3) which corresponds to the irreversible oxygenation of the atomically clean surface. The occurrence of this reaction is consistent with other literature in which oxygen-containing compounds were reported to oxidize a clean germanium surface.' *15 The temperature dependence of the ethylene evolution indicates that the surface possesses sites with varying activities toward bonding with oxygen. The produced at the lowest temperatures corresponds to the formation of irreversibly bonded C2H4
The Journal of Physical Chemistry
oxygen at thc most reactive sites and the evolution at higher temperature indicates a decreasing affinity of the surface for oxygen. The most probable explanation for the decrease in catalytic activity resulting from increases in the maximum sample temperature is the annihilation of catalytic sites by the formation of a surface oxide. An alternative explanation is that the occurrence of a temperature-dependent structural modification of the surface reduces the catalytic activity. Although there is considerable evidence showing that clean germanium surfaces produced a t low temperatures by vacuum cleavage undergo extensive rearrangement when the temperature is increased above 450"K,16 structural modification is not believed to be the controlling factor in the present study because of the results on sample 4-p-Ge. These results were obtained on surfaces which were regenerated after exposure to air by heating at high vacuum torr) and 540' for 20 hr. Such a surface treatment produces a clean annealed surface4 in which structural changes resulting from surface migration of germanium atoms would be virtually complete. The fact that the results on this sample are correlated by the same maximum sample temperature dependence as the other samples shows that the catalytic activity is determined by the conditions of maximum temperature in the presence of reactant and not maximum temperature alone. This indicates that the degree of surface oxygenation exerts a stronger influence on catalytic activity than does surface structure. Analogous results have been observed for adsorption of oxygen on clean germanium surfaces for which surfaces cleaned by high-temperature outgassing possess oxygen adsorption kinetics similar to those for vacuum cleaved ~urfaces.~' The data illustrate the complexity of the interaction between clean germanium surfaces and ethanol. However, the important features of the catalytic results can be interpreted by the reaction sequence CHsCH2OH
+ germanium surface --+ CH&H20H(ads)
(4)
(11) K. M. Sancier, S. R. Morrison, and H. U. D. Wiesendanger, Cata2ysis* 5 * 361 (1966)* (12) G. Ertl, Z . Physik. Chem. (Frankfurt), 46, 49 (1965). (13) V. M. Frolov, E. K. Radzhabli, and S. 2. Roginskii, Kinetika i KaiaZiz, 6 , 504 (1965). (14) R. c. Bracken, J . catalysis, 6, 57 (1966). (15) M. J. Sparnaay, Ann. N . Y . Acad. Sci., 101, 973 (1963). (16) J. J. Lander, G. W. Gobeli, and J. Morrison, J . Appl. Phys., 34, 2298 (1963). (17) M. Green and A . Liberman, J . Phys. Chem. Solids, 23, 1407 (1962).
CATALYTIC AND CHEMICAL PROPERTIES OF CLEAN GERMANIUM SURFACES
CH3CHzOH(ads)
2543
+ Ge(S) +
+ CH3CHz0. + Ge(S) +CH3CH0 + Ge(H) Ge(H) + GeH(S) HZ + Ge(S)
CH~CHZO. GeH(S) (5)
(6)
(7)
Equation 4 corresponds to the adsorption of ethanol on the germanium surface. Equation 5 corresponds to the formation of a covalent bond between hydrogen and a surface germanium atom and an adsorbed radical fragment CH3-CHz0.. The ability of germanium to form such a bond results from the unsaturated bonding of surface germanium atoms. Equation 6 corresponds to the stabilization of CHa-CHZO by elimination of hydrogen and formation of stable acetaldehyde. The hydrogen atom is assumed to be accommodated in some manner by the surface as represented by Ge(H). Equation 7 corresponds to the formation and desorption of hydrogen molecules. Equation 5 is energetically the most difficult in this scheme and is considered the rate-limiting step. The rate expression, eq 1,is consistent with the surface reaction step, eq 5, being the rate-limiting step. A simplified model for the atomic process represented by eq 5 is shown in Figure 7. Attributing the catalytic activity of a germanium surface to its ability to bond covalently with hydrogen is consistent with a number of experimental observations. 1. Covalent bond formation with hydrogen is well established in the chemistry of germanium, in particular tetrahedral GeH4. 2. Germanium surfaces chemisorb hydrogen molecules d i s s o ~ i a t ~ i v e land y ~ ~strongly ~~ bond to atomic hydrogen.19 3. Germanium catalyzes many reactions in which bonds involving hydrogen must be broken (e.g., H2-D2 exchange, formic acid decomposition, ethylene hydrogenation, and ethanol dehydrogenation). 4. The catalytic activity of germanium surfaces is independent of bulk doping for reactions involving hydrogen unless there is the possibility of changes in the surface chemistry resulting from migration of bulk dopant or extremely high doping (-1 atomic yo). This suggests a mechanism involving localized chemical bonding. 5. Oxygen poisons the activity of germanium surfaces. Oxygen is believed t o bond covalently with clean germanium surfaces and saturate the dangling surface bonds.’’ 6. Sites which are most easily oxidized by ethanol possess highest activity toward ethanol dehydrogenation.
(0)Adsorption
of Ethonol
(b) Dehydrogenation of Ethanol
Figure 7. Simplified atomic model for adsorption and catalytic processes on (111) germanium surface.
It should be noted that the independence of catalytic activity on bulk doping does not prove that the catalytic activity is independent of the work function of germanium. The work function of clean germanium surfaces has been shown to be a weak function of bulk doping in several investigations.20 This has been attributed to the existence of a high density of surface states which control the surface electrical properties. The occurrence of these surface states has been associated with the presence of unsaturated bonding of surface germanium atoms. These unsaturated bonds have been considered to be responsible for the dehydrogenation activity of germanium.
Conclusions 1. The catalytic activity of vacuum-crushed germanium powders for the dehydrogenation of ethanol is independent of bulk doping for the samples studied (1020,l O I 9 ~ m n-type, - ~ intrinsic, 1020,l O I 9 ern+ ptype) * 2. Vacuum-cleaved germanium surfaces abstract oxygen from ethanol by the surface chemical reaction
+ Ge(S) +CZH4 + HZ + GeO(S)
CH3CHzOH
3. The catalytic activity of vacuum-crushed germanium powders is governed by the maximum sample temperature because this determines the degree of surface oxygenation. 4. The data are consistent with a model which attributes the catalytic activity of cleaved germanium powders to the ability of an unsaturated surface germanium bond to form a covalent bond with hydrogen. Acknowledgments. This work was supported in part by the Office of Naval Research. C. W. S. held ~~~~~
~~
(18) K.Iamaru, J . Phys. Chem., 61, 647 (1957). (19) K. H. Maxwell and J. Green, J. Phys. Chem. Solids, 14, 94 (1960). (20) G.W.Gobeli and F. B. .411en, Surface Sei., 2, 402 (1964).
Volume 7 1 , Number 8
July 1967
WERNERLESSLAUER AND PETER LAUGER
2544
a National Science Foundation graduate fellowship. Dr. Robert P. Merrill, now Assistant Professor of Chemical Engineering, University of California,
Berkeley, developed the original equipment design and took preliminary data with the ethanol-germaniun system.
Theory of Snakecage Polyelectrolytes
by Werner Lesslauer and Peter Lauger Institute of Physical Chemistry, University of Basel, Basel, Switzerland Accepted and Transmitted by The Faraday Society
(August $0,1966)
Snake-cage polyelectrolytes are modified ion exchangers which consist of a cross-linked polymer network with fixed charges containing trapped linear polyelectrolyte of the opposite charge. It is shown that their properties can be discussed by a relatively simple model. It consists of a pair of parallel plates which are positively and negatively charged. By straightforward mathematical analysis, explicit expressions for the ionic concentrations inside the exchanger are derived. The results are obtained in terms of the Jacobian elliptic functions and are graphically evaluated. The calculated salt absorption of the resin agrees with the experimental values. Furthermore, the electrical and mechanical forces acting on the plates are calculated. In this way, the unusual swelling behavior of snakecage polyelectrolytes can be explained.
1. Introduction In 1957, Hatch, Dillon, and Smith introduced a modified type of ion-exchange resins which they called snake-cage polyelectrolytes. They consist of a crosslinked polymer network carrying fixed charges of one kind and contain physically trapped linear polyelectrolyte molecules of the opposite charge, the crosslinked matrix of the cationic resin constituting the “cage,” in which the linear anionic polymer “snake” is trapped. Snake-cage polyelectrolytes exhibit a number of interesting and useful properties. Two of them are of interest here.2 They reversibly absorb both anions and cations simultaneously from an aqueous solution of low molecular weight electrolytes. The absorbed electrolytes can be removed by merely washing with water. Snake-cage resins swell in concentrated and shrink in dilute salt solutions, whereas the swelling degree of normal ion exchangers decreases with increasing The Journal of Physical Chemistry
salt concentrabion of the surrounding medium. Thus snake-cage polyelectrolytes strongly differ from mixedbed ion exchangers, which are merely a macroscopic mixture of cation- and anion-exchange resin particles. The properties of snakecage polyelectrolytes can be discussed, as will be shown, by means of a relatively simple model. Their molecular structure will be described by a pair of parallel plates which are positively and negatively charged. A similar model with uniform sign of plate charge has already been used for the description of normal ion exchangers3t4 and in the theory of lyophobic colloid^.^ (1) M. J . Hatch, J. A. Dillon, and H . B. Smith, I d . Eng. Chem., 49, 1812 (1957). (2) C . Rollins, L. Jensen, and A. N. Schwartz, Anal. Chem., 34, 711 (1962). (3) L. Lazare, B. R . Sundheim, and H. P. Gregor, J . Phys. Chem., 6 0 , 641 (1956). (4) W. Lesslauer and P. Lauger, 2. Physik. Chem. (Frankfurt), in press.
