GENERAL DISCUSSIONDr, Gert Ehrli& (Srhsnerjady, New York) said ; The question of how well pressurepeaks reflect the state of gas adsorbed on a surface arises quite generally in flashdesorption experiments. For filamentary samples, artifacts can be caused by diffusioninto the bulk or by conversion of a weak binding state to a more stable form duringwarm up. Conversion is usually concomitant with evaporation. Under theseconditions the desorption trace prescnts a qualitatively correct picture of the bindingstates. However, it will not give a quantitative account of the amount of materialpresent prior to observation, or of the correct desorption kinetics. If during evapora-tion of a state, conversion starts but terminates (possibly through saturation of acces-sible sites) before the weaker state is conipletely evolved, then the desorption tracemay show a peak which is not indicative of a separate entity on the surface.Ineither situation the rate of heating plays an important though often neglected role.Evaporation, having the highest activation energy, will be favoured over competingsurface processes by heating the sample at the highest possible rate. With oscillo-graphic recording the only real limit is set by the requirement that the time interval forevolution should be long by comparison with the time of flight through the cell(usually -500 psec). From desorption spectra over a wide range of heating rates,the extent to which the original distribution is perturbed can easily be determined,provided that at low temperatures the rate of conversion does not already predominate.This contingency can be simply explored by adsorption measurements at differenttemperatures. Finally, the peaks in a desorption trace depend both on the rate ofheating and of pumping ; comparison of peak temperatures therefore are useful onlyif the other variables are properly accounted for.Of particular interest in wnnecl;ian with the work of Hayward et al.are the demsp-tion studies of Mimeault and Hansen.1 After hydrogen adsorption at 100°K and-5 x 10-8 mm, they find an additional weak state of binding on tungsten. Theresolution of their detector is unsufficient to distinguish any fine structure ; however,in experiments with deuterium they were able to show that this state does not exchangewith hydrogen and is presumably molecular.On rhodium and iridium, a low tempera-ture state was also isolated. In contrast with tungsten, this did undergo isotopicmixing, suggesting the presence of adatoms.Dr. D. 0. Hayward (Imperial CoZZege) (communicated): It has been pointed outin our paper and also in the comments of Dr. Ehrlich that spurious pressure peaks mayoccur in desorption ‘‘ spectra ” which are not related to separate states of binding ofthe gas on the metal surface. The pressure profile necessarily reflects the state ofsurface occupation in a quantitative manner only when the adsorbed layer hasreached equilibrium prior to the desorption, i.e., there is no possibility of conversior,from weak to strong binding states during the warm-up.Work with many gases onevaporated metal films shows that it is extremely difficult to achieve equilibriumat low temperatures and this probably applies, although to a lesser extent, to filaments.Hence, we agree with Dr. Ehrlich that desorption “ spectra ” obtained by heatingfrom low t5mperatures should be treated with caution.Dr. S. Cerni (Inst. Physic. Chem., Prague) said: In connection with the paper ofBrocker and Wedler, we have also measured the heat of hydrogen adsorption onevaporated nickel film? The evaporation took place at a vacuum of 5 x 10-8 mm Hg1 Mimeault and Hansen, J. Chem. Physics, 1966, in press.2 Cern9, Pontx and Mhdek, J. Catalysis, 1966, 5, 27.10FIG. 1.-Interaction of oxygen with rhodium, as observed in the field emission microscope.(a)Clean Rh ; voltage V = 13.2 kV for emission of 5 x 10-9 A ; (b)-(e) tip at 79°K exposed to oxygenstream at p < 10-8 mm for increasing time intervals. (b) V = 14.1 kV ; (c) 14.7 ; ( d ) 14.8 ; (e)15.6 ; (f) surface warmed to 300°K ; V = 13.9 kV ; (9) clean Rh at 300°K ; V = 13.8 kV ; (h) ex-posed to 0 2 , V = 15.1 ; (i) same surface after continuing interaction at 79°K ; V = 16 kV.[To face page 102GENERAL DISCUSSION 103and at room temperature of the calorimeter wall. The weight of the film was 44 mgand its area was approximately 1000 cm2. The initial heat of adsorption was 25-26kcal/mole and the heat-coverage curve gradually decreased, in agreement with resultof Brennan and Hayes 1 and at variance with that of Brocker and Wedler. In Wedler’spaper the fundamental difference of the results from those of all the other authors whohave studied calorimetrically the Ni-HZ system is accounted for by two factors : thedifference in the vacuum and the wall-tsmperature during the evaporation.I do not think that the difference of our and Wedler’s result is due to contaminationof our films and cleanliness of his : (a) the amount adsorbed on 1 cm2 was in bothcases the same, viz., about 3 x 10-3 pmoles.(b) The most significant source of residualgases above the nickel film is during the evaporation of the filament and therefore thepresence of hydrogen, carbon monoxide and inert gases in the residual gas phaseshould be considered. The pressure did not change with time after the film depositionin my calorimeter; this can be explained neither by the equilibrium pressure ofhydrogen, nor of carbon monoxide (the number of adsorption sites for M2 and COon nickel probably does not differ substantially 2).Therefore in my system the resi-dual pressure of 5 x 10-8 mm Hg was caused most probably by inert gases. ( c ) Thisalso suggests the fact that the nickel filaments used by Brennan and Hayes and by uscame from various sources, so that composition of residual gases probably differed ;the measured heats, however, agreed. (d) Further arguments are available 3-5 infavour of the statement that the film contamination under conditions used in ourcalorimetric work as well as in that of the authors mentioned in the Wedler’s paper,was of no essential significance, not exceeding 1 % of surface metal atoms.The size of nickel crystals doubtlessly depends on the wall temperature during thefilm deposition. However, it is difficult to understand why nickel films deposited at77°K should expose to the gas phase crystallographic planes different from thoseexposed by films deposited at room temperature.Moreover, their 5lms were afterdeposition at 77°K sintered and the heats of hydrogen adsorption were the samewhether the nickel films were tempered at room temperature or at 60°C. ThereforeI think that the differences between the results reported here by Wedler and thosepublished by other authors are not due to a different crystallographic structure of therespective nickel films caused by different wall-temperatures during deposition.Evaporated films are always polycrystalline.It can be expected that with nickelthe heat of adsorption on crystallographic planes with various indices differs, thoughthis effect is with f.c.c. metals less marked than with b.c.c. metals.6 At 273°K theadsorbed hydrogen is most probably mobile.19 7 Under these circumstances theheat-coverage curve of hydrogen on nickel should exhibit a slightly decreasing course,as has been found by several authors. I do not think that the constancy of the heatof hydrogen adsorption reported in his paper can be accounted for by the homogeneityof the film as he suggests and it is difficult to find any plausible mechanism by whichthe reported constancy could be explained.The same applies to the reported increaseof the adsorption capacity of the film in each adsorption-desorption cycle.En conclusion also for oxygen adsorption, which is different in character, the heats of1 Brcnnan and Hayes, Trans. Fara&y SOC., 1964, 60, 589.2 Gundry and Tompkins, Trans. Faraday SOC., 1957, 53, 218.3 Knor and Ponec, CON. Czech. Chem. Comm., 1961, 26, 579.4 Becker, Ado. Catalysis, 1955, 7, 135.5 Wheeler, Structure and Properties of SoZid Surfaces, (ed. Gomer and Smith) (Univ. of Chicago6 Rootsaert, van Reijen and Sachtler, J. Catalysis, 1962, 1, 416.7 Wortman, Gomer and Lundy, J. Chem. Physics, 1957, 27, 1099.Press, 1953), p. 439104 GENERAL DISCUSSIONWedler for nickel 1 and for iron 2 are lower than those obtained by other authors.3-5Apart from the problem of the calorimeter calibration I would ask whether it can beexcluded that too low values of heat are systematically obtained because of (a) theindirect method of evaluation requiring one to ascertain the kinetics of the heatevolution, and (b) the transfer of some part of the evolved heat from the interior of thecalorimeter through the gas phase to the surrounding ice-bath without being registeredby the thermometer winding.Dr.M. W. Roberts (Queen's University, Belfast) said : I do not consider the extentof hydrogen uptake on nickel to be any guide as to the cleanliness of the film surface.For example (see table 1 of our paper), nickel surfaces which have adsorbed a mono-layer of oxygen at - 195" after heating to 22 or 150" in vacuo can then adsorb hydrogenextensively.The heat of hydrogen adsorption may, however, be very different fromthat observed with a clean metal ; the work function change is certainly different.Prof. G. Wedler (Technische Hochschule, Hannover) said : tern? uses the amountof hydrogen adsorbed on 1 cm2 to compare the cleanliness of his films and our films.This method is not valid since the films have different thicknesses and differentroughness factors which cannot be determined exactly enough to compare thecoverages. Furthermore, contaminated or partially precovered films sometimesadsorb even more gas than clean films.6-8Dr. Cernf 9 evaporates his films in a vacuum of 10-7 to 10-8 torr and keeps themin this vacuum up to 18 h before the first doses of gas are admitted.The constantgas pressure represents an equilibrium pressure or the equilibrium between the rate ofdesorption of gas in the apparatus and the rate of pumping. Since the evaporatedfilms act as good getters, they will chemisorb all the active gas. The composition ofthe gas in an ultra-high vacuum system before and after the evaporation of a nickelfilm has been studied by Gentsch10 in this laboratory by means of an omegatron.Before the evaporation the residual gas contained about 40 % CO. After theevaporation>f the film all the CO was gettered. Under these conditions mentionedabove Dr. Cernys films must be contaminated to a noticeable extent even if theresidual gas only contains some small percentage of active gases.The great influence of the surface contamination on chemisorption effects, especi-ally in the system Ni+H2, is clearly demonstrated by the observed changes of theelectric resistance.When ordinary high vacuum conditions (10-6 to 10-7 torr) wereused, only a decrease in the resistance of the nickel films could be observed with thehydrogen adsorption.11 With the improvement of the vacuum conditions the initialincrease of the resistance became more and more determining (ref. (3), (8), (9) and (10)of our paper). Similar observations have been made with the chemisorption of formicacid 12 and water vapour.13Fig. 1 demonstrates that even a residual pressure of 5 x 10-9 torr has a greatinfluence on the adsorption effects in the Ni+H2 system at 273°K.The fully drawn1 Wedler, Z. physik. Chem., 1960, 24, 73.3 Brennan, Hayward and Trapnell, Proc. Roy. SOC. A, 1960, 256, 81.4 Klemperer and Stone, Proc. Roy. Sac. A, 1957, 243, 375.5 Beeck, Adv. Catalysis, 1950, 2, 151.6 Ponec and Knor, Coll. Czech. Chem. Comm., 1961, 26,29.7 Siddiqi and Tompkins, Proc. Roy. SOC. A, 1962, 268,452.8 uinn and Roberts, Trans. Faraday Sac., 1962, 58, 569.9 8 ern?, Ponec and Hlhdek, J. CataZysis, 1966, 5, 27.10 Gentsch, 2. physik. Chem., 1960, 29, 55.11 Suhrmann and Schulz, Z. ph.ysik. Chem., 1954, 1, 69.32 Suhrmann, Kern and Wedler, 2. physik. Chem., 1963, 36, 165.13 Suhrmann, Heras, Heras, Wedler, Ber. Bunsenges.physik. Chem., 1964, 68, 511, 990.2 Wedler, 2. physik.Chem., 1961, 27,388GENERAL DISCUSSION 105lines refer to the nickel film discussed in detail in our paper. It had been evaporatedunder a vacuum of < 2 x 10-10 torr. The exact pressure will have been about 10-11tsrr as can be seen from measurements with an omegatron.1 The broken lines referto the adsorption of hydrogen on a nickel film evaporated under a residual gaspressure of 5 x 10-9 torr. Both the films were treated in exactly the same manner.The rise in resistance, however, is smaller and the heat of adsorption is higher in thesecond case. If these curves (broken lines) are shifted so that the maxima of theresistance coincide (dotted curves), no difference between the behaviour of the twofilms is to be observed at higher coverages.This may indicate that at high coveragesthe same processes will occur. At smaller coverages, however, the heat of adsorptionis clearly higher for the contaminated film.C I1.51 -alo’ - 1 . , . . . . , , , , , ; 10 0.5 I a 0 coverage (lo15 molecules/cm2)F~G. I .-Variation of the resistance of nickel films and of the heat of chemisorption with coverage ofhydrogen at 273°K.residual gas pressure 2 x 10-10 torr ;- - - - - residual gas pressure 5 x 10-9 torr ; . . . . . . curve - shifted by 0.4 x 1015 molecules/cm*.The increase of the heat of adsorption by a partial precoverage of the films hasalso been reported by other authors, e.g., by Dr. Cernf himself in the system Mo + H2after preadsorption of oxygen,2 by Klemperer and Stone (ref. (13) in our paper) in thesystem Ni+02 after preadsorption of hydrogen, by Bagg and Tompkins 3 in thesystem Fe+H2 after preadsorption of CO and in the system Fe+02 after H2 pre-adsorption.Concerning the structure of the films, the crystallographic planes exposed to thegas phase may be different in the films used by the other authors and in the films used1 Gentsch, private communication.2 Knor and Ponec, Cull.Czech. Chem. Comm., 1961, 26, 579.3 Bagg and Tompkins, Trans. Faraday Suc., 1955, 51, 1071106 GENERAL DISCUSSIONby us. tern9 does not consider that we used very thin films, which were rather smooth(roughness factor 1-5-2) whereas tern9 and the other authors used films 10-20 timesas thick with roughness factor of 5-10. These rather porous films may have a greaterperceEtage of high indexed planes in their internal surface and therefore a greaterheterogeneity.The statement of Cerng that all the heats of adsorption reported by us were smallerthan those reported by other authors is not correct.For oxygen adsorption on ironfilms at room temperature Bagg and Tompkins 1 found 71 kcal/mole, we 2 observed100-120 kcal/mole, and Brennan reported a value of 133 kcal/mole (ref. (19) in ourpaper). The dependence of the heat of adsorption of oxygen on nickel on surfacecoverage has never been reported by us. Preliminary measurements of 0 2 adsorptionon a Ni film have only been used to study the properties of the new type of the calori-meter.Concerning the last question of Cernf, the kinetics of the adsorption must beconsidered in the evaluation of the heats of adsorption.An extrapolation of thecooling curve without any correction leads to wrong (too high) values of the heat ofadsorption. A transfer of part of the heat evolved in the calorimeter through the gasphase without registration by the thermometer is impossible, since in our experimentsthe outer jacket of the calorimeter is pamped to a vacuum of < 10-8 torr.DP. D. A. King (ZnzperiaZ CoZZege) (communicated) : Two sets of anomalous resultsobtained by Brocker and Wedler could both be attributed to an underestimation of theamount of gas pumped out of the cell during desorption. First, they report that asthe nickel films approach saturation and an equilibrium pressure is established, highercoverages could apparently be attained by simply pumping gas out of the cell, and thenreadsorbing further doses.Secondly, adsorption in this system proceeds veryrapidly, with an activation energy for adsorption which must be very close to zero, sothat heats of adsorption and desorption should bc equivalent : however, they evaluatecalorimetrically a desorption heat at high coverages as 19 kcal/mole compared withheats of adsorption of - 15 kcal/mole at similar coverages. The accuracy with whichthe amount of gas pumped out of the cell can be determined is limited by difficultiesin the absolute calibration of pressure gauges at low pressures, and the estimation ofvolumes in a system containing pumps. Thus, if this amount had been consistentlyunderestimated by -20 % due to calibration error, the subsequent AR/& values andchemisorption heats determined on readsorption (their fig.3) would be coincident.Similarly, using the same correction factor, the desorption heat would be reducedfrom 19 to 15 kcal/rnole, consistent with the heats of adsorption.Dr. D. Brennan (University of Liverpool) said : It is useful to make some generalcomments about the shapes of film calorimeters and the determination of their heatcapacities. The use of a spherical shape for a calorimeter gives rise to a number ofcalorimetric dificulties, without a compensatory advantage of simplifying the problemof the distribution of the adsorbate within the film. Unless a central diffuser is used,as in the work reported earlier by Hayward and coworkers, to achieve a uniformdistribution of adsorbate prior to the adsorption, the geometry of an inlet tubeconnected to a spherical vessel could be more complex than that of a cylindrical tube.In the absence of uniform distribution of the adsorbate, a cardinal requirement ofany calorimeter is that a given quantity of heat liberated in any part of the calorimetershould give the same response.This is readily achieved in a cylindrical vessel since,with care, uniform wall thickness can be obtained and the sensing wire can also beuniformly wound. Thus, unit length of sensing wire has the same amount of glass1 Bagg and Tompkins, Trans. FaradGy Soc., 1955,51, 1071.2 Wedler, Z.physik. Chern., 1961, 27, 388GENERAL DISCUSSION 107associated with it and, no matter if the film is thicker in one place than in another orthe adsorbate is not uniformly distributed, the calorimetric response will be the samefor a given heat of adsorption no matter where in the calorimeter it is liberated.Now, in a spherical vessel, the amount of glass to be associated with unit length ofwinding for 2 given pitch will depend on the location; the turns near the poles areespecially exceptional. The importance of this effect will depend on how closelywound the sphere is; clearly, the smaller the pitch, the less serious the effect.Thepoint to be made, however, is that even with a wall of uniform thickness, the calori-metric sensitivity is not everywhere the same. This intrinsic non-uniformity is ofimportance when there is non-uniform distribution of adsorbate, but it could also beof importance in the determination of the heat capacity.The quantitative solution of the problem of the distribution of the heat liberatedby current fiowing from small electrodes attached to a large conductor is oftendifficult.For a spherical shell conductor, as for the spherical calorimeter, qualita-tively the best that can be achieved is a symmetrical, as opposed to a uniform, dissipa-tion of heat, and this requires diametrically opposed circular electrodes and uniformconductor. Here, it is virtually impossible to prepare a uniform evaporated metalfilm under these circumstances. If the intrinsic non-uniformity of the calorimeter issignificant, then the measured heat capacity will be a complex function of the diameterof the calorimeter, the number of turns of the resistance thermometer and the distri-bution of the current in the film.Should the electrode assembly not be completelysymmetrical, then it might be, for example, that more heat would be dissipatedequatorially and the derived heat capacity would then be even more unrepresentativeof the calorimeter as a whole.Additionally, it is very difficult to make a sphere with uniform wall thickness andto check that the wall is indeed uniform without first breaking the calorimeter; sinceuniformity of wall is so important, this is a serious deficiency. Again, it is verydifficult to wind the resistance wire uniformly on to such a surface.Regarding thesealing of the resistance wire into the wall, there is some risk that this could disturb theuniformity of the wall; at the same time, the need for such a measure is doubtful,since there is no evidence that the rate of heat transfer through the wall to the wirewound on it is in any way inadequate to the needs of this kind of experiment. If thewire were used as a heater, as it might conceivably be in a calibration run, thenimproved thermal contact could be some advantage, but the limiting factor is thethermal conductance of the glass between turns. Clearly, the details of design ofthese calorimeters are very important.In contrast to these difficulties, the procedures for calibration of cylindrical calori-meters are relatively reliable, if not entirely free of troubles.Two main methodshave been adopted, viz., to dissipate electrical energy in a film deposited in thecalorimeter, or to use the resistance thermometer or additional external winding as aheater. We believe that dissipation of heat in a film is the most satisfactory method inthat it most closely simulates an adsorption measurement. This nethod has beencriticized on the grounds that it is difficult to get electrica! contact with the film but itis our experience that this can be achieved satisfactorily. We have used a number ofdifferent types of contacts, and always the results have been independent of the natureof the contact and the film in cse. The method also has the powerful advantage thatdeterminations can be made for different parts of the tube. There is no validity inthe criticism 1 that, since the boundaries of the film are indistinct, the area of heatliberation is imprecise and the derived values of heat capacity could be a function of1 tern$, Ponec and HIBdek, J. Catalysis, 1966, 5, 27108 GENERAL DISCUSSIONthe current in the film.The only consideration is that the film should be entirelywithin the uniform region of the tube constituting the calorimeter.We are not entirely happy about heat capacity determinations involving externalheaters, particularly the method in which a stationary state is established and then thecurrent switched off. Our attempts to use the resistance thermometer as a heater 1have never given good results because the temperature of the wire was always appreci-ably above that of the wall.The use of a second winding as heater seems to be moresuccessful, but we remain concerned about temperature gradients ; temperatureequilibm ation longitudinally is relatively slow. Finally, the stationary state procedureis so different to that used in an adsorption determination, that, other methods beingavailable, its use in our opinion is unnecessary.There are three points that I would like to make specifically in relation to the paperby Brocker and Wedler. (i) While it is true that our ultimate vacua fall no lower thanthe pressure ranges 10-8 to 10-7 torr, the area of our films is such that there is not thecontamination available to dirty the available surface to any significant extent.Further, where comparision can be made between results obtained on evaporatedmetal films and on small area surfaces maintained in ultra-high vacua, agreement issatisfactory.(ii) The films of Brocker and Wedler are very thin in comparison withthe films used by other investigators. Their roughness factor is only 2, whereas oursis about 10 ; indeed their films are not sufficiently thick to show a smooth dependenceof area on weight. The properties of such very thin films are likely to be perturbed bythe substrate and to be different from those of thicker films. (iii) If the heat of adsorp-tion of hydrogen on nickel were as low as reported by Brocker and Wedler, then it isdifficult to understand why it is not possible to remove the entire adsorbed layer bypumping for about an hour.On the other hand, it is also difficult to understand whya small part of the adsorbed layer can be removed so quickly by pumping, even thoughthe associated heat is reported to be 19 kcal mole-1.Prof. G. Wedler (Technische Hochschule, Hannover) said : The difficulties men-tioned by Brennan are well known to us. Therefore we have critically studied possibleerrors in the determination of the heats of adsorption. After considerable experiencehad been gained in this laboratory, reliable spherical calorimeters can be built.The deviation from the average thickness of the calorimeter bulb has beenmeasured and was less than 3- 5 % for more than 80 % of the bulb. Only the bottomof the vessel and the part near the inlet tube were slightly thicker.This has been takeninto account, when the thermometer wire was sealed on the bulb. The averagedistance between the windings was 3 mm. Care was taken, however, that the amountof glass per cm of the thermometer wire was the same all over the bulb, i.e., a smallerdistance between the windings was used at the bottom and at the top of the bulb.The diameter of the contact foils was 1 cm, the diameter of the bulb 5 cm. There-fore more heat is liberated near the foils than at the “ equator ” of the bulb. Sincethe ratio ‘‘ amount of glasslcm of thermometer wire ” is the same all over the bulb,the non-uniformity of the heat liberation cannot influence the determination of theheat capacity. Furthemore, we found experimentally that using the stationarycalibration method both the warming curve as well as the cooling curve are representedby an exponential law with exactly the same value of r/C.This result indicates thatthe initial differences in the distribution of temperatures are very quickly balanced,presumably due to heat radiation within the bulb. In this respect the sphericalgeometry of our calorimeter is more advantageous than a cylindrical one.Furthermore, (a) the values of r/C obtained by use of the stationary method andthe pulse method differ for the calorimeter used in this work by 6 %. The values of1 Whaba and Kernball, Trans. Faraday Soc., 1953,49, 1351GENERAL DISCUSSION 109r/C obtained from the cooling curves when measuring the heats of adsorption liealways between the values found by the two calibration methods.(b) If the outerjacket is pumped to high vacuum (< 10-8 torr) the experimental value of the coolingconstant r (29 mW/deg.) agrees with the theoretical value (32 mW/deg.), which isobtained with the assumption that all the loss of heat is due to radiation from theuniformly heated bulb. (c) The value of the heat capacity C does not depend on theenergy used in the calibration even if it is altered by a factor of 10. It is also independentof the value of Y, if this is increased by admission of gas into the outer jacket of thecalorimeter, so that heat radiation and heat transfer by the gas phase are responsiblefor the loss of heat. ( d ) The value of C obtained by calibration using the pulsemethod ( e g , 0-63 cal/deg.) agrees well with the value calculated from the weight andthe specific heat of the material of the bulb (0.62 cal/deg. with the same calorimeter).All these results clearly show that the objections made by Brennan with respectto the calorimeter design are not relevant for the spherical calorimeters used in thiswork.The differences in the heat of adsorption of hydrogen on nickel reported byBrennan and the other authors and those found by us are at least 45 %. It is certainlynot possible to explain such a large difference by uncertainties in the determination ofthe heat capacities of the calorimeters.Prof. W. M. H. Sachtler (Amsterdam) said: Wedler mentions that the latticeparameter is slightly shorter for nickel films than for bulk nickel.We have made thesame observation for nickel and copper. With some copper films the lattice parameterwas 6 x 10-3 A shorter than in bulk copper. I would ask whether Wedler attributesthis contraction to the effect of surface tension in small crystals or whether he proposesa different interpretation.Prof. G. Wedler (Technische Hochschule, Hannover) (communicated): In reply toSachtler, we suppose that the contraction of the lattice we have observed with thin filmsof nickel and copper (ref. (23) and (24) of our paper) is due to the influence of the surfacetension. There is, however, no direct proof for this assumption. The effect couldalso be attributed to lattice rearrangements at the surface due to chemisorptionprocesses.However, we observed the lattice contraction for high-vacuum evaporatedfilms irrespective of the vacuum conditions applied during the X-ray diffractionmeasurement. Concerning Dr. King’s question, the possible error in the determina-tion of the amount pumped out of the cell certainly is far smaller than 20 %, since theeffective volumes have been measured by different methods and the calibration of thepressure gauge is reliable in the range of about 10-4 torr. Furthermore, an additionalirreversible sorption of hydrogen by nickel films at high coverages has also beenobserved in previous work using different ultra-high-vacuum apparatus and otherpressure gauges (ref. (3) of our paper).Dr. D. A. King (Imperial College) (communicated) : Observations made during thedeposition and subsequent sintering of nickel films in an ultra-high vacuum systemindicate that the reading on a pressure gauge attached to the cell during deposition isnot necessarily an indication of the degree of contamination of the film, and that thedegree of contamination is reduced by depositing heavy films.The apparatus andprocedure was similar to that used by Hayward, Taylor and Tompkins (this Discussion),background pressures of 1-2 x 10-10 torr being attained. Nickel filaments (JohnsonMatthey “spec pure ”) were outgassed for 20 h at a temperature just below thatrequired for evaporation of metal. When deposition was carried out, at a rate of - 1017 atoms sec-1, with the glass substrate uncooled (i.e., at N 328°K) the cell gaugepressure rose to 10-8 torr and thereafter fell continuously until the gauge was indicat-ing 6 x 10-10 torr after 40 mg had been deposited ; a fall to the “ background ” valueof 2 x 10-10 torr occurred when deposition was terminated.In contrast, when the cel110 GENERAL DISCUSSIONwalls were cooled to 78 or 195"K, no increase in the gauge reading above " back-ground " was observed during film deposition, in agreenent with the observation ofBrocker and Wedler. However, on subsequently warming the films to room tempera-ture after the deposition, with the cell open to the pumps, desorption spectra wereobtained (fig. 1). The dashed curve is the pressure burst obtained with a 27 mg filmdeposited at 78"K, while curves 1-4 are successive spectra obtained by depositing - 15mg at 78"K, warming to room temperature, recooling to 78°K and repeating theprocedure (total film weight 60 mg).Contaminant desorbing over this temperaturerange was considerably reduced by the latter procedure ; however, on finally heatingthe film to 460°K the pressure rose to a maximum of 3 x 10-7 torr.TFIG. 1.An estimate of the degree of surface contamination is difficult to make. From thearea under the desorption curve 1, and the conductance of the exit port, it was calcul-ated that 1014 molecules were desorbed during the warm-up to 290°K. As the filmarea was - l o 3 cm2, this represents a negligible degree of contamination ; however, itis clear from the effect of heating to 460°K that contaminants with higher desorptionheats were present-although it is possible that the contaminants are not concentratedat the metal surface but diffuse from the bulk during the heating procedure.The pressure reading in the gauge during film deposition does not sample the fluxof contaminant at the cell walls, as contaminant atoms or molecules from the filamentmust make at least one collision with the metal film before entering the gauge tubula-tion.In particular, if the sticking coefficient s for contaminant on the metal surfaceis unity (as it is for CO on W, see Gomer's paper and my discussion comment on it)the guage pressure will not rise above the " background " value during deposition,despite the presence of contaminant. Thus, the observation of a pressure rise duringdeposition, as when the substrate is uncooled, does not indicate that the film is morecontaminated but rather that at the higher temperature s < 1 for the contaminants.Thedegree of contamination is, in fact, not reduced by cooling the substrate duringdepositionGENERAL DISCUSSION 11 1Prof. S. Z. Roginskii (Moscow) said : Direct measurements of the differential heatsof chemisorption and of reactions involving chemisorbed molecules are of considerablesignificance. In connection with the conclusion of Brocker and Wedler I show a plot(fig. 1) from our experimental data (Tretyakov et al.) on the effect of H2 chemisorptionon the changes of the work function A$ andon the electroconductivity AK of nickel filmsobtained under very pure conditions atp - 10-10 torr (hydrogen was purified bydiffusion through palladium).The curve forvariations in the work function with coverageshows a maximum and a bend; two extremepoints and a bend majj be seen as well on thecurve for electroconductivity. This is evidencefcr the occurrence of three chemisorptionforms. These facts as well as the differentstrength of adsorption on various metal faces,and the very sharp drop in the rate of ex-change between adsorbed deuterium andordinary gaseous hydrogen, makes it difficultto understand the origin of the observedconstancy in differential heats of adsorption.This is also not consistent with the results ofcertain earlier thermochemical measurements.It would be of interest to find out the reasonfor these discrepancies.Dr.A. A. Hslscher (Amsterdam) said: Inthe adsorption of oxygen on cobalt andnickel Brennan et nl. found a smaller heatof adsorption at low coverage than at highcoverage. Might this be due to an increasein surface energy in the initial stages of-10 - 4P=I*10 id=2.I07 7 -----cIcI0 0.5 1.00FIG. 1.adsorption? Can a difference in surface energy, i.e. in ordering in the surface layersafter adsorption at 78 and 300"K, also be responsible for the fact that at the samecoverage the heat of adsorption at 78 is smaller than at 300"K? Brennan andGraham assume that it is difficult for the cobalt and nickel atoms to move upon adsorp-tion at 78°K. We think, however, that our field ion microscope experiments haveshowlz that the displacement of tungsten atoms upon adsorption of carbon monoxideand nitrogen at 78°K is possible.Would not one expect a.fortiori rearrangements tooccur upon yoxygen adsorption on cobalt and nickel?Dr. S . Cernf (Inst. Physic. Chem., Prague) said: The low value of the initialpoints on the heat-coverage curves of oxygen of Brennan and Graham for tungsten,cobalt and nickel in fig. 2 and 3 of their paper, has been observed also with othersystems. Wedler 1 found the same phenomenon with oxygen on iron, as I have withoxygen on molybdenum and in a more pronounced form with oxygen on nickel.2 Itis important to elucidate whether these low values are merely some experimentalartefacts of ininor importance, or due to some deeper and more substantial cause.The marked effect with oxygen, and its absence with carbon monoxide,3 seem to excludean explanation involving a partial adsorption of the first doses in the cold traps before1 Wedler, Z.physik.Chem., 1961, 27, 388.2 Cernf, Diss. (Institute of Physical Chemistry, Czechoslovak Academy of Sciences, Prague, 1963).3 Brennan and Hayes, Phil. Trans. A, 1965, 258, 347112 GENERAL DISCUSSIONthe calorimeter vessel (the extent of CO adsorption in the traps would be expected tobe the same as that of 0 2 , if not greater). Brennan et al. used cylindrical etchedcalorimeters, Wedler used a spherical calorimeter, and I used cylindrical drawncalorimeters with controlled wall homogeneity.1 This suggests that the observedphenomenon cannot be ascribed to some specific effect of the calorimeter used.Thefact, that both Brennan and Graham, and myself have found this phenomenon in amore pronounced form with low-melting metals (Ni, Go) in comparison with high-melting metals (W, Mo) seems to point to some deeper cause. Mechanisms whichcould be responsible are sinteriiig or rearrangement of the films by serious local risesin temperature 2 which has been mentioned by Holscher, or eventually splitting ofmetal-metal bonds 3 by initial doses. More heat-coverage curves carefully measuredin the initial region for gases with different bond energy and for various metals seemdesirable in order to find the mechanism responsible for this phenomenon.Dr. Gert Ehrlich (Schenectady, New Yurk) said: The behaviour of oxygen at ametal surface is markedly affected by the atomic arrangement of the interface.This isclearly revealed when rhodium is allowed to interact with 0 2 in a field emission micro-scope. At 79"K, this interaction lowers electron emission monotonically ; the majoreffects are concentrated along the zone running from the (100) towards the (1 10) plane.During the initial stages (c), the (110)'s themselves are excluded from change and thedarkening stops at the (430). When after prolonged exposure at 79°K (as typified bye) the oxygen is pumped out and the surface is warmed to 300"K, electron emissionincreases and drastically changes its distribution over the surface. After warming, thepreviously dark zone from (100) to the (1 10) becomes brightly emitting, and insteadthe zones from (100) to (111) recede into darkness.The same type of pattern isobtained by continuous exposure to oxygen at room temperature (h). Adsorption ofadditional oxygen at 79°K (i), after exposure of the surface at room temperature,lowers the overall electron emission ; the (1 10)'s are darkened, and planes in the [ 1 101zone adjacent to the (100) become emitting. The pattern characteristic of adsorptionentirely at low temperatures is not restored.These observations are consistent with the view that at 300°K oxygen brings aboutsome reorganization of the surface, but without any particular specificity for a givensurface feature or plane. At low temperatures, however, a chemisorbed layer isformed that in some way favours planes with kinked steps, such as the (310), at theexpense of less jagged planes like the (221).Part of this chemisorbed material istransformed into the high temperature state on warming ; the remainder is labile andis evolved into the gas phase. Most interesting are the differences in the roomtemperature behaviour of Ni and Rh. On the former,4 the oxygen appears to interactspecifically with the (1 10) planes. On Rh, there are as yet no indications for any suchprefereme.Mr. B. R. Wells (Queen's University, Belfast) said: With reference to fig. 4 ofBrennan's paper, the sequence from 4a to 4b shows an ordered rearrangement ofadsorbed oxygen to a configuration which does not occur in nickel oxides, all withinone atomic layer.This contrasts with what is known about the nickel+oxygensystem in that non-stoichiometey plays an important role. If we are to accept theresults of low-energy electron diffraction studies, we know that in the initial interactionsat - 300°K such phases as Ni70 and Ni30 exist near the gas/metal surface,s and. Quinn1 Cerny, Ponec and HlBdek, J. Catalysis, 1966, 5, 27.2 Brennan and Hayes, Phil. Trans. A , 1965, 5, 27.3 Sachtler and van Reijen, J. Res. Inst. Catalysis, Hokkaido Univ., 1962, 10, 87.4 EhrIich, Ann. N. Y. Acad. Sci., 1963, 101, 722.5 Park and Farnsworth, J. Appl. Physics, 1964, 35, 2220GENERAL DISCUSSION 113and Roberts 1 , 3 have shown that variations in the surface potentials obtained on theoxidation of nickel reflect differences in stoichiometry.Some recent work in thislaboratory 2 indicates that an oxygen deficient phase similar to the Ni30 of Park andFarnsworth 1 may be formed by heating a nickel surface, saturated at 77°K withoxygen, in vacuo to -360°K ; this implies a penetration of - 10 A by the oxygenspecies. Furthermore, this oxygen-deficient phase requires a higher activation energyof formation than the simple 6 : 6 coordination nickel oxide structure. Therefore, ifthis process can take place at -360°K it is questionable to dismiss the so-called" oxide state " on warming from 77 to 273°K. Because the surface does not react asa bulk nickel oxide surface is not a categorical argument against the 6 : 6 coordinationconfiguration in the surface, only that the subsurface metallic nickel is still influencingthe surface characteristics.If an oxygen-saturated nickel surface at 77°K is warmed in UdCUO to 273°K andrecsoled, when further oxygen is added at 77"K, although the work function increaseis only 1.0 eV this increase is caused by only +th to kth of the gas adsorbed initially,and which gave rise to a change of 1.4 eV on the clean metal.2 If we treat this asadsorption on an essentially metallic system, a monolayer of this adsorbed oxygenwould give rise to a work function change of between 5 and 6 eV.A more likelyexplanation is that this is adsorption on a semiconducting oxide, and therefore somesort of " oxide state " must have been formed.Dr. J. W. Geus (Staatsmijnen, Geleen, Netherlands) said: Measurements of theeffects of chemisorption of oxygen on the electrical conductance of evaporatedtungsten films confirm the interpretation given by Brennan et al.for this metal.4 Whenoxygen is admitted at 77°K to a tungsten film, the conductances measured at both 77and 273°K decrease equally until about one oxygen atom per metal surface atom istaken up. Since the decrease in conductance measured at 77°K is the same bothbefore and after heating to 273"K, no major rearrangement of the adsorbed layer iscaused by heating to 273°K. After adsorption of about one oxygen atom per metalsurface atom more oxygen is adsorbed at 77°K without this having an effect on theconductance. On heating to 273"K, the pressure first rises and then slowly decreases,while the conductance decreases symbatically.On recooling, the same decrease inconductance is observed at 77°K. This behaviour is in accordance with the observa-tion of Brennan and Graham, viz., that an activation energy is needed to chemisorbstrongly more oxygen on a tungsten surface covered with a monolayer. In thissecond stage of the adsorption process the decrease in conductance per oxygenmolecule sorbed is clearly smaller. When oxygen is admitted to a tungsten film at273"K, a small fraction of each dose is sorbed beyond a monolayer during migrationof oxygen over the regions of the surface already covered to the less accessible parts ofthe porous tungsten films. Since the effect of the oxygen sorbed beyond a monolayeron the conductance is smaller, the total effect per oxygen molecule adsorbed on theconductance is slightly weaker, when the oxygen is admitted at 273°K.A second remark has to be made about the character of the chemisorptive bondof oxygen on nickel.First, chemical bonds generally have to be interpreted by takingboth covalent and ionic contributions into account. Only when either the covalentor the ionic contribution strongly dominates, a description as covalent or ionic,respectively, is justified.At present the effect of oxygen sorption on the ferromagnetism of nickel no longer1 Quinn and Roberts, Nature, 1963, 200, 648.2 Roberts and Wells, Trans. Faraday SOC., 1966, 62, 1608.3 Quinn and Roberts, Trans. Farday SOC., 1964, 60, 899.4 Geus, Koks and Zwietering, J. Catalysis, 1963, 2, 274114 GENERAL DISCUSSIONyields conflicting information.1 Experimental evidence is available demonstratingthat the magnetic moments of the chemisorbing nickel atoms are decoupled from theferromagnetism of the metal on oxygen sorption. The magnetic moments of thedecoupled nickel surface atoms are oriented in an external magnetic field to a degreemuch smaller than those of the ferromagnetic nickel atoms. Therefore, the niagnetiza-tion decreases on sorption of oxygen. However, when the magnetization is measuredunder conditions where no thermodynamic equilibrium is attained in the time of themeasurement, sorption of oxygen not only decouples the moments of the chemisorb-ing nickel atoms, but also causes the thermodynaniic equilibrium nagnetization to bemore closely approximated.When the latter effect dominates, the apparent rnagneti-zation increases on oxygen sorption, although the ferromagnetism of nickel is de-creased. A decoupling of the moments of the adsorbing nickel atoms from theferromagnetism of the metal by sorption of oxygen points to a donation of electronsfrom the nickel atoms to the oxygen atoms. Consequently, the chemisorption bondhas an appreciable ionic character.This can be concluded also from the results of Lewis,Z who investigated the effectof oxygen sorption of the K X-ray absorption edge of alumina-supported nickel.These results point to a bond between oxygen and a nickel surface analogous to thatexisting in bulk nickel oxide. However, the possibility that a fraction of the nickelparticles is completely oxidized, whereas the remaining part of the nickel has still notreacted with oxygen, is not fully excluded in Lewis’ experiments.1The fact that Park and Farnsworth 3 observed that the photoemission from nickelpartially covered with oxygen follows a Fowler curve, does not necessarily indicatethat bonding of oxygen to a nickel surface is effected by mainly covalent forces.Asstated by MacRae,4 both nickel atoms and nickel-oxygen complexes are present in thenickel surface at low coverages. The emission from the still metallic part of the surfaceobeys the Fowler curve, whilst the pre-exponential factor is strongly decreased. Athigh coverages, Park and Farnsworth no longer observe obedience to the Fowler curve.Dr.A. Frennet (Ecole Roy. Miiitaire, Brussels) said : We have studied the adsorp-tion 5 of krypton and xenon over a wide range of relative pressure ( z 10-5 <p/po < z 0,8) on films of Ni, Cu, Ag, W, Mo, Ti, Ta, Re, Rh, Pt and Pd. We apply two methodsto analyze the experimental isotherms : the B.E.T. and the Dubinin method. TheDubinin equation,6 was recently demonstrated as valid in the low pressure range byHobson.7 For example, if we analyse the adsorption isotherm of krypton on amolybdenum film at 77°K (fig. 1) using the B.E.T. method (fig. 2), three straigh’ L 1’ inesare necessary, over the pressure range. With each of these curves is associated avalue of the number nm of molecules contained in the monolayer and a value of theB.E.T.constant C. The problem is to get a criterion to chose which of these valueshas physical meaning. For this purpose, all the experimental points are plottedfollowing the Dubinin equztion (fig. 3). Two types of experimental points may bedistinguished, those that lie on a straight line, corresponding to the low pressurepart of the isotherm, as predicted by the equation, and those that lie above thestraight line on the left of the figure, and corresponding to the high relative pressurerange part of the isotherm and at the same time at coverages greater than the mono-layer.1 Geus and Nobel, J . Catalysis, accepted for publication.2 Lewis, J. Physic. Chem., 1960, 64, 1103.3 Park and Farnsworth, Surface Sci., 1965, 3, 287.4 MacRae, Surfcce Sci., 1964, I, 319.5 Delaunois, Frennet and Lienard, J.Chim. Physique, 1966, 63, 906.6Dubinin and Radushkevich, Proc. Acad. Sci. U.S.S.R., 1947, 55, 331.7 Hobson, Can. J. Physics, 1965,43, 1934115 GENERAL DISCUSSIONAs Gaines and Cannon 1 noted, the only criterion of B.E.T. applicability is thesaturation of the monolayer, i.e., the relation ( p / p ~ ) " = ~ , , , = - derived from theB.E.T. equation, must be fullfilled. If we now apply the B.E.T. method to the pointscorresponding to a coverage equal or greater than the monolayer as determined in the1+JcI I04 0.2 0-3 0% 04 a6 M 04)PiPo (x lo2)FIG. 1 .-Adsorption isotherm of krypton, at 77°K on a molybdenum film previously saturated withH2 at 273°K.19-0 i-' 121 *120(log Pipd2FXG.2.-The krypton isotherm represented by the Dubinin equation.Dubinin graph, we obtain curve 1 in fig. 2. This curve gives a value of n, = 6-45 x 1018molecules in good agreement with the value of nrn = 6.55 x 1018 molecules given by theDubinin equation.The good agreement between the number of krypton and xenon moleculescontained in the monolayer does not depend upon the nature of the metal film.1 Gaines and Cannon, J. Physic. Chern., 1960, 64,997116 GENERAL DISCUSSIONThus, the quantities of adsorbed krypton and xenon satisfy the relationshowing that, when the monolayer is filled, the adsorbed krypton and xenon do notknow the surface structure of the film. The B.E.T. constant C has small values onall the metals so that on Ti and Ta, where Cis -200, the monolayer is filled at relativepressure of ~5 % ; on Mo, Re, Rh and Pd, where C is 25-50, it is necessary to workat p/po> 10-15 %, and on some metals such as Pt the monolayer is not filled at p/povalues as high as 50 %.(nm)Kr/(%Jxt? = %/ai<r = 1,2070.1 0.7 09 0.4 0.5 0.6 0-7 04PJPOhave to be muItiplied by 10 and 100 respectively.C = 43.6 C = 297 C = 1670FIG.3.-The krypton isotherm represented by the B.E.T. equation. For curve 2 and 3 the coordinates(1) N , = 6.55 X 1018 ; (2) N , = 5.61018 ; (3) N, = 5 . 0 ~ 1018It is remarkable that two methods established from completely different assump-tions, give values of the number of molecules contained in the monolayer in goodagreement. We also believe that many of the isotherms obtained atp/po < few percenton metal films and analyzed using the B.E.T.method give suspicious values of thearea.Dr. D. Brennan (University of Liverpool) said: With regard to the remarks ofFrennet, in comparing the adsorptive capacities of various metals for krypton andxenon1 we were careful not to identify strictly any one parameter with the monolayercoverage. Rather, like parameter was compared with like for each of the adsorbates,with the conclusion that both species have equivalent covering capability. Thisresult was interpreted with good self consistency to mean site adsorption.1 Brennan and Graham, Phil. Trans. A, 1965, 258, 325GENERAL DISCUSSION 117We too have found that the observed isotherm at low pressure could be representedby the equation of Dubinin and Radushkevich.However, the gradient of the graphof log S against [log (P/P0)]2 was always greater for krypton than for xenon and therelative gradients varied from metal to metal so that the values of the intercepts on thelog S axis and the derived Sm values showed no consistency. For example, a largediscrepancy was found for titanium, viz., Sm (Kr) = 29-9 x 1017, Sm (Xe) = 21.2 x 1017atoms (cf. BET. monolayer values of 21.2 x 1017 and 19.8 x 1017 atoms, respectively),whereas close agreement was found for nickel, viz., Sm (Kr) = 12-7 x 1017, Sm (Xe) =12.3 x 1017 atoms (cf. B.E.T. monolayer values of 12.5 x 1017 and 13.2 x 1017 atoms,respectively). Possibly the origin of the difference between our findings and thosereported by Dalaunois, Erennet and Lienard is the fact that these authors used evapor-ated films which had been pre-treated with hydrogen or methane.Concerning the problem of selecting a parameter which can be closely identifiedwith the monolayer coverage, there is no evidence that Sm of the Dubinin-Radush-kevich equation has any utility.On the contrary, Hobson and Armstrong,1 find fornitrogen, argon and helium that obedience to the Dubinin-Radushkevich does notimply values of Snt which can be directly equated to the monolayer coverages (seealso Hobson 2). We believe the existence of a very marked point B means that thecorresponding coverage is close to the monolayer coverage. The B.E.T. equation,applied in the region of the point B, is merely a convenient algebraical device forobtaining a parameter which, under these circumstances must be very close to, if notequal to, the monolayer value.Monolayer values derived from the isotherm atpressures other than corresponding to point B must be regarded with diminishingconfidence as the pressures differ more and more from that of the point B and as thepoint B itself becomes less and less well defined.Dr. V. Ponec and Dr. S. Cernf (Inst. Physic. Chem., Prague) said: The results ofFrennet and his collaborators 3 have again proved the validity of an isotherm of thetype :Experimental data can be compared either with a model of volume filling of the micro-pores of the adsorbent (model I),495 or with a model of a plane monolayer on theheterogeneous surface (model II).6 We introduce the following quantities : E = 2.3RT log (ps/p), wherep, is the saturation pressure,p the equilibrium pressure, E representsthe free molar enthalpy, i.e., the thermodynamic adsorption potential (not thepotential of adsorption forces) ; W is the adsorption volume filled by the adsorbateat the pressure p , is., the volume of filled micropores, as generally accepted ; WQ isthe total adsorption volume (total volume of micropores); na is the number ofadsorbed molecules ; is the number of molecules in the complete monolayer ; urepresents the volume of the adsorbed molecule andco the area of the molecule adsorbedon the surface; k and k’ are constants.MODEL I is characterized by the following conditions :temperature ;log n, = c - B(1og p)’.(a) (aElaT)w=.a, = 0, i.e., the characteristic curve E = f ( W ) is invariant with(b) W = WO exp (- kez), [ Wo # f(T)], i.e., the Dubinin-Radushkevitch equation.1 Hobson and Armstrong, J .Physic. Chem., 1963, 67, 2000.2 Hobson, Can. J. Physics, 1965,43, 1934, 1941.3 Delaunois, Frennet and Lienard, J. Chim. Physique, 1966, in press.4 Dubinin, Chem. Rev., 1960,60, 235.5 Dubinin, in Chemistry and Physics of Carbon, vol. 3, ed. Walker (Dekker, Inc., N.Y.), 1966,in press. 6 Hobson and Armstrong, J. Physic. Chem., 1963, 67,2000118 GENERAL DISCUSSIONMODEL 11 is characterized by the following conditions :(c) (a~/dT),,~ = 0, or (c’) (dE/dT),,, = 0 ;( d ) nu = tl?n exp (- W),(i) The validity of the postulate (a) by itself permits thermodynamics calculationsand also the evaluation of the isotherms of other similar adsorbates from a singleisotherm of a standard adsorbate.(ii) The validity of postulate (a) does not dependon the validity of eqn. (b) and vice versa. Analogous statement holds for model 11.(iii) The condition (c) has substantially different thermodynamic implications com-pared with conditions (a) and (c’): validity of (c) means that the entropy changeconnected with the transfer of molecules from the liquid to the adsorbed state is zero,whereas validity of (a) (and most probably of (c‘) as well) means that by this transferentropy decreases or is zero.For microporous active charcoal and zeolitic molecular sieves a wide validity ofthe postulate (a) has been ascertained and independent information on the structureof these materials suggests that the validity of model I.I9;2 For active charcoal also,eqn.(b) holds well. For macro-porous or non-porous carbons eqn. (e) holds insteadof (b) :This relation after the substitution for E gives the Freundlich equation. Recently itwas ascertained that with materials exhibiting little or no porosity (Ti@, EaS04,glass, etc.) an equation holds which has the form either of (d) or of (e). Thepapers 3-5 seem to show that with non-porous substances the condition (c) is fulfilledand that, most likely, eqn. ( d ) and the validity of model I1 are involved.A final decision concerning the properties of the underlying physical model and theeventual decision of which condition (a), (c) and (c’) actually applies would requiredata for a greater number of adsorbents and a wider range of temperatures.How-ever the Dubinin-Radushkevitch equation is based on the temperature invariancy ofthe characteristic curve E = .f( W), whilst in ref. (4)-(6) the temperature invariancy ofthe characteristic curve E = f(7iU) is postulated. Therefore eqn. (d) describes an essenti-ally different physical situation than the Dubinin-Radushkevitch eqn. (6).The applicability of eqn. (d) for the surface area estimation and the comparison ofresults with those of the BET. equation form a separate problem. Experience hasshown that the B.E.T. equation usually does not hold well for microporous materialsand, moreover, gives non-realistic high values of their surface area.The structure ofthese materials justifies interpretation of experimental data in terms of model I, i.e.,on basis of volume filling of pores. For inaterials having no microporosity thesurface is gradually covered by layers of the adsorbate and thus it is justified tointerpret experimental data on the basis of model I1 or by means of eqn. (d), respec-tively. However, it is misleading to call eqn. (d) the Dubinin-Radushkevitch equationand thus to invoke properties inherent to model I, particularly the idea of volumefilling of the adsorption space. At a single temperature eqn. (b) and ( d ) simultaneouslyhold and lacking additional information it is difficult to decide which of these twocases actually applies.= f(T) ?I.W = WO exp (- k ‘ ~ ) , [ WO # f(T>].1 Dubinin, Chem.Rev., 1960, 60, 235.2 Dubinin, in Chemistry and Physics of Carbon, vol. 3, ed. Walker (Dekker, Inc., N.Y.), 1966,3 Hobson add Armstrong, J. Physic. Chem. 1963, 67, 2000.4 Endow and Pasternak, J. Vac. Sci. Techn., 1966, in press.5 Ricca, Bellardo and Medona, in preparation.in pressGENERAL DISCUSSION 119As can be expected from the physical model forming the basis of the B.E.T.theory, the BET. equation starts to hold from the coverage where the influence ofsurface heterogeneity is weakened and where higher adsorption layers are being formed,i.e., usually with na/nm about 0.8-1 -5. Eqn. ( d ) has been derived for monolayer adsorp-tion on heterogeneous surfaces and, accordingly, holds for the lowest pressures.Thepoint B, the value nm according to the B.E.T. equation and the value nm according tothe eqn. ( d ) are for theoretical isotherms always in the same region, and the agreementof surface areas evaluated in these three ways is therefore not surprising.There cannot be excluded the possibility that both equations, the B.E.T. as well as(d), are but appropriate empirical modes of finding a certain chosen point on isothermswithout any physical meaning. This approach, however, does not call for furtherdiscussion.Dr. J. Miipler, (Inst. hiorgan. Chem. Czechoslovak Acad. Sci., Prague) said;Frennet's finding that o x e / o ~ is constant and does not depend on the nature of theadsorbent has no general validity. In ref.(1) all the measurements of Frennet hasbeen made on the metal films previously saturated with hydrogen or methane, so thatthe variation with the measurements on the clean metal surfaces (see e.g., ref. (2), (3))is not surprising.Prof. R. A. W. Maul (Technischen Hochsclztale, Hannouer) said: Dr. Pone5 hasargued that the Dubinin-Radushkevich (D-R) equation cannot be applied to adsorp-tion phenomena on plane surfaces, since it had been derived in connection withsorption in highly porous carbons. As has been shown by Hobson4 the equationcan be derived by means of the Polanyi potential in connection with a certaindistribution for the adsorption energies. Thus the D-R equation can be effectivelyapplied also to non-porous matsrials as has been shown by a number of recent studies.In a paper by Haul and Gottwald 5 on residence times of rare gas atom adsorbedon Pyrex glass at low surface coverages it has, however, been shown, that, e.g., forxenon at temperatures between 130 and 158°K the intercept of the D-R plot issmaller by a factor of 2 than the B.E.T.monolayer capacity. A similar behaviour hasbeen found by Hobson 4 for argon between 63 and 77"K, whereas in our experimentsbetween 83 and 100°K both values are in good agreement. It is suggested that thiseffect might be explained by surface condensation phenomena occuring at tempera-tures below the two-dimensional critical temperature, the ideal value of which maybe taken as half the three-dimensional critical temperature according to De Boer.6In this case the intercept of the D-R plot can no longer be interpreted as a moiio-layer capacity in the sense of Kaganer.7 On the other hand, as has been mentionedin this Discussion by Dr.Frennet, the Belgian authors 8 have carried out adsorptionexperiments with Mr, Xe and CH4 on a large variety of metal films and found that evenat 78 and 90°K respectively the D-R and B.E.T. values of the monolayer capacity arein good agrcerncnt. The present authors 5 wish to emphasize that this is not incontrast to their findings since the two-dimensional critical temperature on metalsurfaces will presumably be considerably less than the ideal value due to polarizationeffects. Furthemiore with metal surfaces the arrangement of the adsorbed rare gasatoms may be strongly influenced by the geometry of the underlying crystal face.1 Delaunois, Frennet and Lienard, J. Chim. Physique, 1966, in press.2 Brennan, Graham and Hayes, Nature, 1963, 199, 1152.3 Brennan and Graham, Phil. Trans. A, 1965, 258, 325.4 Hobson and Armstrong, J. Physic. Clzern., 1963, 67, 2000.5 Haul 2nd Gottwald, Surface Sci., 4, 1966, in press.6 de Boer, The Dynamical Charccter of Adsorption (Clarendon Press, Oxford, 1953).7 Kaganer, Proc. Acad. Sci. U.S.S.R., 1957, 116, 603.8 Delaunois, Frennet and Lienard, J . Chim. Physique, 1966, 63, 906120 GENERAL DISCUSSIONProf. I(. S. W. Sing (Brunel University) (communicated) : It is generally acceptedthat the application of B.E.T. equation should be restricted to that part of the isothermwhich includes point B. For nitrogen adsorption at - 196" on non-porous hydroxyl-ated silica or alumina, point B corresponds unambiguously to the beginning of themiddle linear region of the isotherm (in accordance with the original designation ofEmmett and Brunauer I), being located 2 at a pressure close to 0-1 PO. The positionwith krypton, argon (and probably also xenon) is complicated and far from clear. Onhydroxylated silica 3 it would appear that the monolayer and multilayer adsorptionprocesses overlap, and that the statistical monolayer is not directly associated with anindistinct point B. On the other hand, on metal surfaces, although the heat of adsorp-tion is comparatively high and likely to lead to the formation of a well-definedmonolayer, the full significance of a characteristic point B is uncertain. According tosome workers,4~ 5 the krypton monolayer is completed on certain clean metals (Ni andCu) at pressures -0.1 PO, and is identified as such by a good point B ; others,6-*however, report a sharp point B with these adsorption systems at much lower pressures( < 0.01 PO). In seeking an explanation for these apparently conflicting results, thereis the possibility of a localized monolayer being formed on certain sites at low pressure,becoming more highly compressed with increase in pressure.9 Also, on a nearlyuniform surface of graphitized carbon,lo or sintered nicke1,ll krypton gives a stepwiseisotherm and the step-height (or point of inflexion) rather than point B appears tocorrespond to the monolayer capacity.121 Emmett and Brunauer, J. Amer. Chem. SOC., 1937, 59, 1553.2 Sing, Chem. Ind., 1964, 321.3 Sing and Swallow, Proc. Brit. Ceram. SOC., 1965, 39.4 Klemperer and Stone, Proc. Roy. SOC. A, 1957, 243, 375.5 Kington and Holmes, Trans. Furaday SOC., 1953, 49, 417.6 Roberts, Trans. Faraday SOC., 1960, 56, 128.7 Anderson and Baker, J. Physic. Chem., 1962, 66, 482.8 Brennan, Graham and Hayes, Nature, 1963, 199, 1152.9 Pierce and Ewing, J. Amer. Chem. SOC., 1962, 84, 4070.10 Amberg, Spencer and Beebe, Can. J. Chem., 1955, 33, 305.11 Fox and Katz, J. Physic. Chem., 1961, 65, 1045.12 Prenzlow and Halsey, J. Physic. Chem., 1957, 61, 1158