Sei sulla pagina 1di 8

Temperature programmed desorption: A statistical rate theory approach

J. A. W. Elliotta) and C. A. Wardb)


Thermodynamics and Kinetics Laboratory, Department of Mechanical Engineering, University of Toronto, 5 Kings College Road, Toronto, Canada M5S 3G8

Received 20 September 1996; accepted 17 December 1996 The equation traditionally used to interpret temperature programmed desorption TPD spectra, the PolanyiWigner equation, does not contain explicitly the coverage and temperature dependence necessary to predict TPD spectra in several important systems including CONi 111 . Herein, the statistical rate theory SRT approach is used to formulate equations for temperature programmed desorption which are then used to examine TPD spectra reported in the literature for CONi 111 . The molecular and material properties for the CONi 111 system have been previously established. One experimental spectrum has been chosen to determine the apparatus constants. The material properties and the apparatus constants are then used in the SRT equations to predict the eight additional TPD spectra for different initial coverages. A critical comparison can then be made between the theory and these eight experimental spectra, since no tting constants were used in these eight cases. The results show that there is clearly qualitative agreement. The SRT equations are then used along with the heat of adsorption to derive an equation for the pre-exponential factor appearing in the PolanyiWigner equation. A prediction is made for the pre-exponential factor that is in agreement with that found empirically. The agreement found between the SRT predictions and the measured spectra indicates that all of the coverage and temperature dependence necessary to predict TPD spectra is given explicitly by the SRT approach. Hence, the experimental support for the SRT approach is enhanced. The SRT equations are then used to predict CONi 111 spectra that would occur if the heating rate were varied. 1997 American Institute of Physics. S0021-9606 97 01212-9

I. INTRODUCTION

At present, temperature programmed desorption TPD spectra are interpreted using the PolanyiWigner1,2 equation for the desorption rate. It has been established3 that this equation, which is based on absolute rate theory,4 does not give explicitly the coverage and temperature dependence necessary to predict TPD spectra in several important systems including5 CONi 111 . Statistical rate theory SRT shows promise of overcoming this difculty since it has been shown to predict the coverage dependence of isothermal adsorption kinetics.610 Statistical rate theory was originally developed to allow the rate of molecular transport across the interface between macroscopic phases of an isolated system to be predicted in terms of the material properties of the two phases.69,1118 Recently, this approach was extended so that it could be applied to a nonisolated system that experienced a change in mass during the kinetic process.10 The extension was then examined by applying it to predict adsorption kinetics in CONi 111 beam-dosing experiments. It was found that the material properties governing adsorption kinetics could be tabulated independently of the kinetic experiments and that all of the coverage dependence in the adsorption kinetics was predicted by SRT. However, when the calculations were
a

compared with the measurements, constants that described the experimental apparatus had to be determined. The fact that these apparatus constants had to be inferred prevented the extension of SRT from being examined as rigorously as wished. Herein, SRT is extended further so that it may be applied to temperature programmed desorption and the resulting equations are used to examine TPD spectra reported by Miller et al.5 for CONi 111 . For the purposes of examining the theoretical developments, this type of experiment has the distinct advantage of being performed repeatedly in the same experimental apparatus, but with different initial surface coverages. As a result, the apparatus constants may be determined from one TPD spectrum and used to predict other spectra. This allows a critical comparison to be made between theory and experiments, since no tting parameters are used in these calculations.
II. THEORETICAL MODEL FOR TEMPERATURE PROGRAMMED DESORPTION

Present address: Department of Chemistry and Materials Engineering, Univ. Alberta, Edmonton, Canada T6G 2G6. b Author to whom correspondence should be addressed; telephone: 416 978-4807, facsimile: 416 978-7322, internet address: ward@me.utoronto.ca J. Chem. Phys. 106 (13), 1 April 1997

In a TPD experiment, a surface is rst exposed to a gas at a particular temperature in order to obtain a specic initial coverage. The surface is then heated in a controlled manner so that the gas desorbs. The gas pressure above the surface is monitored as the surface is heated. The competition between molecules entering the gas phase volume through desorption and leaving the volume through pumping of the experimental chamber creates a pressure spectrum with a characteristic shape. Consider the system shown schematically in Fig. 1 in which a diatomic gas phase is adsorbing under sub 1997 American Institute of Physics 5677

0021-9606/97/106(13)/5677/8/$10.00

Downloaded20Dec2000to128.100.49.167.RedistributionsubjecttoAIPcopyright,seehttp://ojps.aip.org/jcpo/jcpcpyrts.html.

5678

J. A. W. Elliott and C. A. Ward: Temperature programmed desorption TABLE I. Gas phase properties of carbon monoxide Ref. 19 .
g

rad /2
13

r e nm 0.1128

D 0 J/molecule 1.464 10
18

6.394 10

FIG. 1. Schematic diagram of the model system used to predict temperature programmed desorption spectra.

distance of the two atoms, and m 1 and m 2 are the masses of the two component atoms. The gas phase properties are known19 and are given in Table I. represent functions In Eq. 1 , the symbols , b, and that are related to the description of the adsorbed molecules.20 The adsorbed molecules have been treated as quantum mechanical, double harmonic oscillators with a coverage-dependent potential energy.20 There will be six characteristic vibration frequencies associated with the double harmonic oscillator. For CONi 111 , two of these frequencies have been measured using electron energy loss spectroscopy.21 The function is a function of temperature given by
2

monolayer conditions on a single crystal surface with one type of localized bond. We assume that the solid is large enough that it imposes its temperature on the system. This is the same type of system that has been treated isothermally, both as an isolated system68 and as a system with a varying number of molecules and therefore pressure .10 If the coverage is dened to be the number of adsorbed molecules per surface substrate atom, M to be the number of adsorption sites per surface substrate atom, and M to be the number of adsorption sites per unit area, then the SRT equation for the rate of change of coverage with time may be written10 d dt Pe M 2
M M e

exp
1

T
j

2kT
j

, 1

exp

kT

where 1 and 2 are the two characteristic vibration frequencies of an adsorbed molecule that have been experimentally resolved.21 The function b is related20 to the four unknown vibration frequencies and depends on temperature. b T b 0 b 1T b 2T 2 b 3T 3. 4 is related20 to the potential energy of an The function adsorbed molecule and is a function of coverage: c0 c1 c2
2

m 1 m 2 kT P exp b kT

c3

For the CONi 111 system, the coefcients in the functions have been found from measured equilibrium adb and sorption isotherms20 and are given in Table II along with other properties of COi 111 . The equilibrium adsorption cross section8 has been taken to be the surface area divided

exp

b kT

TABLE II. Properties of carbon monoxide adsorbed on Ni 111 . Property M 0 atoms/m M sites/m2 1 rad /2 2 rad /2 b 0 c 0 J/molecule b 1 J/molecule K b 2 J/molecule K2 b 3 J/molecule K3
2

Value 1.86 10 0.57M 0 1.23 1013 5.57 1013 1.4921 10 1.3627 10 2.5408 10 1.7016 10 5.6821 10 2.2721 10 4.8990 10
19

Source geometry Refs. 31 and 20 EELS Ref. 21 EELS Ref. 21


18 21 24

where k is the Boltzmann constant, T is the temperature, and P is the instantaneous pressure in the gas phase. The subscript e on a quantity indicates that it is to be evaluated at equilibrium. The symbol represents a function of temperature and gas phase properties given by 1 exp kT
g

kT
7/2

exp

D0 kT

2
1/2

3/2

27 20 19 19

2m 1 m 2 r 2 m 1 m 2 e

c 1 J/molecule c 2 J/molecule c 3 J/molecule

Ref. 20, obtained by tting the measured data of Refs. 24 and 34 with the procedure presented in Ref. 7.

where g is the characteristic vibration frequency of a gas molecule, D 0 is its dissociation energy, r e is the separation

J. Chem. Phys., Vol. 106, No. 13, 1 April 1997

Downloaded20Dec2000to128.100.49.167.RedistributionsubjecttoAIPcopyright,seehttp://ojps.aip.org/jcpo/jcpcpyrts.html.

J. A. W. Elliott and C. A. Ward: Temperature programmed desorption

5679

by the number of adsorption sites i.e., 1/M .10 Therefore, all of the properties appearing in Eq. 1 have been tabulated for the CONi 111 system.10,20 The application of Eq. 1 to predict the rate of adsorption in an isolated system is straightforward.68 The equilibrium pressure and coverage are found from the isotherm relation.20 P e T,
e e M e

where V is the volume of the gas phase. There will be pumping on the system at some prescribed rate so that is changing with time. From the ideal gas law and conservation of mass d dt rp , V 11

exp

b kT

where r p is the pumping speed volume/time . The pumping speed may be a function of pressure to account for inefciencies of the pumping system. For the results presented here a pumping speed of rp Rp 1 Pb , P 12

As with the beam-dosing adsorption experiments,10 TPD does not occur in an isolated system. Energy is being added to change the temperature of the system and gas molecules are being removed from the system by pumping. We will consider SRT to be valid for an instant in time for which the system has a known temperature, T, and a xed total number of gas molecules, N, divided between the gas phase and the adsorbed phase, denoted by superscripts g and , respectively, N Ng N . 7

One might imagine the system within the dotted line in Fig. 1 to be isolated at time t 0 with initial values of T and N. The system is then allowed to evolve according to SRT for a short period of time, say until t 1 . At t 1 the temperature of the system is changed and some gas molecules are removed. The system is then isolated again and allowed to evolve for another short period of time. In other words, the temperature is stepped in time according to a temperature program and at each step molecules are removed through pumping. As the time periods become innitesimal, Eq. 1 is then integrated along with equations describing the interaction of the system with its surroundings. The additional relations describing the interaction of the system with the surroundings will now be found. In a TPD experiment, energy is added to the system so that the temperature changes in a prescribed manner with time uniformly throughout the system . In most cases a linear temperature prole is used T T 0 R h t, 8

was used, where R p is the constant high pressure pumping speed and P b is the lowest attainable pressure of the vacuum system. Note that the pumping speed is constant unless the pressure is near P b . Next, the equilibrium properties must be evaluated for this system which is not isolated and thus does not have an equilibrium condition. At a given instant in time, the system will have particular values of T and N. Analogously to the treatment in Ref. 10, the equilibrium properties of the system at the given instant are taken to be those to which the system would evolve if it were isolated with the particular values of T and N and an assigned volume V. Because the solid is assumed to be large enough that it imposes its temperature on the system, the equilibrium value of T at each instant is equal to the instantaneous value. In order to obtain the equilibrium pressure and coverage at each step from T and N, we apply the ideal gas law at equilibrium Pe
e

kTM 0

A . V

13

The two equations for the equilibrium pressure, Eqs. 6 and 13 , may be equated to give an equation that may be solved numerically for the equilibrium amount adsorbed, e
e M e

exp

b kT

e e

kTM 0

A . V 14

where R h is the heating rate and t is the time since the beginning of the temperature ramp. To simplify the notation, we write the total number of molecules in a nondimensional format equivalent to coverage N , AM 0 9

where A is the area of the adsorbed phase and M 0 is the number of surface substrate atoms per unit area. Then, combining Eq. 7 with Eq. 9 , the denition of coverage and the ideal gas equation one gets P kTM 0 A , V 10

Once e is known, P e can be found from Eq. 13 . Equations 1 , 8 , 10 , 11 , 13 , and 14 can now be solved simultaneously for the six unknown variables, , T, P, , P e , and e as functions of time. Since this set of equations includes two ordinary, rst-order differential equations, two initial conditions are needed. These would normally be the initial pressure and the initial coverage, P 0 and 0 , respectively. Other variables that need to be recorded for a specic experiment are R h , T 0 , A, V, and R p . If pumping efciency is to be accounted for then P b must also be known. In this study, the solution procedure is as follows. Equations 8 , 10 , and 13 are substituted into Eq. 1 . Equations 1 and 11 are then solved simultaneously using a

J. Chem. Phys., Vol. 106, No. 13, 1 April 1997

Downloaded20Dec2000to128.100.49.167.RedistributionsubjecttoAIPcopyright,seehttp://ojps.aip.org/jcpo/jcpcpyrts.html.

5680

J. A. W. Elliott and C. A. Ward: Temperature programmed desorption

CashKarp RungeKutta routine22 with numerical root-nding23 of Eq. 14 at each function evaluation within the time steps i.e., six times per CashKarp RungeKutta step .

III. DETERMINATION OF THE APPARATUS CONSTANTS FROM A CONi(111) SPECTRUM MEASURED BY MILLER et al.

Temperature programmed desorption spectra were carefully measured by Miller et al. for the CONi 111 system. All of the experimental conditions needed to make a calculation were given in Ref. 5 except for the system volume and the pumping speed. Thus one of the reported spectra that corresponding to the highest initial coverage was selected and used to infer these apparatus constants. The initial pressure for each experiment was assumed to be the lowest attainable pressure for the system which was given in Ref. 5. The pumping speed was assumed to drop off at low pressures according to Eq. 12 . Miller et al. listed initial coverages which they found by integrating the desorption spectra and comparing the result to that of a desorption spectra taken at the saturation coverage. By saturation coverage, it is meant the coverage that would exist at equilibrium at the pressure and temperature at which the dosing exposure was performed. Miller et al. assumed the saturation coverage was 0.5. However, since the saturation coverage is more likely higher than this,24 this assumption has been re-examined and the saturation coverage was inferred from the spectrum corresponding to the highest initial coverage. The nominal initial coverages listed by Miller et al. were then multiplied by the ratio of the inferred saturation coverage to the saturation coverage assumed by Miller et al. Since in Ref. 5, a signal that is proportional to pressure is reported, rather than pressure itself, the pressure scale must also be determined from the spectrum at the highest initial coverage. To establish this scale, the lowest measured pressure for that spectrum was assigned the value of P b and the maximum pressure the peak pressure was assigned a value equal to the peak pressure predicted by SRT for the highest initial coverage. The apparatus constants were taken as those which minimized the following error E 1
p Pc Pb n i 1

FIG. 2. The experimental data shown is from Fig. 4 of Ref. 5. The initial coverage was 0.54. The solid line shows the statistical rate theory prediction using the best values of pumping speed, volume and the pressure scale.

P mi P ci n d

2 p p W Tc Tm .

15

p In the above equation, P c is the calculated peak pressure, P mi is the measured pressure for the ith data point, n is the number of data points, d is the number of parameters being p inferred from the data, W is a weighting factor, T c is the calculated peak temperature or temperature at which the p peak pressure occurs and T m is the measured peak temperature. Minimizing the above error is equivalent to nding the apparatus constants which predict a TPD spectra with a shape closest to the experimental shape with the additional requirement that the predicted peak temperature be the same

as the measured peak temperature. The weighting factor, W, was adjusted until the measured and predicted peak temperatures were in agreement to within 0.1 K. In Fig. 2, the experimental spectrum taken at the highest initial coverage is shown along with the SRT prediction using the best values for the unknown experimental conditions. Note that although SRT predicts the existence of the low temperature shoulder seen in the experimental spectrum, the shoulder is only in qualitative agreement. A lack of agreement at high coverages is expected since it is likely that the shoulder is due to the second type of bonding linear which occurs under this condition25 and only one type of bonding bridge has been explicitly included in the theoretical expression for the chemical potential. The assumptions made in determining the properties20 for CONi 111 amount to treating the small proportion of linear molecules as though they were bridge bonded. Recall20 that the interactions were included by nding the coverage dependence of the potential energy of bound molecules empirically so that the effect of the linearly bonded molecules on the potential energy has been included. Thus, one would expect this model to be more accurate than neglecting the linearly bonded molecules completely, but not as accurate as a full treatment of the heterogeneous bonding. The values inferred for the volume, pumping speed and saturation coverage were 0.72 l , 2.7 l /s, and 0.54, respectively. The value for the saturation coverage is quite reasonable. Using the isotherm relation,20 this coverage may be shown to correspond to the equilibrium coverage at a temperature of 300 K and a pressure of 8.7 10 8 Torr which are reasonable dosing conditions in this type of system. It is not possible to assess whether or not the values for the volume and pumping speed are reasonable since the exact experimental conditions are not known. For example, the rate of pumping in different parts of a large apparatus with a number of obstructions would be different to the rated value of the pump. Also, in the equations these constants are multiplied by other constants that may have errors for example area, A, only occurs in a ratio with V . However, volume and pumping speed are physical parameters that are expected to be constant both as a kinetic experiment proceeds and from experiment to experiment. Thus, we may use the inferred

J. Chem. Phys., Vol. 106, No. 13, 1 April 1997

Downloaded20Dec2000to128.100.49.167.RedistributionsubjecttoAIPcopyright,seehttp://ojps.aip.org/jcpo/jcpcpyrts.html.

J. A. W. Elliott and C. A. Ward: Temperature programmed desorption TABLE III. Experimental conditions used for predictions. Variable Initial coverage, 0 Value 0.0151 0.54 Source nominal values given by Miller et al. Ref. 5 multiplied by 0.54/0.5 nominal value given by Miller et al. Ref. 5 . nominal crystal dimensions given by Miller et al. Ref. 5 . nominal value given by Miller et al. Ref. 5 . nominal value given by Miller et al. Ref. 5 . inferred from spectrum with an initial coverage of 0.54 inferred from spectrum with an initial coverage of 0.54

5681

Heating rate, R h Area, A

6 K/s 2.83 10
5

m2

Initial temperature, T 0 Lowest pressure, P b Volume, V

260 K 5 10 0.72 l
9

Pa

Pumping speed, R p

2.7 l /s

FIG. 4. Predicted and measured peak temperatures. The error bars on the data points represent the error in reading the temperatures from the plots reported in Fig. 4 of Ref. 5. The peak temperature of the spectrum used to determine the unknown apparatus constants is denoted by an open circle. The solid line is the predicted peak temperature from SRT.

values to examine the other spectra recorded by Miller et al.5 to test SRT. The experimental conditions used for the SRT predictions of TPD spectra are summarized in Table III.
IV. COMPARISON OF PREDICTED TPD SPECTRA WITH THE OTHER CONi(111) MEASUREMENTS MADE BY MILLER et al.

Using the conditions shown in Table III, a complete set of spectra were calculated for a variety of initial coverages. The predicted spectra are shown in Fig. 3. Qualitatively, these spectra are in good agreement with the measurements made by Miller et al.5 and others. See, for example, Ref. 26. Specically, the temperature at which the pressure maximum occurs the peak temperature shifts to lower temperatures with increasing initial coverage. At the higher initial coverages, a shoulder appears in the spectra on the low temperature side. In Fig. 4, the predicted peak temperatures the solid line are compared with the experimental peak temperatures for

various initial coverages. The error bars represent the error in reading the peak temperatures from the plots reported by Miller et al.5 Note the expanded temperature scale in Fig. 4. The measured peak temperature corresponding to an initial coverage of 0.54 is shown as an open circle since this spectrum was used to infer experimental conditions. The rest of the theoretical line, however, represents a true prediction of the peak shift. The trend in the peak temperature is well predicted by the SRT equations. In Fig. 5, the predicted peak pressures the solid line are compared with the experimental peak pressures for various initial coverages. The measured peak pressure corresponding to an initial coverage of 0.54 is shown as an open circle since this spectrum was used to infer experimental conditions. The drop in peak pressure as a result of lowering the initial coverage is accurately predicted by SRT. In Fig. 6, predicted and measured spectra are shown for initial coverages of 0.0151, 0.184, 0.324, and 0.518. The spectra have been displaced along the ordinate for clarity. Note that each of the curves shown in Fig. 6 is a prediction with no tting parameters. In all of the predicted spectra, the

FIG. 3. Predicted temperature programmed desorption spectra. The initial coverages used were: a 0.0151, b 0.0346, c 0.0702, d 0.108, e 0.14, f 0.184, g 0.248, h 0.324, i 0.41, j 0.464, k 0.518, and l 0.54.

FIG. 5. Predicted and measured peak pressures. The measurements are those reported in Fig. 4 of Ref. 5 except the unreported pressure scale has been obtained as indicated in Fig. 2. The peak pressure of the spectrum used to determine the unknown pressure scale and other apparatus constants is denoted by the open circle. The solid line is the peak pressure predicted from SRT.

J. Chem. Phys., Vol. 106, No. 13, 1 April 1997

Downloaded20Dec2000to128.100.49.167.RedistributionsubjecttoAIPcopyright,seehttp://ojps.aip.org/jcpo/jcpcpyrts.html.

5682

J. A. W. Elliott and C. A. Ward: Temperature programmed desorption

dicted spectra for an initial coverage of 0.324 and heating rates of 2, 6, and 12 degrees per second are shown. Although there is no experimental data with which to quantitatively compare these predictions, qualitatively the behavior is similar to that found for CO on Ir 110 when the heating rate is varied.27 In particular, for the high temperature peak for COIr 110 , increasing the initial coverage shifts the peak to lower temperatures and increases the peak pressure while maintaining a common high temperature edge which is similar to the behavior shown in Fig. 3. When the heating rate was increased for the COIr 110 system, the peak pressure increased and the high temperature peak shifted to higher temperatures with the high temperature edges being distinct which is similar to the behavior shown in Fig. 7.
VI. COMPARISON OF THE STATISTICAL RATE THEORY APPROACH WITH THE POLANYIWIGNER APPROACH

FIG. 6. Comparison of predicted and experimental TPD spectra for different initial coverages. The experimental spectra are those reported in Fig. 4 of Ref. 5, except that the unreported pressure scale was obtained as indicated in Fig. 2. These spectra have been displaced along the ordinate for clarity. The solid lines are the predictions from SRT. For each case, there were no tting parameters used in making the predictions.

Another way to examine the theoretical approach presented herein is to use SRT to derive an equation for the pre-exponential factor that is inferred when the Polanyi Wigner approach is used. Predictions made using the SRT equation may then be compared with empirical observations.
A. The deciencies of the PolanyiWigner approach

pressure on the high temperature side falls off more rapidly than was measured. This may be the effect of real pumping inefciencies not included in the model. One would otherwise expect the pressure to return to the initial value.
V. PREDICTION OF TPD SPECTRA FOR VARIOUS HEATING RATES

Traditionally, TPD spectra are interpreted using an Arrhenius28 or PolanyiWigner1,2 equation for the desorption rate, dN dt
d

exp

Ed , kT

16

Rather than varying the initial coverage in order to obtain different spectra, an experimental approach that has been used27 for other gassolid systems is to vary the heating rate while keeping the initial coverage constant. In Fig. 7, pre-

FIG. 7. Predicted TPD spectra for an initial coverage of 0.324 and for the different heating rates shown.

where x is the order of the reaction. The above postulated relation is based on absolute rate theory.4,29 Since one of the basic assumptions of absolute rate theory is that equilibrium exist between the adsorbing species in the form of an activated complex and the substrate, the effectiveness of this theory in predicting nonequilibrium effects has often been questioned. When absolute rate theory is applied to adsorption kinetics, the usual procedure is to lump these effects into the pre-exponential factor, d , which must be determined empirically.30 In fact, the desorption energy, E d , appearing in Eq. 16 is dened as that which is inferred when the PolanyiWigner equation is compared with experiment and thus its physical signicance is not immediately clear. In general, both the desorption energy and the pre-exponential factor may depend on coverage and/or temperature.3 They are difcult to infer from experimental desorption spectra without rst making some assumption about their dependencies.1,2 Even once inferred, these quantities are difcult to interpret or use since they are not related to fundamental physical properties. For example, Eq. 16 does not strictly apply to desorption through a precursor intermediate.3 Precursor states have been used in an attempt to explain additional coverage dependence introduced in the rate equations.32,33 In this case one must either assume coverage and temperature dependencies and realize that the kinetic parameters being

J. Chem. Phys., Vol. 106, No. 13, 1 April 1997

Downloaded20Dec2000to128.100.49.167.RedistributionsubjecttoAIPcopyright,seehttp://ojps.aip.org/jcpo/jcpcpyrts.html.

J. A. W. Elliott and C. A. Ward: Temperature programmed desorption

5683

inferred are effective parameters,3 or one must attempt to derive these dependencies from a precursor model. The CO Ni 111 system is one that has been described by a precursor model.31 A rate equation which includes the complete coverage and temperature dependence of all of the kinetic parameters i.e., without having to nd such dependencies empirically from kinetic experiments has not been previously found.

,T

Pe P

M M

T ,

20

where T 1 M 2 exp 7/2 m 1 m 2 kT


g /kT

B. The pre-exponential factor from the statistical rate theory approach

exp
1 /kT

kT

D0 kT

2kT 1 db k dt

2kT

In order to compare the PolanyiWigner equation with that obtained from SRT, consider only the desorption portion d dt Pe M 2
M e M

2 /kT

exp 1 kT

kT Hs . N

exp

kT

m 1 m 2 kT

exp

b kT

. 17

21

Note that since the adsorption rate has been neglected, the above equation and also the PolanyiWigner equation are not valid when the system is very near equilibrium. In order to compare Eq. 17 with Eq. 16 , an assumption must be made about the functional dependence of either the preexponential factor, d , or the desorption energy, E d , so that the effects of these two parameters may be unambiguously separated. Here we will assume that the desorption energy is known and write an explicit expression for the preexponential factor since this factor has a physical meaning that is less obvious. Hence, equating Eqs. 16 and 17 , assuming rst order reaction kinetics5 and solving for d yields 1
d

M 2

m1

Pe m 2 kT P

exp
M M e

Ed kT b exp kT

. 18

Next we assume that the desorption energy is simply the heat of adsorption. A relation for the heat of adsorption for CONi 111 has recently been developed20 qA 7 kT 2
g

D0 1
2

exp
1

2 db dT

kT

b 1

exp

kT

exp

kT

Note that Eq. 20 has been divided into two parts: A temperature-dependent part, , and a coverage-dependent part, the factor that multiplies . The factor is a ratio of equilibrium parameters to nonequilibrium, instantaneous parameters. Historically, it has been understood that one reason that coverage dependence had to be included in the preexponential factor was to account for nonequilibrium effects.30 Eq. 20 gives a theoretical basis for this statement. The only coverage dependence in Eq. 20 appears in a factor which is a ratio of equilibrium to nonequilibrium, instantaneous parameters. This factor would be unity if the system were at equilibrium. This factor also gives insight into why, when the pre-exponential factor is inferred from nonequilibrium experiments, its coverage dependence is found to depend on how the experiment is analyzed and/or performed.5 The signicance of Eq. 20 can be further understood by examining its low coverage limit. In the limit of zero coverage, the factor in Eq. 20 reduces to unity and the equilibrium and instantaneous pressures become equal, leaving d which is a function of temperature and properties which have all been tabulated for CONi 111 .10,20 Thus, without using any information from TPD experiments, the zero coverage pre-exponential factor may be calculated and compared with that inferred empirically from TPD spectra. The value calculated from Eq. 21 is ln d 35.6 3.6 in units of ln s 1 . The largest contribution to the error comes from the error in determining H s / N from calorimetric measurements. The value found empirically5 is ln d 33.0 2.5 which is in agreement, to within the stated errors, with the value predicted by Eq. 21 .
VII. CONCLUSIONS

Hs , N

19

where the last term represents the enthalpy change of the solid per molecule adsorbing. A value for this enthalpy change at 300 K has been inferred20 from the measured calorimetric heat of adsorption. Substituting Eq. 19 into Eq. 18 and simplifying gives

Fundamental equations for predicting TPD spectra have been developed using the SRT approach. The equations found have been tested by comparing the theoretical predictions with measured TPD spectra for CONi 111 . The material properties governing the adsorption of CO on Ni 111 have previously been established over a wide range of cov-

J. Chem. Phys., Vol. 106, No. 13, 1 April 1997

Downloaded20Dec2000to128.100.49.167.RedistributionsubjecttoAIPcopyright,seehttp://ojps.aip.org/jcpo/jcpcpyrts.html.

5684

J. A. W. Elliott and C. A. Ward: Temperature programmed desorption D. A. King, Surf. Sci. 47, 384 1975 . A. M. de Jong and J. W. Niemantsverdriet, Surf. Sci. 233, 355 1990 . 3 X. Guo and J. T. Yates, Jr., J. Chem. Phys. 90, 6761 1989 . 4 A. Clark, The Theory of Adsorption and Catalysis Academic, New York, 1970 , p. 210. 5 J. B. Miller, H. R. Siddiqui, S. M. Gates, J. N. Russell, Jr., J. T. Yates, Jr., J. C. Tully, and M. J. Cardillo, J. Chem. Phys. 87, 6725 1987 . 6 C. A. Ward and M. Elmoselhi, Surf. Sci. 176, 457 1986 . 7 J. A. W. Elliott and C. A. Ward, Dynamics of Gas Adsorption on Heterogeneous Solid Surfaces, edited by W. Rudzinski, W. A. Steele, and G. Zgrablich Elsevier, Amsterdam, 1997 . 8 C. A. Ward and R. D. Findlay, J. Chem. Phys. 76, 5615 1982 . 9 R. D. Findlay and C. A. Ward, J. Chem. Phys. 76, 5624 1982 . 10 J. A. W. Elliott and C. A. Ward, J. Chem. Phys. 106, 5667 1997 , preceding paper. 11 C. A. Ward, J. Chem. Phys. 67, 229 1977 . 12 C. A. Ward, J. Chem. Phys. 79, 5605 1983 . 13 P. Tikuisis and C. A. Ward, in Transport Processes in Bubbles, Drops and Particles, edited by R. Chhabra and D. DeKee Hemisphere, New York, 1992 pp. 114132. 14 C. A. Ward, M. Rizk, and A. S. Tucker, J. Chem. Phys. 76, 5606 1982 . 15 C. A. Ward, P. Tikuisis, and A. S. Tucker, J. Colloid Interface Sci. 113, 388 1986 . 16 C. A. Ward, B. Farabakhsh, and R. D. Venter, Z. Phys. Chem. N.F. Bd 147, S.89101, 727 1986 . 17 F. K. Skinner, C. A. Ward, and B. L. Bardakjian, Biophys. J. 65, 618 1993 . 18 C. A. Ward, R. D. Findlay, and M. Rizk, J. Chem. Phys. 76, 5599 1982 . 19 T. L. Hill, An Introduction to Statistical Thermodynamics Dover, New York, 1986 , p. 153. 20 J. A. W. Elliott and C. A. Ward, Langmuir unpublished . 21 W. Erley, H. Wagner, and H. Ibach, Surf. Sci. 80, 612 1979 . 22 W. H. Press, S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery, Numerical Recipes in C The Art of Scientic Computing, 2nd ed. Cambridge University Press, Cambridge, 1992 , p. 707. 23 W. H. Press, S. A. Teukolsky, W. T. Vetterling, and B. P. Flannery, Numerical Recipes in C The Art of Scientic Computing, 2nd ed. Cambridge University Press, Cambridge, 1992 , p. 362. 24 K. Christmann, O. Schober, and G. Ertl, J. Chem. Phys. 60, 4719 1974 . 25 L. Surnev, Z. Xu, and J. T. Yates, Jr., Surf. Sci. 201, 1 1988 . 26 F. P. Netzer and T. E. Madey, J. Chem. Phys. 76, 710 1982 . 27 J. L. Taylor and W. H. Weinberg, Surf. Sci. 78, 259 1978 . 28 D. P. Woodruff and T. A. Delchar, Modern Techniques of Surface Science, 2nd ed. Cambridge University Press, Cambridge, 1994 , p. 360. 29 K. J. Laidler, Theories of Chemical Reaction Rates McGraw-Hill, New York, 1969 , p. 45. 30 A. Zangwill, Physics at Surfaces Cambridge University Press, Cambridge, 1988 , p. 363. 31 H. Froitzheim and U. Kohler, Surf. Sci. 188, 70 1987 . 32 R. Gorte and L. D. Schmidt, Surf. Sci. 76, 559 1978 . 33 P. Kisliuk, J. Phys. Chem. Solids 3, 95 1957 . 34 O. L. J. Gijzeman, M. M. J. van Zandvoort, F. Labohm, J. A. Vliegenthart, and G. Jongert, J. Chem. Soc., Faraday Trans. 2 80, 771 1984 .
2 1

erages and temperatures.10,20 Constants related to the experimental apparatus were inferred from the experimental spectrum corresponding to an initial coverage of 0.54. The apparatus constants and the tabulated material properties could then be used to make predictions of the spectra obtained at eight other initial coverages. Since no tting parameters were used to predict the other eight spectra, a critical comparison can be made between theory and experiment. The errors in the measurements reported in the literature are not known, so it is not clear that there is any measured disagreement between the measured spectra and those predicted. Clearly there is qualitative agreement in the shapes of the spectra shown in Fig. 6. In addition, the change in both height pressure and location temperature of the spectrum peaks with initial coverage were predicted accurately by SRT. See Figs. 4 and 5, respectively. In addition to assessing the shapes of predicted TPD spectra, a zero-coverage value was predicted for the pre-exponential factor appearing in the PolanyiWigner equation that was in agreement with that found empirically. The good agreement in each of the cases mentioned above indicates that all of the coverage and temperature dependence necessary to predict TPD spectra is given explicitly by the SRT approach. This is signicant because the equation traditionally used to examine TPD, the Polanyi Wigner equation, has been shown not to have adequate coverage and temperature dependence. The additional coverage and temperature dependence has either been added empirically5 or explained using the concept of precursor states.32,33 Since precursor states have not been introduced into the quantum mechanical description adopted in the SRT approach, it appears that precursor states do not necessarily play a major role in determining desorption kinetics in the CONi 111 system. In addition to providing a model for temperature programmed desorption, these results and those of Ref. 10 indicate that nonequilibrium thermodynamics in the SRT sense may be applied to systems that are neither isolated nor in steady state. Hence, the experimental support for the SRT approach is enhanced.
ACKNOWLEDGMENTS

This work was completed with the support of the Natural Sciences and Engineering Research Council of Canada and with an Ontario Graduate Scholarship.

J. Chem. Phys., Vol. 106, No. 13, 1 April 1997

Downloaded20Dec2000to128.100.49.167.RedistributionsubjecttoAIPcopyright,seehttp://ojps.aip.org/jcpo/jcpcpyrts.html.

Potrebbero piacerti anche