Thermodesorption studies on adsorbate-adsorbent interaction

SURFACE

SCIENCE 12 (1968) 221-246 o North-Holland

T~RMODESORPTION

Publishing

Co., Amsterdam

STUDIES ON ADSORBATE-ADSORBENT INTERACTION

V. I. YAKERSON,

V. V. ROZANOV

and A. M. RUBINSHTEIN

Zefinfky Institute of Organic Chemistry, U.S.S.R. Academy of Sciences, Moscow, U.S.S.R. Received 5 February 1968; revised manuscript

received 29 March 1968

A scheme is described for chromatographic thermodesorption apparatus and experimental conditions to obtain information on adsorbate-adsorbent interaction are presented. The effects of temperature conditions, amount of substance, and heating rate are discussed. It was shown that by varying various parameters, information can be obtained on the nature of form of adsorption, activation energy of desorption and on desorption phases. Thermodesorption equations were analyzed for reactions of first order and homogeneous surface. A thermodesorption equation was applied to heterogeneous surfaces with different distribution functions (uniform, exponential, hyperbolic, power). Multiple forms of adsorption and the irreversible nature of chemisorption were established by studying chromatographically the thermodesorption of methyl, and isopropyl alcohols and diisopropyl ether from alumina surface. It was established experimentally that it is possible to determine distribution function and isotherms for adsorption equilibrium on the basis of gas chromatographic investigation of the thermodesorption of benzene from the surface mixed nickei-aluminium oxide catalysts.

1. Introduction FIash-desorptionl), i.e. instantaneous heating by electric current of a metallic filament with an adsorbed fayer permits the study of different metals. The investigations on adsorbate-adsorbent interaction by flash-desorption methods have been confined to refractory metals: the use of non-metallic sorbents encounters difficulties in heat transfer in vacuum. Until now, metals such as tungsten and moIybdenum were mostly used as adsorbent, and simple gases, CO, 02, N2, H21-3), as adsorbates. The majority of solids whose surface properties are of interest in these investigations are either oxides of metals, metals on carriers, or a complex mixture of non-conducting substances; and thermodesorption studies with such a system in vacuum is practically impossible. On the other hand, adsorbates often have a complex chemical structure and can decompose on heating. Gas chromatographic methods of thermal desorption first proposed by Cvetanovic et a1.4*5),have great potentialities in thermodesorption studies. Substances are adsorbed and desorbed in the flow of a carrier gas with high thermal conductivity (helium or hydrogen). Chromatographic methods can 221

222

V.I. YAKERSON,

be successfully adsorbent effectively

V. V. ROZANOV

applied in thermodesorption

AND

A.M. RUBINSHTEIN

studies to investigate

systems, if there is a good procedure, determined by theoretical analysis.

adsorbate-

and the information

can be

2. Experimental 2.1. CHROMATOGRAPHIC ARRANGEMENT TO SEPARATE THERMODESORPTION PRODUCTS Fig. 1 shows the path of the carrier gas. Helium from cylinder 1 via reducer 2 enters column 3, which is filled with molecular sieve activated in helium at 350-1OO”C. Moisture traces are completely removed in column 5 with molecular sieves cooled by liquid nitrogen. If necessary, a cleaning system is included to remove completely oxygen, hydrocarbons and carbon monox-

Fig. I.

Diagram of equipment for thermodesorption

experiments.

ide. The carrier gas, having passed through fine control needle valve and pressure gauge 4, enters the first evaporator 7 through which the samples are introduced. The sample is carried by the gas on the surface of the sorbent in a quartz tube 8 fixed in furnace 9 provided with a thermostating unit and temperature programming 10. The carrier gas, having passed through the adsorbent layer, enters the first detector channel 11, then to the second evaporator 7, through which the sample can also be introduced. Then the gas enters gas-liquid column 12 for the separation of desorption products. Column 13 where reaction chromatography takes place6), is located after the gas-liquid column. The components which have passed through this column are delivered into the second detector channel. A trap cooled by liquid nitrogen is fixed at the outlet to collect the condensate for subsequent chromatographic analysis. A soap film flow meter 14 is fixed between the outlet

THERMODESORPTION

STUDIES

223

and the trap. The thermodesorption chromatogram is recorded on a recorder connected with an integrator to determine the area under the curves. The use of two channels for detecting, as was done in ref. 7, facilitated the quantitative comparison of the areas under the curves and the use of one detector for thermodesorption and analysis. Perhaps, it may be possible to use two detectors in series, one of which measures total desorption phase, while the other detects individual products. 2.2. SIMPLIFIEDARRANGEMENT

FOR THE THERMODESORPTION INVESTIGATIONS*

The carrier gas, after cleaning, enters the comparison channel via the evaporator into the tube with the adsorbent and then to the second detector channel. Thus, the substance is delivered to the detector without separation. The temperature (T,) at the maximum in the thermodesorption curve can be found by calculations of the heating rate which can be determined with the help of a calibrated plot and the chart speed of the recorder. However, if the desorbed substance has a comparatively low boiling point, then a thermostat is not necessary; the detector can be thermostated at room temperature. 3. Experimental procedure The nature and quantity of information depend not only on the instrument design, but also on how the experiment is conducted, including the changes in the following parameters : a) Adsorption temperature (Tads) and interval of temperature programming (temperature conditions); b) Heating rate (b); c) Amount of substance (V). 3.1. TEMPERATURE CONDITIONS 3.1.1.

Conventional

version of thermodesorption

a) Without decomposition of the substance: The usual method for gases, developed by Cvetanovic4* 5) consists in adsorbing different quantities of substance at room temperature with subsequent thermodesorption at different heating rates. Adsorption takes place only at 20°C when the activation energy of adsorption is low; otherwise adsorption would not be detected at all. b) With decomposition of the substance: The studies on thermodesorption in vacuum or in its chromatographic version, as a rule, are carried out with * See fig. 1.

224

V.I.YAKERSON,

V.V.ROZANOV

AND

A.M.RUBINSHTEIN

substances which do not decompose during temperature programming. From the point of view of catalysis studies, the particular case where the substance decomposes is of great interest in spite of the difficulties involved. On the one hand, the reaction can be determined (as one or a number of peaks of the reaction products) and on the other hand the form of adsorption can be determined for a given initial substance. 3.1.2. Partial thermodesorption a) Without lowering the temperature: In this procedure the substance is adsorbed first at a constant temperature higher than the room temperature; after the peak of excess substance has appeared and the base-line is established, the temperature programming system is switched on. This procedure is used when the adsorption temperature does not exceed lOO”C, and the maximum temperature is far from the adsorption temperature. Unlike the previous procedures, the change in Tads reveals the processes requiring definite activation energy of adsorption. b) With temperature lowering: After adsorption at a definite temperature, the temperature is reduced close to room temperature, then linear programming is switched on and thermodesorption is studied. This procedure enables us to find out the Tm-Tads relationship connected with surface heterogeneity (see below), and to separate the peaks which correspond to different forms of adsorption. After having determined the areas of the peak, it is possible to find out how T, depends on 0 (the surface coverage) and the distribution of the introduced substance according to the forms of adsorption. 3.1.3.

Thermodesorption

after saturation

activated adsorption

Thermodesorption, after the substance has been adsorbed at 2O”C, does not reveal the activated forms of adsorption, but thermodesorption after the substance has been adsorbed at higher temperatures does not permit by the nature of low temperature forms of adsorbtion to be judged. However, summary information can be obtained by combining both these methods. Repeated adsorption is conducted by introducing the sample in a series of steps where the temperature is reduced step by step by a definite interval starting from the higher temperature and the last sample is introduced at room temperature. As a result of this, conditions are created for the formation of adsorbed states which require a wide range of activation energies. The entire range of peaks corresponding to different forms of adsorption having different bond strengths with the surface will be observed in the thermodesorption chromatogram. These forms may take place with or without activation.

THERMODESORPTION

STUDIES

225

3.1 A. Step-b_wtep thermode~orpt~on

Superposition of peaks which is characteristic of a real thermodesorption chromatogram, is due to multiplicity of adsorption forms and surface heterogeneity. The second detector channel is used to analyze the desorption products, because of the appearance of so-called “back” peaks, i.e. in a direction opposite to the base line in the chromatogram. If, in spite of the precautions taken, the substance to be desorbed is found simultaneously in both channels, then the form of the elution curve is distorted and even false maxima may appear. Therefore, in order to avoid these hinderances, stepby-step thermodesorption is applied, i.e. the programm to increase the temperature is applied step by step with thermostating in between two increments. The end of each programming is selected slightly above T, of the appearance of the peak. As in ordinary experiments, the maxima of the peaks will be related to the appearance of the peaks corresponding to the forms of the adsorption, but will not be connected to artificial stopping of the increasing temperature; and this makes it possible to apply the thermodesorption equation. By varying the time interval between the “steps” of the experiment, it may be possible to study the question of “back” peaks, stability of T,, etc. And with step-by-step thermodesorption, we can qualitatively estimate the relation between different forms of adsorption, the peaks of which are not well separated in the thermodesorption chromatogram. 3.2. HEATING RATE By changing the heating rate 6, T, can be shifted along the linear dependence plot [IogbTz versus l/T (OK)] and the activation energy of the adsorption and pre-exponential factor can be calculatedl~~ys). Optimum range of heating rates in which the maximum shift of T, is observed (as will be shown below) lies below 25”C/min, consequently there is no necessity to make use of such high heating rates as used in flash methods. Moreover, due to the inertia of the system, it is desirable to measure T, with smaller rates rather than with higher rates of heating. However, by increasing the heating rate, it is possible to contract the width of the peak and to increase the height of the maximum, i.e. to reveal in the thermodesorption chromatogram the peaks corresponding to forms of adsorption which cover a small fraction of the surface. 3.3. SAMPLE QUANTITY The procedure of introducing gaseous samples was described by Cvetanovic and his collaborators*,5). Liquid can be introduced with sufficient accuracy with a microsgringe graduated in 0.06 ~1; however, it is advisable to admit

226

V.I. YAKERSON,

V. V. ROZANOV

AND A.M.

not less than 3 pl. The sample is introduced

RUBINSHTEIN

through

the evaporator,

where

it evaporates rapidly and is carried away by the carrier flow onto the sorbent. Depending on the adsorption temperature, Tads, the sample quantity V and the quantity g of the sorbent, the substance is adsorbed either totally or partially, and then the sorbent is determined by the detector. In the last case, peaks corresponding to weak forms of adsorption appear on the chromatogram much before programming. If peaks do not appear in the chromatogram after the substance is totally adsorbed, then there exists a linear dependence between surface coverage and sample quantity. And this allows the use of the sample quantity V, proportional to 0, or area (S) under the peak in revealing the nature of surface heterogeneity.

4. Results and their discussion 4.1. THEORETICAL ANALYSIS 4.1.1.

Thermodesorption of substance from a homogeneous surface (rejl 9)

In 1958, a theoretical analysis was made of the kinetics of desorption of gases from the surface of solids when the temperature increases linearly for reactions of the first order 8). For first-order reactions the following dependence on the temperature was found for the gas pressure which changes due to desorption from the surface P =

Eexp (-,J+;),

(1)

wherep is the gas pressure in the system, Z the frequency factor, C the evacuation constant in vacuum, EA the activation energy of desorption (in OK; Ed in kcal/mol), T the temperature (in "K), K=ZbEi, b is the heating rate (set/ OK) and

@‘d/T )’

1.

3! 2! l-E;~++(@j$j-...

J _ ex~(+‘,/T) [

The PC/Z versus T curve passes through the maximum. To find out the effect of various parameters and to transform eq. (1) to a form convenient for calculations, we can apply the condition of maximum:

dp

Z d exp [ - (KJ + &IT)]

dT, = C ~~~~~ from which we obtain

the fundamental

o

dT, expression

exp (EA/T,) EA = ZbT,' ,

in thermodesorption (2)

THERMODESORPTION

where 7’,‘,is the temperature

227

STUDIES

(in OK) at the maximum

of the curve. An analo-

gous expression was derived in ref. 10, but by a different method. Refs. 1, 8 and 10 partially discuss the effect of various parameters in eq. (2). Let us now examine the effect of Z, T,,, and b on EA and on each other. RedheadlO) discussed the relation of E,, (in kcal/mol) on T, for different values of b. However, since b is usually known and Z is assumed it is expedient to find out the relation between Ed and T,,, for different Z. Let us fix Z in the interval 109-1014 see-’ and b=8 sec/“K (in the optimum interval b = 3-17 set/ OK, Ed varies by 2%). The relation between Ed and T, is approximately given by the equation E, = KT,, (3) where Ed is expressed in kcal/mol, and T, in “K. The factor K is strongly dependent on Z, but weakly on T, (fig. 2). The relation of Kcp with 1ogZ is approximately linear. The difference in Ed for the extreme values is greater,

60

0

-150 -50

+50 +I50 +250 +350 +450

r,‘c

Fig. 2. Dependence of activation energy of desorption En (in kcallmol) on the maximum temperature T, (in “C) of the thermodesorption peak with different frequency factors Z. (1) Z= 109 set-l; (2) Z= lOlo see-I: (3) Z= IOllsec-l; (4) Z- 1012set-l; 15) Z= lOI set-I ; (6) Z = lOI set-l.

the higher is the temperature T, and at 500°C it is equal to 16 kcal/mol. The effect of Z is rather more significant than that of b, because Z can change by more than an order of five, but b changes at most by an order of two, and the optimum interval of b is less. The relationship between Ed and T, allows the determination of the various forms of adsorption. In fact, each maximum in the thermodesorption chromatogram determines a specific form of adsorption with respect to both energy (Ed) and quantitative aspect (area under curve). Since T, is determined with an accuracy of + 1 “C, Ed can be deter-

228

V.I. YAKERSON,

V. V. ROZANOV

AND A.M.

RUBINSHTEIN

mined to an accuracy of 0.1 kcal/mol; and when two different forms of adsorption are distinguished only by 3.5 kcal/mol by activation energy, then the position of T’,‘,will shift by 50°C; in an actual experiment, this difference may perhaps be still smaller. Hence we can conclude that those very differences which are concealed in a classical adsorption experiment can be revealed by thermodesorption experiments. In order to find out the dependence of T, on b, and at the same time to determine Ed and Z, eq. (2) should be expressed in logarithms. If we take Ed/T, = 35 and Z= 10i3, we obtain after simple transformations log T,, = 4.8 + 0.5 (log&

- log b),

where the numerical coefficients depend on Z and the ratio Ed/T,. The relation between log T, and logb is linear for some specific Ed (ref. 11). Let us now consider how T,,, changes depending on the heating rate b (fig. 3). It is clear that the greater the heating rate, the smaller the shift of T, in the high temperature region (the form of the curve is typical for any

43

410 cl Fig. 3.

Dependence of maximum temperature of thermodesorption peak T, (in “C) on heating rate 6’ (in Tjmin) at Ed = 50 kcal/mol and 2 = 1013set-I.

values of Ed and Z). The greatest changes in T, occur at relatively low heating rates (up to 20”C/min). When the values of b are great, we apparently find that T, is independent of b, and there is no possibility of determining Ed and Z. Since the interval of change of T,,, contracts as Ed decreases, the smaller is T,, the greater should be the interval of change in heating rates for a simultaneous determination of Ed and Z.

THERMODESORPTION

4.1.2.

229

STUDIES

Thermodesorption qf substance from a heterogeneous surface

Until now we have examined the thermodesorption from a homogeneous surface of a solid according to the first-order reaction, when the activation energy of desorption (Ed) does not depend on the coverage, i.e. all surface sites are energetically equivalent. Let us now consider different cases of heterogeneous surfaces assuming, as in the general theory of adsorption on heterogeneous surfaces12,13), that Ed but not 2 varies with the coverage, i.e. the adsorption entropy is the same for different sites of the surface. In the end we shall examine the possibilities of simultaneous changes in Ed and Z depending on the coverage with which they are connected by the relation of the compensation effect. 1) Uniform distribution and logarithmic isotherm of adsorption equilibrium By making the activation changes (E&

use of the linear relation (Brensted-Polanyi-Temkin) for energy of desorption EA (in OK) (ref. 13), let us write the and (E&, with respect to the coverages 8, and f12 as follows: (E&,

= (EA)o - We,

(E&)0, = (E& 0, > 01,

(E&I

(4a) (4b)

- PC& ’ (E&z 9

where C is a constant, and the factor p indicates the fraction of change in activation energy of desorption with respect to the changes in heat of adsorption; (Ed),, is the activation energy of desorption at 0=0. Let us now write the fundamental equation in thermodesorption

exp(9 and taking

the logarithms,

(E& = bZT,‘, m

we have

(E& ~+ln(E&=lnbZ+2lnT,, T, that is (E& = T, [ln bZ + 2 In T, - In (E&l, and with the help of (3) we get (E& = T, (In bZ + In T, - In K), where K, as was shown in ref. 9, weakly depends For coverages 8, and 8,:

(5)

on T,.

(E&,

= T,, (In bZ + In T,, - In K,) ,

(54

(E&

= T,, (In bZ + In T,, - In KJ .

(5b)

230

V.I. YAKERSON,

Combining

V. V. ROZANOV

AND

A.M.

RUBINSHTEIN

eqs. (4a) and (5a), (4b) and (5b), we get (_I?& - pC8, = T,, (In bZ + In T,, - In K,),

(64

(E&

(6b)

- PCB, = Tmz(In bZ + In T,, - In K2).

It can be shown that at T, the quantities from the equality

within brackets

are equal. In fact,

lnbZ+lnT,,-lnK,=lnbZ+lnTm2-lnK,, it follows that ln (L,IKr)

= ln

(T,,I&) .

(7)

Using the data of ref. 9 and fig. 2, we can show that if TmI and T,,,, are separated by 500°C then at Z= 10r4-10’ set-r, eq. (7) is accurate up to 5%. By subtracting eq. (6b) from eq. (6a) and taking into account the relation (7) we obtain: (8)

DC(& - 0,) = T,, - T,,.

This equation is more accurate and valid than the equation derived earlier in ref. 14. The isobar of adsorption equilibrium in this case has the following formra) t’zf&,,,-C’T,

(9)

where C’ is a constant, T the temperature. Substituting eq. (9) in the relations (4a) and (4b), and bearing in mind that T is the temperature at which adsorption takes place in the thermodesorption experiment (Tad,), we get 3

(104

- bc&nax + PCC’T,, ads.

(lob)

(EL&, = (J%)o - flC&,,,, + PCC’T,, &h, From

= (E&

eqs. (5a), (lOa), (5b) and (lob), Cc’(T~,ads

-

We can also derive the relations eqs. (lOa) and (lob) the activation relation (3). Consequently, when therm of adsorption equilibrium

ads

and with the help of (7), we find

Tz,ads)

=

Tm,

-

Tm,.

(8) and (11) by the energy is connected the distribution is is logarithmic, T,

(11)

following method: In with T, by the linear uniform and the isoand Tads are linearly

connected. As was shown earlier, the surface coverage is determined by the area under the desorption curve, which in turn is linearly connected with the quantity (V) of the sample introduced only if the peak of excess substance does not appear until1 the moment of programming. Therefore it follows from eqs. (4) and (3) that T, will drop linearly as S or Vincreases. Thus, we obtain the plot for the uniform distribution and for the logarithmic isotherm of adsorption equilibrium which is given in fig. 4.

THERMODESORPTION

231

STUDIES

2) Exponentially heterogeneous surface a) Exponential distribution and power isotherm of adsorption equilibrium. In this case the linear dependence between activation energy and 8,, O2 takes the following formis) (-CJ)B~= (E&

- Pn ln 0, 9

(124

KJB~ = (Gh

- fin In 0, .

U2b)

TmI

!.YL s (V)

Fig. 4.

Subtituting

Subtracting

Tm versus Tads, and Tm versus S(V).

(12a) and (12b) in (5a) and (5b) respectively,

we obtain

(E&

- j3n In fJ1 = T’,, (In bZ + In T,, - In K,),

(134

(E&

- j?n In t& = 7& (In bZ + In T,, - In K2).

Wb)

(13b) from (13a) and with the help of (7) we find n ln(t?,/O,)

= T,, - Tml.

(14)

Eq. (14) expresses the relation between T, and 8 more accurately expression derived earlier (in ref. 14). The isobar of the adsorption rium has the following formis) 8 = O,,,,, exp (By substituting

From

this expression

Cl/T).

than the equilib-

(15)

in eqs. (12a) and (12b) (here T= Tad,), we get

(E&,

= (E:)o - bn ln %,,, + n C’ Ti,

ads 9

(164

(E&z

=

ads.

Wb)

(%)o

-

b

In kax

+

nC’T,,

(5a), (5b), (16a), (16b), and with the help of (7), we find BnC; (T,, ads - q, ads) = T,, - T,z.

(17)

Thus, T, grows linearly with Tads, logs and log V (the last two quantities are proportional to loge). Hence the plot looks as given in fig. 5.

Fig. 5.

Tm versus

Tads

and Tm versus logs (log y).

V.I. YAKERSON,

232

V. V. ROZANOV

AND

A.M.

RUBINSHTEIN

b) Exponential distribution and negatively powered isotherm of adsorption The linear dependenceIs) of Ed on 8, and 8, is as follows:

equilibrium.

(EL&, = (E&

+ C” In (1 - 0,))

(E&2 = (E&

+ C” In (1 - 0,).

By substituting in the above equations eqs. (5a) and (5b), and then subtracting C”[ln(l By making

the values of (E&, and (E&, eq. (18b) from (18a), we get

- 0,) - ln(1 - e,)] = T,, - Tmz.

use of the expression

for the isobars

of adsorption

(19)

equilibrium

(1 - 0) = (1 - %,,,) exp(C’P’), where T= Tads, let us substitute

from

13) (20)

the values of 8 from (20) in (18a) and (18b):

(E&

= (E&), + C” In (1 - e,,,)

+ C’C”TI, ads,

(214

(E&

= (E&),, + C” In (1 - O,,,) + C’C”T2, ads.

@lb)

By substituting the values of (E&, and (E,& from eqs. (5a) and (5b) in eqs. (21a) and (2lb), and then subtracting the transformed eq. (21b) from the transformed eq. (21a) and making use of relation (7), we get C’C”(T,ads

-

The graphical representations of 0 we make use of S or V) for the powered isotherm of adsorption Here Sre, is the ratio of the area coverage; Vrel is the ratio of the sponding

to monolayer

Gads)

=

L,

-

(22)

Tmz.

T,,, versus Tads, and T, versus 6’ (instead of exponential distribution and the negatively equilibrium take the form given in fig. 6. to an area corresponding to a monolayer amount of substance to an amount corres-

coverage.

Tmi.(Tmi/ log (I-V,,,) log (I-S,,,)

l dS

Fig. 6.

3) Hyperbolic

T, verws

TBdS and T,

versus

distribution and bilogarithmic

Let us write the linear relationrs)

log(l

~ Vm) [lOg(l ~

%I)].

isotherm of adsorption equilibrium

for Ed at 0, and G2:

ln(E&,

= ln(E&

- 0,/C”,

(23a)

ln(E&,

= ln(E&

- 0,/C”.

(23b)

THERMODESORPTION

By substituting

the values

of (E,&

233

STUDIES

and (&&

from eqs. (5a) and (5b) in

eqs. (23a) and (23b), and then subtracting the transformed eq. (23b) from the transformed eq. (23a) with the help of relation (7) we obtain: C” (0, - 0,) = In(T,JT,,)

.

(24)

the value of 0 from eq. (25)

If the above operations are made after substituting of the isobar of adsorption equilibrium ls) e=e,,,-C’lnT,

(25)

where T= Tads, we find WC”)

(26)

ln(Tr, adslG,ads) = ln(T,,/Tk).

Thus, in the case of hyperbolic distribution and bilogarithmic powered isotherm of adsorption equilibrium, the graphical relation between T, and Tadsand between T,,, and S(V) will be as given in fig. 7.

\ 1

Fig. 7.

logTm versus

and logTm versus log,9 (log V).

lOgTads

powered isotherm of adsorption equi-

4) Power distribution and logarithmically librium Let us write the linear relation

of (Ed), (ref. 13)

(E& = (E& and the equation

of the isobar 8=

From

log s (log V)

-

c”e’““+ l))

of adsorption

&

[(E&

(27)

equilibrium13)

- C’T”+‘].

(27) and (28) it follows that (E& = (E&

Therefore

- C” &I

(E’& _ gI

Tn+‘]“(“+“.

we have the plot given in fig. 8. Thus the observed

linear relation-

234

V.I. YAKERSON,

V. V. ROZANOV

AND

A.M. RUBINSHTEIN

ship between T, and Tadsindicates that the distribution is either uniform or exponential and it permits us to select the proper distribution depending upon the relation between T, and S(V), lnS(ln V) or ln(1 -S) [ln(lV)]. 5) Surface heterogeneity and compensation efSect14) Let us consider

the case in which EA and Z are connected

by the relation

In2 = yEA.

(29)

Then T,, =

I

\ sl/h+l)

_

E: ~________~ lnb+yE&+2lnT,-1nEA’

(“r/(n+l))

(30)

T, versus W *+I (W

Fig. 8.

n+l ).

This particular case can be well studied by substituting the values of Ei in eq. (30). When EA changes from 18.5 to 54 kcal/mol, T, changes by 500°C (Z= 1013) for a homogeneous surface; but when the surface is heterogeneous the change in T, is only by 100°C if the relation (29) is valid. However, in such a case the relation between Ed and T, instead of being linear [eq. (3)] is, in fact, curvilinear. As T,,, increases, the interval between the maxima of thermodesorption peaks will contract. For two different substances, desorbed from one and the same surface, or for two different forms of desorption of one substance on one surface, or for one and the same substance desorbed from two different surfaces when conpensation effect takes place and eq. (29) holds, we have for thermodesorption the following equations y

(314

exp (E&JT,J E& = b&T,, , In (Z,/E&,) = In (Z2/E&) = y .

@lb)

exp(E&lT,,)

From

these equations,

E& = b&T’,,

by simple algebraical ~

_

log

(29)

operations, z2

T,,

)I

z2

we get

hzz,

z, logZ,

.

THERMODESORPTION

235

STUDIES

5. Experimental results 5.1. MULTIPLE FORMS OF ADSORPTION AND IRREVERSIBLE NATURE OF CHEMSORPTION

5.1.1. Thermodesorption

of alcohols,from

the surface of alumina15)

The experiments were carried out with 0.156 g of sorbent with Sspec = 500 m’/g. A column (4 cm in length and 4 mm inner diameter) was partially filled with sodium chloride (0.2-0.25 mm) sprayed with stationary phase 1% polyethylene glycol 4400 was used as separating column between the first and second detector channels. The reaction column filled with alumina to 1 mm length, frac. 0.25-0.5 mm was connected after the separating column. This column prevented the initial substances from entering the second channel, but it allowed the gaseous decomposition products into the second channel (“back” peaks). The samples were introduced with a microsyringe. a) Thermodesorption of methyl alcohol. The experiments were carried out with Al,O, dehydrated at 450-500°C and with adsorbed methanol. The following peaks were observed in the thermodesorption chromatogram (fig. 9): the peak 1 appeared immediately after the sample was introduced; if

I

T=const. Fig. 9.

J

I

170

I.

340

_

T,C

Thermodesorption chromatogram of methyl alcohol on alumina surface, dehydrated at 45G5OWC (b’ = 12.7 “C/min, V= 5 ~1, Tads = 45 “C).

Tadswas higher than 60°C and the sample introduced in sufficient quantity, then the peak 2 was weakly developed. However, whenever peak 1 existed, peak 2 was distinctly visible. Peak 3 (T,= 19O’C) appeared with a clear-cut maximum and the peaks 2 and 4 (T, = 370 “C) were located at the left and

236

V.I. YAKERSON,

V. V. ROZANOV

AND

A.M. RUBINSHTEIN

right side of peak 3 respectively. Finally, the peak 5 is a “back” peak. By using the step thermodesorption method, it was possible to confirm that this peak 5 is the “back” peak of peak 4. T, of all the peaks did not change in the interval 5-10 ml/min of carrier gas rate. When the amount of methanol was reduced from 5 ~1 to 1.25 ~1, the situation was simple: the peaks 1, 2 and 3 disappeared gradually. However, under partial thermodesorption conditions with lowering of temperature (the samples were introduced at 100, 150,250 and 300°C with subsequent lowering of temperature down to 50 “C and then the increasing programming was used) the picture was different. T, of peak 4 does not depend on Tads (fig. lo), T, of peak 3 does not change in the interval 50-100°C of Tads, but when the temperature is 150°C it shifts

L

100

Fig. 10.

200

300

T&,-c

The dependence of maximum temperature of one of the thermodesorption on the temperature of adsorption of methyl alcohol on alumina surface.

peaks

to 250°C and at higher temperatures it disappears altogether; the peaks 1 and 2 in this series of experiments developed only one peak which appeared immediately after methanol was introduced. A combination of partial and step thermodesorption (a series of experiments where the substance is introduced at 50°C and a subsequent increase of temperature up to the so-called temperature of partial thermodesorption 120, 170,240 and 270 “C is followed by a temperature lowering to 50°C and a subsequent increase of programming) permitting us to fix the position of T,,, belonging to peak 4 so that it does not change; the position of T, of peak 3 remains constant (Tm= 330°C) in the interval 240-270°C of partial thermodesorption, but at lower temperature values of T, decreases. Thus, methyl alcohol is easily chemisorbed on alumina at 20 “C with the formation of all the four forms of adsorption. Peaks 1 and 2 correspond to weakly adsorbed methyl alcohol, desorbed in unchanged chemical form, Ed of peak 2 (at Z= 1Or3 set-‘) is equal to 28 kcal/mol. Peak 3 corresponds to a more stable form of adsorption (which covers a large fraction of the surface), and dehydration of methanol takes place through this form of adsorption. According to the data given by Svetlanov and Flidr6), dehydration of methyl

THERMODESORPTION

alcohol

on A&O3 has a pre-exponential

STUDIES

equal

237

to 2.4 x 10’ in the interval

195-285 “C. With the help of the thermodesorption equation (2), it was calculated that for the form for which T,= 19O”C, the value of Ed (in this case it is the activation energy or reaction) was equal to 22 kcallmol, and this value coincides with that given in ref. 16. The sensitivity of T, for peak 3 with respect to Tadsor the temperature of partial thermodesorption indicates that the surface is obviously heterogeneous in this case; and this well agrees with the considerable area under peak 3 in the chromatogram as a result of high surface coverage in comparison with other forms of adsorption. Finally, the form of adsorption corresponding to peak 4 is the most stable one with respect to which Al,O, behaves as a quasi-homogeneous surface. This form of adsorption is not an intermediate in dehydration; however, side reactions leading to the formation of gaseous decomposition products do occur as a result of decomposition of this surface compound. b) Thermodesorption of isopropyl alcohoP). Three peaks (fig. 11) are observed in the thermodesorption chromatogram, the last of which is a mV

I

r=~on+i 100 Fig. 11.

250

CT-

Thermodesorption chromatogram of isopropyl alcohol on alumina dehydrated at 450-500°C (b’ = lO”C/min, V= 2.5 ~1, Tads = 48°C).

surface,

“back” peak, and the first peak appears immediately after the sample is introduced in the thermostated condition with Tadsstarting from 30°C. T, of peak 2 is equal to 183 “C. On a sample with the surface hydrated at 20 “C the general picture does not change, but T, of peak 2 shifts by 15-25°C in the region of higher temperatures. Thus, as in the case of methyl alcohol, adsorption takes place very rapidly practically with any activation energy. A part of the substance is desorbed in an unchanged form (peak l), and

238

V.I.YAKERSON,

V.V.ROZANOV

AND

A.M.RUBINSHTEIN

another part is chemically adsorbed and desorbed with the formation of propylene which passes through both the cahnnels of the thermal conductivity cell. Desorption occurs simultaneously with the reaction and its E,, is 31 kcal/mol at Z= 1013 set-‘). This magnitude is very close to the activation energy of reaction in the adsorbed layerr?). On a hydrated surface the value of Ed is slightly higher by a few calories. The increase in Ed on a hydrated surface and the increase in Ed with the amount of previously adsorbed water indicates that the competition of alcohol and water is the same in one and the same place on the surface. If water is introduced after the alcohol has been adsorbed on a dehydrated surface, the result is the same as in the case of adsorption of alcohol on a hydrated surface. These results supplement the infrared spectroscopic data on isopropanol adsorbed on aluminaIs). The fact that the activation energies of desorption of methyl alcohol and isopropyl alcohol coincide with activation energies of dehydrationi6*t7) is, in our opinion, not accidental. Perhaps the thermodesorption method during desorption of reaction permits the determination of the activation energy of the reaction. 5.1.2. Thermodesorption

of diisopropyl

Fig. 12 shows the thermodesorption on dehydrated surface of alumina;

ether from

the surface oj’ alumina 15)

chromatogram of diisopropyl ether ether gives two forms of irreversibly

mV

I I

T.cmrt.kj$o Fig. 12.

1 I

’

350

TLC

Thermodesorption chromatogram of diisopropyl ether on alumina dehydrated at 456500°C fb’ = 14”C/min, V = 2.5 ~1, Tads = 45 “C).

surface,

THERMODESORPTION

chemisorbed

substances

239

STUDIES

which are decomposed

with the formation

of gaseous

products. The two peaks corresponding to these forms of adsorption are characterized by T, at 140 and 160°C and the corresponding values of Ed are 28 and 29.5 kcal/mol respectively (Z= 1013 set-r). The two “back” peaks, which have practically the same heights and areas as the direct peaks, indicate that each of these forms undergoes decomposition with the formation of gaseous products. When ether is adsorbed on an Al,O, surface hy drated at 20 “C, T, of peak 1 decreases to 105 “C, while T,,, of peak 2 increases to 200°C. The last magnitude is characteristic of adsorption of isopropyl alcohol. We can suppose that diisopropyl ether in the presence of water gives one of the forms of adsorption as given by isopropyl alcohol. 5.2. SURFACE HETEROGENEITY AND TYPE OF DISTRIBUTION FUNCTION Recently, gas chromatographic elution methods of determining adsorption characteristics of solids are widely used to study the adsorbate-adsorbent interactions lg, 20). H owever, there arises one main difficulty in interpreting the results, viz. the question whether the output characterizes the behaviour of the substance in physical adsorption or in chemical adsorption. The previous6,ai) g as c h romatographic investigations on the adsorption of a number of saturated, unsaturated and aromatic hydrocarbons on a chromia-alumina potassium catalyst show that the output curve represents only the physical adsorption. In fact, these data reflect only the specificity of physical adsorption, while chemisorption interaction defied experimental determination. The studies on the well known system 22), viz. surface of y-alumina-water vapour by gas adsorption chromatography corroborate this point of view, because the values of the adsorption heat determined from the temperature dependence of the correct retention volumes in the range 160-45O”C corre-

I I I L

Fig. 13.

Thermodesorption

chromatogram of water on alumina surface (b’ = 7S”C/min, v=2 x lo-sg, B=O.l).

240

spond

V.I. YAKERSON,

V. V. ROZANOV

to very weak interactions.

AND

A.M.

The limitations

RUBINSHTEIN

of gas adsorption

elution

methods in investigating adsorbate-adsorbent interaction caused by the necessity of establishing adsorption equilibrium do not arise in chromatographic methods of thermodesorption; consequently, it is possible to study both weak and strong forms of adsorption, and surface heterogeneity, over a wide range. 5.2.1. Thermodesorption of water from the alumina surjbce23) When ethers and alcohols are thermodesorbed from alumina surface in a definite temperature interval, the adsorbate is decomposed and water is one of the decomposition products. But in the ordinary chromatographic investigations on adsorption of water, the output curve represents only the weak physical adsorption ““). Various forms of adsorption on alumina and surface heterogeneity were investigated by the different thermodesorption methods described above. In linearly increased programming procedures, only one peak with an extended tail at higher temperatures was observed (fig. 13). When the amount of water V in calculating the surface coverage exceeded 0.4, then the peak corresponding to excess substance appeared long before the increased programming was switched on. As the amount of water was increased from 0.0012 to 0.006 g (0=0.06-0.27), T, sharply dropped from 387 to 128 “C, and then it remained constant (fig. 14). In partial thermodesorption with decrease of temperature, when Tadsof water was changed from 94 to 306°C T,,, increased linearly from 206 to 416°C (fig. 15). Therefore, we T,V

,

xlo-

200.

Fig. 14.

PlotofT,,,versus

Von the thermodesorptionchromatogramof

water on alumina.

THERMODESORPTION

241

STUDIES

can conclude that water is easily adsorbed on y-A1,03 surface dehydrated at 4%500°C practically without any activation energy. The forms of adsorbed water with respect to bond strength are rather varied. The calculation of Ed with the thermodesorption equation (2) (Z is assumed to be equal to lOI3 set-‘) indicated that when 6 changes from 0.06 to 0.27, Ed drops from

TCrdJ.lc

250 Fig. 15.

Plot of T, versus Tads for the~od~orption

of water from ahunina surface.

46 to 26 kcal/mol. Depending on the quantity of the sample, a part of the substance is strongly bonded and the other part is less strongly bonded. Thus, when the amount of water is large (8>0.4), part of the water appears long before the beginning of the programming (at 20-5O”C), and afl the thermodesorption curves are stretched over the entire chromatogram right mY

I

~~~ I

I I I

r=const. Fig. 16.

30

Thermodesorption

100 chromatogr~ Y= 8 X 1o-3

150

200

250

300 r,*c

of water on alumina surface @J’= 14”C/min, g, Tsds = 33 “C).

242

V.I. YAKERSON,

V. V. ROZANOV

AND

up to 450 “C (the upper limit of programming).

A. M. RUBINSHTEIN

Here, a fraction

of the water

is desorbed at all temperatures, but maximum desorption takes place at 126°C (fig. 16). These results correspond to desorption from different sites differing in bond strength; moreover a wide range of Ed values can be observed. When the amount of water is reduced (and therefore 0), water will be adsorbed on sites with higher heat of adsorption, and desorption will occur when T,,, is higher. An analogous picture is observed when adsorption is conducted at higher temperatures: a part of the substance appears immediately after the sample is introduced (not strongly adsorbed water), and another part when programming is applied. Thus, when Tads is equal to 94 “C, the remaining water on the surface is desorbed with Ed = 33 kcal/mol, and when Tads= 306 “C, with Ed = 48 kcal/mol (Z is taken to be 10’ 3 set- ‘). The observed linear relation between T,,, and Tads, as was pointed out earlier, shows that the distribution is exponential or uniform. The values of Ed so obtained are at the same time the values of the adsorption heat if we take into consideration that water is adsorbed without activation. The wide range of Ed values illustrates the heterogeneity of the surface. However, we should take into account that in this case both desorption of water and dehydration of the very surface take place. And this is very well corroborated by spectroscopic data on the state of water on the adsorbent-oxide surfacez4). 5.2.2. Thermodesorption catalysts

of benzenefrom

the surface of nickel-aluminium

oxide

a) Benzene on 40% NiO-Al,O, catalyst. It is shown from fig. 17 that when heating is programmed, the output curve is characterized by one straight symmetric peak with T,,= 190°C. Before programming is started, excess substance develops a complex peak with a shoulder on the thermodesorption chromatogram. Benzene is adsorbed and desorbed without decomposition. The fact that benzene is adsorbed only in one form, and that is does not

Fig. 17.

Thermodesorption chromatogram of benzene on 40% NiO-A1203 (b’ = 13.6”C/min, V = 1.8 ~1 CsHs, TzdS= 8O”C, 9 = 2.5 g catalyst).

catalyst

THERMODESORPTION

243

STUDIES

decompose while being adsorbed or desorbed, is a very convenient prerequisite to verify experimentally the dependencies derived earlier between the parameters of the equations characterizing the various distribution laws and the characteristics obtained in thermodesorption studies. The general relation between T, and Tadsis a linear one (fig. IS). The plot of log T, versus T, lc 420 t

60 I.

.~

0 Fig. 18.

60

Plot of T,,, versus

120

._,..

180

,’

240 300

360 420

Twt,,*c

for thermodesorption of benzene from the surface of 40 % NiO-A1203 catalyst.

Tads

indicating that the distribution is hyperbolic, and that the isotherm of adsorption equilibrium is bilogarithmic, is not well defined. However, plots of log T, versus S, as well as of log T, versus I’, are evidently curvilinear, and according to our analysis, the distribution should be other than hyperbolic. In fact, the question here is to select between exponential and uniform distributions. As we had shown earlier, when the distribution is uniform, then not only the relation between T, and T& is linear, but also that between T, and S, and that between T, and V; but the last two plots log T,,,

200.

I Fig. 19.

\

0

,;

i 2 3 4 9s Plot of Tm versus togs for thcrm~~orption of benzene from the surface of 40% NiO-Al203 catalyst.

244

V. 1. YAKERSON,

V. V. ROZANOV

AND A.M. RUBINSHTEIN

are curvilinear. This excludes uniform distribution with the logarithmic isotherm of adsorption equilibrium with the linear change of activation energy of adsorption and of the desorption depending on surface coverage. In order to select between the isotherms of adsorption equilibrium when the distribution is exponential, it is necessary to analyze how the points are distributed on the plots T, versus logs, T, versus log V, T, versus log (1 - I&,), and Y&versus log(1 - Sre,). Unlike the last two cases, the points on the T,,, versus logs curve (fig. 19) and on the T,,, versus log V curve are well described by the equation of a line. And this confirms that the distribution is exponential and the isotherm is a power one. b) Benzene 0~170% NiO-A1203 catalyst. The thermodesorption chromatogram (fig. 20) points out two forms of adsorption on the catalyst surface at

Fig. 20.

Thermodesorption chromatogram of benzene on 70% NiO-Al203 (6’ = 13,6”C/min, V= 2.0 ~1, Taa,= 6O”C, 9 = 1.3 g catatyst).

catalyst

T, = 175 and 321 “C. The excess amount appears as a non-symmetrical peak with a maximum and a shoulder. The data can be analyzed correctly only for the first low temperature peak. As in the case of a 40% NiO-Al,Os catalyst, a linear relation between T, and Tads(fig. 21) coordinates and the weak curvilinear curve of log T, versus log Tads are observed. The log T,,, versus Tadscurve may represent a hyperbolic distribution, but the two plots Tm ‘C

300 250

a 0

200

~

Fig. 21.

too

200

300

Tadz..T

Plot of T, versus Tads for thermodesorption of benzene from the surface of 70% NiO-AI203 catalyst.

THERMODESORPTION

STUDIES

245

of log T,,, versus S and log T, versus V exclude this possibility. Uniform distribution with logarithmic isotherm for adsorption equilibrium is also excluded by experimental data, since it is not possible to obtain a linear relation on the plot of T, versus Sand of T, versus V. Thus, the only solution is an exponential distribution with a power or negative power isotherm for adsorption equilibrium. In fact, the linear relation between T, and Tads is also confirmed by the fact that the points on the plots of T, versus logs (fig. 22) and T, versus log V are defined by a straight line equation; the curvilinear relationships between T, and log(1 -S,,J and between T, and log

350-

250.

150.

Fig. 22.

Plot of T, versus logs for thermodesorption of benzene from the surface of 70 ‘A NiO-A1203 catalyst.

(1 - V,,,) decidedly

favour the selection of an exponential distribution with power isotherm for adsorption equilibrium, with logarithmic change of activation energies with respect to coverage, and by the power equations of

the kinetics of adsorption and desorption. The large number of experiments (-70) carried out in a wide range of parameters reliably determine the distribution function and the type of isotherm adsorption equilibrium. Thus, the nature of surface heterogeneity in thermodesorption methods is determined by plotting - 12 graphical relations, where the basic variables are Tads, V, S and T,. References 1) 2) 3) 4) 5) 6)

J. Ehrlich. Advan. Catalysis 14 (1963) 271. V. N. Ageev and N. I. Ioniv, Zh. Tekhn. Fiz. 35 (1965) 2109. Yu. K. Ustinov, V. N. Ageev and N. I. Ionov, Zh. Tekhn. Fiz. 35 (1965) 1106. Y. Amenomiya and R. J. Cvetanovic, J. Phys. Chem. 67 (1963) 144, 2046, 2705. Y. Amenomiya, J. H. B. Chenier and R. J. Cvetanovic, J. Phys. Chem. 68 (1964) 52. V. G. Berezkin, AnaIyticaZ Reaction Gus Chromatography (Nauka, Moscow, 1966) (in Russian).

246

V.I. YAKERSON,

V. V. ROZANOV

AND

A.M.

RUBINSHTEIN

7) V. I. Yakerson, L. I. Lafer, L. A. Gorskaya and A. M. Rubinshtein, Neftekhimiya 5 (I 965) 264. 8) A. W. Smith and S. Aranoff, J. Phys. Chem. 62 (1958) 687. 9) V. I. Yakerson and V. Ya. Danyushevsky. Bull. Acad. Sci. USSR, Div. Chem. Sci. (1967) 197. 10) P. A. Redhead, Trans. Faraday Sot. 57 (1961) 641. 1 I) P. A. Redhead, Vacuum 12 (1962) 203. 12) S. 2. Roginsky, Adsorption and Catalysis on Heterogeneous Surfaces (Akad. Nauk SSSR, Moscow, 1949) (in Russian). 13) S. L. Kiperman, Introduction to Kinetics of‘ Heterogeneous Catalytic Reactions (Nauka, Moscow, 1964) (in Russian). 14) V. I. Yakerson, Bull. Acad. Sci. USSR, Div. Chem. Sci. (1967) 199. 15) V. I. Yakerson. L. 1. Lafer and A. M. Rubinshtein, Bull. Acad. Sci. USSR. Div. Chem. Sci. (1967) 195. 16) E. B. Svetlanov and R. M. Flid, Zh. Fiz. Khim. 40 (1966) 3055. 17) 0. V. Krilov, Zh. Fiz. Khim. 39 (1965) 2657. 18) V. 1. Yakerson, L. 1. Lafer and A. M. Rubinshtein, Dokl. Akad. Nauk SSSR 174 (1967) Ill. 19) G. Shay, Theoretische Grundlagen der Gaschromatographie (Berlin, 1961). 20) A. V. Kiselev and Ya. I. Yashin, Gas Adsorption Chromatography (Nauka, MOSCOW, 1967) (in Russian). 21) V. 1. Yakerson, L. 1. Lafer, L. 1. Gorskaya and A. M. Rubinshtein, Bull. Acad. Sci. USSR, Div. Chem. Sci. (1964) 1638. 22) V. I. Yakerson, Izv. Akad. Nauk SSSR, Ser. Khim. (1967) 1364. 23) V. I. Yakerson and A. M. Rubinshtein, Izv. Akad. Nauk SSSR, Ser. Khim. (1967) 1367. 24) V. I. Yakerson, L. 1. Lafer and A. M. Rubinshtein, Zh. Prikl. Spektroskopii 5 (1966) 360.