Adsorption and surface heterogeneity

Adsorption and surface heterogeneity

SURFACE SCIENCE 24 (1971) 391-403 8 North-Holland Publishing Co. ADSORPTION AND SURFACE HETEROGENEITY * G. F. CEROFOLINI Uficio Consulenze Scient...

549KB Sizes 1 Downloads 144 Views

SURFACE SCIENCE 24 (1971) 391-403 8 North-Holland Publishing Co.

ADSORPTION

AND SURFACE

HETEROGENEITY

*

G. F. CEROFOLINI Uficio Consulenze Scientifiche, via Pier della Francesca 74,201W Milano, Italy

Received 5 June 1970 The problem, of finding the energy distribution function of a heterogeneous surface, is considered here. In order to solve the problem, we approximate the effective local isotherm (supposed to be Langmuir) with the nearest one, in the Lagrangian sense, in a given functional class. With this local isotherm, and Dubinin-Radushkevich (DR) as the overall one, we compute the energy distribution function. With this result, we are able to compare the developed theory with the Polanyi one. 1. Introduction A surface, heterogeneous with respect to physical adsorption of a given gas, can be considered as consisting of a family of non-interacting, energetically homogeneous, zones. If B(p, q) is the local isotherm, i.e. the law that relates the coverage 8 to pressure p in the zone with binding energy q, then the isotherm for the whole surface is given by +a, WP) =

s

ebdcP(ddq,

0

where cp(q)dq is the fraction

of surface with binding

energy between

q and

q+dq. The most important problem arising from eq. (1) is the following: “given the local isotherm 6(p, q) and the overall isotherm 9(p), determine the distribution function that satisfies eq. (1)“. In the following, we shall consider only coverages less than one monolayer, for which lim

O(p,q)=l,

p++C0

lim

9(p)

= 1.

p++a,

* Part of this work has been supported by a SAES Getters SpA grant. 391

(2)

(3)

392

G. F.

Substituting relationships condition for q(q),

CEROFOLINI

(2) and (3) in eq. (1) we get the normalization +m

s

cP(ddq= 1,

0

which suggests that eventual solutions of eq. (1) belong to the functional space L’ (0, + 00)~ consisting of the summable functions on the positive semiaxes (0, + a). If no solutions exist in this space, they must be looked for in the larger distribution space, as shown by a simple example: if 9 (p) = 0 (p, 4), then the solution of eq. (1) is cp(q) = 6 (q-q) which corresponds a homogeneous surface with binding energy q. In this paper we shall consider as local isotherm the Langmuir one

O(P,4) = ~~

P

p+PLev(-q/W’

and as overall isotherm a(P)=

the DR onei-3)

expC- B(RT In(P~/P>>“I =P i

]

G p. y

-=P’PO?

where the symbols have the usual meaning. Both eqs. (4) and (5) satisfy relationships (2) and (3); furthermore Langmuir local isotherm satisfies the following asymptotic condition

(5)

the

i.e., Langmuir’s isotherm behaves at low pressures as Henry’s isotherm. In contrast with this, the DR isotherm does not have this property; the meaning of this is discussed elsewhere4). The problem of finding the distribution function q(q) satisfying eq. (1) when the local isotherm is the Langmuir isotherm, and the overall isotherm is DR isotherm, has not yet been solved and in order to find approximated solutions, two methods have been developed. In the first method, the overall isotherm is approximated by anothers), obeying analytical requirements such that the computation technique described by Sipsc) can be used. In the second method, the local isotherm is replaced by a kernel which, considered as a function of p, contains a discontinuity or an angular point 4,‘~ 8) (i.e., a point where the first derivative is discontinuous). So doing, the first kind Fredholm integral equation (1) is easily reduced to a Volterra integral equation. whose solution method is well knowng).

393

ADSORPTION AND SURFACEHETEROGENEITv

In this paper

we prefer the second method

since the overall

isotherm

is

the only sure datum* of our problem. As a consequence of our choice a preliminary problem raises up, i.e., to choose - in a given functional class - the kernel which is the best approximation

to the Langmuir

local isotherm. 2. The variational problem

Let us consider ditions :

a family

(9

of local isotherms

H'(p,q)=

satisfying

the following

con-

H(P, 4)-=P G P'(4),

I 1

GP>

p'(4);

p’(q) is a “condensation pressure”, and is - for physical reasons - a monotonically decreasing function of q. The variational problem we consider is the following: “find the function p=p’ (q) such that the Lagrangian distance between H’(p, q) and 0 (p, q):

(ii)

d [H’, O] =

sup

HO. +=Jj

W’(PT 4) - B(P, 41

for all q”.

is minimum,

2.1. CONDENSATION LOCAL ISOTHERM The condensation

local

isotherm

K(PY 4) = 40

F

Oc=P G Pc(d,

0

%(p,q)

r____________________________-___

-P

(7)

1 l~P>Pc(q).

:

0

Fig. 1.

is the following*)

-

(-$)

P/P,

sy-&)

The Langmuir isotherm compared with the nearest condensation

local isotherm.

* The DR isotherm has been tested for various gas-surface systems and for coverages 9 between 10e6 and 10-r; for lower values of 9 the isotherm should behave as the Henry isotherm*); for greater values of 9 the isotherm falls in the BET range, i.e., on the surface there is a multi-layer adsorption. As a general rule, the DR isotherm holds true for heterogeneous surfacesla~4.11-1*) and this is the experimental background of our considerations.

394

G. F. CEROFOLINI

From fig. 1 it is easily seen that the distance

between

8 (p, q) which are continuous

(for all local isotherms

I-I, and 0 is minimum

* with respect top) when

e(Pc(4)t4) = 4. Replacing

in the implicit

equation

PC(q)= so that the best approximating K(P,

pL

(8) the expression

O-=P

(4) we obtain

exp(-- q/W, is ** :

local isotherm

4) =

(8)

< pLexp(- q/RT),

(7’)

i l-=p>PLexp(-q/W.

2.2. ASYMPTOTICALLYCORRECTLOCALISOTHERM Now we consider

the following

H,(PT4) =

adsorption

isotherm

PIPLexp(-q/RT)~pdP,(4),

1 l

which is called “asymptotically

-=P ’

correct”

P,(4),

since it satisfies conditions

(2) and

(6).

Fig.2. The Langmuir isotherm compared with the nearest asymptotically correct local isotherm. * If the local isotherm is the Hill-De Boer isotherm, then the value of the condensation pressure is suggested by different considerations*). ** Hobson and Armstrong4) considered the following condensation pressure P (4) =

PO em

(-

q/RT).

(9’)

With this choice (which shows the correct functional dependence on q) the condensation on the zone with zero binding energy occurs at the vapour pressure of the adsorbate (it is to be remembered that the DR isotherm holds true at temperatures lower than that of vapour condensation). In such a manner, with the choice (9’), Hobson and Armstrong have also taken into account the imperfect nature of the gas.

ALISORPTTON

AND SURFACE

395

HETEROGENEITY

From fig. 2 it is easily seen that the function which solves the variational problem is implicitly given by Pa(q) ~ exp(q/RT) PL

=

1,

from which we obtain pa(q) = pL exp (-

q/RT).

(11)

Under condition (1 l), eq. (10) becomes (P/P,) exp(qlRT)

* P 6 pL exp (-

-=P>PLexp(-q/RT).

q/RT),

(10’)

An isotherm of type (10) has already been considered by Hobson’), who also imposed a condition of type (11) for condensation pressure. However, the choice (11) in the work of Hobson is an “ad hoc” hypothesis and is not upheld by considerations of variational nature. The condensation pressure pa(q) [=pc(q)] satisfies the requirement (ii), so that no other conditions must be imposed. 3. Distribution function In this section we shall compute the energy distribution function corresponding to local isotherms RC(p, q) and ii,(p, q) 3.1. CONDENSATIONLOCALISOTHERM

Let us consider the inverse function of pc (q) Q(P) = RT In(p,/p),

and express the overall isotherm as a function of Q. From eq. (5) we obtain

exp[I-

B(Q

-

QJ'l-=Q

2 Qo,

*Q
(12)

where

Q. =RTln(pL/po).

(13)

Replacing in eq. (1) the best condensation local isotherm, eq. (7’), and the DR isotherm, eq. (12), we obtain +CO

s

4oc(ddq = g(Q),

Q

396

0.F.

which can be solved by partial

the distribution

.~. 0

function

CEROFOLINI

differentiation

is reported

with respect

to Q:

in fig. 3.

._I_._

2

I

3

4

5

Fig. 3. The distribution function according to the condensation approximation: the function is compared with a Gaussian one (dashed line). This is constructed according to the following rules: the maximum of the Gaussian function is the same of the Maxwellian one, and the width of the curve computed at a half of the maximum height is the same of the Maxwellian function.

3.2. ASYMPTOTICALLYCORRECTLOCALISOTHERM From eqs.(l)and(1O’)we have

where Q(p) is the inverse function of p, (q). Expressing, by means of eq. (1 l), p as a function

of Q we obtain

exp(-~~)jexp(~)U.~q)dq + ? rp,(cddq=g(Q), (14) 0

Q where ,i?(Q) is given by eq. (12). By partial differentiation of eq. (14) with respect following first kind Volterra integral equation exp(-,e,)jexp(~T),.(q)dq=-RT~, 0

to Q we obtain

the

ADSORPTION

AND SURFACE

which reduces to a second kind entiation

Volterra

391

HETEROGENEITY

integral

equation

by partial

differ-

with respect to Q:

This equation

is solved in Appendix (Q)

cp

=

_

RT

B

A and its solution

is

a’So_ aS(Q)

(16)

x’

aQ2

i.e., for eq. (12)* - 2B(Q- Q~)"lewCB(Q- Qo>'l +2~(Q-Qo>ex~[-~(Q-Qo)21-=Q~QO~ 0 *Q
4. Average binding energy In this section, we shall limit our attention to the condensation local isotherm, since results achieved with this method are much simpler than those of the asymptotically correct local isotherm method. In particular we will focus our attention on the average binding energy that in previous works495) has been found to be Cj=(*n)+B-% For the first model considered

(17)

here the average

& =

Q,, it is

qC =

(Qo+u)2Buexp(-Bu2)du=Qo

+m

q4oo(q) dq = j-

0

energy is

Qo+(h)*@,

in fact, if we put p=q+CC s

binding

2Bu s

0

0 -l-co

x exp(-

Bu’) du +

2~2~’ exp(-

Bu’) du = Q. + (@)‘B-*

.

s

0

We can evaluate

Q, as follows.

The expression

p,_ = n,(2rcmk,T)+

of pL is

v. ,

where its is the number of sites per true unit area, m is the ad-atom mass, constant, and v. is the vibration frequency of the ad-

k, is the Boltzmann

* It is to be noted that the problem of finding the distribution function is simply solved in both cases, and the solution methods can be applied to overall isotherms different from the DR. An example of such application will be considered in Appendix B.

398

G. F. CEROFOLINI

atom perpendicular of p. as follows:

to the surface. We can give an approximated

= pL !L’ 5’ n,

expression

exp

~0

where vb, ni and qs represent respectively the vibration frequency perpendicular to the surface, the surface density and the heat of sublimation, all referred to the adsorbate in the bulk (=liquid) phase. Since it is ni/n,- 1 we obtain , which, compared

with eq. (13), gives Q. = qs + RT ln(v,/vb).

If we suppose that the adsorbate is in the liquid phase we obtain Then the expression of the average binding energy becomes

Q, =qs.

& = qs + ($T)+B-f.

(18)

But, if the adsorbate can be considered as a liquid, we can compare our result with those of the Polanyi potential theoryls). According to this theory, by using the Clausius-Clapeyron equation, we obtain the isosteric heat of adsorptionl6) : 4: = qs + [ln(1/9)]2 from which, by repeated of adsorption :

*, we obtain

integrations qt=

the average isosteric

qs + 2B-f.

From fig. 4 it is seen that the following qst > & )

Bet,

qst N qc )

(19)

relationships (because

$’ 9 $RT),

must hold: both are satisfied, as it is seen by comparing Let us compare now in greater detail the advantages respect to the Hobson

and Aimstrong

eqs. (18) and (19). of our model with

one, and to the Polanyi

one.

* In fact, if we put y = l/9, we obtain 1 [In (l/@12 d9 = / - i2 (111y)~dy = [b (lny)2 1 + 2 / _ $ lny dy = +m +CO 0 +CC

=2)[+yj +m

+ j-i2dd=2. +m

heat

ADSORPTION

AND SURFACE

399

HETEROGENEITY

The latter theories both predict an adsorption binding energy which depends on the nature of the adsorbent surface only through B-*, a parameter which - according to our considerations (see comments to fig. 3) gives the energetic heterogeneity of the surface. In particular, if we consider

Fig. 4.

Comparison

between isoteric heat of adsorption

and binding energy.

the adsorption of argon, we have B -*N 1 kcal/mole for practically all adsorbent solids: this implies (in those theories) binding energies practically constant for all systems; this fact is hard to believe. Furthermore two simple experimental evidences regret the quoted theories. The first evidence is the following: the Hobson and Armstrong choice of the condensation pressure (9’) implies the expression (17) for the average binding energy; the experimental value of Bw3 is, however, too low to give a reasonable accord with computer determinationsr7), at least in one case: i.e.14) argon adsorption on nickel. The second evidence is the following: the Polanyi theory gives an average adsorption binding energy always greater than q,, while BET measurements report that the adsorption binding energy can be lower or greater than qs. In our model, for a suitable choice of v,, we are able to remove all these difficulties. Appendix A

A second kind Volterra integral equationg) Q

s(Q)-~SF(Q,q)e(q)dq=g(q) 0

400

G. F. CEROFOLINI

has the solution

0

where

G(Q,~.)= +f ~"-1F,(Q,~), ll=l

(A21

F,(Q,q)=F(Q,q),

I

Q

F,(Q, 4)=

s 4

F,-,(Q, t)F(t,q)G

n> 1

From eq. (15) it is seen that if

(A3)

g(q)~-~Td2gce, aQ2



A = l/(RT)) we have the equation which interests us. For eq. (15) evaluating the sum (A2) with the conditions

(A3) we obtain

+g[fT+&(g)2+...}=

G(Q,q;rl)=exp(-CR?){1

1,

for which, from eq. (Al), we have eq. (16):

q(Q)=-RT-+--

’ _RTa2~(Q')dQ, =RTaZg(Q)

a's(Q) 1 aQ RT s

ag(Q>

aQ' -aQ*

,Q'z

0

OBSERVATION

If we consider

the functions O(P) = 1 - W),

K(PT 4) = 1 - B(PT 419 from eq. (I), under

normalization +m

condition

for q(q), we have

s

K(PY4)cP(d& = O(P).

(A4)

0

Now we approximate

K (p, q) with the kernel K(P,

K'(p,q)= () i

4) -z=P G JYq) v -=p>P(q),

(A9

ADSORPTION

AND SURFACE

401

HETEROGENEITY

where P (q) is a monotonically decreasing function of q, and then invertable. If Q=Q(p) is th e inverse function of P=P(q), eq. (A4) with kernel (A5) becomes Q(P) K(P,

4) v(q)

dq

ew

= O(P).

s

0

Considering Q as independent eq. (A6) becomes

variable

and expressing

p as a function

of Q,

Q

s

&Q, 4)cp(4)dq = e(Q)1

(A71

0

where

e(Q)=+(Q)).

R(Q,4)= K@'(Q),41,

By partial differentiation with respect to Q eq. (A7) becomes second kind Volterra integral equation Q

&Q,

Q)

cp(Q)+

s

R,(Q,

4) cp(4) dq =

the following

36(Q) mae-

>

0

whose solution method has been described before. It is to be noted that the two models considered in this work derive from particular applications of the general method described here. This method seems to us the most general one able to give approximated solutions of eq. (1). Appendix B It is claimedi*~4~is~s0) that in the low pressure limit the DR isotherm also reduces to the Henry isotherm. In this Appendix we suppose that effectively the DR isotherm reduces to the Henry one, and considering the asymptotically correct local isotherm, we shall determine self-consistently the pressure at which there is the transition between the DR and Henry isotherms. The overall isotherm we consider is the following &.cxp(Q/W c=Q>QH9 g(Q)= exp[-B(Q-Qo)Z1~QH>Q~Qo, -+Qo>Q, i 1 where k and QH are the parameters which we want to determine. Applying eq. (16) (which has been deduced independently by the overall

402

isotherm

G. F. CEROFOLINI

expression)

in the Henry range, we obtain q,a (Q)

So the distribution

= _

function

a’s(Q) _aS(Q! =

RT

aQ'

aQ

0.

becomes

0

( cP’(Q)=

-+Q~QH, 2B(Q - Qo)‘l+ 2B(Q xexp[-B(Q-QQ,)‘]+Q,>Q>Q,, 0 -=Qo'Q-

@BRTEl-

But (Pi must be normalized near p=O:

independently

Qo)}

from the behaviour

of 9 (p)

QH

l=

s

~~ce,dQ=[-RT~~~]Q=Q~-ia(Q,)-I)

0

=

2BRT(Q,

_

and this is just an equation 2BRT(Q,

,-B(QH-Q@

Q,)

{,-'J(QH-Qo)*

_

I},

for Qu _

Qo)

e-B(Q~-Qo)2 = ,-WQH-QO)~,

which is easily solved, and whose solutions

QH=m, The last result

_

is of present

are

QH - Q. = 1/2BRT. interest

and can be written

as a function

of

&~=Q~-Qo=RTln(p,/po): &H= 1/2BRT. The result (Bl) has been proposed by Hobson and Armstrong4) and was used by Schramls) in thermodynamical computations: strated that eq. (Bl) strictly follows from our formalism.

(Bl) as an ansatz, we demon-

References 1) 2) 3) 4) 5) 6)

M. M. Dubinin and L. V. Radushkevich, Dokl. Akad. Nauk SSSR 55 (1947) 331. M. G. Kaganer, Dokl. Akad. Nauk SSSR (1957) 251. M. M. Dubinin, J. Am. Chem. Sot. 81 (1959) 235. J. P. Hobson and R. A. Armstrong, J. Phys. Chem. 67 (1963) 2000. D. N. Misra, Surface Sci. 18 (1969) 367. R. Sips, J. Chem. Phys. 16 (1948) 490; R. Sips, J. Chem. Phys. 18 (1950) 1024. 7) J. P. Hobson, Can. J. Phys. 43 (1965) 1934; J. P. Hobson, Can. J. Phys. 43 (1965) 1941. 8) L. B. Harris, Surface Sci. 10 (1968) 129;

ADSORPTION

9) 10) 11) 12) 13) 14) 15) 16) 17)

18) 19) 20)

AND

SURFACE

HETEROGENEITY

L. B. Harris, Surface Sci. 13 (1969) 377; L. B. Harris, Surface Sci. 15 (1969) 182. see e.g.: F. G. Tricomi, Znregrul Equations (New York, 1957); A. E. Taylor, Introduction to Functional Analysis (New York, 1958). N. Hansen, Vakuumtechnik 11 (1962) 70. N. Endow and R. A. Pastemak, J. Vacuum Sci. Technol. 3 (1966) 196. R. Haul and B. A. Gottwald, Surface Sci. 4 (1966) 334. F. Ricca, R. Medana and A. Bellardo, Z. Physik. Chemie NF, 52 (1967) 276. A. Schram, Suppl. Nuovo Cimento 5 (1967) 276. M. Polanyi, Verh. Deut. Physik. Ges. 10 (1916) 55; M. Polanyi, Trans. Faraday Sot. 28 (1932) 316. A. Schram, Suppl. Nuovo Cimento 5 (1967) 309. The adsorption binding energy can be evaluated with the aid of potential reported in: F. Ricca, Suppl. Nuovo Cimento 5 (1967) 339; using the value of binding energy of the pair Ni-Ar reported in: J. Lorenzen and L. M. Raff, J. Chem. Phys. 52 (1967) 339. T. L. Hill, Advan. Catalysis 14 (1952) 211. J. P. Hobson, J. Vacuum Sci. Technol. 3 (1966) 281. S. E. Hoory and J. M. Prausnitz, Surface Sci. 6 (1967) 377.

403

maps