In this, the first of a series of papers, we lay the foundations for appreciation of chemical surfaces as D‐dimensional objects where 2≤D<3. Being a global measure of surface irregularity, this dimension labels an extremely heterogeneous surface by a value far from two. It implies, e.g., that any monolayer on such a surface resembles three‐dimensional bulk rather than a two‐dimensional film because the number of adsorption sites within distance l from any fixed site, grows as lD. Generally, a particular value of D means that any typical piece of the surface unfolds into mD similar pieces upon m‐fold magnification (self‐similarity). The underlying concept of fractal dimension D is reviewed and illustrated in a form adapted to surface‐chemical problems. From this, we derive three major methods to determine D of a given solid surface which establish powerful connections between several surface properties: (1) The surface area A depends on the cross‐section area σ of different molecules used for monolayer coverage, according to A∝σ(2−D)/2. (2) The surface area of a fixed amount of powdered adsorbent, as measured from monolayer coverage by a fixed adsorbate, relates to the radius of adsorbent particles according to ARD−3. (3) If surface heterogeneity comes from pores, then −dV/dρ∝ρ2−D where V is the cumulative volume of pores with radius ≥ρ. Also statistical mechanical implications are discussed.

1.
G. A. Somorjai, Chemistry in Two Dimensions: Surfaces (Cornell University, Ithaca, 1981).
2.
By deviations we mean smooth, but not necessarily small deformations. So, for example, deformations leading to a sphere are not excluded.
3.
A lucid discussion of this point, actually much in the spirit of the fractal concept, can be found in Ref. 15, p. 523.
4.
Referenced as parts I, II, and III.
5.
(a) B. B. Mandelbrot, Fractals—Form, Chance, and Dimension (Freeman, San Francisco, 1977);
and (b) B. B. Mandelbrot, Fractal Geometry of Nature (Freeman, San Francisco, 1982).
6.
D.
Avnir
and
P.
Pfeifer
,
Nouv. J. Chim.
7
,
71
(
1983
).
Parts of the current work (parts I and II) have been presented at the Fifth European Conference on Surface Science, Gent, 1982;
and at the Israel Chemical Society Annual Meeting, Tel Aviv, 1982.
See also
P.
Pfeifer
,
D.
Avnir
, and
D.
Farin
,
Surf. Sci.
126
,
569
(
1983
);
and D. Avnir, D. Farin, and P. Pfeifer, in 3S 1983 Symposium on Surface Science, edited by P. Braun, G. Betz, W. Husinsky, E. Söllner, H. Störi, and P. Varga (Technische Universität, Vienna, 1983), p. 233–236.
7.
See Ref. 5(a), p. 54–55.
8.
(a)
M. E.
Fisher
,
Essays Phys.
4
,
43
(
1972
);
(b)
K. G.
Wilson
,
Sci. Am.
241
,
140
(
1979
).
9.
From a mathematical viewpoint, definition (1) of nonstandard dimension is not the most general one (for example, the limit need not exist). See also Ref. 5(a), pp. 287 and 300. But it is certainly sufficient for all our purposes.
10.
For example, for a finite number of points, N(r) remains constant for sufficiently small r. whence D = 0. For a straight line of length L. N(r) is the smallest integer >1/2Lr−1 so that D = 1.
11.
See part II. For instance, example 1 mimics anisotropic crystal‐surface roughness. Example 2 anticipates highly porous amorphous solids.
12.
This shows that, while curves with D − 1 substantially > 0 have to be quite irregular, the same value for the surface analog D − 2 is attained by comparably mild irregularity if isotropic.
13.
See Ref. 5(b), p. 145, for a graphic. In Ref. 5(a), p. 164–167, this was called Sierpinski sponge.
14.
Thus, Mandelbrot’s paradigmatic example of coast lines having D≈1.25 lives in a range of 1–1000 km (Ref. 5). Our surface yardsticks will vary from 16 Å2 to as high as 180 000 Å2. But indirectly inferred values like D≈1.65 for the backbone wiggliness of proteins [
H. J.
Stapelton
,
J. P.
Allen
,
C. P.
Flynn
,
D. G.
Stinson
, and
S. R.
Kurtz
,
Phys. Rev. Lett.
45
,
1456
(
1980
)]may not be easily attributed to a specific such range.
15.
A. W. Adamson, Physical Chemistry of Surfaces, 3rd ed. (Wiley, New York, 1976).
16.
K. K. Unger, Porous Silica (Elsevier, Amsterdam, 1979).
17.
R. K. Her, The Chemistry of Silica (Wiley, New York, 1979).
18.
C.
Pierce
, and
B.
Ewing
,
J. Phys. Chem.
68
,
2562
(
1964
).
19.
A. L.
McClellan
and
H. F.
Harnsberger
,
J. Coll. Interface Sci.
23
,
577
(
1967
).
20.
M. W.
Cole
,
D. R.
Frankl
, and
D. L.
Goodstein
,
Rev. Mod. Phys.
53
,
199
(
1981
).
21.
AEROSIL (Degussa), CAB‐O‐SIL (Cabot). See also part II.
22.
Since global deformation of the surface does not affect the asymptotics (2), any other macroscopic shape is equally admissible.
23.
Back to a single particle, this can also be expressed as “area‐volume relation” (Ref. 5(b), Chap. 12): [fixed‐yardstick surface area]1D∝[volume]13.
24.
The situation described in Table II is not exhaustive. For example, by scaling up the feth approximant of the Koch surface (Sec. III) by a factor of 3K−1(k = 1,⋯,10) one obtains a series of surfaces that are similar within resolution of squares of length r0 = 1. On the kth one of these, Eq. (8) holds [with Eq. (4)] for r0⩽r⩽rmax = 3k−1, and essentially only for this range. Yet, for these surfaces, Table II remains also true if, in columns 4 and 1, r0 is replaced by any r0*<r0. Thus, when surfaces are not similar within the resolution afforded by the fixed yardstick (here r0*, provided it is sufficiently below r0), the dependence n vs R may detect a D>2 even with a yardstick below the associated fractal regime. But the absence of such similarity may also prevent the dependence n vs R from detecting the proper D>2, even if the yardstick is in the fractal regime. An example is obtained from assembling k3 Menger sponges of unit length (Sec. HI) to a “cube” of length k(k = 1,⋯10). Within resolution r0 = 1, these cubes are indistinguishable from true cubes and hence similar within this very resolution. On any one of these surfaces, however, Eq. (8) holds with D = 2.72⋯ for 0<r⩽r0. While under any fixed yardstick of radius r0*⩽r0, Eq. (13) holds with D = 3.
25.
Menger’s sponge may serve again as a model: When breaking it into pieces, one expects fragmentation along the “faces” that connect the 8 corner to the 12 bridging cubes. So in the first grinding step, the original sponge of unit length breaks into 20 sponges of length 1/3, each of which is (strictly) similar to the original one, etc.
26.
It is instructive to see this for the Koch curve: If the tip of the curve is chosen as origin, no such deformations are necessary. For other choices, some peripheral wings of large‐l segments have to be inverted to achieve similarity.
27.
Formally, Eq. (14) can also be understood as a “mass‐radius relation” (Ref. 5).
28.
For D = 2, Eq. (16) is to be read as 0(−dv/dσ)σ−1dσ<∞.
29.
An alternative proof of Eq. (16) considers channels to and from pores as negligible compared to the cavities themselves (ink‐bottle pores). Thus, while all pores are treated as closed, their walls may be arbitrarily irregular. One then invokes the “diameter‐number relation” (Ref. 5(b), p. 118–121) to conclude that the number of pores with radius ⩾σ is proportional to σ−D, where D is the total surface’s dimension. Thus, the volume of pores with radius between σ and σ+dσ is σ3σ−D−1, which is Eq. (16).
30.
Cross reference also, G. A. Somorjai and M. A. Van Hove, Adsorbed Monolayers on Solid Surfaces (Springer, Berlin, 1979).
This content is only available via PDF.
You do not currently have access to this content.