I-Ray: The Scherrer For Particle Size

Download as pdf or txt
Download as pdf or txt
You are on page 1of 5

NOVEM HER 15, 1939

PHYSICAL REVIEW

VOLUME 56

The Scherrer Formula for I-Ray Particle Size Determination


Department

A. L. PATTERsON
of Physics, J3ryn Mawr College, Bryn Maser, Pennsylvania
(Received July 24, 1939)

An exact derivation of the Scherrer equation is given for particles of spherical shape, values
of the constant for half-value breadth and for integral breadth being obtained. Various approximation methods which have been used are compared with the exact calculation. The
tangent plane approximation of v. Laue is shov n to be quite satisfactory, but some doubt is
cast on the use of approximation functions. It is suggested that the calculation for the ellipsoidal
particle based on the tangent plane approximation will provide a satisfactory basis for future
work.

1.

of their validity by comparison


calculation.

INTRODUCTION

" 'N

1918, P. Scherrer' showed that, when


parallel monochromatic radiation falls on a
random oriented mass of crystals, the diffracted
beam is broadened when the particle size is small.
By an approximation method he obtained an
of the
expression for the half-value breadth
diffracted beam in the form
~

2. EXACT CALCULATION OF J(X)


Following v. Laue' Eq. (28) and confining our
attention to the contribution from one wholenumbered point (A.= 2k, ~) we may write, in the
notation of I (Eq. (17c) etc. )

B=E7/(L

cos x/2),

(1)

in which ) is the wave-length of the incide nt


the linear dimension of particle, X/2 t he
x-rays,
Bragg angle and X a. numerical constant for
which he obtained the value 2(ln 2/vr)"*=0. 93.
Since then, various workers' ' using different
approximation methods and different definitions
for the breadth B have obtained different values
for the constant E. As a result those interested
in using the relation (1) for the determination of
particle size have rightly been in doubt as to its
correct value.
The results of the preceding paper in this
issue' include as a special case the interference
function for a spherical particle. For this particle, the calculation of the distribution in angle
of the diffracted beam can be carried through
exactly, and a value of the Scherrer constant can
methods of
be obtained. The approximation
other authors are also applicable in this case, and
consequently it is possible to obtain an estimat e

~(x) =

)t "I

Scherrer, Gottinger Nachrichten {1918);cf. also R


Zsigmondy, E'olloidchenzie (3rd Ed. 1920), p. 394.
2 N.
Zeits. f. Physik 31, 439; 33, 648 (1925)
' M. Seljakow,
v. Laue, Zeits. f. Krist. 64, 115 (1926).
4 C. C. Murdock, Phys. Rev. 35, 8 {1930).
~
L. Bragg, The Crystalline State, Vol. 1, p. 189.
' W.
F. W Jones, Proc. Roy. Soc. A166, 16 (1938).
7 A. L. Patterson,
p; 972, hereafter cited as I.

+(&') I'/(&'IHI'~o)

XdAidA2dA3,

'P.

with the exact

=+A;b;,
A ~),
U; = 3II;(2~&;

&IHI

(2a)
(2b)

(2c)

(2a) is taken between


the spheres kIHI and (b+Ak) IHI. J(x) is then
a quantity proportional' to the intensity of the
x-rays scattered through an angle y by a random
oriented mass of crystal particles of uniform size,
each of which has a shape function %(LT,).
We shall confine ourselves here to the simple
case* in which the lattice is cubic of translation
@=1/b and Mi=I)du= M&=M. We may then
write
(3a)
PA, b;I =b PA = p2
in which the integration

2irgbfb;

2
I

= 4vrmb~gb;2 = pi,

2~+Is b QA, b, '=R '/Ild'a'


Ro' 3PQ(27rh;
A, )'= Q U.'
I

(3b)

(3c)
(3d)

F. W. Jones, reference 6, p. 40, has pointed out that


the "constant" implicit in (2a) is constant only for small
variations of y, i.e. , in the neighborhood of a wholenumbered point. It also contains factors which depend on
the particle size, and cannot be used directly in a discussion
of the distribution of particle size.
* These conditions are not necessary for the integration
of (2a). More general conditions can be set up, but so far
no practical use has been made of them.

978

SCHERRER FORMULA
Under such conditions, we may transform the
integral (2a) to polar coordinates (p, 8, q&) with
the vector
(cf. v. Laue, ' Eq. (28a) and Eq. (8)
below) as polar vector, when it takes the form

(0+6 k) IHt

)=JI
k

II

Jt Jl

*)I'/ 'I

Following

6=(gh;b~)/I Ph;b;I.
v. Laue, ' Eq. (32), we may

(8)
then write'

IHI

2~

Ir

gent planes to these spheres at the points i'o H


and (&+LB) H G, in which
is a unit vector in
the direction of the vector gk;b; i.e.,

)(p'

J(,) = Jt

sin 8dpd8:dy.

e(R)

oRdR,

5 /v]

The integration with respect to


immediately and we obtain

can be carried

2x

J(x) =6k

I+(U;) I'

sin 8d6dp,

in which the assumption of cubic symmetry and


the conditions (3) are no longer required. Instead
we must, however, write

(4)

R'=+M

subject to the condition

n=

Ro'/lf'a'=O'I HI'+ po'

2foI

HI p
q cos I1.

(5a)

To integrate this expression, we must be able to


write the U; in terms of the polar coordinates. If,
as in the case of a spherical particle, +(U,) is a
function of Ro alone and is independent of rp (cf. I
Table I), we can integrate directly with respect
to y, and make the change of variable

3Pa'k H
I

sin ada

pg

= RodRo

(2&)

(2~k; A;)'=QU

(10a)

'I ZL(b'6)/~ O'I'

(1ob)

Ke

note that in the special case of cubic symmetry and under the conditions (3) the quantity
8/g takes the form ll/rI=A=2Makb.
W. L. Bragg' was led to make a further
of the discussion by considering
simplification
only the intensity of the x-rays reHected while
the vector H is in the same direction as the
vector G. With this assumption we have simply

(5b)

J(x) = +(&Is) I'


I

We then have (omitting a factor which is constant for a given whole-numbered point)

J(x) = t

I@(Ro) I'RodRo,

in which

5=sin (xo/2)
o

sin

(X/2),

(7a)

=sin (xo/2)+sin (x/2),

(7b)

6 = 2Makb,
Q = 23''ah~.
The exact evaluation

of this integral
spherical particle will be given below.

3.

CALCULATION

Instead of
(2a)
(Io+rQ) IHI,
integral taken
integral

fI(A, ) = C, exp ( (oIoR')

(7c)

fo(A;)

(7d)
for the

J(X)

BY APPROXIMATION
METHODS

OF

this approach is confessedly approximate, it has the advantage of being applicable to


all possible forms of the function 4'(U~).
Since the interference function I%'(U;) I' is not
in general a function of R alone, v. Laue' (Eqs.
(26) and (27)) made use of approximation functions for the interference function in (9). These
functions were of the type

Although

attempting the evaluation of the


between the spheres kIHI and
v. Laue approximates it by an
over the region between the tan-

(12a)

= Co(coo'R'+1) '

(12b)

In discussion with the author some years ago


v. Laue has also suggested a third approximation
fun tion

fo(A;)

ooooR')
Co(1

In each the constant

R'(coo
Rg )cv3

co

'
(12c)

is chosen so that the

9 B. E. Warren, Zeits. f. Krist.


99, 448 (1938}has given
a simplified discussion of some of v. Laue s analysis.

A. L. PATTERSON

980
integral breadth

~(x) = (s/&)'I t:s (&/~)


(&le) cos (o/~))'+(~/v)' n' (oln) I,

W dehned by

W=[f(0)) ')tJtJI f(A, )dA4dAodAo

(13)

(cf. I, Eq. (18)) has the same value for the approximation function and for the interference
function to be approximated.
Jones' (Section V (2)) has also made use of the
approximation functions (12a) and (12b), but he
chooses values for au such that the approximation
function and the interference function give the
same value for the integral breadth (in the scale
of 8) for the function J(x) for the axial planes.

4.

EXACT AND APPROXIMATE CALCULATIONS


FOR SPHERICAL PARTICLES

From the interference function for ellipsoidal


particles (I, Ta, ble I, formula 5) under the special
conditions (3) we obtain the interference function
for a spherical particle of radius 3Ia. It has the
form
I

+(~o) ' = (9/Zoo)


I

(sin Ro

'. (14)
Ro cos Ro)

For this function the integrals (6) and (9) can


be evaluated exactly, and we can therefore obtain
an insight into the nature of the tangent plane
approximation. We can also estimate the accuracy of the Bragg computation. Furthermore,
functions of the
we can set up approximation
three types (12), using both the v. Laue and the
Jones criteria for evaluating 4o; and apply both
the exact and the tangent plane calculations to
them. It is thus possible to use the exact calculation as a test of the validity of the various
methods and to estimate their
approximation
value in cases for which the exact calculation
cannot be carried through.
We substitute (14) in (6) and after integration
by parts we obtain exact form~

J(X) = 6

4t

(sin

4L(sin

6 icos 6)o+iV sin'

Q Q cos P)'+Q'

sin'

while the Bragg approximation

(15)

The integral (9) which follows from the tangent


plane approximation then obviously takes the
form
~ The scales have been chosen for (15}, (16), and {17)
so that they all have a maximum value unity.

leads to

We are now in a position to compare the two


approximation methods with the exact calculation for a spherical particle of a cubic crystal
(8/g=A). We note f4rst that for all practical
purposes, the expressions (15) and (16) become identical. The function y 'I (sin y
y cos y)o
+y' sin' y) has as its slope (4/y ) (sin y
y cos y)', and is therefore a monotonic decreasing function of y whose slope is small for large
values of y. Since
will usually be large compared with 0, the second term of (15) will merely
result in a very small reduction in the background
intensity due to the first term. In general its
effect can be neglected, although in special cases
(e.g. , for x small) it may have to be taken into
account. It seems therefore possible to give
strong support to v. Laue's use of the tangent'
plane approximation. There is nothing in the
analysis to lead one to suppose that the case of
the spherical particle is in any way peculiar as
far as this assumption is concerned, and it is to
be expected that the approximation will be just
as good for particles of other shapes.
The integral breadths B~ and the half-value
breadths B~~o (in the scale of 8) for the functions
(16) and (17) are given in Table I together with
the corresponding value of the Scherrer constant'
X and X~/2. It is seen that for the sphere, the
Bragg approximation
agrees with the exact
calculation within 10 percent. This is unfortunately not close enough to enable us to place
immediate trust in the results obtained by this
approach for discussions of particle shape. It
should be noted that values of the Scherrer

6)
Q).

(16)

Exact (16)
Bragg (17)

"I.

TABLE

4.189

3.770

I. Scherrer

constants.

1.333
1.200

3.477
3.630

1.107
1.155

in the Scherrer equation is then the diameter of


the spherical particle. For particles of the same volume,
the value obtained by Scherrer would correspond to 1.15
(cf. Murdock, reference 4, p. 20) and I, Eq. (25). Note also
that 3 is redefined.

SCHERRER FORM ULA


constant can be calculated by the Bragg method
for all the functions given in the previous paper
(I, Table II and Eq. (23)) and that the integral
breadth 8 and the Scherrer constant E can be
obtained to this approximation for any particle
for which the curve of cross-sectional areas can
be set up. A detailed examination of the nature
of the Bragg approximation should therefore be
of great value for discussions of particle shape.
It is hoped that in the future, tangent plane
calculations can be carried out for some of the
exact interference functions which have been
obtained (I, Tables I and II).
We can now test the approximation functions
(12ac). We note that both the exact (6) and the
tangent plane (9) integrals can be evaluated for
functions of this type. We shall confine ourselves
here to the tangent plane integral since it has
been set up under more general conditions and
has been shown to lead to almost the same results
as the exact integral. We may use both the v.
Laue and the Jones method for evaluating the
constants co. The values obtained and the corresponding values of the Scherrer constants X
and XI~2 are given in Table II.
These results show that the use of approximation functions in integrations of the type involved
in deriving the Scherrer formula are by no means
to be trusted. We see that the Laue approximation which has been made to give identical
values for the integral breadth in the reciprocal
space leads to very divergent values for the
constant E, one measure of the integral breadth
in the scale of x, while the values for X~~2
(another measure of that breadth) show relatively good agreement with one another. This
latter result can only be described as fortuitous.
The Jones approach forces the equality of the
constant E for the various functions, but leads
of course to quite different values for the integral
breadth in the reciprocal space, and' to quite
discordant values for the constant E~I2. The
reason for this lies of course in the behavior of
these functions at infinity. The identity of the
integral breadths is no guarantee for the identity
of any other property of these curves. Furthermore, two curves which have been matched by
any of the criteria discussed above, cannot be
expected to agree after being subjected to an

"

"Cf. M. v. Laue, Ann. d. Physik 26, 59 (1936).

TABLE

II. Sckerrer

V. LAUE

fg(A;)

f2(A;)
fs(~s)

constants from approximation

JONES

0.455
0.550
0.305

0.423
0.750

0 255

V. LAUE

JONES

1.333
1.241 1.333
1.817 1.333
1.115 1.333

functions.

V. LAUE

JONES

1.107
1.166 1.252
1.157 0.849
1.133 1.353

infinite integration of the nature of that involved


in the derivation of the Scherrer formula.
In the discussion of the effect of the size and
shape of the sample and of the nature of the
radiation such as has been given by v. Laue' and
Brill and Pelzer" it seems that the use of approximation functions is inevitable. The results of the
present paper indicate that the conclusions drawn
from such an analysis should be examined critically. It seems that the approximation functions
must be chosen with much more care, particularly
with respect to their behavior at infinity. Perhaps the more purely empirical approach suggested by Jones' will provide the best means for
making allowance for the dimensions of the
sample. It is, however, of great importance that
the fundamental discussion given by v. Laue' be
placed on a sound basis.
CONCLV SION

In the approximation methods of v. Laue' and


Jones' the use of the expression (10b) is equivalent to the assumption of an ellipsoidal particle.
They compare their approximation functions for
such a particle with the exact interference function for a particle which is a parallepipedon in
shape, and obtain various values for co which lead
to the values for the Scherrer constant given in
Table II. The results of the present paper seem
to indicate that it would be safest to assume that
the particle is ellipsoidal at the start and to use
the values for the Scherrer constant obtained for
the exact function for an ellipsoidal particle.
Then, in a case in which the Scherrer formula is
directly applicable, or after the application of an
empirical analysis such as that given by Jones, '
the values of obtained from a given crystalloof the
graphic form would be representative
direction
the
in
mean dimension of the particle
"R. Brill, Zeits. f. Krist. 68, 387 (1928); R. Brill and

H. Pelzer, ibid. 72, 398 (1929); 74, 147 (1930), etc.

RANDALL

FUSON,

982

normal to the faces of that form. If the assumption of an ellipsoidal particle is well-founded, the
breadth should vary in the way suggested by
(lob). If not, the departures from the ellipsoidal
shape can be examined in the light of Table II
of the preceding paper' and an indication of the
actual particle shape can be obtained.

NOVEM BER 15, 1939

AND

DENNISON

The work reported in this and the preceding


paper was commenced at the Massachusetts
Institute of Technology. It is a pleasure for the
writer at this time to express his thanks to Professor J. C. Slater for the privilege of working in
his laboratory, and to Professor B. E. Warren for
many valuable discussions.

PHYSI CAL REVIEW

VOLUM E 56

The Far Infra-Red Absorption Spectrum and the Rotational Structure of


the Heavy Water Vapor Molecule
NELsoN FUsoNi
University

H, M.

RANDALL,

AND

D. M.

DENNIsoN

of Mickif, un, Ann Arbor, 3Eichigan


(Received September 7, 1939)

An investigation has been made of the spectrum of heavy


water vapor (D20) in the region from 23@, to 135@. The
instrument used was a self-recording spectrograph of large
aperture, using echelette gratings, vacuum thermopile, and
a system of filters, shutters, and reststrohlen plates to remove higher order spectral impurity. The radiation path
in the spectrograph could be evacuated. From this research
the experimental positions and relative intensities of 210
pure rotation absorption frequencies were obtained. Absorption maxima were located with an accuracy of about
0.05 cm '. Lines 0.5 cm I apart were partially resolved,
higher resolution and dispersion being of little advantage
since the true width of these absorption lines was of this
same order of magnitude. The energy levels of a zeroth-

I.

020. In more recent years the extension of


the spectrum of H~O into the region of pure rotation frequencies has been made. 4 One of these
latter studies' succeeded in establishing the rotational energy levels of the ground state with high
accuracy. The present investigation was undertaken in the attempt to parallel this H&O analysis
with a similar study of D20.
The problem has been to map the far infra-red
pure rotation absorption spectrum of heavy
362 (1934); E. Bartholome and K. Clusius, Zeits. f. Elec.
upon

INTRODUCTION

amount of work has been done on


A LARGE
the spectral analysis of ordinary water
in

its vapor state. Information concerning the molecule as a whole is largely limited to studies of the
absorption frequencies occurring in the infra-red
spectrum. The
region of the electromagnetic
vibration-rotation
bands of H~0 have been
examined a number of times. ' Since the discovery2
of the heavy hvdrogen isotope, H' 'often called
deuterium, D), similar studies' have been made
f,

~ Now

at Rutgers University.
' W. W. Sleator, Astrophys.

J. 48, 125 (1918); W. W.


Sleator and E. R. Phelps, Astrophys. J. 52, 28 (1925);
R. Mecke, Zeits. f. Physik 81, 313, 445, 456 (1933); L. G.
Bonner, Phys. Rev. 46, 458 (1935); E. Ganz, Ann. d.
Physik 28, 445 (1937).
~
Urey, Brickwedde and Murphy, Phys. Rev. 40, 1
(1932).

order approximation to the D20 asymmetric rotator mole= 11, and


cule were computed through quantum number
corrected for zero point vibration and centrifugal stretching
in the ground state. A comparison of the positions and
intensities- of the experimental data with those of the
transitions between these "key" levels showed a rather good
agreement. These levels were therefore corrected to fit the
data, and checked for consistancy by means of series
regularities and combination relations. In all, 111 distinct
energy levels based on the experimental data were computed. A graph of the experimental data contrasted wit'h a
similar graph of the transitions based on these corrected
levels gives a clear picture of the success of the analysis.

3
J. W. Ellis and B. W. Sorge, J. Chem. Phys. 2, 559
(1934); T. Shidei, Phys. and Math. Soc. of Japan Proc. 10,

Chem. 40, 529 (1934); E. F. Barker and W. %. Sleator, J.


Chem. Phys. 3, 660 (1935);L. Kellner, Proc. Roy. Soc. 1SQ,

a410 (1937).
4 H.
Rubens, Berliner Ber. S: 8 (1931);H. Witt, Zeits. f.
Physik 28, 245 (1924); M. Czerny, Zeits. f. Physik 34, 232
(1925); J. Kuhne, Zeits. f. Physik 84, 722 (1933);N. Wright
and H. M. Randall, Phys. Rev. 44, 391. (1933); Barnes,
Benedict and Lewis, Phys. Rev. 4"/, 918 (1935).
~ Randall,
Dennison, Ginsberg and Weber, Phys. Rev.
52, 160 (1937) (this paper will hereafter be referred to as
RDGW).

You might also like