The Scherrer Formula Revisited

Detlef-M. Smilgies, CHESS

The ubiquitous Scherrer formula [P. Scherrer, Göttinger Nachrichten (1918)]

D = {K*lam} / {cos(th) dtth}

has been the workhorse of  x-ray diffraction methods for determining the average size of scattering objects D from the width (fwhm) of the diffraction peak Dtth, and aside from Bragg's law may be the most commonly referred to formula in x-ray diffraction. The original derivation has been refined in various other papers and textbooks. Nowadays the prefactor K is typically used between the bounds 0.9 < K <1. Using the modern language of Fourier transforms we provide an easy way to understand the shape of the Bragg reflection and to derive a value for K.

Warren provides a straightforward, effective way to K: using the scattering from a finite number of Bragg planes, K is determined from the FWHM of the sin(Nx)/sin(x) function. In a more formal way, that lends itself to further generalization, we shall rederive the result in the language of Fourier transforms:

For small crystallites of micron size, i.e. smaller than the dynamic extinction length, the kinematic approximation can be safely applied. In this case the scattering intensity distribution is given the squared modulus of the Fourier-transform of the electron density. For an infinitely extended lattice this Fourier transform is an infinite series of delta functions centered on the reciprocal lattice with the prefactor for each reflection given by the structure factor of the unit cell. In order to describe a finite crystal we follow Patterson and introduce a shape function which is 1 inside its bounds, and 0 elsewhere. The product of the shape function and the infinite lattice yields a finite crystal of a shape given by the shape function. Using the convolution theoreme, the intensity distribution of a reflection of a finite crystal is then given by the convolution of the delta functions at the reciprocal lattice points and the Fourier transform of the shape function. Thus the shape function Fourier transform defines the shape and width of the reflections.

A finite slab is cut out of an infinite number of Bragg planes by multiplying the Bragg planes with a cut-off function

rect(z) = 1, z2 < R2; 0

The Fourier transform of rect(z) is the sin(x)/x function with x=qz. The product in real space transforms to a convolution in reciprocal space and the sharp Bragg peaks get smeared out in z-direction by sin(x)/x. From this construction it is clear that this approach should work well for lamellar systems.

Similarly we can define a circular and a spherical domain:

circ(x,y) = 1, x2+y2<R2; 0

sphere(x,y,z) = 1,x2+y2+z2<R2;0

The Fourier transforms of these cut-off functions are well-known in small-angle scattering as the formfactor of a disk and of a sphere, resp.

Circ(xi)= {2J1(xi)/xi}

...

Anisotropy

The scattering from an arbitrary ellipsoid can be obtained by realizing that any linear transformation S (rotation, stretch, shear) applied to the shape will be related to the distortion of the 1/2 value ellipsoid in q-space:

  xi = qR  = qT R = qT S-1 S R = (S-1T q)T (SR)

i.e. when the rela space distortion is given by S

The scattering intensities are then given by the modulus of the form factors and are plotted in Fig 1. Note that the period of the characteristic cut-off oscillations changes as a function of dimension, where we note that the 3 cut-off functions introduced correspond to the abstract spheres in 1,2, and 3 dimensions.

K-factor ....

spherical grains: K=1.15 most common case, usually K<=1 is used

anisotropic shapes in soft materials:


less anisotropic:
ellipsoid: sphere(:  xi=sqrt(a2 x2 + b2 y2 + c2z2) : a,b,c=half axes

1) oriented / crystalline: In anisotropic lattices often the particle shape is anisotropic, too, as the growth or crystallization along non-equivalent directions is different. For the characterization of anisotropic shapes in crystalline materials, only radial scans should be used for Scherrer analysis, since tangential scans are effected by mosaicity and texture, and do not reflect pure shape information. As soon as the powder reflections are indexed, the half-axes of the shape ellipsoid can be determined from the low-index reflections. Care need to be taken that the powder is not strained, as this will also effect the peak widths. A good indication for strain is, if the peak width changes in a series of reflections along the same radial vector, such as (00L) or (HH0). If shape is the only determining factor, the FWHM of such reflections should be identical. The width from a variety of different radial directions can then be fitted to a shape ellipsoid x2/a2 + y2/b2 + z2/c2 = 1 with a,b,c, the half axes of the shape ellepsoid. Rotations may have to be includud, if the shape does not have a symmetry axis parallel to any low-index reflection. The K value 1.15 for the spherical domains would apply here.   

anisotropic shape
For more anisotropic domains rod-like : J1 * sinc, H >> R  (for instance carbon nanotubes and other fibers)
disk-like: J1 * sinc,  R >> H (for instance clay platelets) In these cases the K factor for rods (0.9) applies for meridional reflections along the axes (rods) of perpendicular to the basal plane (platelets), while in the equatorial direction the disk factor K=1.05 should be applied.

As special but common case in soft materials are lamellae and hexagonal cylinders, for instance in block copolymers or surfactant mesophases. The linear K=0.9 value should be applied for the lamellar reflections, while the disk K=1.05 value would apply for the hexagonal reflections.





The derivation given here simplifies Patterson's original treatment making use of the modern language of Fourier transforms, but reproduces his result for a spherical domain of K=1.15. A more or less spherical shape may be the most commonly encountered domain shape. Hence we would like to make a point that, unless there is strong reason to believe that the domain shape is strongly non-spherical, Patterson's K value of 1.15 should be applied. In special situations where the scattering object can be considered essentially 1D or 2D use of the other factors can be made.





C. C. Murdock, Phys. Rev. 35, 8 (1930)
B. E. Warren, Zeitschr. f. Krist. 99, 448 (1938)
A. L. Patterson, "The Diffraction of X-Rays by Small Crystalline Particles",  Phys. Rev. 56, 972 - 977 (1939)
A. L. Patterson, "The Scherrer Formula for X-Ray Particle Size Determination", Phys. Rev. 56, 978 - 982 (1939)
B. E. Warren, "X-ray diffraction", Dover