Determination of Atomic Positions and Refinement of the Structure

6.1 Trial Structure from Qualitative Clues

In simple cases a trial structure can often be constructed on the basis of the variety of qualitative
clues, without any calculation being required. The unit cell size, and particular symmetry, plus an
approximate idea of the molecular shape, often led to a reasonable trial structure. If a structure is
composed of nearly planar molecules arranged in parallel layers, the structure problem is
essentially a two-dimensional one and can usually be solved by packing considerations. The
molecules must avoid the point symmetry elements, they must not approach too close to one
another, and they must leave very little empty space. It is in fact very impressive to see how
efficiently space is filled in a structure. Such a procedure may yield two or three possible
arrangements to test. If some atoms are restricted to spatial positions then the problem is simpler
because we have merely to determine how the molecules are oriented around these spatial
positions.

A semi-quantitative application of these packing considerations is has proved to be useful

for crystals composed of molecules of known shape. The essential idea is to rotate and translate
the independent molecule in the unit cell until the sum of the intermolecular repulsions is
minimized. This process is carried out automatically by a computer program and is rapidly
successful at least for crystals composed of rigid hydrocarbon molecules. In more general cases,
packing considerations are insufficient to indicate the structure, and additional information is
necessary.

Experience and knowledge of the structures of related materials are obviously helpful in
guessing reasonable trial structures. For example, hydrogen bonds may be expected wherever
they are possible, and this suggests the parts of adjacent molecules that should approach closely
together. Also the anisotropy of almost any quantity may provide a clue to the molecular
orientation. In most cases, still more powerful clues have to be included to arrive at a trial
structure such as those from the` Patterson function.

6.2 The Patterson Function:

One powerful method of the structure determination is the study of the Patterson function. With
no assumptions about the structure, we can construct the following function, discovered by
Patterson (1934-1935):

( ) ∑ | ( )| [ ( )]
 ∑ | ( )| ( ) (6.1)

This is just like the Fourier series representation of the electron density ( ) except
that | | replaces , as | ( )| | ( ̅ ̅ )|
̅ , the imaginary terms in the exponential functions
necessarily cancel, so only the real cosine terms are included in the second version. The fact that
the variables are taken as rather than as serves merely to indicate a difference in the
physical interpretation. The first important point is that the Patterson function can be evaluated
directly from experimental measurements, without any knowledge of the structure that has
produced the diffraction pattern.

6.2.1 SIGNIFICANCE OF THE PATTERSON FUNCTION: The simplest approach is to ask

what is the nature of the hypothetical structure that would have the | | values for its structure
factors; ( ) would be the density of scattering matter in this structure. Now, in terms of the
structure factors of the unknown actual structure for which

( ⃗) ∑ ( ⃗ ⃗⃗) (6.2)

we have

| ( ⃗ )| ( ⃗) ( ⃗) ∑ [ ⃗ (⃗ ⃗ )] ( )

This expression has the same functional form as that for ( ⃗ ) in Eq.6.2. Accordingly,
| ( ⃗ )| is the structure factor for the hypothetical structure composed of pseudo atoms with
atomic scattering factors located at positions ⃗ ⃗ ⃗ ⃗⃗ ⃗ where
and so on. Consequently, ( ) represents this hypothetical structure; so it exhibits peaks
that correspond to all inter atomic vectors ⃗ ⃗ in the real structure (all displaced to a common
origin) and that are of a size that increases with increasing scattering factors of the two atoms
joined by the inter atomic vector. Since those terms in the double sum with correspond to
zero-length inter atomic vectors, there is a large peak at the origin of the Patterson function.

A two-dimensional schematic illustration of the relation of the Patterson function to the

electron density is given in Fig. 6.1. One unit cell of the structure is outlined, and the contents
of the eight surrounding cells are shown; the small circles represent the centers of the atoms. In
the Patterson function the point at the center may be regarded as the origin. The double circle
there represents the zero-length inter-atomic vectors that are between each of the two atoms in
the unit cell of the structure and itself. Other double circles arise from inter-atomic vectors
between atoms related by lattice translations. The single circles in the Patterson function
represent vectors between two atoms that are not related by lattice translations. The single circles
are approximately one half the height of the double circles because they represent single vectors
rather than overlapping pairs. Even from this simple example it is apparent that the Patterson
function is periodic with the same unit cell as the structure, and that large peaks build up at the
origin(s).

Fig. 6.1: Relation b/w structure ρ and Patterson Function Pfor a simple 2-D case

6.3 The Heavy Atom Method

For crystals in which one or a few atoms (the heavy atoms) scatter more strongly than any of the
other atoms (the light atoms) several techniques are available to aid the in the determination of
the structure. All require the use of the Patterson function to locate the heavy atoms. The inter
atomic vectors between pairs of heavy atoms stand out prominently in the Patterson function, so
the positions of these atoms can be readily determined if there are only one or two independent
heavy atoms. We assume then that the heavy atom positions are known.

The simplest way to use the heavy atom positions applies to cases in which the heavy
atoms dominate the scattering from the crystal, that is, when the structure factor contribution
from all the light atoms is less than the contribution from the heavy atoms, for most reflections.
Then the heavy atom contribution usually determines the sign of a reflection, in centro-
symmetric structures, or gives an approximate value for the phase, in non centro-symmetric
structures. A Fourier synthesis of the electron density, with phases determined by the heavy
atoms, usually reveals the rest of the structure. In this case the light atoms cannot be located as
precisely as usual because the structure factors are relatively insensitive to their positions. The
majority of reported structure determinations have been accomplished by the heavy atom
method.

6.4 Agreement Of Observed And Calculated Structure Factors:

If the magnitudes of the structure factors calculated for some trial structure are in sufficiently
good agreement with the observed magnitudes, the trial structure is likely to be correct. We have
to examine what is “sufficiently good agreement” and find criteria to determine when a
modification of a structure is an improvement. Agreement must be better for reflections at small
scattering angles than at large ones, because the small angle reflections are from planes with
large inter-planar spacing so they are less sensitive to atomic positions than are reflections at
larger angles. Large values of | | the observed magnitude, must be matched by large values of
| |, the magnitude of the calculated value, but small | | need not be matched so closely by
small | |. The reason is that a large magnitude of the structure factor can only be achieved by
having many atoms contributing in phase, for example, by ∑ ⃗ ⃗ with most of
the cosine values near +1; then a small change in ⃗⃗ causes only a small change in . On the
other hand, a small magnitude requires either exact cancellation of contributions from many
atoms or most cosine values near 0, and in either case a small change in any atomic position
makes a relatively large change in the structure factor.

A quantitative measure of the agreement between | | and | | is needed, as well as the

qualitative considerations just outlined. The conventional measure is the discrepancy index

∑|| | | ||
∑| |

where the sums are over all reflections, or over all reflections of measurable intensity.
This quantity has no precise theoretical significance, but at least it should normally become
smaller as the agreement improves. Because it is easy to evaluate, it is commonly used to follow
the progress of refinement from a crude trial structure to the accurate final structure .