Waves diffracted from a crystal gives rise to diffraction spots, where each spot corresponds to a point

in the reciprocal lattice and represents a wave with a particular amplitude and a relative phase. But in
practice, the detector counts the number of photons hitting it which gives an intensity for every “pixel”
on the detector. Intensity is proportional to the square of the amplitude of the diffracted wave, but it
says nothing about the phase. As such, phase information is lost. This is a big problem, because in order
to calculate the inverse FT, both amplitude and phase information at every spot is needed:

1
𝜌(𝑥𝑦𝑧) = ∑|𝐹(ℎ𝑘𝑙)| ∙ 𝑒 −2𝜋𝑖[ℎ𝑥+𝑘𝑦+𝑙𝑧−𝛷(ℎ𝑘𝑙)]
𝑉
ℎ𝑘𝑙

Here 𝛷(ℎ𝑘𝑙) is the phases, which are lost during data collection and 𝐹(ℎ𝑘𝑙) are amplitudes which are
calculated from the intensities obtained during data collection. 𝜌(𝑥𝑦𝑧) is the function of electron density
in real space.

The following diagram is used as an aid to explain why exactly phases are needed:

On the left, the four balls represent the positions of atoms with respect to the Bragg planes given by a
particular diffraction angle. The relative phase depends on the relative distance of the atoms between
the planes that define a phase angle of 0. On the right, is shown the atoms’ contributions to the overall
scattering. The magnitude of the complex number is the amplitude while the angle made with the real
line is the phase. (𝑧 = 𝐴 ∙ 𝑒 𝑖𝛷 ; Amplitude = |𝑧| = 𝐴, Phase = 𝛷) Here the complex numbers are
colour-coded in the Argand plane and placed head to tail to show how their addition results in the overall
phase and amplitude (black arrow). The complex number representing the overall scattering (black
arrow) from a particular set of Bragg planes is termed the structure factor, 𝐹. Applying the inverse FT
to the structure factors we obtain an electron density map. From this, it is clear that if we lack phase
information, we have incomplete information about the structure factors and so we cannot work back
to electron density. This is, in essence, the phase problem.

Scientists have come up with several methods to overcome the phase problem, i.e. methods to recover
phases. One of these is molecular replacement (MR). MR works because if a “guess” places the atoms
in about the right place, then the calculated phases will be approximately correct and a useful electron
density map can be computed by combining the observed amplitudes with the calculated phases. Here
the “guess” is made from the structure of a closely-related protein whose structure has already been
determined. To successfully carry out MR, the model structure needs to be placed with the correct
orientation and position in the unknown unit cell. Orientation requires three rotation angles, whilst
placement requires three translational parameters. Here a function called the Patterson map (which
results from the inverse FT of only intensities – or is it amplitudes?) is useful in determining these. The
Patterson function generates a map of vectors between atoms. Intramolecular vectors depend only on
the orientation of the molecule, so these are used to determine the rotation function. Intermolecular
vectors depend on both orientation and position, so once orientation is determined from the
intramolecular vectors, the intermolecular vectors are used to determine the translation function. After
the model is correctly placed, phases can be calculated and combined with the observed amplitudes to
give a map of electron density for the protein under investigation.

There exist also several direct methods, namely isomorphous replacement and anomalous scattering.
The main concept underlying isomorphous replacement is the idea that perturbation of the structure
factors allows us to make some deductions about possible phase values. Heavy atoms are used in
isomorphous replacement because they contain a much greater number of electrons than those found in
proteins. It can be shown that the contribution to overall scattered intensity of an atom is proportional
to the number of electrons it contains. The change in intensity resulting from the addition of one heavy
atom to a molecular structure is hence significant enough to measure. By comparing the differences in
scattered intensities of a normal crystal and one containing heavy atom(s), one can compute a Patterson
map which can be used to compute the location of heavy atoms in the crystal and hence their
contribution to the structure factors. From this, possible phase values can be calculated and tested with
the amplitude information to generate an electron density map.

Meanwhile, anomalous dispersion is based on the use of anomalous scatters, atoms that do not obey
Friedel’s law. In essence, Friedel’s law states that for diffraction spots with opposite Miller indices –
that is, if one is (ℎ𝑘𝑙) the other is (ℎ̅𝑘̅ 𝑙)̅ – the structure factors are complex conjugates. That is, the
amplitude component is the same, but the phase is reversed (seen as a horizontal reflection of the
complex number on the Argand plane). Diffraction spots with opposite Miller indices are appropriately
named Friedel mates. Anomalous scatters do not obey Friedel’s law because the frequency of oscillation
induced by the x-ray radiation is too similar with their natural frequency of oscillation. This causes a
shift in both amplitude and phase of the induced oscillation (for reasons I do not understand), a
phenomenon called anomalous scattering. The result of this is that the anomalous scatter will affect the
structure factors of Friedel mates asymmetrically. Because the anomalous scattering effect depends on
the frequency of oscillation being similar to the natural frequency of oscillation of the anomalous
scatter, by collecting data at several wavelengths, it is possible to obtain phase information. This is
known as Multiple Anomalous Dispersion (MAD).