One group of white lines which has not been fully taken advantage of are those of the 3rd transition series metals from Hf upwards in the Periodic Table . Fig. shows and values calculated from fluorescence measurements which were performed on the X31 beam-line across several of the edges of the 3rd transition series metals plus that for Lu. The measurements were made using the Si(311) monochromator with a bandpass of . The samples are listed in the figure caption.
Figure: and values for the edges of some of the 3rd transition series metals plus Lu. The values were calculated from fluorescence measurements on samples of (from left to right) , , , , , - and . The energy scale is not calibrated and is therefore only accurate to within a few . The three elements missing from the diagram are Ta, Os and Au but one could expect them to display intermediate behaviour compared to their neighbours. A trend in the maximum height of the white lines is visible in the transition series metals with Hf having the largest feature and Hg having the smallest. The trend is representative of the change in the density of empty states in the 5d band of the transition metals. Hf has only two electron in its 5d band and as we progress through the series one electron is added per element. Pt has nine electrons and Hg has a filled band.
All of the edges shown in Fig. except for Hg and Pt show large white line features. Also important is the fact that the presence of two steep inflection points on the white lines mean that the corresponding curves possess a local maximum and minimum. The large differences between the values of at these two positions, denoted here as in principle allows the possibility of phasing reflections by making measurements at three energies separated only by a few as suggested by Phillips  where the particular case of caesium white lines was discussed. Values for the maximum at the white line peak, , and the maximum observed difference in due to the two inflection points, are tabulated in Table for the data shown in Fig. where a significant white line is present. Also shown are the FWHM of the white lines (measured by taking the base of the line at the theoretical value of just above the edge in energy) and the energy difference between the local maxima and minima in the curves, .
Table: Parameters describing the height and widths of the white lines for some of the 3rd transition series metals where white lines are present. Values are given for the FWHM line width, the energy difference and difference between the maximum and minimum values of the curve about the white line and the maximum value of at the white line.
It is observed that Hafnium presents the best possibility for obtaining large contrast in the anomalous scattering correction terms in the immediate vicinity of the edge. The absorption edge is characteristic of a transition. The size of the white line at the edge is dependent on the density of empty states in the 5d band. The trend observed in the heights of the white lines from Hf through to Hg is indicative of the increasing number of electrons in the 5d orbital from Hf, with two electrons, to Hg, with 10 electrons (a filled orbital). The width and height of the white lines are however also dependent on the chemical state of the ions. The number and type of ligands bound to the metals can act to distort the outer electronic bands and this is reflected in the white line.
The 3rd transition series metals from Ir downwards have generally not been used for isomorphous replacement although there have been a few cases involving derivatives with the complex ions ,  and   . The ion is in fact known to bind very strongly to cis-diol groups present in DNA and RNA structures.
The reason for the preferential use of heavy atoms such as Pt, Au and those with still larger atomic numbers for MIR is not necessarily one of specificity of binding or good isomorphism between the derivative and native structures since several of the slightly lighter transition metals are known to bind in a similar fashion  . The reason is more likely to be that these metals have more electrons and produce greater contrast in the scattering of the native and derivative structures.
Paradoxically the metals of the periodic table from Ir downwards in atomic number are the most promising for the application of the MAD techniques and if derivatives can be obtained then criteria for isomorphism of the native and derivative structures can be slackened considerably since all measurements in the MAD method can be made on the same crystal form and species if not the same crystal sample. The edges of these metals compared to those of the lanthanides occur within a higher range of energy () making them favourable because of the reduced absorption of the X-rays by the sample crystal and by air.
The large rates of change of and with X-ray energy around the white lines mean that for optimised diffraction work, at the white line maximum and the inflection points, the requirements for accurate X-ray energy selection and stability are great. For example at the rising inflection point of the Hf edge the rate of change of with energy is . A similar rate of change of is observed at the white line maximum.
With these thoughts in mind it seemed suitable to attempt to obtain derivative protein crystals using one of the above mention heavy atoms from Hf to Ir for the purpose of assessing the use of the calibration apparatus for energy stabilisation and to decide on a suitable strategy for X-ray energy selection given the typical shapes of the curves when a significant white line is present, an important choice appearing to be whether the third wavelength chosen to obtain contrast in the value should be at the falling inflection point, where the necessary absorption corrections to the data may be more straightforward given the proximity of the three experimental energies (the case for the Tb experiment of Kahn et al ), or at a point far above the edge where greater contrast can be achieved (as in the Ho work of Weis et al ). In addition one question which remains unanswered is whether a narrow band pass approach to MAD work should be taken at all. Would it be better simply to use a broader bandpass at points away from the absorption edge and take advantage of the comparatively larger flux on the sample?