In the olivine structure, there are two slightly different sites: M1 and M2. The M1-O bond has a length 2.16 Å. M2 is slightly larger (2.19 Å) and is slightly distorted from the regular geometry. It is the usual site for Mn and Ca, while M1 is the usual site fro Ni. Fe has an intermediate size, however, so may be found in either site. At high temperature, high thermal energy should make it possible to find Fe in either site. At lower temperature, however, it might preferentially prefer one or the other, in which case it could be used as a thermometer.

Experiments were first performed at MIT in 1964 to plot the absorption of olivine from the near UV to the near IR. A new technique was used, based on the assumption that the occupancy site of Fe could be determined by examining the spectrum produced by Fe+2's green color. The presence of two different sites would be expected to split the octahedral sites into 5 discrete levels (more closely spaced for M1), which could then be detected by the slight difference in absorbance. Two double peaks were observed, and they were attributed to Fe in the M1 and M2 sites. Because the M2 absorption was much stronger, it was concluded that M2 M1 for Fe.

The problem was subsequently studied at Virginia Polytechnic Institute and the University of Chicago using X-ray diffraction. In this experiment, olivine crystals were irradiated with 60
m crystals. Because of a quantum mechanical parameter known as the structure factor, Fe is predicted to scatter X-rays better than Mg. An application of Bragg's law would then lead to the relative occupancies. These experiments found M1 = M2.

A third approach was pursued by the Carnegie Institute Geophysical Lab in Washington, DC. The experimenters made use of the Mössbauer effect. 57Co beta decays into 57Fe, which subsequently emits a gamma ray. Because the entire crystal recoils, the gamma ray is emitted with an energy slightly smaller than predicted. Since this gamma ray does not have the correct energy, it cannot be absorbed by 57Fe at rest. By Doppler shifting the 57Fe, however, an absorption can be obtained. Since the M1 and M2 sites have slightly different energies, they will require slightly different Doppler shifts and can therefore be distinguished in the plot of velocity vs. absorption. These experiments found that M1 M2.

The X-ray experiment was repeated with new refined X-ray methods in the 1970s. These improved the accuracy of the previous experiments, and confirmed the Mössbauer results by finding M1
M2. It was then shown that more light is absorbed in a distorted site than an undistorted one. The absorption curves must therefore be calibrated to account for this fact, so the raw curves tell nothing about absolute ratios.

The result that M1 M2 contradicts expectations, since the M2 site should have a greater crystal field stabilization energy. Since measurements were made at room temperature, however, they were not necessarily representative of the energies at the crystallization temperature 1200 . The effects of raising temperature on the energy of the two sites were studied as a spin-off of the technology developed for the Apollo program. When energy measurements were made at high temperature, the angles changed such that the M1 distortion increased at a faster than that of the M2 site between 24 to 710 . Therefore, extrapolating to higher temperatures, there must be a point where the M1 distortion is greater than the M2 distortion. At crystallization temperature, Fe would therefore crystallize in the lower energy M1 site and be locked in as cooling occurred.

Minerals with an olivine structure include montecellite = CaM₂MgM1SiO₄ and olivine = (Mg,Fe)₂SiO₄.