Molecular weights of some oxides


The chemical composition of rocks and minerals is usually reported in weight percent of oxides of relevant major rock-forming elements and minor elements. Here's an example from my "Impact melts and granulites in the lunar meteorite PCA 02007" abstract:

Analysis reported in weight percent oxides

Why are chemical compositions reported in weight percent oxides? One method of wet chemical analysis, gravimetric analysis, involves precipitating oxides of various elements from solutions of rock powder and weighing the precipitate (so it's natural to report measurements in weight percent oxides--percent since we're interested in concentrations, not absolute amounts, of elements in a sample). At present, most mineral analyses are done using an electron microprobe. Why then continue to report compositions in weight percent oxides? I'll explain generally how electron microprobes work and then, in light of their operation, give what I think is a compelling explanation.

The electron microprobe measures x-ray counts at some wavelength. (Inner shell electrons of an atom of some element are ejected by an electron beam; outer shell electrons "fall" to take their place and emit an x-ray with an energy, or equivalently a wavelength, characteristic of that element). The higher the concentration of some element, the more x-ray counts per second at some characteristic wavelength of that element; that is, the number of counts/second at a characteristic wavelength of some element from some sample is directly proportional to the number of counts/second at the same wavelength from another sample, and the proportionality constant is simply the ratio of the concentrations of the element in each sample. If we measure the number of counts/second at a characteristic wavelength of some element on a "standard"--a chemical compound of known composition--we can determine the concentration of that element in a sample of unknown composition by measuring the counts/second at the same characteristic wavelength. In practice, to accurately measure composition, there are a number of physical, or "matrix", corrections that have to be applied, which take into account x-ray absorption, x-ray fluorescence, x-ray background, etc.

Electron microprobe analysis (EMPA) has several major advantages over wet chemical analysis. 1) It's much faster. Typical count times at a characteristic wavelength are around 15-30 seconds, microprobes usually have four or five spectrometers which can count concurrently, and usually only 10-15 elements are analyzed; counting the time for the crystals in wavelength-dispersive spectrometers to shift positions to measure different characteristic wavelengths, a single analysis takes 2-3 minutes. Compare this to a day for a wet chemical analysis. 2) Since the electron beam can be focused to a 1 micron pencil, and the area excited by the electron beam and emitting x-rays is not (necessarily) much larger, the electron microprobe is able to analyze the chemical composition of domains only several microns in size: this is important since minerals are frequently inhomogenous at larger scales (due to zoning, exsolution, or alteration). In fact, even rocks can be inhomogenous at this scale: the matrix of impact-derived breccias and some meteorites is made of grains of sub-micron size, and presolar grains in meteorites can be less than 10 nanometers in size! 3) EMPA is non-destructive: a precious lunar or meteoritic sample doesn't have to be powdered to determine its chemical composition.

One major failing of the electron microprobe is that it's difficult to directly measure the concentrations of light elements in a sample, including oxygen. Oxygen is an extremely important element; it makes up nearly a third of the Earth by weight, and is present in nearly all rock-forming minerals (all silicates and oxides). It's by far the most important anion in most minerals. In many types of rocks (including some meteorites and moon rocks) there are no other anions present. The atoms in the unit cell making up any given mineral are charge-balanced (otherwise the mineral would have an enormous net charge). So for many rocks and minerals (particularly olivine, pyroxene, and the feldspars) we can assume that each species of cation gets paired with an appropriate number of oxygen anions to maintain charge balance. So, for instance, if we measure the weight percent of Si (almost certainly existing as Si4+) in a sample, we can convert that into the mole fraction Si; charge balance requires 2 O2- for each Si4+, so we know that there's 2x the mole fraction of Si of O in the sample neutralizing the Si4+, and we can back-calculate into moles and finally weight percent of SiO2, a charge-neutral component making up the rock. We repeat this for each cation and end up with a list of oxide weight percents. Now, if our analysis was good, and our assumptions about what we were analyzing was justified, these oxides, these charge-neutral components, will sum to 100 wt. % (since they cover every cation, and account for all the oxygen present).

Of course, no analysis is perfect: there's experimental error, some minor or trace cations aren't measured, etc., so the sum probably won't be 100 wt. %. For metals, sulfides, or silicates with H2O, F-, Cl-, etc., the assumption that all anions are oxygen is unjustified, so the oxide sum won't be 100 wt. %, either. For minerals with cations existing with different valences than assumed, the oxide sum again won't be 100 wt. % (although, assuming a perfect analysis, the deviation from 100 wt. % can be used to back-calculate how much of an element exists in one valence state of two--Fe2+ vs. Fe3+, for example).

In short, analyses are reported in weight percent oxides because doing so 1) is traditional, and 2) makes it easy to pick out the bad analyses (since they're different from 100 wt. %--usually every analysis below 97-98 wt. % is considered bad). Since oxygen is usually determined by stoichiometry (the method I described above) rather than measured independently with the microprobe, it's less misleading to report concentrations in oxides (which suggests oxygen was determined by stoichiometry) than report concentrations of all elements with a separate concentration for oxygen (which suggests oxygen was measured independently). I hope that's a satisfying answer to the question posed above.

It's useful to have an accurate table of the molecular weights of oxides of the rock-forming elements for processing microprobe data to determine the crystal chemistry of minerals. Below I've reproduced the table of oxide weights found in the first edition of Deer, Howie, and Zussman's "An Introduction to the Rock Forming Minerals". The value of O used in this table is exactly 16 (compare S, 32.066, and SO3, 80.066); probably the most precise molecular weight available was used for each element. I added Ti2O3 myself: trivalent Ti is rare in nature, but it comes up in my senior thesis, so why not add it to these tables? There are actually two tables below. The first table lists only oxides of the major rock forming elements. The second table lists oxides of all major and minor rock forming elements.

OxideMolecular weight
Al2O3101.94
CaO56.08
FeO71.85
Fe2O3159.70
K2O94.20
MgO40.32
Na2O61.982
SiO260.09

OxideMolecular weight
Al2O3101.94
B2O369.64
BaO153.36
BeO25.013
CO244.010
CaO56.08
CeO2172.13
Ce2O3328.26
CoO74.94
Cr2O3152.02
CuO79.54
FeO71.85
Fe2O3159.70
H2O18.016
HfO2210.5
K2O94.20
La2O3325.84
Li2O29.88
MgO40.32
MnO70.94
MnO286.94
Mn3O4214.42
Na2O61.982
NiO74.71
Nb2O5265.82
P2O5141.95
PbO223.21
Rb2O186.96
SO380.066
Sc2O3137.92
SiO260.09
SnO134.70
SrO103.63
Ta2O5441.90
ThO2264.05
TiO279.90
Ti2O3143.7
UO2270.07
U3O8842.21
V2O5181.90
Y2O3225.84
ZnO81.38
ZrO2123.22


Back to wvaughan.org
Will Vaughan. Last revision December 29, 2010.