High-resolution valence-band XPS spectra of the nonconductors quartz and olivine

High-resolution core-level and valence-band x-ray photoelectron spectra (XPS) of the nonconductor silicates, quartz, and two olivines (${[\mathrm{Mg},\mathrm{Fe}]}_2\mathrm{Si}{\mathrm{O}}_4$, Mg rich and Fe rich) have been obtained with the Kratos magnetic confinement charge compensation system which minimizes differential charge broadening. Observed total linewidths are about $1.3\mathrm{eV}$ for several of the peaks in these spectra, much narrower than previously obtained. The quartz and olivine valence-band spectra are very different, even though the dominant contributions to both come from molecular orbitals of $\mathrm{Si}{\mathrm{O}}_4$ groups. High-quality calculations of the band structure using pseudopotential density functional theory with generalized gradient approximation (GGA) for quartz yield a theoretical XPS valence-band spectrum in the Gelius approximation that is in excellent agreement with the experimental spectrum. $\mathrm{GGA}+U$ (where $U$ is the on-site Coulomb energy) calculations on Mg-rich and Fe-rich olivines yield theoretical XPS spectra in good qualitative agreement with experiment. The valence-band spectral contributions are readily assigned using the calculated partial density of states. For example, the Fe $3d$ $t_{2g}$ and $e_g$ orbitals in $M1$ and $M2$ sites of the olivine are located at the top of the olivine valence band and are readily resolved in both observed spectra and theoretical calculations. The valence-band spectrum of quartz is about $3\mathrm{eV}$ more dispersed than the valence bands of the olivines and considerably greater than the $1.4\mathrm{eV}$ difference calculated previously. The difference in dispersion arises mainly from two effects. The dispersion of the quartz valence band is primarily a response to the nature of metal atoms in the second coordination sphere of the central Si atom of $\mathrm{Si}{\mathrm{O}}_4$ tetrahedra. For example, the $\mathrm{Si}-\mathrm{O}$ moiety is more negatively charged in olivine than in quartz because of the transfer of electrons from Mg to $\mathrm{O}-\mathrm{Si}$. The calculations indicate very small Mg $2s$ contributions to the forsterite $\left({\mathrm{Mg}}_2\mathrm{Si}{\mathrm{O}}_4\right)$ valence band. Hence all Mg $2s$ electrons effectively reside on O atoms (O $2p$ orbitals) and the $\mathrm{Mg}-\mathrm{O}$ bond is ionic. Polymerization (network formation) also contributes to dispersion in that it results in large splittings in the Si $3s$ and Si $3p$ partial density of states. Fe $3d$ electrons of olivines are of both “bonding” and “nonbonding” according to the calculations. In the Mg-rich olivine, the great majority of Fe $3d$ electrons are nonbonding and give rise to separate Fe $3d$ $t_{2g}$ and $e_g$ peaks at the top of the valence band. The same no-bonding contributions are observed in Fe-rich olivine, but O $2p$ contributions near the top of the valence band indicate appreciable $\mathrm{Fe}-\mathrm{O}$ bonding.

Figure 1

Core-level XPS spectra of quartz: (a) Si $2p$ and (b) O $1s$. Residuals from the least-squares fit are shown below the spectra.

Figure 2

Core-level spectra of Mg-rich olivine ${\left({\mathrm{Mg}}_{0.87}{\mathrm{Fe}}_{0.13}\right)}_2\mathrm{Si}{\mathrm{O}}_4$: (a) Si $2p$, (b) O $1s$, and (c) Mg $2p$.

Figure 3

The “O $2s$” and valence-band XPS spectra (VBXPS) of (a) quartz and (b) Mg-rich olivine.

Figure 4

The VBXPS spectra (dotted lines) of (a) quartz at two different orientations, (b) Mg-rich olivine, and (c) Fe-rich olivine. The calculated XPS spectra (solid and dashed lines) for the quartz and two olivines of similar composition to the two natural olivines are given above the experimental spectra. The intensity of each state was corrected for cross section and each band broadened by $0.37\mathrm{eV}$ to match the instrumental resolution. The calculated quartz spectrum was aligned to the experimental spectrum of quartz, and the calculated olivine spectra (solid lines) were aligned to the experimental energies of peak 1 (the Fe $3d$ band). The calculated forsterite spectrum (dashed line) was aligned to the experimental energies of peak 5 (the Fe $3d$ band).

Figure 5

(Color online) Partial density of states for the valence band of quartz: (a) oxygen partial density of states and (b) silicon partial density of states. Note that the Si intensities have been exaggerated by 5 times relative to the O intensities.

Figure 6

(Color online) Partial density of states for the valence-band region forsterite: (a) oxygen partial density of states, (b) silicon partial density of states, and (c) magnesium partial density of states. Note the exaggeration of the Si (5 times) and Mg (30 times) intensities relative to O.

Figure 7

(Color online) Partial density of states for the valence band of Mg-rich olivine ${\left({\mathrm{Mg}}_{0.87}{\mathrm{Fe}}_{0.13}\right)}_2\mathrm{Si}{\mathrm{O}}_4$ from the $\mathrm{GGA}+U$ calculation: (a) oxygen partial density of states, (b) silicon partial density of states, (c) magnesium partial density of states, and (d) iron partial density of states. Note the exaggeration of the Si and Mg intensities relative to O intensities.

Figure 8

(Color online) Partial density of states for the valence band of fayalite from the $\mathrm{GGA}+U$ calculation: (a) oxygen partial density of states, (b) silicon partial density of states, and (c) iron partial density of states. Note the exaggeration of the Si (5 times) intensity relative to O.

Figure 9

(Color online) A comparison of the partial density of states of the Mg-rich olivine from (a) GGA calculation and (b) $\mathrm{GGA}+U$ calculation.

Figure 10

(Color online) The Fe $3d$ density of states from the GGA and $\mathrm{GGA}+U$ calculations: (a) the Fe $3d$ density of states as a function of spin state for both calculations and (b) the number of summed Fe $3d$ electrons for both calculations.