High temperature study of octahedral cation exchange in olivine by neutron powder diffraction
Science; Washington; Mar 22, 1996; Henderson, C M B; Knight, K S; Redfern, S A T; Wood, B J;

Volume:	271
Issue:	5256
Start Page:	1713
ISSN:	00368075
Subject Terms:	Temperature Neutrons Ions Geology

Abstract:
Time-of-flight, neutron powder diffraction to 1,000 degrees C gives accurate octahedral site occupancies and intersite distribution coefficients for MnMgSiO4 and MnFeSiO4 olivines.

As olivine

(Mg,Fe,Mn) sub 2 SiO sub 4

is the major constituent of the Earth's upper mantle, its physical properties dominate the deep Earth's geophysical and geochemical properties down to the 410-km seismic discontinuity, which is attributed to the transitions of olivine to beta- and gamma-spinel-type polymorphs. A common simplifying assumption is that olivine is near ideal, with Mg and Fe fully disordered over M1 and M2, the two octahedral sites. Intracrystalline M-site partitioning would, however, be expected to modify olivine's thermodynamic stability, the diffusion rates of M-site metals, and (potentially) its elastic parameters. Furthermore, if significant partitioning does occur, M-site occupancies might also provide a means of using olivine as a petrogenetic indicator for thermometry and speedometry in a wide range of rocks.

Crystal chemical studies of olivines, primarily at room temperature (T) and pressure (P), show that divalent cations tend to order preferentially between M2, the larger site, and M1, the more distorted site; for example, Fe, Ni, and Zn into M1, and Mn and Ca into M2 (1). Early attempts to determine the T dependence of partitioning involved study at room T and P of samples quenched from high T, mainly in the Mg-Fe (2) and Mg-Mn (3) olivine systems; contradictory data were obtained, likely because M1-M2 reequilibration occurred during quenching. It has been shown that Mg-Fe intersite equilibration at 1000degC occurs in about 1 s, that reordering upon cooling occurs on a time scale of about 10 ms, and that blocking Ts appear to be between 500deg and 800degC, depending on the cooling rate (4). Rapid exchange has recently been confirmed in a theoretical study of ordering kinetics (5). The typical experimental quenching rates are too slow to freeze-in the equilibrium high-T ordering states. In situ experiments are, therefore, essential for studying ordering in such mineral systems. No significant changes in Fe-Mg ordering were detected up to 800degC by in situ powder x-ray diffraction (XRD) (6). By contrast, high-T XRD, single crystal studies (7) suggest that Fe is slightly ordered into M1

(Equation omitted) = 1.1 to 1.2 (8)

, and that with increasing T this ordering increases

for example, (Equation omitted) 1.37 at 900degC (9)1. However, recent high-T neutron diffraction, single crystal work (10) shows that (Equation omitted) increases from 1.02 at room T to about 1.46 at 800degC, followed by a progressive decrease, falling to 0.19 at 1300degC. is trend indicates that Fe orders into M1 up to about 950degC, but this behavior is reversed at higher Ts with substantial partitioning into M2.

More precise structural information for olivines at high Ts is necessary to obtain clearer K sub D versus T trends. The precision to which Fe-Mg (Equation omitted) may be determined by x-ray methods is limited both by the rapid fall-off of form factors and the inherent limits on obtaining data at short scattering vectors, whereas the precision of neutron time-of-flight diffraction measurements is constrained by the neutron scattering contrast between Fe and Mg nuclei. We have used a different approach, namely, investigating the process of cation exchange in a broader context using Mn analogs. Aspects of the fundamental characterization of cation exchange thus obtained can be used to interpret the expected behavior of Fe-Mg nonideality. To this end, time-of-flight, neutron powder diffraction data for olivine analogs were obtained at the ISIS facility, Rutherford Appleton Laboratory, using the POLARIS diffractometer (11). We have studied synthetic FeMnSiO sub 4 and MgMnSiO sub 4 olivines (12) to take advantage of the large contrast in coherent neutron scattering lengths between the pairs Fe-Mn and Mn-Mg

Fe, 9.54 fm; Mg, 5.375 fm; Mn, -3.73 fm (1 fm = 10 sup -15 m)

. We were also able to obtain data over a wide range of scattering vectors, allowing good discrimination of occupancies and displacement factors so that precise site occupancy factors and K sub D values for Fe-Mn and Mg-Mn olivines were obtained (13) (Table 1 and (Fig. 1). (Table 1 and Fig. 1 omitted)

By combining measured K sub D values for the two samples at the same T, the T dependence for (Equation omitted) was approximated as

(Equation 1 omitted)

(Equation 2 omitted)

(Equation 3 omitted)

where K sub D = 1 for complete disorder.

The measured (Equation omitted) value of 4.21 for the Fe-Mn olivine (Fig. 2A) at room T confirms earlier observations that Fe is ordered onto M1 and Mn onto M2 (1); this K sub D represents the degree of order quenched-in during sample cooling. With increasing T, no significant change occurs until about 400degC, where K sub D begins to show a small increase, reaching 5.03 at 500degC. This increase occurs because at such Ts, the sample approaches the equilibrium order. With further increase to 1000degC, there is a steady, slightly nonlinear decrease in (Equation omitted) to 2.70. Upon heating, the (Equation omitted) value at 600degC is the same (4.21) as that measured at room T at the start of the experiment, suggesting that the blocking T for Fe-Mn exchange during the rapid quench from the sample synthesis T was close to 600degC. The decreasing (Equation omitted) with increasing T corresponds to the sample becoming less ordered.

Upon stepwise cooling (at about 0.4deg per second) down to 500degC, (Equation omitted) values coincide with those measured on heating. This trend defines a blocking T as low as 500degC, consistent with the slower cooling rate of our experiments allowing equilibration at lower T than occurred during quenching after synthesis. Figure 3 shows idealized paths for such ordering patterns and demonstrates the dependence of blocking T and low-T K sub D value on cooling rate. (Figure 3 omitted) Data points between 550deg and 1000degC, when plotted on an Arrhenius-type diagram of In K sub D versus 1/T, define an energy for Fe-Mn intersite exchange of 10.1 +/- 0.3 kJ/mol. Data for the Mn-Mg olivine sample follow similar trends (Fig. 2B), but (Equation omitted) is larger than (Equation omitted) with a value of 7.5 at room T; thus, Mg (in the Mg-Mn sample) is more strongly ordered onto M1 than Fe (in Fe-Mn olivine). (Figure 2B omitted) The Mg-Mn sample becomes less ordered as T increases. The ordering path on heating is similar to that depicted in Fig. 2A, implying similar blocking Ts for Mg-Mn and Fe-Mn exchange. Data between 500deg and 1000degC define an exchange energy for Mn-Mg of 15.7 +/- 0.9 kJ/mol. It therefore seems unlikely that geothermometry based on Fe-Mn and Mg-Mn exchange on M1-M2 would be possible for samples equilibrated above 500degC.

The (Equation omitted) values for the Fe-Mg system derived from the equilibrium partitioning data for the other two samples above 500degC (Fig. 2) are <1, suggesting that Fe is ordered onto M2 and Mg onto M1. Although this deduction contrasts with some high-T results (7, 8), it is in line with the single crystal neutron results above 900degC (10); further high-T work on Fe-Mg olivines is required to cast further light on this phenomenon. The absolute values deduced here should be treated with caution as subtle differences in the crystal chemical relations for the Fe-Mg olivine system could affect the exact order properties, but the T dependence is thought to be reliable. The deduced (Equation omitted) values between 550deg and 1000degC define an energy of Fe-Mg exchange of 4.8 +/- 1.0 kJ/mol.

The low exchange energies reflect the rapid M1-M2 exchange, with the lowest energies expected to correspond to the fastest exchange rates. Thus, exchange rates should decrease in the order Fe-Mg > Fe-Mn > Mg-Mn. This implies that the hopping energies of intersite diffusion decrease in the order Mn > Mg > Fe, and hence, the relative values of the calculated volume diffusion coefficients in olivine decrease as D sub Fe > D sub Mg > D sub Mn (14). For the Fe-Mn sample at low T, refinement of isothermal diffraction data collected in successive 30-min periods shows no significant structural differences, suggesting that intersite exchange equilibration occurs on a time scale of minutes (perhaps much less). Thus, K sub D values determined at room T for Fe-Mn olivines (15) quenched from high T, and presumably for quenched Mg-Mn olivines (16), must be treated with caution.

The results of our in situ neutron diffraction studies on Mn analogs have wider implications for Fe-Mg natural olivines. The increasing degree of order up to about 900degC observed in high-T single crystal measurements on natural Fe-Mg olivines (7, 9, 10) might originate from the same processes as seen in our Fe-Mn data below 450degC, that is, the samples move toward equilibrium degrees of order at Ts lower than the blocking T experienced during natural cooling. In addition, the fact that K sub D values fall off the equilibrium K sub D versus T trend at low Ts (Fig. 3) offers the possibility of obtaining information on cooling rates of natural samples.

Recent structural studies of Fe-Mg order-disorder arrangements over octahedral sites in amphiboles are based on structural analyses carried out at room T on quenched samples (17), but no T dependence was detected between 600deg and 750degC. Such data have been used to develop sophisticated thermodynamic mixing models (18), but before such models can be assessed, it is necessary to establish whether Fe-Mg intersite exchange is very rapid, as in olivine, or significantly slower, as in orthopyroxene (19). The contrast in Fe-Mg intersite exchange kinetics is evident when the exchange energy estimated here for olivine (4.8 kJ/mol) is compared to the value of 25 kJ/mol for orthopyroxene (20). Further high-T in situ structural work is necessary to elucidate the effects of octahedral site order-disorder on structural and thermodynamic properties of mafic minerals in general.

REFERENCES AND NOTES

1. G. E. Brown Jr.. in Orthosilicates, P. H. Ribbe, Ed. (Reviews in Mineralogy, vol. 5, Mineralogical Society of America, Washington, DC, 1980), pp. 275-381.

2. F. Princivalle, Mineral. Petrol. 43, 121 (1990); G. Ottonello, F. Princivalle, A. Della Giusta, Phys. Chem. Miner. 17. 301 (1990).

3. T. Akamatsu et al., Phys. Chem. Miner. 16, 105 (1988).

4. N. Aikawa, M. Kumazawa, M. Tokonami, ibid. 12, 1 (1985).

5. T. Akamatsu and M. Kumazawa. ibid. 19, 423 (1993).

6. G. Artioli, M. Bellotto; B. Palosz, Powder Diffr. 9, 63 (1994).

7. J. R. Smyth and R. M. Hazen, Am. Mineral. 58, 588 (1973); G. E. Brown and C. T. Prewitt. ibid., p. 577.

8. Distribution coefficients for intersite order in Fe-Mg olivines are formulated as (Equation omitted) = (Fe/Mg) sub M1 /(Fe/Mg) sub M2 , (molar ratios).

9. T. Motoyama and T. Matsumoto, Mineral. J. 14, 338 (1989).

10. G. Artioli, R. Rinaldi, C. C. Wilson, P. F. Zanaui, Am. Mineral. 80. 197 (1995); R. Rinaldi and C. C. Wilson, Solid State Commun. 97, 395 (1996).

11. R. I. Smith, S. Hull, A. R. Armstrong, Mater. Sci. Forum 166-169, 251 (1994).

12. We prepared the Fe-Mn sample by mixing Fe sub 2 O sub 3 , MnO, and SiO sub 2 and crystallizing the mixture at 1150degC for three periods of 12 hours each, and the Mn-Mg olivine was prepared from MgO, MnO, and SiO sub 2 crystallized at 1400degC for three periods of 4 hours each; both were ground between heat treatments. Both samples were buffered in CO sub 2 -CO gas measures about 1 log unit more oxidizing than the Fe-FeO buffer and quenched within the reducing gas atmosphere to below 500degC in 1 min. About 3.5 g of each was prepared. Both samples were studied by x-ray powder diffraction and were found to be phase pure, with narrow peaks showing that they were well crystallized and homogeneous. Each sample was loaded into a thin-walled quartz glass tube, with silica wool separating the sample from either Fe-FeO buffer (Fe-Mn sample) or Ni-NiO buffer (Mn-Mg). The tubes were then evacuated and sealed, placed in vanadium cans, and mounted inside a vanadium furnace assembly, the whole of which was evacuated to a residual P of about 0.05 Pa. For the Fe-Mn sample, isothermal diffraction patterns were collected for 2 hours each upon heating and for 1/2 hour each upon cooling; Mg-Mn data were collected for 1 hour upon heating. Structural data were obtained by Rietveld refinement. For the Fe-Mn sample, at each T between 400deg and 750degC, diffraction patterns were collected in four 1/2-hour periods. Some of these isothermal data sets were refined separately, but no differences in site populations were observed.

13. Relative errors on occupancy factors for the Fe-Mn olivine are about 0.5%, equivalent to errors in K sub D values of only 1 to 1.5%. Data for the Mg-Mn sample are less precise. with errors in K sub D values of about 3 to 4.5%.

14. M. Miyamoto and H. Takeda, Nature 303, 602 (1983).

15. H. Annersten, J. Adetunji, A. Filippidis, Am. Mineral. 69, 1110 (1984).

16. C. A. Francis and P. H. Ribbe, ibid. 65, 1263 (1980).

17. M. Hirschmann, B. W. Evans. H. Yang, ibid. 79, 862 (1994).

18. M. S. Ghiorso, B. W. Evans, M. M. Hirschmann, H. Yang. ibid. 80, 502 (1995).

19. J. A. Sykes-Nord and G. M. Molin. ibid. 78, 921 (1993).

20. G. M. Molin. S. K. Sawena, E. Brizi, Earth Planet. Sci. Lett. 105, 260 (1991).

21. We thank Rutherford Appleton Laboratory for research grant RB6035 and R. Rinaldi and an anonymous reviewer for constructive critical comments.

13 October 1995; accepted 1 February 1996

C.M.B. Henderson, Department of Earth Sciences, University of Manchester, Manchester M13 9PL, UK.

K.S. Knight, ISIS, Rutherford Appleton Laboratory, Oxon OX11 OQX, UK.

S.A.T. Redfern, Department of Earth Sciences, University of Cambridge, Cambridge CB2 3EQ, UK.

B.J. Wood, Department of Geology, University of Bristol, Bristol BS8 1RJ, UK.

* C.M.B. Henderson, To whom correspondence should be addressed.