Synchrotron FTIR imaging of OH in quartz mylonites

Previous measurements of water in deformed quartzites using conventional Fourier transform infrared spectroscopy (FTIR) instruments have shown that water contents of larger grains vary from one grain to another. However, the non-equilibrium variations in water content between neighboring grains and within quartz grains cannot be interrogated further without greater measurement resolution, nor can water contents be measured in finely recrystallized grains without including absorption bands due to fluid inclusions, films, and secondary minerals at grain boundaries. Synchrotron infrared (IR) radiation coupled to a FTIR spectrometer has allowed us to distinguish and measure OH bands due to fluid inclusions, hydrogen point defects, and secondary hydrous mineral inclusions through an aperture of 10 μm for specimens > 40 μm thick. Doubly polished infrared (IR) plates can be prepared with thicknesses down to 4–8 μm, but measurement of small OH bands is currently limited by strong interference fringes for samples < 25 μm thick, precluding measurements of water within individual, finely recrystallized grains. By translating specimens under the 10 μm IR beam by steps of 10 to 50 μm, using a software-controlled x−y stage, spectra have been collected over specimen areas of nearly 4.5 mm2. This technique allowed us to separate and quantify broad OH bands due to fluid inclusions in quartz and OH bands due to micas and map their distributions in quartzites from the Moine Thrust (Scotland) and Main Central Thrust (Himalayas). Mylonitic quartzites deformed under greenschist facies conditions in the footwall to the Moine Thrust (MT) exhibit a large and variable 3400 cm−1 OH absorption band due to molecular water, and maps of water content corresponding to fluid inclusions show that inclusion densities correlate with deformation and recrystallization microstructures. Quartz grains of mylonitic orthogneisses and paragneisses deformed under amphibolite conditions in the hanging wall to the Main Central Thrust (MCT) exhibit smaller broad OH bands, and spectra are dominated by sharp bands at 3595 to 3379 cm−1 due to hydrogen point defects that appear to have uniform, equilibrium concentrations in the driest samples. The broad OH band at 3400 cm−1 in these rocks is much less common. The variable water concentrations of MT quartzites and lack of detectable water in highly sheared MCT mylonites challenge our understanding of quartz rheology. However, where water absorption bands can be detected and compared with deformation microstructures, OH concentration maps provide information on the histories of deformation and recovery, evidence for the introduction and loss of fluid inclusions, and water weakening processes. Published by Copernicus Publications on behalf of the European Geosciences Union. 1026 A. K. Kronenberg et al.: Synchrotron FTIR imaging


1
Introduction 116 117 Quartz mylonites sheared at middle to lower levels of the continental crust exhibit 118 microstructural and textural evidence of dislocation creep, a process that is widely 119 believed to require water weakening in framework silicates. The effects of water 120 on dislocation creep of quartz, including the nucleation, glide, climb and recovery 121 of dislocations, and recrystallization are well known from: 1) experimental studies 122 of natural crystals, in which water was introduced into grain interiors (e.g., Griggs,123 1967; Blacic, 1975Blacic, , 1981FitzGerald et al., 1991), 2) studies of synthetic and 124 natural quartz varieties with large initial water contents (e.g., Griggs and Blacic, 125 1965;Hobbs, 1968 quartz water contents have been reported that show a trend of decreasing OH 160 content towards the center of a high grade shear zone (Finch et al., 2016). Maps of 161 OH content, constructed from FTIR spectra of deformed granitic rocks (Seaman et 162 al., 2013) show compelling relationships between water content and 163 microstructures generated during deformation, suggesting that water contents are 164 reduced during recrystallization and partial melting. 165 Much as deformation microstructures and textures provide a link between 166 our understanding of deformation mechanisms activated in deformation 167 experiments at high laboratory strain rates and the mechanisms governing 168 plasticity and creep of shear zones at low natural strain rates (Snoke et al., 1998;169 Heilbronner and Barrett, 2014), IR spectroscopy can provide a link between our 170 understanding of water weakening in the lab and in nature. In this paper, we 171 report on methods of FTIR spectroscopy to characterize OH absorption bands and 172 image OH contents in quartz (and other nominally anhydrous minerals) at higher 173 resolution than is possible using conventional instruments, coupling synchrotron 174 IR radiation with FTIR spectroscopy. We then apply these methods to mylonitic Scotland (Christie, 1963;Law et al., 1986Law et al., , 2010  parallel to foliation, with grain shape aspect ratios up to 50:1 to 100:1 and smooth 233 undulatory extinction between crossed polarizers that have been described as 234 quartz ribbons (Bonney, 1883;Christie, 1960Christie, , 1963Weathers et al., 1979;Law et 235 al., 1986). At the margins of the larger quartz grains, more equant, finely 236 recrystallized grains overprint these elongate high-strain grains, with the 237 proportion of new to old grains increasing structurally upwards towards the Moine 238 Thrust plane (Christie, 1960;Weathers et al., 1979;Law et al., 1986). Rare grains 239 of feldspar and quartz aligned in "mechanically strong" orientations are relatively 240 equant, and appear as augen or globular quartz grains. Quartz c-axes exhibit 241 strong lattice preferred orientations in both deformed old grains and recrystallized 242 grains, with symmetrical Type 1 (Lister, 1977) cross-girdle fabrics at distances > 243 150 mm beneath the thrust plane and increasingly asymmetric cross-girdle to 244 single girdle fabrics closer to the thrust plane (Law et al., 1986(Law et al., , 2010. These 245 fabrics reflect general flattening strains accommodated by quartz basal and prism 246 slip, with variations in estimated flow vorticities and partitioning of strain between 247 original and recrystallized grains (Law et al., 2010;Law 2010 Law, 2014). 312 The small recrystallized quartz grain sizes (16-9 µm), subgrain sizes, and  Oriented rock chips were first mounted on a glass thin section plate using 383 CrystalBond 509 (supplied by Aremco, NY) and a low-temperature hot plate. The 384 top surfaces of samples were ground by hand against large glass plates using a 385 sequence of silicon carbide (400 mesh size) and alumina powder slurries of 386 decreasing particle size (600 mesh and 9.5 µm). Grinding with each grit size was 387 carried out until the thickness of the sample, measured by a micrometer, was 388 reduced by three times the grit dimension of the previous step (e.g., 3 µm grit was 389 used to remove 3 x 9.5 µm = 29 µm). Samples were cleaned ultrasonically before 390 advancing to the next grit size. Ground surfaces were then polished using 391 polishing cloths (Buehler Texmet and Microcloth) mounted on high-speed laps 392 and a sequence of alumina powder slurries with particle sizes of 3 and 0.3 µm, 393 using the same measurement technique and testing the finish at each step by visual 394

inspection. 395
Polished specimens were removed from the glass plate and remounted, this 396 time on their polished surfaces, applying gentle pressure to sample centers to 397 promote uniform thickness of the CrystalBond mounting medium (~20 µm). 398 Sample plates were cut parallel to the first surface, and the second (cut) surface 399 was ground and polished by the same methods outlined above. For the second 400 surface, a micrometer was used to measure the compound thickness of sample, 401 glass plate, and mounting medium, checking thickness at sample extremities, to 402 adjust the grinding procedure for thickness uniformity. Sample thicknesses were 403 determined from the compound sample-glass plate thickness, the glass plate 404 were removed from the glass thin section over a low-temperature hot plate, and the 418 CrystalBond resin removed using acetone. 419 Once quartz IR plates, 4 to 150 µm thick, are removed from the glass thin 420 section plate, they are very fragile, and we found that they did not tolerate 421 ultrasonic agitation. As a result, we soaked samples in acetone baths several 422 times, dissolving CrystalBond resin for 30 minutes for each bath, exchanging the 423 acetone in the beaker and repeating this procedure three times. Even without 424 agitation, some samples did not survive when air bubbles were caught beneath 425 them, leading to specimen warping and disintegration. However, with sufficient 426 care, loss rates were low, and doubly polished IR plates were prepared without 427 impregnating resins with thicknesses of 150 µm to 4 -8 µm. Central Thrust quartz spectra tend to have smaller OH absorption bands (Fig. 2c), 456 with some grains showing small sharp bands at 3595, 3482, 3431, 3408, and 3379 457 cm -1 due to hydrogen interstitial defects (Kats, 1962;Aines and Rossman, 1984) 458 and less common grains with a larger broad band at 3400 cm -1 . Finely dispersed 459 micas are less common in these coarser-grained mylonites, but some quartz grains 460 also exhibit a 3600 cm -1 OH band due to micas (Fig. 2d). Aside from differences in the Nicolet and Bruker FTIR instruments and 462 software, the most significant difference between these facilities is the IR source, 463 so that OH absorption measurements with the Nicolet FTIR and its conventional 464 IR source could not practically be made with apertures < 50-100 µm, while OH 465 absorption bands could routinely be made with the Bruker FTIR and synchrotron 466 IR source through a 10 µm aperture. In both cases, the IR sources are nearly 467 isotropic, leading to differences in those OH absorption bands that are dichroic in 468 quartz and mica grains of varying orientation. Small sharp OH bands due to 469 hydrogen interstitials of quartz between 3595 cm -1 and 3379 cm -1 absorb most 470 strongly for vibrational directions perpendicular to the quartz c axis (Kats, 1962), 471 and these bands vary according to the c-axis orientation of each quartz grain 472 relative to the unpolarized IR beam. OH absorption bands at ~3600 cm -1 of micas 473 are strongly dichroic (Beran, 2002), and they vary according to mica grain 474 orientations relative to the unpolarized IR beam. However, the primary OH band 475 of interest for water-weakening of milky quartz, at 3400 cm -1 , is broad and Direct comparisons of intragranular water of a given quartz grain using the 495 two FTIR facilities and a common aperture size could not be made because of 1) 496 the poor signal-to-noise ratio of spectral measurements through a 10 µm aperture 497 with the broad globar IR radiation, and 2) the inability to make spectral 498 measurements for a 100 µm-apertured area with the narrow synchrotron IR beam 499 (which is not much broader at the microscope stage than the 10 µm aperture). 500 Samples with uniform OH contents might serve as standards to compare OH 501 absorption bands for different spectrometers, irrespective of aperture size, but our 502 observations indicate that OH contents of the quartz grains in the mylonites we 503 measured are highly variable. Measurements through finer apertures led to larger 504 variances in OH absorption bands, within grains as well as between neighboring 505

grains. 506
We also observed larger amplitude interference fringes in spectra measured 507 with the synchrotron IR source than those observed in spectra measured with the 508 globar IR source for a given sample. These fringes are caused by internal 509 reflection within doubly polished, parallel-sided IR plates, and they have larger 510 magnitudes for the highly collimated IR synchrotron beam than for the broad, 511 confocal globar IR radiation. many spectra we collected for quartz mylonite plates (Fig. 2c, Grains 1 and 8 where n is the mean refractive index of quartz (n = 1.55) and δυ is the measured 552 peak-to-peak fringe spacing (Stuart et al., 1996;Griffiths and de Haseth, 2007). 553 The two measures of thickness were in agreement within resolution (~ 5%) for a 554 given IR plate and location within the specimen, with thick IR specimens (~100 555 µm ) showing real variations in local thickness of + 10 µm and thin samples 556 varying in local thickness from 4 to 8 µm. 557 IR spectra were collected and integrated absorbances of OH bands were 558 measured above an assumed straight-line background, where backgrounds were fit 559 and integration limits were chosen at the same wavenumber values, from ~3705 560 cm -1 to 2880 cm -1 for OH bands of quartz grains and ~3702 cm -1 to 3544 cm -1 for 561 OH bands of micas (Fig. 3a). We have confidence in our ability to separate OH 562 absorption bands due to fluid inclusions in quartz and due to micas in Moine 563 Thrust samples, because IR spectra of coarse muscovite grains in Moine Thrust 564 mylonites consist of a single OH absorption band at 3620 cm -1 (Fig. 3b). This 565 strong OH stretching band is well known from multiple spectroscopic studies of 566 muscovite (Beran, 2002; Tokiwai and Nakashima, 2010a,b; Kallai and Lapides, 567

2015). 568
Our ability to distinguish OH absorbance due to fluid inclusions and micas 569 in Main Central Thrust samples, by contrast, is poor. In addition to the primary 570 OH absorption bands of muscovite (Fig. 3c) and biotite (Fig. 3d) grains at 3638 571 cm -1 and 3614 cm -1 , respectively, Main Central Thrust muscovite spectra show 572 smaller OH bands at 3815, 3311, 3146, and 3035 cm -1 and biotite spectra show 573 shoulders at both sides of the primary OH absorption band (at ~3679 cm -1 and 574 ~3561 cm -1 ) and significant OH bands at 3258, 3043, and 2829 cm -1 . All of these 575 OH bands are anisotropic, and complexly so. OH band absorbances of micas will 576 therefore vary with crystallographic orientation within an unpolarized IR beam. 577 The primary OH band of Main Central Thrust muscovite at 3638 cm -1 , measured 578 in polarized IR radiation, is strongest when the vibration direction E is parallel to 579 the basal plane (001), consistent with idealized hydrogen positions in dioctahedral 580 micas and the polarization of OH bands normally reported for muscovite (Beran, 581 2002). However, our polarized IR measurements of muscovite OH bands at 3311, 582 3146, and 3035 cm -1 indicate that they are more nearly isotropic. The primary OH 583 band of Main Central Thrust biotite grains at 3614 cm -1 is strongest when E is 584 perpendicular to (001), consistent with hydrogen positions of trioctahedral micas 585 and OH band polarizations observed for phlogopite (Beran, 2002). However, the 586 OH bands at 3258, 3043, and 2829 cm -1 are only weakly polarized and in the 587 opposite sense of the primary 3614 cm -1 band. 588 As a result, we only feel confident in our determinations of OH 589 absorbances of quartz grains in Main Central Thrust mylonites (Fig. 2c), when 590 mica inclusions are absent (3600 cm -1 band is undetectable). When the 3600 cm -1 591 band is present (Fig. 2d), we cannot readily interpret absorption bands of 592 molecular water or hydrogen defects in quartz; at best, quartz OH contents are 593 c (in molar ppm, OH/10 6 Si) = 1.05 Δ (in cm -2 ).
(3) 607 608 To the extent that spectra include sharp OH absorption bands due to hydrogen 609 point defects, this calibration will overestimate OH concentrations due to H 610 interstitials, given that the value of k for these bands is larger than for molecular 611 water (Kats, 1962   Integrated absorbances Δ (A/t in cm -2 ) were also measured for the 3600 cm -629 1 OH band of micas but no attempt was made to convert these to OH 630 concentrations. Muscovite grains of Moine Thrust samples show simple OH 631 absorption spectra with one prominent band at ~3600 cm -1 (Fig. 3b), but integrated 632 absorbances for this band cannot be converted to muscovite (or OH) content were also chosen to image distributions of micas, plotting log 10 (Δ in cm -2 ) for the 698 integrated absorbance of the ~3600 cm -1 OH band, using a similar key for the 699 contour interval, where cool colors correspond qualitatively to high mica contents 700 and warm colors correspond to low mica contents. 701 702 4 Results 703 704 Synchrotron radiation coupled with the FTIR spectrometer and microscope at the 705 NSLS I has enabled spectral measurements for sampling volumes that are smaller 706 by a factor of 100 than is possible with a conventional FTIR microscope system. 707 OH band measurements for quartz through a 10 µm aperture are comparable to 708 OH band measurements made using a conventional FTIR spectrometer-  absorption band at 3600 cm -1 due to mica inclusions (Fig. 4a), just as those 730 measured for larger sampling volumes (Fig. 2a, b). Given the more intense IR 731 radiation of the synchrotron source, we did not encounter any losses in spectral 732 quality due to signal-to-noise ratios for the smaller sample volumes measured. 733 However, interference fringes were more apparent using the synchrotron-FTIR noise ratio of our measurements continue to be acceptable to measure the small 768 OH absorption bands in very thin IR plates, interference fringes increase in size as 769 the IR plate thickness t is decreased. Interference fringes become very large at t < 770 25 µm and OH bands cannot be detected at t = 13 µm and 6.5 µm (Fig. 4b, c). Optical microstructures of 100 µm-thick IR plates are poorly resolved by 792 comparison with those imaged in 30 µm-thick sections (Fig. 1c, d), with 793 interference colors that reflect greater optical retardation, grain boundaries that are 794 not as clearly defined, and greater numbers of overlapping grains. Remarkably, 795 ultrathin IR plates of quartz mylonites, just 4-8 µm thick (Fig. 1f), continue to 796 exhibit contrast between grains and within grains, with first-order black to grey 797 birefringence that can be enhanced by increasing image contrast. first integrated absorbance (as illustrated in Fig. 3a) to represent the best measure 838 of absorbance Δ OH qtz of water and hydrogen defects of quartz grains. 839 These values were plotted spatially for Moine Thrust and Main Central 840 Thrust samples, and contoured on common logarithm scales, to form high-841 resolution images (Figs. 6 -9) of OH absorbance of quartz (converting to molar 842 ppm, OH/10 6 Si) and OH absorbance of micas (Δ OH musc in cm -2 ). In all cases, we 843  (Fig. 6). The lack of 3600 cm -871 1 OH absorbances within globular quartz grains indicate that finely dispersed 872 micas in original relict quartz grains are absent, while large 3600 cm -1 OH 873 absorbances at quartz grain extremities indicate that micas are localized at quartz 874 grain boundaries (Fig. 6a, b). Broad band OH contents of quartz of >1000 ppm, 875 thought to be sufficient for water weakening, are present in undeformed and 876 deformed ribbon quartz grains (Fig. 6 c, d) with very large water contents 877 (>10,000 ppm) marking a healed crack in the quartz augen, made up of a planar 878 array of fluid inclusions. OH contents due to fluid inclusions are also very large at 879 the globular augen quartz boundaries, coincident with high mica concentrations, 880 and in some ribbon and recrystallized quartz grains. With a sample plate thickness 881 of 56 µm, these maps represent OH absorbances within the interiors of larger 882 quartz grains, while OH absorbances of fine micas and recrystallized quartz 883 represent composite spectra of polycrystalline fault rock. However, we are 884 confident that the quartz OH contents reflect fluid inclusions, even in these fine-885 grained regions, because of the simple spectral quality of muscovite (Fig. 3)

in the 886
Moine Thrust samples and our ability to distinguish between the 3400 cm -1 and 887 3600 cm -1 OH absorbances. 888 High-strain ribbon quartz grains in the Moine Thrust samples have large 889 OH contents (>1000 ppm) comparable to those of water-weakened synthetic and 890 milky quartz (samples SG-10 and SG-8, Figs. 7 and 8, respectively), with some 891 reductions in OH at recrystallized margins of original grains. Micas, as imaged by 892 the 3600 cm -1 OH absorbance (Fig. 7a, b), continue to be highly localized at the 893 grain boundaries of deformed quartz grains, with a mixture of fine-grained mica 894 and quartz grains providing evidence for some redistribution during 895 recrystallization (Fig. 8a, b). The broad 3400 cm -1 OH absorbance in marginal 896 recrystallized regions surrounding ribbon quartz grains are locally smaller than 897 those of the original deformed quartz grains in some regions (Fig. 7 c, d) while 898 broad OH absorbances of recrystallized quartz continue to be large where mica 899 contents (as evidenced by the 3600 cm -1 band absorbance) are large (Fig. 8 c, d). 900 FTIR maps of coarse-grained Main Central Thrust mylonites yield spectral 901 measurements of individual grains of quartz, muscovite and biotite, even for 902 relatively thick IR plates and larger step sizes (Fig. 9). Small, sharp absorption 903 bands of quartz grains yield OH contents of ~100 ppm, with only local regions of 904 quartz with larger OH contents near contacts with coarse-grained muscovite and 905 biotite grains. Quartz grain interiors generally lack the absorbance band at 3600 906 cm -1 ; thus, there is no evidence for finely dispersed micas within these coarse-907 grained deformed (and recrystallized) quartz grains. Contours of integrated OH 908 absorptions at 3400 cm -1 are considerably larger for coarse-grained micas and near 909 their contacts with quartz grains. However, these bands cannot be attributed to 910 fluid inclusions where they coincide with large 3600 cm -1 mica bands, given that water contents reduced during high temperature deformation by a variety of 951 processes, from pipe diffusion Jansen, 1990, 1994;Hollister, 1990;  Central Thrust samples have little or no detectable molecular water (Fig. 9).  The physical means by which we attempted to reduce interference fringes 1031 (rotating the sample within the IR beam, and changing numerical aperture of the 1032 IR objective) were also unsuccessful. However, interference fringes might be 1033 laboratory experiments (e.g., Heard and Carter, 1968;Blacic, 1975;1077 Christie, 1984). Given sufficiently high temperatures, it is possible that quartz 1078 may deform at tectonic strain rates without critical hydrogen defects at 1079 dislocations and water weakening (Kilian et al., 2016). However, this implies that 1080 we have not measured flow laws for appropriately dry quartzites that we can apply 1081 to amphibolite conditions and natural strain rates. Alternatively, water may have 1082 been lost from quartz interiors following deformation. It is also possible that 1083 hydrogen defects that enhance dislocation motion at high temperatures and natural 1084 strain rates may be sourced from grain boundaries or micas, diffusing over longer 1085 distances than are possible at greenschist conditions or laboratory strain rates.