Deformation‐induced graphitization and muscovite recrystallization in a ductile fault zone

A suite of slate samples collected along a 2 km transect crossing the Lishan fault in central Taiwan were evaluated to assess the role of ductile deformation in natural graphitization at lower greenschist facies metamorphic conditions. The process of natural aromatization, or graphitization, of an organic precursor is well established as a thermally driven process; however, experimental studies have shown that the energy provided by deformation can substantially reduce the activation energy required for graphitization. This study provides a natural example of deformation‐induced graphitization. A strain gradient approaching the Lishan fault was established by scanning electron microscope imaging and X‐ray diffraction analysis of phyllosilicates and quartz that showed an increase in the strength of slaty cleavage development via dissolution‐precipitation processes. Thermal conditions were constrained to be near isothermal using calcite‐dolomite geothermometry. Raman spectroscopic results from carbonaceous material, including D1‐full width‐at‐half‐maximum (FWHM), G‐FWHM, Raman band separation (RBS), and a lesser‐known vibrational mode B2g‐FWHM, showed robust linear trends across the same sampling transect. However, the G‐FWHM parameter showed a trend opposite of that expected from thermally driven graphitization. The Raman results are interpreted to reflect a strain‐driven reduction in graphite crystallite size (decrease in G‐FWHM) but improvement in structural ordering in individual coherent domains. A multiple linear regression with an R2 value of 0.92 predicts the graphite D1‐FWHM values from the XRD‐derived ratio of muscovite populations and muscovite microstrain, demonstrating the concomitant recrystallization of silicates and carbonaceous material across the strain gradient, despite the disparate processes accommodating the deformation. This study demonstrates the role of deformation in natural graphitization and provides a new perspective on the use of graphite as a geothermometer in strongly deformed greenschist facies rocks.


| INTRODUCTION
The general increase in the structural ordering of carbonaceous material (CM) with diagenetic and metamorphic temperature have led to the widespread application of the crystallinity of CM as a geothermometer (Beyssac et al., 2002).The general irreversibility of the graphitization, or aromatization process (Buseck & Beyssac, 2014;Luque et al., 1993), make this an attractive method for monitoring peak temperatures in metamorphic terranes.The molecular structure of CM is also used in assessing source rock maturity in oil and gas exploration (e.g., Dow, 1977;Liu et al., 2013;Sauerer et al., 2017) and in defining thermal histories for tectonic reconstructions (e.g., Beyssac et al., 2002Beyssac et al., , 2007;;Chen et al., 2019;Rahl et al., 2005).
The process of graphitization is understood to proceed through the expulsion of heteroatoms from organic precursors and progressive formation of hexagonal aromatic carbon rings, eventually forming three-dimensionally stacked layers of graphene sheets (Grew, 1974;Landis, 1971;Pasteris & Wopenka, 1991).Experimental work by Bustin et al. (1995) showed that temperatures far exceeding crustal conditions (2800 C) are required for the initiation of anthracite graphitization under ambient pressure conditions.In the same study, experimental results from anthracite deformed under coaxial stress conditions at 6 kbar and 900 C still did not yield significant aromatization, whereas experiments performed at lower temperatures (600 C) but with high shear stress resulted in significant structural ordering.These findings, and those of others (Bustin et al., 1995;Mastalerz et al., 1993;Moris-Muttoni et al., 2023;Nover et al., 2005), as well as thermodynamic extrapolations (Bonijoly et al., 1982;Ross & Bustin, 1990), have demonstrated that strain energy, that is, the energy represented by stretched molecular bonds and localized radicals formed from bond breakage, lowers the activation energy required for graphitization in laboratory experiments, and by proxy, in natural systems.
Despite the clear experimental data demonstrating that strain associated with non-hydrostatic stress reduces the activation energy required for graphitization by an order of magnitude relative to thermal energy (Bonijoly et al., 1982;Nakamura et al., 2017;Ross & Bustin, 1990), numerous graphite geothermometers that use X-ray diffraction (XRD) data (e.g., Luque et al., 1993), and more recently Raman spectroscopic data (e.g., Henry et al., 2019, and references therein), have been proposed and applied across a wide range of geologic studies of metamorphic terranes (e.g., Beyssac et al., 2007;Kouketsu et al., 2019).Generally, temperatures calculated from these results agree with other published work constraining peak metamorphic conditions, raising the question of whether strain derived from ductile deformation processes plays a significant role in natural graphitization, and the relevance of laboratory results to understanding natural systems in this context.
Several studies have investigated the impact of ductile deformation on graphitization in naturally deformed rocks through various approaches.These studies used structural modelling (Diessel et al., 1978), detailed petrographic characterization combined with known geologic conditions (Barzoi, 2015;Barzoi & Guy, 2002;Muirhead et al., 2021), quantitative strain analysis (Kedar et al., 2020), and mineral equilibrium-based geothermometers (Barzoi, 2015;Diessel et al., 1978) to help constrain the other geologic parameters that can influence graphitization.The results from these studies together provide evidence that the strain associated with deformation-induced cleavage development, characteristic of regionally deformed metamorphic terranes, promotes carbon atomic ordering and could lead to anomalous graphite geothermometry temperatures.A recent publication by Schito et al. (2023) reviewed Raman spectroscopy-based studies of graphitization from a variety of temperature-strain-temporal conditions and provided a research pathway to a kerogen-to-graphite kinetic model.Schito et al. (2023) noted the persisting knowledge gaps on the pressure-temperature conditions for regional metamorphic settings where pressure and deformation become an important factor in the graphitization processes.The Schito et al. (2023) study and those discussed above demonstrate the need for more work in this area of study, specifically on rocks metamorphosed and deformed in lower-greenschist facies conditions where the role of deformation in graphitization is likely to be of importance but has gone largely unstudied.
Here we seek to evaluate the role that ductile deformation plays in the process of graphitization within a ductile fault zone.We present results from slates collected across the Lishan fault zone, a lower greenschist facies ductile shear zone in central Taiwan.The similarity of slate lithologies on either side of the Lishan fault provides a natural laboratory to evaluate the role of strain in similar rock types, nearly independent of metamorphic grade.By integrating results from scanning electron microscopy (SEM), whole-rock XRD data, calcitedolomite geothermometry, and Raman spectroscopy, we establish trends of phyllosilicate recrystallization and increasing graphitization across a shear zone, at near-constant thermal conditions.Through statistical evaluation of these data, it is shown that the progress of muscovite recrystallization through dissolutionprecipitation during slaty cleavage development predicts CM structural ordering, indicating that the different deformation-induced processes involved are concomitant.These data offer insight into how strain resulting from ductile deformation impacts the graphitization process in naturally deformed rocks and further inform our understanding of the use of graphite crystallinity as a geothermometry tool.

| GEOLOGIC SETTING
The island of Taiwan is the product of the late Cenozoic and ongoing arc-continent collision of the Luzon Arc of the Philippine Sea plate with the Eurasian plate.The suture between these two plates, the Longitudinal Valley fault, marks the boundary where oblique convergence has resulted in the southward migration of accretion and deformation (Figure 1a; Ho, 1986;Suppe, 1984).West of the Longitudinal Valley fault, Taiwan is divided into several N-S striking tectono-stratigraphic units separated by thrust faults and differentiated by their ages and metamorphic grade (Figure 1a).From west to east, the Eurasian continental margin is divided into several distinct belts.The Western Foothills contain folded but unmetamorphosed continental sediments.The Hsüehshan Range is comprised of Eocene and Oligocene age rocks and on the eastern side, the lower part of the Paileng Formation includes the Tachien member, a thick bedded basal sandstone and quartzite, which is overlain by interbedded slates and sandstones (Brown et al., 2012).The Central Range is divided into the Tertiary cover sequence on the western side, the Backbone Slate Range, and the higher-grade basement rocks of the Eastern Central Range (Figure 1a).The metamorphic history occurred in two stages: a Late Cretaceous contact metamorphic event, the 'Nanao orogeny' affecting the Central Range, and the regional Pliocene-Pleistocene 'Penglai' or Taiwan orogeny (e.g., Wintsch et al., 2011) that caused the shortening of all of the above tectonostratigraphic units.
The area examined here is in the central region of the island along the steeply westward dipping Lishan thrust fault, which separates slates of the Hsüehshan Range from the slates of the Backbone Slate Range (Figure 1b).The trace of the Lishan fault is noted by a N-S striking valley in central Taiwan, but poor surface exposure and the juxtaposition of notably similarly slaty rocks on either side of the fault zone have led to varying interpretations of its tectonic significance (Chen, 2000;Yue et al., 2005).However, mapping (Brown et al., 2012) and geophysical studies (Bertrand et al., 2009;Brown et al., 2022 (2012).Sampling transect from this study is marked with a red box and white circles.Chen et al., 2015;Wu, 1978;Wu et al., 2004) have led to a general understanding that the Lishan fault cuts out at least $25 million years of stratigraphy and in the location of this study, marks the boundary between the middle Miocene Lushan Formation of the Backbone Slate Range and the middle Eocene Paileng Formation of the Hsüehshan Range.The same boundary is characterized by a several hundred-meters wide, steeply dipping ductile shear zone (Brown et al., 2012(Brown et al., , 2022) ) and is a potentially seismically active boundary in Taiwan (Kuo-Chen et al., 2015).
Previous work put loose constraints on the peak metamorphic conditions on either side of the Lishan fault.Graphite geothermometry and partially reset detrital zircon (U-Th)/He ages from the Northern (Beyssac et al., 2007;Chen et al., 2011Chen et al., , 2019;;Simoes et al., 2012) and Central cross-island highways (Beyssac et al., 2007;Simoes et al., 2012), and the sparse points in between (Beyssac et al., 2007), consistently show that thermal conditions in the Lushan Formation on the eastern side of the Lishan fault did not exceed 330 ± 50 C. Thermal constraints to the west of the fault in the Paileng Formation are limited by lack of data.However, two graphite geothermometry datapoints close to this study's field area suggest that within 1-2 km due west of the Lishan fault, thermal conditions reached $370 ± 50 C (Beyssac et al., 2007).Brown et al. (2012) was done along an $2 km transect across the Lishan fault following outcrops exposed by the Zhuoshui River, south of Wujie in Ren'ai Township (Figure 1b).Orientations of penetrative bedding-parallel cleavage were measured at all 10 sample locations.To avoid the possibility of overlooking early weak cleavages in these rocks, we refer below to the dominant cleavage as 'S n '.

| Powder X-ray diffraction
Aliquots from each sample were hand ground to <40 mesh and micronized in a McCrone mill in order to achieve a homogenous particle size of $10 μm.Powders were top loaded in a Ti-cavity specimen mount and analysed on a Bruker D8 diffractometer equipped with a Sol-X point detector and Cu anode in the Department of Geological Sciences at Indiana University, Bloomington, IN.X-ray scans were run from 2 to 70 2θ with a 0.02 2θ step-size and count time of 2 s/step.Identification of major phases was performed using Bruker's EVA software and the International Centre for Diffraction Data Powder Diffraction File database.Rietveld refinement (TOPAS, Bruker) was used to quantify mineral abundances, unit-cell parameters, and non-instrument related peak broadening caused by microstrain and crystallite size.Raw diffraction data are reported in the Supplemental Information, Data S1.

| Raman spectroscopy
Raman spectroscopic analysis of graphite was performed on thin sections prepared from rock samples cut perpendicular to the dominant cleavage plane.A final polish using a 1 μm diamond paste was used on all samples.A thorough study by Lünsdorf (2016) assessed the Raman spectral attributes of CM induced by various degrees of surface polishing.Their study evaluated a suite of CM bearing rocks, ranging from sedimentary organic matter to semi-graphite that formed under similar metamorphic conditions to the present study.Their results demonstrated that down to a 1 μm diamond grit polish, Raman parameter values [D1/G amplitude ratio, G-peak location, scaled total area (STA)] showed no consistent variability with polishing grit size.Furthermore, Beyssac et al., 2003 reported that in poorly organized CM where the R2 area ratio (D1/G + D1 + D2) is higher than 0.5, there is no significant difference in the Raman parameters from polished CM relative to an unpolished CM.The mean and standard deviation R2 value for our sample set is 0.67 ± 0.01, well above the R2 = 0.5 threshold.Based on these studies and our results, Raman data were collected from CM exposed at the surface of polished thin sections.In recognition of the body of literature documenting the spectral artefacts due to polishing (e.g., Ammar & Rouzaud, 2012;Beyssac et al., 2003;Kouketsu et al., 2019;Pasteris, 1989), we collected Raman data from CM below a transparent grain from three of the samples and compared the spectra with the surface CM spectra and found no significant differences.
Spectra (n = 14-16 per sample) were collected on a Horiba Xplora Plus Raman microscope system at the U.S. Geological Survey (USGS) Raman Spectroscopy Lab in Reston, VA.Experimental parameters included using a 532 nm laser with perpendicular polarization, 1200 groove mm À1 spectral grating, 100 μm confocal pinhole diameter, 100 μm spectrometer slit, 75 μW laser power at the sample surface, and a 100 Â objective with a numerical aperture of 0.9.Each analysis was collected across a spectral range of 50-2664 cm À1 in three accumulations of 15 s.The laser spot size was conservatively estimated to be 1 μm in diameter at the sample surface.Spectra were fit across a range of 500-2400 cm À1 using a linear background function and a sum of Lorentzian peak functions.
The raw spectral data are reported in Data S2.Previously published graphite geothermometry data (Beyssac et al., 2007) from nearby locations confirm lower greenschist facies metamorphic conditions, thus this study used the peak fitting method reported by Kouketsu et al. (2014) which was developed for rocks that experienced similar lower greenschist facies metamorphic conditions.Peak fitting results were averaged for each sample, and we report three Raman parameters [D1 peak full-widthat-half-maximum (FWHM), G-FWHM, and Raman Band Separation (RBS)] typically evaluated in these thermal conditions to assess graphitization (Henry et al., 2019, and references therein).Data were analysed using the peak fitting module in the IGOR Pro software suite.
In the first order region between 1000 and 2000 cm À1 , the Raman spectra of CM are divided into two regions commonly referred to as the D-band (or disorder) region and the G-band (or graphite) region (Figure 2a).In the CM evaluated here, three D-bands were evaluated including the D4-band at $1200 cm À1 , the D1-band at $1340 cm À1 , and the D3-band at $1500 cm À1 .The D4 and D3 bands are relatively minor and are only present in poorly organized CM, disappearing with increasing aromatic organization (Wopenka & Pasteris, 1993).The D1-and G-bands, the two main peaks in CM spectra, represent in-plane vibrations within aromatic rings, where the D1 band is commonly attributed to distortions in the aromatic structure due to the presence of heteroatoms or lattice defects (Beny-Bassez & Rouzaud, 1985) and the G-band is related to vibrations of aromatic carbon in organized graphene sheets with E 2g2 symmetry (Tuinstra & Koenig, 1970).
A second, rarely investigated spectral region in geologic research was identified between 800 and 900 cm À1 where we observed two low-intensity peaks that correspond to the Raman out-of-plane B 2g band located at 865 cm À1 (Kawashima & Katagiri, 1999;Tan et al., 2004;Zhai et al., 2016) and the in-plane D 000 band located at 857 cm À1 (Figure 2b; Tan et al., 2001;Zhai et al., 2016).Due to the low signal-to-noise ratio in this region, all spectra from each sample were averaged from 700 to 1000 cm À1 and fit across the same range with a cubic background and a sum of two Lorentzian peaks (Figure 2b).In this study we only report results from the Raman Shift (cm -1 ) Intensity (cts.) Raman Shift (cm -1 ) Example Raman spectrum of carbonaceous material (red) and associated sum fit (black) for the four main vibrational modes using a least-squares fitting regression, Lorentzian peak shape, and linear background.Blue curves in lower plot are the fits for the individual peaks.D1-full width at half maximum (FWHM), G-FWHM, and Raman band separation (RBS) are noted with black dashed lines.(b) Example 700-1000 cm À1 region generated (red curve) from averaging all spectra from a single sample.The B 2g and D 000 peaks were fit using a cubic background and Lorentzian peak shape.Blue curves represent the individual peak fits, and the black curve is the sum fit.
B 2g peak, as the D 000 peak was not discernable in some samples with lower signal-to-noise values.Finally, multiple linear regression analysis using a least-squares approach was performed with the Microsoft Excel Analysis ToolPak to predict Raman CM parameters from XRD data.

| Electron microprobe analysis
Electron microprobe data were collected on an automated five-spectrometer JEOL 8900 electron microprobe at the USGS Reston Microbeam Lab in Reston, VA and are reported in the Supplemental Information, Data S3.
The microprobe was operated using Probe for EMPA software (Donovan, 2015).Five samples were selected for analysis based on XRD results that indicated the presence of both calcite and dolomite.A 15 keV accelerating voltage was used for all samples and the beam current was set to 10 nA with a spot size of 10 μm to minimize beam damage.Examples of analysed calcite grains in contact with Fe-dolomite are shown in the Supplemental Information, Data S4.Data were collected on thin sections polished to a 1 μm finish.The spectrometer configuration, count times, approximate detection limits, and primary standards are given in the associated USGS data release (Stokes et al., 2023).

| Electron petrography
Back scattered electron (BSE) and coloured cathodoluminescence (CL) images were collected under high-vacuum conditions on a FEI Quanta FEG 400 SEM equipped with a GATAN Chroma CL detector in the Department of Geological Sciences at Indiana University, Bloomington, IN.High-resolution colour CL scans (1000 Â 1000 pixels, 1000 μs dwell time) were collected using a 10-12 kV accelerating voltage, 1-3 numerical aperture, and a 3-5 μm spot size.

| RESULTS
All results, including raw data, are available in the supporting information tables, as well as in the associated USGS data release at https://doi.org/10.5066/P96B9NJV(Stokes et al., 2023).

| Field relationships
The Lishan fault is an abrupt fault that puts the middle Miocene Lushan Formation to the east against the Eocene Paileng Formation to the west (Brown et al., 2012).Rocks on either side of the fault are dominated by slates and interbedded sandstone and minor carbonate beds.Folds are present but not common.Fieldwork across a 2 km transect showed that bedding was transposed into the slaty cleavage orientation (Figure 3b; Brown et al., 2017), and cleavage was increasingly well developed to the west (Figure 3c,d).At the eastern-most locations (e.g., 3D3, Figure 3a) the coarser grained, more quartz-rich layers tend to be fractured and healed by quartz or carbonate veins whereas the finer-grained, more micaceous layers have pervasive cleavage and fracture parallel to the bedding (and cleavage).Some rocks (e.g., sample 4C2A) contain intensely deformed septarian carbonate concretions of variable sizes and aspect ratios (Figure 3c).

| Powder X-ray diffraction
Ten slate samples were processed for whole-rock powder XRD analysis and the results of the Rietveld refinements are presented in Table 1.All samples are rich in quartz and muscovite, with quartz ranging from 30 to 50 wt.%and muscovite contents ranging from 25 to 42 wt.%.
The XRD data showed broad and variable 00ℓ muscovite peak positions and shapes.Petrographic results show that multiple muscovite populations are present and based on these observations, we modelled the muscovite with two 2 M1 phases to accommodate the peak broadening derived from, at least in part, coexisting muscovite populations.Quantitative evaluation of the muscovite 004 peak for the sample suite shows a linear correlation for both muscovite peak position (R 2 = 0.78) and FWHM (R 2 = 0.56) across the transect, crossing the Lishan fault (Figure 4).Here, the 004 peak position trends towards higher 2θ values (or lower d spacings), and FWHM values decrease from east to west.
In order to model peak broadening with geologically reasonable results, two dioctahedral 2 M1 micas, muscovite (a) and muscovite (b) (hereafter referred to as Ms (a) and Ms (b); Table 1), were used in the Rietveld refinements.Unique but overlapping limits were placed on the unit-cell parameters for both muscovite phases to prevent coalescence of the two mica fits during refinement, as is common when trying to model multiple populations of the same mineral with the same unit cell constraints.Unit cell limits for Ms Although not commonly reported in geologic literature but often quantified in Rietveld refinements, these two parameters are used to evaluate the size and structural coherency of X-ray scattering domains, commonly referred to as crystallites.The microstrain parameter also provides insight into compositional heterogeneity and disorder in individual mineral phases.Refinement results show that Ms (a) is the dominant phase, making up >94% of the total muscovite.Crystallite size was not found to be a significant contributor to peak broadening, but microstrain values were large and variable for Ms (a), ranging from 0.28 in sample 4F to 1.0 in sample 3C (Table 1).

| Petrography
The slates in these rocks are dominated by quartz and muscovite in sub-equal amounts, with lesser amounts of chlorite, feldspars, pyrite, and carbonates (Table 1).Most grains are < 50 μm in diameter requiring electron petrography to show the textural relationships among grains.
Here we also characterize the slaty cleavage, so that its development can be compared to the graphitization of CM.
Quartz grains are 20 to 50 μm long with aspect ratios of 0.5 to 0.2 and are set in a matrix of fine-grained muscovite and chlorite (Figure 5a).Cathodoluminescence imaging shows that the cores of some pale blue luminescent grains are fractured and extended (Figure 5b, c).The cracks are filled with dull red to dark blue luminescent quartz, a colour similar to the overgrowths of these grains, and to other entire grains.The equant shape of many quartz grains, only observed in CL images, suggests that they are detrital, where longer grains may host metamorphic overgrowths, or strain shadows (after Passchier & Trouw, 2005).Albite grains are similarly equant with luminescent pink cores and dull brown overgrowths in CL (Figure 5b).
Muscovite occurs in three textural varieties.The first generation of muscovite occurs as detrital grains which are typically less than 20 μm long and 2-3 μm thick (white arrows, Figure 5d) but may exceed 50 μm length (Figure 5e).Rare polycrystalline grains up to 70 μm long  (Figure 5f) suggest a metamorphic provenance for these grains.Some single detrital grains appear transposed into the overprinting cleavage orientation where they may be distinguished from cleavage-forming muscovite grains by their larger size (Figure 5d,e).
A second generation of muscovite occurs as intergrowths with chlorite forming chlorite-muscovite 'stacks' (CMS).These typically occur with chlorite > muscovite where cleavage planes may be at a high angle to the penetrative cleavage (e.g., Figures 5g,h and 6c,d,f).Rare grains up 100 μm thick and 300 μm long occur at a high angle to the S n cleavage (Figures 5h and 6b).Their near random orientation is consistent with the view that they form as pre-or early syntectonic porphyroblasts (Passchier & Trouw, 2005).
Cleavage-forming muscovite is clear in the field where its preferred orientation gives the rocks a strong fissility and a conspicuous sheen on cleavage planes that define the metamorphic fabric, S n .Individual flakes are 1-2 μm thick and rarely longer than 10 μm (Figure 5).These flakes truncate all earlier grains, including detrital quartz (Figure 5a-c, i) and muscovite (Figures 5d,f and  6e), CMS (Figures 5g,h,k and 6d,f), and carbonate rhombohedra (Figures 5i and 6a).These coalesce to form anastomosing networks of folia that are hundreds of μm long but only a few flakes thick (Figure 5l) and cradle larger quartz and feldspar grains and CMS.
Calcite and dolomite are present in many samples and both were present in quantities >1 wt% in five of the 10 samples (Table 1).Mg-dolomite occurs as polygonal to subhedral grains 10-15 μm in diameter and is ubiquitously overgrown by Fe-rich dolomite, typically subrhombohedral in shape (Figures 5i,j and 6a and Data S4).Calcite occurs as irregularly shaped masses up to 100 μm long and tends to include round grains of quartz (Figures 5j and 6a).Instances of calcite and dolomite in contact show grain boundaries that are typically straight (e.g., Data S4e) and the rarity of colocation and nature of the boundary shapes suggests they are in equilibrium.Dolomite and calcite grains are commonly replaced by the cleavage forming micas.Evidence for this dissolution is preserved by the truncations of internal features in the carbonates such as the boundaries between Mg-dolomite cores and Fe-dolomite rims (Figures 5i,j and 6a) and quartz inclusions in large calcite grains (Figure 5j).

| Carbonaceous material textures
CM particles of varying size are disseminated throughout all samples and were evaluated using reflected light microscopy.CM particles are irregular in shape but are often elongate parallel to the direction of the cleavage plane defined by phyllosilicates.Most CM particles are <20 μm in length, but rare grains exceeding 100 μm in length also occur.Figure 7 shows representative images of CM textures from samples 4D and 3D3.Reflected light images from both samples demonstrate the alignment of CM within and parallel to the bedding parallel cleavage.Several grains appear to have been pulled apart (Figure 7b), but it is unclear if this texture is inherited from low-temperature maturation processes (e.g., solid bitumen migration) or developed during deformation.Coarser CM particles with lower aspect ratios than those observed in the fabric-proximal grains tend to occur next to clasts of CMS (Figure 7c) or large grains of quartz or carbonate.In general, the textural occurrence, shape, and size of CM in these samples are similar to those of organic macerals in unmetamorphosed marine shales (e.g., Hackley & Lewan, 2018;Wei et al., 2021).In clayrich shales, similar to the protolith of the slates studied here, organic grains tend to be concentrated with clayrich zones due to marine depositional processes and can have high aspect ratios, parallel to the sedimentary fabric.It follows that the similar textures observed in our samples are at least in part inherited from a sedimentary shale protolith and that the CM is of organic origin.

| Raman spectroscopy
Averaged peak fitting results from Raman spectroscopic analysis are summarized in Table 2. Standard deviation values calculated from the averaged first-order region peak results are <3%, relative to the average, indicating little intrasample heterogeneity with respect to CM crystallinity.Spectral peak fitting from the 800-900 cm À1 region was performed on a single spectrum generated by averaging all spectra from each sample.The signal to noise ratio was improved, but precision errors associated with the peak-fitting results from this region are still relatively large (i.e., $±20%).
Raman parameters RBS, D1-FWHM, G-FWHM, and B 2g -FWHM are strongly correlated with sample longitude, that is, along the east to west transect, perpendicular to the trace of the Lishan fault (R 2 = 0.90-0.96; Figure 8).Values for RBS, D1-FWHM and B 2g -FWHM decrease linearly westward, whereas G-FWHM increases (Figure 8).Using these Raman parameters, three different graphite geothermometers appropriate for use in lower greenschist facies metamorphic conditions were evaluated (Kouketsu et al., 2014;Rahl et al., 2005; Table 2).The two Kouketsu geothermometers are derived from linear empirical relationships with the D1-FWHM and G-FWHM peak parameters, leading to temperature trends that mimic the trends of the Raman results (Table 2).The Rahl et al. (2005)  graphite geothermometer uses both peak area height data for the D1, D2, and G-bands, thus the calculated temperatures for this dataset using this geothermometer are mostly constant across the transect (Table 2).

| Calcite-dolomite geothermometry
In order to constrain metamorphic temperatures across the sample transect, calcite-dolomite geothermometry (Anovitz & Essene, 1987) was applied.The 1-sigma standard deviations of the averaged temperatures for each sample were assigned as the error (Table 3).Calcite and dolomite are present in six of the ten samples, five of which contain more than 1 wt.% of both (Table 1).In these samples, calcite grains typically have straight grainboundaries when in contact with Fe-rich rims of the dolomite rhombohedra suggesting that neither appear to be forming at the expense of the other and are in thermodynamic equilibrium (Data S4).
Temperature estimates were calculated using electron microprobe data collected from the calcite grains in contact with the Fe-rich rims of dolomite rhombs.The Fe and Mg atoms per formula unit (apfu) values are based on pure calcite (CaCO 3 ) and reported errors are the 1-sigma standard deviation values (Table 3).Average values of Mg increase from 0.010 ± 0.001 apfu to 0.014 ± 0.002 apfu from east to west along the transect.The Fe values range from 0.017 ± 0.008 apfu in sample 4C2A to 0.023 ± 0.003 apfu in sample 3D2.Calculated temperatures are shown in a box and whisker plot in Figure 9 and average values for each sample are reported in Table 3. From east to west, the calculated average temperatures are similar for samples 3D2, 3D3, and 3C at 235 ± 50 C and show a slight increase to 285 ± 32 C for sample 4C2A.The westernmost sample, 4F, shows the highest temperature of 320 ± 32 C but, in consideration of the standard deviations, is only slightly higher than the eastern-most samples (3D2 and 3D3).

| Constraints on metamorphic conditions
Calcite-dolomite geothermometry constrains the metamorphic temperatures experienced by the rocks studied here.Textural relationships between the carbonate grains and the cleavage suggest that they crystallized just before or during early stages of ductile deformation, based on preservation of euhedral to subhedral grain shape of the zoned dolomite rhombs coexisting with the occasional truncation features of carbonate grains by the micaceous cleavage (Figure 5i, j).Four samples from the eastern section of the transect, all from the Lushan Formation (3D2, 3D3, and 3C), show nearly indistinguishable temperatures of 235 ± 50 C whereas the western most sample, 4F, across the Lishan fault in the Paileng Formation, yields a temperature of 320 ± 32 C (Figure 9).These results are similar in trend to published graphite geothermometry data in the region of our study area, where peak metamorphic temperatures in the Hsüehshan Range to the west of the Lishan fault tend to be higher than those F I G U R E 5 Back scattered electron (BSE) and cathodoluminescence (CL) images of typical microstructures in the slates of the Lushan and Paileng formations.(a) Quartz and albite grains (dark grey) set in a matrix of anastomosing S n folia (yellow dashed lines) of fine grained muscovite flakes (medium grey) that truncate (Tr) detrital quartz and albite grains.(b) CL image of the region shown in (a).Most quartz grains show a dak maroon colour, but some grains contain a detrital quartz core (pale to bright blue).One detrital quartz core is fractured (Fr), and others have generations of overgrowths (og) characterized by dark blue-grey quartz followed by dark maroon quartz, often in strain shadows (ss).Truncations of detrital core and overgrowths are indicated (Tr).  in the Backbone Slate Range to the east of the fault (Beyssac et al., 2007).
Results from three different graphite geothermometers show opposing trends across the transect and the underlying reasons for this will be discussed in detail in later sections.However, temperatures calculated following the Kouketsu et al. (2014) D1-FWHM geothermometer (Table 2) are in statistical agreement with the calcite-dolomite geothermometry results (Table 3).While these different geothermometers agree, the D1-band records the structural improvement of the CM through the annealing of defects and removal of heteroatoms (Beny-Bassez & Rouzaud, 1985), and while this process is well documented as being induced by thermal energy (e.g., Pasteris & Wopenka, 1991), laboratory studies have shown that the energy imparted by strain can also yield similar results (e.g., Ross & Bustin, 1990).
Like other studies investigating the role of deformation in natural graphitization (e.g., Barzoi, 2015;Barzoi & Guy, 2002;Kedar et al., 2020;Muirhead et al., 2021), the role of thermal energy must be thoroughly evaluated, thus seismic frictional heating must be ruled out as a heat source.In this study area, the ductile deformation associated with the Lishan fault appears to be accommodated by a wide (1-2 km) ductile shear zone (Figures 3c,d, 5, and 6;Brown et al., 2012), with little to no evidence of focused seismic slip and associated high-strain rate required for widespread thermal heating (Kaneki et al., 2016).The linear nature of the data presented (Figures 4 and 8) and petrographic observations of ductile deformation structures (Figures 5 and 6) are consistent with crossing a wide strain gradient.Moreover, calcitedolomite and D1-FWHM geothermometry results do not show a profile of steeply increasing temperature as would be expected from shear heating at the fault line.

| Slaty cleavage development
The microstructures described above allow the origin of the slaty cleavage to be explored.All samples show the development of at least some slaty cleavage.The intensity of this S n cleavage increases to the west as reflected in the increasing amount of small muscovite grains and lesser chlorite flakes that define the cleavage, and by the increasing contiguity of muscovite flakes defining the folia.However, the intrasample development is heterogeneous, with microlithons showing very little cleavage, whereas the surrounding folia show increasingly well-developed cleavage (compare Figures 5a,e,i from the same sample).The trend of mica recrystallization parallels a trend of increasing aspect ratio of quartz grains elongate parallel to the cleavage.At the grain scale, muscovite and chlorite flakes in the cleavage folia truncate the boundaries between quartz and plagioclase cores and their overgrowths.This texture is evidence for the cleavage post-dating both the fracturing of the cores of the detrital quartz grains and of the crystallization of the overgrowths (Figure 5b,c).Cleavage-forming muscovite grains also truncate detrital muscovite (Figures 5d  and 6e) and muscovite in the CMS (Figures 5g,h,k and  6d,f).Carbonate grains are similarly shortened against the cleavage where the boundary between dolomite cores and Fe-dolomite overgrowths are truncated (Figure 5i,j).In all cases, the removed parts of grains against these surfaces are parallel to the cleavage, that is, those surfaces face the shortening direction.In contrast, the surfaces facing the elongation direction (parallel to the cleavage) are either unmodified or contain metamorphic overgrowths in strain shadows (Figures 5b,c,l and 6b,c).
These truncations and associated overgrowths constitute strong evidence for pressure solution of grains facing the shortening direction and precipitation in extensional sites, contributing to the growing body of literature documenting slaty cleavage development in lower greenschist facies conditions (e.g., Akker et al., 2021;Ho et al., 1996;Kisch, 1991;Knipe, 1981;Lee et al., 1986;Wenk et al., 2020;White & Knipe, 1978;Wintsch et al., 1993).In particular, the petrography shows that both detrital muscovite grains and muscovite in the CMS are dissolved in favour of the cleavage-forming muscovite defining S n .The progress of these muscovite dissolution-precipitation F I G U R E 6 BSE images of textures and microstructures in a sequence of slates from east to west showing the westward increase in development of the S n cleavage (yellow dashed line).(a) Carbonate grains in a matrix of quartz all truncated by a poorly developed S n cleavage.Fe-dolomite overgrowth around a dolomite core embays marginal quartz.Other subhedral grains of Fe-dolomite embay detrital albite grains.(b) A large CMS in a matrix of quartz, muscovite, and chlorite.Cleavage is poorly developed but anastomoses around the CMS.(c) A lens-shaped microlithon containing a CMS with strain shadow overgrowths of quartz and minor chlorite.A relatively rare band of quartz with minor chlorite is sandwiched between two closely spaced muscovite folia at the top of the image.(d) A polycrystalline CMS truncated by two muscovite-rich cleavages.(e) A strongly foliated slate with almost all quartz grains sheathed in muscovite-and chlorite-rich folia.Examples are highlighted with yellow dashed lines.Truncations are common (some indicated with a white arrow).(f) Closely spaced muscovite folia defining S n that truncates quartz grains and CMS.Abbr: Ab: albite, Dol: dolomite, Fe-Dol: Fe-rich dolomite, Ms: muscovite, Qz: quartz, ss: strain shadow, fr: fracture, tr: truncation, BSE: back scattered electron, CMS: chlorite-muscovite stack.
reactions is documented by the east to west evolution of the XRD muscovite results (Figure 4).The linear shift of the 004 muscovite peak to a lower 2θ peak position (larger d-spacing), narrowing of the peak shape, as well as a general decrease in the Ms (a) microstrain parameter (Table 1), together reflect the dissolution of older muscovite populations out of chemical and textural equilibrium with lower greenschist facies ductile deformation conditions, and precipitation of the S n muscovite population.The increase in the S n muscovite population relative to the detrital and CMS populations is strong evidence that the transect of this study crosses a strain gradient.

| Strain-induced graphitization in a ductile shear zone
The data discussed to this point establish a lower greenschist facies strain gradient crossing the Lishan fault, documented by XRD results and corroborated by SEM observations of mica recrystallization via a strain-induced dissolution-precipitation process.The Raman parameters from CM also show strong linear trends across the strain gradient, however interpretations of the trends across the gradient are complicated by deviations in our data from the typical trends observed for Raman peaks with increasing metamorphic grade.A summary by Henry et al. (2019) of published Raman results and associated graphite geothermometry data documents that with increasing structural ordering by thermal processes above $250 C, G-FWHM, RBS, and D1-FWHM values should decrease with increasing temperature.However, our data clearly deviate from this correlation of thermal graphitization in that G-FWHM increases where the other parameters decrease (Figure 8).These deviations in Raman peak parameters from thermal graphitization behaviour have important consequences for the application of metamorphic geothermometers based on the Raman response of CM in deformed terranes.To illustrate these consequences, we evaluated our Raman data against three widely applied Raman metamorphic geothermometers.We focused on the study by Kouketsu et al. (2014)  these results to values generated from the graphite geothermometer provided in Rahl et al. (2005) which uses peak parameters from the D1-, D2-, and G-bands.
A comparison of our Raman data against the Kouketsu dataset is provided in Figure 10 where D1-FWHM is plotted against G-FWHM for the samples from this study and a subset of data from Kouketsu et al. (2014).Temperatures for both datasets were calculated using the two thermometers provided in the Kouketsu et al. (2014) study, labelled G and D1 in Figure 10 and listed in Table 2.One of the eastern-most, and least deformed sample from our transect (3D2), shows D1-and G-FWHM values and corresponding calculated temperatures (D1: 241 ± 30 C; G: 253 ± 30 C) notably consistent with the Kouketsu sample from similar thermal conditions (D1: 237 ± 30 C; G: 247 ± 30 C).However, outside of the three eastern-most samples (3C, 3D3, and 3D2), our data deviate linearly from the Kouketsu dataset towards higher G-FWHM with decreasing D1-FWHM.This divergence is reflected in the calculated D1-FWHM based temperatures that show a modest linear east to west increase from 241 ± 30 C to 322 ± 30 C, whereas the G-FWHM based temperatures show a slight decrease, but overall little change across the transect ( 253  the temperatures calculated for sample 3D2 but are consistent (327-339 ± 50 C) across the transect, reflecting the inversely related G-and D1-peak parameters in our dataset.These observations highlight the potential for misleading graphite geothermometry results from deformed rocks, at least at thermal conditions near 300 C.
In lieu of graphitization driven by thermal processes, we suggest microstructural changes to sample CM induced by strain to explain the Raman spectra behaviour across the sample set.Previously published transmission electron microscope (TEM), XRD, and reflectance data from graphite and coal samples deformed in laboratory experiments under different stress conditions reveal the complex nature of the graphitization process and the role that strain plays at the nanometre scale (Bonijoly et al., 1982;Bustin et al., 1995Bustin et al., , 1995;;Ross & Bustin, 1990).The process involves the heterogeneously distributed collapse and coalescence of pores, and the rotation and alignment of basic structural units, effectively re-arranging graphene sheets to form threedimensional graphite domains.These processes of structural improvement were unequivocally more efficient in deformation experiments, and our data provide evidence showing the manifestation of deformation and strain energy for naturally deformed CM.
Our findings of G-FWHM increasing with decreasing D1-FWHM values, concomitant with a strain gradient, may be explained by a decrease in the average graphite T A B L E 3 Electron microprobe data collected from calcite.Averaged X Mg-calcite and X Fe-calcite values are reported as atoms per formula unit (apfu) based on CaCO 3 .Temperatures were calcuated based on the equations provided in Anovitz & Essene, 1987. Errors   crystallite diameter documented by increasing G-FWHM (Maslova et al., 2012) but improved structural coherency (decreasing D1-FWHM) as a function of increasing ductile deformation.While the available data from the earlier laboratory studies evaluating the role of deformation on graphitization are limited and direct correlations with our study are unclear at this stage, the Raman results presented here suggest that the importance of strain energy for graphitization documented in laboratory settings may be extrapolated to naturally deformed rocks, and potentially monitored by the Raman response of CM.
Finally, an analysis of the B 2g vibrational mode in graphite is not, to our knowledge, reported in geologic literature but has been utilized by nuclear science researchers to characterize graphite recrystallization during irradiation (Kawashima & Katagiri, 1999;Tan et al., 2004;Zhai et al., 2016).The 'out of plane' B 2g mode becomes Raman active when slight shifts in the crystal structure result in lower symmetry (Zhai et al., 2016) and was first reported where Raman data were collected along the edge planes of highly oriented pyrolytic graphite (HOPG) crystals (Kawashima & Katagiri, 1999).In the case of graphite from deformed metamorphic rocks, we expect the B 2g peak to be most likely observed in spectra collected from samples oriented perpendicular to the dominate cleavage plane, the typical sample orientation for petrographic studies.B 2g -FWHM values from this study decrease from east-to-west across transect, indicating an improvement in the out-of-plane, or c-axis parallel, structural coherency.In other words, the decrease in the B 2g parameter across the low-to-high strain gradient is consistent with coarsening through an increase in the number of stacked graphene layers and improved alignment per crystallite of graphite.
Based on the Raman data presented in this study, we propose a simple conceptual model in Figure 11 to help illustrate our summarized interpretations from this study.With increasing degree of ductile deformation across our sample transect, graphitization may occur by improved in-plane structural coherency, a decrease in the average graphite crystallite diameter, an increase in the number of stacked graphene sheets, and their improved alignment.

| Unified graphite-muscovite response to ductile deformation
The linear nature of the CM Raman results across the transect and the similarly linear muscovite XRD results prompted further investigation of the relationship.We applied multiple linear regression modelling to further explore the correlative response to strain for these different materials.We present a regression with the most correlative datasets, however, numerous trials with different combinations of variables (e.g., different minerals and Raman parameters) were performed to ultimately maximize the R 2 value, minimize the difference between the R 2 and the adjusted R 2 , and minimize the Significance F which describes the predictive power of the regression.An optimized regression is presented in Figure 12 in which D1-FWHM graphite values are predicted with high confidence (adjusted R 2 = 0.92) by a linear equation that includes Ms (a) microstrain data and the wt.% ratio of the two muscovite populations (Table 1).The Significance F value was two orders of magnitude lower (0.0003) than a common threshold of 0.05, signifying the validity of the model as a strong empirical predictor.The p-values, a term that tests the null hypothesis of the coefficients for the independent variables Ms (a) microstrain and muscovite ratio, were similarly low (<0.0003).This confirms their independent utility as predictors of D1-FWHM.G-FWHM was also evaluated as a dependent variable, and while the adjusted R 2 value (0.61) was lower than the regression model predicting D1-FWHM, the model was still statistically significant (significance F = 0.01; pvalues = <0.01;Data S5).The predictability of the D1-FWHM and G-FWHM Raman parameters from muscovite XRD results that record recrystallization due to strain-induced cleavage development provides further evidence for the influence of deformation in the processes of graphitization as recorded in the CM structure.
The degree to which the bulk crystallinity of muscovite can be predictive of the bulk CM crystallinity is surprising considering the different processes accommodating the deformation.The replacement of earlier generations of muscovite by S n muscovite occurs along grain boundaries and involves the aqueous dissolution of the former and the precipitation of the latter.These grain-scale observations across the strain gradient are similar to those made by numerous studies exploring mechanism of slaty cleavage development and documenting the associated mineral textures (e.g., Akker et al., 2021;Ho et al., 1996;Knipe, 1981;Wenk et al., 2020;White & Knipe, 1978;Wintsch et al., 1993).In contrast to the strain-induced muscovite recrystallization outlined above, graphitization of organic precursors occurs by in situ processes of aromatic organization and expulsion of heteroatoms, hydrocarbons, CO, and CO 2 (Mastalerz et al., 1993;Wintsch et al., 1981;Wopenka & Pasteris, 1993).The observed textures of CM and the low intrasample variability in Raman results indicate that precipitation of the CM from carbonrich fluids is unlikely in the rocks evaluated by this study.However, it is possible that grain boundary fluids may play a ubiquitous role in facilitating in situ aromatization, where grain boundary fluid-derived H + quenches free radical sites and cleaves non-aromatic components (Hackley et al., 2022;Lewan, 1997) and provides both a medium and pathway for their expulsion.More work is warranted to understand the role of grain boundary fluids in graphitization at these thermal conditions, but what is made clear from this study is that deformation activates graphitization and dissolution-precipitation processes equally.

| Structural implications for the Lishan fault
The Lishan fault separates Eocene rocks from Miocene rocks across a sharp contact (Brown et al., 2012).The carbonate thermometry shows that the Eocene rocks reached $325 C on the west, whereas the Miocene rocks reached $250 C on the east (Figure 9).If the geothermal gradient proposed by Lee et al. (2022) of 30-50 C/km is correct for this area, and the geothermometry values are accurate given the large errors, then the displacement across the fault could reflect a net displacement of $ 2 km.However, the S n cleavage crosses the Lishan fault smoothly without interruption.The cleavage also truncates carbonate grains from which the temperature estimates were derived.It follows that the cleavage-forming event occurred after the eastern rocks had reached their $250 C temperatures, and thus, after most (or probably all) of the west-side-up displacement had taken place.Given that the thick quartzites of the Tachien member, the lower part of the Paileng Formation, west of the Lishan fault would act as a buttress to deformation, it is not surprising that this ductile pressure-solution strain would occur in the same region as the discrete Lishan fault.However, the ductile strain would be more widely distributed but increase towards that buttress.This study presents several lines of evidence documenting a ductile strain gradient in slates across the Lishan fault in central Taiwan, providing a natural laboratory to evaluate the role of strain in graphitization at lower greenschist facies conditions.Linear trends of muscovite XRD-peak data paired with SEM observations of truncationreplacement textures record fluid-assisted dissolutionprecipitation processes facilitating the slaty cleavage development.Across the transect, Raman results from CM were similarly linear, but divergent from typical trends expected from thermally induced graphitization.These results are interpreted to reflect the unique influence of ductile deformation on the graphitization process.The similarity of the carbonate geothermometry results and the D1-FWHM temperatures suggest that the D1-band Raman results reflect the crystallinity due to thermal energy, however this study suggests that in these rocks, the decrease in D1-FWHM due to the removal of heteroatoms and lattice defects is, at least in part, the result of strain energy.The more apparent impact of the strain gradient on the Raman response of CM is the increase in G-FWHM, which is interpreted as a reduction in the average graphite crystallite diameter.Three different graphite geothermometers using different Raman peak information yielded increasing, decreasing, and static temperature profiles across the transect, highlighting the potential, albeit subtle, complexities arising from evaluating CM crystallinity in naturally deformed rocks.
Multiple linear regression analysis revealed that the progressive process of in situ graphitization monitored by D1-FWHM and G-FWHM can be predicted by the XRD muscovite data which records recrystallization by dissolution-precipitation mechanisms.The findings from this study provide insight into the correlative effects of deformation in different materials as well as a new perspective on strain induced graphitization.

F
I G U R E 1 (a) Geologic map of Taiwan modified from Fisher et al. (2007).CP: Coastal Plain; WF: Western Foothills; HR: Hsüehshan Range; BS: Backbone Slate Range; ECR: Eastern Central Range; CR: Coastal Range; LF: Lishan fault; LVF: Longitudinal Valley fault; Fm: formation.(b) Geologic map inset modified from Brown et al.
ty clea v a g e c l e a v a g e | | b e d d i n g L is h a n fa u lt N F I G U R E 3 (a) Sampling locations across an E-W transect along a tributary feeding into the Zhuoshui River valley.Approximate location of the Lishan fault (Brown et al., 2012) is shown with a white dashed line.(b) Poles to dip and dip direction are plotted on a quarter lower hemisphere projection stereonet.Semitransparent blue and grey circles represent the average and standard deviation (1 sigma) of the metamorphic (black) and sedimentary (blue) fabric measurements.(c) Field photo of a septarian concretion deformed parallel to the slaty cleavage.Dashed lines outline the slaty cleavage around the concretion.(d) Field photo showing interbedded quartz-and carbonate-rich layers.The ++ quartz, + quartz, and +mica indicate relative amounts of quartz versus mica based on field observations.Lighter layers with higher amounts of quartz and/or carbonate tend to have crosscutting veins (white bands).
U R E 4 X-ray diffraction results for the 004 muscovite peak.(a) 004 peak position and full width at half maximum (FHWM) values are plotted against longitudinal geographic location.The approximate location of the Lishan fault is shown with a black dashed vertical line.Coarse dashed lines are the linear regression fits for each data set with associated correlation coefficient (R 2 ).(b) 004 muscovite peaks for the western (4F) and easternmost (3D2) samples demonstrating the differences in peak shape and position across the transect.Error bars on data are not visible as they are smaller than the symbols.
Legend on next page.
(c) BSE and companion CL image of fractured detrital quartz grain showing quartz overgrowths in the strain shadow (ss) and truncation of the core-overgrowth contacts by muscovite defining S n .White dashed line shows the core-overgrowth boundary invisible in paired BSE image.(d) BSE image showing detrital muscovite flakes (white arrows) aligned to the upper left in a host of detrital quartz grains cradled by E-W S n cleavage defining muscovite.(e) BSE image of a less common large detrital muscovite grain (DMG).(f) BSE image of composite detrital muscovite grain truncated along its margins by S n .(g) BSE image of chlorite (pale grey)-muscovite (medium grey) stacks (CMS) with quartz (dark grey) overgrowths in the strain shadows, truncated by muscovite and chlorite defining S n .(h) BSE image of a microlithon containing a chlorite porphyroblast truncated by muscovite and chlorite that define S n .Overgrowths of quartz with randomly oriented chlorite flakes and CMS form on the margins of the porphyroblast in the strain shadow.(i) BSE image of a weakly zoned euhedral dolomite porphyroblast with a euhedral Fe-dolomite overgrowth, all truncated by muscovite defining the S n foliation.(j) BSE image of dolomite grain with Fe-dolomite overgrowth, and a calcite grain with quartz inclusions.Both are truncated by muscovite that defines S n cleavage.(k) BSE image muscovite defining S n cleavage cutting CMS and detrital quartz and muscovite (white arrow) grains.(l) BSE image and semitransparent CL image overlay showing well-developed (S n ) foliation (one example shown in dashed yellow line) cradling each detrital quartz grain and overgrowths.Strain shadows of randomly oriented (left) and moderately oriented (right) quartz and muscovite sandwich the larger quartz grain, upper centre.Abbr: ab: albite, Cal: calcite Chl: chlorite, Dol: dolomite, Fe-Dol: Fe-rich dolomite, Ms: muscovite, Qz: quartz, ss: strain shadow, fr: fracture, tr: truncation, CMS: chlorite-muscovite stack, DMG: detrital muscovite grain.
U R E 6 Legend on next page.
as they used the same laser wavelength and provided two linearly empirical geothermometers based on D1-FWHM and G-FWHM (referred to as D2-FWHM in the Kouketsu et al. study).Additionally, we compared Reflected-light images showing the textures of graphite observed in two samples.(a) Reflected-light image of a welldeveloped cleavage band in sample 4D.CM is light tan.(b) Locations of CM from image (a) are shown in red.Traces of the chlorite-muscovite fabric are noted with black dashed lines.(c) and (d) reflected-light images from sample 3D3.CM particles are coloured translucent red.Traces of the chlorite-muscovite cleavages are noted with white dashed lines.Dol: dolomite, CMS: chlorite-muscovite stack, H: hole, Ms: muscovite, Chl: chlorite, CM: carbonaceous material, Qz: quartz.
± 50 C to 216 ± 50 C).The temperatures generated from the Rahl et al. (2005) equation are $90 C higher thanT A B L E 2 Mean Raman parameter values from each sample.Standard deviations (SD; 1-sigma) are reported for averaged D1-full width at half maximum (FWHM), G-FWHM, and Raman band separation (RBS) values from 14 to 16 spectra per sample.B 2g peak data are from a single spectrum generated by averaging all spectra per sample.A peak-fitting precision error (1-sigma) is reported for B 2g -FWHM.Graphite geothermometry results were calculated using the D1-FWHM and G-FWHM based (referred to as D2 in Kouketsu et al., 2014) equations with errors from Kouketsu et al., 2014 and Rahl et al., 2005.Summary of Raman data presented in Box and whisker plot summarizing the calcitedolomite geothermometry results.Circles represent the mean temperature for each sample, boxes represent the first and third quartile above and below the mean, and the whiskers represent the maximum and minimum values of the dataset.The approximate location of the Lishan fault is noted with a fine-dashed black line.U R E 1 0 G-FWHM versus D1-FWHM results for the data presented in this study (black circles, grey text) and a subset of data fromKouketsu et al. (2014) (italic black text and green squares).Temperatures were calculated using the D1-FWHM and G-FWHM (referred to as D2-FWHM byKouketsu et al.)  based equation inKouketsu et al. (2014) and have an error of ±30 C and ±50 C, respectively.Large arrows note the general trend of the two datasets towards higher temperature (green arrow) and increasing deformation (black arrow).
r e a s i n g d u c t i l e d e f o r m a t i o n c-axisF I G U R E 1 1 Simplified atomscale conceptual model derived from the interpretations of the Raman data presented in this study.Across the strain gradient, or with an increase in ductile strain (black arrow), G-FWHM increases indicating reduction in the average graphite crystallize size but a general increase in the structural coherency (decreasing D1-FWHM) and number of stacked graphene layers (decreasing B 2g -FWHM).Crystallographic c-axis noted to indicate growth of graphite domains in the out-of-plane direction.
U R E 1 2 Multiple linear regression model predicting D1-FWHM (Raman parameter) from muscovite Mu (a) microstrain and Mu (a)/(a + b) ratio values from Rietveld refinements of X-ray diffraction data.Red arrow notes east (E) to west (W) nature of the trend.Dashed black line is the regression line.Solid grey line is the 1:1 line.

Table 2 ,
plotted against longitudinal geographic location.The approximate location of the Lishan fault is noted with a fine-dashed black line.Coarse dashed lines are the best fit line for each dataset and R 2 values are listed on graph.RBS: Raman band separation; FWHM: full width at half maximum.