Samples were deemed to be subject to background strain only if they conformed to two criteria: (1) strain fabric at the sample site was parallel to the regional strain fabric in that area, and (2) the sample strain fabric was visually interpreted to be of similar intensity as the regional fabric in that unit. This interpretation, for the purposes of initial sample selection, was based on field observations and confirmed through inspection of micro-scale structures in thin section (note that most organic material was located between calcite grains and within seams of insoluble material). Practically, this meant that the sample was not part of a shear zone or a fault. Background samples were collected at distances of greater than 10 m from such localised high-strain zones. Where a high-strain zone was diffusely bounded, with a gradual return to background levels, the area was avoided entirely for the purpose of background sampling. Since the entire field area is part of a fold–thrust system, avoiding localised zones of high strain significantly limited potential sample sites. In total, 22 background samples from 15 different sites were included in the study, distributed approximately evenly within the intervening stratigraphy, between strained sites.

Four strained sample sites were selected (Fig. 2), three of which are centred around thrusts (Tenneverge, Salvadon, and Finive) and one in the Emaney shear zone. They are described in detail below. Displacements across thrusts and shear zones are estimated from cross-sections based on field mapping.

Raman spectroscopy measures the wavelengths of radiation produced by inelastic (Raman) scattering during the de-excitation of electrons in different molecular bonds, in this case focussing on those involved in different forms of organic carbon. In rocks, organic carbon can take on a range of nanostructures, depending on many factors during both deposition and burial: these include, but are not limited to, initial kerogen type, peak temperature, and the strain conditions to which the rock has been exposed. In the initial stages of burial, the carbon will have a nanostructure resembling that of kerogen (Thrower, 1989; Beyssac et al, 2002a; Rouzaud et al., 2015). As temperatures start to increase, the carbon nanostructure breaks down into smaller fragments as bonds are broken. With the application of strain (Kwiecinska et al., 2010; Kitamura et al., 2012, 2018; Savage et al., 2014; Furuichi et al., 2015) or very high temperatures (Wopenka and Pasteris, 1993; Oberlin et al., 1999; Schito et al., 2017), these fragments are aligned into parallel sheets, approaching a graphitic nanostructure. Graphitisation has not occurred in any samples in this study.

A total of 62 samples were crushed and powdered before being treated with 10 % HCl to remove inorganic carbon and therefore improve the signal-to-noise ratio when obtaining Raman spectra (Pasteris, 1989; Salver-Disma et al., 1999; Beyssac et al., 2002b; Mostefaoui et al., 2008; Muirhead et al., 2012). The residue was then rinsed and dried at room temperature to avoid thermal alteration. Using a Renishaw InVia Raman spectrometer at the University of Aberdeen, a 514 nm laser was targeted at individual grains in the residual powders; the laser power was <0.3 mW at the sample, and spot size was 1–2 µm. Each run comprised three co-adds of 5 s acquisitions to produce a single spectrum for analysis. This process was carried out on 10 individual grains from each sample.

Backscattered radiation was recorded, deconvolved, and analysed using Renishaw WiRE 3.4 software (Fig. 3). Using in-built software functions, noise reduction was first carried out on each spectrum (Fig. 3a), before baseline removal was performed using a cubic spline interpolation, which was user-guided (Fig. 3b). Finally, a Gaussian curve fit was applied to the two visible peaks in the spectrum (Fig. 3c), as in Bonal et al. (2006), and the Raman spectra parameters (peak intensity, position, width and area) were recorded (Fig. 3d; Quirico et al., 2009; Olcott Marshall et al., 2012). This process was carried out three times for each spectrum to minimise the error involved in the user-guided baseline removal process, resulting in 30 analyses per sample. Data points presented in this study therefore represent an average derived from 30 spectra per sample.

Different Raman spectral parameters are used in combination to determine the carbon nanostructure in a sample. Figure 3d highlights five key spectral parameters that can be calculated from the two curves fitted to the D and G peaks.

Peak area (A). The height and therefore area of a single peak are affected by signal strength, but comparing the areas beneath the D and G peaks negates this issue. The most common ratio comparison is known as R2, which is calculated as A[d] / (A[g] + A[d]).

Figure 4 shows a schematic summary of the changes in Raman spectral parameters with increasing temperature and strain. At low maturities, as amorphous carbon degrades with increasing temperature, the D peak increases in intensity (Tuinstra and Koenig, 1970; Levine, 1993; Oberlin et al., 1999). This increases the ID/IG ratio (Fig. 4a, i). Subsequently, as higher maturities are reached, carbonaceous fragments align into sheets and the G peak becomes more intense, decreasing ID/IG (Muirhead et al., 2012, 2017; Buseck and Beyssac, 2014). Generally, a decrease in ID/IG is observed when strain is applied to relatively low-maturity organic carbon (Fig. 4a, ii; Kwiecinska et al., 2010; Kitamura et al., 2012; Furuichi et al., 2015), but brittle fragmentation has been reported when mature, nearly graphitic carbon is subject to low-temperature strain, resulting in an ID/IG increase (Nakamura et al., 2015; Kirilova et al., 2018).

In addition to changing intensity, both the D and G peaks shift towards lower wavenumbers as the material approaches complete graphitisation, but the D peak shifts more significantly (Wopenka and Pasteris, 1993; Beyssac et al., 2002; Quirico et al., 2009). This causes the RBS to change (e.g. Zhou et al., 2014; Schmidt et al., 2017; Schito and Corrado, 2018; Henry et al., 2019). Figure 4b(i) shows an increase in RBS with increasing temperature at higher maturities, but little is known about how the parameter changes at low maturities such as in this study. The effect of strain has not been studied extensively, although Kuo et al. (2017) reported an increase in RBS with the application of strain (Fig. 4b, ii).

Figure 4c(i) illustrates how the R2 (curve area) ratio has been shown to correlate with temperature (Beyssac et al., 2002; Aoya et al., 2010; Nakamura et al., 2015; Chen et al., 2017; Kirilova et al., 2018; Henry et al., 2019). It follows a pattern similar to that of ID/IG with temperature, but little work has been done to establish whether this similarity extends to strain (Fig. 4c, ii).

It is generally accepted that the width (measured as full width at half-maximum, FWHM) of the D and G peaks changes with temperature (Zeng and Wu, 2007; Aoya et al., 2010; Kouketsu et al., 2014; Zhou et al., 2014; Hu et al., 2015; Bonoldi et al., 2016; Chen et al., 2017). However, the nature of the change varies depending on the thermal and barometric conditions, along with the nature of the organic starting material. At relatively low temperatures (<300 ∘C), these studies report a decrease in FWHM[d] with increasing temperature (Fig. 4d, i), which also correlates with an increase in ID/IG. FWHM[d] is thought to undergo very little change, if any, when exposed to differential strain (Ammar et al., 2011; Hu et al., 2015; Fig. 4d, ii).

Therefore, comparison of the relative intensities, positions, widths, and areas of the D and G peaks is a common method of assessing the extent to which a rock has been heated, which can be correlated with geological processes such as burial depth, contact metamorphism, and exposure to hot fluids. By comparison, the investigation of the effect of strain on these parameters is in its infancy (Kwiecinska et al., 2010; Kitamura et al., 2012; Furuichi et al., 2015; Kuo et al., 2017; Kedar et al., 2020).

ID/IG shows a general trend from lower values (0.3–0.4) in the upper stratigraphy to higher values (0.7–0.8) in the lowest stratigraphy (Fig. 5a) in background samples. This trend correlates with the depth through the thrust stack, and the graph of ID/IG with depth (Fig. 5a, ii) highlights the gradual increase towards higher values approaching the basal thrust. The Morcles thrust flattens towards the NW end of the cross-section, coinciding with a more vertical trend in ID/IG values. This trend appears to be disrupted across thrust planes and shear zones in the strained samples. In the case of the Salvadon thrust, the samples taken 10 m above and below the thrust plane are consistent with the regional trend, with values of 0.377 and 0.417, respectively. However, ID/IG values on the thrust plane are 25 %–30 % lower, with an average value of 0.249. There is a similar though slightly less significant drop on the Tenneverge thrust plane from 0.552 and 0.541 to 0.469, which is a decrease of around 18 %. In the case of both the Salvadon and Tenneverge thrusts, there is a gradual reduction in ID/IG approaching the thrust plane, which occurs more abruptly and closer to the thrust plane in the hanging wall than the footwall where the change is more gradual. On the Finive thrust plane there is a drop to 0.632 from surrounding values of 0.739 and 0.772. The Emaney shear zone, which is not immediately associated with a major fault, exhibits values of 0.624 and 0.627 within the shear zone itself, with higher values (0.772 and 0.888) outwith the shear zone, which is a similar change as observed at the thrust localities.

Raman band separation (RBS) varies through the stratigraphy (Fig. 5b), with what appears to be a prevalence of values >265 in the upper stratigraphy and <265 in the lower sequence. The average error associated with each sample is ±4, suggesting that the change through the stratigraphic sequence is not significant. Unlike ID/IG, there is no distinct shift on the thrust planes (Fig. 6); instead the RBS value on the Tenneverge and Salvadon thrust planes sits between the values of samples taken immediately above and below the thrust. There is a small shift to slightly higher values in the Emaney shear zone (263/261 from 255/259 below the shear zone), but these values are accompanied by an approximate error of ±2, which is comparable to the magnitude of the difference between samples.

R2 (the area ratio; Fig. 5c) shows little overall change with depth towards the Morcles thrust, although there is a weak trend towards higher values with increasing depth within individual thrusted packages. Outwith the strained samples, there is a prevalence of values <0.6 towards the top of the stratigraphy and >0.6 in the lower half, but many values are accompanied by errors as high as ±0.1. There is a marked drop on each thrust plane and within the shear zone, similar to the behaviour seen in ID/IG. On the Salvadon thrust, there is a decrease from 0.572 (20 m below the thrust plane) and 0.568 (10 m above the thrust) to 0.460 on the thrust plane. Here, the pattern resembles that of the ID/IG change for this thrust, with a more gradual change approaching the thrust from the footwall than in the hanging wall. On the Tenneverge thrust plane the R2 value is 0.486 compared to 0.503 at a distance 0.5 m above and 0.551 at 5 m below (note the different distances are due to the extent of the shear zone bounding the thrust, which is thicker in the footwall). In the case of the Finive thrust, R2 is 0.512 on the thrust plane, whilst above and below it, R2 sits at 0.580 and 0.555, respectively. Finally, there is also a decrease in R2 moving from 10 m below the Emaney shear zone (0.560 and 0.564) to within it (0.549 and 0.529). The average error for each sample was ±0.078, so only the Salvadon thrust samples exhibit a change with a greater magnitude than this, but importantly (at least within the immediate vicinity of the strained samples) the direction of change is consistent.

Only FWHM[d] is shown in Fig. 5d, as FWHM[g] varies significantly and shows no discernible trend in our data (the reader is referred to https://zenodo.org/record/4771951#.Yt5t33bMI2w). There appears to only be a small shift, if any, on thrust planes and in the shear zone. There is, however, a slight general decrease in FWHM[d] with depth towards the Morcles thrust (Fig. 5c) and therefore a reverse correlation with ID/IG, as illustrated in Fig. 7.

At low thermal maturities, increasing temperatures cause a rise in ID/IG (Fig. 4a; Dietrich and Casey, 1989; Muirhead et al., 2012; Sauerer et al., 2017). In the Haut Giffre, we estimate peak burial temperatures to be 150–250 ∘C based on a standard geothermal gradient of 25 ∘C km-1 and an estimated burial depth of 6 km at the top of the exposed stratigraphic pile (Pfiffner, 1993; Kirschner et al., 1999; Austin et al., 2008). Therefore, it is reasonable to suggest that the increase in the ID/IG ratio towards the basal Morcles thrust is associated with increasing peak temperature and hence maximum burial. This fits with the observations, for example, of Schito et al. (2017), wherein a similar trend of increasing ID/IG with depth is seen through some 4 km of core containing siliciclastics from the lower Congo Basin, Angola, at temperatures up to 170 ∘C.

There is a significant drop in ID/IG on thrust planes and in shear zones, with as much as a 40 % decrease in the ratio values on the Salvadon thrust plane compared to the surrounding stratigraphy (0.417 to 0.279, a difference of 0.168) and 30 % difference between samples in the Emaney shear zone compared to those adjacent (0.888 to 0.624, a difference of 0.264). There are several possible causes of this drop.

A lower peak temperature on the fault plane than the surrounding rock would reduce the ID/IG value on the fault plane. However, there is no plausible mechanism to explain how this would occur.

Raman band separation (RBS) is reported to increase with increasing temperature (Fig. 4b; Zhou et al., 2014; Bonoldi et al., 2016; Sauerer et al., 2017) and should therefore increase with depth towards the basal thrust in our study. Pressure also affects peak positions (Ross and Bustin, 1990; Bustin, 1995; Huang et al., 2010). However, the trend is weak in our RBS data. If frictional heating on fault planes were the primary control on changes in RBS, and we assume an approximate instantaneous slip magnitude of ∼1 m, then it would be expected that temperatures could rise by >400 ∘C (Savage et al., 2014). This should be enough to produce a shift in RBS which is greater than the general variation we see in our samples. However, Nakamura et al. (2019) report that in addition to temperature, RBS is sensitive to lithology and the effects of fluids, which may explain the variable results we see in this study.

D-peak width (FWHM[d]) exhibits a reverse trend as that of ID/IG, decreasing slightly with depth towards the basal thrust. This supports experiments by Zeng and Wu (2007), who observed a decrease in FWHM[d] with increasing temperature, although their experiments were on samples at 300 ∘C and above. Zhou et al. (2014) also observed a decrease in FWHM[d] with increasing ID/IG in solid bitumen, similar to the trend seen in this study. Studies on coal approaching a magmatic contact by Chen et al. (2017) indicate that both FWHM[d] and FWHM[g] decrease with increasing temperature, but in this study we only observe a decrease in FWHM[d], whilst FWHM[g] changes very little. This is also observed in the case of other starting materials including kerogen (Henry et al., 2019). However, since kerogen (and organic carbon in general) can take on such a wide variety of forms, it is easily possible that a trend observed in one case may not be strong in another.

Unlike ID/IG, there is little change in FWHM[d] across fault planes or shear zones, which suggests that the two parameters are not directly related. For samples affected by polishing during sample preparation, it has been noted that Raman spectral peak widths are less influenced than peak intensities (Ammar et al., 2011; Hu et al., 2015), suggesting that FWHM[d] does not change significantly due to shearing. In light of our results, it may be possible to extend this suggestion to shearing on fault planes and in shear zones. This is supported by the results of fine-scale transects across strained localities (Fig. 6), wherein the error and general variation in FWHM[d] seems to outweigh any significant shift on fault planes or in shear zones.

R2 (the area ratio) shows a trend similar to that of ID/IG, only not as pronounced. Note that the range in R2 values is lower than that of ID/IG due to the normalised denominator used to calculate R2 (Eq. 1), so a trend weaker than ID/IG is expected. There is a drop in R2 on thrust planes and within the Emaney shear zone; this drop is 6 %–20 % (a difference of 0.035; see Fig. 5c) compared to 15 %–40 % for ID/IG. However, the percent change in R2 on fault planes is comparable to the total change in R2 between the upper and lower sections of the stratigraphy (though this is subject to a high degree of variation).

Since peak area is a product of peak intensity and peak width, it follows that R2 is dependent on ID/IG and FWHM[d]. These two parameters have opposing trends with depth through the sequence, resulting in a general dampening of any R2 trend. However, in strained samples, ID/IG tends to drop, whilst FWHM[d] remains unchanged. This means that R2 also drops and makes this parameter more sensitive to strain-related spectral changes than to burial trends.

There are numerous geothermometric equations which use Raman spectral parameters to calculate predicted maximum temperatures. These have been developed for different geological settings, each applicable to a particular carbon type, predicted temperature range, and methodology (such as laser wavelength and deconvolution method). Such examples include but are not limited to Lahfid et al. (2010), Kouketsu et al. (2014), Wilkins et al. (2018), Schito and Corrado (2018), and Muirhead et al. (2019). However, our work demonstrates that strain has an effect on certain Raman spectral parameters, with some more affected than others. This could have significant implications for the results of Raman geothermometers. Here we consider one geothermometric equation – that of Schito and Corrado (2018) – as an example in order to demonstrate how strained rocks might impact results.

This equation uses (in order of decreasing significance) the following Raman spectral parameters: ID/IG (intensity ratio), RBS (Raman band separation), FWHM[d] (D-peak width), FWHM[g] (G-peak width), A[d] (D-peak area), and A[g] (G-peak area) to give a vitrinite reflectance value (%Ro equivalent) of 0.3 to 1.0. Ro=-3.0211+0.33633(ID/IG)+0.01251(RBS)+0.0024823(FWHMd)-0.000000376(Ad)-0.0033158(FWHMg)+0.000000595(Ag) The result is then converted to an approximate temperature value using the equation proposed by Barker and Pawlewiscz (1986). T1=ln⁡%Roeq+1.680.0124 Applying this to our results gives a calculated temperature range across all the samples of 79 to 104 ∘C, with a 10 ∘C increase in local temperature on the Salvadon thrust plane and a 5–10 ∘C decrease in the Emaney shear zone; otherwise, little change is noted in the strained samples.

An actual temperature increase due to frictional heating would explain any increase in temperature on a fault plane but may not account for observations in a ductile shear zone. Further to this, frictional heating could easily raise temperatures to values outside the calibration range of the Schito and Corrado (2018) equation, giving erroneous results using this equation and necessitating the use of a different geothermometer in these samples. However, a transient temperature rise during frictional heating on a fault plane may be too short-lived to promote spectral changes (Bustin, 1983; Fulton and Harris, 2012; Kitamura et al., 2012; Furuichi et al., 2015). Inconsistencies in the magnitude of temperature change can potentially be explained by differing thicknesses of active slip (Raboniwitz et al., 2020) or a low slip magnitude in a single event (Polissar et al., 2011; Savage et al., 2014, 2018; Raboniwitz et al., 2020). However, these do not explain the distinct drop in apparent temperature on the Salvadon thrust plane or the elevated temperature values within the Emaney shear zone, where frictional heating should not play a role.

The most significant term in the equation is ID/IG, and our data show that ID/IG is strongly affected by strain-related spectral changes. It therefore follows that the equation should be sensitive to strain, but the fact that not all strained samples produce calculated temperature shifts of the same direction or magnitude suggests that the process is more complex than simply strain or temperature having an effect. Regardless of cause, however, if a strained sample can produce a difference in calculated temperature of 10 ∘C in a stratigraphic sequence with an overall temperature range of only 25 ∘C, then context is important when estimating temperatures using this method. For example, if using these temperature data to reconstruct a burial history, then a strained sample might be “out” by over a kilometre, or it might give the correct value. It is therefore important that more work is done to calibrate Raman geothermometers in rocks which have undergone strain in natural environments.