Concentrations of major (Si) and minor (Al and Ti) elements in quartz grains were all determined by CAMECA SXFive electron probe micro-analyzer at the Institute of Geology and Geophysics, Chinese Academy of Sciences (IGGCAS) in Beijing, China. Accelerating voltage was 15 kV and beam currents were 30 and 300 nA for major and minor elements. The peak counting times were 20s for Si and 120s for Al and Ti. Rhodonite, albite and rutile were used as standards for calibration. Standardization was performed with a beam current of 20 nA and with a peak counting time of 20s. Data were obtained by the X-PHI matrix correction (Merlet, 1994). Oxygen was calculated by cation stoichiometry and included in the matrix correction. The following detection limits were based on a 3σ estimate of the measured background variance: ∼210 ppm for Si, ∼12 ppm for Al and ∼13 ppm for Ti.

A large spot size of 50 μm was used for all analyses in order to involve the composition of substantial rutile needles in the measurement, because TiO₂ in these rutile needles might have once occupied by pre-existing quartz that is before rutile exsolution (See Section Discussion). The natural smoky quartz crystal was used as a secondary standard, which has reproducible Ti concentration (57 ± 4 ppm; Audétat et al., 2015). Our averaged Ti concentration (N = 55) of the quartz standard is 58 ± 3 ppm, which is consistent with the recommended value within error. Our analyzed quartz grains in the matrix are all far away from other major Ti-bearing minerals, such as rutile and biotite, to avoid secondary fluorescence effect.

The Cathodoluminescence (CL) images of quartz were observed using a Nova NanoSEM 450 (FSEM) scanning electron microscope at the IGGCAS. The Nova NanoSEM 450 was equipped with a Gatan Instrument “MonoCL14” detector. An accelerating voltage of 10 kV and a beam current of 1.1 nA were used during the analysis, and the working distance was 27–29 mm.

The titanium concentration in quartz was first proposed as a geological thermometer in 1987 (Ostapenko et al., 1987). The thermodynamic principle of the TitaniQ thermometer depends on the titanium equilibrium between quartz and rutile, which can be expressed in an exchange reaction: TiO₂ ^Qz = TiO₂ ^Rt. The calibration experiments suggest that titanium solubility in quartz depends on the pressure and temperature conditions in the case that quartz crystallizes coexisting with rutile, and titanium contents of quartz (in ppm by weight) would increase exponentially with T. Similar to other trace-element thermometers, TitaniQ thermometer has several calibrations (Wark and Watson, 2006; Thomas et al., 2010; Huang and Audétat, 2012; Thomas et al., 2015). Within these models, the experiments were set to high pressure (5–20 kbar) and high temperature (700–940°C) conditions in the Thomas et al. (2010) model, which is the most suitable for the metamorphic conditions of our studied samples. Hence, we used the model calibrated by Thomas et al. (Thomas et al., 2010) in this study. And the pressure we have chosen for all calculations is 8 kbar based on couples of previous studies in the eastern Khondalite Belt (Figure 2; e.g., Santosh et al., 2007a; Santosh et al., 2012; Jiao et al., 2013a; Yang et al., 2014; Li and Wei, 2016; Li and Wei, 2018).

When applying the TitaniQ thermometer, the activity of TiO₂ should be accurately estimated. The value is generally 1 if rutile exists in the studied rocks, which means that the system is TiO₂ saturated (Ghent and Stout, 1984). There are also methods for calculating TiO₂ activity if the rocks do not contain rutile, which are mainly based on different solution models and compositions of Fe-Ti oxides (Hayden and Watson, 2007; Hildreth and Wilson, 2007; Kularatne and Audétat, 2014).

All samples in this study contain rutile, although the grain size and volume percentage vary. For example, five samples from south of Heling’er (samples 20HL49, 20HL59, 20HL64, 20HL73, and 20HL89) contain few tiny rutile. A study suggested that fluid remains saturated with respect to rutile once the fluid is rutile-saturated, and subsequent rutile growth is not sufficient to drive the liquid to rutile-undersaturation (Acosta et al., 2020). Although the absence of rutile does not necessarily imply quartz growth from a rutile-undersaturated liquid, researcher interpreted it as a liquid that was near saturation but below the critical supersaturation required to nucleate new rutile crystals (Acosta et al., 2020). Because of the aforementioned reasons, we used a TiO₂ activity of 1 for all samples.

TitaniQ temperature mapping was conducted using ArcGIS with the inverse distance weighted interpolation method.

A total of 463 spots were analyzed in 44 samples. Analyses with unacceptable total SiO₂ content (>105 or <95) were discarded. We finally obtained 417 valid data points (Supplementary Tables S1, S2), among which 182 were quartz inclusions within garnet from 42 samples (Supplementary Table S1), and 235 were of matrix-type quartz from 43 samples (Supplementary Table S2).

Within quartz inclusions of garnet, Ti concentrations range from 40 to 550 ppm. We calculated the average Ti concentrations for each sample, which range from 58 to 420 ppm, and four data points of extremely high Ti concentrations far from the others within several samples (samples 20HL34, 20HL59, 20HL64, and 09TGS01) were not included and regarded as outlier data (Figure 3A). Temperatures calculated by the TitaniQ thermometer for these quartz average Ti concentrations range from 654°C (sample 20HL89) to 891°C (sample 18LC01), all below UHT conditions (Figure 4A). The maximum Ti concentrations of each sample range from 65 to 550 ppm, the corresponding TitaniQ temperatures range from 665 to 934°C, and the highest temperatures are from samples in the Liangcheng and Zhuozi areas (Figure 4B).

Within matrix-type quartz, Ti concentrations range from 30 to 1,000 ppm. The calculated average Ti concentrations for each sample, range from 100 to 528 ppm, and six data points of extremely high Ti concentrations far from the others within several samples (i.e., samples 18ZZ01, 20HL104, 18TG01, 18TG04, and 09XH32) were not included, and were regarded as outlier data (Figure 4C). Temperatures calculated by the TitaniQ thermometer for these quartz average Ti concentrations range from 685 to 917°C (Figure 4D), and some of them have recorded UHT metamorphic conditions. The maximum Ti concentrations of each sample range from 156 to 680 ppm (Figure 4C), and the corresponding TitaniQ temperatures are 760–970°C, and the highest temperatures are also from samples in the Liangcheng and Zhuozi areas (Figure 4D).

We compared the difference in average Ti concentrations between the matrix-type quartz and quartz inclusions for each sample, without considering the outlier data (Figure 5A). The matrix-type quartz contains Ti concentrations <180 ppm higher than that of quartz inclusions, and only five samples show an inverse trend (Figure 5A). Therefore, the temperature calculated for the matrix-type quartz is generally <120°C higher than that of quartz inclusions (Figure 5B).

We also compared the difference in maximum Ti concentrations between the matrix-type quartz and quartz inclusions for each sample, without considering the outlier data (Figure 5C). The matrix-type quartz also contains Ti concentrations <300 ppm higher than that of quartz inclusions. The temperature calculated for matrix-type quartz is generally <160°C higher than that of quartz inclusions (Figure 5D).

The CL images revealed types of quartz internal structures not observed in the photomicrographs. We divided quartz according to internal structure into four types: 1) Non-structure with homogeneous luminescence, 2) linear structure with low luminescence, 3) sealed fractures, and 4) reticular structures.

The non-structure with homogeneous luminescence is the simplest CL image in our samples and displays uniform luminescence without a special structure (Figure 6). It is more common in quartz inclusions than in matrix-type quartz. Quartz inclusions with non-structure contain low Ti concentrations, generally 100–200 ppm, and TitaniQ temperatures are 700–800°C (Figures 6A,B). Matrix-type quartz with non-structure has different Ti concentrations, which could be as low as 225 ppm with a TitaniQ temperature of 805°C (Figure 6D) and as high as 490 ppm with a TitaniQ temperature of 917°C (Figure 6F).

A linear structure with low luminescence occurs in both quartz inclusions and matrix-type quartz (Figure 7). These dark linearities within quartz generally correspond to the rutile needles shown in the photomicrographs. Rutile is rare and barely visible under photomicrographs but distinct in CL images. The occurrence of rutile needles is more common in quartz inclusions than in matrix-type quartz but generally occurs in the cores of the grains. Quartz Ti concentrations within rutile needle-rich domains are commonly high (>250 ppm), and the corresponding temperatures are >820°C (Figure 7).

Healed fractures characterized by healed microcracks and dark patches occur in both quartz inclusions and matrix-type quartz (Figure 8). The dark patches grew around and extended along the cracks. Healed fractures generally occur simultaneously with rutile needles (i.e., linear structures with low luminescence). Thus, quartz with healed fractures also has high Ti concentrations of >200 ppm and corresponding temperatures of >790°C (Figures 8B,D).

Reticular structures only occur in a few quartz grains (Figure 9). An irregular pattern of differential CL forms an enigmatic reticular microstructure. It is likely that a large single quartz crystal with bright luminescence is separated by dark stringers into numerous sub-crystals. However, this structure occurs differently in quartz inclusions and matrix-type quartz; in the former, the dark stringers are relatively blurry and disappear in the rim of the grain (Figure 9B). By contrast, the stringers are more fully developed in the matrix-type quartz (Figure 9D). Quartz grains with this internal structure generally contain low Ti concentrations of <250 ppm and corresponding temperatures of <820°C (Figure 9).

Studies have constrained the metamorphic age and P–T path of aluminous gneisses/granulites in the eastern Khondalite Belt. Zircon U-Pb dating and detailed petrological investigations in combination with phase equilibria modeling have constrained the timing of UHT metamorphism at 1.94–1.80 Ga, following the prior HP metamorphism (Santosh et al., 2007b; Jiao et al., 2013b; Li and Wei, 2016). The P–T paths involved post-peak cooling and decompression stages, although the prograde stages were variable (Figure 2; Santosh et al., 2007a; Santosh et al., 2012; Yang et al., 2014; Li and Wei, 2016; Li and Wei, 2018). We interpret the quartz formation based on the aforementioned metamorphic evolution history.

Quartz inclusions in garnet generally contain less Ti concentrations than matrix-type quartz, and the TitaniQ temperature is <160°C lower (Figure 5). This result suggests that the quartz inclusions probably grew during prograde metamorphism. Garnet may protect quartz inclusions from later resetting. However, quartz inclusions have a wide range of Ti concentrations, which may represent quartz inclusions that formed during different periods of prograde metamorphism, or are reset by later retrogression. Quartz inclusions with homogeneous luminescence CL images have low Ti concentrations, which may form at relatively low temperatures during the early prograde period, and others with rutile needles have high Ti concentrations and may form at relatively higher temperature conditions of the prograde metamorphism.

Matrix-type quartz was dominant in the studied rocks. The grains with homogeneous luminescence CL images and rutile needle structures that contain high Ti concentrations probably formed during the near-peak or early cooling stage. It was suggested that the high-temperature quartz that crystallized from the residual melt during early cooling is characterized by non-structure homogeneous luminescence CL (Storm and Spear, 2009).

The occurrence of rutile needles has been observed in Ti-rich quartz interiors of medium- to high-grade metamorphic rocks, such as quartz in migmatites and UHT metamorphic rocks (Cherniak et al., 2007; Sato and Santosh, 2007; Storm and Spear, 2009). Two mechanisms have been proposed to explain these features: 1) The entrapment of preexisting acicular rutile by growing quartz and 2) the exsolution of high-Ti quartz into low-Ti quartz and acicular rutile. In our samples, rutile needles are generally restricted to grain cores and are consistent with quartz crystallographic orientations, which strongly proves the second mechanism (Storm and Spear, 2009; Thomas and Nachlas, 2020). The dark “spots” in the CL images observed by Cherniak et al. (2007) were also interpreted as the feature of rutile exsolution during the post-peak cooling.

For both quartz inclusions and matrix-type quartz, especially the latter without garnet protection, some grains contain low Ti concentrations, which suggest the occurrence of later resetting. Healed fracture structures exist in both quartz inclusions and matrix-type quartz, which are similar to filled fractures in plutonic quartz and fluid-controlled quartz recovery in granulites (Seyedolali et al., 1997; Van den Kerkhof et al., 2004). Quartz first formed with non-healed fractures owing to the stresses imposed on the grains and decrepitation of fluid inclusions caused by the pressure difference between fluid inclusions and host quartz, probably by tectonic settings, such as uplift. Subsequently, the fractures were filled with secondary quartz or quartz nuclei by precipitation of non-luminescing SiO₂. Quartz with such structures commonly contains high Ti concentrations and yields high TitaniQ temperatures. Thus, we propose that quartz with such fractures within our samples grew or recrystallized during post-peak nearly isothermal decompression.

The occurrence of reticular structure in quartz is less common in our samples, and quartz with such structures contains relatively low Ti concentrations and accordingly records low TitaniQ temperatures. Many studies have verified a positive correlation between Ti concentrations in quartz and CL intensity (Sprunt et al., 1978; Müller et al., 2002; Spear and Wark, 2009). Thus, the Ti concentrations are heterogeneously distributed. Reticular zoning might indicate the pathway of fluid infiltration.

In summary, we posit that matrix-type quartz with extensive rutile exsolution most probably grew during the near-peak to the early cooling stage and can better constrain the near-peak conditions of the studied samples than inclusion-type quartz. Therefore, the variation in the maximum TitaniQ temperatures of matrix-type quartz in these samples can better show the thermal regime of the studied region.

Another major question can be posed to interpret the meaning of TitaniQ temperature: Can the matrix-type quartz that formed during near-peak or early cooling preserve the original Ti concentrations? Mechanisms for quartz Ti re-equilibrium are diffusion (e.g., both lattice and grain boundary diffusion) and recrystallization during deformation and/or retrogression.

Experimental studies have shown that Ti diffusion is slow in quartz (Cherniak et al., 2007; Jollands et al., 2020). According to Cherniak et al. (2007), the diffusion distance of Ti-in-quartz is 0.05, 2, and 5 μm at 0.001, 1.0, and 10 Ma at 500°C. However, the diffusion distances increased to 16 μm, 500 μm, and 1,600 μm over the same time range at 800°C. Nevertheless, experiments under anhydrous conditions suggest that grain boundary diffusion is 3–4 orders of magnitude faster than lattice diffusion (Bromiley and Hiscock, 2016). Therefore, the grain boundary acts as an efficient conduit for Ti equilibrium, especially in fine-grained rocks. Studies have shown an increasing trend of Ti concentrations from the core toward the rim in quartz inclusions of garnet, interpreted as an exchange with the host garnet (Spear et al., 2012), but we did not recognize such features in our samples.

Many studies have suggested that dynamic recrystallization is an efficient mechanism for Ti re-equilibrium, especially fast grain boundary migration, which is crucial for Ti re-equilibrium at high temperatures (Stipp et al., 2002; Grujic et al., 2011). Generally, recrystallized quartz, which has well-developed facets impinging on the preliminary quartz, results in cuspate, serrated, and irregular grain boundaries of preliminary quartz (Thomas and Nachlas, 2020). Typical serrated boundaries from matrix-type quartz shown in Figures 5C,D suggest incomplete re-equilibration and would contain low Ti concentrations; thus, we did not select domains near the serrated boundaries for analysis.

Despite the aforementioned factors, the highest TitaniQ temperatures (ca. 960–970°C) are broadly consistent with the reported temperatures calculated by other methods in the studied region (Santosh et al., 2007a; Jiao et al., 2013a; Jiao et al., 2013b; Li and Wei, 2016; Li and Wei, 2018), for example, 966–1,096°C by a two-feldspar thermometer in Xuwujia, 850–1,000°C by a Zr-in-rutile thermometer in Tuiguishan, 940–1,030°C by Ti-in-zircon and a ternary feldspar thermometer in Nantianmen and Xiaonangou, ∼975°C based on phase equilibria modeling in Xumayao, and 930–1,050°C by pseudosection in Hongsigou (Jiao et al., 2011; Jiao and Guo, 2011; Liu et al., 2012; Zhang et al., 2012; Yang et al., 2014). This result confirms the validity of the TitaniQ thermometer in high-grade rocks, and TitaniQ temperatures may represent the minimum estimates of the peak temperatures.

Other possible influences on our calculated temperatures were 1) the pressure estimation, 2) the location of selected analytical spots, and 3) sample bias. We set a pressure of 8 kbar for all calculations in this study. Pressure at 1 kbar higher would result in a 30°C higher calculated temperature. The analyzed domains with abundant rutile needles contain higher Ti concentrations than the domains without rutile needles; therefore, if any, we generally selected domains with the most abundant rutile needles for analysis. However, quartz and rutile have different crystal structure, which result in different characteristic X-ray efficiencies. This effect might influence the measured Ti concentrations. The EMP analytical spots are actually in two-dimension, which may underestimate the real volume percentage of rutile needles in quartz. Finally, we must acknowledge that in some samples we probably missed the quartz grain that contains the highest Ti concentrations, and in this case, the maximum TitaniQ temperatures for such samples would be underestimated.

In Figure 10A, the mapping is based on the temperature calculated from the maximum Ti concentrations of matrix-type quartz in each sample; for comparison, that in Figure 10B is based on the temperature calculated from the average Ti concentrations of matrix-type quartz in each sample. Sample localities are marked in the picture to show the distribution of the samples and the effect of sample bias.

Both temperature mapping results show that the large-scale and extremely hot regions are surrounded by Liangcheng, Heling’er, and Zhuozi. The maximum TitaniQ temperatures were >900°C, suggesting the occurrence of UHT metamorphism. This phenomenon is consistent with those of several UHT metamorphic localities reported in this area (e.g., Liu et al., 2012; Jiao et al., 2013a; Jiao et al., 2013b). This hottest area was also consistent with the region where extensive charnockite occurred (Figure 1). Our results also show that Tuguiwula area, where typical UHT granulites with diagnostic Spr + Qz and Opx + Sil + Qz mineral assemblages have been reported, are hot, but the scale of the hot region is limited and surrounded by a relatively cold region. It might imply that the surrounding area is probably located at a higher crustal level. Alternatively, it is noticed that we have not collected samples from Zhuozi to Jining and further to Tuguiwula because of the thick sediment cover. The variation of temperatures in this region is probably caused by sample bias. Therefore, it is also possible that the whole area of Heling’er-Liangcheng-Tuguiwula-Jining-Zhuozi have experienced UHT metamorphism. It requires confirmation in future research.

In contrast, the Youyu and Xinghe areas are not as hot as the Liangcheng-Heling’er-Zhuozi area, although the major mineral assemblages are mostly the same. The maximum TitaniQ temperature is below 900°C, implying that these two areas probably do not experience UHT metamorphism. Alternatively, the analyzed quartz probably crystallized during the later stage of post-peak cooling. Quartz from these areas generally shows a non-structure with homogeneous luminescence structures, and rutile needles cannot be observed.

The thermal regime of the eastern Khondalite Belt, established by the TitaniQ thermometer in this case, is consistent with the results of another independent study using Zr-in-rutile thermometry mapping (Qi et al., unpublished). This particular case study further confirms the validity of the TitaniQ thermometer in recognizing UHT metamorphism and constraining its thermal conditions and spatial scales.

Several models have been proposed to interpret the formation of UHT metamorphism in the eastern Khondalite Belt, including radiogenic heat production (Jiao et al., 2013a), post-collisional lithosphere extension due to asthenospheric upwelling (Zhao, 2009; Li and Wei, 2018; Jiao et al., 2021), ridge subduction (Santosh and Kusky, 2009; Peng et al., 2011; Santosh et al., 2012), mantle plume (Santosh et al., 2008), and emplacement of mantle-derived mafic magma (Huang et al., 2019). In addition, a back-arc setting has also been proposed to explain UHT metamorphism in other regions (Brown, 2006; Brown and Johnson, 2018). Our study certifies the regional distribution of UHT metamorphism in the eastern Khondalite Belt, and the large-scale hottest region is the Liangcheng-Heling’er-Zhuozi area, which is consistent with the occurrence of extensive charnockite. This result implies that the formation of UHT metamorphism is closely related to charnockite. This extremely hot metamorphism and magmatism might be the coupling effect of the same mechanism, which probably formed in the root of a Paleoproterozoic large hot orogeny. However, we could not distinguish these UHT formation models in this study, and the potential model should be able to explain the spatially close relationship between UHT metamorphism and charnockite.