Fracture, Dissolution, and Cementation Events in Ordovician Carbonate Reservoirs, Tarim Basin, NW China

Ordovician carbonate rocks of the Yijianfang Formation in the Tabei Uplift, Tarim Basin, contain deeply buried (>6000m), highly productive oil and gas reservoirs associated with large cavities (>10m). Previous workers inferred that large cavities are paleocaves (paleokarst) formed near the surface and subsequently buried. Alternately, caves may have formed by dissolution at depth along faults. Using 227 samples from 16 cores, we document textures and cement compositions bearing on cavity histories with petrographic, high-resolution scanning electron microscopy (SEM), isotopic, and fluid inclusion microthermometric observations. Results show that dissolution occurred at depth and was caused by (1) acidic fluids derived from Middle-Late Silurian and/or Devonian-Permian hydrocarbon generation and maturation, (2) high-temperature fluids, of which some were associated with Late Permian igneous activity, and (3) Mg-rich fluids that accompanied Jurassic-Cretaceous deformation and the formation of partially open fractures and stylobreccias (fault breccias). The relative paragenetic sequence of the structure-related diagenesis suggests seven stages of fracturing, dissolution, and cementation. Mottle fabrics in the Yijianfang Formation contain argillaceous carbonate-rich silt and are bioturbation features formed within the marine environment. Those mottled fabrics differ from clearly karstic features in the overlying Lianglitage Formation, which formed by near-surface dissolution and subsequent infilling of cavities by allochthonous sediment. Mottle fabrics are crosscut by compacted fractures filled with phreatic-vadose marine cements and followed by subsequent generations of cement-filled fractures and vugs indicating that some fractures and vugs became cement filled prior to later dissolution events. Calcite cements in fractures and vugs show progressively depleted values of δO documenting cement precipitation within the shallow (~220m), intermediate (~625m), and deep (~2000m) diagenetic environments. Deep (mesogenetic) dissolution associated with fractures is therefore the principal source of the high porosity-permeability in the reservoir, consistent with other pieces of evidence for cavities localized near faults.

The origin of the deep pore networks in carbonate rocks of the Yijianfang Formation in the central Tarim Basin is debated [4,5,[21][22][23]. Karstic features are present in Middle Ordovician carbonates within the nearby Lunnan and Tahe oilfields [2,[24][25][26][27]. Here, cavities have been linked to a hiatus of as much as 120 m.y. prior to deposition of Silurian siliciclastic rocks [2,25,28]. In the paleostructurally lower Halahatang area, however, a well-defined unconformity is absent [29,30]. The main evidence for near-surface karst is the so-called mottled fabric, a texture very common within the Yijianfang Formation. This feature has been interpreted to be nearsurface paleokarst breccia or sediment-filled vugs/cavities [31][32][33][34] but has recently been brought into question [22,35]. Additionally, regional structural interpretation and fault mapping in the Halahatang area reveal a network of conjugate strike-slip fault systems that developed damage zones which has a good agreement with the high well production [36,37]. Wu et al. [37] hint that there is much more dissolution developed along the fault damage zones, suggesting that fluid pathways formed in those areas facilitate the later dissolution in the fault zones. The fracture-cave reservoirs are quite different from the high-matrix reservoirs and meteoric karst reservoirs [9,38] but are strongly heterogeneous reservoirs that are close to mesogenetic dissolution or hydrothermal karstification along fault zones [14,21,39].
Dissolution processes partially or completely eliminate original microstratigraphic and textural records. Precipitated cements on the other hand, especially precorrosion and postdissolution minerals filling cavities and fractures, may record information on the elemental and isotopic chemistry of dissolving fluids involved in structural diagenetic events [40]. Cements provide information on the physical and chemical conditions of the reservoir during different stages of diagenesis, as well as the composition of fluids, and allow obtaining information that would otherwise be unavailable from empty voids. Sequences of crosscutting cements and fluid inclusions trapped within them can provide insights into the origin, types, and pathways of fluids as well as depths of dissolution and cementation events over time [41][42][43][44]. We used petrography, scanning electron microscopy (SEM), C, O, and Sr isotopes, and fluid inclusion microthermometry of core samples from 16 wells in the Halahatang oilfield to establish the paragenetic sequence, temperature conditions, and origins of the fluids and events that led to the formation of cryptic mottle textures as well as dissolution vugs and fractures within these deeply buried (5500 to 7500 m) Ordovician carbonate reservoirs.
Here, we show that deep (mesogenetic) dissolution associated with fractures is the principal source of currently open cavities in the reservoir. Mottle fabrics are not the result of penetrative karstic processes but represent burrows that were later overprinted by mesogenetic dissolution. We found no evidence for pervasive near-surface dissolution. We document seven stages of overprinted fracturing and dissolution, with evidence for precipitation of cements from the marine to deep diagenetic environments. Stable isotopic analyses tied to geothermal gradients indicate most of the dissolution that predated each additional fracture-dissolution event occurred within the mesogenetic environment (220-2000 m). Dissolution was caused in various stages by a range of fluids, including meteoric and organic acids and hightemperature and Mg-rich fluids.

Geological Setting
The Tarim Basin is the largest cratonic area in western China covering an area of nearly 600,000 km 2 ( Figure 1). The basin is bounded by the Tian Shan Mountains to the northwest, Kuluketage region to the northeast, Kunlun Mountains to the south and southwest, and Altyn Fault Zone to the southeast ( Figure 1). Its Precambrian crystalline basement represents a fragment of the Rodinia supercontinent that has undergone long-term geological evolution from the Sinian (latest Neoproterozoic) to the Neogene [45,46]. Except for flood basalts of the Tarim Igneous Province at~290 Ma [47] (Figure 1), the evolution of the Tarim Basin is characterized by almost continuous sedimentation since the Neoproterozoic. Long-term polycyclic tectonic evolution of the Tarim Basin resulted in the formation of three paleohighs (Tabei, Central Tazhong, and Southeast Tadong) and four depressions (Kuqa, Manjiaer, Southwest Tangguzibasi, and Southeast Tadong) [45,[48][49][50][51][52][53][54] (Figure 1). During the Late Sinian-Early Ordovician period, the Tarim Basin was under extension resulting in a series of normal faults and horstgraben structures [45,55,56]. The passive continental margin of the paleo-Altun and paleo-Kunlun ocean in the south of Tarim was converted to an active continental margin in the Middle Ordovician, experiencing strong compression from south to north and forming a large-scale Middle Caledonian thrust belt [45,55,56]. Ordovician carbonate successions were exposed and partially eroded due to uplift related to the Caledonian and Hercynian orogenies (Ordovician-Permian), generating a series of unconformities and/or disconformities (Figure 2), and pervasive karst events directly related to hiatuses [25,57]. Following closure of the paleo-Altun, paleo-Kunlun, and paleo-Tianshan oceans in the Late Hercynian, Tarim turned into intracontinental basins in the Indosinian and experienced continuous deposition and burial. The Indonesian-Yanshanian (Triassic-Cretaceous) was characterized by intensive faulting in hillside structural belts, whereas in the Himalayan (Cenozoic), faulting concentrated in the foreland basin, closely related to intensive uplift, thrusting, strike-slip thrust, or thrust strike-slip of surrounding orogenic belts [51,58]. The basin has experienced a period of rapid burial since 5 Ma [51].
The Tabei Uplift in the northern Tarim Basin ( Figure 1) comprises a series of Sinian-Devonian marine, Carboniferous-Permian marine-terrigenous, and Triassic-Quaternary terrestrial sedimentary rocks [25,59] (Figure 2). Hydrocarbon reservoirs are mainly distributed in Ordovician carbonate rocks. During this time, platform margins extended for more than 3000 km, migrating intermittently and producing complex sequence architectures due to the interaction of basin tectonism, topography development, and sea level change [59]. The Halahatang oilfield (Figure 3(a)) is located in the Halahatang Depression, and it is bordered by the Luntai Uplift to the north, the Shuntuoguole Low Uplift to the south, the Lunnan Low Uplift to the east, and the Yingmaili Low Uplift to the west [60][61][62]. Other important oilfields in the Tabei Uplift include Tahe and Lunnan oilfields to the east [57,63]. In the Halahatang oilfield, the Ordovician system comprises the Middle-Lower Ordovician Yingshan

Materials and Methods
We collected 227 samples of the Yingshan, Yijianfang, Tumuxiuke, and Lianglitage Formations from 16 cores across the study area (Figure 3(b)). Of these, 262 billets were selected for petrographic analysis by standard optical microscopy, with emphasis on the Yijianfang Formation. Thin sections were half stained with Alizarin Red S and potassium ferricyanide in order to facilitate the identification of ferroan and nonferroan calcite, as well as dolomite [64]. Forty thin sections were examined under cathodoluminescence microscopy (optical-CL) using a Reliotron III Cathodoluminescence attachment operated at 10-18 kV gun potential and 0.5-0.6 V beam current. Textural characteristics of limestones and diagenetic cements were investigated using a Zeiss Sigma high-vacuum field emission scanning electron microscope (HV FE-SEM). Carbon-coated samples (~15 nm) were imaged under SEM-CL with a Gatan MonoCL4 detector operated at 5 kV and 120 μm aperture. To investigate the morphology and distribution of nanometer-scale pores, we followed the sample preparation protocol described by Loucks et al. [65]. We prepared 10 × 10 mm cubes in 8 samples using the standard X-sectional argon ion milling methods [66]. Samples were then coated with 5 nm of iridium and analyzed using a FEI Nova Nano 430 FE-SEM. Operating conditions were 10-15 kV, spot size of 3-5 nm, and aperture of 30 μm.

Geofluids
Re filaments with tantalum fluoride and 0.05 M phosphoric acid and subsequently analyzed on a Thermo Fisher Triton thermal ionization mass spectrometer in a static mode. Intensity of 87 Sr of 8 V ðusing 10 -11 ohm resistorsÞ ± 5% was maintained for 8 blocks of 20 cycles with 8second integration time. The 87 Sr/ 86 Sr ratio was corrected for mass fractionation using 87 Sr/ 86 Sr = 8:375209 and an exponential law.
Fluid inclusion microthermometry of 20 doubly polished thin sections was conducted using a gas flow heating-freezing stage (FLUID, Inc.-adapted, U.S. Geological Survey-type) mounted on an Olympus BX-51 microscope. The stage was calibrated using CO 2 ice melting and H 2 O ice melting and the critical homogenization temperature of synthetic fluid inclusions [69]. Fluid inclusion petrography and microthermometry followed the procedures of Fall and Bodnar [70]. Liquid-vapor homogenization temperatures (T h ) were determined to be ±1.0°C by thermal cycling using temperature steps of 0.5°C [71] (Supplementary Material III).  (Figure 4(a)). These facies are best developed in the northern part of the study area ( Figure 3(b), A through I wells), and albeit less abundant, they are also common in the southern part ( Figure 3(b), J through P wells). These lithofacies indicate deposition in shallow water with high-energy environment within a wave-dominated shelf. The Tumuxiuke Formation is mud-dominated, mainly composed of bioclast wackestones and peloid wackestones (Figure 4(b)), suggesting that the depositional setting was relatively low energy and deeper marine water during a transgressive period. The basal part of the Lianglitage Formation mostly consists of ooid grains and bioclast grainstones (Figure 4(c)), indicating that the basal part might have been deposited as ooid shoals or bioclast shoals. Upwards, the Lianglitage Formation changes

Geofluids
into peloid intraclast packstone and/or bioclast wackestone, and peloid wackestone is also locally present, indicating a low-to moderate-energy depositional environment. The upper intervals contain more fine-grained siliciclastic sedi-ments, indicating the incorporation of terrigenous sediments into the carbonate system. Grainstones, packstones, and wackestones of all formations contain little primary porosity, mostly in the form of intraparticle and intercrystalline pores < 150 μm in diameter (Figure 4(c)). Intraparticle pores mostly occur within skeletal fragments such as echinoderms and bryozoans. In the Lianglitage Formation, intraparticle pores are most abundant and localized in a grainstone interval, where they are completely filled with bitumen ( Figure 4(c)). Interparticle microporosity is negligible in unaltered host rock.

Karstic Features.
Clear-cut karstic breccias that contain external sediment were only observed in the Lianglitage Formation. Karstic breccias include crackle breccia, mosaic breccia, clast-supported chaotic breccia, and matrix-rich clast-supported chaotic breccia (Figure 4(f)). Most breccia clasts have a similar lithology, are light red in color, range from very fine to very coarse pebbles in size, and are subangular to subrounded in shape (Figure 4(g)). The matrix is commonly dark brown to reddish brown and contains silt quartz, clays, and iron oxides in sharp contact with clasts ( Figure 4(h)). Rounded mottle fabrics in the Yijianfang Formation comprise internal, orange to red luminescent, argillaceous carbonate-rich silt ðCs0aÞ < 20 μm in size ( Figure 5(d)). Primary (<3%) and secondary (see below) intergranular micropores and nanopores within mottle fabrics comprise as much as~10% of the volume of the rock. Most pores are less than 20 μm in size, contain up to 3% illite, and are lined or filled with bitumen (Figures 5(e) and 5(f)). In contrast, vertical and horizontal channels are filled with various mixtures of fragmented fossils and dark-brown calcite mudstone matrix, containing intergrown illite/chlorite, quartz, and phosphates      ). Intergranular micropores in channels constitute <10-15% of the volume of the channel and are partially infilled with 5-10% illite, kaolinite, and bitumen ( Figure 5(i)). Ordovician host rocks, including mottle fabrics, are crosscut by several types and generations of partially to fully cemented fractures, stylolites, and dissolution vugs ( Table 1). The classification below was established for the Yijianfang Formation, but on the basis of textural and geochemical similarities, we extend this classification to similar features present in the Yingshan, Tumuxiuke, and Lianglitage Formations.

Fractures.
On the basis of crosscutting relationships and the type of cement fill, we define seven groups of openingmode fractures, from oldest (F1) to youngest (F7) ( Table 1). Bed-perpendicular and bed-oblique calcite-filled fractures (F1) with irregular walls and apertures between 0.5 and 1 cm occur mostly in the uppermost part of the Yijianfang and Tumuxiuke Formations ( Figure 6(a)). Calcite cement fill (Cc1) in these fractures consists of equant to bladed subhedral crystals grading rapidly to palisade morphology. Palisade calcite crystals are 300 μm to 1 mm long and 10-50 μm wide and can develop up to 3 cm thickness ( Figure 6 F3 fractures are present within the Yingshan and Yijianfang Formations and to a lesser extent in the Tumuxiuke and Lianglitage Formations. This group includes bedperpendicular and bed-oblique calcite-and bitumen-filled fractures with straight walls and apertures of 100-500 μm (F3a in Figure 6(f)), slightly wider (500 μm-1 cm) fractures of similar characteristics that may preserve partially open porosity (F3b in Figure 6(f)), and planar bitumen-filled microfractures with apertures < 100 μm (F3c). Calcite cements in F3 fractures are zoned, dark red to bright orange in luminescence, bladed to blocky rim cements, and 50 to 200 μm in size, with euhedral terminations (Cc3a) (Figures 6(g)-6(i)). Some F3 fractures preserved partial porosity that was subsequently infilled by bitumen (Figures 6(g)-6(i)).
The fourth group of fractures (F4) is common in the Yijianfang Formation and less abundant in the Tumuxiuke and Lianglitage Formations. This group encompasses planar bed-perpendicular and bed-oblique fractures rang-ing between 0.5 mm and 2 cm in aperture with irregular margins and calcite+bitumen±minor late pyrite cement (F4a in Figures 6(j) and 6(l)) and vertical microfractures that terminate~1 cm away from horizontal stylolites and are subparallel to stylolite teeth (F4b in Figure 6(k)). F4 fractures are filled with blocky calcite crystals, 50 μm to 1 mm in diameter, which show red luminescence (Cc4 in Figures 6(k)-6(l)).
Fractures in the fifth group (F5) are only present in the Yingshan and Yijianfang Formations. F5 fractures are bedperpendicular and/or bed-oblique to bedding and have irregular walls and apertures between 1 mm and >7 cm (Figure 7(a)). F5 fractures are filled with euhedral fluorite, barite, and calcite cement (Cc5), up to 5 cm in size, in variable proportions (Figure 7(b)), and are associated with calcite and celestite host rock replacement in cores L, N, and K (Figure 3(b)). Calcite and celestite mineralization is nonpenetrative and limited to a few millimeters to centimeters away from F5 fractures walls. Pores around celestite crystals are commonly less than 100 μm in size (Figure 7(c)). In most cases, fluorite and barite predate calcite cementation (Figures 7(a) and 7(b)), and although textures are ambiguous, celestite replacement appears to be closely related to fluorite and barite mineralization. In a few examples, barite and fluorite appear to be coeval, whereas in others, fluorite was followed by calcite and finally barite precipitation. Cc5 in F5 fractures is blocky, 200 to 250 μm in size, and shows red to dull red luminescence (Figure 7(b)). Silica mineralization (microcrystalline quartzchert) occurs along some F5 fracture walls (Figure 7(a)). Within silicified areas, pores of various shapes are as much as 400 μm long (Figure 7(d)).
F7 fractures are present in all formations. This group comprises bed-perpendicular and bed-oblique calcite-filled fractures ranging between 50 μm and 2 mm in aperture (F7a in Figures 7(g) and 7(h)), tension gashes (F7b in Figure 7(i)) associated with stylolites oriented at an angle to bedding (S2, see below), and bedding-oblique to beddingperpendicular stylolitic and sheared fractures (F7c in Figure 7(j)). F7a fractures are filled with calcite cement forming bridges that span across fracture walls Cc7a in Figure 7(h). Bridges are evidenced by fracture wall-parallel trails of fluid inclusions (see below) indicating a crack-seal mechanism of cementation during fracture opening (synkinematic cement; discussed by Lander and Laubach [72] and Ukar and Laubach [73]). Bridges are not remnants of dissolved calcite but indicate calcite accumulation that depends on thermal exposure and renewal of the fracture surface area during repeated fracture opening [72]. F7a fractures may preserve as much as 25% residual porosity. Tension gashes (F7b) tip out~1 cm away from S2 stylolites and are subparallel to stylolite teeth (Figure 7(i)). F7b fractures are filled with blocky calcite cement (Cc7b) and dolomite cement (Dc1) (Figure 7(i)). Cc7b ranges between 50 μm and 1 mm in size and appears unzoned and light luminescent under SEM-CL. Dc1 forms 8 Geofluids
Bed-perpendicular to bed-oblique stylolites (S2) are locally abundant, best developed in J, L, M, N, O, and P cores (Figure 3(b)). Randomly oriented S2 give the rock a pseudo- 4.1.6. Cavities and Vugs. Pores larger than the average grain size (~20 μm) are abundant in host rocks of all formations. Vugs are usually between 0.5 and 2 mm in diameter, although some exceed 1 cm (Figure 8(a)), and have irregular edges supporting dissolution as their formation mechanism. Many vugs are associated with or terminate at fractures, indicating a temporal and genetic relationship among fractures and vugs (Figure 8(a)). Crosscutting relationships between different types of vugs could not be established, but on the basis of the association between fractures and vugs and similarities in mineral fill, we describe five generations of vugs formed synchronously with F2-F7 fractures.
Although textural evidence indicates that mottle fabrics were not open voids in the sense of karstic breccias such as those found in the Lianglitage Formation, mottle fabrics are the earliest dissolution feature in the Yijianfang Formation.   12 Geofluids

Geofluids
We designate these as the first generation of vugs (V1). V2 vugs overprint mottle fabrics and are infilled with blocky, nonluminescent to bright orange luminescent calcite (Cc2; Figure 6(e)), whereas V3 vugs are partially infilled with bladed to blocky rim cements (Cc3a) and bitumen (Figures 8(b) and 8(c)) and V4 vugs contain red luminescent calcite (Cc4; Figure 8(a)). No V5 vugs were identified in cores. V6 and V7 vugs are infilled with coarse blocky calcite cement (Cc6 and Cc7b, respectively) (Figures 7(k) and 8(d)). V3 are the most abundant type of vug in the Yijianfang Formation and are also present in the Yingshan and Tumuxiuke Formations, within the karstic horizon of the Lianglitage Formation. V3 are present in A-I cores, coinciding with the location of mottle fabrics in the northern part of the study area (Figure 3(b)). In fact, F3, V3, S1, and mottle fabrics show a close relationship (Figures 8(e) and 8(f)); where an F3 and/or S1 encounter a dark mottle, porosity (secondary) increases giving the rock the aspect of pseudobreccia as described by Fu [35]. In some cases, this secondary dissolution affects the area immediately surrounding the mottle fabric (Figure 8(g)). The pore space within mottle fabrics was subsequently infilled with bitumen giving these areas their characteristic dark color (Figure 8(h)).
Cc3b and Cc4 that precipitated in fractures show dissolution corrosion (Figures 8(b) and 8(i)) providing evi-dence of dissolution-enhanced fractures. Corroded Cc4 calcite is usually associated with metal sulfides (pyrite mineralization) and kaolinite (Figures 8(i) and 8(j)). This dissolution event resulted in the formation of as much as 15% secondary porosity.

Geochemistry. Stable isotopic and fluid inclusion analyses
were conducted in order to shed light on the origin of the fluids that led to the precipitation of fracture-and vugfilling cements. Geochemical analyses are summarized in Table 2 and Supplementary Materials I, II, and III.

Carbon and Oxygen Isotopes. Yingshan and Yijianfang
Formation host rocks yielded δ 18 O values ranging from -7.7 to -4.7‰VPDB and δ 13 C values ranging from -0.1 to +1.6‰VPDB (Figure 9). δ 18 O of Tumuxiuke Formation carbonates ranges from -6.7 to -5.9‰VPDB and δ 13 C from -0.6 to +1.7‰VPDB, whereas δ 18 O of Lianglitage Formation carbonates ranges from -8.7 to -4.3‰VPDB and δ 13 C varies from -0.6 to +3.2‰VPDB. Red clasts within Lianglitage karstic breccias have δ 18 O ranging from -7.3 to -6.4‰VPDB and δ 13 C ranging from +1.2 to +1.7‰VPDB (Figure 9).  14 Geofluids  (Figure 9). The isotope equilibrium temperature equation of Craig [75] can be applied to calculate approximate temperatures of cement precipitation. We use the heaviest stable isotopic value of marine fluids obtained from host rocks and isotopic values reported for Middle Ordovician carbonates [76] (Figure 9). Precipitation of Cc0 within the host rocks is inferred to have occurred at~25°C, whereas Cc1 precipitated at <20°C, Cc2 at~30°C, Cc3 at <45°C, and Cc4 at <80°C. The presence of exotic minerals accompanying Cc5 in F5 fractures indicates that the system was open and that δ 18 O SMOW of the parental fluid probably had a different composition than the value assumed for Ordovician marine water so this method was rendered unreliable for these and subsequent cements.  (Figure 9). Due to their small size, we have no Sr data for Cc7 cements.

Fluid Inclusion Microthermometry.
Most fracture and vug cements in all formations trapped primary aqueous fluid and hydrocarbon (oil) inclusions in fluid inclusion assem-blages (FIA) occurring along various generations of growth zones or close-packed clusters within a single growth zone or in the core of host grains ( Figure 10). Several FIAs of secondary aqueous and hydrocarbon FIs were also observed. The aqueous inclusions range from 5 to 25 μm in the long diameter, but most are <10 μm. The FIAs contain either single-phase liquid inclusions only or two-phase liquidvapor inclusions. In a few cases, we observed inclusions containing also a solid phase which was typically opaque black or light amber, interpreted to potentially be a solid hydrocarbon phase (wax or bitumen). The solids appear only sporadically in a few inclusions within the same assemblage, so their phase ratios (liquid-vapor-solid) vary. Variations in phase ratios of inclusions within these FIAs are interpreted to indicate that the solid was trapped accidentally in the inclusions and not as a result of precipitation from the fluid. A summary of homogenization temperatures (T h ) and salinity of primary two-phase aqueous inclusions observed in various cements is presented in Figure 11 and Table 2.
Overall, the aqueous fluid inclusions show a wide range of T h from~50°C to 165°C for all cements, with temperature variation ranging from~2°C to 40°C within individual FIAs. Ice melting temperatures also show a wide range from 1.1 to 11.1°C, corresponding to salinities of 0.4 to 19 wt% NaCl equivalents [77]. The eutectic (first ice melting) temperature of the inclusion fluids was observed at around -50°C, indicating that the aqueous fluid trapped in the inclusions can be adequately described as a H 2 O-NaCl-CaCl 2 -(±MgCl 2 )-type fluid. Melting of other phases at this temperature, such as antarcticite [77], was not observed due to the small size of the inclusions. However, one sample contained fluid inclusions in which we observed the melting of two solid phases during the freezing-heating runs. The first solid melted at temperatures ranging from -22.6°C to -21.2°C, interpreted to be hydrohalite, and the second melted at temperatures ranging from -12.6°C to -4.8°C, interpreted to be water ice. The melting of these two phases allowed us to determine that the aqueous fluid contained 0 to 3.9 wt% CaCl 2 and 7.7 to 15.7 wt% NaCl [77]. An analyzed F5 fracture shows a clear sequence of cement precipitation with zoned fluorite forming first, followed by barite and calcite (Cc5). All these cements trapped primary aqueous fluid inclusions that show a systematic temperature variation from~80°C to 110°C in fluorite, from~120 to 140°in barite, and~60 to 100°C in late Cc5 calcite (Figure 11). The salinity of the fluids shows a similar trend that mirrors temperatures, ranging from 9.6 wt% to 10.5 wt% NaCl equivalents in fluorite, up to~19 wt% NaCl equivalents in barite, and from 5.7 to 14.7 wt% NaCl equivalents in Cc5 (Figure 11).
Primary oil inclusions are present within Cc3, Cc4, and fluorite cements in all three formations but are absent in early Cc0 host rock, Cc1 fracture, and Cc2 vug cements. We have no primary oil inclusion data for Cc5 and Cc6 cements. FIAs of oil inclusions show both heterogeneous and apparent homogeneous liquid-to-vapor ratios within the same FIA at room temperature. Oil inclusions in FIAs with homogeneous liquid-to-vapor ratios show T h ranging from 44 to 75°C ( Table 2). The fluorescence color of the oil inclusions under UV light ranges from yellow to green in Cc3 calcite and F5 17 Geofluids fluorite cements of the Yijianfang Formation, and green to bright blue in Cc3 cements of the Lianglitage and Tumuxiuke Formations as well as subsequent Cc7 cement of the Yijianfang Formation (Figure 11).

Discussion
Discerning cavities formed by shallow (epigenetic) dissolution from those due to deep-seated (mesogenetic) dissolution is challenging since by definition, the cavities themselves cannot be sampled. In cases where cavities are small enough to be sampled by core (vugs), voids preserve scant evidence of their origin. In many cases, however, vugs and associated fractures within the studied carbonates contain cement linings and fills that can be used to obtain information about the physical and chemical conditions during the various stages of dissolution and cementation that affected these rocks. Information about temperature, pressure, stress, and fluid compositions during deformation [40][41][42][43][44] is recorded in cements and fluid inclusions trapped within them. Cemented vugs and dissolutionenhanced fractures identified in this study are cut by subsequent generations of cemented vugs and fractures, indicating cements recorded conditions prior to and following dissolution. The sequence of structural diagenetic features identified in this study, and in particular the cements that precipitated within them, can be used to constrain the evolution of physical, chemical, and mechanical conditions that led to the formation of hydrocarbon-hosting cavities in the Halahatang area.
Although Middle-Lower and Upper Ordovician carbonate rocks show many similarities in their structural diagenetic evolution, not all features are present in all formations. Our sample suite allowed establishing the most complete 18 Geofluids paragenetic sequence for the Yijianfang Formation, which is presented below (Figure 12).  [78] for Middle Upper Ordovician marine carbonates (Figure 9). Therefore, host rock cements (Cc0) most likely precipitated from marine pore fluids. However, 87 Sr/ 86 Sr ratios of the Tumuxiuke and Lianglitage Formations are slightly higher than the values expected for Upper Ordovician marine carbonates, probably due to the clay mineral content of the matrix in the mudstones and wackestones [79]. Low salinities, similar to those of seawater, of some fluid inclusions in Cc0 host rock cements support precipitation within the marine diagenetic realm ( Figure 11). However, T h -considered   Figure 12: Summary of the diagenetic sequence in the studied Ordovician carbonates. 20 Geofluids minimum trapping temperatures [80]-and salinities of some of the inclusions are highly variable and much higher than it would be expected for a shallow marine environment. Moreover, fluid inclusion temperatures are much higher than those calculated on the basis of oxygen isotopic contents. The high variability in T h and salinities supports the inference that the high temperatures and salinities of such fluid inclusions are the result of reequilibration with burial fluids during progressive burial to higher depths [81,82]. The reequilibration can be a combination of adjustment of the inclusions to new temperature and pressure conditions by stretching or changing the volume of the inclusions or by partial infiltration of new fluids, mixing with the inclusion fluids. The inferred reequilibration of inclusions suggests that caution must be taken if temperatures are correlated with burial thermal history models for timing constraints (e.g., [83]). Our study shows that mottle fabrics are neither the result of penetrative surficial karstic processes nor deep diagenetic features but are burrows that were subsequently overprinted by mesogenetic dissolution. Channel-shaped mottle fabrics infilled with fragmented fossils support the interpretation that these are calcite-filled burrows characteristic of early diagenesis [84]. Cements in mottle fabrics and host rocks have similar geochemical characteristics indicating pores formed due to minor dissolution of the host rock. 87 Sr/ 86 Sr ratios higher than those of the host rocks may be due to the alteration of illite and kaolinite within intergranular pores of Cs0 [64,85,86] (Figure 9). Our results are broadly compatible with previous interpretations of mottled fabrics being the result of selective diagenetic alteration and subsequent infiltration of bitumen [35]. The irregularity of mottle fabric walls and their distribution may reflect dissolution caused by hydrocarbon-bearing fluids that upwelled from underlying hydrocarbon source rocks [22].

Stage 1: Marine Phreatic Diagenesis.
Uplift processes, such as those experienced in the area during the Caledonian and Hercynian orogenies, typically lead to reduced temperatures and pressure and may be accompanied by substantial changes in pore-fluid composition due to deeper penetration of meteoric fluids from elevated recharge areas [87]. Such fluids can travel along fractures and cause dissolution, forming solution-enhanced fractures with irregular walls such as F1. Palisade calcite cements (Cc1) and internal sediment (Cs1) within F1 fractures are characteristic of precipitation within the phreatic-vadose level [8,41,88]. However, δ 18 O and δ 13 C isotopic values indicate precipitation from marine fluids [40]. Infiltrated meteoric waters could mix with pore waters so that fluids would show low, but nonzero, salinities. Low salinities of fluid inclusions in Cc1 cements within the overlying Tumuxiuke Formation support marine precipitation, but the high T h suggest reequilibration during progressive burial [81]. Using a geothermal gradient of 28°C/km for the Tabei Uplift during the Middle Ordovician [89], calculated temperatures of <20°C indicate that Cc1 cements precipitated very near the surface, consistent with precipitation in the Middle Ordovician [90][91][92][93] (Figure 13).

Stage 2:
Marine to Shallow Burial Diagenesis. This stage was characterized by the onset of burial and formation of compaction-related fractures and stylolites (S1) as well as dissolution. Cc2 cements in F2 fractures and V2 vugs show δ 13 C values similar to those of the host rock, whereas δ 18 O values of these cements are slightly more depleted (Figure 9). This depletion suggests that Cc2 precipitated from marine fluids during progressive burial of the rock [94]. Estimated temperatures of 30°C indicate precipitation of Cc2 at~220 m depth, within the shallow burial diagenetic environment [95]. Fluid inclusions within Cc2 in the Yijianfang and overlying formations show a range but generally higher temperatures, suggesting, again, reequilibration of the inclusions at higher burial depths. The 87 Sr/ 86 Sr ratio of these cements is similar to that of the host rocks (Figure 9), indicating high rock-fluid interaction [96]. A comparison with a representative burial history curve for the study area indicates that Middle Ordovician strata attained those depths around the end of the Ordovician ( Figure 13). The lack of primary oil inclusions in Cc2 or earlier cements agrees with this timing, preceding the first stage of hydrocarbon generation in the Silurian [1,61]. It is likely that similar fluids that caused stage 1 dissolution continued to cause corrosion of carbonates at this stage.

Stage 3: Intermediate Burial
Diagenesis. This stage was characterized by the formation of F3 fractures and V3 vugs. The spatial association between F3, V3, and mottle fabrics indicates V3 most likely formed by the enlargement of dissolution-prone, more porous, diagenetically altered areas (mottle fabrics). Cc3 that precipitated in F3 fractures and V3 vugs shows depleted δ 18 O values compared to the host rock indicating burial and precipitation from low-to medium-temperature formation waters (Figure 9). Calculated temperatures of <45°C indicate that Cc3 cements precipitated at~625 m depth, within the intermediate burial diagenetic environment [17,95], in contrast to the higher temperatures of fluid inclusions in overlying formations ( Figure 11). The similarity of δ 13 C values between Cc3 and the host rock indicates that carbon was buffered by the host rocks [94]. The 87 Sr/ 86 Sr ratio, similar to that of the host rock, also indicates the prevailing interaction between the fluid and the host limestone ( Figure 9).
Cc3a filling V3 vugs and F3 fractures contains primary oil inclusions, indicating that oil was present in the system at the time of precipitation of these cements. Oil presence is further supported by bitumen in group 3 fractures and vugs. The first oil accumulation in the study area, derived from Cambrian source rocks [90], dates to the Late Caledonian-Early Hercynian period [1,61]. This timing of oil generation and inferred depth for Cc3 cements is compatible with the formation and partial cementation of group 3 features in the Silurian (see Ukar et al. [93]). Heavy crude oil in the study area is rich in H 2 S, and sulfur content may be as high as 2% [63]. A rise in the temperature of the reservoir caused by burial would trigger thermal decomposition of sulfur compounds to produce more H 2 S [1]. Sulfur-rich oil can mix with water to create H 2 SO 4 in a process similar to that which created Carlsbad Caverns in the Guadalupe Mountains (Texas, USA) [97]. Fluids, including hydrocarbons and acidic fluids associated with sulfur-rich oil, flowed along F3 fractures and S1a stylolites causing localized dissolution and formation of vugs. Host rock reworking and dissolution associated with burrowing that led to the formation of early diagenetic mottle fabrics generated up to 1-3% of porosity within these areas, allowing for infiltration of dissolving fluids and localization of subsequent dissolution around dissolution-prone mottle fabrics. Where an F3 and/or S1a encountered a dark mottle, porosity (secondary) increased giving the rock the aspect of pseudobreccia [35].  (Figure 9). Cc2 to Cc4 follow a general trend of slight increase in δ 13 C and decrease in δ 18 O that suggests rock-water interaction at progressively more elevated temperatures. Calculated temperatures of <80°C indicate that Cc4 precipitated at~2000 m depth, within the deep burial diagenetic environment [17,95]. The similarity of δ 13 C values between Cc4 and host rocks indicates that carbon was buffered by the host carbonate [94]. However, high 87 Sr/ 86 Sr values of these calcites suggest incorporation of radiogenic Sr derived from older rocks such as evaporitic Cambrian strata [98,99], from the Precambrian basement [100,101], and/or alteration of feldspars, clays, and other clastic terrigenous minerals present within deeper strata [102].
Cc5 and Cc6 show more depleted δ 13 C and/or heavier δ 18 O, departing from the general trend of increasing temperatures ( Figure 9) and pointing to fluid mixing in a hydrologically open system [16,40]. Precipitation of noncarbonate minerals along F5 (fluorite, barite, and celestite) supports changes in fluid chemistry and/or temperature. Infiltration of high-temperature fluids is supported by generally high FI temperatures within F5 and F6 cements ( Figure 11). The partial dissolution experienced by some F4 fractures could have been caused by such high-temperature fluids, possibly from a deeper source. This is supported by the parallel and systematic change of temperatures and salinity, showing increasing then decreasing trends, from earlier fluorites to late calcite cements in F5 fractures. The oxidation of pyrite associated with F4 and stylolites would be accompanied by a pH decrease creating an acidic environment that would drive dissolution [103,104]. Corrosive fluids may have also been associated with subsequent precipitation of fluorite [16].
A 400 ± 37 Figure 13: Estimated timing and depths of occurrence of the diagenetic stages described in the text in the context of a burial history model calculated for one of the wells in our study area (burial history curve from Zhu et al. [61]; well HA601-4). 22 Geofluids filling by the end of the Devonian, consistent with inferred depths of calcite precipitation, reached by Ordovician strata at that time ( Figure 13). 87 [91,92,105] suggesting that Cc6 acquired its high Sr isotope ratios by incorporating radiogenic Sr derived from magmatic fluids [64,79,84,86,106]. Dolomite (Dc1) and calcite cement (Cc7) mineralization associated with F7 fractures, V7 vugs, and stylobreccias (S2, fault breccias) indicates the presence of Mg-rich fluids in the system following Permian activity. The penetrative deformation recorded by these structures most likely occurred in the Jurassic-Early Cretaceous ( [93]).
Younger Cc7 cements in the Yijianfang Formation contain primary oil inclusions with higher API gravity values (dull to light blue color;~40-45 API°gravity) than older Cc3 and fluorite cements (yellow to green color;~30-35 API°gravity), which is consistent with increased oil maturity in the system over time (Figure 11). The UV light fluorescence color of oil inclusions in Cc3 and Cc4 cements in the overlying Tumuxiuke and Lianglitage Formations is more mature (green to light blue) than that in equivalent features within the Yijianfang Formation and is similar to that in Cc7 within the Yijianfang Formation. This trend is the opposite of that expected for variations in thermal maturity of oils with depth [107]. Without a constraint on the composition of the oils, possible explanations for the variations in the UV light fluorescence color of the oils between formations include locally occurring factors such as biodegradation, water washing, and migration fractionation ( [107]) within a potentially weakly interconnected fluid system across the clay-rich Tumuxiuke Formation at the time of Cc3 cementation.

Dissolution and Fluid
Pathways. In the Tabei Uplift, a seismic interpretation of the top Ordovician illustrates an erosional topography and seismic geomorphologic patterns related to the unconformity, with numerous sinuous fluvial channels and canyons, fluvial valleys, sinkholes, and tower karsts and hills [2,108,109]. Intensive karstification was hypothesized to be related to a prolonged sea level fall and subaerial exposure, which coincided with a major glaciation at the end of the Ordovician [110].
However, in the Halahatang area, the only breccias we encountered that resemble karstic fillings are the ones present in the Lianglitage Formation. Localization of silt quartz, clays, and iron oxides to within brecciated areas indicate that such sediments were most likely transported into cavities during subaerial exposure (epigenetic dissolution). Similar sediment-filled cavities have been extensively documented in cores of the Lianglitage Formation in other study areas [27,111]. Moreover, the presence of widespread near-surface paleokarst is evident in seismic patterns, where topographic relief of 100s of meters resembling the modern topography of the Guilin active karst region is visible at the top of Upper Ordovician carbonates. The fracture-cavity network has been related to incised valleys, the result of an ancestral river at the top of the Lianglitage Formation [112].
Karstic breccias are absent in the Yijianfang Formation in our study area. Early diagenetic features here are mottle fabrics, but these show similar isotopic and lithologic compositions as the host rock indicating that they formed in situ due to minor reworking and dissolution associated with burrowing. Primary porosity associated with near-surface mottle fabric dissolution was <3%, similar to the 1-3% of the matrix. The 5-12% porosity reported for the reservoir [113] must have therefore been caused by subsequent dissolution. Textural and isotopic evidence shows that the majority of the dissolution recorded by Yijianfang Formation strata occurred within the intermediate and deep diagenetic environments (mesogenetic dissolution). Although the Lianglitage and Tumuxiuke Formations contain evidence for mesogenetic dissolution, some of the structural diagenetic features identified in the Yijianfang Formation are absent in the overlying formations (Table 1). This difference in structural diagenesis between the Lianglitage and Yijianfang Formations suggests a hydrologic isolation by the presence of caprock of the compact muddy limestone stratum of the Tumuxiuke Formation in between [25,57].
However, petrographic, SEM-CL, and isotopic compositions of vug-and fracture-filling cements of many of the identified structural diagenetic features are the same across the Tumuxiuke Formation indicating that the carbonate rocks below and above this capping horizon were somehow connected, in agreement with observations by Zhu et al. [114]. The close association between fractures and dissolution vugs indicates corrosive fluids flowed along fractures, some of which must have extended vertically across formations for tens to hundreds of meters. Diagenetic cements that precipitated along stylolites indicate that these also acted as pathways rather than barriers to flow.

Multiple Shallow and Deep Dissolution Events.
Multiple karst events related to surficial exposure and shallow to deeply penetrating groundwater are widespread in Paleozoic carbonate rock sequences, and groundwater can penetrate to great depths and cause dissolution (e.g., [115]). Dissolution along faults is also widely accepted [18,20,116,117]. But the evidence we found of repeated dissolution and precipitation primarily at great depth and not necessarily associated with exposure surfaces or groundwater contemporaneous with exposure raises a question of how useful the term "paleokarst" is to describe both (1) features with clear linkages to surficial or near-surface processes and (2) dissolution at great depth owing to corrosive fluids penetrating along faults and fractures that may have uncertain or no connection with karst topography or groundwater. We contend that the structures we describe are probably not usefully categorized as paleokarst, since processes operating at depth may markedly differ in characteristics and timing of dissolution and may have patterns that differ substantially from those governed by paleotopography and other near-surface controls. For example, corrosive fluids might catalyze fault slip and fault porosity evolution as a consequence of chemicalmechanical interaction (e.g., [118]). We prefer to identify 23 Geofluids and categorize dissolution features with a descriptive terminology.
Dissolution at deep burial depths on a massive scale (up to 20%) has been doubted [119,120]. Ehrenberg et al. [119] and Ehrenberg [120] argued that burial dissolution in carbonates would not contribute significantly to a net increase in carbonate reservoir porosity. Furthermore, they stated that pore waters highly undersaturated with respect to calcite would in most cases be neutralized very rapidly during migration before reaching carbonate reservoirs and that consequently dissolution would be limited to a narrow reaction front along fractures in case acidic waters would reach the reservoir level. But in the case of Halahatang, the prevalence of active faults resulted in multistage reactivation of faults both in shear and in extension, and the formation of porous fault rock and dissolution-enhanced cavities, as evidenced by stylobreccias and associated fractures and vugs as well as fault rocks exposed in nearby outcrops ( [93]) and cavities imaged near faults [23,36,37,121]. On the basis of fault porosity and abundance and core observations, Ukar et al. [93] suggest that local mesogenetic porosity enhancement of as much as 5-10% is plausible. Both porous fault rock and dissolution-enhanced cavities and fractures could provide hydrocarbon storage and migration pathways and be responsible for decameter-scale bit drops and large volumes of drilling mud losses.

Conclusions
Petrographic, geochemical, and microthermometric results from~230 samples from 16 wells across the Halahatang oilfield indicate that Middle Ordovician rocks underwent multiple stages of fracturing, dissolution, and mineralization.
Unlike the overlying Lianglitage Formation where nearsurface karstic breccias are widespread, in the Yijianfang Formation we found no evidence for cave-fill sediments that would have resulted from penetrative epigenetic karstic processes. Near-surface diagenetic features in the Yijianfang Formation are restricted to mottle fabrics filled with allochthonous carbonate silt and cements formed as a result of bioturbation (stage 0). Primary porosity associated with near-surface dissolution is similar to that of unaltered host rocks (1-3%), arguing against a significant increase in porosity-permeability related to this event. Subsequent secondary dissolution caused porosity within mottle fabrics to increase by as much as 10%.
Crosscutting relationships between sequences of partially cemented fractures and vugs that postdate mottle fabrics and stable isotopic compositions of calcite cement fillings indicate that multiple subsequent structural diagenetic stages occurred within the shallow to deep burial diagenetic environments. δ 18 O analyses tied to geothermal gradients indicate cement precipitation within the shallow (~220 m, stages 1-2), intermediate (~625 m, stage 3), and deep (~2000 m, stages 4-7) diagenetic environments. The presence of oil fluid inclusions in cements at stage 3 and younger confirms the presence of hydrocarbons in the system at the time of void-filling mineral precipitation.
Stages 0-2 dissolution was associated with possible infiltration of meteoric fluids, whereas stages 3-4 were likely caused by acidic fluids associated with Middle-Late Silurian and Devonian-Permian hydrocarbon generation. Stage 5 was associated with fluorite and barite mineralization with a possible deeper fluid source. An isochron age of 400 ± 37 Ma confirms stage 5 dissolution and subsequent cementation occurred by the end of the Devonian. Stage 6 calcite has elevated 87 Sr/ 86 Sr as a result of external fluid infiltration associated with Permian igneous activity. Stage 7 stylobreccias and fractures were associated with Mg-rich fluids and Jurassic-Cretaceous deformation. Fluctuating, but high (80-120°C), homogenization temperatures and generally high salinities (5-20% NaCl equiv.) of fluid inclusions trapped within late calcite cements indicate potential reequilibration of the fluid inclusions with deeper, higher-temperature fluids. Close association of fractures with vugs of all stages and mineralization along stylolites indicate that they both acted as corrosive-fluid pathways.

Data Availability
The data that support the findings of this study are provided in the supplementary material.

Conflicts of Interest
No conflict of interest was declared.