Petrogenesis of a Large-Scale Miocene Zeolite Tuff in the Eastern Slovak Republic: The Nižný Hrabovec Open-Pit Clinoptilolite Mine

Zeolites occur in several geologic settings of differing ages, but nearly all mineable deposits are of Cenozoic age, particularly in the United States. The predominant zeolite precursors worldwide are acid to intermediate volcanic glasses from ashfall tuffs interbedded with lacustrine sediments, from vitroclastic tuffs in volcanic rocks or from hydrothermally altered vitrophyres and vitroclastic tuffs. Clinoptilolite occurrences unrelated to volcanic ash are less common and typically linked to feldspar, feldspathoid, biogenic silica, or clay mineral alteration (Eyde and Holmes, 2006). Sheppard (1973) identified six scenarios of zeolite formation: (1) in hydrologically closed (saline/alkaline) systems, (2) in open (freshwater/groundwater) systems, (3) during burial diagenesis/metamorphism, (4) by hydrothermal activity, (5) in deep marine environments, and (6) by weathering of soils.

The world production of natural zeolites in 2018 is estimated at ~1.1 million metric tonnes (Mt), with China producing 300 kt, Korea 120 kt, New Zealand 100 kt, United States 95 kt, Turkey 70 kt, Cuba 57 kt, and Jordan 20 kt; all other countries are summed up to a production of 350 kt (U.S. Geological Survey, 2019). More than several hundred occurrences of natural zeolites are reported in the western United States alone; around 20 are productive clinoptilolite deposits (Eyde and Holmes, 2006).

One of the economically most important natural zeolite occurrences in the world is the deposit around Nižný Hrabovec in the Slovak Republic (Fig. 1A), with reserves of ~150 Mt of clinoptilolite tuff (Chmielewská, 2014). From this deposit, more than 170 kt of high-grade clinoptilolite tuff were extracted in 2018 by the company Zeocem a.s. from the open-pit mine near the village of Nižný Hrabovec, close to Vranov nad Topl’ou, about 30 km northeast of Košice.

Clinoptilolite, which is mined in the Nižný Hrabovec deposit, is a heulandite-type zeolite mineral that is a nontoxic microporous hydrated aluminosilicate. It is characterized by a stable 3-D framework of linked SiO₄ and AlO₄ tetrahedra, with intervening open channels and cages. Partial substitution of Si⁴⁺ by Al³⁺ results in an overall negatively charged crystal lattice, largely neutralized with diverse exchangeable extraframework cations (commonly K⁺, Na⁺, Ca²⁺, and/or Mg²⁺) that are loosely bound within the open pore structure (Bogdanov et al., 2009). Water molecules also occupy open framework cavities (Inglezakis and Loizidou, 2012).

These lattice properties, combined with the resulting sorptive character, make clinoptilolite an important, highdemand commodity, accounting for the predominant part of natural zeolite production (Chmielewská, 2014). Clinoptilolite is used as a building and decor stone, in cement and concrete aggregates, as a paper and rubber filler, for soil conditioning and fertilizing in horticulture and agriculture, and in treating gas and exhaust gas, water and wastewater, and organic and nuclear waste (cf. Mumpton, 1999; Reháková et al., 2004; Eroglu et al., 2017). It is also important both for animal husbandry (cf. Valpotić et al., 2017) and for medical purposes in humans (cf. Colella, 2011). Chemically modifying the surface of clinoptilolite and loading the framework with different cations allows many further uses (e.g., Inglezakis and Zorpas, 2012).

Few whole-rock analyses of the Hrabovec tuff have been published in the international literature (Šamajová, 1997; Chmielewská et al., 2002; Reháková et al., 2004; Chmielewská, 2014). However, these are mostly limited to single specimens, with general descriptions of the mineralogical composition. Chmielewská (2014) presented whole-rock and trace element data from one sample, while Šamajová (1979) used scanning electron microscopy (SEM) to investigate the formation of the zeolites.

This study presents a detailed and integrated petrological and geochemical investigation of one of the world’s largest actively mined clinoptilolite tuff deposits. Whole-rock mineralogical and major and trace element compositions, as well as Sr and Nd radiogenic isotope analyses with in situ electron probe microanalysis (EPMA) measurements and SEM observations on clinoptilolite-bearing rocks, sampled along a ~80-m vertical profile, are presented here. The aim of the study is to determine how the deposit formed in order to reveal the potential precursor materials of clinoptilolite and the conditions under which the zeolitization processes occurred. High-resolution EPMA mapping allows us further to present entirely new insights into the mineralogical conversion from a calc-alkaline vitric tuff to a clinoptilolite-bearing natural commodity.

The East Slovak basin is an intramontane basin, lying at the junction of the Western, Central, and Eastern Carpathians, which are parts of the Alpine-Himalayan orogen (Fig. 1A). The basin evolution was strongly controlled by collisional processes during the Carpathian orogeny, followed by thermal subsidence and extension in the Pannonian domain (Schmid et al., 2008). Note that the Central Paratethyan stage names are used here (e.g., Piller et al., 2007; Hohenegger et al., 2014); the relationship of these to the formal terminology of the International Commission on Stratigraphy (www.stratigraphy.org/) is explained in Figure 2. 

Tectonically, the East Slovak basin is a back-arc basin comprising shallow-marine to continental, in part volcanogenic, sediments with a total thickness of up to 7 km (Kováč et al., 1995). The history of the basin comprises several shallowing-upward sequences that can be related to an early Miocene depocenter in the northwest, an early to lower-middle Miocene depocenter in the center, and an upper-middle to late Miocene depocenter in the southeast of the basin (Janáček et al., 1969). An initial NS-compressive regime was followed by Ottnangian (middle Burdigalian) uplift, facilitated by dextral strike-slip zones. Karpatian pull-apart subbasins developed further into northeast-southwest extensional basins in the early and middle Badenian. In the late Badenian, the tectonic evolution of the East Slovakian basin was characterized by extension along normal and strike-slip faults, resulting in an attenuation of the crust that was associated with a dramatic increase in the geothermal gradient (Konečny et al., 2002; Pécskay et al., 2006).

During the middle and late Sarmatian (late middle Miocene), the East Slovak basin was affected by intense calc-alkaline volcanic activity, including explosive eruptions of silicic magmas (Harangi et al., 2007). These were generated by back-arc transtension/extension affecting a relatively thick continental crust during subduction, coupled with asthenospheric mantle upwelling during the early to middle Miocene (Kováč et al., 1995; Konečny et al., 2002; Pécskay et al., 2006). Harangi et al. (2007) suggested that lithospheric extension was the main cause of the calc-alkaline magmatism in the Western-Central Carpathians, with subduction having an indirect influence. During the Pannonian (late Miocene), the East Slovakian basin represented the northern, marginal part of the Pannonian basin system. The last tectonic stage of the basin evolution was characterized by Pliocene to Pleistocene NE-SW–directed compression (Kováč et al., 1995).

Along the northwest margin of the East Slovak basin, a lower Badenian (16.3–15.03 Ma) shallow-marine sequence crops out (Nižný Hrabovec Formation; 500–600 m thick), comprising fine-grained sandstones and mudstones grading into coarse-grained sandstones overlain by volcanogenic sediments of the Hrabovec tuff (Šamajová, 1979; Varga et al., 1985; Kováč et al., 1995). This tuff, which has a thickness of ~100 m in a 7-km-long belt running from the villages of Pusté Cemerné in the southeast, through Nižný Hrabovec, Kucín, Majerovce to Vranov nad Topl’ou in the northwest (Fig. 1A; Šamajová, 1979; Kováč et al., 1995), was strongly affected by diagenetic-anchimetamorphic processes soon after deposition. This resulted in the formation of authigenic clinoptilolite and cristobalite (Šamajová, 1997). The tuff is overlain by the Middle Badenian Vranov Formation (800–1,000 m thick), comprising neritic to shallow-marine sandstones and mudstones (Varga et al., 1985; Kováč et al., 1995).

The outcrop of the Hrabovec tuff studied here lies in the Nižný Hrabovec open-pit mine, which was 350 m long east-southeast to west-northwest and 150 m wide north-northeast to south-southwest, with an overall depth of 80 m (Figs. 1, 2).

The structurally lowermost part of the mine, best seen at the northwestern corner, exposes SW-dipping meter-scale bedded, but otherwise structureless, brownish coarse-grained sandstones, with millimeter-sized detrital white micas forming the lower part of the lower Badenian Nižný Hrabovec Formation. Above the Hrabovec tuff, which forms the upper part of the Nižný Hrabovec Formation, a layered sequence of decimeter-thick sandstones with interbedded mudstones (Vranov Formation; middle Badenian) has the same bedding orientation as the sediments below the tuff, dipping at ~45° to the south-southeast (Figs. 2, 3A). Graded bedding in sandy layers of the Vranov Formation records fining-upward sequences, with load casts at the base of the sandstone beds, and occasionally preserves current ripples, suggesting deposition in a turbiditic environment.

Between these siliciclastic units, the ~100-m-thick clinoptilolite tuff of the Hrabovec Formation crops out, constituting a very pale blue-green to green rock, weakly stained orange by fluids adjacent to faults or fractures. Some parts of the lower-most level on the north side of the mine show a rhythmic layering with an orientation parallel to the over- and underlying sedimentary bedding (Figs. 2, 3B). This layering is seen more by the refraction of brittle fractures cutting the layering than by significant compositional changes. Elsewhere, the rock breaks easily, sometimes with a very poor conchoidal fracture, suggestive of a homogeneous, ultrafine grain size. The only postdepositional but prelithification structures in the tuff are locally abundant conjugate deformation bands seen as thin (<1 cm wide) slightly darker bands oriented essentially orthogonal to the margins of the deposit; these probably formed during an early stage of tuff diagenesis. The tuff is cut by four faults or fault zones up to a meter thick that only slightly offset the base of the Vranov Formation (Fig. 2). No quartz veins or evidence of quartz mineralization within the fault zones were observed.

The topographic background used for the geologic map of the Nižný Hrabovec mine is an orthophoto mosaic draped onto a high-resolution digital elevation model (DEM; Fig. 2). Both the orthomosaic and the DEM were calculated with the photogrammetric software Agisoft PhotoScan Professional version 1.4.3 (www.agisoft.com) from 514 aerial pictures taken with a DJI Mavic Pro drone on May 23, 2018. For the automatic double grid mission, the Android mobile application Pix4DCapture 4.4.0 was used, with 70% overlap between the pictures and a 70° inclination of the camera. The constant flight height was 20 m above the uppermost level of the mine. The georeferenced orthophoto mosaic (UTM34 N) was made 50% transparent and draped over the DEM using QGis version 2.18.13 (Figs. 1B, 2).

Representative ~1-kg samples were collected, avoiding material that showed or was near deformation bands, fractures and joints, faults, or Fe-stained areas. Three samples were taken on each of the six mine levels, crushed, homogenized, and prepared for all further analytical procedures. Due to the active open-pit mining at several levels and the dip of the tuff horizon (~40° to the southwest), a pseudovertical sampling profile from the bottom to the top of the mine was collected (Fig. 1B).

Powder X-ray diffraction (XRD) patterns were obtained using a Bruker D-8 Advance diffractometer (at Glock Health, Science and Research GmbH) equipped with a Cu source. Tube conditions were 40-kV voltage and 40-mA current. Measurements were performed on powdered specimens in top-loaded sample holders using a θ-2θ configuration, scanning angles from 6° to 70°, and 0.01° step size at 0.5 s/step. Phase evaluation was done with the Bruker software DIFFRAC.EVA (release 2016). For quality control, the reference sample NIST SRM1976a (sintered alumina) and an intralaboratory clinoptilolite reference sample were analyzed before each batch run.

Whole-rock major and trace elements were analyzed by Bureau Veritas Commodities Canada Ltd., in Vancouver, British Columbia. Inductively coupled plasma-atomic emission spectroscopy (ICP-AES) and ICP-mass spectrometry (ICP-MS) measurements were performed after sample fusion with a mixture of lithium-tetraborate, followed by dissolution of the fused sample material in nitric acid. Accuracy and precision of the rock analyses as well as the method detection limit were assured using international reference materials provided by Bureau Veritas (STD GS311-1, STD GS910-4, STD OREAS25A-4A, STD OREAS45E, STD SO-19), as well as reference materials provided by Glock Health, Science and Research GmbH (CRM Zeolite CGL-010 and USGS CRM RGM-2); see Appendix.

Sample preparation procedures and analyses for the acquisition of whole-rock Sr and Nd isotope ratios were performed through Bureau Veritas Commodities Canada Ltd. High-purity reagents were used for digestion of the samples as well as the Sr-Nd column chemistry. The final cuts were measured on a multicollector-ICP-MS (Thermo-Finnigan Neptune). Samples were digested in HF/HNO₃/HCl mixtures and dried down in 6N HCl with boric acid to stabilize fluorides. An aliquot of the sample was processed for Sr separation using the Sr-Spec resin (Eichrom Technologies Inc.) in HNO₃ media. Rare earth elements (REEs) were separated from the matrix using cation resin, and Nd was further purified from the REEs with the Ln resin using 0.18N HCl. To follow the accuracy, precision, and correctness of the sample preparation process, as well as the measurements, for Sr, the results for the certified reference materials USGS AGV1, USGS BHVO1, GSJ JA1, and GSJ JB1a were obtained; for Nd, the La Jolla and IH standards were used. The La Jolla standard was determined at ¹⁴³Nd/¹⁴⁴Nd = 0.511847 ± 0.000030 (2σ, n = 3), consistent with literature data. The SRM 987 Sr standard was determined at ⁸⁷Sr/⁸⁶Sr = 0.710290 ± 0.00001 (2σ, n = 10). As the studied rocks are young, isotopic data are presented and discussed as present-day values; initial values (age corrected to 15 Ma) are given as well (Table 1).

Mineral spot analyses as well as high-resolution element mapping were performed on a Cameca SX-100 EPMA (at Department of Lithospheric Research, University of Vienna) on polished carbon-coated thin sections. Machine specifications, working conditions, and measurement procedures, as well as analytical errors, are documented in detail in Tschegg et al. (2010). Augite, orthoclase, and albite standard minerals were repeatedly analyzed to control the quality of the data. Two measurement setups were established and tested prior to analysis—one with a 6-μm defocused beam technique for feldspars and micas and a second with a 10-μm defocused beam for the clinoptilolite crystals and matrix measurements. To avoid secondary emission effects while analyzing the matrix, the beam was situated only in areas where no individual (visible in backscattered electron [BSE]) mineral phases or porosity were present. To assess the spatial distribution of elements within both the matrix and the clinoptilolite grains, high-resolution element mapping was performed using EPMA. The high-resolution element maps were made with a focused beam at 20-kV accelerating voltage and 40-nA current, and scanning times per spot were 0.2 s. Si, Al, Ca, and Ba were analyzed on one wavelength-dispersive (WD) spectrometer each, and Na, K, and Fe were analyzed simultaneously on the energy-dispersive (ED) detector. Scanning time was ~5 h for each element map.

SEM observations for the petrologic work were done with a Tescan Vega Series SEM (at Glock Health, Science and Research GmbH) on carbon-coated rock chips (diam ~3–5 mm) that were mounted on carbon tabs placed on aluminum pins. Beam conditions were 20 kV and 15 nA.

The XRD spectra summarize the whole-rock mineral composition of the tuff (Fig. 4). A heulandite-type zeolite is the major constituent, with cristobalite and feldspars present in relatively small amounts, as well as accessory biotite and quartz. The major peak of the zeolite phase at 9.85° 2θ describes the (0 2 0) plane (Miller indices) and the second major peak at 22.45° 2θ the (4 0 0) plane. Other zeolite peaks are at 11.18° (2 0 0), 28.11° (4 2 2), and 30.10° (4 4 0) 2θ. The roughly similar heights of the two major zeolite peaks, at (0 2 0) and (4 0 0), indicate the presence of clinoptilolite; in heulandite, the (0 2 0) reflection would be far more intense compared to the (4 0 0) peak (Armbruster, 2001). The main biotite peak (0 0 1) is at 8.82° 2θ, the main cristobalite peak (0 1 1) at 21.87°, the quartz peak (1 0 1) at 26.65°, and the main feldspar peak (0 0 2) at 28.03° 2θ. Minor amounts of cristobalite and feldspar, as well as the traces of biotite and quartz, differ slightly between samples, but the variations show no systematic mineralogical changes within or between the six levels of the mine. The content of amorphous phases, potentially indicated by a slight enhancement of the background signal between 20° and 30° 2θ (Ostrooumov et al., 2012), is negligible. No XRD evidence for other natural zeolite minerals like analcime (Na₁₆[Al₁₆Si₃₂O₉₆] · 16H₂O), mordenite (Na₈[Al₈Si₄₀O₉₆]·24H₂O), or erionite ([Na₂Ca₆K]₉ [Al₉Si₂₇O₂₇] · 27H₂O) has been seen in any of the studied samples (typical unit cell formulas from Ulmanu, 2012).

The samples from Nižný Hrabovec show a very homogeneous major element composition, with small compositional ranges (Table 1). The whole-rock concentrations of SiO₂ are very high, ranging from 74.3 to 77.6 wt %; Al₂O₃ ranges from 12.7 to 14.4 wt %, CaO from 2.7 to 3.6 wt %, K₂O from 3.1 to 4.3 wt %, and Na₂O from 0.3 to 2 wt %. The remaining oxides are present only in very small amounts, with minor variations (1.1–1.6 wt % Fe₂O₃, 0.5–0.8 wt % MgO, 0.1–0.2 wt % TiO₂, and <0.1 wt % P₂O₅). The concentration of MnO is below the detection limit.

A plot of whole-rock compositions on the total alkali versus silica (TAS) rock classification diagram (Le Maitre, 1989) shows that the samples fall in the rhyolite field, with high SiO₂ (74.3–77.6 wt %) and Na₂O + K₂O concentrations from 4.1 to 5.5 wt % (Fig. 5A). On the Na₂O + K₂O versus FeO versus MgO (AFM) diagram (Irvine and Baragar, 1971), the samples form a poor linear trend toward the alkali-rich end of the calc-alkaline series (Fig. 5B). In a plot of SiO₂ versus K₂O, most of the samples lie within the high-K series field with only two samples lying within the medium-K calc-alkaline series field, consistent with a rhyolitic (>70 wt % SiO₂) composition (Peccerillo and Taylor, 1976; Fig. 5C). In the ACNK system (molar whole-rock Al₂O₃, silicate CaO, Na₂O, and K₂O; Fedo et al., 1995), some of the studied samples show a weak trend slightly separate from the magmatic line (points 1–8), reflecting potential geochemical changes during formation of the clinoptilolite tuff with regard to the alteration systematics of Fedo et al. (1995; Fig. 5D).

Forty-four trace elements were analyzed, including REEs. Of the trace elements, eight are below the detection limit for some or all samples (Be, V, Cr, Co, As, Ag, Cd, Tl; Table 1). The large ion lithophile elements (LILE) Ba, Sr, and Pb range from 664 to 891 ppm, 151 to 431 ppm, and 1.6 to 16.7 ppm, respectively. In contrast, Rb shows limited variation, ranging from 93 to 117 ppm. The high-field-strength elements (HFSE) also show homogeneous compositions, with Nb ranging from 8.3 to 9.4 ppm, Ta from 0.7 to 1.0 ppm, Zr from 96 to 166 ppm, Th from 8.7 to 11.1 ppm, U from 3.0 to 4.3 ppm, and Hf from 2.9 to 4.4 ppm.

Total rare earth elements (∑REEs) vary only slightly in concentration from 117 to 141 ppm. The light REEs (LREEs) La and Ce have concentrations ranging from 28 to 36 and 50 to 57 ppm, and Nd and Sm from 18 to 23 and 3 to 4 ppm, respectively. The heavy REE (HREE) concentrations range from 2 to 3 and 0.3 to 0.6 ppm for Yb and Lu, respectively. Chondrite normalized REEs (Sun and McDonough, 1989) show a very uniform pattern, except for sample R1805-10/9 (level 2), which shows a slight enrichment of HREEs (Fig. 6A). A negative Eu anomaly (Eu/Eu*=EuN$SmN⋅GdN$⁠) is present in all samples, with values between 0.42 and 0.60. On the chondrite-normalized spider diagram (Sun, 1980), the data again are very homogeneous, with significant negative anomalies of Nb, Ta, and Ti (Fig. 6B). Apart from three samples (all from level 3), the samples have a weak negative Sr anomaly. In general, the rocks reflect a composition that is strikingly similar to the composition of the upper continental crust (Taylor and McLennan, 1985), with normalized REE values (when normalized to upper continental crust) between ~0.75 and 1.1.

Values of ⁸⁷Sr/⁸⁶Sr range from 0.70868 to 0.70892, whereas values of ¹⁴³Nd/¹⁴⁴Nd range from 0.512443 to 0.512514. The performed age correction to 15 Ma results in marginally lower initial values, which range from 0.70840 to 0.70851 for ⁸⁷Sr/⁸⁶Sr_(i) and 0.512436 to 0.512508 for ¹⁴³Nd/¹⁴⁴Nd_(i) (Table 1).

In Figure 7, the samples show no obvious trend but resemble the Miocene to Pleistocene calc-alkaline rock series of the East Carpathians and the central segment of the Carpathian-Pannonian region, which are slightly less enriched in radiogenic Nd ratios than the West Carpathian series (Fig. 7; Mason et al., 1996; Seghedi et al., 2004; Harangi et al., 2007). The small variance of the Sr and Nd isotopes parallels the small variance in trace element systematics.

The samples have the typical texture of an aphanitic acidic tuff, with larger mineral phases embedded in an extremely fine grained matrix (Fig. 8A, C). Predominantly clinoptilolite, K-feldspar, plagioclase, and, in trace amounts, quartz and biotite were identified by energy dispersive spectrometry (EDS)-EPMA; this is consistent with the mineral assemblage determined by XRD analysis (Fig. 4). Feldspars, quartz, and biotite occur as irregular broken fragments with no obvious crystal growth, recrystallization, or zonation features. Fragment size is typically around 50 to 150 μm; rarely, bigger grains have been seen. Well-developed, idiomorphic clinoptilolite either occurs as microcrystalline pseudomorphs of glass shards (Fig. 8A, C) or as bundles of monoclinic microcrystals growing in voids, cracks, or vesicles (Fig. 8B-D). For both textures (shard replacement and well-developed crystals in voids), cleavage planes and, more commonly, grain boundaries are clearly visible; single crystal lengths are mainly around 10 to 30 μm.

EPMA were obtained on all mineral phases (except quartz) as well as on the very fine grained matrix. Representative spot analyses of clinoptilolite, feldspars, and biotite as well as the matrix are summarized in Tables 2 through 6.

Clinoptilolite grains have SiO₂ and Al₂O₃ contents from 67.6 to 72.3 wt % and 12.5 to 13.4 wt %, respectively, giving Si/Al ratios from 4.5 to 5.1 (Fig. 9A, B; Bish and Boak, 2001). Concentrations of Na₂O and MgO of clinoptilolite are generally low (0.0–0.3 and 0.4–0.9 wt %, respectively), with CaO and K₂O being the major cations in the structural network, having concentrations between 3.5–4.3 and 2.3–3.6 wt %, respectively (Fig. 9A, B). A ternary plot in terms of Na-(Ca + Mg)-K shows that the compositions plot within the clinoptilolite-Ca field, just slightly overlapping into the K clinoptilolite field (Fig. 9C). The clinoptilolite is essentially devoid of Mn, Fe, Ba, and Ti. No compositional differences were found between the crystals forming the shard pseudomorphs and those growing in voids.

The matrix is too fine grained to observe individual crystal phases, even with high-resolution imaging techniques (Fig. 8A-D). However, EPMA shows that the matrix has essentially the same composition as the clinoptilolite phases, with slightly higher SiO₂ concentrations (60.8–74.0 wt %), similar Al₂O₃ contents (10.3–13.4 wt %), and slightly lower K₂O (2.1–3.6 wt %) and CaO (2.24–3.64 wt %) contents compared to the pure clinoptilolite phase (Tables 2, 3; Fig. 9A-C). For Na₂O, Al₂O₃, CaO, MnO, FeO, and TiO₂, a broader variation of composition can be observed in the matrix. The low totals (88.7–93.0 wt %) for the clinoptilolite analyses are due to the water incorporated within the lattice.

High-resolution element maps of two representative areas, in which both coarse clinoptilolite intergrowths and the surrounding fine-grained matrix are visible, are shown in Figure 10. The absence of striking gray shade differences shows the compositional similarity of the coarse clinoptilolite and finegrained matrix. However, within the matrix, areas of higher Si concentration (dispersed white spots) compared to the uniform clinoptilolite areas are visible (Fig. 10B); digitally subtracting Al from Si enhances the Si signal, such that a relatively uniformly dispersed Si phase becomes visible (Fig. 10H). These white areas were not resolvable into individual grains, even with EPMA, reflecting their extremely small grain size. The lack of difference in Al, Ca, K, and Na in the coarse clinoptilolite compared to the matrix is striking (Fig. 10C-F), although the matrix is slightly more enriched in Fe (Fig. 10G). Figure 10I through L illustrates two clinoptilolite crystal patches embedded in the matrix at higher magnification; an accumulation of extremely fine grained Si phases is particularly visible at the boundary between the two large clinoptilolite patches and matrix (Fig. 10J), with Al and Ca lacking in these areas (Fig. 10K, L). The element distribution within clinoptilolite is strikingly homogeneous (indicated by the uniform gray shade).

Feldspars comprise K-feldspar and plagioclase. Potassium feldspar consists of orthoclase (Ab₀Or₁₀₀An₀) (Fig. 9D), in which sometimes micrographic quartz segregations can be recognized. Next to major fault zones, orthoclase is partly intergrown with quartz and biotite and in healed fractures with plagioclase. Plagioclase compositions vary from mainly andesine to oligoclase (Ab_43–61Or_3–7An_32–53). As observed for orthoclase, plagioclase exhibits neither magmatic growth (core-rim) compositional zonation nor any signs of corrosion or alteration processes at the rims.

The biotite grains in the tuff paragenesis have MgO contents from 8.1 to 8.9 wt %, FeO from 22.3 to 25.6 wt %, TiO₂ from 3.7 to 4.6 wt %, and BaO from 0.5 to 1.2 wt %. The biotites preserve well-developed (0 0 1) cleavage planes and are completely undeformed and unaltered.

The data gained from whole-rock mineralogical, major and trace element (including REEs), and radiogenic isotopic analyses, as well as the petrographic and mineral geochemical observations, all indicate that the clinoptilolite tuff mined at Nižný Hrabovec is of volcano-sedimentary origin. The remarkable homogeneity within the monomineralic deposit is reflected by the lack of larger systematic geochemical variations in mineral or geochemical composition along the sampled profile (Tables 1–6; Figs. 5–7, 9).

The ACNK plot indicates the degree of alteration by postdepositional processes (Fig. 5D; Fedo et al., 1995). Igneous rocks generally lie on the line defined by points 1 through 8 (Fig. 5D), although due to magmatic variations not all do so exactly. Typical alteration to clay minerals results in enrichment in Al, driving whole-rock compositions toward illite compositions in ACNK space and then toward A (solid lines, Fig. 5D; Fedo et al., 1995) with the alteration quantified by the chemical index of alteration (CIA = molar [Al₂O₃/(Al₂0₃ + CaO + Na₂O + K₂O)] · 100) of Nesbitt and Young (1982). Subsequent or contemporary metasomatic alteration by K-rich fluid drives compositions toward K (K-enrichment trend; Fig. 5D; Fedo et al., 1995).

Samples 10/3 and 10/5, with CIA values of 52%, lie on the magmatic trend and likely show a tuff geochemistry in ACNK space (red crosses, Fig. 5D), which resembles the precursor composition. Samples 10/13 and 10/14, with the lowest CN fractions, have CIA values of 57% (blue crosses, Fig. 5D). Projection from the K axis through these samples to the alteration trend suggests they were initially altered to a CIA value of 62%, with subsequent K metasomatism reducing the CIA to 57%. However, this two-stage alteration model requires samples that were unaffected by the initial alteration to be also unaffected by K metasomatism. As this is implausible, a simpler model, with alteration occurring in a single event, is preferred, with alteration not toward illite but directly toward a K-richer composition (NH trend; Fig. 5D). Also implausible is the assumption of having a parental magma related to higher amounts of K-feldspar phenocrysts in the precursor rocks of those samples, as this would shift the compositions toward the K-feldspar composition in the ACNK system.

Direct observation of K-rich alteration is generally absent, apart from orthoclase healing some fractures within a few centimeters of fault zones. As the samples for geochemical analyses were taken well away from all brittle structures, this influence can be excluded. There is no chemical zoning or rim corrosion visible in plagioclase phenocrysts, and there is no evidence for sericitization or the presence of clay minerals. Possibly, ultrafine-grained parts within the matrix were more readily K metasomatized. If the alteration seen in ACNK space was contemporaneous with the K-feldspar healing fractures, then it occurred postzeolitization, after the tuff had lithified.

In summary, the ACNK system illustrates that some of the studied samples, with CIA values of 52%, likely reflect the original tuff composition, whereas samples with the maximum CIA value of 57% suggest that in a few parts of the mine slight geochemical changes in the form of K metasomatism have occurred.

On the TAS and AFM plots, the samples with the lowest and highest CIA values plot unsystematically within the overall poor negative trend in the rhyolite field (Fig. 5A, B). That the high and low CIA value samples lie within the cluster of data points in these two plots is due to the concomitant decrease of Na₂O as K₂O increases, giving an almost constant total whole-rock alkali content (Table 1). In contrast, on the plot of SiO₂ versus K₂O, samples with the lowest and highest CIA values lie at opposite ends of the trend (Fig. 5C). Only the samples with the lowest CIA values, representing the ones that are closest to the precursor tuff composition (with respect to their major element composition), lie within the medium-K calc-alkaline series field; all others reflect high-K calc-alkaline affinity.

Trace element (including REE) compositions corroborate the essentially homogeneous nature of the tuff; the chondrite-normalized patterns are characteristic of highly evolved magmas, formed in continental-arc settings (Fig. 6A, B; Winter, 2001). The observed REE patterns resemble middle Miocene calc-alkaline rocks from the Central Slovakian volcanic field (Harangi et al., 2007). The negative Eu anomalies present in all samples are in general linked to plagioclase fractionation during magma genesis and the potential subsequent segregation of the feldspar phases (Fig. 6A). The enrichment of LREEs in combination with the observed Nb-Ta-Ti anomalies can be interpreted as subduction-related geochemical signatures, typical for active continental margins (Fig. 6B; Pearce, 1982, 1983). This is consistent with previous interpretations of the overall geotectonic setting as a continental back-arc basin (Kováč et al., 1995). The uniformity of HFSEs in contrast to the LILEs, which differ slightly in their concentrations, is probably not related to the magmatic record of the tuff but to processes taking place during zeolitization. HFSEs are generally relatively immobile, and their concentrations are essentially insensitive to alteration or diagenetic processes, while LILEs are highly mobile in hydrous low-T environments, and, hence, their concentrations may change during hydrothermal alteration (Hawkins, 1995). REEs are reported to be relatively stable during hydrothermal alteration of rhyolitic rocks (Pandarinath et al., 2008).

Strontium and neodymium radiogenic isotope ratios show the same tendency as the whole-rock geochemical and mineralogical data: a strikingly narrow range in values, mirroring the extremely homogeneous character of the rock succession throughout the mine (Fig. 7). With mean ⁸⁷Sr/⁸⁶Sr and ¹⁴³Nd/¹⁴⁴Nd values of 0.70880 and 0.512463, respectively, an enriched source with a large contribution from crustal sources is indicated.

The lack of clear magmatic trends between the isotopic ratios and major and trace elements makes more involved petrogenetic interpretations of the parental magma formation (such as degrees of partial melting, melt fractionation, mixing, and/or assimilation) extremely model dependent and, in effect, quite unreliable.

No nonzeolitized parts of the tuff have been found in the area. Therefore, performing mass balance calculations of the vitric precursor composition (parental magmatic source) and the composition of the tuff after zeolitization is not possible. However, considering the physicochemical homogeneity of the zeolite deposit and the lack of evidence for major alteration, shown by the whole-rock analyses as well as the mineral chemical compositions, it seems likely that the zeolitization process was essentially isochemical. The analyzed samples have major element compositions comparable to other highly evolved mid-Miocene magmas in the Carpathian-Pannonian region (Fig. 5C; Harangi et al., 2007). Even with small compositional differences between the parental volcanic glass and the subsequent clinoptilolite phase, no major effects would be expected concerning the whole-rock composition (Ogihara, 2000).

The homogeneous nature of the Hrabovec tuff in the studied mine is particularly striking when it is compared to some other natural zeolite occurrences. For example, Karakaya et al. (2015) documented an occurrence in Turkey that has a mineral paragenesis and formation history similar to those observed in this study. However, the clinoptilolite phase that occurs there is in spatially very restricted (small-scale) layers that are intercalated with precursor rocks affected by strongly varying degrees of alteration and thus other natural zeolite and authigenic minerals.

The only zeolite mineral detected in the studied tuff is clinoptilolite; although isostructural with heulandite, XRD peak intensity patterns, as well as EPMA measurements that consistently give Si/Al values ≥4, clearly indicate the presence of clinoptilolite (Gottardi and Galli, 1985; Bish and Boak, 2001; Figs. 4, 9; see Tables 2, 3). XRD evidence for other zeolite minerals, such as analcime, mordenite, or erionite, has not been found, although it is reported from other tuff outcrops in the Nižný Hrabovec area (Šamajová, 1997). The minor and nonsystematic variation in whole-rock compositions and the uniform whole-rock mineralogical composition as well as mineral chemistry of the deposit indicate that stable and homogeneous physicochemical conditions operated during the zeolitization process.

Generally, zeolites originate from volcanic or aluminosilicate-bearing materials undergoing physicochemical reactions, including dissolution-precipitation processes, with aluminum- and silica-rich volcanic glasses being the common precursor of the zeolite minerals (Hay and Sheppard, 2001). Changes in the ambient temperature and pressure conditions, fluid composition, and fluid-flow rate directly influence the geothermal as well as the geochemical gradient and consequently the reaction products (Iijima, 1980; Hay and Sheppard, 2001). The combination of the outlined physical and chemical parameters, the mineralogical composition of the precursor material, and the time frame of the ongoing reaction decides which zeolite minerals form. Clinoptilolite crystallization commonly occurs when volcanic glass from high-silica tuffaceous rocks is replaced during postdepositional diagenesis through devitrification reactions (Ogihara, 2000).

The studies of both Šamajová (1979) and Chmielewská(2014), as well as the present investigation, have shown that clinoptilolite formation within the Hrabovec tuff resulted from alteration of a massive, homogeneous rhyolitic volcanogenic sediment in the middle Miocene. Stable ambient physicochemical conditions and involvement of an isochemical fluid during zeolitization are suggested by the striking homogeneity of the clinoptilolite tuff deposit.

EPMA and element maps show that the minerals from the original rock paragenesis (K-feldspar, plagioclase, biotite, and quartz) are uniformly distributed within the matrix and have consistent compositional ranges. They do not show any growth zonation and lack any signs of postmagmatic alteration. The authigenic clinoptilolite minerals also have relatively uniform compositions (Fig. 9), suggesting a geochemically very homogeneous glass precursor.

For the coarse-grained clinoptilolite, the formation process included dissolution of the volcanic glass by pore fluids under ambient pressure-temperature conditions, followed by precipitation of tabular clinoptilolite crystals up to ~30μm. The replacement of glass shards and growth into open pore spaces by clinoptilolite is illustrated in Figure 8B through D. Element maps of Si, Al, Ca, K, Na, and Fe document the uniform composition of the newly formed crystals (Fig. 10). Similarly, the matrix crystallites of clinoptilolite (Fig. 8) formed through dissolution-precipitation reactions.

Analogue models of clinoptilolite genesis and occurrences were described by Iijima (2001) and Utada (2001). These models suggest that clinoptilolite formed in a diagenetic environment, subsequent to burial to several hundred meters depth and ambient temperatures between 44° and 84°C. As no volcanic glass remnants have been detected in the studied tuff, we conclude that the reaction was fast enough to completely replace the volcanic glass in the precursor tuff. However, both the burial rate and the geothermal gradient were too low in the sampled section to generate other zeolite minerals such as analcime, laumontite, or heulandite, which would, particularly in the absence of clinoptilolite, indicate deeper burial conditions (Utada, 2001). Apart from the chemical composition of the vitric precursor, the development of a specific natural zeolite mineral is mainly dependent on the ambient pressure and temperature conditions, combined with the time available for crystal growth (Eyde and Holmes, 2006). Although experimental and field data show slightly different trends, temperature seems to be the most important parameter concerning the stability fields of natural zeolites. For example, at temperatures <44°C, no zeolites form, whereas at temperatures >84°C, the stability field of clinoptilolite is exceeded, with analcime, heulandite, and laumontite stable up to 123°C. Above 123°C, laumontite is the only stable zeolite (Utada, 2001).

Deep borehole data from an area south of Nižný Hrabovec, however, reveal a vertical zonation from clinoptilolite stability in the shallower parts to analcime occurrence in deeper (central) parts of the basin (Šamajová and Kuzvart, 1992; Šamajová, 1997). As the Hrabovec tuff layer in the central basin is nearly 100 m thick but buried to greater depth, the increasing temperature with depth is reflected by the presence of analcime rather than clinoptilolite, typical for large zeolite deposits formed through burial diagenesis of volcanogenic sediments (Šamajová, 1997; Hay and Sheppard, 2001).

Although the XRD scans indicate that cristobalite is a minor component of the rock (Fig. 4), only traces of quartz have been observed directly in thin sections as a crystalline silica phase. However, the Si element mapping shows abundant very fine grained well-dispersed silica micrograins within the tuff matrix (Fig. 10B). Subtracting the Al signal from Si signal results in an intensified attenuation of the clinoptilolite signal, enhancing the silica signal (Fig. 10H). This silica is most likely a metastable cristobalite precipitate, forming a microcement phase within the matrix (Fig. 10B, H), as well as around larger zeolitized glass shards, derived from the excess SiO₂ produced during the clinoptilolite-forming reaction (Broxton et al., 1987; Chipera and Apps, 2001; Snellings et al., 2008). Snellings et al. (2008) suggested that the close association of cristobalite and clinoptilolite indicates a closed system, in which excess SiO₂ is not removed from the system. The slightly higher Si/Al ratios in the matrix (mean Si/Al 5.2, σ=20) compared to the coarse clinoptilolite (mean Si/Al 4.87, σ = 0.14) may also be reflecting the presence of a fine-grained SiO₂ phase within the matrix (Fig. 9A, B). Whether this SiO₂ phase is only low cristobalite is difficult to determine, taking the analyzed multiphase system and the general analytical problems of distinguishing the SiO₂ polymorphs into account. The microsilica found could, at least in part, consist of opal C crystallites, which are a disordered and/or a transitional form of cristobalite microcrystallites that replaced amorphous silica (Elzea et al., 1994; Elzea and Rice, 1996). However, this crystallographic distinction has no implications for the formation of the Nižný Hrabovec tuff. Note that such microsilica will most likely be present in all clinoptilolite deposits of volcanogenic-sedimentary origin, except those in which a significant fluid flow has occurred during and/or after the clinoptilolite-forming reaction, removing it in solution. In such an open system, the formation of a homogeneous, predominantly monomineralic and therefore economically relevant deposit is improbable. No evidence for such flow, such as quartz veins, has been seen in the Hrabovec tuff at the mine studied.

Few workers have investigated radiogenic isotopes in zeolite deposits; the available studies, however, mainly focused on age dating and are restricted to a very small number of isotopic systems (e.g., Karlsson, 2001). Gundogdu et al. (1989) tried to decipher zeolite genesis using ⁸⁷Sr/⁸⁶Sr isotope ratios, suggesting that the observed limited range in Sr isotope values was linked to zeolite minerals evolving in a closed system with a volcanogenic Sr source. This is consistent with our interpretation of clinoptilolite formation at Nižný Hrabovec, as the ranges of radiogenic Sr and Nd isotope ratios in the tuff are minimal, suggesting a homogeneous volcanic source.

The essentially homogeneous chemical and mineralogical composition, in combination with the lack of clastic or carbonate layers within the tuff, implies that a large, well-equilibrated magma chamber erupted, producing a significant volume of homogeneous volcanic ash in a very short time period—most likely in one major eruptive event.

Geochemically, the analyzed tuff is comparable to the geographically and tectonically neighboring Miocene to Pleistocene calc-alkaline rock series of the central segment of the Carpathian-Pannonian region and East Carpathians, respectively (Figs. 5C, 7). These continental arc-related rocks are interpreted as being linked to regional lithospheric (back-arc) extension and subduction processes (Mason et al., 1996; Seghedi et al., 2004; Harangi et al., 2007). The exact volcanic source of the Hrabovec tuff is unknown. In the area west of Nižný Hrabovec in the Slanské vrchy mountains, as well as to the east in the Vihorlatské vrchy mountains, intense Miocene volcanic activity leading to diverse stratovolcanic facies was described by Žec et al. (2002). A middle Miocene rhyolite complex with volcanic centers was also described ~50 km south of Nižný Hrabovec, close to Viničky (Fig. 1; Lexa et al., 2014). However, no compositional data for these volcanogenic complexes are available. Pécskay et al. (2002) linked the Hrabovec tuff to calc-alkaline volcanic activity in the region of northeastern Hungary to northwestern Romania.

The conclusions of this study are as follows:

1. A maximum CIA value of 57 % (from an unaltered value of 52 %) indicates that very minor postdepositional geochemical changes (K metasomatism) occurred in some parts of the lower Badenian (16.30–15.03 Ma; ≅ Langhian stage, Miocene) 100-m-thick Hrabovec tuff (eastern Slovak Republic). No systematic variations in tuff compositions are seen within the investigated Nižný Hrabovec open-pit mine, mirroring the striking homogeneous nature of this clinoptilolite tuff.

2. The tuff has a rhyolitic composition and, to the largest part, a high-K calc-alkaline affinity, similar to Miocene to Pleistocene calc-alkaline continental arc-related rocks from the Carpathian-Pannonian region, which are interpreted as related to regional lithospheric (back-arc) extension and influence subduction processes.

3. REE compositions are similar to those of the upper continental crust, with chondrite-normalized LREE enrichment and negative Eu anomalies between 0.42 and 0.6. Chondrite-normalized spider diagrams show negative anomalies for Nb, Ta, and Ti and weak positive Sr anomalies, consistent with geochemical signatures of active continental margins.

4. Mean ⁸⁷Sr/⁸⁶Sr and ¹⁴³Nd/¹⁴⁴Nd ratios are 0.70880 and 0.512463, respectively. These reflect an enriched magma source, with a distinct crustal contribution.

5. The essentially homogeneous (geochemical and mineralogical) whole-rock and radiogenic isotope compositions of the clinoptilolite tuff and the lack of clastic or carbonate interlayers in the tuff indicate that they were deposited in a marine environment over a very short time interval, likely in one large eruptive event.

6. XRD and microprobe analyses show that clinoptilolite is the only zeolite present in the tuff.

7. EPMA and SEM observations show that clinoptilolite formed idiomorphically by replacing glass shards in situ and/or growing into open pore spaces and by replacing extremely fine grained ash particles in the tuff matrix.

8. Finely dispersed matrix microsilica is likely authigenic cristobalite, which is the excess Si released during the zeolite formation.

The authors thank Glock Health, Science and Research GmbH for covering travel and analytical costs; Peter Kováč (Zeocem a.s.) for supporting our field work and coordinating the blasting program around it; Franz Kiraly and Theodoros Ntaflos (University of Vienna) for assisting with EPMA and specimen preparation; Joachim Fendrych and Dietmar Nagl (Glock Health, Science and Research GmbH) for helping in the lab and for providing SEM pictures; and Michael Bizimis (University of South Carolina) and Wencke Wegner (University of Vienna) for helping with radiogenic isotopic data interpretation. The comments of Ralf Halama, an anonymous referee, associate editor Paul Spry, and Editor Larry Meinert substantially improved the manuscript.

Cornelius Tschegg is operations manager and chief geologist at Glock Health, Science and Research GmbH (Austria), which he joined in 2012. He received his master’s and doctoral degrees from the University of Vienna in 2005 and 2008, respectively. During his studies and as a postdoc, Cornelius specialized on the multianalytical investigation of diverse geomaterials, ranging from ultramafic to calc-alkaline volcanic rocks, from unconsolidated clastic sediments to metamorphic, highly deformed ultramylonites, and from altered carbonate rocks to ancient ceramics. His main interest is combining whole-rock and mineral geochemical methods with high-resolution imaging techniques, allowing a broad range of research tasks to be tackled.