Physico-mechanical properties of halite and gypsum–mudstone crushed rock backfills of the Mercia Mudstone Group Open Access

It is well known that the Mercia Mudstone is highly heterolithic, with significant and often abrupt variations in facies depending on the location and depth (e.g. Howard et al. 2008). Whilst the specific facies expected at a DGR site are not known, we targeted our sampling locations to capture a range of DGR-relevant lithologies. This included an evaporite (halite) and a mixed evaporite–fine grained clastic rock (gypsum–mudstone).

Halite was obtained from the Winsford mine in Cheshire, UK. The halite was mined from the Bottom Bed (formerly known as Hundred-foot salt) of the Northwich Member of the Mercia Mudstone Group. The Northwich Member is a Mid Triassic (Anisian) mudstone of continental/lacustrine origin containing interbeds of clastic and evaporitic strata formed from flash flooding, wind-blown sediments and evaporation (Howard et al. 2008). The samples provided comprised pink to clear halite aggregate that had been cut by a continuous miner machine and passed through a sizing crusher to reduce the size of larger fragments to assist materials management below ground (C. Hilton, Compass Minerals pers. comm. 2023) (Fig. 1a). The only additions were small volumes of water aerosol for dust suppression and cooling, which would be comparable to requirements during the construction of a DGR in rock salt. Samples were transported via courier to the University of Plymouth, where they were stored in polythene bags to maintain natural moisture content.

Gypsum and mudstone samples were collected from a coastal outcrop of the Upper Triassic Mercia Mudstone located in the Bristol Channel near Watchet, Somerset, UK (Whittaker and Green 1983). The predominantly red mudstone cliffs are intersected by numerous centimetre-thick, occasionally anastomosing, gypsum veins, decimetre-thick nodules with some massive gypsum bands and grey-green siltstones (see Philipp 2008 for a detailed analysis). The mudstone sampled was of weak–medium strength, red-brown, with occasional grey-green patches and poorly bedded. Watchet was selected for sampling in this study as it presented a fresh (actively eroding) cliff exposure, where fine-grained mudstone-dominated rocks and gypsum veins were readily accessible. Samples were obtained using a geological hammer, sealed in sample bags to minimize moisture loss and stored at room temperature until experimental preparation. Weather conditions were dry during sample collection (Fig. 1b).

Upon receipt, the crushed rock was characterized to determine physical properties (particle size distribution and natural moisture content), as well as elemental and structural properties using a combination of X-ray fluorescence (XRF), and optical and scanning electron microscopy.

Elemental analysis was undertaken using a NITON Xl3t 950 pXRF in bench-top mode, samples were powdered to pass through a 75 µm sieve and placed within polyethylene cells with a 3.5 µm mylar film base.

Scanning electron microscopy coupled with energy-dispersive spectroscopy (SEM-EDS) was utilized to investigate the microstructure and elemental composition. SEM-EDS analysis was undertaken on the Zeiss Sigma 300 LV FE SEM at the Plymouth Electron Microscopy Centre, University of Plymouth. The elemental composition was determined using two Oxford Instruments Ultimax 65 mm² detectors, and data were processed using Oxford Instruments’ Aztec software. Halite was polished using an oil-based polishing solution. Halite analyses were completed following beam stabilization on Cu tape, at ×1000 magnification. Calibrations were within an acceptable 2% margin. Analyses of the halite samples themselves were undertaken at 20 kV accelerating voltage and a working distance of 8.5 mm. Scratches (3 µm) remained in the halite after polishing with 1 µm solution and were attributed to the soft nature of the material. Inclusions within the halite did not show the 3 µm scratches.

The bulk halite material received from Winsford showed a wide distribution of particle sizes ranging from less than 0.6 mm up to 63 mm, as determined by sieving analysis according to BS 1377-2:2022. Figure 2 sets out the particle size distribution (PSD) for the bulk ‘as-received’ material from the mine, indicating a likely PSD of GDF rock backfills should typical mining methods be deployed.

As-received PSD required reducing for laboratory-scale experimental purposes, in accordance with BS 1377-7:1990, such that further crushing was undertaken. Particles exceeding 6.7 mm in diameter were incrementally crushed by cast iron pestle and mortar. The resulting particles were sieved after each round of crushing using woven stainless steel test sieves of 6.7 mm mesh size. This iterative crushing and sieving process was repeated until the entire sample passed through a 6.7 mm sieve. The gypsum and mudstone samples underwent similar crushing using the same equipment, followed by sieving through a 6.7 mm mesh size to obtain uniform particles. The processed halite, gypsum and mudstone samples were further manually sieved using test sieve mesh sizes of 0.63 and 2.0 mm. The sieved fractions were stored in sealed containers at room temperature.

Unconfined compressive strength (UCS) testing required the production of cylindrical specimens from the granular stock, formed within a steel compaction vessel. The compaction procedure involved combining two grain size categories of crushed rock (60% with a grain size of 0.63–2 mm and 40% with a grain size of <0.63 mm), compaction of the mixture within a cylindrical steel tube 38 mm in diameter and 80 mm in length, and subsequent extraction of the specimens for analysis using a hydraulic press. Halite was compacted to target densities of 1.72, 1.73, 1.78 and 1.83 g cm⁻³, with a ±0.05 g cm⁻³ tolerance. The compaction pressure required to achieve each target density was determined experimentally by incrementally increasing the pressure on trial specimens. The final dimensions and mass of each compacted specimen were measured using callipers and an analytical balance, the density was calculated and recorded. Specimens were stored in airtight containers at room temperature until required for mechanical testing.

Samples for direct shear testing of granular halite and mudstone–gypsum mixtures had a maximum particle size of 2 mm (60% with a grain size of 0.63–2 mm and 40% with a grain size of <0.63 mm). Pure halite samples were tested to reflect a DGR hosted in a massive halite deposit, whereas the mudstone–gypsum mixtures reflect a DGR in mudstones that typically contain variable quantities of gypsum veins. Mudstone and gypsum were mixed at ratios of 100:0, 80:20, 60:40, 40:60, 20:80 and 0:100 by mass to investigate how composition impacted backfill strength. Halite was tested without blending. The samples were homogenized by hand and then placed in a shear box with 60 kg load for 24 h to achieve compaction prior to testing. One replicate sample was prepared for each mixture ratio using the same compaction procedure described above, with a target density of 1.6 g cm⁻³ for the gypsum–mudstone mixtures and 1.5 g cm⁻³ for halite.

UCS testing of cylindrical compacted-granular halite specimens, c. 37.8 mm in diameter, was undertaken. The samples were trimmed to achieve flat surfaces on the top and bottom, resulting in heights between 70 and 77 mm, thus maintaining a 2:1 height to diameter ratio as specified by BS 1377-7:1990. UCS testing was undertaken on a load-frame at Structural Soils Ltd, Bristol, UK.

Shear strength testing was conducted using direct shear apparatus to evaluate the mechanical behaviour and shear resistance of halite and composite gypsum–mudstone samples under varied vertical loads and mixture compositions. The test procedures involved sample preparation, consolidation under vertical loads, the application of horizontal shear displacement and measurement of shear force response. Shear strength testing was conducted on halite samples prepared at a compaction density of 1.50 g cm⁻³ using the 60 × 60 mm direct shear apparatus at the University of Plymouth, UK. This density was selected to evaluate the shear strength response at a lower compaction density than explored by the UCS, using the same crushed halite mixture as the previous tests. Direct shear testing was also conducted on compacted gypsum–mudstone mixtures to determine composition-dependent shear strength properties. The tested sample mixtures were prepared at compositions ranging from 100% gypsum to 100% mudstone in 20% incremental steps, as previously set out. All samples were compacted to a consistent density of 1.60 g cm⁻³. Two replicate samples were compacted and tested for each mixture ratio. The direct shear behaviour was evaluated under 19.95, 33.57, 47.20 and 60.82 kPa normal stresses.

Elemental analyses are presented in Table 1 for the halite, mudstone and gypsum. While XRF cannot resolve detecting lighter elements such as Na, the Cl content is consistent with the expected NaCl. Additional petrographical and SEM analysis verified the primary halite mineralogy and microstructure. Ca, S, Si and Fe were also present as minor constituents.

Gypsum analyses showed major Ca (200 000 mg kg⁻¹) and S (130 000 mg kg⁻¹) concentrations matching its elemental composition (CaSO₄·2H₂O). Si was detected with a concentration of 3200–4000 mg kg⁻¹. Minor amounts of K, Sr, Fe, Cl and Ti of 75–900 mg kg⁻¹ were also present.

Analysis of mudstone samples identified high Si, Al, K, Ca and Fe concentrations ranging from 20 000 to more than 100 000 mg kg⁻¹, reflecting an aluminosilicate mineral content. Intermediate Ti concentrations of c. 4000 mg kg⁻¹ and c. 1000 mg kg⁻¹ of Cl were also measured. The mudstone had higher levels of Rb, Ba and transition metals (Pb, Zn, Mn, Cr and V) compared to the evaporite rocks.

The average halite moisture content of 0.95% is consistent with halite's anhydrous mineral structure. The gypsum had an elevated moisture content of 17.60% attributable to its hydrated mineral structure containing interstitial water molecules (CaSO₄·2H₂O). The measured moisture content of 17.60% compares reasonably well with the calculated 20.97% water expected based on gypsum's dihydrate chemical structure. In contrast, the mudstone had a low moisture content of 1.92%. The water content may have been slightly affected by storage in sealed polythene bags, which could have contained moist air.

Analysis of the mudstone thin section under a petrographic microscope revealed a very fine-grained clay-rich matrix with sparse larger inclusions up to c. 40–150 $μ$m (Fig. 3a). The matrix exhibited a preferred clay mineral orientation that is likely to be related to depositional or diagenetic processes. Accessory grains included sub-angular to subrounded quartz, plagioclase feldspar and calcite in the mudstone fabric based on birefringence patterns. Minor gypsum mineral veins c. 50–80 $μ$m thick were observed cross-cutting the matrix (Fig. 3b), potentially associated with hydration-induced hydrofracturing and fluid flow (Philipp 2008). In halite samples (Fig. 3c), cleavage planes were visible in some grains, and both single crystals and polycrystalline aggregates were observed. No definitive accessory mineral phases were identified. The halite grains showed typical anisotropic birefringence colours and extinction under crossed polars. The gypsum thin section showed a coarse-grained equigranular texture composed predominantly of tabular to elongated gypsum crystals 80–200 $μ$m in length (Fig. 3d). These grains exhibited prominent cleavage and rare bending. Flower-like gypsum crystal aggregates up to 200 $μ$m wide were also observed. Tiny (<10 $μ$m) opaque inclusions were occasionally noted. The gypsum grains were largely isotropic with very rare occurrences of low birefringence silicate accessories of the order of c. 5–15 $μ$m. These fine accessories did not show a preferred orientation. There was no evidence of chemical weathering or dissolution features.

SEM imagery is presented in Figure 4, with further analysis by SEM-EDS presented in the Supplementary material and summarized below.

Halite analysed by SEM-EDS confirmed semi-quantitatively NaCl in the pink (97.7%) and clear (95.5%) materials. Minor constituents of Ca–S–O and Si–O are below detection but may correspond to impurities of anhydrite and quartz (see Figs S1 and S2 and Table S1in the Supplementary material).

The SEM-EDS analysis (Fig. S3 in the Supplementary material) showed that the gypsum is compositionally dominated by O, Ca and S peaks, corresponding with the CaSO₄·2H₂O elemental analysis observed in the XRF analysis of the bulk sample. Minor inclusions of the Sr–S–O- bearing celestine (SrSO₄) are present within the gypsum (see Fig. S4 and Table S2, Spectrum 1 in the Supplementary material). Further analysis of inclusions revealed silicate phases compositionally like quartz and/or feldspars, which are likely to be present as accessory mineral grains (see Figs S5 and S6 in the Supplementary material).

The SEM-EDS analysis of mudstone (Figs S7 and S8 in the Supplementary material) revealed a heterogeneous mixture of mineral phases, consistent with the fine-grained mudstone composition. Further analysis revealed the presence of zircon (see Table S3, Spectrum 9 in the Supplementary material, which contains impurities of Na, Al, Ca, Sc, Fe, Yb, and Hf that fall below the detection limit), quartz (see Table S3, Spectrum 10 in the Supplementary material), and phases compositionally consistent with calcite and/or dolomite. In addition, signals corresponding to Al, Si, O, K and Na are likely to correspond to feldspars or clay mineral phases. Iron oxide and/or oxyhydroxide phases were indicated by Fe and O signals (for further details see Figs S9–S15 and Table S3 in the Supplementary material).

The UCS testing provides insights into the relationship between compaction density and mechanical strength for crushed salt backfill. Table 2 and Figures 5–8 summarize the UCS test results.

The UCS test results demonstrate that higher compaction densities yield substantially increased strength in the crushed halite specimens. The lowest density sample tested at 1.72 g cm⁻³ exhibited strain-hardening-type behaviour, ultimately failing at just 220 kPa UCS (Fig. 5). However, increasing the compaction density to 1.73 g cm⁻³ significantly boosted the UCS to 3500 kPa (Fig. 6). This sharp change in strength and strain properties between a compaction density of 1.72 and 1.73 g cm⁻³ is indicative of a step change in cohesive forces between particles; however, further testing and post-mortem material analysis would be required to confirm this. Further elevating the compaction density to 1.78 g cm⁻³ enhanced the UCS to 5700 kPa (Fig. 7), showing the considerable strength gains achievable through controlled densification. The maximum achieved compaction density of 1.83 g cm⁻³ produced the highest UCS of 6600 kPa (Fig. 8), confirming that maximizing the density yields the strongest crushed salt backfill specimen.

These experimental trends of strength increasing with density agree with past studies by Bechthold et al. (2004), which showed similar strength gains in compacted crushed salt backfills. Unlike the low-density sample, the higher-density specimens failed by shear fracturing at ductile strains above 1%, demonstrating their superior structural integrity compared to the brittle failure of the 1.72 g cm⁻³ sample. Nonetheless, further repeat testing would be required to gain a more robust understanding of the relationship between compaction density and UCS in order to inform backfill design.

In the post-closure phase of a DGR in a ductile evaporite or mudstone, the host rock and overburden have the potential to creep: for instance, in the proximity of unlined vaults, and where concrete tunnel linings eventually degrade and fail. This creep will apply progressive loading to emplaced vault and tunnel backfill that will naturally undergo additional compaction, resulting in increased density and, consequently, enhanced strength. For example, experimental observations at a depth of 800 m in the Asse mine (Zechstein series salt dome, Germany) saw backfill stresses reaching 2–4 MPa over 8 years after emplacement in unlined drifts (Blanco-Martín et al. 2016).

Samples were subjected to progressive normal stresses of 19.95, 33.57, 47.20 and 60.82 kPa. The mean peak shear stresses at the applied normal stresses are presented in Figure 9.

Analysis of the results for Test 1 gave a friction angle of 39.49° and cohesion of 0.86 kPa. Test 2 yielded a very similar friction angle of 39.70° and cohesion of 0.92 kPa. Regression analysis of these data provided r² values of 0.88 and 0.93 with a linear fit, with a significantly better fit achievable with a curve (the simple polynomial used ‘y = 0.018 × 2–0.6x + 25.6’ yields an r² value of >0.99), indicating a non-linear relationship between normal stress and peak shear stress in the crushed halite. Possible explanations for this behaviour could include cohesion of halite particles during the initial compaction phase or effects of interparticle friction (Massoudi and Mehrabadi 2001); however, further investigation is required to confirm this and to determine the mechanistic relationship between backfill properties and compaction-related strength gain. It was also observed that the samples showed brittle post-peak strain softening in the shear stress v. displacement response (see Fig. S16 in the Supplementary material). The angle of internal friction of crushed halite (39°–40°) measured in this work sits slightly below the range of 43°–53° reported by Bechthold et al. (2004) on crushed rock salt from the Asse mine in Germany, potentially reflecting the low compaction density in this study.

Pure gypsum showed the highest peak shear resistance, while pure mudstone had the lowest. In between, the declining peak shear stress with reducing gypsum percentage is illustrated in Figure 10.

Friction angles declined linearly from 41.30° to 28.98° with reducing gypsum percentage (Table 3). In contrast, cohesion rose from 0.75 to 6.67 kPa as the mudstone fraction increased (Table 4). This indicates that a higher gypsum content provides greater frictional shearing resistance, while a higher mudstone content enhances cohesive strength.

While the gypsum–mudstone composite samples showed composition-dependent shear strength, no directly comparable literature studies on similar crushed rock mixtures were found for quantitative comparison. However, the experimental results did agree qualitatively that phyllosilicate-rich mudstones exhibit higher cohesive strength, while gypsum provides greater frictional resistance (Bowles 1996).

This study explores the effect of compaction density and material composition on the strength behaviour of crushed rock backfill mixtures being considered for deep geological repositories (DGRs) for nuclear waste. Increased materials strength was achieved through controlled compaction and by blending materials; however, the nature of the strength gain varies depending on the materials tested.

UCS testing of compacted halite at a density of 1.83 g cm⁻³ resulted in a UCS of 6600 kPa, more than an order of magnitude higher than a specimen compacted to 1.72 g cm⁻³. Direct shear testing of granular halite at a density of 1.5 g cm⁻³ determined an angle of internal friction of 39°–40°, reflecting a relatively high level of friction between particles at this low density. Increasing normal stress resulted in a non-linear, apparently exponential, increase in peak shear strength for the halite. Following DGR closure, stress redistribution and creep of the host rock will further compact backfill to even higher densities, and thus shear strengths, exceeding those tested experimentally.

Data presented for the gypsum–mudstone mixtures provided new insights into the potential to engineer shear strength by blending rock types. Direct shear testing of composites from 100% gypsum to 100% mudstone revealed that mixtures rich in gypsum provide greater frictional shearing resistance, while mixtures richer in mudstone promote more cohesive strength. However, higher normal stresses yield greater peak shear stress in more gypsum-rich samples, despite lower levels of cohesion.

The data presented demonstrate the capacity to engineer targeted mechanical properties through controlled compaction density and the composition of crushed rock mixtures. The backfill density and strength will significantly influence underground engineering design choices for the DGR, including the geometry of subsurface structures, emplacement planning (e.g. backfill slope angles) and disposal system reinforcement requirements. The experimental density–strength trends determined in this research provide an important basis for modelling crushed rock barrier properties and behaviour over extended durations.

Further experimental and modelling work is required to reduce remaining uncertainties, particularly around the long-term consolidation behaviour and property evolution of crushed salt barriers over the extended time frames involved in deep geological disposal, including appropriate boundary conditions and the influence of scale.

The authors would like to thank Compass Minerals (Winsford) for the provision of crushed halite, and Structural Soils Ltd for testing support and access to facilities. Jenny Wiggins and Vannia Santos Durndell, University of Plymouth, are thanked for their technical support. In addition, the constructive review feedback has improved the quality of this paper, so thanks are extended to the reviewers. This research forms part of MSc research projects undertaken by C.C. Umeaghadi and T. Lucock.

CCU: conceptualization (equal), data curation (lead), formal analysis (equal), methodology (equal), writing – original draft (lead), writing – review & editing (supporting); TL: data curation (supporting), investigation (supporting), writing – review & editing (supporting); RH: methodology (equal), supervision (equal), writing – review & editing (equal); JB: investigation (equal), writing – review & editing (supporting); MTB: conceptualization (lead), supervision (lead), writing – review & editing (lead).

This work received no specific grant from any funding agency in the public, commercial, or not-for-profit sectors.

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

All data generated or analysed during this study are included in this published article (and, if present, its supplementary information files).