Rock fracturing by subglacial hydraulic jacking in basement rocks, eastern Sweden: the role of beam failure

ABSTRACT Dense networks of dilated fractures occur locally in the upper 5–15 m of bedrock in basement gneisses in eastern Sweden. Near Forsmark, pre-existing sub-horizontal fractures have been jacked open and filled with water-lain sediment, likely during the latest Weichselian glaciation. Despite extensive previous research, it is uncertain whether subglacial hydraulic jacking led to the generation of new fractures, in addition to reactivation of pre-existing ones. Re-analysis of historic photos from excavations near the Forsmark power plant indicates formation of two types of new fracture. Firstly, rock fragments were broken off the main fracture surfaces as existing fractures were jacked open. Secondly, fracture analysis shows that whilst few subvertical fractures occur above tight sub-horizontal fractures, a higher density of vertical fractures occurs above dilated sub-horizontal fractures, suggesting new formation. We apply a model of beam failure theory, borrowed from structural engineering, to constrain potential new fracture generation, using assumptions based on measured water pressure fluctuations from beneath the Greenland Ice Sheet. This modelling shows that beam failure is a plausible mechanism for the generation of new vertical fractures during a subglacial water fluctuation cycle under a range of realistic glaciological conditions. This implies that hydraulic jacking can result in further in situ disruption and brecciation of the shallow rock mass, decreasing the rock mass strength and increasing its hydraulic conductivity. Altogether, hydraulic jacking of existing fractures and the formation of new vertical fractures results in effective subglacial mechanical weathering of the shallow rock mass.


Introduction
The near-surface formation of fracture networks in basement rocks represents a form of mechanical weathering of the shallow rock mass. In non-glacial environments, this is recognised as a common phenomenon with implications for the evolution of the "critical zone", relevant for groundwater use, engineering properties, as well as the long-term weathering and erosion of basement terrains (Brantley et al. 2007;Holbrook et al. 2019;Collins et al. 2020). In subglacial environments, the formation or growth of rock fractures has been investigated in theoretical papers (e.g., Morland & Morris 1977;Iverson 1991;Hallet 1996;Zoet et al. 2013), and has been documented in sedimentary rocks (Phillips et al. 2013), but convincing observational evidence for the subglacial formation of new fractures in basement rocks is sparse.
In basement gneiss of eastern Sweden (Fig. 1), dense networks of dilated fractures are locally abundant in the upper 5-15 m of bedrock ( Fig. 2A). Excavations related to the construction of the Forsmark nuclear power plant in the 1970s, as well as later test-excavations, revealed sub-horizontal fractures, up to 170 m long in section, that are dilated and filled by 1-80 cm of laminated silt and sand (Stephansson & Ericsson 1975;Carlsson 1979;Leijon 2005;Forssberg et al. 2007). The dilation of subhorizontal fractures and their sediment fill suggest that hydraulic jacking by overpressured water occurred during Pleistocene glaciations. Studies on fracture coatings, such as chlorite, epidote and iron staining, indicate that most fractures were first formed during the Precambrian and Palaeozoic (Sandström et al. 2008;2009), so that Pleistocene hydraulic jacking largely involved the reactivation of pre-existing fractures. However, some vertical fractures appear free of coatings and may have formed during Pleistocene glaciation. Despite extensive studies on various aspects of bedrock fractures at Forsmark (e.g., Carlsson 1979;Follin et al. 2007;Forssberg et al. 2007;Martin 2007;Moon et al. 2020), it is not clear if some new (vertical) fractures in the shallow subsurface were generated during hydraulic jacking by subglacial meltwater.

Origin of shallow fractures in basement rocks
An increase in fracture density in the near surface is observed in many basement terrains composed of massive rock; associated subhorizontal fractures are termed sheeting joints (also: exfoliation joints). Their near-surface occurrence suggests a component of "unloading", combined with horizontal stresses (Jahns 1943;Twidale 1973;Ziegler et al. 2013); Martel (2017) showed that topographic curvature as well as ground water pressure facilitate the formation of sheeting joints. Sheeting joints are known to form near the Earth's surface during the first exhumation of massive rock masses, for instance in young granite domes (Sierra Nevada, USA, e.g., Martel 2017), or during development of deep glacial troughs (e.g., in the Swiss Alps; Ziegler et al. 2013). Shallow rock masses that have only experienced a single exhumation episode are dominated by subhorizontal sheeting joints with only few vertical connecting fractures. As an example, the Dharwar Craton in south India has been exhumed by weathering and erosion more or less continuously since the Cretaceous, driven by rift-shoulder uplift, as India separated from other Gondwana continents (e.g., Gunnell 1998). Sheeting joints in felsic orthogneiss near Mysore (Fig. 2B) are well developed down to c. 12 m, are tens of metres long, and form a broadly anastomosing network, with very few connecting vertical fractures. The rock slabs delineated by the joints have a length: height ratio of >10:1.
In eastern Sweden, the first exhumation of the basement rocks occurred during the Mesoproterozoic, followed by several cycles of burial and re-exhumation (Stephens 2010;Japsen et al. 2016). These events did not result in deep erosion of the basement (Hall et al. 2019a, b), so that earlier formed shallow fracture networks were not destroyed by erosion but instead were re-activated. The shallow basement in Sweden has a more complex loading/unloading history than first-exhumation rock masses, and the resulting fracture networks are more complex (compare Fig. 2A and B). This implies that vertical, near-surface, fractures in eastern Sweden may be younger, secondary features, not primarily related to exhumationrelated sheeting jointing, and possibly generated subglacially below the latest Weichselian ice sheet.

Setting and previous work
Most of eastern Sweden consists of crystalline basement, comprising various felsic gneisses with subordinate amphibolite, formed between 1.9 and 1.8 Ga (e.g., Stephens 2010). The rocks cooled and were partly exhumed soon thereafter, after which all deformation occurred in the brittle regime (Sandström et al. 2008). In east-central Sweden, the first near-surface exhumation of the basement occurred in the Mesoproterozoic, prior to burial by "Jotnian sandstone". In the early Neoproterozoic and Palaeozoic at least two further cycles of loading and unloading occurred during deposition and removal of Sveconorwegian and Caledonian foreland basins; near surface exhumation occurred prior to burial by Cambro-Ordovician sediments (Stephens 2010;Japsen et al. 2016;Hall et al. 2019b). The basement along the western shore of the Baltic Sea was likely then reexposed only during the Pleistocene (Hall et al. 2019a;Hall & Van Boeckel 2020). Further repeated loading and unloading occurred beneath 1-3 km thick Pleistocene ice sheets (Carlsson & Olsson 1982a;Svensk Kärnbränslehantering 2010). A dominant horizontal NW-SE compressive stress system affects the rock mass in eastern Sweden; close to the surface these horizontal stresses exceed the vertical stresses (Carlsson & Olsson 1982a, b;Martin 2007).
Fractures are commonly coated with minerals such as chlorite, epidote, corrensite, quartz, calcite and hematite. Detailed analysis has shown that these were formed during at least three phases, the first two during Proterozoic, and the last during the Palaeozoic; most fractures were thus formed well before Quaternary glaciation (Sandström et al. 2008;2009;Sandström & Tullborg 2009). Fractures were re-opened  to depths of 500 m under the presently prevailing stress regime (Moon et al. 2020). Carlsson (1979) provided detailed counts on the number of coated vs uncoated fractures for different joint sets: most fractures (75-90%) are coated with chloriteand thus are old features; however, up to 25% of subvertical E-W trending joints in the power plant excavations at Forsmark were uncoated (Carlsson 1979). Deeper in bedrock, only very few (<5%) of fractures are uncoated (Claesson Liljedahl et al. 2011). Other temporary excavations in the Forsmark area also record uncoated vertical fractures (Hermanson et al. 2003;Leijon 2005;Lagerbäck et al. 2005 (Carlsson 1979;Leijon 2005). Sediment fill in bedrock fractures consists mainly of silt, with minor sand and diamicton, and some rock fragments (Carlsson 1979;Leijon 2005). The silt-sand fills show locally laminations indicating deposition by flowing water (Stephansson & Ericsson 1975;Carlsson 1979;Leijon 2005). Microfossil and pollen content of the silt sediment in the fractures is similar to that of nearby till deposits (Stephansson & Ericsson 1975;Robertsson 2004), consistent with sourcing of the sediment from till at the ice-bed interface. Compression tests on the sediment fill suggest it was subjected to pre-consolidated load of 3400-3900 kPa (Carlsson & Olsson 1976;Carlsson 1979), and compressed by considerable effective stress, requiring substantial ice thickness. Stephansson & Ericsson (1975) envisaged fracture dilation by freeze-thaw action during an interglacial, and subsequent sediment fill during ice cover. This model requires the dilated fractures to remain open for a long time, which is difficult to reconcile with the observation of sharp edges of differentially uplifted but abraded rock blocks (Fig. 3c) near Forsmark (Forssberg et al. 2007). In contrast, Pusch et al. (1990) envisaged a scenario wherein hydraulic jacking occurred proglacially, with overpressure created by ice retreating over a frozen bed, resulting in strong hydraulic gradient between the ice sheet and the foreland, leading to overpressure in the glacial foreland. This model relies on a permafrost layer functioning as an aquitard, to retain high water pressures in the glacial foreland. This model is at odds with the high preconsolidation loads as documented by Carlsson & Olsson (1976) and other various strands of evidence (Hall et al. 2020) that indicate that hydraulic jacking occurred subglacially, rather than proglacially.
Potential groundwater overpressure conditions in the bedrock below Forsmark have been studied and discussed extensively (e.g., Talbot 1990Talbot , 1999Talbot , 2014Hökmark et al. 2010;Lönnqvist & Hökmark 2013;Hökmark & Lönnqvist 2014). The normal hydraulic gradient of a sloping ice front (e.g., Boulton et al. 1993) may lead to overpressure in a marginal or proglacial setting, in particular if the water is "locked up" below a low conductivity layer, like permafrost. However, no strong water pressure fluctuations are expected from these mechanisms. In contrast, in the ablation zone of the western Greenland Ice Sheet pressure measurements in boreholes (with ice thickness between 150-800 m) have shown strongly fluctuating water pressures on a seasonal and daily basis at the icebed interface (Andrews et al. 2014;Claesson Liljedahl et al. 2016;Wright et al. 2016;Harper et al. 2019). Water pressures in boreholes varies from 80 to 110% of overburden pressure. In moulins, in regions that are well-connected to the bed, Andrews et al. (2014) measured water-pressure fluctuations between 60 and near 100% overburden pressure on a daily basis during the melt season. Hall et al. (2020) suggested that such repeated overpressure events operating beneath the latest Weichselian Fennoscandian ice sheet were responsible for the hydraulic jacking of subhorizontal fractures in eastern Sweden. In addition, they interpreted the jacking of subhorizontal fractures as the first step towards glacial ripping, an erosional process that involves jacking, disintegration and transport, ultimately resulting in boulder spreads that are widespread in lowland Sweden.
The aim of this paper is to test the hypotheses that: (i) some vertical fractures in the shallow basement of eastern Sweden may have been newly generated during and following hydraulic jacking of pre-existing subhorizontal fractures by fluctuating water pressures beneath the latest Pleistocene Fennoscandian Ice Sheet, and (ii) such newly formed vertical fractures contribute to the fracturing of the shallow rock mass and thus constitute an enhanced mode of subglacial mechanical weathering. We re-analyse historic photos from the Forsmark excavations in terms of fracture density for both subhorizontal and subvertical fractures. We then apply classic beam theory borrowed from structural engineering to test if new vertical fractures can be generated during subglacial water pressure fluctuations.

Methods
To test how representative the Forsmark subhorizontal fractures are for east Sweden, a number of quarries and sections were visited (Fig. 1) and described qualitatively. Historic photos from the canal and test excavations at Forsmark were kindly provided by Svensk Kärnbränslehantering AB (SKB). Many of these photos have an accurate scale (ranging pole). Observations on sediment fill, fracture distribution and brecciation were made from these photos. For further fracture analysis, the historic photos were imported in a GIS and georeferenced against an artificial, but accurately scaled 1 m grid. Fracture networks were digitised from these images, giving a quantified 2D dataset of a vertical section of particular fracture sets (full details of the method in Palamakumbura et al. 2020). Once the fracture networks are captured as lines in a GIS, various statistical operations can be applied to derive data on fracture density or spacing and fracture dip. Fracture density is reported as fracture length divided by surface area, in m −1 , or the number of fractures per unit length, also in m −1 , following Singhal and Gupta (2010). Fracture spacing is then simply the reciprocal of fracture density, in m. Fracture dilation is identified visually from the photograph and attributed to the corresponding digital fracture trace. For the modelling, we apply classic beam theory from structural engineering, and adapt this for subglacial conditions, explained in that section.

Variations in shallow fracture networks in eastern Sweden
In the various quarries studied across Uppsala and Gävleborg counties ( Fig. 1; Table 1), fracture networks are highly variable in character, fracture density and fracture aperture, and substantial changes can occur over short distances. As an example, in the Uppsala Krossen quarry in granitic gneiss, one section shows subhorizontal fractures with pronounced curvature, connected by steeper fractures (Fig. 4A). The curved subhorizontal fractures are akin to sheeting joints but do not follow the topographic surface (cf. Martel 2017), rather, they terminate against gently dipping fracture zones that separate the different sections (this may be regarded as "pseudotopography" delineating solid bedrock, which would have the same mechanical effect as true topography in the models of Martel (2017)). An adjacent section, in contrast, shows an orthogonal fracture pattern, with subvertical and subhorizontal fractures (Fig. 4B). A third section is dominated by a "crisscross" pattern of gently-dipping fractures. None of the fractures is discernibly dilated or jacked. In other quarries such as Alunda Krossen quarry (Fig. 4C), long (>10 m) subvertical fractures dominate, with shorter (<2 m) interconnecting shallow-dipping fractures. Again, no fracture is discernibly jacked.
In contrast, at the Svagberget quarry, SW of Iggesund (Fig. 4D), a more or less continuous sub-horizontal fracture at 4-6 m depth separates coherent bedrock beneath from a rock mass dominated by open fractures, with apertures ≥5-10 cm. The upper rock mass is strongly disrupted, even brecciated. Sediment was observed along the main subhorizontal fracture, although close examination was too risky, given the unstable nature of the quarry face.
Although this sample is non-systematic (systematic analysis of fracture distribution in basement terrain with extensive forest and soil cover is not possible), about a quarter of the visited quarries possess open, jacked fractures, close to rockhead, whilst the remainder has tight fractures (Table 1). However, long, dilated sub-horizontal fractures with thick sediment fills (as exemplified by Fig. 3A, B at Forsmark) are comparatively rare.

Additional observations on fracture fill and fracture networks from historic photos
Observations from the historic photos from Forsmark, in addition to those previously reported (Carlsson 1979;Leijon 2005;Forssberg et al. 2007), include the following: In one section (Fig. 5A), crackle and mosaic brecciation (terminology following Morrow 1982) occurs 1 m below the top-rock surface in a zone of closely spaced subhorizontal fractures. In this c. 0.5 m thick zone, the rock mass has clearly been dilated and disrupted to such a degree that blocks have lost coherence, leaving a loose mass of blocks 10-50 cm across, with the creation of substantial void space in between. The rock below appears not to be affected by jacking or dilation. The rock above still shows coherence, but has more vertical fractures than the rather massive rock below the zone. In another section (Fig. 5B), similar forms of brecciation affect the entire top 3-4 m of bedrock.
In a later test excavation ('Drill Site 5ʹ; Leijon 2005), a partly sediment-filled fracture shows angular rock fragments (1-20 cm long) that appear to be broken off from the roof of the fracture (Fig. 5C). The rock fragments are partly mixed with fine sediment. On the right (yellow arrows), the rock fragments appear little displaced or rotated, whilst to the left (blue arrow) some rock fragments are substantially rotated and displaced.
In another test excavation (AFM 001264; Forssberg et al. 2007), old vertical fractures, coated with iron oxide are clearly seen (1) on Figure 5D. A possible new fracture is visible on the right, although this could be machine damage during excavation (2). Brecciation into angular rock fragments occurs at two levels. At the upper level (3), brecciation is focused where subvertical and subhorizontal fractures interact, with new formation of short fractures. At the lower level, the subhorizontal fracture shows in detail (Fig. 5E) a sediment fill of silt and sand up to 15 cm thick. The sediment is laminated, but the laminations are disturbed and folded (5). The roof of the fracture is uneven, and the soft-sediment deformation was likely caused by uneven vertical compression during fracture closure, following a water-pressure drop. On the right (6), small (1-2 cm) rock fragments are mixed with sediment. However, some open fractures at the Forsmark excavation show very little brecciation ( Fig. 2A). Here a long (>20 m) fracture has opened by c. 50 cm without signs of brecciation and development of new vertical fractures above it; however, this fracture has a continuous fill of silt and sand.

Fracture analysis from historic photos
Fracture analysis from the historic Forsmark cooling water intake canal photos show a general increase in fracture density towards rock head (Fig. 6), as reported by Carlsson (1979). In our study, we calculated fracture density separately for subvertical and subhorizontal fractures. In two sections (SKB-003 and SKB-006; Fig. 6A, B), the subvertical and subhorizontal fracture densities increase in tandem, from <1-2 m −1 at the base of the section to 2-4 m −1 at the top of the sections, with the highest fracture density caused by a concentration of subhorizontal fractures, with numerous subvertical fractures, 1-3 m below rockhead. These two sections also show the most pronounced dilation along subhorizontal fractures (blue arrows). A quarry section MK-22 from the Svagberget quarry, SW of Iggesund (Fig. 6H), was also digitised and shows similar high densities for subvertical and subhorizontal fractures. The top 4-6 metres of the section is clearly dilated and the rock mass disrupted. Sediment was observed (from a distance -unstable quarry face) in the subhorizontal fracture that underlies this disrupted rock mass.
A nominal or proxy length:height ratio of resultant blocks or slabs (Fig. 7) was calculated by dividing horizontal and vertical fracture spacing (the reciprocal of the horizontal and vertical fracture density). This shows that for section MK22 and SKB-006 the length:height ratio remains below 5 for their entire thickness and that some sections show a marked decrease in length:height towards the top, whilst sections SKB-036 and SKB-057 -which show no dilation at all along subhorizontal fractures -show very long length:height ratios.
Altogether, it appears that an increase in subvertical fracture density in the top few metres of the rock mass is present above dilated subhorizontal fractures, but is absent if subhorizontal fractures are not dilated. This suggests that at least some vertical fractures were formed as a result of dilation of the underlying subhorizontal fractures.

Results -Modelling of beam failure
The modelling here assumes that hydraulic jacking and dilation of subhorizontal fractures were driven by repeated water pressure fluctuations with a daily to seasonal frequency as measured at the base of the Greenland Ice Sheet (Andrews et al. 2014;Claesson Liljedahl et al. 2016;Wright et al. 2016;Harper et al. 2019). Following Wright et al. (2016), relative water pressure is defined here as the fraction of overburden pressure P w /P i (where water pressure is P w , and cryostatic pressure is P i ). If the water pressure equals overburden pressure (P w /P i = 1), the ice is at (local) flotation, and the effective pressure (P w -P i ) is zero. Now consider a rectangular slab of rock at the base of an ice sheet, underlain by a transmissive subhorizontal joint (Fig. 8A), that is uplifted ("jacked") by overpressured groundwater, by a few centimetres (Fig. 8B). As the water pressure drops, the slab will be lowered again (Fig. 8C). Now consider that the slab will not be lowered back in its exact original place, but is "hanging", supported at its ends, perhaps by a broken rock fragment (circled in Fig. 8B, C), such as observed in some of the test excavations near Forsmark (Fig. 5C, D, E). The rock slab has now become a bridge or beam (Fig. 8C). This beam is subjected to a bending stress, caused by the load provided by its own weight plus any load of the overlying ice. As the tensile strength of rock is much lower than its compressive strength, when the stresses are too high, the slab or beam may fail in a tensional manner near its base (Fig. 8D). A new subvertical fracture is then formed.
This problem in its simplest form equates to the structural engineering problem of a beam supported at both ends under a uniform distributed load (Fig. 8E). The maximum stress σ max that such a beam is subjected to (compressive at the top and tensional at the base) is given by (EngineeringToolBox 2009): where h r is the height of the beam, L the length of the beam, q tot is the total load and I is the Area Moment of Inertia. The Area Moment of Inertia I for a rectangular crosssection is given by (EngineeringToolBox 2008): so that (taking the width of the beam, w = 1): The total load q tot is the sum of the load of the overlying rock and the ice, minus the water pressure at the base: where q r is the load by the rock beam, q i , the load of the ice, P w the water pressure, g the gravity constant, h i the thickness of the ice and ρ i and ρ r the density of ice and rock respectively. Since it is convenient to express this in terms of relative water pressure (P w /P i ), equation [4] is rewritten as (full derivation is given in the Appendix): Thus, the maximum stress at the base of a rock beam as a function of the height and length of the beam and ice thickness and relative water pressure is given by: The maximum length of the beam can then be given as: ffi ffi ffi ffi ffi ffi ffi ffi ffi ffi ½σ max p = 0:75 gh r ρ r þ gh i ρ i 1 À P w P i The tensile yield strength of the felsic granodioritic gneiss at Forsmark is in the order of 10-15 MPa (Glamheden et al. 2007). (In comparison, the compressive strength of the same rocks is in the range of 150-250 MPa). These values are for intact rock without micro fractures or other flaws, so that the tensile strength of 10-15 MPa should be seen as a maximum. In nature, tensile failure is thus likely somewhat lower. The results (Fig. 9A) show tensile stresses increasing exponentially with the length of the beam, for different scenarios (beam height h r = 1 and 5 m; ice thickness h i = 200, 400 and 600 m; relative water pressure P w /P i at 0.8 and 0.9). For instance (left hand line), a 1 m high beam, under 600 m of ice, with a relative water pressure of 0.8, will fail if longer than 3-4 m. Under a wide range of conditions, a 1 m high beam will fail if longer than 2 to 7 m. In contrast, a thick (5 m) beam, under relatively thin ice and high water pressure will not fail at less than 10 m length.
The maximum length for different parameters is plotted against beam height (Fig. 9B), for a fixed tensile strength of 10 MPa. This shows that beams (rock slabs) have the tendency to fracture into short, stubby blocks with a length: height ratio between 4:1 and 3:1. This compares reasonably well with the block length:height ratios as seen in the Forsmark sections, which show ratios >5:1 for sections without dilated fractures and <5:1 for section with (or above) dilated subhorizontal fractures (Fig. 8). Overall, the results show that tensile beam failure during low water pressure events (following jacking by high water pressure) is a plausible mechanism for the production of new vertical fractures in long rock slabs.

Discussion
The fracture analysis on the historic photos from the Forsmark excavations shows that:  This suggests that hydraulic jacking and fracture dilation along subhorizontal fractures locally resulted in the generation of new subvertical fractures. Observations on the historic photos show that the development of angular rock fragments in open sub-horizontal fractures at Forsmark is a common occurrence. This implies that after a high-pressure jacking event, a fracture may remain "jammed" open, leaving some void space. After a water pressure drop and increase in effective pressure, the rock slab above the fracture then functions mechanically like a beam. The modelling shows it is plausible that, if a rock slab above a dilated or jammed-open fracture functions as a beam, that beam may fail after a hydraulic jacking event as water pressure drops and effective pressure increases. New subvertical fractures and a dense shallow fracture network can thus be generated subglacially. Conversely, the modelling shows that it is not plausible for long, intact rock slabs to survive multiple water pressure fluctuations: they have a strong propensity to break up, under a wide range of realistic conditions. This conclusion is supported by the observation of brecciation associated with long subhorizontal factures near Forsmark and elsewhere. This implies that it is not plausible for long, open fractures to survive a glaciation (cf. Stephansson & Ericsson 1975), and it is more likely that jacking and further disruption occurred during the same glaciation, and likely in a relatively short time frame.
Nevertheless, one such long dilated fracture without additional fracturing or brecciation does occur at Forsmark, but this fracture is filled with 30-50 cm sediment. A solution to this apparent conundrum is that the sediment fill supported the load of the overlying rock slab, like a type of cushion, so that the overlying rock slab did not function like a beam, and thus preventing further break-up of this slab. This is supported by the high pre-consolidation loads measurements by Carlsson (1979). However, this load-bearing role of sediment fill would only be possible if sediment deposition occurred during an overpressure cycle. As such this is logic in that an overpressure event is likely to result in vigorous subglacial water flow that can transport sediment into the dilated fractures. Deposition would occur when water velocity drops after peak pressure (e.g., Phillips et al. 2013). The approximate coeval development of dilated fractures and sediment deposition is supported by the observation of mixed rock fragments and sediment fill (Fig. 6).
The generation of dilated, fracture networks in the shallow rock mass at Forsmark thus likely followed the following steps: (1) An overpressure event occurs at the ice-bed.
(2) Water under high-pressure is squeezed into a preexisting subhorizontal fracture. If water pressure exceeds overburden pressure, the fracture is jacked open, and the overlying rock slab is lifted upwards. (3) Locally, rock fragments may break off the roof or the sides of the dilated fracture; this keeps the fracture "jammed" open after water pressure drops; (4) Sediment, transported with the water, may be deposited in the dilated fracture, likely soon after peak pressure (e.g., Phillips et al. 2013). (5) As water pressure drops, the effective pressure and hence the load of overlying rock and ice increases, and the horizontal fracture tends to close. (6) Now there are two possibilities: a. the overlying rock slab functions mechanically like a beam, the combined load of rock and ice breaks the rock slab, and new subvertical fractures are generated, b. If sufficient sediment is available and deposited in the fracture, the sediment may take the load and becomes compressed, but beam failure may not occur; (7) Repetition of the above cycle will result in either (i) increasing number of subvertical fractures, resulting in a disrupted rock mass (e.g., Figs. 2A, 3A, 4D), or -more rarely -(ii) very wide (>10 cm) dilated fractures filled with laminated sediment (e.g., Fig. 3B).
The rarity of the wide sediment-filled fractures as seen in Forsmark overlain by intact rock suggests that jacking and subsequent beam failure was generally more common than the sediment deposition and cushioning. Sediment availability (presumably sourced from loosely consolidated till at the icebed nearby) is in principle independent from overpressure, so it is possible to have a jacking event by water with little sediment. In that case, fracture jacking does occur, but the fracture will not be filled with sediment but with water, and subsequent water-pressure drops can easily lead to beam failure and the creation of new fractures.

Implications
Hydraulic jacking followed by beam failure and the generation of new vertical fractures can create a dense fracture network in the shallow bedrock. This process represents a form of mechanical weathering, operating subglacially. The jacking and fracturing will thus lower the rock mass strength of the shallow bedrock, so that it is easier to erode, be it by plucking or by glacial ripping (Hall et al. 2020). The jacking and fracturing also greatly increases in hydraulic conductivity of the upper rock mass. There is no a priori reason that the same mechanism would not operate in other rocks. In sedimentary rocks, bedding planes may play the role of sheeting joints, in allowing access for water pressure to build up beneath slabs of rock. However, the resultant blocks will, in most rocks, be smaller than in basement rocks, and may well be eroded quickly and turn to till. It is thus possible that jacking and fracturing is a major mechanism of subglacial sediment production in sedimentary rocks (Bukhari et al. 2021).
Finally, the deep overpressure as studied, modelled and discussed widely (Talbot 1990(Talbot , 1999(Talbot , 2014Hökmark et al. 2010;Lönnqvist & Hökmark 2013;Hökmark & Lönnqvist 2014) involves residual pore pressure and relies on the hydraulic gradient of the ice sheet as a whole, combined with low hydraulic conductivity of the bedrock mass. Such low conductivity retards pressure equilibration ("drainage") during ice advance or retreat and may thus lead to overpressure. We do not think such mechanisms are responsible for the shallow hydraulic jacking discussed herein because (i) the upper rock mass is characterised by a high conductivity fracture network, that is unlikely to be able to build up sufficient residual pore pressure; (ii) the laminated sediments suggest repeated overpressure events, whilst release of residual pore pressure is likely only as a "one-off" event. Rather, it is likely that the highfrequency, high-magnitude water pressure fluctuations that have been documented to occur beneath the Greenland Ice Sheet (Andrews et al. 2014;Claesson Liljedahl et al. 2016;Wright et al. 2016;Harper et al. 2019) are responsible. Overpressure by such water pressure fluctuations at the icebed is limited to about 10% above cryostatic pressure (e.g., Claesson Liljedahl et al. 2016), and can thus only effectively jack open rock to a depth of c. 3% of ice thickness (rock being c. 3 times denser than water). They are thus restricted to the uppermost 10s of metres of the bedrock, consistent with observations. Thus, shallow (<10-30 m) and deep hydraulic jacking (if occurring) are likely driven by very different mechanisms.

Conclusions
Near Forsmark in east-central Sweden, pre-existing subhorizontal fractures have been jacked open and filled with waterlain sediment, creating dense networks of dilated fractures in the upper 5-10 m of basement gneisses. Re-analysis of historic photos from excavations in the shallow rock mass around the Forsmark power plant shows that there is a higher density of vertical fractures above dilated sub-horizontal fractures. Subvertical fractures are less abundant above tight subhorizontal fractures. Rare dilated fractures with thick sediment fill do not show subvertical fractures. Vertical fractures may thus have been newly formed during hydraulic jacking.
It is suggested that the hydraulic jacking is caused by the same high-frequency water pressures fluctuations that have been measured at the base of the Greenland Ice Sheet.
Modelling applying beam theory, borrowed from structural engineering, shows that beam failure is plausible for rock slabs longer than 5-10 m, if they are jacked up and lowered down, under fluctuating water pressures. Beam failure is thus a plausible mechanism for the generation of new vertical fractures above jacked subhorizontal fractures. This explains the dense fracture network (and smaller size of joint-bound blocks) in the shallow subsurface, and explains previous observations that subhorizontal fractures are coated (and hence pre-Quaternary), whereas some subvertical fractures are uncoated and may have developed during the last glaciation. Beam failure modelling also implies it is not plausible that long slabs of rocks are jacked and lowered and remain intact. The exception is if the void space below the jacked slab is filled with sediment, so that the load overlying rock and ice distributed across a wider area of soft sediment. This is only possible if sediment deposition occurred coevally with the jacking. are thanked for comments on an earlier version of the manuscript. Stefan Bergman and one anonymous reviewer are thanked for constructive comments. MK, RP and CA publish with permission of the Executive Director of BGS.
Funding by Swedish Nuclear Fuel and Waste Management Company (SKB) is gratefully acknowledged.

Disclosure of potential conflicts of interest
No potential conflict of interest was reported by the author(s).

Funding
This work was supported by the by Swedish Nuclear Fuel and Waste Management Company (SKB) [n/a].