Geophysical early warning of salt precipitation during geological carbon sequestration

Sequestration of industrial carbon dioxide (CO2) in deep geological saline aquifers is needed to mitigate global greenhouse gas emissions; monitoring the mechanical integrity of reservoir formations is essential for effective and safe operations. Clogging of fluid transport pathways in rocks from CO2-induced salt precipitation reduces injectivity and potentially compromises the reservoir storage integrity through pore fluid pressure build-up. Here, we show that early warning of salt precipitation can be achieved through geophysical remote sensing. From elastic P- and S-wave velocity and electrical resistivity monitoring during controlled laboratory CO2 injection experiments into brine-saturated quartz-sandstone of high porosity (29%) and permeability (1660 mD), and X-ray CT imaging of pore-scale salt precipitation, we were able to observe, for the first time, how CO2-induced salt precipitation leads to detectable geophysical signatures. We inferred salt-induced rock changes from (i) strain changes, (ii) a permanent ~ 1.5% decrease in wave velocities, linking the geophysical signatures to salt volume fraction through geophysical models, and (iii) increases of porosity (by ~ 6%) and permeability (~ 7%). Despite over 10% salt saturation, no clogging effects were observed, which suggests salt precipitation could extend to large sub-surface regions without loss of CO2 injectivity into high porosity and permeability saline sandstone aquifers.

CSMe aimed at analysing the distribution of CO 2 , brine, and CO 2 -induced precipitation of salt crystals in the rock using micro-X-ray computed tomography (XCT) scans. In the first step, 17 pore volumes (PV) of brine were injected through the rock at a flow rate of ~ 0.2 cm 3 min -1 , followed by a XCT (lasting ~ 24 h). Next, CO 2 gas www.nature.com/scientificreports/ (CO 2(g) ) was injected into the sample at ~ 0.2 cm 3 min -1 continuously for 24 h (~ 200 PV), followed by a second XCT scan (~ 24 h). From the XCT, we developed a four-phase segmentation analysis to obtain the CO 2 , brine, quartz grain and salt volumetric fractions. Then, the sample was dried, cut and prepared into thin sections to compare with the original sample to assess the total salt content in the porous medium. Finally, the sample was subjected to X-ray diffraction (XRD) and scanning electron microscopy with energy dispersive spectroscopy (SEM-EDS) analysis to assess the precipitation of secondary minerals resulting from the CSMe test (see Supplementary Information). For the thin section analysis, we developed a three-phase (pore space, quartz grain and salt) segmentation analysis.
In the NOCe, we measured geophysical (ultrasonic compressional and shear, P and S, wave velocities and electrical resistivity) and hydro-mechanical (axial strains) properties, under three flooding conditions: (i) initial 25% NaCl brine flow-through the brine saturated sample (~ 10 PV); (ii) CO 2(g) injection flow displacing the brine from the saturated sample (drainage episode; ~ 75 PV); and (iii) 25% NaCl brine flow through the partially CO 2 saturated sample (imbibition episode; ~ 30 PV). During the test, we alternated between two flow (Q) regimes: active measuring periods when Q was above 0, and interludes (Q = 0) without data collection. The test was repeated twice, named hereafter as NOCe Test-1 and Test-2, respectively.

Results and discussion
Salt saturation estimate (CSMe). XCT image analysis (Fig. 2) shows that CO 2 and brine occupied 42% and 50% of the total pore volume, respectively, after CO 2 injection (i.e., second scan), yielding a salt saturation (S NaCl ) of ~ 8% (Table 1). During the scanning the vessel was sealed, limiting the brine salinity to the dissolved ions within the pore water. Considering that the XCT scan lasted ~ 24 h, S NaCl likely increased during scanning and therefore 8% represents a lower bound of S NaCl in the CO 2 -brine-rock system after two days of CO 2 exposure (i.e., one day of CO 2 injection and one for the scan).
Theoretically, using mass balance considerations, in our closed system, one single halite crystal of density ρ NaCl = 2100 kg m -3 formed from the brine (S brine , with density ρ brine = 1190 kg m -3 and salinity X NaCl = 25% wt) occupying the available pore space post-test (i.e., 1 -S CO2 ) would fill 8.2% of the total pore volume (i.e., S NaCl(theoretic) = (1-S CO2 ) · X NaCl · ρ brine /ρ NaCl ). However, thin section analysis ( Fig. 3) reveals the subsequent drying process led to a final S NaCl(observed) ~ 19% (± 2% derived from 2D-thin section to 3D transformation 28 ). This analysis indicates that the salt precipitates around the rock grains mainly as an agglomerate of halite microcrystals, as previously observed 6 . This finding is corroborated by SEM-EDS analysis (Fig. 4) and XRD post-testing, which also shows NaCl crystals sometimes coexisting with nahcolite (NaHCO 3 ) as subsidiary (further information in Supplementary Information). For practical reasons, we adopt the physical properties of the major component observed (i.e., halite) for the whole salt fraction and estimate a salt-aggregate micro-porosity of ~ 56.7% ± 2%, resulting from the ratio between theoretical and thin-section estimated S NaCl post-test (ϕ NaCl,micro = 1 -S NaCl(theoretic) / S NaCl(observed) ). This micro-porosity fraction is below the image resolution for both the XCT and thin section processing. Accounting for this micro-porosity, the S NaCl determined from XCT scan drops to 3.5%, indicating the salt precipitation was still in an early stage after two days exposed to CO 2 . Geophysical data analysis (NOCe). The NOCe CO 2(g) injection test lasted approximately five days (~ 115 h) and it was repeated twice, with an effective flow-through time of ~ 40 h that led to ~ 110 pore volumes (PV) throughput each (Fig. 5). Both tests show similar trends for all the measured properties except for the strains. Around 2% hysteresis effect of P-and S-wave velocities (V P and V S ) suggests a change in the physical properties of the sample during the experiment. On the other hand, the transport properties of the sample remained undiminished (under the experimental conditions), as indicated by an almost constant pore pressure gradient (ΔP p ) throughout the NOCe for each pore fluid (i.e., brine flow in E1 brine and E5; CO 2 from E1 CO2 to E4) within the range of flow rates 0-2 cm 3 min -1 .
The CO 2 arrival (episode E1 CO2 after 9 PV of brine flow; Fig. 5), led to a resistivity increase of ~ 50%. Electrical resistivity tomography (ERT) images ( Fig. 6) show preferential pathways localized peripherally by the inlet-outlet ports on the axial platens (diagonally opposite one other); the invariant differential pore pressure for changing flow-rates suggests the preferential pathways remained active through the test 29 . This annular fluid distribution provides very little variations on V P and V S . Hence, the brine was first mainly drained from the edges of the sample whereas the CO 2 is progressively saturating the resident brine towards the centre. Water evaporation rate into the CO 2 stream is very low during this step 23 . This behaviour might indicate early salt precipitation acting as a lateral permeability barrier for CO 2 advance towards the central part of the sample 18 .
After the initial gas-pushing water piston fluid substitution effect 30 , each interlude leads to a redistribution of the two-phase fluids due to capillary effects that enhances salt precipitation 7, 26 . During the interlude E1-E2, in the absence of effective pressure variations, the axial strain (ε a ) increased (more significantly during the first test; Fig. 5). Since our quartz-rich sandstone is barely reactive to CO 2 , we interpret sample dilation as being caused by local salt-induced volumetric increase effects, previously observed in CO 2 -brine tests in sandstones under hydrostatic confining conditions 31 and in air-brine systems under uniaxial loading 32 . In this regard, it has been recently demonstrated 33 that salt crystals may exert pressures against pore walls up to ten times higher than the confining pressure used here (up to 150 MPa at the 50 μm observation microscale). Sample dilation continued during episode E2 accompanied by a weak decrease in both V P and V S by ~ 2%, and a resistivity increase by ~ 25%, in agreement with previously reported data 5,31,34-37 . Overall, E1 and E2 have similar pore fluid distributions (Fig. 6), and small geophysical variations, which can be explained by minor mechanical-chemical changes in the outer part of the sample.
The most significant changes in the elastic properties of the rock sample occur after the second interlude (from E2 to E3; Fig. 5), with V P and V S increasing by ~ 3% and V S ~ 0.5%, respectively, in both tests. These changes Scientific RepoRtS | (2020) 10:16472 | https://doi.org/10.1038/s41598-020-73091-3 www.nature.com/scientificreports/ contrast with the positive deformation of the sample (i.e., inflation) and the theoretical lowering of V P for an increasing CO 2 content 5, 36 . However, our results agree with the scarce data reported about changes of elastic wave properties associated with salt precipitation in high porosity sandstone 8, 38 . During E3 and E4, resistivity, strains,  Saturation uncertainties. In a natural reservoir, a constant brine salinity scenario should be expected provided that salt clogging does not occur; in our test, the brine salinity progressively decreases as salt saturation increases. Since we calculate pore fluid saturations from the measured electrical resistivity and Archie's laws 31,35 , we must first consider any changes in brine salinity caused by the CO 2 drying process. The observed changes in bulk resistivity are mainly related to (i) a fluid substitution effect (i.e., high resistive CO 2 displacing the original brine in the porous medium), and (ii) a progressive decrease of the original brine salinity with CO 2 (i.e., ionsdepleted residual brine due to salt precipitation). Each of these two factors reduces resistivity and affects the transformation of electrical resistivity into saturation differently. Table 1. XCT scan and thin section processed results. a General volume fraction error of 1% and saturation uncertainty of ± 0.01. b NaCl phase contains NaCl plus others isotropic minerals.  −m , Archie's laws for fully and partially saturated porous media 40 , respectively. In these expressions, R b is the bulk resistivity of the sample with subscripts 0 and i indicating full and partial saturation, R w is the resistivity of the pore fluid, ϕ the porosity, S w the brine saturation, and m and n are the cementation and saturation exponents, empirically derived for each particular rock.

CT-Run (CSMe)
The dissolved CO 2 (less than 5% vol. at the experimental conditions 41,42 ), is invisible to our resistivity and ultrasonic detection tools 35 , and therefore neglected in the saturation calculations based on Archie's expressions above. This simple transformation is valid for invariable porous framework systems, i.e., in the absence of dissolution/precipitation phenomena. However, the XCT scanning showed a minimum S NaCl (aggregate microcrystalline) of 8% for our experiments. CO 2 -induced salt growth progressively changes the bulk electrical properties of the rock, by changing both the porosity and brine resistivity. Therefore, to calculate S w , the combination of the first and second Archie's laws has to be modified to account for each state of the rock with respect to the original values (subscripts i and 0, respectively), as follows: In Eq. (1), we consider constant cementation (m) and saturation (n) exponents, because the salt aggregates observed in our CSMe test are suspended at the CO 2 -brine interface (Fig. 2). However, the cementation exponent is likely to increase in more advanced drying stages. Note that the change in water resistivity between stages is explicitly considered in Eq. (1), but the change in porosity due to salt precipitation is not. The micro-metre salt crystals are surrounded by highly conductive water films feeding the crystallization process 6 .
At laboratory conditions (20 ºC), if we consider a closed system where all the original brine contributed its ions to halite formation, the R w would have increased according to the expression 43 R w = 4 × 10 5 /TS 0.88 , with S being the brine salinity (in ppm) and T the temperature (in Fahrenheit). However, some backflow of brine towards the injection point 17 , particularly during interludes, might have partially contributed to salinity recovery. We consider the transition between a backflow completely renewing the resident brine (i.e., constant salinity), to no backflow leading to a progressively NaCl-depleted brine (Fig. 7). In our test, the degree of brine saturation estimated from bulk resistivity data should lie between both cases.
Combining elastic and electrical properties. Salt can form either away from the mineral grains (noncementing salt), or at grain contacts bridging mineral grains (cementing salt). Image analysis suggests a transition from non-cementing salt at early stages of salt formation (XCT images; Fig. 2) to cementing salt after fully drying (thin sections; Fig. 3). Early detection of this transition in the field using geophysical monitoring techniques is essential to initiate mitigation strategies in a timely manner, before both injectivity and ultimately the mechanical integrity of the reservoir are compromised (as suggested by the mechanical hysteresis observed in our data).
The bulk elastic properties of rocks are sensitive to the rock frame and pore fluid elastic properties 44 , and to the distribution of fluids in the porous medium for multiple fluids (e.g., uniform/patchy 45 ). Hence, to infer such a transition in the salt distribution from our geophysical data, we can use the so called uncemented and cemented sand models 46,47 , and introduce them into Biot-Stoll poroelastic formulation [48][49][50] to calculate frequencydependent P and S wave velocity and attenuation. This approach has been applied previously to understand and assess quantitatively the impact of marine methane hydrate on these properties 51, 52 . Due to the similarities in the distribution of hydrate and salt in the pores, we adopt these models using the physical properties (bulk and shear moduli, and density) of halite. www.nature.com/scientificreports/ Using these model predictions, we can explain the observed evolution of the CO 2 -brine-salt-sandstone system from changes in the compressional and shear wave velocities with brine saturation, with both the NOCe Test-1 and Test-2 showing very similar results (Fig. 8). We assume that non-cementing salt only modifies the bulk modulus of the fluid. In contrast, cementing salt modifies both the shear and bulk moduli of the dry granular frame. Hence, V S can be significantly affected by cementing salt but not by non-cementing salt (only minor changes can occur due to changes in bulk density of the rock). www.nature.com/scientificreports/ For P-waves, the change in V P from E1 brine (brine flow) to E1 CO2 (CO 2 flow) can be explained by the presence of up to ~ 4% of solid non-cementing micro-crystalline NaCl into the pore space during Test-1 (~ 2% for Test-2), partially reabsorbed during the interlude E1-E2 (Fig. 8a). The most significant change in V P occurs from E2 to E3. During this interlude, the drying through evaporation is likely to be enhanced by capillary and diffusive mechanisms, leading to a higher crystallization rate 7,26 . The model indicates a sharp S NaCl increase from E2 (S NaCl ~ 1%) to E3 (S NaCl ~ 13% in Test-1 and ~ 14% in Test-2). The brine remaining in the pore space is then NaCl-depleted and salt crystallization stabilizes, leading to a (CO 2 -brine) fluid substitution stage with some salt reabsorption phenomena at the end of the episode E4.
For S-waves, V S reflects an early rock weakening effect during the transition from brine to CO 2 injection (V S drops from E1 brine to E1 CO2 ), likely associated with poorly attached silica cement or grains, removed during early stages of the CO 2 flooding due to sudden changes in the bulk properties (i.e., early salt nucleation) of the pore fluid (Fig. 8b). This early weakening effect is observed in both NOCe tests, but more significantly in Test-1. To isolate the CO 2 -induced salt precipitation effect after this early weakening, we project the modelling results to fit the data after E1 CO2 (grey bands and grey lines in Fig. 8b) by considering the final V S value of each test as V S,0 . From E1 CO2 to E3, the V S data scatter (within the ultrasonic measurements error band) over the whole range of results for the non-cementing salt modelling (i.e., S NaCl from 0 to 14%). The clear increase in V S from E3 to E4 www.nature.com/scientificreports/ can be explained by considering a transition (red arrows in Fig. 8) from only non-cementing salt in E3, to noncementing salt with a small fraction (0.1%) of cementing salt in E4 (red curve in Fig. 8); grey solid curve for the projection in Fig. 8b). Finally, brine is injected into the sample, and rapidly replaces (and dissolves) the salt-rich CO 2 -brine mixture in the pore space. For a precise quantification of the evolution of the three pore phases, we would need to account for the progressive ions depletion of the original brine, which could be significant, as shown in Fig. 7; this would require monitoring of pore water conductivity during the CO 2 flooding, which is missing in this experiment. Instead, we have added one single horizontal error bar to reflect the uncertainty associated with the increasing pore water resistivity for each episode (E1-E5).
The transition from non-cementing to cementing salt (E4) occurs after a substantial S NaCl increase from E2 to E3 of (from ~ 2 to ~ 12% in Test-1 and up to ~ 14% in Test-2) well-defined by ~ 3% increase in V P (Fig. 8a). This rapid S NaCl increase occurs in the absence of CO 2 flow (interlude E2-E3). By contrast, during flowing episodes, changes in rock properties are better explained by fluid substitution mechanisms as data points within each episode evolve towards higher CO 2 saturations.

Insights of salt distribution in geological reservoirs.
From the constant response of our pore pressure sensors for different flow rate conditions all through the NOCe tests, we conclude that any clogging effects due to salt induced porosity reduction were insignificant. The distribution of gas-induced salt precipitation in a porous medium can be described through the Peclet (P e = LV/D), and Damköhler (Da = Lk r / V for Pe > 1 53 ) numbers (reference length, L; effective flow velocity, V; diffusion coefficient of salt in water, D; kinetic rate constant, k r ). Pe determines the importance of advective (Pe > 1) and diffusive (Pe < 1) transport 54 , while Da describes to which extent the dissolution/precipitation processes are dominated by fluid velocity or mineral reactivity 53 . For Pe > > 1, salt precipitation would occur close to the injection zone, whereas Pe < < 1 would lead to a homogeneous distribution 23 ; the crystallization process is controlled by diffusion when Da > > 1 55 . In our closed system, the calculation of Da is not straightforward, as it varies with the ions concentration decay in original brine with the increasing S NaCl . In the NOCe, the Pe changes from the order of 10 2 (advective regime) during flowing episodes (i.e., V > 0), to null during interludes (i.e., V = 0), and the system is completely controlled by the diffusion (i.e., Da tends to infinity). Since this change in Pe is expected to enhance the homogenization of salt precipitation, it explains the S CO2 and S NaCl backwards transition from E1 CO2 to E2 during Test-1 (Fig. 8a).
Overall, upon detection of CO 2 -induced salt precipitation in a reservoir formation through seismic and electromagnetic geophysical surveys, reducing or stopping the CO 2 injection is advisable for high porosity and permeability reservoirs and low brine salinity (e.g., Sleipner CCS field 35 ). However, stopping or reducing the CO 2 injection (i.e., minimum Pe; maximum Da), might have undesirable effects for reservoirs with lower porosity and permeability and higher brine salinity 8, 56 . In these latter reservoirs, the higher capillarity of the rock could lead to self-enhancing salt growth phenomena 6 , rather than salt reabsorption and homogenization, which would extend the salt precipitation to large sub-surface regions. In turn, this would reduce the effectiveness of common clogging mitigation techniques, such as fresh water flooding, mainly focused on the surroundings of the injection well. www.nature.com/scientificreports/ Furthermore, salt would be prone to fill (micro-) pores and start cementing the rock, thus modifying rock mechanical properties. During the GCS post-injection stage, natural aquifer recharge might lead to re-dissolution of salt crystals, this time with permanent changes in rock mechanical properties, as suggested by the hysteresis shown by the axial strain (only ~ 5% recovery) during the NOCe, together with the V P and V S drop and the increase of the original porosity and permeability, during NOCe Test-1 and Test- 2. conclusions During a set of CO 2 flow-through experiments using a synthetic sandstone of well-known petrophysical properties, we have found evidence of CO 2 -induced salt precipitation from X-ray micro-CT imaging and of detectable geophysical signatures (elastic waves and electrical resistivity) associated to this phenomenon. Our results show for the first time how elastic wave velocities, when combined with electrical resistivity measurements, can be used to detect early stage salt precipitation during CO 2 injection into reservoir sandstones. Based on our experimental results, we conclude: 1. The precipitation of salt induced by CO 2 injection into brine saturated sandstones significantly affects measurable geophysical properties of the original rock formation, i.e., compressional and shear wave velocity, and electrical resistivity. During early stage injection, salt precipitates away from grain contacts (non-cementing), www.nature.com/scientificreports/ then starts to cement the rock grains after a certain threshold salt saturation is achieved (10% in our experiment). 2. The precipitation of salt triggers detectable changes in compressional wave velocity (up to 3%) that can be differentiated from pore fluid substitution effects with the aid of electrical resistivity, based on the latter's sensitivity to electrically insulating gas phases like CO 2 . While compressional wave velocity is sensitive to both non-cementing and cementing salt precipitation, shear waves velocity is predominantly sensitive to the fraction of cementing salt. 3. During CO 2 -induced salt precipitation, pore fluid salinity changes have to be considered to successfully estimate the evolution of CO 2 , brine and salt saturations from electrical resistivity. 4. CO 2 -induced salt precipitation may lead to dilation of the rock frame, even in high porosity and permeability sandstones that are barely reactive to CO 2 . Hence, early detection of salt precipitation is crucial to preserve reservoir injectivity properties and mechanical integrity.

Materials and methods
Rock samples and pore fluid. We created a homogeneous synthetic sandstone of high (effective) porosity (ϕ = 0.29; He-pycnometry), and high absolute permeability k abs ~ 1660 mD (N 2 -permeameter), to avoid saltinduced clogging during CO 2 injection. Following the manufacturing procedure described in Falcon-Suarez, et al. 57 , we used sorted, coarse grained (diameter > 500 µm) sand and low cement-to-grain ratio to ensure high porosity and permeability (with a final dry density ρ d ~ 1826 kg m -3 ) and to avoid clay conductivity and clogging effects. The composition of the sample was ~ 97% quartz, ~ 3% albite (X-ray diffraction, XRD, analysis with a Philips X'Pert pro XRD-Cu X-ray tube), which ensures the applicability of the original Archie's laws without applying corrections related to the presence of clay minerals in the rock. From the original sample, we extracted two core plugs of (sample A) 2 cm length, 1.2 cm diameter, and (sample B) 2 cm length, 5 cm diameter, for the CSMe test and for the NOCe tests, respectively. We used a 25% NaCl synthetic brine solution prepared with deionized water in both tests, keeping NaCl concentration slightly below the saturation point to promote rapid crystallization during CO 2 -induced evaporation.
To ensure the saturation condition, the samples were first oven-dried, then saturated with the experimental brine via imbibition under vacuum conditions, and finally flushed at 5 MPa with the experimental brine solution to enhance the dissolution of remaining air in the system. At the pore pressure used in this test, the salinity of the brine has neglectable effect on the mutual solubility of CO 2 and brine.

CSMe setup and test configuration. The brine saturated sample A contained in an in-house developed
Aluminium pressure vessel was pressure equilibrated for one night at the target P c (25 MPa) and P p (5 MPa) conditions, and left for one night for stress-strain equilibrium. Then, the test started with a brine flow episode at 0.2 cm 3 min -1 , up to completing 17 PV of brine flow-through. Immediately after, the first XCT-scanning was carried out lasting ~ 24 h. Next, CO 2 was injected in the sample at variable flow rate (0.1-0.2 cm 3 min -1 ) continuously during 24 h (Peclet's number Pe ~ 60), reaching ~ 200 PV of collected pore fluid volume downstream. Immediately after, the second XCT-scanning was conducted.
The pressure vessel setup consisted of an aluminium tube sealed with commercial high-pressure fittings, and three flow-through lines for confining, and inlet and outlet pore pressure. The sample A was jacketed and plugged with in-house manufactured endcaps, allowing pore fluid flow-through while isolating the sample from the confining hydraulic oil. Pressure vessel connecting lines were flexible to ensure free twisting of the vessel within the XCT instrument during XCT image acquisition. We performed the measurements using a Xradia Micro XCT-400. Tomographic images were collected in increments of 0.1 degrees of rotation with a background reference image generated from 10 averaged individual scans preceding installation of the pressure vessel. The reference image was subtracted from the measurement images to give adjusted values for the sample and vessel without background effects. Optimal tomography angles were calculated using specialist software XM-Controller 8. 1.7546 (XRadia Inc.; https ://www.zeiss .com/micro scopy /int/produ cts/x-ray-micro scopy ), and measurements were obtained with 0.5 × and 4 × lens yielding resolutions of approximately 25 and 3 microns, respectively. For this study, we only used the latter (high resolution image set). Tomographic images were generated following beam hardening and centre shift corrections to ensure optimal image quality in the reconstructed images, using Avizo software version 9. 5.0 (https ://www.fei.com/softw are/avizo -user-guide ). The XCT-images were segmented using a 3D weka semi-automated segmentation 58 to characterize the four phases of interest: CO 2 , brine, quartz and the rest, which included ore minerals and salt (Fig. 2). First, we selected a sub-sample (~ 1.5 mm 3 ) and obtained the four phases; second, we extrapolated the results to the whole sample, following the procedure described in Callow, et al. 59 .
After the test, the sample was left to dry under atmospheric conditions. Then, thin sections were obtained from both the original and the tested samples, to assess the total salt content in the porous medium. The images were analysed using Fiji-ImageJ software version 2.0.0-rc-68/1.52 g (https ://fiji.sc) to measure the total salt in the sample. The blue-resin method used to make the thin sections allowed the segmentation of the obtained in plane-polarized light images into four domains: pores (blue), grains (light grey), cement (brownish), and salt (black regions: isotropic salt crystals and ore minerals). The processing consisted of a threshold filtering to define the pores (ϕ) and salt (S) fractional areas, and then transform it to grey scale to calculate the salt saturation as S NaCl = S / (ϕ + S).

NOCe setup and test configuration.
The rig is configured around a triaxial cell core holder under accurate control (ISCO pumps) of confining and pore pressure 31 . The rubber sleeve inside the vessel has 16 stainless steel electrodes connected to an electrical resistivity tomography data acquisition system 31 www.nature.com/scientificreports/ electrode configuration, each run collects 208 individual (tetra-polar) measurements, which are then inverted using a variation of the software EIDORS (version 3.1) MATLAB toolkit (https ://eidor s3d.sf.net). Axially, two (confining) platens house ultrasonic pulse-echo sensors, isolated from the sample and the rest of the rig by two polyether ether ketone (PEEK) buffer rods of well-defined acoustic impedance and low energy loss. This configuration provides a reliable delay path within the frequency range 400-1000 kHz, which enables the identification of top/base sample reflections for calculating ultrasonic P-and S-wave velocities and attenuations using the pulse-echo technique 18,19 , with ± 0.1% velocity precision and ± 0.3% accuracy (95% confidence), while up to ± 5% for the attenuation. The signals are displayed on a digital oscilloscope LeCroy Wavesurfer 4000HD (https ://teled ynele croy.com/) and stored on the computer for spectral analysis assembly using MATLAB R2017a (https :// www.mathw orks.com). For this test, we processed the ultrasonic data from Fourier analysis of broad band signals to compare the ultrasonic properties of our three samples at a single frequency of 550 kHz. The sample B was equipped with 350 Ohms electrical strain gauges, epoxy-glued on the lateral side of the sample, to record axial (ε a ) strains during the tests in continuous (every two seconds), with an accuracy of ± 10 -6 m m -1 (i.e., < 1.5% error for the deformation range in our experiment; Fig. 5). The strain gauges limited the ERT interpretation to the vertical plane (i.e., 2D) elongated 90° with respect to the gauges position. Along with the strains, the pore pressure was also monitored with two additional pressure (piezoresistivity) sensors located directly up-and downstream of the sample. The injection of both the brine and CO 2 were carried out under controlled pore pressure downstream (5.5 MPa) and outlet flow. To avoid early compaction effects, the sample was initially placed into the triaxial vessel at the test conditions during four days, and then subjected to minimum brine flow-through (0.01 cm 3 min -1 ). Injected confining fluid and the strains indicated mechanical stabilization occurred after two days. After the first test, the confining and pore pressure were dropped by keeping constant the differential pressure (P diff = P c -P p = constant, to minimize stress-induced damages), down to atmospheric P p conditions. Then, first, the sample was newly saturated via vacuum imbibition under confining inside the vessel for ~ 3 h; next, under continuous brine flow, both P c and P p were again increased at constant P diff , first up to P p = 10 MPa (i.e., P c = 30 MPa) during 1 h flow to enhance dissolution of any remaining CO 2 bubble, and finally to the original test conditions (P c = 25 MPa and P p = 5 MPa). Once at the test conditions, the CO 2 flow-through test was repeated, replicating the episodes of the first one. After the NOCe Test-2, the sample was batched in DIW for a week, and then dried to measure the final (subscript f) porosity (ϕ f = 0.306; He-pycnometry) and permeability (k abs,f ~ 1771 mD; N 2 -permeameter).

Elastic wave modelling: rock with salt in the pores. Our modelling of elastic wave parameters follows
Ecker, et al. 51 approach that considers two idealized models for solid phase precipitation in the pores (gas hydrate in their case and salt in this work). We combine the stiff uncemented and cemented sand models 46,47 using MAT-LAB (R2017a) to calculate the frame properties of the dry rock, and applied Biot-Stoll 48-50 to calculate P and S-wave velocity and attenuation over the whole brine saturation range at a frequency of 550 kHz. Fluid density, viscosity and compressibility at the experimental pressure, temperature and salinity conditions were calculated using the High Pressure International EOS for brine 60 and data from the National Institute of Standards and Technology (https ://webbo ok.nist.gov/chemi stry/fluid /) for CO 2 .