Gravity-dependent rate sensitivity in granular intrusion: microgravity experiments and simulations

Abstract

Understanding the shear and intrusion rheological behavior of granular materials under reduced gravitational conditions is crucial for applications in planetary exploration and submarine earthquake engineering. To this end, it is important to understand whether gravity affects the drag forces on objects intruding granular media and, if so, quantify these effects. We have studied this issue experimentally in 1 g and 0 g conditions, the latter using the Beijing Drop Tower. Measuring the resistive forces on a cylinder moving at a constant speed through a granular bed, we find that gravity plays a significant role – the resistive forces increase significantly with cylinder speed in 0 g, while increasing much more slowly in 1 g. We use Coupled Eulerian-Lagrangian (CEL) simulations that support our results. We attribute this behavior to the increasingly dominant effect of pressure-sensitive frictional forces in microgravity with increasing fluidity in these conditions. Other than the significant implications for extraterrestrial exploration, our findings suggest that constitutive modeling of flow in microgravity should differ significantly from that in 1 g.

Introduction

Granular materials exhibit complex rheological behavior that is sensitive to external conditions. In particular, effects of gravitational forces have significant implications for space geotechnical engineering and planetary exploration. Previous studies, including NASA’s Mechanics of Granular Materials (MGM) triaxial shear tests, observed anomalously high peak friction angles in Ottawa sand under microgravity^1–3. This was attributed to enhanced interlocking, hyper-plasticity, and shear band formation under low granular pressure. Couette shear experiments^4–6 and constitutive models^7–10, parameterized by the inertial number, also revealed pressure-dependent shear rheology, featuring a transition from quasi-static to viscous-dominated flow in microgravity’s low-pressure regimes. While resistive forces on objects embedded in and intruding into granular beds have been studied in 1 g conditions^11–14, their behavior in microgravity remains underexplored. Specifically, prior investigations have been primarily either numerical simulations^6,15 or used density-matching methods¹⁶ to mimic low-gravity conditions^17,18. While those studies suggested that resistive forces are very sensitive to low granular pressure^19,20, traditional critical state soil mechanics models fail to capture the observed rate-dependent behavior in microgravity^21,22.

To fill this gap, we have conducted experiments in the Beijing Drop Tower to directly measure resistive forces on a cylinder moving through a granular bed of polypropylene (PP) beads. The experiments in microgravity, 10⁻³g, were repeated in 1 g. We find a striking contrast between the two—while resistive forces in microgravity exhibit strong dependence on the cylinder speed, they are nearly speed-independent in 1 g. We attribute this to enhanced viscous frictional forces under low granular pressure¹⁵, where the inertial number governs the transition to a liquid-like rheological state. We have validated this behavior by Coupled Eulerian-Lagrangian (CEL) simulations, employing an advanced rheological constitutive model, which help elucidating the pressure-sensitive mechanisms that drives the rate-dependence. We believe that this study provides key insight into granular rheology under microgravity, which advances the understanding of granular flow in low-gravity environments, and it is therefore relevant to lunar and Martian soil mechanics.

Methods

Experimental

The experimental setup is sketched in Fig. 1a. A rectangular container of dimensions 350 mm × 180 mm × 300 mm was filled to height of 190 mm with a granular bed of polypropylene (PP) beads of density 900 kg/m³, mean particle diameter 3 mm, packing fraction 0.617, and bulk density 555 kg/m³. In this study, we selected polypropylene (PP) beads because of their low density, which is essential for measuring the forces encountered in both 1 g and 0 g conditions within the sensor’s measurement range. An embedded cylinder of length 180 mm with eight 20 mm-wide double-hole beam force sensors mounted along the cylinder was suspended on a guide rail attached to a slider, enabling horizontal movement through the granular bed. The motion speed was computer-controlled and ranged from 35 to 100 mm/s. The eight force sensors mounted along the embedded cylinder were employed to measure resistive forces at various depths. The cylinder’s bottom was positioned 50 mm above the container base.

Fig. 1. Schematic diagram of the experimental setup and process.

a Experimental setup (all measures are in mm). b The timeline of each drop tower experimental run.

Experiments were conducted in the Beijing Drop Tower, whose height is 116 m, providing a microgravity environment of 10⁻³ g for approximately 3.6 s during a free fall. The experimental capsule, of a platform diameter of 850 mm, housed the setup, a controller, and a high-speed camera to record motion. Each experimental run followed the timeline shown in Fig. 1b. At t = 0 s, a stepper motor initiated the controlled motion of the embedded cylinder; at t ≈1 s, the capsule was released, entering the microgravity phase; the motor was deactivated at t = 2 s; and the capsule entered the recovery phase at t ≈ 4.6 s, concluding the run.

Measurements of the resistive forces were taken during runs at different cylinder speeds. The resistive forces were measured in stage I at various depths before capsule release in 1 g. In Stage II, under microgravity conditions, forces were recorded as the capsule was falling. Stage III provided baseline zero-force measurements. Strong force fluctuations at the end of Stage III marked the capsule’s impact with the recovery basket, at which point the run ended. A typical set of measurements is shown in Fig. 2 for a run at a cylinder speed of 50 mm/s.

Fig. 2.

A typical run data at a drag speed 50 mm/s.

For each depth, the resistance data from Stage 2 and Stage 3 were averaged separately to determine the corresponding resistance values for these stages. The difference between the Stage 2 and Stage 3 averages was then regarded as the cylinder’s pulling resistance under 0 g conditions at that specific depth. Simultaneously, the resistance data from the final 0.5 s of Stage 1 were averaged to represent the resistance for Stage 1, with its deviation from the initial value defined as the cylinder’s pulling resistance under 1 g.

Additionally, the standard deviation of the data from Stage 1 and Stage 2 at each depth was calculated, serving as the error estimate for the 1 g and 0 g resistance values, respectively.

Numerical simulations

To investigate the differences in behavior between 1 g and microgravity, we carried out numerical simulations of the interaction between an embedded cylinder and a granular bed, using the CEL finite element method. We have chosen the CEL method as it provides a natural computational platform for implementing our unified constitutive model within a continuum mechanics framework, and as it can simulate large-deformation fluid-structure interaction problems, making it particularly suitable for our experiments.

The computational domain for the numerical simulation is designed to directly correspond to the dimensions of the experimental setup. The Eulerian domain inside the experimental box is set to 200 mm in height, exceeding the initial granular material height of 140 mm. The intruding beam dimensions correspond to the physical configuration used in the experiments, ensuring a one-to-one representation. The temporal discretization used in our simulations is adapted based on the specific requirements of the CEL modeling framework to ensure stability and accuracy in capturing the interactions between the cylinder and the granular material. The simulation time step is fixed at 10⁻⁵ s. The spatial discretization includes a refined grid for both the Eulerian and Lagrangian structures. The Eulerian grid near the cylinder is refined to 5 mm, gradually coarsening to 10 mm at the outer boundary. The Lagrangian grid for the intruding beam is aligned with these dimensions to accurately capture the interactions. We have assumed the surfaces to be “smooth and rigid,” meaning that penetration into the surfaces is prohibited in the normal direction, and the frictional shear force is set to zero. This definition aligns with the mathematical representation of hard contact in the simulations. The walls of the numerical domain are treated similarly as smooth, rigid surfaces, ensuring that no penetration occurs. We acknowledge that friction between the granular material and the walls can play a role in the observed resistive forces; however, our calibration tests show that including or omitting wall friction has minimal influence on the primary results. To reduce computational costs, the 50 mm gap between the cylinder tip and the container base (Fig. 1a) was omitted. We recognize that omitting this gap may have implications for the simulation outcomes, particularly regarding downward particle movement and resulting resistive forces. To address this concern, we conducted a calibration analysis. If we ignore the 50 mm gap below, the calculation results only affect the resistance of the bottom segment—making it slightly larger—but do not influence the resistance calculations for the segments above. The main reason is that the most particles can still flow around the rod and continue downstream without the gap below.

The CEL method, commonly applied to fluid-structure and solid-structure interactions, was implemented in Abaqus via the VUMAT explicit algorithm, incorporating a constitutive model called Tsinghua-Midi Sand model (TMS), which had been previously validated against granular rheology experiments in microgravity⁸.

The TMS constitutive model decomposes the total shear stress into quasi-static and viscous components, as shown in Fig. 3 and described in Eqs. 1–6. The quasi-static shear stress depends on nonlinear shear stiffness, derived from soil mechanics data, and the viscous shear stress follows a modified μ(I) rheology law. I is the inertial number, which reflects granular pressure and shear strain rate effects¹⁰. Key model parameters are listed in Table 1. Figure 3 illustrates how shear strain (γ) relates to the effective friction coefficient (μ) and its constituents. This γ–μ plane, augmented with shear-rate information, is the standard framework for describing the flow of sands and other granular media. The upper horizontal line μ₁ marks the boundary between two dense-flow regimes: below μ₁ the material deforms steadily and moderately, whereas above μ₁ the flow accelerates and becomes unsteady. The latter regime—excluded from the present study—is distinguished by the loss of sustained inter-particle contacts and by momentum transfer that is dominated by binary collisions. The lower, curved line OC corresponds to the quasi-static mobilized friction coefficient (μ_m). This coefficient encodes the stress–strain response when deformation is slow; its upper limit, μ_s, coincides with the Drucker–Prager critical-state yield surface. Thus, μ_m <μ_s identifies a quasi-static condition, while μ_m > μ_s signals the onset of moderate shear flow. In steady states where both μ > μ_s and γ > γ_s, the “viscous” contribution to stress is assumed to depend on the shear rate (γ̇) and the effective isotropic pressure. In most seismic and dynamic-geotechnical problems, granular soils are loaded transiently (e.g., at point A) and never reach the steady condition represented by point C; instead, they progress toward an intermediate shearing state B. Consequently, their behavior is governed by transient, rather than steady, shear-flow characteristics.

Fig. 3. The TMS constitutive model concept⁸.

The green area is the domain of steady state flow described by the MiDi rheological model. The blue area is the domain of transient flow described by the transient rheological models.

Table 1.

TMS model main parameters

Item	Parameter	Symbol(unit)	Value
Basic physical parameters	Glass bead density	${\rho }_s(\mathrm{kg}/{\mathrm{m}}^3)$	880
	Initial void ratio	$e_0$	0.621
	Bulk density	$\rho (\mathrm{kg}/{\mathrm{m}}^3)$	543
	Average particle diameter	$d\left(\mathrm{mm}\right)$	3
Quasi-Static part parameters	Critical state friction coefficient	${\mu }_s$	0.32
	Shear stiffness parameter, related to the maximum shear stiffness $G_0$	$A_0$	6000
	Characteristic value of critical state strain related to the power-law change with pressure $p$	${\gamma }_{r_1}$	0.0005
	Characteristic strain value where stiffness $\frac{G}{G_0}=0.5$	${\gamma }_{r_2}$	0.0003
	Bulk modulus	$K\left(\mathrm{MPa}\right)$	100
Flow part parameters	Maximum friction coefficient for steady flow	${\mu }_1$	0.6
	Reference inertial number	$I_0$	0.012

The total stress tensor can be decomposed into an isotropic stress tensor, a quasi-static shear stress part and a viscous shear stress part:

$${\sigma }_{\mathit{ij}}\left({\epsilon }_{\mathit{ij}},{\dot{\epsilon }}_{\mathit{ij}}\right)=p\left({\epsilon }_{\mathit{kk}}\right){\delta }_{\mathit{ij}}+{\tau }_{\mathit{ij}}^s\left({\epsilon }_{\mathit{ij}}\right)+{\tau }_{\mathit{ij}}^v\left({\epsilon }_{\mathit{ij}},{\dot{\epsilon }}_{\mathit{ij}}\right),$$ 1

where ${\epsilon }_{\mathit{ij}}$ is the strain tensor, ${\dot{\epsilon }}_{\mathit{ij}}$ is the strain rate tensor, ${\epsilon }_{\mathit{kk}}$ is the volumetric strain, $p{\delta }_{\mathit{ij}}$ is the effective isotropic stress tensor, ${\delta }_{\mathit{ij}}$ is the Kronecker operator tensor, ${\tau }_{\mathit{ij}}^s$ is the quasi-static shear stress tensor, and ${\tau }_{\mathit{ij}}^v$ is the viscous shear stress tensor.

The effective isotropic stress $p$ is computed by using a linear volumetric stiffness$K$ and the quasi-static shear stress ${\tau }_{\mathit{ij}}^s$ is computed by using a nonlinear shear stiffness $G\left(\gamma \right)$. The effective isotropic stress $p$ and the quasi-static deviatoric stress ${\tau }_{\mathit{ij}}^s$ can be expressed as:

$$p=K{\epsilon }_{\mathit{kk}},$$ 2

$${\tau }_{\mathit{ij}}^s=G\left(\gamma \right){\gamma }_{\mathit{ij}},$$ 3

$$G\left(\gamma \right)=A\left(\gamma \right)\frac{P_a}{{\left(1+e\right)}^3}{\left(\frac{p}{P_a}\right)}^{m\left(\gamma \right)}$$ 4

where $A\left(\gamma \right)=\frac{A_0}{1+\frac{\gamma }{{\gamma }_{r_2}}}$ and $m\left(\gamma \right)=m_0+\frac{1-m_0}{1+\frac{{\gamma }_{r_1}}{\gamma }}$ are fitting functions. $K$ is the bulk modulus and ${\gamma }_{\mathit{ij}}=2({\epsilon }_{\mathit{ij}}-\frac{1}{3}{\epsilon }_{\mathit{kk}}{\delta }_{\mathit{ij}})$ is the engineering deviatoric strain tensor.

In the quasi-static component, the mobilized friction coefficient ${\mu }_m$ and the second invariant of the quasi-static shear stress $|{\tau }^s\mathit{|}$ can be expressed as:

$${\mu }_m=\frac{\left|{\tau }^s\right|}{p},\left|{\tau }^s\right|=\sqrt{\frac{1}{2}{\tau }_{\mathit{ij}}^s{\tau }_{\mathit{ij}}^s}$$ 5

In the viscous component, the friction coefficient ${\mu }_v$ and the deviatoric stress ${\tau }_{\mathit{ij}}^v$ can be expressed as:

$${\mu }_v\left(I\right)=\frac{{\mu }_1-{\mu }_m}{1+\frac{I_0}{I}},{\tau }_{\mathit{ij}}^v={\mu }_v\left(I\right)p\frac{{\dot{\gamma }}_{\mathit{ij}}}{\left|{\dot{\gamma }}_{\mathit{ij}}\right|}$$ 6

The quasi-static mobilized friction coefficient, ${\mu }_m$, is determined by the strain state, whereas the viscous friction coefficient, ${\mu }_v$, is determined by both the strain state and the strain rate. In contrast to original μ(I) rheology, the quasi-static and viscous stresses are only coupled in our model via Equation 5. The viscous friction coefficient ${\mu }_v$ depends on both the quasi-static mobilized friction coefficient, ${\mu }_m$, and the inertial number $I=\left|{\dot{\gamma }}_{\mathit{ij}}\right|d\sqrt{\frac{{\rho }_s}{p}}$, with d the bed particle diameter and ${\rho }_s$ the granular density. This allows the model to account for transient rheological states shown in Fig. 3.

Results

Experimental observations on resistive forces

In Fig. 4a we plot the dependence of the resistive force on depth in 1 g. It shows that the local resistive force (per sensor segment) depends only weakly on the speed but increases linearly with buried depth, consistent with a lithostatic pressure gradient acting on the fixed segmental area. Figure 4b shows a completely different behavior. The resistive forces are two orders of magnitude lower in microgravity than in 1 g and increase more pronouncedly with the cylinder speed. While the forces increase somewhat with depth, this increase is weaker and noisier.

Fig. 4. Resistive force vs buried depth for different drag speed, which reflects two orders of magnitudes difference in the y-axes.

a 1 g and b 0 g.

In Fig. 5, we display the relationship between resistive forces, sum of the measured across all sensors, and cylinder speed in 1 g (red squares) and microgravity (blue circles). The inset of Fig. 5 presents the results normalized by ΔF/F (35 mm/s). In 1 g, the forces increase slightly from ~7 to ~9 N across the speed range, while in microgravity, forces rise 2.5-fold from ~0.06 to ~0.15 N, highlighting pronounced speed sensitivity. The normalized result presented in the inset of Fig. 5 highlights the significant speed sensitivity observed in microgravity.

Fig. 5.

Total resistive forces versus drag speed in 1 g (blue square) and in 0 g (red triangle): with the inset for normalized result defined as ΔF / F(35 mm/s), where ΔF is the difference of the total resistive force at drag speed v compared to F(35 mm/s), and F(35 mm/s) is the force at the lowest tested speed (35 mm/s).

Numerical simulations and validation

In Fig. 6, we compare simulated (hollow symbols) and experimental (solid symbols) resistive forces at various depths, normalized by the experimental force at 130 mm depth and 100 mm/s cylinder speed. The figure shows that, in 1 g, both the simulated and experimental resistive forces show minimal variation with cylinder speed across all depths. This is consistent with a dominant quasi-static behavior. In microgravity, both the simulated and experimental results exhibit a significant increase in the resistive force with cylinder speed, which indicates a viscous-dominated flow. In Fig. 7, we present the dependence of the normalized total resistive forces (summed across depths) in 1 g (red symbols) and in microgravity (blue symbols) on the cylinder speed. These confirm a mild increase in 1 g and a steeper increase in microgravity, aligning well with the experimental trends. Nevertheless, there is an underestimation in the numerical simulations relative to the experimental measurements in microgravity condition. This deviation suggests that the numerical model does not capture accurately plasticity due to local granular rearrangements

Fig. 6. Normalization of the resistive force conducted by the experimental force at 130 mm depth and 100 mm/s cylinder speed to ensureresults are presented in a scalable and comparative format, facilitating clearer analysis across varied experimental conditions.

a 1 g and b 0 g.

Fig. 7.

Normalization of the resistive force conducted by 12 N and 0.25 N to ensure results are presented in a scalable and comparative format, facilitating clearer analysis across varied experimental conditions in 1 g and 0 g, respectively.

The findings illustrated in Figs. 4, 5, and 6 reflect the complex interactions between the intruding beam and the granular material, particularly under varying gravity conditions. The observed trends indicate that increasing the depth of intrusion leads to significant changes in the resistive forces experienced by the beam in both gravity scenarios. This phenomenon can be attributed to the increase in hydrostatic pressure with depth, even at the drop tower microgravity levels of 10^-2g to 10^-3g. This mechanism is a well-known physical principle encountered in both granular materials and fluids.

A key finding is that the resistive force is nearly independent of the drag speed under 1 g conditions at certain depths. In contrast, under microgravity, it exhibits a much stronger rate dependence, providing a coherent link between our experimental and numerical findings. The underlying physics aligns with the inertial number-based granular rheology theory, which compares microscopic and macroscopic timescales.

Since pressure p depends on the gravity level, while shear rate relates to drag speed, the same drag speed results in different inertial number values across various gravity conditions. Under typical 1 g conditions, the drag speeds are small enough to achieve a rate-independent scenario. However, in microgravity, the drag speeds become significant enough as the typical microscopic timescale extends or ‘slows down’. This phenomenon of microscopic ‘slow down’ under microgravity conditions provides valuable insights into granular intrusion flow.

Discussion

Our experiments and simulations highlight the key role that gravity plays in the rheological behavior of granular materials, particularly in the dependence of resistive forces on the speed of intrusion. In microgravity, granular pressure is very low, as shown in Fig. 10. This means that the inertial number$,\mathit{I}=\left|{\dot{\gamma }}_{\mathit{ij}}\right|d\sqrt{\frac{{\rho }_s}{p}}$, is much higher, which enhances the viscous frictional forces, which, in turn, dominate the granular rheology. Then, following from Equation $6$, this leads to a pronounced rate dependence of the viscous shear stress, ${\tau }_{\mathit{ij}}^v$. This is unlike in 1 g, where the granular pressure suppresses the inertial number, resulting in a quasi-static-like flow with much milder dependence on the intruder speed.

Fig. 10. Granular pressure contours (logarithmic scale) at a depth of 130 mm 1 s after the cylinder intrudes at 100 mm/s.

a 1 g and b 0 g.

Figure 8 illustrates the granular speed field at a depth of 130 mm 1 s after the cylinder intrudes at 100 mm/s. In microgravity, the granular speed field extends over a wider region than in 1 g. This enhanced flowability is the direct result of the strongly suppressed inter-granular forces and hence granular pressure. In Figs. 9 and 10, we show contour maps of the shear strain rates and the granular pressure, respectively. Due to the pressure’s and strain rate’s wide range, it is shown on logarithmic scale. The same color scales were employed in the corresponding plots to facilitate comparisons between the two gravitational conditions (1 g and 0 g). These observations are clearly the result of the enhanced fluidity of the granular bed in microgravity, which magnifies the effect of the intrusion speed on resistive forces.

Fig. 8. Granular speed fields at a depth of 130 mm 1 s after the cylinder intrudes at 100 mm/s.

a 1 g and b 0 g.

Fig. 9. Granular shear strain rate contours (logarithmic scale) at a depth of 130 mm 1 s after the cylinder intrudes at 100 mm/s.

a 1 g and b 0 g.

This effect is also well seen by plotting the simulated viscous shear stress, ${\tau }_{\mathit{ij}}^v$, normalized by the total shear stress, against cylinder speed, in Fig. 11. The viscous shear stress exhibits a significantly stronger dependence on the cylinder speed in microgravity. In 1 g, the lower inertial number suppresses the viscous contribution, leading to nearly speed-independent resistive forces (Figs. 6a and 7). All these results show that it is the pressure-sensitive rheological behavior that underlies the difference in the observed dependence on speed between microgravity and 1 g.

Fig. 11. Normalized viscous shear stress against drag speed in 1 g and in 0 g.

‘5 d’ means the averaged resistive forces within the five times cylinder diameter domain and ‘all range’ means the averaged resistive forces within entire sand box.

The Froude number ($F_r=v/\sqrt{\mathit{gL}}$), a dimensionless parameter that normalizes velocity in relation to gravity, appears suitable for analyzing our results, particularly regarding the rate dependence of the resistance force under both 1 g and microgravity conditions. Normalizing velocities through the Froude number may provide complementary insights, especially for comparing inertial-gravitational balances across different g-levels.

For example, in 1 g, our drag speeds (35–100 mm/s) produce Fr values of approximately 0.08–0.23 (using L = 20 mm, the diameter of the cylinder). In microgravity (10⁻³ g), the same speeds result in Fr values of about 2.50–7.14, indicating a regime where inertial forces significantly dominate over gravity. This finding is consistent with the observed increased fluidity and broader granular speed distributions illustrated in Fig. 12.

Fig. 12.

Normalization of relative resistanceΔF / F(35 mm/s) versus Froude number in both 1 g and 0 g.

This normalization reveals the shift in the “effective” velocity range necessary to observe rate effects. To maintain comparable Fr values across conditions, microgravity experiments would require substantially lower speeds (around 1–3 mm/s for Fr values of approximately 0.08–0.23). However, the constraints of our setup, notably the limited 3.6 s of microgravity time, dictated the selected speed range to effectively capture measurable motion and forces.

The systematic deviation of the numerical resistive forces from the experimental results, highlighted in Figs. 6 and 7, warrants further investigation into improving the numerical modeling. Potential causes may be non-local granular flow effects, as reported in Couette flow studies²³, although it is much more likely that the problem lies in inaccurate modeling of local plasticity, which results from local microstructural rearrangements in the medium as the intruder passes through. The problem may be remedied by including explicit effects of the dynamics of the local fabric tensor to complement the μ(I) rheology⁶. Other reasons for the deviation may be that the Eulerian description ignores strain history and precise volume changes in quasi-static shear. These limitations highlight the need for better constitutive models and numerical methods to fully capture the complex rheology of granular materials in microgravity.

Using the Beijing Drop Tower, we conducted experiments to measure resistive forces on an embedded cylinder moving through a granular bed both in 1 g and in microgravity, 10⁻³g. Our results show that there are markedly different rheological behaviors. The resistive forces exhibit much stronger dependence on the cylinder speed in microgravity, increasing 2.5-fold from 0.06 N to 0.15 N when the speed increases from 35 to 100 mm/s. In contrast, the increase in 1 g is very mild, from 7 N to 9 N over the same range of speed variation. Simulations using the CEL simulations, with the TMS constitutive model, corroborate these findings. The difference is traced to the enhanced fluidity of the granular bed in microgravity, which can be quantified by the significant increase of the inertial number at the low granular pressure. This effect is suppressed in 1 g and the forces are almost steady-state-like, increasing much more slowly with intrusion speed.

To conclude, analysis of granular speed fields, shear strain rates, and granular pressure distributions highlights significant differences between responses to intruding objects in 1 g and in microgravity. In the latter, the very low granular pressure enhances liquid-like flow and rate-dependent viscous shear stress. Our findings elucidate the sensitivity of the rheology of granular materials to external pressure, which has significant ramifications for modeling soil mechanics in low-gravity environments, such as lunar and Martian surfaces. While there is an overall conceptual agreement between our experimental and numerical observations, we find systematic deviations which suggest limitations of the numerical modeling in capturing local microstructural rearrangements. We propose that these limitations warrant further development of constitutive models.

This study lays down a basis for understanding granular rheology in microgravity, offering insights into the complex interplay of granular pressure, inertial number, and viscous friction. As such, our results should encourage better design of granular flow models for extraterrestrial applications and pave the way for future experiments, including those on the Space Station, to explore granular shear rheology under sustained low-gravity conditions.

Author contributions

M. Hou and Q. Wu carried out the experiment and the data analyses; X. Cheng, S. Yang, X. Ze and Hao Li conducted the modeling and the simulation; M. Hou, X Cheng and R. Blumenfeld wrote and edited the manuscript. M. Hou led the project.

Data availability

The data that support the findings of this study are available from the corresponding author, M. Hou, upon reasonable request.

Competing interests

The authors declare no competing interests.