Thermo-viscous feedback in heat transport through fracture-matrix systems with finite thermal capacity

arXiv:2610.22531 · physics.flu-dyn, physics.geo-ph · Submitted 2026-10-15 · Read on arXiv

Listen

Transcript

Introduction to the show: ident: Physics. The day in general physics: instruments, methods and results across the field.

Kai: I'm Kai.

Mira: And I'm Mira, here with Lev.

Kai: We're looking at the paper: "Thermo-viscous feedback in heat transport through fracture-matrix systems with finite thermal capacity".

Mira: Here's the gist of the paper. In geothermal systems, injected fluid flows along preferential fracture pathways and exchanges heat with matrix blocks The reduced model gives parsimonious semi-analytical predictions with physically based parameters, verified against the nonlinear solution Thermo-viscous Feedback Mechanism The work analyzes thermo-viscous feedback in a planar fracture–matrix system where viscosity controls discharge through the integrated hydraulic resistance, while finite capacity enters through an effective single-timescale memory fixed by initial uptake and total capacity At fixed pressure difference, heating accelerates and broadens the front, whereas cooling delays and sharpens it The terminal discharge depends only on the endpoint viscosity ratio, while matrix uptake, storage and contrast determine its approach time At fixed rate, viscosity alters only the pressure drop, not the thermal response Governing Equations and Models The problem is formulated over a planar fracture of length L and half-aperture b, bounded by two identical rock-matrix domains of thickness hm Heat transport in the fracture follows the macroscopic two-dimensional advection–diffusion model The matrix temperature satisfies a diffusion equation where the matrix is impermeable, so heat is transported there by conduction alone Temperature and heat flux are continuous at the fracture–matrix interface, with conductivities ki = ΓiκT i The fluid is incompressible and Newtonian, with dynamic viscosity µ(Tf) controlling the flow Hydraulic Closure and Viscosity Dependence The hydraulic closure is determined by the local dimensionless hydraulic mobility mH(x, t), which depends on the temperature field Under a fixed pressure difference, the normalized velocity U(t) is given by U(t) = 1 / L integral x/L mH(x, t) - 1 The local dimensionless hydraulic mobility is mH(x, t) = 3 / b cubed integral z squared mµ

Θf (x, z, t): dz Temperature variations across the aperture enter the hydraulic response through the Poiseuille-weighted local mobility mH, whereas longitudinal variations combine through the harmonic average in Eq.

Kai: (2.19) Finite Matrix Response and Timescales The finite matrix can be represented through its response to the interfacial heat flux, introducing a wall-exchange kernel Hbm(qx, s) The slab response in Eq.

Mira: (2.25) is the classical bounded-conduction transfer function Characteristic timescales include the matrix relaxation time t1 = 1 / νex The ratio t1/ta defines the matrix relaxation number Π1 = t1/ta Finite matrix storage is measured by the matrix capacity number Cm = σhm / b Perturbation Hierarchy and Reconstruction The analysis uses a perturbation hierarchy separating matrix storage from the hydraulic response it induces The exponential velocity reconstruction of Eq.

Kai: (3.11) controls front-moment errors through the velocity error in a front-moment limit model with rapid transverse equilibration The reconstructed velocity urec is defined by urec / u0 = exp λU1 + lambda squared U2 - 1/2 U Key Findings and Comparisons The integrated hydraulic resistance links viscosity to discharge, with a terminal velocity fixed by the injected-to-initial viscosity ratio under fixed pressure difference Matrix diffusion and capacity, together with the contrast itself, set the timescales over which that endpoint is approached The operating protocol selects the hydraulic observable affected by viscosity: discharge at fixed pressure difference, the fracture pressure drop at fixed rate Under strong matrix uptake, the isoviscous response collapses onto a two-variable similarity surface The tempered effective memory preserves early diffusion and total storage and provides elementary crossover laws The exact slab resolves the planar geometry and supplies the comparisons in Sections 5.6 and 5.7 Model Sensitivity Analysis The comparison between the tempered closure and the exact slab shows that tempering for elementary crossover laws is useful, while the slab is better for quantitative timing, tails and spreading The first-order RMS discrepancy remains below 5% for λ ≤ 1, Π1 ≥ 1 Constitutive sensitivity affects breakthrough timing more than outlet RMS; quantitative timing at large temperature contrasts can use Vogel in the nonlinear solver Applications and Limitations The framework provides a controlled baseline for separating these mechanisms from geometrical, mechanical, buoyant, and multiphase effects and for subsequent upscaling Applications include breakthrough prediction and interpretation of pressure histories at prescribed rate The physical assumptions delimit the use of this baseline, as roughness, deformation and reactions require geometry and exchange updates The evidence comprises analytical and numerical verification, the calibrated numerical benchmark of Okoroafor and Horne (2022) for response shape Numerical Implementation Details The solver uses backward differentiation formulas (BDF) with Newton iteration The numerical implementation uses a sparse semi-discrete transport operator and the first-order expansion of Eq.

Paper summary: Kai: So, we just talked about the general idea of this paper, which is that injected fluid flows along fractures and exchanges heat with the surrounding rock matrix. Mira And what this work claims is that reduced models often neglect temperature-dependent viscosity and assume unlimited matrix storage in the matrix. Kai This paper analyzes thermo-viscous feedback where viscosity controls discharge through integrated hydraulic resistance, while finite capacity enters through an effective single timescale memory fixed by initial uptake and total capacity. Mira It sounds like they're showing how heating accelerates and broadens the front at fixed pressure difference, but cooling actually delays and sharpens it instead.

Mira: What matters is how they handle that feedback loop; at fixed pressure difference, heating accelerates and broadens the front, whereas cooling delays and sharpens it instead. Kai That's a crucial distinction because it shows that the terminal discharge depends only on the endpoint viscosity ratio under fixed pressure difference, while matrix uptake, storage, and contrast determine how fast you get to that endpoint.

Lev: I wonder about the timescales they are using; they mention characteristic timescales like t1 equals one over nuex squared, which relates to matrix relaxation time, and the ratio of t1 over ta defines a matrix relaxation number pi one. Kai And that ratio, pi one, tells you how much storage matters compared to the transport timescale.

Mira: The perturbation hierarchy they use separates the matrix storage from the hydraulic response it induces, and they use an exponential velocity reconstruction of eq three point one one to control front-moment errors through the velocity error in a front-moment limit model with rapid transverse equilibration. Kai So they are essentially trying to fix errors in the way heat moves across that fracture interface by looking at how fast things equilibrate perpendicular to the flow direction.

Lev: When you get down to the matrix response, they introduce a wall-exchange kernel hbm(qx, s) which represents the slab response in eq two point two five, which is the classical bounded-conduction transfer function

Conclusion: Kai: So we're wrapping up this look at "Thermo-viscous feedback in heat transport through fracture-matrix systems with finite thermal capacity". Mira Basically, this paper is about how temperature changes the fluid flow through a fracture and how that back-and-forth affects the whole system.

Lev: It looks like they're focusing on that viscosity changing the discharge, which is important because in real hardware, you can't just assume constant properties. Kai Yeah, it’s about how temperature drives the fluid speed through a crack and that speed feeds back into how much heat gets trapped in the rock matrix.

Mira: The authors are looking at these finite capacity effects, which means they aren't just looking at an infinite reservoir of stuff to store heat in. Lev That’s key because real systems have limits on how much thermal energy they can hold before things get weird.

Kai: It suggests that the way you measure the flow or the pressure drop really depends on how much heat the matrix has already soaked up. Mira And it points to a situation where heating doesn't just make things faster, it can actually slow down a process if cooling kicks in later.

Lev: If you were building this on a real chip or something, you’d need to be careful because the timing of the thermal response is what really matters for error correction. Kai Exactly. The next thing we look at is how this specific feedback mechanism plays out under different conditions, like when you change the rate of heating versus cooling.

Department of Civil, Chemical, Environmental, and Materials Engineering at the University of Bologna · Department of Energy Science & Engineering at Stanford University · Univ Rennes, CNRS, Géosciences Rennes

physics.flu-dyn, physics.geo-ph

Submitted: 2026-10-15

Updated: 2026-10-28

Comments: 24 pages. Published version in Advances in Water Resources. Open access under CC BY 4.0

License: http://creativecommons.org/licenses/by/4.0/

Importance score: 90/100

The gist: In geothermal systems, injected fluid flows along preferential fracture pathways and exchanges heat with matrix blocks The reduced model gives parsimonious semi-analytical predictions with physically

Key concepts

Thermo-viscous Feedback Mechanism: This describes the interplay where viscosity controls fluid discharge through hydraulic resistance, and finite thermal capacity influences the system's memory. Heating speeds up flow, while cooling slows it down, leading to complex changes in the transport front over time.
Hydraulic Closure and Viscosity Dependence: The local mobility of fluid flow depends on temperature. This relationship is captured by a dimensionless hydraulic mobility function that incorporates the fluid's dynamic viscosity. Temperature variations across the fracture aperture directly affect how easily fluid moves through it.
Finite Matrix Response and Timescales: The matrix acts as a storage element with finite capacity, introducing characteristic relaxation times. These timescales, such as the matrix relaxation time ($t_1$), determine how long it takes for the system to reach its final state, which is governed by the ratio of this time to other relevant timescales.

Terminology

Benchmark — 23× in this paper · explained in 2 episodes
Verification — 10× in this paper · explained in 2 episodes
Accuracy — 9× in this paper · explained in 1 episode
Reconstruction Error — 5× in this paper · explained in 2 episodes
Anisotropy — 3× in this paper · explained in 4 episodes
Consistency — 3× in this paper · explained in 3 episodes
Parameterization — 3× in this paper · explained in 2 episodes
Trace — 3× in this paper · explained in 2 episodes
Diffusion Model — 2× in this paper · explained in 10 episodes
Feedback Loop — 2× in this paper · explained in 2 episodes
Calibration — 1× in this paper · explained in 9 episodes
Decoupling — 1× in this paper · explained in 2 episodes
Fidelity — 1× in this paper · explained in 2 episodes
Generative AI — 1× in this paper · explained in 2 episodes
computational cost — 1× in this paper · explained in 2 episodes

Summary

In geothermal systems, injected fluid flows along preferential fracture pathways and exchanges heat with matrix blocks The reduced model gives parsimonious semi-analytical predictions with physically based parameters, verified against the nonlinear solution

Thermo-viscous Feedback Mechanism

The work analyzes thermo-viscous feedback in a planar fracture–matrix system where viscosity controls discharge through the integrated hydraulic resistance, while finite capacity enters through an effective single-timescale memory fixed by initial uptake and total capacity At fixed pressure difference, heating accelerates and broadens the front, whereas cooling delays and sharpens it The terminal discharge depends only on the endpoint viscosity ratio, while matrix uptake, storage and contrast determine its approach time At fixed rate, viscosity alters only the pressure drop, not the thermal response

Governing Equations and Models

The problem is formulated over a planar fracture of length L and half-aperture b, bounded by two identical rock-matrix domains of thickness hm Heat transport in the fracture follows the macroscopic two-dimensional advection–diffusion model The matrix temperature satisfies a diffusion equation where the matrix is impermeable, so heat is transported there by conduction alone Temperature and heat flux are continuous at the fracture–matrix interface, with conductivities ki = ΓiκT i The fluid is incompressible and Newtonian, with dynamic viscosity µ(Tf) controlling the flow

Hydraulic Closure and Viscosity Dependence

The hydraulic closure is determined by the local dimensionless hydraulic mobility mH(x, t), which depends on the temperature field Under a fixed pressure difference, the normalized velocity U(t) is given by U(t) = 1 / L integral x/L mH(x, t) - 1 The local dimensionless hydraulic mobility is mH(x, t) = 3 / b cubed integral z squared mµ[Θf (x, z, t)] dz Temperature variations across the aperture enter the hydraulic response through the Poiseuille-weighted local mobility mH, whereas longitudinal variations combine through the harmonic average in Eq. (2.19)

Finite Matrix Response and Timescales

The finite matrix can be represented through its response to the interfacial heat flux, introducing a wall-exchange kernel Hbm(qx, s) The slab response in Eq. (2.25) is the classical bounded-conduction transfer function Characteristic timescales include the matrix relaxation time t1 = 1 / νex The ratio t1/ta defines the matrix relaxation number Π1 = t1/ta Finite matrix storage is measured by the matrix capacity number Cm = σhm / b

Perturbation Hierarchy and Reconstruction

The analysis uses a perturbation hierarchy separating matrix storage from the hydraulic response it induces The exponential velocity reconstruction of Eq. (3.11) controls front-moment errors through the velocity error in a front-moment limit model with rapid transverse equilibration The reconstructed velocity urec is defined by urec / u0 = exp λU1 + lambda squared U2 - 1/2 U

Key Findings and Comparisons

The integrated hydraulic resistance links viscosity to discharge, with a terminal velocity fixed by the injected-to-initial viscosity ratio under fixed pressure difference Matrix diffusion and capacity, together with the contrast itself, set the timescales over which that endpoint is approached The operating protocol selects the hydraulic observable affected by viscosity: discharge at fixed pressure difference, the fracture pressure drop at fixed rate Under strong matrix uptake, the isoviscous response collapses onto a two-variable similarity surface The tempered effective memory preserves early diffusion and total storage and provides elementary crossover laws The exact slab resolves the planar geometry and supplies the comparisons in Sections 5.6 and 5.7

Model Sensitivity Analysis

The comparison between the tempered closure and the exact slab shows that tempering for elementary crossover laws is useful, while the slab is better for quantitative timing, tails and spreading The first-order RMS discrepancy remains below 5% for λ ≤ 1, Π1 ≥ 1 Constitutive sensitivity affects breakthrough timing more than outlet RMS; quantitative timing at large temperature contrasts can use Vogel in the nonlinear solver

Applications and Limitations

The framework provides a controlled baseline for separating these mechanisms from geometrical, mechanical, buoyant, and multiphase effects and for subsequent upscaling Applications include breakthrough prediction and interpretation of pressure histories at prescribed rate The physical assumptions delimit the use of this baseline, as roughness, deformation and reactions require geometry and exchange updates The evidence comprises analytical and numerical verification, the calibrated numerical benchmark of Okoroafor and Horne (2022) for response shape

Numerical Implementation Details

The solver uses backward differentiation formulas (BDF) with Newton iteration The numerical implementation uses a sparse semi-discrete transport operator and the first-order expansion of Eq. (3.5) uses the leading-order state to determine velocity The solver for hydraulic histories uses backward Euler (BDF1) with Newton iteration

Literature Context and Benchmarks

The analysis is benchmarked against published numerical benchmarks, such as the calibrated numerical benchmark of Okoroafor and Horne (2022) for response shape Literature contrasts are listed in Appendix D.5 The literature inputs and parameter context are provided in Appendix D.5 The fixed-rate Vogel check integrates its viscosity on the same calibrated leading-order temperature profile as Fig. 11(c), using 4000 longitudinal points and 250 logarithmic times over 0.02–5.

Improvements for AI systems

Modeling of enhanced geothermal systems can be improved by incorporating a finite matrix memory kernel that accounts for finite-matrix memory and thermal capacity accessible to the flow is finite. This allows AI systems to predict long-term thermal breakthrough accurately by distinguishing between semi-infinite and finite storage responses.
Development of reduced-order modeling algorithms can be enhanced by using an exponential velocity reconstruction that preserves the exact terminal discharge and bounds front-moment errors within the front-moment limit model. This capability enables AI to generate more reliable predictions for outlet temperatures even when using simplified models.
AI systems can be refined to incorporate viscosity-dependent feedback loops by implementing a perturbation hierarchy that separates matrix storage from the hydraulic response it induces. This allows the system to explicitly model how heating accelerates and broadens the front, whereas cooling delays and sharpens it at fixed pressure differences.
The AI can be trained on a framework where different closures are compared, such as the exact slab versus tempered closure, allowing for quantitative assessment of which closure is more appropriate for specific scenarios like breakthrough timing, tails and front spreading.
AI systems can perform sensitivity analysis on constitutive laws by using the defined metric Emu, which measures the difference between viscosity laws, such as comparing the exponential and Vogel solutions across different temperature contrasts. This allows researchers to determine whether a specific viscosity model is more sensitive to changes in breakthrough timing or outlet RMS.

Abstract

In geothermal systems, injected fluid flows along preferential fracture pathways and exchanges heat with matrix blocks. Reduced models often neglect temperature-dependent viscosity and assume unlimited matrix storage, although the thermal capacity accessible to the flow is finite. This work analyzes thermo-viscous feedback in a planar fracture-matrix system where viscosity controls discharge through the integrated hydraulic resistance, while finite capacity enters through an effective single-timescale memory fixed by initial uptake and total capacity. The reduced model gives parsimonious semi-analytical predictions with physically based parameters, verified against the nonlinear solution. A perturbation hierarchy separates matrix storage from the hydraulic response it induces, while an exponential velocity reconstruction preserves the exact terminal discharge and bounds front-moment errors within the front-moment limit model. At fixed pressure difference, heating accelerates and broadens the front, whereas cooling delays and sharpens it. The terminal discharge depends only on the endpoint viscosity ratio, while matrix uptake, storage and contrast determine its approach time. At fixed rate, viscosity alters only the pressure drop, not the thermal response. Finite thickness controls the transition to capacity-limited equilibration, and strong matrix uptake collapses the isoviscous response onto a two-variable similarity surface.