Preprint
Article

This version is not peer-reviewed.

Nonequilibrium Thermodynamics of Coffee Brewing Through Coupled Heat and Mass Transfer, Chemical Potential Gradients, and Entropy Production

Submitted:

11 November 2026

Posted:

13 November 2026

You are already at the latest version

Abstract
Coffee brewing is a transient solid–liquid extraction process in which thermal transport, intraparticle diffusion, interfacial dissolution, and fluid motion interact within a heterogeneous porous medium. Although existing mathematical descriptions have addressed coffee extraction kinetics, double-porosity transport, and effective equilibrium desorption, a unified thermodynamic formulation connecting these mechanisms remains desirable. This study develops a theoretical framework for coffee brewing based on nonequilibrium thermodynamics, chemical-potential gradients, coupled heat and mass transfer, and entropy production. The coffee bed is represented as a porous continuum containing a mobile aqueous phase and a heterogeneous solid phase, while intraparticle transport and solid–liquid interfacial exchange are incorporated through thermodynamically constrained constitutive relations. Species conservation, energy conservation, Darcy-scale fluid transport, and the local entropy balance are formulated consistently. A closed, isothermal immersion model is further derived to establish the monotonic decrease of total Gibbs free energy during spontaneous extraction. Linearization around the equilibrium state yields analytical expressions for the characteristic relaxation modes of a multicomponent extraction system. Dimensionless groups are introduced to distinguish thermal, diffusive, interfacial, and advective limitations. The resulting formulation provides a theoretical basis for interpreting coffee brewing as a coupled irreversible process without introducing unverified experimental data or claiming empirical validation.
Keywords: 
;  ;  ;  ;  ;  ;  ;  ;  ;  

1. Introduction

Coffee brewing is a physicochemical operation in which water interacts with a porous and chemically heterogeneous material produced through the roasting and grinding of coffee beans. The resulting beverage is not simply water containing a uniformly dissolved substance. It is a multicomponent liquid extract generated through the simultaneous occurrence of wetting, dissolution, diffusion, interfacial transfer, convection, and, depending on the brewing method, pressure-driven flow. The chemical composition of the resulting liquid changes continuously during preparation, while the physical structure of the coffee particles and the temperature distribution within the brewing system influence the rate and extent of extraction.
The extraction process is affected by several interacting variables, including the particle-size distribution, particle porosity, roast characteristics, coffee-to-water ratio, brewing temperature, solvent composition, contact time, flow rate, bed permeability, and the hydrodynamic conditions established by the brewing apparatus. These parameters cannot always be treated independently. A change in particle size, for example, can modify the accessible interfacial area and the internal diffusion length while simultaneously changing the hydraulic resistance of a packed bed. Similarly, changing water temperature affects not only the thermodynamic properties of the dissolved constituents but also the viscosity, density, and transport properties of the liquid phase.
The physicochemical complexity of coffee extraction has been discussed in the literature through several complementary approaches. Cordoba Castro and colleagues reviewed the influence of brewing conditions on the physicochemical characteristics and flavor of coffee beverages, emphasizing temperature, pressure, water quality, particle size, and coffee-to-water ratio as important determinants of extraction behavior [1]. Their analysis also identified the need for a more detailed physical interpretation of mass- and energy-transport phenomena. This observation is particularly relevant to the present study because the purpose here is to develop a formulation in which thermal and compositional changes are treated as interacting thermodynamic processes rather than as unrelated experimental variables.
An important contribution to the mathematical understanding of coffee brewing was made by Moroney and colleagues, who developed an experimentally validated multiscale extraction model for roasted and ground coffee [2]. Their description treated the coffee bed as a saturated, static, doubly porous medium and distinguished rapid release from damaged or readily accessible regions of coffee grains from slower extraction associated with the internal structure of the particles. The model established that observed extraction profiles can arise from multiple physical mechanisms operating on different timescales. Its significance extends beyond the particular experimental conditions because it provides a mathematical structure for relating particle-scale phenomena to the macroscopic behavior of a coffee bed.
In subsequent work, Moroney and colleagues examined coffee extraction kinetics in a well-mixed system using the double-porosity framework [3]. Their analysis connected dissolution and transport within the coffee grains to the observed extraction behavior in both dilute suspensions and packed beds. A further asymptotic study considered the dominant mechanisms governing extraction from a packed bed and demonstrated the relevance of intraparticle diffusion and advective transport through the bed [4]. These investigations establish that a satisfactory mathematical description must distinguish between the mobility of compounds within coffee particles and their transport through the surrounding liquid.
A different but complementary modeling perspective was introduced by Liang, Chan, and Ristenpart, who developed a pseudo-equilibrium desorption model for full-immersion brewing [5]. Their model related total dissolved solids and extraction yield to an effective equilibrium constant and demonstrated that an equilibrium-based description can provide useful relationships for immersion brewing. The distinction between this approach and a transient transport model is important. A final or effective equilibrium state does not independently specify the time-dependent thermal and mass-transfer processes through which the system approaches that state. Moreover, an apparent plateau in a bulk measurement does not necessarily imply that every individual chemical constituent has reached its own phase equilibrium.
Experimental investigations of espresso extraction provide further evidence of the interdependence of brewing parameters. Schmieder and colleagues examined the influence of flow rate, particle size, and temperature on the extraction kinetics of trigonelline, caffeine, 5-caffeoylquinic acid, and total dissolved solids [6]. Their results illustrate that the behavior of a brewed coffee cannot always be represented by one extraction curve or one universal rate coefficient. The extraction response depends on the identity of the compound and on the interaction between the operating conditions. This observation supports the use of species-specific variables in mathematical models.
The chemical composition of brewing water also has thermodynamic significance. Hendon and colleagues investigated the role of dissolved cations in coffee extraction and discussed interactions between mineral ions and coffee-derived compounds [7]. Their work demonstrates that the solvent cannot always be treated as chemically inert pure water. The presence of dissolved ions can influence molecular interactions, solution activities, and extraction behavior. Consequently, the thermodynamic state of the liquid phase may depend on water composition as well as on the concentration of extracted coffee constituents.
Despite these developments, the mathematical description of coffee brewing still presents an opportunity for a more explicit connection between extraction kinetics and thermodynamic irreversibility. A mass-transfer model can predict the movement of solutes, but it does not automatically identify the chemical-potential differences driving that movement. An energy-balance model can describe temperature evolution, but it does not by itself determine the entropy generated by thermal and compositional transport. Similarly, an equilibrium desorption model can describe an effective final state without necessarily resolving the local thermodynamic forces operating during the preceding transient process.
Nonequilibrium thermodynamics provides a general framework for making these connections. Its central idea is that irreversible processes can be described through the relationship between thermodynamic forces and associated fluxes. Heat conduction is driven by temperature-related forces, diffusion is driven by chemical-potential differences, and interfacial transfer is driven by the thermodynamic tendency of a constituent to move between phases. When more than one irreversible process occurs simultaneously, cross-coupling may arise. Onsager’s reciprocal relations provide a theoretical basis for describing such coupling in systems sufficiently close to equilibrium and subject to the assumptions of linear irreversible thermodynamics [8]. A systematic treatment of the entropy balance, transport processes, and thermodynamic forces is provided by de Groot and Mazur [9].
The application of these ideas to coffee brewing requires a distinction between local and global equilibrium. A porous coffee bed may be sufficiently close to local thermodynamic equilibrium for temperature, chemical potential, and other state variables to be defined within a representative volume. Nevertheless, the entire brewing system is generally not at equilibrium. Temperature gradients may persist, soluble substances may continue to move from the coffee particles into the liquid, and pressure-driven flow may transport dissolved species through the bed. Local equilibrium is therefore a modeling assumption that permits the use of thermodynamic state functions; it must not be interpreted as a claim that extraction itself is an equilibrium process.
Another important distinction concerns the thermodynamic driving force for extraction. Concentration gradients are often used in mass-transfer models because they provide a practical description of diffusion. However, the chemical potential of a constituent depends on its activity, standard state, temperature, and the composition of the surrounding mixture. A concentration-based model can therefore be interpreted as a restricted approximation to a more general chemical-potential formulation. This distinction becomes important when coffee extract is sufficiently concentrated, when water composition changes, or when interactions among dissolved components influence their activities.
The objective of the present study is to formulate a theoretical description of coffee brewing in which heat transfer, intraparticle transport, interfacial extraction, liquid-phase diffusion, hydrodynamic transport, and entropy production are connected within one framework. The study develops species-specific balance equations, introduces chemical-potential-based interfacial transfer laws, examines the thermodynamic admissibility of coupled heat and mass fluxes, and derives analytical consequences for a closed immersion system. It also establishes dimensionless groups that can be used to identify the relative importance of thermal, diffusive, interfacial, and advective mechanisms.
The proposed formulation is intended as a theoretical extension of established coffee extraction models, not as a replacement for their experimentally supported descriptions. Its principal contribution is the organization of the relevant processes through a common thermodynamic structure. The formulation does not claim experimentally measured coefficients, universal extraction constants, or validation under specific brewing conditions. Instead, it establishes the equations, assumptions, analytical restrictions, and parameter dependencies that a future quantitative implementation would need to satisfy.

2. Materials and Methods

The present investigation is an analytical and theoretical study. No new coffee samples were prepared, no brewing experiments were conducted, and no laboratory measurements or sensory evaluations were generated. The mathematical formulation is developed from conservation of mass, conservation of energy, chemical-potential relations, porous-medium transport, and the local entropy balance. Established principles of nonequilibrium thermodynamics are used as the theoretical foundation, while the coffee-specific model is constructed by combining the solid–liquid character of extraction with the multiscale geometry of a porous coffee bed.
The system is represented by a macroscopic porous domain Ω containing roasted and ground coffee particles and an interconnected liquid phase. The liquid occupies a fraction ε of the bulk volume, while the remaining volume is associated with the solid coffee structure and its internal void regions. The macroscopic bed may be considered fully saturated during the principal derivation. This assumption simplifies the representation of the liquid phase and is not intended to describe every stage of initial wetting or partially saturated brewing.
The liquid-phase velocity, measured as a superficial volume flux, is denoted by u. The liquid pressure is p, the temperature is T, and the porosity is ε. The specific external surface area of coffee particles per unit bulk bed volume is denoted by as. The dry coffee mass per unit bulk volume is ρb. For each extractable constituent i, the liquid concentration is ci, the solid-phase molar loading is qi, and the associated chemical potentials in the solid and liquid phases are μ i s and μ i l , respectively.
The constituent index i is retained throughout the formulation. This is necessary because the chemical constituents of roasted coffee differ in their molecular sizes, solubilities, diffusivities, activity coefficients, and interactions with the solid matrix and solvent. Although total dissolved solids provide a useful aggregate measure of extraction, they do not uniquely determine the distribution of individual constituents. A species-resolved formulation is therefore adopted as the fundamental description, with aggregate quantities introduced only when useful.
The model is developed at two related scales. At the macroscopic scale, the coffee bed is represented by a porous continuum with effective transport properties. At the particle scale, internal diffusion and the transfer of solutes from the particle interior to the surrounding liquid are treated through effective intraparticle and interfacial transport relations. These two descriptions are connected through the interfacial mass flux. This approach follows the general logic of volume averaging and multiscale transport theory, in which microscopic processes are represented through effective macroscopic equations [2,9,11].
The main assumptions are as follows. The liquid mixture is described by local thermodynamic state variables. The solid coffee matrix is treated as a heterogeneous material with effective chemical potentials for extractable components. The principal formulation neglects chemical reactions, degradation, and volatile partitioning, although extension terms for these processes are discussed. The liquid is treated as a continuum with effective transport coefficients, and the coffee bed is assumed to retain a fixed geometry during the basic derivation. Finally, the baseline thermal model assumes local thermal equilibrium between the solid and liquid phases, so that they share a common local temperature. The consequences of relaxing this assumption are considered separately.
The local thermal-equilibrium approximation is appropriate only when interphase heat transfer occurs sufficiently rapidly compared with the macroscopic thermal evolution of the system. It is not a universal condition for every brewing configuration. If the thermal equilibration time between a coffee particle and the surrounding liquid is not negligible, separate solid- and liquid-temperature equations are required. The present study uses the common-temperature formulation as the principal model because it allows the coupling between chemical extraction and thermal transport to be established without introducing additional unmeasured interphase parameters.

2.1. Thermodynamic State Variables and Chemical Potential

For a chemical species i in a phase α, where α may denote the solid-associated environment or the liquid phase, the chemical potential is written as
μ i α T , p , x = μ i ∘ , α T , p + RT ln a i α .
Here, μ i ∘ , α is the standard chemical potential in the selected phase-specific reference state, R is the universal gas constant, and a i α is the dimensionless activity of the constituent. The activity is related to the composition through an activity coefficient. For a mole-fraction representation,
a i α = γ i α x i α ,
where γ i α is the activity coefficient and x i α is the corresponding mole fraction.
In a dilute liquid solution, activity may be approximated using a concentration-based standard state:
a i l = γ i l c i c ∘ ,
where c∘ is the reference concentration. This relation is a convenient way of writing activity in terms of concentration. It does not imply that every coffee extract behaves as an ideal dilute solution. In concentrated extracts or multicomponent systems, the activity coefficient may vary with temperature and composition.
The chemical potential of a component associated with the coffee matrix requires a different interpretation. A compound may be present in a readily soluble region, associated with a polymeric structure, trapped in a cellular region, or located in an internal pore. Its effective solid-phase chemical potential must represent the thermodynamic state of that constituent in its local environment. The quantity μ i s is therefore treated as an effective partial molar Gibbs energy associated with the extractable inventory rather than as the chemical potential of a pure crystalline solid.
The transfer driving force is defined by
Δ μ i = μ i s − μ i l .
With this convention, Δμi > 0 indicates that transfer from the solid-associated state into the aqueous phase is thermodynamically favored. When Δμi < 0, the opposite direction is favored. At interfacial chemical equilibrium,
μ i s = μ i l ,     Δ μ i = 0 .
The equality is defined using consistent phase-specific standard states. It does not require the numerical standard chemical potentials in the two phases to be identical.
For an idealized dilute liquid phase, an effective equilibrium concentration c i * can be defined by the condition that the liquid chemical potential at c i * equals the current solid-phase chemical potential. The liquid chemical potential at the actual concentration ci is then
μ i l c i = μ i l c i * + RT ln c i c i * . Using μ i s = μ i l c i * , the transfer driving force becomes
Δ μ i = RT ln c i * c i .
Equation (6) is valid under the ideal-dilute approximation and when c i * is defined consistently with the current solid state and temperature. In a non-ideal solution, the more general relation is written in terms of activities:
Δ μ i = RT ln a i * a i l .
These expressions demonstrate why concentration and thermodynamic driving force are related but not interchangeable. The concentration difference c i * − c i is a useful kinetic approximation near equilibrium, while the logarithmic activity ratio expresses the thermodynamic driving force more generally.
The temperature dependence of the equilibrium state may be expressed through the standard Gibbs energy of transfer. Defining
Δ G i ∘ = μ i ∘ , l − μ i ∘ , s ,
the equilibrium activity ratio is
K i = a i l a i s = exp − Δ G i ∘ RT .
If the standard enthalpy of transfer is approximately constant over a restricted temperature interval, the temperature dependence of the equilibrium constant follows the van’t Hoff relation:
∂ ln K i ∂ T = Δ H i ∘ R T 2 .
Equation (10) provides a formal connection between temperature and the thermodynamic distribution of a constituent between phases. It does not imply that increasing temperature necessarily increases the measured extraction yield under all brewing conditions. The final amount extracted also depends on transport resistance, water composition, the changing solid matrix, and the hydrodynamic history of the system.

2.2. Multiscale Representation of the Coffee Particle

A coffee particle contains regions with different accessibility and transport characteristics. The outer surface, damaged cellular regions, connected pore spaces, and internal particle regions do not necessarily contribute to extraction at the same rate. The double-porosity model developed by Moroney and colleagues provides an established mathematical representation of rapid and slow extraction mechanisms within coffee grains [2,3,4].
The present framework retains this physical distinction but expresses the transfer mechanisms in terms of thermodynamic potentials. For an effective particle-scale description, let ξ denote the position inside an individual particle, and let c p , i ξ , t be the concentration of species i in the internal pore liquid. The internal porosity is ϕp, and qp,i represents the remaining solid-associated inventory per unit particle volume.
An internal dissolution rate Rd,i is introduced to represent the transfer of material from the solid matrix into the internal pore liquid. The local particle-scale balances are
∂ q p , i ∂ t = − R d , i ,
and
∂ ϕ p c p , i ∂ t + ∇ ξ ⋅ j p , i = R d , i .
Here, jp,i is the internal diffusive molar flux. The sum of Equations (11) and (12) gives
∂ ∂ t q p , i + ϕ p c p , i + ∇ ξ ⋅ j p , i = 0 .
Equation (13) expresses the conservation of the constituent within the particle, provided that no chemical reaction or additional source is present.
The internal dissolution process is represented by a chemical-potential-based relation:
R d , i = L d , i μ i s − μ i p T ,
where μ i p is the chemical potential of the constituent in the internal pore liquid and L d , i is a non-negative phenomenological coefficient. The relation is a near-equilibrium constitutive approximation. If the process operates far from equilibrium, the coefficient may depend on the state variables and the relation may need to be replaced by a nonlinear kinetic law.
The internal pore-liquid flux is described by
j p , i = L p , i − ∇ ξ μ i p T .
The coefficient L p , i represents effective particle-scale mobility. It includes the influence of internal porosity, tortuosity, and the local structure of the pore network. Under restricted ideal-solution and isothermal conditions, Equation (15) can be reduced to a Fick-type diffusion equation.
At the external particle surface, the internal pore liquid exchanges material with the surrounding liquid phase. Let n be the outward normal to the particle surface. The boundary condition is
j p , i ⋅ n = r b , i ,
where rb,i is the outward molar transfer rate per unit particle surface area. The external interfacial relation is
r b , i = L b , i μ i p − μ i l T ,
with L b , i ≥ 0 .
Equations (14)–(17) describe three successive physical stages: release from the solid-associated matrix, transport through the internal pore liquid, and transfer across the external particle–liquid interface. These stages are not necessarily equally limiting. A particle with a large internal diffusion resistance may exhibit slow extraction even when external interfacial transfer is rapid. Conversely, a particle with readily accessible internal constituents may be controlled by the external liquid film or the local chemical-potential difference.
The model does not require the three transport stages to have identical characteristic timescales. Instead, the corresponding mobilities and boundary conditions determine which process dominates under a given set of conditions. This is consistent with the general multiscale interpretation of coffee extraction established in previous modeling work [2,3,4], while introducing a common thermodynamic description for the associated driving forces.
In a simplified effective model, the internal and external resistances can be combined into an overall transfer resistance. If the successive transport stages are approximately linear, quasi-steady, and carry the same molar flux, an effective conductance may be represented schematically by
1 L i , eff = 1 L d , i + 1 L p , i + 1 L b , i ,
where the coefficients must first be expressed on a consistent flux and thermodynamic-force basis. Equation (18) is a series-resistance approximation, not a universal law for arbitrary heterogeneous particles. In a detailed particle model, the internal spatial gradients should be solved explicitly rather than replaced by a single effective coefficient.

2.3. Macroscopic Conservation of Dissolved Species

At the bed scale, the coffee matrix and the liquid phase are represented through volume-averaged quantities. The species balance in the liquid phase is
∂ ε c i ∂ t + ∇ ⋅ u c i + J i = a s r ¯ b , i .
Here, u is the superficial liquid velocity, Ji is the diffusive flux relative to the selected liquid reference velocity, and r ¯ b , i is the particle-surface transfer rate averaged over the relevant particle population. The effective surface area as is expressed per unit bed volume. If all particles have identical surface properties, the average may be replaced by the representative particle value.
The corresponding coarse-grained solid inventory balance is
∂ ρ b q i ∂ t = − a s r ¯ b , i ,
provided that qi represents the total amount of species i retained within the particles per unit dry coffee mass, including the internal partitioning represented in the particle model. Equations (19) and (20) are the bed-scale representation of the particle-to-liquid exchange.
Adding the two equations gives
∂ ∂ t ρ b q i + ε c i + ∇ ⋅ u c i + J i = 0 .
Equation (21) is the macroscopic conservation law for species i. The internal dissolution and particle-scale transport processes determine the value of the net surface flux, but they do not create or destroy the constituent. The transfer term appears with opposite signs in the solid and liquid balances and therefore cancels in the combined equation.
The particle-resolved equations and the effective bed-scale equations are alternative levels of description linked through volume averaging. They should not be added as independent duplicate inventories. When the internal particle equations are solved explicitly, their surface flux provides the source term in the liquid balance. When internal processes are not resolved, their combined effect is represented through the effective coefficient in Equation (20).
For a closed immersion system without external mass exchange, integration of Equation (21) over the total volume yields
d dt ∫ Ω ρ b q i + ε c i dV = 0 .
This relation states that the total amount of the constituent in the coffee solid and liquid phases remains constant when chemical reactions, evaporation losses, and external mass flows are absent.
For a flowing brewing system, the integral conservation law becomes
d dt ∫ Ω ρ b q i + ε c i dV = − ∫ ∂ Ω u c i + J i ⋅ n   dA .
The right-hand side represents the net outward transport of the constituent through the boundary. This expression is particularly relevant to percolation and espresso brewing, where dissolved material is continuously transported out of the coffee bed.
The total extraction fraction of constituent i can be defined as
X i t = 1 − ∫ Ω ρ b q i t   dV ∫ Ω ρ b q i 0   dV .
The corresponding extracted amount is
M i , ext t = ∫ Ω ρ b q i 0 − q i t dV .
These relations describe the depletion of the solid-phase inventory. They do not directly describe beverage concentration because the latter also depends on the liquid volume and the distribution of dissolved material.
For a selected group of soluble compounds, an aggregate extraction yield may be expressed as
E t = ∑ i ∈ S M i , ext t M i M coffee , dry ,
where Mi is the molar mass of constituent i, and S denotes the set of species included in the chosen analytical definition. This definition makes explicit that aggregate extraction yield depends on the selected species set and the method used to quantify them.
Total dissolved solids, by contrast, are an aggregate concentration measure. Under a simple mass-fraction definition,
w TDS = m dissolved m beverage .
If the beverage consists only of water and dissolved solids, then
m dissolved = w TDS 1 − w TDS m water .
The conversion between total dissolved solids and extraction yield therefore requires a mass balance and a precise definition of the water mass included in the calculation. Water retained by spent coffee grounds can influence the inferred extraction yield. Liang and colleagues discuss this issue within their full-immersion equilibrium model [5]. The present framework retains the distinction between solid depletion, beverage concentration, and phase equilibrium.

2.4. Chemical-Potential-Driven Liquid Transport

The liquid-phase diffusive flux is associated with spatial differences in chemical potential. A general thermodynamic expression for an independent species flux is
J i = ∑ j L ij m X j ,
where L ij m are phenomenological transport coefficients and Xj are the corresponding thermodynamic forces.
For a multicomponent liquid mixture, the diffusive fluxes are not all independent when they are measured relative to the barycentric velocity. A convenient formulation uses one reference component, denoted by n, and considers n − 1 independent diffusion fluxes. Their thermodynamic forces are
X i = − ∇ μ i − μ n T .
The chemical-potential difference relative to the reference component accounts for the compositional constraint on diffusion. This is important for a coffee extract because the liquid contains multiple dissolved substances rather than a single isolated solute.
Under an ideal dilute approximation and at constant temperature, Equation (29) may reduce to a Fickian relation:
J i = − D i eff ∇ c i .
The effective diffusion coefficient depends on the properties of the liquid and the porous medium. It may incorporate molecular diffusion, tortuosity, constriction, and the fraction of the pore space available for transport.
For a non-ideal mixture, the chemical-potential gradient contains contributions from more than the concentration gradient of the same species. Expanding the gradient gives
∇ μ i = ∂ μ i ∂ T c ∇ T + ∑ j ∂ μ i ∂ c j T , c k ≠ j ∇ c j .
The expression includes a direct temperature contribution and cross-compositional contributions arising from the dependence of activity coefficients on mixture composition. Therefore, the transport of one constituent may be influenced by the concentration gradients of other dissolved constituents.
This does not automatically imply a large cross-diffusion effect under every brewing condition. It establishes that a model based solely on independent Fickian diffusion is a restricted approximation. The magnitude of the additional terms depends on the thermodynamic properties of the liquid mixture and must be evaluated using appropriate activity-coefficient data.
The dependence of coffee extraction on water composition provides a relevant example. Dissolved cations can interact with coffee-derived compounds, influencing their molecular environment and potentially changing the thermodynamic conditions associated with dissolution [7]. In a more detailed model, the chemical potentials would therefore depend on the concentrations of water-derived ions as well as the extracted coffee constituents.

2.5. Thermal Energy Balance

The local temperature field is determined by heat storage, heat conduction, heat carried by moving liquid, and the thermal contributions associated with phase transfer and hydraulic dissipation.
Under the local thermal-equilibrium approximation, the solid and liquid phases share a common temperature T. The effective volumetric heat capacity is written as
C eff = 1 − ε ρ s c p , s + ε ρ l c p , l ,
where ρs and cp,s are the effective density and specific heat capacity of the solid phase, and ρl and cp,l are the corresponding properties of the liquid phase.
The macroscopic energy equation is expressed as
C eff ∂ T ∂ t + ρ l c p , l u ⋅ ∇ T = − ∇ ⋅ J q ' + ∑ i = 1 n − 1 h ‾ i J i + Q ext + Φ D − a s ∑ i Δ H i tr r ¯ i .
Here, Jq' is the reduced heat flux, h ‾ i is the appropriate partial molar enthalpy associated with the diffusive transport of species i, Qext represents external volumetric heating or cooling, and ΦD is the energy dissipated through pressure-driven flow. The quantity Δ H i tr denotes the enthalpy change associated with transfer of constituent i from its solid-associated state into the liquid phase, with a positive value corresponding to an endothermic transfer under the selected convention.
The reduced heat flux is separated from the enthalpy carried by diffusing species. This distinction is essential in multicomponent nonequilibrium thermodynamics because heat conduction and species diffusion may be coupled, and the energy transported by diffusion should not be counted twice. The reduced heat flux may be written in a simple Fourier approximation as
J q ' = − k eff ∇ T ,
where keff is the effective thermal conductivity of the porous coffee–water medium.
When the enthalpy carried by species diffusion is small relative to the bulk sensible-energy transport, its divergence may be neglected as a first approximation. Equation (34) then reduces to the commonly used effective heat equation:
C eff ∂ T ∂ t + ρ l c p , l u ⋅ ∇ T = ∇ ⋅ k eff ∇ T + Q ext + Φ D − a s ∑ i Δ H i tr r ¯ i .
The interfacial enthalpy term represents the thermal effect of transferring constituents between phases. Its magnitude is not assumed to be known. A parameterized implementation would require compound-specific or effective enthalpy data. If the thermal contribution of dissolution is negligible relative to the dominant heat-transfer terms, the corresponding term may be omitted, but that approximation should be stated explicitly.
The Darcy dissipation term is defined by
Φ D = − u ⋅ ∇ p ,
for the simplified pressure-driven flow description without gravitational contributions. For a passive porous medium, this term is non-negative when the velocity and pressure gradient satisfy Darcy’s law. It represents the conversion of mechanical energy associated with pressure-driven flow into internal energy.
The energy equation therefore contains several mechanisms that should not be confused. Heat conduction responds to the temperature gradient, advection transports sensible energy with the moving liquid, diffusion may carry partial molar enthalpy, dissolution may contribute an interfacial enthalpy change, and pressure-driven flow dissipates mechanical energy. The relative magnitudes of these terms depend on the geometry and operating conditions.
The local thermal-equilibrium approximation can be relaxed when the solid and liquid temperatures differ appreciably. In that case, separate energy equations may be written:
C s ∂ T s ∂ t = ∇ ⋅ k s eff ∇ T s + h sl a sl T l − T s + Q s ,
C l ∂ T l ∂ t + ρ l c p , l u ⋅ ∇ T l = ∇ ⋅ k l eff ∇ T l + h sl a sl T s − T l + Q l .
The terms hsl and asl represent the interphase heat-transfer coefficient and interfacial area, while Qs and Ql include appropriately partitioned thermal contributions. The interphase heat-transfer terms cancel when the two energy equations are added, as required by energy conservation. Their associated entropy production, however, remains non-negative:
σ sl T = h sl a sl T s − T l 1 T l − 1 T s ≥ 0 ,
for hsl ≥ 0 and positive absolute temperatures. This equation shows how the two-temperature model is connected to the same thermodynamic framework as the single-temperature approximation.

2.6. Hydrodynamic Transport Through the Porous Coffee Bed

In immersion brewing, the macroscopic liquid velocity within the coffee bed may be small after the initial wetting and agitation phases. In percolation and espresso brewing, however, liquid transport through the coffee bed is an essential part of the extraction process. A thermodynamic model that includes only diffusion cannot describe the movement of dissolved compounds through a continuously flowing bed.
At sufficiently low pore Reynolds numbers, the superficial liquid velocity may be represented by Darcy’s law:
u = − K η ∇ p ,
where K is the permeability tensor and η is the dynamic viscosity of the liquid. For an isotropic medium, K = KI, where K is the scalar permeability.
The permeability may depend on particle size distribution, packing structure, compaction, and the evolution of the bed during wetting. If the coffee bed is assumed mechanically rigid, permeability is treated as constant. For a deformable bed or a system in which fines migrate and modify the pore network, an evolving permeability law would be required.
Substituting Darcy’s law into Equation (37) gives
Φ D = η K u 2 ≥ 0
for an isotropic bed. For an anisotropic permeability tensor,
Φ D = η   u ⋅ K − 1 u ≥ 0 ,
provided that K is positive definite.
The flow field also influences the transport of dissolved constituents through the advective term in Equation (19). Increasing the liquid velocity changes the rate at which solutes are carried away from the bed, modifies local concentration profiles, and can change the relative importance of external and internal mass-transfer resistances. The effect of flow rate is therefore not represented solely by a change in contact time.
For espresso brewing, the coffee puck may also undergo mechanical compression and changes in pore structure. In such a situation, the assumption of constant permeability may become inadequate. A more complete model would couple the pressure field to porosity and permeability through a constitutive relation:
K = K ε , σ eff , d p ,
where σeff is an effective mechanical stress and dp represents a characteristic particle-size measure. Equation (44) is an illustrative constitutive dependence, not a calibrated permeability law. Its purpose is to identify the additional physical variables required when bed deformation is important.

2.7. Thermodynamic Forces and Entropy Production

The second law of thermodynamics requires that the local production of entropy associated with irreversible processes be non-negative. For a continuum system, the entropy balance can be written as
∂ s ∂ t + ∇ ⋅ J s = σ ,
where s is the local entropy density, Js is the entropy flux, and σ is the entropy-production rate per unit volume.
For the multicomponent liquid mixture, the appropriate thermodynamic forces depend on the choice of reference velocity and the definition of the heat flux. Using a reduced heat flux Jq' and n − 1 independent diffusive fluxes, a representative local entropy-production expression is
σ tr = J q ' ⋅ ∇ 1 T − ∑ i = 1 n − 1 J i ⋅ ∇ μ i − μ n T .
The first term describes entropy production due to heat conduction. The second represents the contribution of independent diffusion processes. The use of chemical-potential differences relative to a reference component accounts for the constraint on the sum of diffusive fluxes.
The local thermodynamic contribution from transfer between the coffee solid and liquid phases is
σ int = a s ∑ i r ¯ i Δ μ i T .
The pressure-driven contribution is
σ D = − u ⋅ ∇ p T .
For a system described by the simplified Darcy relation, this becomes
σ D = η T u ⋅ K − 1 u ≥ 0 .
The total entropy production, neglecting additional reaction and mechanical terms, is therefore expressed as
σ = σ tr + σ int + σ D .
Equation (50) identifies separate physical contributions from thermal transport, diffusion, interfacial extraction, and hydraulic dissipation.
Under Fourier conduction,
Jq' = −keff∇T,
and the thermal contribution becomes
σ q = J q ' ⋅ ∇ 1 T = k eff T 2 ∇ T 2 ≥ 0 .
The positivity follows from keff ≥ 0 and T > 0.
For a diagonal diffusion approximation,
J i = L ii m − ∇ μ i − μ n T ,
with L ii m ≥ 0 , the contribution of each independent species is
σ i = L ii m ∇ μ i − μ n T 2 ≥ 0 .
For interfacial transfer, a local linear relation is introduced:
r ¯ i = L ii I Δ μ i T ,     L ii I ≥ 0 .
Consequently,
σ int = a s ∑ i L ii I Δ μ i T 2 ≥ 0 .
Equations (51), (53), and (55) demonstrate that the selected constitutive relations are thermodynamically admissible within their linear-response assumptions.
The interfacial mobility need not remain constant under all extraction conditions. A nonlinear generalization can be written as
r ¯ i = L i I Δ μ i , T , c Δ μ i T ,
provided that
L i I Δ μ i , T , c ≥ 0 .
This condition preserves the non-negativity of the interfacial entropy-production contribution, even when the mobility depends on the state. It does not guarantee that the resulting kinetic law is quantitatively correct; its parameterization requires suitable physical or experimental information.
The total entropy production can therefore be expressed in a compact form:
σ = k eff T 2 ∇ T 2 + X m T L m X m + X I T L I X I + η T u ⋅ K − 1 u
where Xm denotes the independent bulk diffusion forces and XI represents the interfacial chemical-potential forces. The matrices Lm and LI must be positive semidefinite in their respective thermodynamic-force spaces. Equation (58) is the central thermodynamic admissibility condition of the proposed formulation.

2.8. Coupled Heat and Mass Transport

When thermal and diffusive processes occur simultaneously, a constitutive model that treats the corresponding fluxes as fully independent may be incomplete. Near local equilibrium, the fluxes may be represented as linear functions of the complete set of thermodynamic forces.
For one thermal flux and a set of independent diffusion fluxes, the coupled relation is
J q ' J 1 ⋮ J n − 1 = L qq L q 1 … L q , n − 1 L 1 q L 11 … L 1 , n − 1 ⋮ ⋮ ⋱ ⋮ L n − 1 , q L n − 1 , 1 … L n − 1 , n − 1 X q X 1 ⋮ X n − 1 ,
with
X q = ∇ 1 / T ,     X i = − ∇ μ i − μ n T .
The diagonal entries describe direct transport responses. The off-diagonal entries represent cross-coupling between heat and mass transport. Depending on the thermodynamic and microscopic properties of the system, a thermal force can contribute to a species flux, while a compositional force can contribute to the reduced heat flux.
The presence of cross-coefficients does not imply that every such effect is important in coffee brewing. The coefficients must be determined for the material and state under consideration. They should not be assigned arbitrary values merely to make the model appear more comprehensive.
Under the assumptions of Onsager’s reciprocal relations and an appropriate choice of conjugate fluxes and forces, the phenomenological matrix satisfies
L ij = L ji .
This symmetry is subject to the applicable microscopic reversibility and time-reversal conditions [8,9]. It should not be imposed indiscriminately on arbitrary empirical transport coefficients.
The total entropy production associated with the coupled transport matrix can be written as
σ m = X T LX .
Thermodynamic admissibility requires
X T L s X ≥ 0   for   all   X ,
where
L s = 1 2 L + L T
is the symmetric part of the matrix. In the absence of the relevant symmetry-breaking effects, Onsager reciprocity gives a symmetric matrix directly. The positivity condition is a separate restriction imposed by non-negative entropy production.
For a two-process case involving one reduced heat flux and one independent species flux, Equation (63) requires
L qq ≥ 0 ,     L ii ≥ 0 ,
and
L qq L ii − L qi   2 ≥ 0 .
Equation (66) limits the strength of the cross-effect. A coupling coefficient cannot be chosen independently of the direct transport coefficients without potentially violating the second law.
For coffee brewing, the physical interpretation of this coupled formulation is that the temperature field and the concentration field may influence one another through both thermodynamic properties and transport responses. A temperature difference modifies the chemical potential and mobility of dissolved constituents. Conversely, species transport can contribute to the total energy flux through the partial molar enthalpy carried by diffusion. The influence of these mechanisms should be determined from the constitutive properties rather than assumed to be quantitatively dominant.
The model also clarifies the distinction between a thermodynamic coupling and an empirical correlation. If an increase in temperature coincides with a change in extraction rate, this observation alone does not demonstrate an Onsager cross-effect. The correlation may arise from changes in viscosity, diffusivity, solubility, flow resistance, or interfacial mobility. A genuine cross-coefficient is a term in a constitutive relation between appropriately defined thermodynamic forces and fluxes, and its interpretation requires the assumptions of the linear theory.

3. Results

3.1. Conservation and Consistency of the Coupled Model

The first result of the formulation is the local conservation of each extractable constituent. The solid- and liquid-phase equations are
∂ ρ b q i ∂ t + a s r ¯ i = 0 ,
∂ ε c i ∂ t + ∇ ⋅ u c i + J i − a s r ¯ i = 0 .
Adding these equations eliminates the interfacial transfer term and yields
∂ ∂ t ρ b q i + ε c i + ∇ ⋅ u c i + J i = 0 .
This result is independent of the specific constitutive equation used for r ¯ i . The interfacial rate determines the speed of redistribution between phases, but mass conservation is imposed by the structure of the balance equations themselves.
For a closed system, the spatial integral of Equation (69) remains constant. For an open system, changes in the inventory are determined by the boundary flux. This establishes a direct mathematical distinction between immersion and percolation brewing without requiring different fundamental conservation laws.
The multiscale formulation introduces additional internal variables, but it does not change the conservation principle. When particle-scale dissolution, internal diffusion, and external surface transfer are resolved separately, the total amount of a species is conserved across the internal pore liquid, remaining solid inventory, and surrounding bulk liquid. The surface flux is the only transfer term connecting the individual particle to the macroscopic liquid domain.

3.2. Thermodynamic Direction of Interfacial Extraction

The chemical-potential difference determines the thermodynamic direction of interfacial transfer. Under the adopted sign convention, the local linear relation is
r i = L i I Δ μ i T .
If L i I > 0 , the sign of ri is the same as the sign of Δμi. Consequently, a positive chemical-potential difference produces transfer from the solid-associated state to the liquid phase, while a negative difference produces the opposite transfer.
This relationship is more general than an extraction law expressed solely through concentration. In the ideal dilute limit, Equation (6) gives
r i = L i I R ln c i * c i .
When ci is close to c i * , the logarithm can be linearized:
ln c i * c i ≈ c i * − c i c i * .
Therefore,
r i ≈ L i I R c i * c i * − c i .
Equation (73) has the form of a conventional linear mass-transfer relation,
r i = k i m c i * − c i ,
with an effective coefficient
k i m = L i I R c i * .
The result demonstrates that a concentration-difference mass-transfer relation can emerge as a near-equilibrium approximation to a chemical-potential-based law. The latter is more general because it retains the activity and thermodynamic reference-state dependence.
The effective equilibrium concentration c i * is not constant in every situation. It may change as the remaining solid inventory is depleted, as the temperature varies, or as the chemical composition of the water changes. A model that assumes a constant c i * over a complete brewing process therefore introduces an additional approximation.

3.3. An Analytical Result for Closed Immersion Brewing

A closed immersion system provides a useful setting for deriving analytical consequences of the thermodynamic formulation. Consider a single extractable species in a well-mixed system containing a fixed mass of dry coffee Ms, a liquid volume Vl, and a total solid–liquid interfacial area A. The temperature and pressure are assumed constant, and no liquid or solute crosses the external boundary.
Let q(t) denote the molar amount of the species remaining in the solid per unit mass of dry coffee, and let c(t) denote its dissolved concentration. The mass balances are
M s dq dt = − Ar ,
V l dc dt = Ar .
Adding these equations gives
d dt M s q + V l c = 0 .
Thus,
M s q t + V l c t = M s q 0 + V l c 0 .
Solving for the liquid concentration gives
c t = c 0 + M s V l q 0 − q t .
The liquid concentration is therefore determined by the solid depletion when the total system is closed.
The interfacial transfer relation is
r = L I Δ μ q , c , T T .
Combining Equations (76), (80), and (81) yields
dq dt = − A L I M s T Δ μ q , c 0 + M s V l q 0 − q , T .
An equilibrium state qeq is defined by
Δ μ q eq , c eq , T = 0 ,
together with the mass-conservation condition
M s q eq + V l c eq = M s q 0 + V l c 0 .
The existence and stability of this state depend on the thermodynamic properties of the phases and the available material inventory. Equation (83) defines phase equilibrium for the selected constituent, not necessarily an operationally desirable brewing endpoint.
Near equilibrium, the chemical-potential difference can be expanded around the equilibrium state. Along the mass-conservation constraint,
Δ μ q ≈ H q − q eq ,
where
H = d Δ μ dq eq
is the effective thermodynamic curvature along the constrained path. A locally stable equilibrium requires
H > 0 .
Substitution into Equation (82) yields
dq dt = − A L I H M s T q − q eq .
The solution is
q t − q eq = q 0 − q eq exp − t τ ext ,
where the characteristic relaxation time is
τ ext = M s T A L I H .
Equation (90) is an analytical result of the reduced model. It shows that the relaxation time depends on the solid inventory, temperature, interfacial area, interfacial mobility, and local thermodynamic curvature. It is not a universal extraction-time formula. In particular, L I and H may vary with temperature, composition, particle structure, and the state of extraction.
The exponential behavior in Equation (89) follows from linearization near equilibrium. It should not be interpreted as a claim that the complete extraction curve of real coffee is exponential. Early-stage release, multiscale intraparticle diffusion, evolving activity coefficients, and changing temperature can all produce non-exponential behavior. The result identifies the local relaxation structure of the proposed model near a stable equilibrium state.

3.4. Free-Energy Dissipation During Spontaneous Extraction

The reduced immersion system permits a further result concerning the total Gibbs free energy. Let gs(q,T,p) be the effective Gibbs free energy per unit mass of dry coffee associated with the solid-phase inventory and let gl(c,T,p) denote the corresponding liquid contribution per unit volume. The total Gibbs free energy is
G tot = M s g s q , T , p + V l g l c , T , p .
At constant temperature and pressure, the differential of this function is
d G tot = M s μ s   dq + V l μ l   dc .
Using the mass balances,
M s dq dt = − Ar ,     V l dc dt = Ar ,
the time derivative of total Gibbs free energy becomes
d G tot dt = − A μ s r + A μ l r .
Therefore,
d G tot dt = − Ar μ s − μ l .
By the definition of the transfer driving force,
d G tot dt = − Ar Δ μ .
Substituting the linear interfacial relation gives
d G tot dt = − A L I T Δ μ 2 ≤ 0 .
Equation (97) establishes that the Gibbs free energy decreases monotonically during spontaneous extraction within the reduced isothermal closed-system model, provided that the interfacial mobility is non-negative. The equality holds when the interfacial driving force vanishes or when the interfacial mobility is zero.
This result provides an explicit connection between the extraction kinetics and the second law. The process does not need to be interpreted as a sequence of equilibrium states. Rather, the system evolves through a succession of nonequilibrium states while its available Gibbs free energy decreases under the stated assumptions.
The corresponding interfacial entropy-production rate is
S ˙ int = Ar Δ μ T .
Using Equation (81),
S ˙ int = A L I T 2 Δ μ 2 ≥ 0 .
Multiplying Equation (99) by the constant temperature gives
T S ˙ int = A L I T Δ μ 2 .
Combining Equations (97) and (100) yields
d G tot dt = − T S ˙ int .
Equation (101) expresses the dissipation of Gibbs free energy in terms of entropy production for the closed isothermal model. It is a direct consequence of the balance equations and the chosen constitutive law. It should not be extended unchanged to a non-isothermal or open brewing system because heat and matter exchange across the external boundary introduce additional terms.
The result also clarifies why the thermodynamic equilibrium state should not be confused with the sensory optimum. The equilibrium condition identifies the state at which the net thermodynamic driving force for transfer vanishes within the selected model. The sensory optimum is a separate question involving the composition of the beverage and human perception. There is no general thermodynamic principle requiring maximum extraction or complete equilibrium to correspond to the preferred sensory outcome.

3.5. Multicomponent Extension and Relaxation Modes

The single-species result can be extended to a system containing n independently tracked constituents. Let
q = q 1 , q 2 , … , q n T , and define the vector of solid-to-liquid chemical-potential differences as
Δ μ = Δ μ 1 , Δ μ 2 , … , Δ μ n T .
The interfacial flux relation becomes
r = L I Δ μ T ,
where LI is a phenomenological interfacial mobility matrix. The matrix should be positive semidefinite in the relevant thermodynamic-force space. If cross-coupling between individual interfacial transfer processes is neglected, it may be approximated as diagonal.
The vector form of the solid inventory balance is
M s d q dt = − A r .
Along the set of mass-conservation constraints, the total Gibbs free energy gradient is
∇ q G tot = M s Δ μ .
The evolution equation can therefore be written as
d q dt = − A M s 2 T L I ∇ q G tot .
The corresponding free-energy derivative is
d G tot dt = − A M s 2 T ∇ q G tot T L I ∇ q G tot .
If LI is positive semidefinite,
d G tot dt ≤ 0 .
This is the multicomponent counterpart of Equation (97). It shows that the monotonic decrease of free energy does not depend on a single-solute assumption, although a fully quantitative prediction requires the thermodynamic curvature and the interfacial mobility matrix for the selected species.
Near equilibrium, define the vector of deviations
δq = q − qeq.
The chemical-potential difference is approximated by
Δ μ ≈ H δ q ,
where H is the constrained thermodynamic Hessian:
H = ∂ Δ μ ∂ q eq .
For a stable local equilibrium, the relevant constrained Hessian is positive definite. Substitution into Equation (104) gives
d δ q dt = − A M s T L I H δ q .
The solution is
δ q t = exp − A M s T L I H   t δ q 0 .
The eigenvalues of the product LIH determine the relaxation rates. For positive-definite H and symmetric positive-semidefinite LI, the nonzero eigenvalues are non-negative. The corresponding relaxation times can be expressed as
τ k = M s T A λ k ,
where λk denotes a positive eigenvalue of LIH, with the dimensional interpretation determined by the definitions of the species and transport coefficients.
This result implies that a multicomponent extraction system can possess several relaxation modes rather than a single universal extraction timescale. One mode may correspond predominantly to the relaxation of a rapidly transferred species, while another may be associated with a slower constituent or a strongly coupled combination of species. The modes are mathematical properties of the coupled linearized system; identifying them with particular coffee molecules requires experimental or computational parameterization.

3.6. Dimensionless Analysis and Dominant Transport Mechanisms

The governing equations contain transport coefficients with different physical dimensions. Dimensionless groups provide a systematic way to compare their relative importance without assigning unverified numerical values.
Let L be a characteristic length of the coffee bed, U a characteristic superficial liquid velocity, D i eff an effective liquid-phase diffusion coefficient, keff the effective thermal conductivity, and Ceff the effective volumetric heat capacity.
The mass-transfer Peclet number is defined as
P e i = UL D i eff .
A small Pei indicates that diffusion is important relative to advection over the selected length scale. A large Pei indicates that bulk liquid flow can transport the dissolved species more rapidly than molecular diffusion over the same scale. The interpretation depends on the chosen geometry and characteristic velocity; an immersion system with negligible mean flow may require a diffusion-based timescale instead.
The thermal Peclet number is
P e T = ρ l c p , l UL k eff .
This group compares advective thermal transport with conduction. A large value indicates that thermal transport by the flowing liquid can be important relative to conduction through the effective porous medium. A small value indicates that conduction is comparatively important.
The thermal diffusivity is
α eff = k eff C eff ,
and the corresponding thermal diffusion timescale is
τ T ∼ L 2 α eff .
The characteristic internal diffusion timescale of a solute in a particle of radius Rp is
τ D , i ∼ R p 2 D p , i eff ,
where D p , i eff is the effective intraparticle diffusion coefficient.
If the linearized interfacial law can be expressed in a conventional concentration-based form,
r i = k i , int c i * − c i ,
then ki,int has dimensions of velocity. The ratio between the particle diffusion resistance and interfacial resistance can be characterized by
B i m , i = k i , int R p D p , i eff .
This mass-transfer Biot-type number compares the external interfacial transfer rate with internal particle diffusion. A small value indicates that external transfer is comparatively restrictive, while a large value indicates that internal diffusion can become more important.
For a packed bed, an interfacial Damköhler-type group can be defined as
D a i , bed = a s k i , int L U .
This ratio compares the rate of transfer into or out of the liquid phase with advective transport through the bed. Its precise interpretation depends on the geometry and the effective mass-transfer coefficient.
The relative magnitude of the thermal and extraction timescales can be described by
Π T = τ T τ ext .
When ΠT≪1, the temperature field relaxes relatively rapidly compared with the extraction process. A quasi-steady thermal approximation may then be appropriate, subject to the external boundary conditions. When ΠT≫1, thermal evolution can remain important during extraction and should be retained explicitly in the coupled model.
A corresponding diffusive-to-extraction timescale ratio is
Π D = τ D τ ext .
These quantities provide a framework for classifying possible rate-limiting behavior. If internal diffusion is much slower than the interfacial process, the interior of the coffee particle can become the principal transport resistance. If the interface is slower, the bulk liquid may remain relatively undersaturated while extraction proceeds at a rate controlled by the solid–liquid boundary.
The dimensionless analysis also clarifies why one universal rate coefficient is unlikely to describe every brewing method. A stationary immersion system, a gravity-driven filter bed, and a pressure-driven espresso puck differ in their characteristic velocities, particle-bed geometry, liquid residence times, and thermal boundary conditions. Their dimensionless groups can therefore occupy different regimes even when the coffee material is identical.
The dimensionless groups are analytical constructs rather than experimentally evaluated values in the present study. Their numerical determination requires measurements or justified estimates of permeability, diffusion coefficients, heat capacity, thermal conductivity, interfacial mobility, and effective extraction length scales.

3.7. Boundary Conditions for Different Brewing Configurations

The governing equations are general at the local continuum level, but their boundary conditions depend on the brewing configuration. The distinction between immersion and percolation brewing is particularly important because it changes whether the liquid phase is treated as approximately stationary or as a continuously flowing medium.
For full immersion brewing, the macroscopic liquid velocity may be approximated as negligible after the initial mixing process. The liquid-phase mass balance then becomes
∂ ε c i ∂ t + ∇ ⋅ J i = a s r ¯ i .
At an impermeable boundary, the normal species flux satisfies
n ⋅ J i = 0 ,
unless there is a specified filtration, evaporation, or sampling operation. The thermal boundary may be approximated as adiabatic or represented by an external heat-transfer condition:
− n ⋅ k eff ∇ T = h T T − T ∞ .
For percolation brewing, water enters the bed through an inlet and the extracted liquid leaves through an outlet. A prescribed inlet concentration can be applied:
c i , in ≈ 0
for species not initially present in the incoming water. The inlet temperature is specified by the boundary condition:
T = T in
when the inlet temperature is controlled. The outflow boundary must permit the advective transport of dissolved compounds. Depending on the numerical formulation, an appropriate convective outflow condition can be adopted rather than imposing an arbitrary fixed concentration at the outlet.
For espresso, the flow field is driven by a substantial pressure difference across the coffee puck. Darcy’s law may provide a first approximation, but the validity of this approximation depends on the local flow regime and the mechanical state of the puck. If pressure-induced deformation, fines migration, or channel formation occurs, permeability should be allowed to evolve. The model must also recognize that the liquid velocity need not be spatially uniform. A nonuniform pressure field can generate different local residence times and extraction histories.
The boundary conditions for immersion, filter, and espresso brewing should therefore not be interpreted as identical. What remains common across the configurations is the local thermodynamic structure: species transfer responds to chemical-potential differences, heat transfer responds to temperature-related forces, and the entropy-production rate must remain non-negative.

3.8. Thermodynamic Interpretation of Extraction Yield and Beverage Composition

The proposed model distinguishes between the amount of a constituent extracted from the coffee solid, the concentration of that constituent in the beverage, and the chemical potential determining its transfer. This distinction is important because these quantities respond differently to changes in the water-to-coffee ratio, liquid retention, and boundary flow.
The extraction fraction Xi measures the relative depletion of species i from the solid matrix. The dissolved concentration ci measures its distribution in the liquid volume. The chemical-potential difference Δμi represents the thermodynamic driving force for net transfer between the phases. These quantities are connected through mass conservation and the thermodynamic equations, but no one quantity uniquely determines the others without additional information.
For example, increasing the liquid volume in a closed immersion system can reduce the concentration of a dissolved constituent through dilution while leaving the total amount extracted unchanged at a particular instant. The change in concentration can alter the liquid chemical potential and consequently influence the subsequent extraction rate. Therefore, dilution may influence the future driving force even though it is not itself a direct measure of solid depletion.
The equilibrium desorption model of Liang and colleagues provides a useful comparison for this distinction [5]. Their model establishes relationships among brew ratio, total dissolved solids, and equilibrium extraction yield under its effective-equilibrium assumptions. The present formulation considers the transient chemical-potential and transport processes that determine how the system evolves toward an equilibrium or operational endpoint. The two approaches can be related by using an equilibrium model to specify a limiting state, while the nonequilibrium model describes the dynamic approach to that state.
The present framework does not imply that a higher extraction yield necessarily produces a beverage with greater sensory quality. Extraction yield is a bulk mass metric, while the chemical composition of the liquid depends on the selective transfer of individual constituents. The relative proportions of acids, alkaloids, carbohydrates, phenolic compounds, and other molecules may differ between brews with similar aggregate dissolved-solids values. Sensory perception introduces additional nonlinear relationships that are outside the present thermodynamic formulation.

3.9. Theoretical Model Implications

Several implications follow directly from the equations developed above.
First, the temperature field should not automatically be treated as a fixed input. In a transient brewing system, temperature affects the chemical potential, effective diffusion coefficients, viscosity, interfacial mobility, and thermal boundary flux. Unless the thermal timescale is demonstrably shorter than the extraction timescale, an isothermal approximation may conceal an important part of the process.
Second, the chemical potential offers a more general description of extraction driving force than concentration difference alone. In an ideal dilute solution, the chemical-potential relation reduces to a concentration-based form near equilibrium. Outside this limiting regime, activity coefficients and cross-compositional effects may influence the transfer tendency.
Third, the porous particle and the macroscopic bed should be treated as distinct transport domains. The intraparticle diffusion length and the bed-scale advective length are different physical quantities. A model that combines them into one apparent diffusion coefficient may be useful for fitting data but can obscure the mechanisms responsible for the observed behavior.
Fourth, the pressure field contributes to the extraction process through liquid transport and hydraulic dissipation. In filter and espresso brewing, the flow field affects the residence time and the local concentration history. In a rigid-bed Darcy approximation, pressure drop and permeability determine the superficial velocity. If the bed deforms, the hydraulic relation must be extended.
Fifth, an equilibrium endpoint and the path toward that endpoint represent different levels of description. The closed immersion derivation demonstrates how a chemical-potential driving force can lead to free-energy dissipation and relaxation toward equilibrium. However, practical brewing often terminates before true equilibrium is reached, and some systems remain open to continuous liquid flow.
Finally, the model provides a mathematical basis for identifying several possible kinetic limitations. A process may be dominated by internal particle diffusion, dissolution at the solid–pore-liquid boundary, external particle-to-liquid transfer, or advection through the bed. The dimensionless groups introduced above can organize these mechanisms into a common regime description. Determining the dominant mechanism for a specific brewing configuration requires quantitative parameter estimates and, ultimately, comparison with experimental measurements.

4. Discussion

The central objective of this study was to connect the physical description of coffee extraction with the thermodynamic principles governing irreversible transport. The resulting formulation does not depend on a single empirical extraction curve. Instead, it is built from chemical-potential differences, conservation laws, transport coefficients, boundary conditions, and entropy production.
A comparison with earlier coffee extraction models is necessary to clarify the scope of the contribution. Moroney and colleagues established a double-porosity model in which rapid extraction from readily accessible regions and slower transfer from the internal particle structure contribute to the macroscopic extraction profile [2,3,4]. The present formulation retains the physical relevance of these mechanisms. Its difference is that the transfers are represented through chemical-potential differences and linked explicitly to entropy production. The existing double-porosity structure can therefore be viewed as a possible transport architecture into which the thermodynamic formulation may be incorporated.
The particle-scale equations introduced here are not intended to imply that every microscopic pore or cellular structure has been resolved. The variables qp,i, cp,i, and ϕp represent effective fields at a particle-scale continuum level. The exact internal geometry of roasted coffee can be substantially more complex, and the validity of a homogeneous particle-scale description depends on the selected representative volume and the degree of structural averaging. The strength of the approach is that it allows the relevant transfer stages to be distinguished without requiring an atomistic description of the entire coffee particle.
The equilibrium desorption model of Liang and colleagues provides an important comparison because it describes the effective endpoint of full-immersion extraction [5]. Their model uses a species-averaged equilibrium parameter and derives relationships between brew ratio, dissolved-solids concentration, and extraction yield. The present framework instead focuses on the transient thermodynamic structure. It defines the chemical-potential difference that drives transfer and provides an analytical description of the local approach to equilibrium. In this respect, the equilibrium and nonequilibrium formulations address different questions. The former provides a reduced description of an effective final state, while the latter describes the thermodynamic forces and dissipation associated with the extraction path.
The experimental findings of Schmieder and colleagues also provide a useful context [6]. Their study measured the extraction kinetics of selected espresso constituents under variations in flow rate, particle size, and temperature. The results demonstrate the compound-specific character of extraction and the interaction between brewing conditions. The theoretical model developed here explains why a single global extraction coefficient is not expected to describe all components under all operating conditions. Different species can have different activity coefficients, internal diffusion mobilities, interfacial transfer coefficients, and thermal responses.
The influence of dissolved mineral ions on coffee extraction, examined by Hendon and colleagues, is another reason for adopting a chemical-potential framework [7]. In a simplified model, the liquid phase may be treated as a dilute aqueous solution with a fixed activity coefficient. However, the actual chemical environment of dissolved coffee constituents can depend on the ionic composition of the brewing water. An extension of the present model could include mineral ions as additional independent species, with activity coefficients and complexation reactions. The resulting chemical-potential gradients would then depend on both coffee-derived solutes and water-derived components.
The entropy-production formulation is a particularly important element of the proposed theory. A mass-balance model may conserve species while still employing constitutive coefficients that are thermodynamically inconsistent. By writing the local entropy-production rate in terms of conjugate fluxes and forces, the model places restrictions on the permissible transport coefficients. The requirement that the symmetric part of the phenomenological matrix be positive semidefinite is not merely a mathematical convenience. It is the condition that prevents the linear model from predicting negative entropy production for admissible thermodynamic forces.
The pressure-driven contribution is similarly important. In a porous coffee bed, hydraulic transport is not independent of extraction. The flow field carries dissolved constituents from the particle surfaces and influences the concentration gradients that develop throughout the bed. Darcy’s law provides a useful low-Reynolds-number approximation, and the corresponding pressure dissipation contributes non-negatively to entropy production. However, the rigid-bed assumption may be inadequate in some espresso configurations, where compaction and local variations in permeability can occur. In such cases, the pressure and mechanical fields should be coupled.
An additional consideration is the validity of linear irreversible thermodynamics. The Onsager formulation is based on near-equilibrium transport and the existence of well-defined local thermodynamic states [8,9]. Practical brewing may involve large initial temperature differences between hot water and room-temperature coffee, large local concentration differences, and changing pore-scale conditions. The linear constitutive relations should therefore be understood as local approximations, not as universal laws for the complete brewing process.
The nonlinear extension introduced through a state-dependent mobility coefficient is one possible response to this limitation. Requiring the mobility to remain non-negative preserves the sign of the interfacial entropy-production term. However, entropy admissibility alone is insufficient to establish quantitative accuracy. A nonlinear law must still be justified by physical arguments or by comparison with suitable data. More complex kinetic laws may be needed if dissolution, adsorption, molecular association, or changes in pore accessibility become dominant.
The same caution applies to the cross-coupling terms in the transport matrix. The existence of an Onsager coefficient does not demonstrate that a particular cross-effect is significant in coffee brewing. A temperature gradient may alter extraction through changes in viscosity and diffusivity without generating a measurable heat-of-transport effect. The phenomenological matrix is therefore a framework for organizing possible couplings, not evidence that all off-diagonal coefficients are large or nonzero.
The closed-batch analytical result provides a more specific contribution. Under constant temperature and pressure, the proposed interfacial law makes the total Gibbs free energy a monotonically decreasing function of time. The result is derived directly from mass conservation and the relationship between transfer rate and chemical-potential difference. The linearized model then yields a relaxation spectrum for a multicomponent system. This structure is useful because it separates the thermodynamic driving force from the mobility controlling the rate of approach to equilibrium. Two systems can have similar equilibrium states but different relaxation times if their interfacial mobilities or transport resistances differ.
The analytical relaxation result should not be confused with the complete extraction kinetics of a real brew. The derivation assumes a closed, well-mixed liquid volume, fixed temperature, fixed interfacial area, constant interfacial mobility near equilibrium, and a single stable equilibrium state. Real coffee brewing may violate several of these assumptions. The initial release from damaged particle surfaces can be rapid; slower material may become available through internal diffusion; the temperature may change; and liquid may be continuously added or removed. The exponential solution is therefore a local result for the approach to equilibrium, not a global fitting equation for every brewing method.
A further distinction concerns the Gibbs free energy used in the closed-system derivation. At constant temperature and pressure, Gibbs free energy is the appropriate thermodynamic potential for the corresponding isothermal–isobaric description. In a non-isothermal system, the total entropy and energy balances must be considered together. The monotonic decrease of a single global Gibbs free energy cannot be assumed without accounting for external heat exchange, pressure work, and the boundary conditions. For an open brewing system, the relevant balance must include the chemical and energy fluxes through the inlet and outlet.
A quantitative implementation of the model would require several categories of property data. These include particle-size distributions, internal porosity, bed porosity, permeability, effective thermal conductivity, heat capacity, liquid viscosity, species-specific diffusion coefficients, interfacial area, activity coefficients, and transfer enthalpies. For a multicomponent system, thermodynamic data would also be required to determine the relationship between species concentrations and chemical potentials. Such data are not available at a sufficiently complete level for every relevant compound and brewing condition. The theoretical framework therefore identifies the parameters that need to be estimated rather than assigning them unsupported numerical values.
Experimental validation would require time-resolved measurement of temperature and concentration during brewing. A suitable study could compare selected constituents in sequential fractions from a packed bed, as done in espresso extraction kinetics research [6], while also measuring the temperature profile and pressure drop. For immersion brewing, a controlled experimental arrangement could record temperature and bulk composition as functions of time. In either case, the thermodynamic model would need to be parameterized using independently estimated transport properties and then compared with measurements not used in parameter fitting.
The present work does not perform such an experiment. Its results are analytical consequences of the proposed governing equations. This distinction is essential because a mathematically consistent theoretical model is not automatically a validated predictive model. Validation would require systematic comparison between model predictions and independently obtained data, including uncertainty analysis and an assessment of parameter identifiability.
The sensory implications of the framework should also be treated cautiously. The chemical composition of brewed coffee is an important determinant of beverage characteristics, but the model does not calculate sensory response. Differences in extraction histories can produce different compound distributions, even when bulk total dissolved solids are similar. Nevertheless, predicting flavor intensity, preference, bitterness, acidity, or perceived balance would require a separate physicochemical and sensory model.
Several extensions follow naturally from the present formulation. The first is the inclusion of a detailed double-porosity particle structure, in which accessible surface material and slowly accessible internal material are represented through separate state variables. The second is an evolving-permeability model for pressure-driven brewing systems. The third is the inclusion of water-derived ions and complexation reactions through a species-resolved thermodynamic model. The fourth is a two-temperature formulation when solid–liquid thermal equilibration is not sufficiently rapid. Finally, the framework could be coupled to chemical-reaction terms for constituents that undergo transformation during brewing.
These extensions should be developed only when the associated physical properties can be defined and the corresponding equations can be closed. Adding more variables does not automatically improve a model. A useful extension must introduce a physical mechanism that is relevant to the selected brewing configuration and must provide a way to estimate or identify the additional parameters.
The theoretical framework presented here is therefore best understood as a structured starting point for a class of nonequilibrium coffee-brewing models. Its strength lies in the compatibility of the conservation laws, the chemical-potential driving forces, and the entropy-production constraints. The equations provide a common language for comparing immersion and percolation systems while preserving the distinction between their different flow conditions and boundary requirements.

5. Conclusions

Coffee brewing has been formulated as an open, heterogeneous, and nonequilibrium thermodynamic process in which heat transfer, intraparticle diffusion, interfacial extraction, liquid-phase transport, and pressure-driven flow may occur simultaneously. The proposed framework combines species-specific conservation equations with chemical-potential-based transfer relations and an entropy balance that imposes thermodynamic restrictions on the constitutive coefficients.
The main result is a unified description of the thermodynamic driving forces associated with extraction and transport. The chemical-potential difference between the solid-associated and liquid states determines the direction of interfacial transfer. In an ideal dilute limit, this relation reduces to a concentration-based driving force, showing how conventional mass-transfer expressions can emerge as approximations to a more general thermodynamic formulation.
The multiscale description distinguishes processes inside individual coffee particles from the transport of dissolved compounds through the macroscopic bed. The energy equation connects the evolving thermal field with heat conduction, advection, diffusion-associated enthalpy transport, interfacial transfer, and hydraulic dissipation. Darcy-scale flow provides a first approximation for pressure-driven brewing, subject to the limitations of rigid-bed permeability and the assumed flow regime.
The entropy-production analysis establishes the conditions under which heat and mass transfer, interfacial extraction, and pressure-driven flow are thermodynamically admissible. In particular, the mobility and transport matrices must satisfy non-negativity conditions, and the symmetric part of the coupled phenomenological matrix must be positive semidefinite.
For a closed, isothermal immersion model, the total Gibbs free energy is shown analytically to decrease during spontaneous extraction. Linearization near a stable equilibrium produces exponential relaxation for one constituent and a spectrum of relaxation modes for a multicomponent system. These results identify the relationship between thermodynamic curvature, interfacial mobility, and extraction timescales without requiring experimentally fitted numerical values.
The dimensionless formulation provides a basis for distinguishing thermal, intraparticle diffusion, interfacial, and advective regimes. It also clarifies why the same extraction coefficient or operating condition cannot be expected to describe all brewing methods and all coffee constituents.
The proposed model is theoretical and has not been experimentally validated in the present study. Its quantitative use requires thermodynamic property data, transport coefficients, appropriate boundary conditions, and comparison with time-resolved measurements. Future work should focus on parameter estimation, numerical solution of the coupled equations, and controlled validation under immersion and percolation conditions. Within these limitations, the framework provides a consistent thermodynamic basis for studying coffee brewing beyond purely empirical extraction descriptions.

Supplementary Materials

No supplementary experimental datasets, laboratory measurements, or computational simulations were generated for this theoretical study. The mathematical formulation and analytical derivations necessary to understand the proposed framework are presented in the manuscript.

Author Contributions

Conceptualization, S.G.; methodology, S.G.; theoretical investigation, S.G.; formal analysis, S.G.; mathematical formulation, S.G.; writing—original draft preparation, S.G.; writing—review and editing, S.G. The author has read and agreed to the published version of the manuscript.

Funding

This research received no external funding.

Institutional Review Board Statement

Not applicable. This study did not involve human participants, animals, or identifiable personal data.

Data Availability Statement

No new experimental datasets were generated or analyzed during this theoretical study. All mathematical formulations and analytical derivations supporting the conclusions are included in the manuscript. No external experimental dataset was used.

Acknowledgments

The author acknowledges the contributions of earlier research in coffee extraction modeling, porous-medium transport, chemical thermodynamics, and irreversible processes, which provide the scientific foundations for the theoretical development presented in this work.
Editorial transparency note: Any use of generative artificial intelligence in the preparation, editing, mathematical development, or language revision of the manuscript should be disclosed in accordance with the applicable Preprints.org and journal policies. The author is responsible for independently verifying the mathematical formulation, scientific interpretations, and bibliographic information and for approving the final submitted manuscript.

Conflicts of Interest

The author declares no conflict of interest.

Abbreviations

as
External particle interfacial area per unit bed volume
ai
Thermodynamic activity of species i
A
Total solid–liquid interfacial area in the reduced immersion model
Bim,i
Mass-transfer Biot-type number
ci
Bulk liquid concentration of species i
cp,i
Intraparticle pore-liquid concentration of species i
c i *
Effective equilibrium concentration of species i
Ceff
Effective volumetric heat capacity
D i eff
Effective liquid-phase diffusion coefficient
D p , i eff
Effective intraparticle diffusion coefficient
Dai,bed
Bed-scale interfacial Damköhler number
E
Extraction yield
FoT
Thermal Fourier number
Gtot
Total Gibbs free energy
H
Constrained thermodynamic curvature near equilibrium
Jq'
Reduced heat flux
Ji
Diffusive molar flux of species i
K
Hydraulic permeability
keff
Effective thermal conductivity
ki,int
Effective interfacial mass-transfer coefficient
L
Characteristic bed length
L i I
Interfacial phenomenological mobility
LI
Interfacial mobility matrix
Ms
Dry coffee mass
Mi
Molar mass of species i
p
Liquid pressure
Pei
Mass-transfer Peclet number
PeT
Thermal Peclet number
qi
Solid-phase molar loading
qp,i
Intraparticle solid-associated inventory
R
Universal gas constant
Rd,i
Intraparticle dissolution rate
ri
Interfacial molar transfer rate
s
Entropy density
σ
Local entropy-production rate
T
Absolute temperature
τD
Characteristic diffusion timescale
τext
Characteristic extraction relaxation time
τT
Characteristic thermal timescale
u
Superficial liquid velocity
Vl
Liquid volume
Xi
Extraction fraction of species i
X
Vector of thermodynamic forces
αeff
Effective thermal diffusivity
Δ G i ∘
Standard Gibbs energy of transfer
Δ H i tr
Enthalpy change of transfer
Δμi
Solid-to-liquid chemical-potential difference
ε
Bed-scale liquid porosity
η
Dynamic viscosity
ϕp
Intraparticle porosity
ρb
Dry coffee mass per unit bed volume
ρl
Liquid density
ρs
Effective solid density
ΠD
Diffusive-to-extraction timescale ratio
ΠT
Thermal-to-extraction timescale ratio
ΦD
Hydraulic dissipation rate per unit volume

References

  1. Cordoba Castro, N.M.; Ruiz Pardo, R.Y.; Moreno Moreno, F.L.; Fernandez Alduenda, M.R. Coffee extraction: A review of parameters and their influence on the physicochemical characteristics and flavour of coffee brews. Trends Food Sci. Technol. 2020, 96, 45–60. [Google Scholar] [CrossRef]
  2. Moroney, K.M.; Lee, W.T.; O’Brien, S.B.G.; Suijver, F.; Marra, J. Modelling of coffee extraction during brewing using multiscale methods: An experimentally validated model. Chem. Eng. Sci. 2015, 137, 216–234. [Google Scholar] [CrossRef]
  3. Moroney, K.M.; Lee, W.T.; O’Brien, S.B.G.; Suijver, F.; Marra, J. Coffee extraction kinetics in a well mixed system. J. Math. Ind. 2016, 7, 3. [Google Scholar] [CrossRef]
  4. Moroney, K.M.; Lee, W.T.; O’Brien, S.B.G.; Suijver, F.; Marra, J. Asymptotic Analysis of the Dominant Mechanisms in the Coffee Extraction Process. SIAM J. Appl. Math. 2016, 76, 2196–2217. [Google Scholar] [CrossRef]
  5. Liang, J.; Chan, K.C.; Ristenpart, W.D. An equilibrium desorption model for the strength and extraction yield of full immersion brewed coffee. Sci. Rep. 2021, 11, 6904. [Google Scholar] [CrossRef]
  6. Schmieder, B.K.L.; Pannusch, V.B.; Vannieuwenhuyse, L.; Briesen, H.; Minceva, M. Influence of Flow Rate, Particle Size, and Temperature on Espresso Extraction Kinetics. Foods 2023, 12, 2871. [Google Scholar] [CrossRef]
  7. Hendon, C.H.; Colonna-Dashwood, L.; Colonna-Dashwood, M. The Role of Dissolved Cations in Coffee Extraction. J. Agric. Food Chem. 2014, 62, 4947–4950. [Google Scholar] [CrossRef]
  8. Onsager, L. Reciprocal Relations in Irreversible Processes. I. Phys. Rev. 1931, 37, 405–426. [Google Scholar] [CrossRef]
  9. de Groot, S.R.; Mazur, P. Non-Equilibrium Thermodynamics; Originally published by; Dover Publications: New York, NY, USA; North-Holland: Amsterdam, 1984. [Google Scholar]
  10. Kondepudi, D.; Prigogine, I. Modern Thermodynamics: From Heat Engines to Dissipative Structures, 2nd ed.; John Wiley & Sons: Hoboken, NJ, USA, 2014. [Google Scholar] [CrossRef]
  11. Whitaker, S. The Method of Volume Averaging; Kluwer Academic Publishers: Dordrecht, The Netherlands, 1999. [Google Scholar] [CrossRef]
Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content.
Copyright: This open access article is published under a Creative Commons CC BY 4.0 license, which permit the free download, distribution, and reuse, provided that the author and preprint are cited in any reuse.