WorksPhysics and Astronomy
Thermodynamics of Non-equilibrium Diffuse-Interfaces in Mesoscale Phase Transformations
Abstract
We present a new phase-field formulation for the non-equilibrium interface kinetics. The diffuse interface is considered an integral of numerous representative volume elements (RVEs), in which there is a two-phase mixture with two conserved and two non-conserved concentration fields. This particular way of separating concentration fields leads to two distinct dissipative processes besides the phase boundary migration, i.e., trans-sharp-interface diffusion between different phases within a single RVE and trans-diffuse-interface diffusion between different RVEs within a single phase. A two-part mechanism is proposed to balance the driving forces and energy dissipation for diffusionless and diffusional processes in RVEs. Moreover, the classical interface friction and solute drag can be recovered by integrating the diffuse interface. Compared with the sharp interface theory, the present diffuse interface can reproduce the atomic simulated partial-drag self-consistently in thermodynamics.
Motivation.—Many inhomogeneous systems at the mesoscale consist of domains of well-defined phases separated by distinct interfaces [1]. Due to the coexistence of individual and cooperative atomic motions [2, 3, 4, 5, 6], the dissipative processes within the interfaces are usually more complex than those in the bulks. Therefore, to quantitatively predict phase transformations, it is vital to uncover how various dissipative processes partition the thermodynamic forces [6] and develop suitable models for describing interface kinetics from near to far from equilibrium.
As the most fundamental approach, classical sharp interface models can be directly applied to transformation kinetics with simple morphology [7] and serve as the standard or rule for other simulation techniques, e.g., phase-field modeling, cellular automation, and level-set methods [8]. However, sharp interface models fail in the kinetics of interfaces far from equilibrium. A simple but illustrative example is the unusual "partial drag" [9] in the rapid solidification of highly undercooled melts. Specifically, although sharp interface models have been extended to non-equilibrium cases [10, 11, 12, 13, 14] based on the classical irreversible thermodynamics [15, 16, 17], the predicted driving force and energy dissipation significantly deviate from those of experiments [18] and atomic simulations [19, 20, 21, 22]. Therefore, a thermodynamically consistent description of far-from-equilibrium interfaces is still lacking.
The failure of sharp interface models is likely attributed to the diffuse nature of many interfaces. The phase-field (PF) theory has become a mainstream means of modeling diffuse interfaces by introducing a series of fields representing the material's properties with the interface [23, 24]. As the continuum limit of microscopic models [24], the classical mesoscale PF models should inherently reflect the averaged properties of diffuse interfaces. However, mainly focusing on the sharp interface limit [25, 26, 27], they are more like an effective sharp interface model for capturing complex pattern formations. By contrast, there is no clear understanding of the dissipative processes and their thermodynamics inside the diffuse interface. Therefore, this letter aims to develop a PF model self-consistently in thermodynamics, with transparent dissipative processes inside the diffuse interface to represent the far-from-equilibrium kinetic.
Model.— This model starts from a classical free energy formulation of a two-phase binary system, i.e., , in which is used to interpolate the free energy density , and are coefficients for the double-well potential and gradient energy, respectively. The overall concentration is .
As shown in Figure. Figure 1(a), the diffuse interface is considered an integral of numerous representative volume elements (RVEs), in which "RVE1" and "RVE2" are adjacent. The dashed curve denotes the average phase fractions rather than orderly arrangement of two phases. For a single RVE in Figure. Figure 1(b), the domains of and are separated by a distinct phase boundary as a "sharp interface", whose kinetics is controlled by phase boundary migration (red arrows) and short-range atomic exchange (green arrows), referred to as "trans-sharp-interface diffusion" in this work. Furthermore, between adjacent RVEs, there is a long-range diffusion in the same phase (blue arrows), which is defined as "trans-diffuse-interface diffusion" currently.
Figure 1. (a) Diffuse interface and the internal representative volume elements (RVEs); (b) Three dissipative processes between and within RVEs, where the blue arrows means the long-range diffusion in the same phase while the green arrows are the short-range exchange between different phases.
Due to the atomic exchange, it is natural to see that the separated concentration fields should be non-conserved. Thus, we here divide the instantaneous into two fields: a conserved and a non-conserved , leading to a diffusion equation with additional sources or sinks, i.e.,
where and () are the fluxes of trans-diffuse-interface diffusion and trans-sharp-interface diffusion in Figure 1(b), respectively. Note that it is the trans-sharp-interface diffusion, together with phase boundary migration, that determine the "sharp-interface" phase transformation. Therefore, there should be (or ), making the mass conservation yield
which is also similar to the classical sharp interface models [9, 10, 11, 12, 13, 14]. Addition, it should be note that a similar understanding of internal dissipative processes has been previously used in phase-field models [28, 29], but the present Eqs. (1) and (2) are treating their kinetic coupling in an exactly different way.
Now, let us consider a phase transformation in a closed system, the change rate of total free energy is given by [30]
where is the Lagrange multiplier for the constraint of Eq. (2). Based on the extremum principle of linear thermodynamics [17], the total free energy dissipation is then expressed as a quadratic form of four fluxes in Eq. (3)
where is the phase-field mobilities, and are the atomic mobilities for and , respectively. Combining Eq. (2) with for four fluxes [17] can eliminate the and solve the evolving equations as [30]
in which are the diffusion potentials, is an effective phase-field mobility, and are the atomic mobilities for and , respectively [30], where is the lattice distance and is the interface width.
Additionally, one can see that Eqs. (5)–(7) follow the Onsager’s reciprocal relationship [15, 30].
Different from existing models [28, 29], the present Eq. (7) indicates that the trans-diffuse-interface diffusion of is also driven by the chemical gradient of another because is being changed by the trans-sharp-interface diffusion (vice versa). Thus, and are dependent in Figure 1. Substituting Eqs. (6) and (7) into Eq. (1) will offer the temporal evolutions of , , and [30].
Thermodynamics of sharp-interface.—In a single RVE, there is a "sharp-interface" type phase transformation governed by the response functions of interface velocity [Eq. (5)] and concentrations [Eq. (6)]. According to Eq. (2), the fluxes of trans-sharp-interface diffusion are given by
where is the material composition crossing the sharp interface, instead of the real solid composition . Substituting Eq. (8) into the non-conserved terms of Eq. (3), the net driving force for a RVE is
which is the same as the sharp-interface models [11, 12, 13, 14, 15]. It means that is consumed to drive a diffusional transformation from to and two fluxes being transferred, for which " ", " ", and " " are three conjugated forces cross-coupled with three fluxes [15]. However, in order to distinctly understand different dissipative processes and their thermodynamics, it is necessary to find independent fluxes and forces.
Considering a steady state in RVEs, the independent driving force for phase boundary migration according to Eq. (5) yields as
It suggests that there are two types of diffusionless transformation for phase boundary migration, and only the part migrates as Hillert et al.’s viewpoint [6]. Thus, a "two-part" concept will be used in latter discussions, i.e., Part I () and Part II (). Substituting of Eq. (8) into Eq. (6), there will be two independent trans-sharp-interface fluxes [30], whose changes in free energy are
and is the same as Eq. (9) from three conjugated forces and fluxes. Additionally, Eqs. (11) and (12) can be further extended as Part I: , , ; and Part II: , , and , whose physical meaning will be illustrated later.
Now, we interpret the Eqs. (10)–(12) with the following "two-part" phase transformation mechanisms, whose schematic depiction is shown in Figure. Figure 2(a), and Figure. Figure 2(b, Figure 2c) exhibit the free energy changes of different dissipative processes.
Figure 2. (a): The "two-part" schematics for “sharp-interface” type phase transformation in RVEs; (b) and (c): Graphical constructions of molar free energy changes for Parts I and II, respectively. Parts I and II are divided by the dashed lines in (a). In each part, the phase boundary is represented by the black solid line, while the background colors represent the lattice concentrations of and . The (1.1-1.4) and (2.1-2.4) denote the kinetic processes of Part I and II, respectively.
For Part I (), the phase boundary first diffusionless transform from to [(1.1) in Figure. Figure 2], during which two trans-sharp-interface fluxes and their corresponding free energy changes are compensated, according to Eq. (6). Then, the metastable starts desorbing atoms to approach [(1.2)], which also produces a moving tendency of the phase boundary. Concurrently, the small changes in concentration and phase boundary induce two non-compensated trans-sharp-interface fluxes [(1.3) and (1.4)], providing an opposite moving tendency of the phase boundary. As a result of (1.2)-(1.4), the phase boundary is not moved. Alternatively, the net material transferred from product to the sharp interface equals that transferred from the interface to parent , i.e., , so there are no retained atoms moving the phase boundary. The net free energy change, Eq. (11), is actually the same as Hillert et al.’s work [12].
For Part II (), similarly, (2.1)-(2.3) first make desorb atoms into , while the phase boundary is static due to the offset by . The net change in free energy is given by Eq. (12). Then, the phase boundary migration makes the metastable phase transform into in a diffusionless way [(2.4) in Figure 2]. Note that previous sharp-interface models [11, 12, 13, 14, 15] can also be reinterpreted by the present "two-part" mechanism, whose proportion is determined by atomic mobility for trans-sharp-interface diffusion and long-range diffusion fluxes at the phase boundary [30]. Additionally, a recent experiment by neutron total scattering [5] should be mentioned. It is found that there exist "atomic groups or cells" in the interface, analogous to the present RVE concept, and the phase transformation shows a diffusion-assisted displacive manner as the Part II.
Thermodynamics of diffuse-interface.—Now, it is necessary to explore the relationship between the present PF model with the classical diffuse interface theories by Cahn [31] and Hillert and Sundman [32]. Using a standard coordinate transformation for the steady state, Eq. (1) is transformed into . is the conserved concentration at a reference point, at which . Then, apply to Eqs. (5)-(7) and multiply by , , and on both sides. After integrating the whole diffuse interface, the balance of the (absolute) total driving force and energy dissipation is given by
The superscripts “” and “” denote the and boundaries of the diffuse interface, respectively.
It is shown that is analogous to that of sharp-interface , and can be similarly called as the material composition crossing the diffuse interface. If assuming the effective phase-field mobility in Eq. (5) as a constant, Eq. (13) can be further simplified by combining Eq. (2), yielding as
where is the interface energy. One can see that the first and second terms on the right-hand side respectively correspond to the interface friction/migration () and solute drag () of the classical diffuse interface theories [31, 32, 33]. Moreover, is related to the intrinsic interface mobility by
.
Partial-drag in rapid solidification. To validate the present PF model in representing non-equilibrium interface kinetics, we apply it to the 1-D rapidly solidified Al-Cu alloys. Almost all model parameters [30] are the same as Haapalehto et al.’s molecular-dynamic (MD) simulations [21], e.g., phase diagram, diffusion coefficients, interface thickness, and partition coefficient . Figure 3(a) and (Figure 3b) exhibit the concentration profiles and their corresponding non-equilibrium , where there is an increasing solute trapping with interface velocity . The complete solute trapping at a finite velocity is reproduced by modifying the of Eq. (7) with an effective coefficient [30], which accounts for the far-from-equilibrium diffusion in the very high velocity regime [34]. After reaching , the diffusionless solidification occurs, where the solute drag disappears in Figure 3(c), and thus is linearly proportionate to the undercooling in Figure 3(d).

Figure 3. (a) and (b): Present composition profiles and corresponding partition coefficient; (c) and (d) Energy dissipation and interface temperature. Haapalehto et al.’s results [21] of MD simulations and sharp-interface model (CGM) are listed for comparison.
Finally, we can revisit the atomic simulated "partial drag" phenomenon [19, 20, 21, 22] with the present diffuse interface model. We starts from a “sharp-interface” approximation with and in Figure 4(a). Because of the negligible diffusion in bulk solids, in Eq. (13).
Figure 4. (a) Schematic depiction of an approximated sharp interface for present PF model; (b) Comparison of simulated with "full drag" and "partial drag" models [19].
The corresponding is thus given by the black arrow in Figure 4(b) as in the previous full-drag sharp-interface models [9, 10, 11, 12, 13, 14]. However, as shown in Figure 3(c) and Figure 4(b), the simulated is very close to "partial-drag" sharp interface model [9], i.e., . The coefficient "0.5" surprisingly coincides with the value recently summarized from almost all existing atomic simulations [22]. It indicates that the present work can reproduce the non-equilibrium interface conditions self-consistently in thermodynamics, while the recent "partial drag" sharp interface model [11] is a convenient method for approximation. In the future, it is important to explore a strategy of quantitative extending the present PF model to a size scale available for mesoscale simulations.
The work was supported by the Research Fund of the State Key Laboratory of Solidification Processing (NPU), China (Grant No. 2020-TS-06, 2021-TS-02) and Natural Science Basic Research of Shannxi (Program No. 2022JC-28).
References
- [1]H. Emmerich, Adv Phys 57 1 (2008).
- [2]K.L. Merkle, L.J. Thompson, Phys. Rev. Lett. 88 3 (2002).
- [3]L. Bourgeois, N.V. Medhekar, A.E. Smith, M. Weyland, J.F. Nie, C. Dwyer, Phys. Rev. Lett. 111 046102 (2013).
- [4]J. Wei, B. Feng, R. Ishikawa, T. Yokoi, K. Matsunaga, N. Shibata, Y. Ikuhara, Nat. Mater. 20 951 (2021).
- [5]Y. Cui, Y. Zhang, L. Sun, M. Feygenson, M. Fan, X.L. Wang, P.K. Liaw, I. Baker, Z. Zhang, Mater. Today. Phys. 24 100668 (2022).
- [6]M. Hillert, Acta Mater. 47 4481 (1999).
- [7]M. Gouné, F. Danoix, J. Ågren, Y. Bréchet, C.R. Hutchinson, M. Militzer, S. van der Zwaag, H. Zurob, Mat. Sci. Eng. R. 92 1 (2015).
- [8]Du, Y., Schmid-Fetzer, R., Wang, J., Liu, S., Wang, J., & Jin, Z. (2023). Computational Design of Engineering Materials: Fundamentals and Case Studies. In Cambridge University Press.
- [9]C.A Hareland, G. Guillemot, C.A. Gandin, P.W. Voorhees, Acta Mater. 241 118407 (2022).
- [10]M. Buchmann, M. Rettenmayr, Scripta Mater. 50 507 (2004).
- [11]M. Hillert, M. Rettenmayr, Acta Mater. 51 2803 (2003).
- [12]M. Hillert, J. Odqvist, J. Ågren. Scripta Mater. 50 547 2004.
- [13]H. Wang, F. Liu, H. Zhai, K. Wang, Acta Mater. 60 1444 (2012).
- [14]W. Kuang, H. Wang, J. Zhang, F. Liu, Acta Mater. 120 415 (2016).
- [15]L. Onsager, Phys Rev 37 405 (1931).
- [16]Ziegler, H. (1983). An introduction to thermodynamics.
- [17]F.D. Fischer, J. Svoboda, H. Petryk, Acta Mater. 67 1 (2014).
- [18]J.A. Kittl, P.G. Sanders, M.J. Aziz, D.P. Brunco, M.O. Thompson MO. Acta Mater. 48 4797 (2000).
- [19]Y. Yang, H. Humadi, D. Buta, B.B. Laird, D. Sun, J.J. Hoyt, M. Asta, Phys. Rev. Lett. 107 025505 (2011).
- [20]H. Humadi, J.J. Hoyt, N. Provatas, Phys. Rev. E. 93 010801(R) (2016).
- [21]M. Haapalehto, T. Pinomaa, L. Wang, A. Laukkanen, Comp. Mater. Sci. 209 111356 (2022).
- [22]E.A. Antillon, C.A. Hareland, P.W. Voorhees, Acta Mater. 248 118769 (2023).
- [23]I. Steinbach, Annu. Rev. Mater. Res. 43 89 (2013).
- [24]L.Q. Chen, Y. Zhao, Prog. Mater. Sci. 124 100868 (2022).
- [25]A. Karma, Phys. Rev. Lett. 87 115701 (2001).
- [26]S.G. Kim, Acta Mater. 55 4391 (2007).
- [27]K. Ji, E. Dorari, A.J. Clarke, A. Karma, Phys. Rev. Lett. 130 026203 (2023).
- [28]I. Steinbach, L. Zhang, M. Plapp, Acta Mater. 60 2689 (2012).
- [29]H. Wang, F. Liu, G.J. Ehlen, D.M. Herlach, Acta Mater. 61 2617 (2013).
- [30]Supplementary Materials, in preparation.
- [31]J.W. Cahn, Acta Metall. 10 789 (1962).
- [32]M. Hillert, B. Sundman, Acta Metall. 24 731 (1976).
- [33]M. Alkayyali, F. Abdeljawad, Phys. Rev. Lett. 127 175503 (2021).
- [34]H. Wang, P.K. Galenko, X. Zhang, W. Kuang, F. Liu, D.M. Herlach, Acta Mater. 90 282 (2015).


