Phase-field modeling of diffusion controlled phase ... - KTH Mechanics
Phase-field modeling of diffusion controlled phase ... - KTH Mechanics
Phase-field modeling of diffusion controlled phase ... - KTH Mechanics
You also want an ePaper? Increase the reach of your titles
YUMPU automatically turns print PDFs into web optimized ePapers that Google loves.
1330 I. Loginova et al. / Acta Materialia 51 (2003) 1327–13393. Solute drag <strong>modeling</strong> <strong>of</strong> the g→atransformation in Fe–CIn the solute drag <strong>modeling</strong> the thickness (δ) <strong>of</strong>the interface is considered small enough comparedto the curvature <strong>of</strong> the interface to approximate the<strong>diffusion</strong> <strong>field</strong> inside the interface as planar, seee.g. in ref. 2. The steady-state solution <strong>of</strong> Eq. (7)will be a good approximation inside the interfacebecause it is so much thinner than the distance ittravels. Eq. (7) takes the formv(uVC u a C) J C (14)mwhere v is the interface migration rate. The flux isgiven by Eqs. (8) or (9) and is set to 0 in the growinga <strong>phase</strong>. For a given combination <strong>of</strong>v and u a C the solution <strong>of</strong> Eq. (14) yields a concentrationpr<strong>of</strong>ile across the interface and thus also theC content on the g side <strong>of</strong> the interface, i.e. u g/aThe dissipation <strong>of</strong> Gibbs energy due to <strong>diffusion</strong>inside the interface, expressed per mole <strong>of</strong> Fe anddefined as positive, is given byG diffm V mvdC .dm CJ C dz (15)dzBy combining Eqs. (14) and (15) and makinguse <strong>of</strong> ∂ G m /∂ u C m C one obtainsG diffm (u C u a C) d ∂ G mdz (16)dz∂ u CdIn order to perform calculations the variation inthermodynamic properties and mobility inside the<strong>phase</strong> interface must be known. In the solute dragtheory G m is postulated as a function <strong>of</strong> both distanceand composition. Several choices are possible.One choice is to use Eqs. (3) and (13) insidethe interface but rather than Eqs. (4) and (5) onesimply postulates that g(f) 0 and p(f) z/d.This approach was taken by Hillert and Sundman[2] and will be used in the present report.The interface has a finite mobility due to interfacialfriction and <strong>of</strong>ten a linear relation betweeninterface migration rate and driving force isobserved experimentally, i.e.v M V mG i m (17)where G i m is the driving force needed to movethe interface and M is the interfacial mobility.G i m is defined positive for a spontaneous reactionand is the Gibbs energy dissipated by the interfacefriction. The total Gibbs energy dissipated in theinterface is thus given by the sum <strong>of</strong>G diffm and G i m. The dissipation must be suppliedfrom the total driving force available over the interface.It is given per mole <strong>of</strong> atoms and expressedin terms <strong>of</strong> the individual chemical potentials onp. 152 in ref. [13]. By instead introducing theGibbs energy per mole <strong>of</strong> Fe, G m , and its firstderivative we obtainG totm G g mG a m(u g/aC u a C) ∂Gg m∂ u g C(18)where G totm is defined as positive for the consideredreaction to occur. For a given migrationrate v we may, by combining Eqs. (9), (14) and(16)–(18) find the u a C and u g/aC that makes the dissipatedGibbs energy exactly matchG totm . In thelimit <strong>of</strong> low migration rates u a C and u g/aC willapproach the local equilibrium values predicted bythe <strong>phase</strong> diagram but at high velocities they willapproach each other.The solute-drag <strong>modeling</strong> <strong>of</strong> the interfacial reactionsmay be combined with a treatment <strong>of</strong> C <strong>diffusion</strong>in g ahead <strong>of</strong> the interface. By such anapproach it is possible to describe the gradual deviationfrom local equilibrium as the interfacemigration rates increases. In principle such <strong>modeling</strong>could be based on numerical methods as inthe DICTRA s<strong>of</strong>tware [14] or semi-analyticalmethods as the Green function formalism usedrecently by Enomoto [15]. For the sake <strong>of</strong> simplicitywe will here take a simpler approach basedon the linear-gradient approximation. For the thickening<strong>of</strong> a grain-boundary precipitate it yields thefollowing expressionv D (u g/aC u gC ) 2(19)2(u a CuC g/a )(u a CuC g )where is half the thickness <strong>of</strong> the grain-boundaryprecipitate and u gC is the carbon content in g faraway from the interface, i.e. the initial content <strong>of</strong>