Freezing liquid u2 II Melting ice u1 T = Ts liquid u1 Tl I ice u2 I II Moving boundary problems Heat equation (with convection) with phase change For melting: (I) Liquid: (k1 u1x )x = ρ1 c1 u1t + ρ1 c1 νu1x | {z } possible convection ν = bulk velocity (II) Solid (k2 u2x )x = ρ2 c2 u2t Initial condition u1 = u2 = T0 at t = 0 For t > 0, conditions for melting or freezing freezing u1 = Ts on boundary melting u1 = T` on boundary (u1 = temp of solid) (u1 = temp of liquid) u1 = u2 = Tf (fusion temperature at interface), at x = X(t) The location of the interface X(t) changes in time X(t) is an unknown in the problem. ν = convection in fluid as a result of volume change (melting/freezing) The change of volume at the interface = A∆X ⇒ Aν∆T = V2 − V1 = m ⇒ν = 1− I conduction heat out ∆x ρ2 ρ1 1 ρ2 − 1 ρ1 dX dt i mass is constant In many cases, assume ρ2 = ρ1 (or ρ2 ≈ ρ1 ) ⇒ no convection One more condition is necessary, to determine X(t) Energy balance at X(t) II conduction heat in h x ∂u1 ∂u2 −k1 + k2 ∂x {z ∂x} | heat transfer/unit time ∆t In a volume of ∆X ∆t = m` mass · latent heat dX from ⇐ solid to liquid dt dX from liquid [−k1 u1x + k2 u2x ] = −ρ1 ` ⇐ to solid dt [−k1 u1x + k2 u2x ] = ρ2 ` 62 ! So −k1 ∂u2 ∂u1 + k2 = ∂x ∂x ±ρ2,1 `X 0 (t) (` = latent heat) | {z } +(melting) −(freezing) In most of what we do we take ρ2 = ρ1 Then the system of equations is (with const. coeff’s) 0 < x ≤ X(t) D1 u1xx = u1t D2 u2xx = u2t at the interface ( Dj = kj /ρj cj x > X(t) u1 = u2 = Tf at x = X(t) −k1 u1x + k2 u2x = ±ρ`X 0 (t) at x = X(t) + i.c.’s & b.c.’s Note: this problem is nonlinear! To see this, consider first the interface condition at x = X(t) 0= du dTf = dt dt = ⇒ ∂u ∂x ∂u + at x = X(t) ∂x ∂t ∂t ∂u ∂X ∂t =− ∂u ∂t ∂x Now substitute in second interface condition u1x [−k1 u1x + k2 u2x ] = ∓ρ`u1t nonlinear! Constant coeff, Neumann problem (freezing) D2 u2xx = u2t D1 u1xx = u1t u1 = u 2 = T 0 t>0 lim u2 = T0 x→∞ T0 I II t=0 u1 = Ts at x = 0 u1 = u2 = Tf at x = X(t) k1 u1x − k2 u2x = ρ`X 0 (t), ` = latent heat We know we can convert the pde to an ode using similarity transformation for diffusion (heat) equations u2 Tf Ts u1 X(t) Let z = c √xt in region II u1x c2 c c x u1z √ , u1xx = u1zz , ut = −u1z 3/2 t 2 t t 1 c 1 [D1 c2 u1zz ] = − u1z z ⇒ t t 2 = 63 x Let c = √1 2 D1 u1z = e−z ⇒ u1zz + 2zu1z = 0 2 u1 = C 1 to satisfy u1 = Ts at x = 0 (z = 0) u 1 = C1 Z Z z 0 2 e−z dz + C2 02 e−z dz 0 + Ts Similarly, for u2 , let Z ξ x ξ= √ √ u2 = E 1 2 D2 t u → T0 as ξ → ∞ (x → ∞) u2 = T 0 − E 1 Z ∞ ξ 02 e−ξ dξ 0 + E2 02 e−ξ dξ 0 At the interface: T0 − E 1 u2x |x=X u1x |x=X Note, Z ∞ 2 √X D2 e −ξ 02 0 dξ = C1 √ t 2 1 − X = E1 √ e 4D2 t 2 D2 t 2 1 − X = C1 √ e 4D1 t 2 D1 t Z 2 √X D1 t 0 02 e−z dz 0 + Ts = Tf k1 u1x − k2 u2x = ρ`X 0 (t) X √ 2 tD appears in both condi√ tions: This suggests, take X = |{z} 2K tD1 ⇒ C1 Z K 0 E1 q 02 e−z dz 0 = −Ts + Tf Z∞ 02 e−ξ dξ 0 = T0 − Tf D1 K D2 2 2 − p k1 (Tf − Ts )e−K k (T0 − Tf )e−K D1 /D2 √2 + = ρ`K D1 √ RK 02 ∞ R 0 2 D2 2 D1 0 e−z dz e−ξ 2 dξ 0 Finally, q D1 D2 Tf − T s u1 = T s + R K −z 2 dz 0 e Z 2 √x 0 Z∞ T0 − T f u2 = T 0 − q R∞ D1 D2 e−z 2 dz K 64 D1 t 2 √x D2 t 2 e−z dz 2 e−z dz Equation for K ⇒ determines behavior of √ interface (const. t) increases as Tf − Ts ↑ X(t) decreases as T0 − Tf ↑ X0 t Note: Similarly solutions do not always work! e.g. for finite domains, varying media, 2 phases present initially and b.c. of different forms, can not use similarity solution Stefan problem Simplified version - (without similarity solutions) In the previous problem, some simplications occur for T 0 ∼ Tf Then u2 ∼ T0 (E1 = 0) and we solve the equation only for solid T0 − T s u1 ∼ T s + R K 0 Equation for K: e−z 2 dz Z 2 √x D1 t 0 2 e−z dz 2 p k1 (Tf − Ts )e−K = ρ`K D1 √ R K −z 2 2 D1 0 e dz For water: Tf − Ts is small ⇒ use Taylor series expansion for ⇒ K 2 ∝ Tf − Ts ⇒ Z K 0 2 e−z dz ∼ 0 + K slow movement for small Tf − Ts For time-varying media, and Tf −Ts 1, we can use a similar approach (Lunardini, 1981). Consider the problem with convection, two-phase initial condition, melting problem. Here we neglect the variation in the solid region. liquid X(0) = X0 6= 0 solid Equation for liquid: u1 X D1 u1xx = u1t + (1 − u2 ρ2 0 ρ1 )X (t)u1x u1 (x, 0) = T0 , u1 (X, t) = Tf , u1 (0, t) = Ts −k1 u1x = ρ1 `X 0 (t) at x = X Now we nondimensionalize to simplify, and indicate the small parameter ST (Stefan number), related to Tf − Ts 65 u1 − T 0 X D1 t x , θ= ξ= τ= 2 X0 Ts − T f X0 X0 1 u1x = θy (Ts − Tf ) X0 D1 D1 u1t = θτ 2 (Ts − Tf ), Xt = ξτ 2 X0 X X0 0 D1 D1 D1 ρ2 ⇒ θyy (Ts − Tf ) = θτ 2 (Ts − Tf ) + (1 − )ξτ θy 2 · (Ts − Tf ) 2 X0 X0 ρ1 X0 ρ2 ⇒ θyy = θτ + (1 − )ξτ θy ρ1 y= Interface condition D1 ξτ X0 = −(Tf − Ts )θy |y=ξ 1 k X0 ρ1 ` Substitute in θ equation ⇒ θyy = θτ + (1 − ρ2 /ρ1 )ST θy |y=ξ c (Tf − Ts ) ` ≡ ξ(0) = 1, θ(y, 0) = 0, Ts − T 0 ≡ θs θ(0, τ ) = Ts − T f k =c ST 1, ρ1 D1 θ(ξ = 1, τ ) = Tf − T 0 ≡φ Ts − T f ST is the Stefan number. Using a perturbation expansion in S T θ ∼ θ 0 + S T θ1 + . . . ξ ∼ ξ 0 + S T ξ1 + . . . ⇒ to leading order ignore convection to leading order ST 1 z }| { ρ2 θ0yy = θ0τ + (1 − )ST θ0y |y=ξ ρ1 θ0 (y, 0) = 0, θ0 (0, τ ) = θs θ0 (1, τ ) = φ At the interface: ξ0τ = −ST θ0 y|y=ξ ⇐ keep ST term otherwise we get only constant for ξ 0 The interface equation suggests the new time scale τ̂ = S T τ =0 ⇒ z }| { θ0yy = ST θ0 τ̂ , neglect for ST 1 66 θ = θ(y, τ̂) ξ0τ̂ = −θ0y |y=ξ , θ(y, 0) = 0 θ(0, τ ) = θs θ(ξ0 , τ ) = φ ξ(0) = 1 Leading order approximation, θ0 is linear in y θ0 = θ s − ξ0τ̂ = + ξ0 = Note: q y · (θs − 1) ξ0 1 (θs − 1) ξ0 θ0 (ξ0 , τ ) = φ0 = 1 φ= | {z } Tf − T 0 Ts + (Tf − Ts ) − T0 = Ts − T 0 Ts − T 0 2(θs − 1)τ̂ + 1 θs − 1 = Ts − T f Tf − T 0 Ts − T 0 − = =φ Ts − T f Ts − T f Ts − T f Other problems where similarity solutions work: (diffusivity) When thermal conductivity varies linearly with temperature we can also use similarity solutions: I x = X(t) II equations: solid liquid 0 < x < X(t) u1t = (D1 u1x )x ρ1 x > X(t) u2t + (1 − )Xt u2x = (D2 u1x )x ρ2 u1 − T s D1 = D01 1 + β1 T − Ts 0 u2 − T s D2 = D02 1 + β2 T0 − T s If we again make the transformation z = √xt √ X = 2λ D01 t c x c ⇒ u1x = u1z √ u1t = − 3/2 u1 2t t 2 u1 − T s c z 1 + β1 D01 u1z ⇒ − u1z = 2t T0 − T s z t Let c = √1 , 2 D01 θ1 = u1 −Ts T0 −Ts ⇒ −2zθ1z = ((1 + β1 θ1 )θ1z )z 67 Note, when β = 0, we’re back to the Neumann problem, and the solution was in terms of error functions. Here we define the modified error function, call it φ 1 , which satisfies ((1 + β1 φ1 )φ1 z )z + 2zφ1z = 0 with φ1 (0) = 0, then ⇒ u 1 = T s + C 1 φ1 x √ 2 D01 t θ1 = 0 at u1 = Ts at x = 0 · (T0 − Ts ) We can even solve for C1 , √ Since u1 = Tf at x = X(t) = 2λ D01 t Then Tf = Ts + T0 − Ts C1 φ(λ) x φ1 2√D t Tf − T s 01 ⇒ C1 = ⇒ u1 = Ts + (Tf − Ts ) (T0 − Ts )φ(λ) φ1 (λ) We still have to determine λ from the interface condition Similarly, for u2 : x Let ζ = d √ , substitute for X(t) t . √ u2ζ ρ1 2 λ D01 d u2 − T s d2 √ u2ζ = + (1 − ) . √ D02 −ζ u2ζ 1 + β2 2t ρ2 t T0 − T s ζ t 2 t Let u2 −Ts T0 −Ts = θ2 , d = √1 2 D02 ρ1 ⇒ 2 1− ρ2 p λ D01 d − 2ζ θ2ζ = ((1 + β2 θ2 )θ2ζ )ζ Again we have an ode, with λ as a constant. Let the solution be φ2 , with φ2 → 1 as ζ → ∞ Then (θ2 → 1 as u2 → T0 ) and as x → ∞ θ2 = 1 − C2 (1 − φ2 (ζ)) Note: We can also write the equation for φ 2 in the same form as the equation for φ1 . p ρ1 )(λ D01 d) − ζ ρ2 = [(1 + β2 θ2 )θ2y ]y Let − y = (1 − ⇒ −2yθ2y 68 and u2 = (T0 − Ts )θ2 + Ts At the interface: " Tf = (T0 − Ts ) 1 − C2 1 − φ2 !!# √ λ D01 √ + Ts D02 √ Once we have u2 , u1 , can get an equation for λ from X = 2λ D01 t, using the interface condition. √ ρ1 `λ D01 √ − k2 u2x =− k1 u1x √ √ t x=2λ D01 t x=X(t)=2λ D01 t We have to compute φ1 , φ2 , φ01 , φ02 for each possible value of λ and substitute to find the root in the interface condition. " !# √ (Tf − Ts )φ01 (λ) 1 k2 D 01 √ √ k1 − (T0 − Ts )C2 φ02 λ √ φ1 (λ) 2 D01 2 D02 D02 p = −ρ1 `λ D01 69