Week 13: Parabolic equations
We now turn our attention to parabolic equations. To this end, suppose L and f are as in Week 7 (in divergence or nondivergence form). A linear second-order parabolic equation is an equation of the form
∂tu+Lu=f,where u=u(x,t) is a function u:Ω×I→R, I⊂R an interval. Here we allow the coefficients aij, bi and c of L to depend on t as well, i.e. they are also functions on Ω×I.
Motivation
Parabolic equations arise naturally in the study of heat conduction. For example, the work of Fourier implies that in an isotropic medium Ω⊂Rn, given an initial temperature distribution u0:Ω→R, the temperature u(x,t) at x∈Ω and time t>0 should solve the initial-value problem
∂tu−Δut↘0limu(⋅,0)=0=u0;as we will see, the solution to this problem is uniquely determined by u0 and appropriate boundary values of u, which corresponds to the temperature distribution outside Ω. As t→∞, one would expect the temperature of Ω to reach an equilibrium so that formally, the limit u∞:=limt→∞u(⋅,t) should solve the equation −Δu∞=0. This suggests that a study of parabolic equations could lead to insight into elliptic equations and be another route to proving existence of solutions.
Another motivation for parabolic equations is the variational aspect of elliptic problems (see Week 7). Before describing this connection, consider the classical problem of finding the minima of some f∈C∞(Rn). As we know, if x0 is a minimiser of f, then Df(x0)=0. One way of finding x0 is the method of steepest descent (or negative gradient method), which essentially boils down to considering the system of ODE
x˙(t)x(0)=−Df(x(t)), t>0=givenand analysing the limit limt→∞x(t), which ought to lead to a minimiser x0, if one exists (and if x(t) is defined for all t!). That this is promising is suggested by the fact that
dtdf(x(t))=−∣Df∣2(x,t)≤0so that f∘x is non-increasing. Note also that the solutions x0 to Df(x0)=0 are equilibrium points of the above ODE system.
Going back to the variational formulation of the equation Lu=0 in the case where bi≡0 and u vanishes on ∂Ω, recall that solutions to this equation arise as critical points of the energy
E(u)=21∫Ωi,j∑aij∂iu⋅∂ju+cu2in the sense that the `derivative of E at u in direction φ∈C0∞(Ω)' vanishes for all φ, which we write formally as:
⟨∇E(u),φ⟩:=dtd∣∣∣∣t=0E(u+tφ)=⟨Lu,φ⟩=0,these inner products being the L2(Ω) inner product. In analogy with the negative gradient method, we now look for appropriately regular u:Ω×[0,∞[→R vanishing on ∂Ω×[0,∞[ solving the initial-value problem
∂tu(x,t)u(⋅,0)=−∇E(u(x,t)), (x,t)∈Ω×]0,∞[=given.As with the negative gradient method, we compute that for such a function u,
dtdE(u)=dtd(21BL(u,u))=⟨Lu,∂tu⟩=−∣∂tu∣2≤0so that u indeed tends to decrease the energy associated to the problem. Here we again expect that as t→∞, u(⋅,t) should appropriately tend to a solution to Lu=0.
Weak maximum principle, comparison principle and uniqueness
In this section we consider classical solutions u∈C2,1(Ω×]0,T[)∩C0(Ω×[0,T]) to the Dirichlet initial-boundary value problem
∂tu+Luu(⋅,0)u(x,t)=0 on Ω×]0,T[=u0=g(x) for (x,t)∈∂Ω×[0,T[.(*)for given u0∈C0(Ω) and g∈C0(∂Ω×[0,T]), where we shall assume that L is an elliptic operator in nondivergence form:
Lu=i,j=1∑naij∂i∂ju+i=1∑nbi∂iu+cuwith aij,bi,c∈C0(Ω×[0,T]) for all i,j∈{1,…,n}. Let QT:=Ω×]0,T[ and define the parabolic boundary of QT as
∂′QT:=(∂Ω×[0,T])∪(Ω×{0}).As before, we say that u is a subsolution (or supersolution) to the problem (*) if the first equation is replaced with the inequality ∂tu+Lu≤0 (resp. ∂tu+Lu≥0).
Theorem 13.1 (weak maximum principle for c≡0). Suppose u∈C2,1(Ω×]0,T[)∩C0(Ω×[0,T]) is a subsolution to the initial-boundary value problem (*) with c≡0. Then
QTmaxu=∂′QTmaxu=max{maxg,maxu0}.If instead u is a supersolution to (*), then
QTminu=∂′QTminu=min{ming,minu0}.
Proof. We first consider the case where u is a subsolution and ∂tu+Lu<0.
-
Fix ε>0 small and let (x0,t0)∈QT−ε be such that maxQT−εu=u(x0,t0). We now consider cases.
-
If (x0,t0)∈QT−ε, then we must have ∂tu(x0,t0)=0, Du(x0,t0)=0 and D2u(x0,t0) non-positive definite, which altogether imply that
∂tu+Lu≥0,contradicting the assumed inequality.
-
If (x0,t0)∈Ω×{T−ε}, then we again have Du(x0,t0)=0 and D2u(x0,t0) non-positive definite, and since u(x0,t0)≥u(x0,t0+h) for all small h<0, we obtain that ∂tu(x0,t0)≥0, again leading to the inequality
∂tu+Lu≥0,contradicting our assumption.
We therefore deduce that (x0,t0)∈∂′QT−ε. Therefore, for any small ε>0 and any (x,t)∈QT−ε, we have that
u(x,t)≤∂′QT−εmaxu≤∂′QTmaxu.Since for any (x,t)∈QT we may find an ε0>0 such that (x,t)∈QT−ε for all ε<ε0, we deduce that
QTmaxu=QTsupu≤∂′QTmaxu.Since ∂′QT⊂QT, the opposite inequality is also true and we are done.
Now suppose ∂tu+Lu≤0 and consider v(x,t):=u(x,t)−εt for fixed ε>0. We immediately see that
∂tv+Lv=−ε+∂tu+Lu≤−ε<0so that we fall into the preceding case, whence for all (x,t)∈QT,
u(x,t)−εt≤QTmaxv=∂′QTmaxv=(x,t)∈∂′QTmax(u(x,t)−εt)≤∂′QTmaxu.Taking the limit ε↘0, we arrive at u(x,t)≤max∂′QTu, i.e. supQT≤max∂′QTu, whence we are done. For the supersolution case, replace u with −u.
Remark 13.2. It follows immediately using the same techniques that if u solves (*) with T=∞, then
Q∞supu=∂′Q∞supuand
Q∞infu=∂′Q∞infu.These expressions are of course only finite if g is appropriately bounded.
Just as with elliptic equations, we also have a maximum principle in the case where c≥0.
Theorem 13.3 (weak maximum principle). Suppose u∈C2,1(Ω×]0,T[)∩C0(Ω×[0,T]) is a subsolution to the initial-boundary value problem (*) with c≥0. Then
QTmaxu≤∂′QTmaxu+.If instead u is a supersolution to (*), then
QTminu≥−∂′QTmaxu−.In particular, if u solves (*), then maxQT∣u∣=max∂′QT∣u∣.
Proof. Suppose u is a subsolution. As in the case where c≡0, we first assume that ∂tu+Lu<0, and we may as well suppose u∈C2,1(Ω×[0,T]). If maxu≤0, then we are clearly done, so suppose maxu=u(x0,t0)>0. If this point is in the interior of QT or in Ω×{T}, we deduce as before that ∂tu(x0,t0)≥0, Du(x0,t0)=0 and D2u(x0,t0) is nonpositive definite so that
(∂tu+Lu)(x0,t0)=∂tu(x0,t0)−i,j=1∑naij(x0,t0)∂i∂ju(x0,t0)+i=1∑nbi(x0,t0)∂iu(x0,t0)+c(x0,t0)u(x0,t0)≥0,which yields a contradiction. In the more general case, we consider v(x,t):=u(x,t)−εt and note that if u has a positive maximum, then so does v for sufficiently small ε. This then establishes the claim after taking ε↘0 as in the case where c≡0. The supersolution case follows from replacing u with −u. Finally, if u is a solution, then it is both a subsolution and supersolution so that the inequality minQT∣u∣≤max∂′QT∣u∣ (and thus the equality) immediately follows.
Remark 13.4. In contrast to the elliptic case, maximum principle-type theorems may still be established even if c is not necessarily nonnegative. For instance, if u is a supersolution to (*) with u0,g≥0, then the function v(x,t)=u(x,t)esupc⋅t satisfies
∂tv+(L−c)v=esupc⋅t(∂tu+Lu)+(supc−c)⋅esup∣c∣⋅tu≥0,which allows us to conclude using Theorem 13.3 that v≥0, which in turn implies that u≥0. If instead u is a subsolution and u0,g≥0, we deduce similarly that u≤0. In this case boundedness of c is crucial and follows as a result of our assumptions.
Corollary 13.5 (comparison principle). Suppose that ui∈C2,1(Ω×]0,T[)∩C0(Ω×[0,T]) is a solution to the initial-boundary value problem
∂tui+Luiui(⋅,0)ui(x,t)=fi on Ω×]0,T[=u0i=gi(x) for (x,t)∈∂Ω×[0,T](**)fi∈C2,1(Ω×]0,T[), u0i∈C0(Ω) and gi∈C0(∂Ω×[0,T]) (i∈{1,2}). Suppose furthermore that the inequalities
f1u01g1≤f2≤u02≤g2hold. Under these conditions, u1≤u2 on Ω×[0,T].
Proof. Set u:=u1−u2. It follows immediately that u is a subsolution to (*) with initial-boundary data u0≤0 and g≤0, whence the maximum principle (cf. Remark 13.4) implies that
u1−u2=u≤max{maxu0,maxg}≤0on QT, which establishes the claim.
Remark 13.6. The comparison principle implies in particular that a function u∈C2,1(QT)∩C0(QT) is uniquely determined by the values of ∂tu+Lu and u∣∂′QT.
Harnack inequality
Just as with elliptic equations, we have the following Harnack inequality.
Theorem 13.7 (parabolic Harnack inequality). Let U⋐Ω be connected and t1,t2∈]0,T] with t1<t2. There exists a constant C>0 depending only on U, t1, t2 and L such that whenever u∈C2,1(QT) is a nonnegative solution to ∂tu+Lu=0 on QT,
Usupu(⋅,t1)≤C⋅Uinfu(⋅,t2).Strong maximum principle
Suppose throughout this section that Ω is connected.
Theorem 13.8 (strong maximum principle for c≡0). Suppose u∈C2,1(Ω×]0,T[)∩C0(Ω×[0,T]) is a subsolution to the initial-boundary value problem (*) with c≡0 and maxQTu=u(x0,t0) for some (x0,t0)∈QT. Then u is constant on Qt0.
If instead u is a supersolution to (*), then if minQTu=u(x0,t0) for some (x0,t0)∈QT, we must have that u is constant on Qt0.
Proof. As usual we suppose u is a subsolution. Let
Ω′={x∈Ω:∀t∈]0,t0[, u(x,t)=u(x0,t0)}This set is clearly closed in Ω. We shall now show that it is open and nonempty. To this end, fix a open ball U⋐Ω with x0∈U and let v∈C2,1(RT)∩C0(RT) be a solution to the initial-boundary value problem
∂tv+Lvv∣∂′RT=0 on RT=u∣∂′RT,where RT:=U×]0,T[. By the comparison principle, we have that u≤v≤u(x0,t0) so that v(x0,t0)=u(x0,t0) as well. Now, since u(x0,t0)−v≥0 also solves ∂tv+Lv=0, we may employ the Harnack inequality on some V⋐U connected with x0∈V to deduce that for any t<t0,
Vsup(u(x0,t)−v)≤C(t)Vinf(u(x0,t0)−v)≤C(t)(u(x0,t0)−v(x0,t0))=0so that v≥u(x0,t0), i.e. v≡u(x0,t0) so that u∣∂′Rt0≡u(x0,t0). Since V∋x0 is otherwise arbitrary, it follows that u≡u(x0,t0) on Rt0, i.e. U⊂Ω′. Ω′ is therefore nonempty and, by the preceding argument, open in Ω. By connectedness, we must have Ω′=Ω.
As in the elliptic case, we also obtain a strong maximum principle in the case where c≥0 under appropriate sign conditions.
Theorem 13.9 (string maximum principle for c≥0). Suppose u∈C2,1(Ω×]0,T[)∩C0(Ω×[0,T]) is a subsolution to the initial-boundary value problem (*) with c≥0 and maxQTu=u(x0,t0)≥0 for some (x0,t0)∈QT. Then u is constant on Qt0.
If instead u is a supersolution to (*), then if minQTu=u(x0,t0)≤0 for some (x0,t0)∈QT, we must have that u is constant on Qt0.
Proof. Let Ω′ be as in the preceding proof and again suppose u is a subsolution. Note that if u(x0,t0)=0, the above proof applies mutatis mutandis, so suppose that u(x0,t0)>0. Again let U⋐Ω be an open ball with x0∈U and suppose v∈C2,1(RT)∩C0(RT) solve the initial-boundary balue problem
∂tv+(L−c)vv∣∂′RT=0 on RT=u+∣∣∂′RT.By the weak maximum principle, we see that 0≤v≤u(x0,t0). Moreover, on the set {u≥0}, we have that ∂tu+(L−c)u≤−cu≤0 so that by the comparison principle, u≤v on {u≥0}∩RT, thus also on all of RT, whence v(x0,t0)=u(x0,t0). Thus, since u(x0,t0)−v≥0 solves the equation ∂tv+(L−c)u=0, we again employ the Harnack inequality on V⋐U connected with x0∈V to obtain that for any t<t0,
Vsup(u(x0,t)−v)≤C(t)Vinf(u(x0,t0)−v)≤C(t)(u(x0,t0)−v(x0,t0))=0so that v≥u(x0,t0). Therefore, as before, we must have v≡u(x0,t0) on Rt0 so that u≡u(x0,t0)>0 on {u≥0}∩∂′(V×]0,t0[), whence we must have u≡u(x0,t0) on all of ∂′(V×]0,t0[. We therefore deduce as before that u∣Rt0≡u(x0,t0) i.e. U⊂Ω′ so that arguing as before, Ω′=Ω.