Citation: Hitoshi Ishii. The vanishing discount problem for monotone systems of Hamilton-Jacobi equations. Part 1: linear coupling[J]. Mathematics in Engineering, 2021, 3(4): 1-21. doi: 10.3934/mine.2021032
[1] | Hitoshi Ishii . The vanishing discount problem for monotone systems of Hamilton-Jacobi equations: a counterexample to the full convergence. Mathematics in Engineering, 2023, 5(4): 1-10. doi: 10.3934/mine.2023072 |
[2] | Juan-Carlos Felipe-Navarro, Tomás Sanz-Perela . Semilinear integro-differential equations, Ⅱ: one-dimensional and saddle-shaped solutions to the Allen-Cahn equation. Mathematics in Engineering, 2021, 3(5): 1-36. doi: 10.3934/mine.2021037 |
[3] | Giacomo Ascione, Daniele Castorina, Giovanni Catino, Carlo Mantegazza . A matrix Harnack inequality for semilinear heat equations. Mathematics in Engineering, 2023, 5(1): 1-15. doi: 10.3934/mine.2023003 |
[4] | Stefano Almi, Giuliano Lazzaroni, Ilaria Lucardesi . Crack growth by vanishing viscosity in planar elasticity. Mathematics in Engineering, 2020, 2(1): 141-173. doi: 10.3934/mine.2020008 |
[5] | Daniele Del Santo, Francesco Fanelli, Gabriele Sbaiz, Aneta Wróblewska-Kamińska . On the influence of gravity in the dynamics of geophysical flows. Mathematics in Engineering, 2023, 5(1): 1-33. doi: 10.3934/mine.2023008 |
[6] | François Murat, Alessio Porretta . The ergodic limit for weak solutions of elliptic equations with Neumann boundary condition. Mathematics in Engineering, 2021, 3(4): 1-20. doi: 10.3934/mine.2021031 |
[7] | Claudio Canuto, Davide Fassino . Higher-order adaptive virtual element methods with contraction properties. Mathematics in Engineering, 2023, 5(6): 1-33. doi: 10.3934/mine.2023101 |
[8] | La-Su Mai, Suriguga . Local well-posedness of 1D degenerate drift diffusion equation. Mathematics in Engineering, 2024, 6(1): 155-172. doi: 10.3934/mine.2024007 |
[9] | Quoc-Hung Nguyen, Nguyen Cong Phuc . Universal potential estimates for $ 1 < p\leq 2-\frac{1}{n} $. Mathematics in Engineering, 2023, 5(3): 1-24. doi: 10.3934/mine.2023057 |
[10] | Stefano Biagi, Serena Dipierro, Enrico Valdinoci, Eugenio Vecchi . A Hong-Krahn-Szegö inequality for mixed local and nonlocal operators. Mathematics in Engineering, 2023, 5(1): 1-25. doi: 10.3934/mine.2023014 |
Dedicated to Italo Capuzzo Dolcetta with friendship, respect, and admiration on the occasion of his retirement.
We consider the weakly coupled m-system of Hamilton-Jacobi equations
λvλ+Bvλ+H[vλ]=0 in Tn,(Pλ) |
where m∈N, λ is a nonnegative constant, called the discount factor in terms of optimal control. Here Tn denotes the n-dimensional flat torus, H=(Hi)i∈I is a family of Hamiltonians given by
Hi(x,p)=maxξ∈Ξ[−gi(x,ξ)⋅p−Li(x,ξ)],(H) |
where I={1,…,m}, Ξ is a given compact metric space, g=(gi)i∈I∈C(Tn×Ξ,Rn)m and L=(Li)i∈I∈C(Tn×Ξ)m. The unknown in (Pλ) is an Rm-valued function vλ=(vλi)i∈I on Tn, B:C(Tn)m→C(Tn)m is a linear map represented by a matrix B=(bij)i,j∈I∈C(Tn)m×m, that is,
(Bu)i(x)=(B(x)u(x))i:=∑j∈Ibij(x)uj(x) for (x,i)∈Tn×I. |
We use the abbreviated notation H[vλ] to denote (Hi(x,Dvλi(x))i∈I. The system is called weakly coupled since the i-th equation depends on Dvλ only through Dvλi but not on Dvλj, with j≠i. Problem (Pλ) can be stated in the component-wise style as
λvλi+∑j∈Ibij(x)vλj+Hi(x,Dvλi)=0 in Tn, i∈I. |
We are mainly concerned with the asymptotic behavior of the solution vλ of (Pλ) as λ→0+. Asymptotic problems in this class are called the vanishing discount problem, in view that the constant λ in (Pλ) appears as a discount factor in the dynamic programming PDE in optimal control.
Recently, there has been a keen interest in the vanishing discount problem concerned with Hamilton-Jacobi equations and, furthermore, fully nonlinear degenerate elliptic PDEs. We refer to [1,7,10,12,19,20,23,24,25,27] for relevant work. The asymptotic analysis in these papers relies heavily on Mather measures or their generalizations and, thus, it is considered part of the Aubry-Mather and weak KAM theories. For the development of these theories we refer to [14,16,17] and the references therein.
We are here interested in the case of systems of Hamilton-Jacobi equations and, indeed, Davini and Zavidovique in [12] have established a general convergence result for the vanishing discount problem for (Pλ). We establish a result (Theorem 9 below) similar to the main result of [12]. In establishing our convergence result, we adapt the argument in [23] (see also [18]) to the case of systems, especially, to construct generalized Mather measures for (Pλ). Regarding the recent developments of the weak KAM theory and asymptotic analysis in its influence for systems, we refer to [5,6,26,28,29,30,33].
The new argument, which is different from that of [12], makes it fairly easy to build a generalized Mather measure for systems in a wide class. One advantage of our argument is that it allows us to treat the case where the coupling matrix B in (Pλ) depends on the space variable x∈Tn. As in [20,23], our approach is applicable to the system with nonlinear coupling of fully nonlinear second-order elliptic PDEs, but we restrict ourselves in this paper to the case of the linearly coupled system of first-order Hamilton-Jacobi equations. Another possible approach for constructing generalized Mather measures is the so-called adjoint method (see [5,15,19,27,33]).
This paper is part 1 of our study of the vanishing discount problem for weakly coupled systems of Hamilton-Jacobi equations and deals only with the linear coupling and with compact control sets Ξ. These restrictions make the presentation of our results clear and transparent. In part 2 [20], we remove these restrictions and establish a general convergence result extending Theorem 9 below. Sections 5 and 6 are devoted to the study of ergodic problems of the form Bu+H[u]=c, where c∈Rm is an unknown as well. Also, thanks to the linearity of the coupling, our results on the ergodic problems are applied to extend the scope of Theorem 9. On the other hand, the role of the ergodic problem, with general right-hand side c, is not clear at least for the author in the vanishing discount problem for the systems with the nonlinear coupling.
In this paper, we adopt the notion of viscosity solution to (Pλ), for which the reader may consult [2,4,8,31].
To proceed, we give our main assumptions on the system (Pλ).
We assume that H is coercive, that is, for any i∈I,
lim|p|→∞minx∈TnHi(x,p)=∞.(C) |
This is a convenient assumption, under which any upper semicontinuous subsolution of (Pλ) is Lipschitz continuous on Tn.
We assume that B(x)=(bij(x)) is a monotone matrix for every x∈Tn, that is, it satisfies
for any x∈Tn, if u=(ui)i∈I∈Rm and uk=maxi∈Iui≥0, then (B(x)u)k≥0. (M) |
This is a natural assumption that (Pλ) should possess the comparison principle between a subsolution and a supersolution.
In what follows we set, for λ≥0,
Bλ=λI+B, |
and (Pλ) can be written as
Bλvλ+H[vλ]=0 in Tn. |
We use the symbol u≤v (resp., u≥v) for m-vectors u,v∈Rn to indicate ui≤vi (resp., ui≥vi) for all i∈I.
The following theorem is well-known: see [13,22] for instance.
Theorem 1. Assume (C) and (M). Let λ>0. Then the exists a unique solution vλ∈Lip(Tn)m of (Pλ). Also, if v=(vi),w=(wi) are, respectively, upper and lower semicontinuous on Tn and a subsolution and a supersolution of (Pλ), then v≤w on Tn.
Henceforth, let 1 denote the vector (1,…,1)∈Rm.
Outline of proof. We follow the line of the arguments in [22]. Although [22] is concerned with the case when the domain is an open subset of a Euclidean space, the results in [22] is valid in the case when the domain is Tn.
Choose a large constant C>0 so that the constant functions ±C1 are a supersolution and a subsolution of (Pλ), respectively. (See also (2.3) below.) According to [22,Theorems 3.3,Lemma 4.8], there is a function vλ=(vλi)i∈I:Tn→Rm such that the upper and lower semicontinuous envelopes (vλ)∗ and vλ∗ are a subsolution and a supersolution of (Pλ), respectively. By the coercivity assumption (C), we find (see [9,Theorem I.14], [21,Example 1]) that the functions (vλi)∗ are Lipschitz continuous on Tn. Let R1>0 be a Lipschitz bound of the functions (vλi)∗. To take into account the Lipschitz property of (vλi)∗, we modify the Hamiltonian H. Fix any M>0 so that
max(x,ξ,i)∈Tn×Ξ×I|gi(x,ξ)|<M, | (1.1) |
and choose constants N>0 and R2>0 so that
Hi(x,p)≥M|p|−N for (x,p,i)∈Tn×BR1×I, |
and, in view of (1.1),
Hi(x,p)≤M|p|−N for (x,p,i)∈Tn×BR2×I. |
Define G=(Gi)i∈I∈C(Tn×Rn)m by
Gi(x,p)=Hi(x,p)∨(M|p|−N). |
By the choice of R1, it is easy to see that (vλ)∗ is a subsolution of
λu+Bu+G[u]=0 Tn. | (1.2) |
Also, since G≥H, vλ∗ is a supersolution of (1.2). Observe furthermore that, if |p|≥R2, then
Gi(x,p)=M|p|−N for (x,i)∈Tn×I, |
the functions Gi are uniformly continuous on Tn×BR2, and hence, for some continuous function ω on [0,∞), with ω(0)=0,
|Gi(x,p)−Gi(y,p)|≤ω(|x−y|) for (x,y,p)∈(Tn)2×Rn,i∈I. |
The last inequality above shows that G satisfies [22,(A.2)], which allows us to apply [22,Theorem 4.7], to conclude that (vλ)∗≤vλ∗ on Tn and, moreover, that vλ∈Lip(Tn)∗. Similarly, we deduce that the comparison assertion is valid. Thus, vλ is a unique solution of (Pλ).
Regarding the coercivity (C), the following proposition is well-knwon.
Proposition 2. The function given by (H) satisfies (C) if and only if there exists δ>0 such that
Bδ⊂co{gi(x,ξ):ξ∈Ξ} for(x,i)∈Tn×I, | (1.3) |
where co designates "convex hull" and Bδ denotes the open ball with origin at the origin and radius δ.
Outline of proof. Set C(x,i)=co{gi(x,ξ):ξ∈Ξ}. Assume that (1.3) is valid for some δ>0 and observe that
Hi(x,p,u)≥maxξ∈Ξ−gi(x,ξ)⋅p−max(x,i,ξ)∈Tn×I×ΞLi(x,ξ)=maxq∈C(x,i)−q⋅p−max(x,i,ξ)∈Tn×I×ΞLi(x,ξ)≥supq∈Bδ−q⋅p−max(x,i,ξ)∈Tn×I×ΞLi(x,ξ)=δ|p|−max(x,i,ξ)∈Tn×I×ΞLi(x,ξ), |
which shows that (C) holds.
Next, assume that (1.3) does not hold for any δ>0. Then there exists (xk,ik)∈Tn×I for each k∈N such that
B1/k∖C(xk,ik)≠∅. |
For each k∈N select qk∈B1/k∖C(xk,ik) and rk∈C(xk,ik) so that rk is the point of C(xk,ik) closest to qk. (Notice that C(xk,ik) is a compact convex set.) Setting νk=(qk−rk)/|qk−rk|, we find that
νk⋅(q−rk)≤0 for q∈C(xk,ik). |
Sending k→∞ along an appropriate subsequence, say (kj)j∈N, we find that there are a unit vector ν=limj→∞νkj of Rn, r=limj→∞rkj∈Rn and (x,i)∈Tn×I such that
r∈C(x,i) and ν⋅(q−r)≤0 for q∈C(x,i). |
If r≠0, then we have ν=−r/|r|, since limk→∞qk=0, and the inequality above reads
ν⋅q≤−|r|<0 for q∈C(x,i). |
These observations imply that for t>0,
Hi(x,−tν)=maxξ∈Ξtgi(x,ξ)⋅ν−minξ∈ΞLi(x,ξ)≤−minξ∈ΞLi(x,ξ), |
which shows that (C) does not hold. This completes the proof.
The rest of this paper is organized as follows. In Section 2, we recall some basic facts concerning monotone matrices. In Section 3, we study viscosity Green-Poisson measures for our system, which are crucial in our asymptotic analysis. We establish the main result for the vanishing discount problem in Section 4. We study the ergodic problem (P0) in the cases when B is irreducible, and B is a constant matrix, respectively, in Sections 5 and 6, and combine the results with the analysis on the vanishing discount problem of Section 4.
Here we are concerned with m×m real matrix B=(bij)i,j∈I.
Let ei denote the vector (ei1,…,eim), with eii=1 and eij=0 if i≠j.
Lemma 3. Let B=(bij) be a real m×m matrix. It is monotone if and only if
bij≤0 ifi≠j and ∑j∈Ibij≥0 fori∈I. | (2.1) |
We remark that if B satisfies (2.1), then
bii=∑j∈Ibij−∑j≠ibij≥0. | (2.2) |
Proof. We assume first that B is monotone. Since
1i=1=maxj1j>0, |
By the monotonicity of B, we have
0≤(B1)i=m∑j=1bij1j=m∑j=1bij for i∈I. | (2.3) |
Similarly, if i≠j and t≥0, then we have 1=(ei−tej)i=maxk∈I(ei−tej)k and hence,
0≤(B(ei−tej))i=bii−tbij, |
from which we find by sending t→∞ that
bij≤0. |
Hence, (2.1) is satisfied.
Next, we assume that (2.1) holds. Let u∈Rm satisfy
uk=maxi∈Iui≥0. |
Then we observe that, since uk≥uj for all j∈I,
(Bu)k=∑j∈Ibkjuj=bkkuk+∑j≠kbkjuj=bkkuk+∑j≠kbkjuk=uk∑j∈Ibkj≥0. |
Thus, B is monotone.
Lemma 4. Let u∈Rm and C≥0 be a constant. Let B be an m×m real monotone matrix. Then we have
B(u−C1)≤Bu≤B(u+C1). |
Proof. Using Lemma 3, we see that
(B1)i=∑j∈Ibij≥0 for i∈I, |
which states that B1≥0. It is then obvious to compute that
B(u+C1)−Bu=CB1,Bu−B(u−C1)=CB1 and CB1≥0 |
and therefore,
B(u+C1)≥Bu≥B(u−C1). |
For λ≥0 we write F(λ) for the set of all (ϕ,u)∈C(Tn×Ξ)m×C(Tn)m such that u is a subsolution of
Bλu+Hϕ[u]=0 in Tn, |
where Hϕ=(Hϕ,i)i∈I and
Hϕ.i(x,p)=maxξ∈Ξ(−gi(x,ξ)⋅p−ϕi(x,ξ)). |
In the above, since ϕ is bounded on Tn×Ξ, if H satisfies (C), then Hϕ satisfies (C).
Lemma 5. The set F(λ) is a convex cone in C(Tn×Ξ)m×C(Tn)m with vertex at the origin.
Proof. Recall [3,Remark 2.5] that for any u∈Lip(Tn)m, u is a subsolution of
Bλu+H[u]=0 in Tn |
if and only if for any i∈I,
(Bλu)i(x)+Hi(x,Dui(x))≤0 a.e. in Tn, |
and by the coercivity (C) that for any (ϕ,u)∈F(λ), we have u∈Lip(Tn)m.
Fix (ϕ,u),(ψ,v)∈F(λ) and t,s∈[0,∞). Fix i∈I and observe that
(Bλu)i(x)+Hϕ,i(x,Dui(x))≤0 a.e. in Tn,(Bλv)i(x)+Hψ,i(x,Dvi(x))≤0 a.e. in Tn, |
which imply that there is a set N⊂Tn of Lebesgue measure zero such that
(Bλu)i(x)≤g(x,ξ)⋅Dui(x)+ϕi(x.ξ) for all (x,ξ)∈Tn∖N×Ξ,(Bλv)i(x)≤gi(x,ξ)⋅Dvi(x)+ψi(x,ξ) for all (x,ξ)∈Tn∖N×Ξ. |
Multiplying the first and second by t and s, respectively, adding the resulting inequalities and setting w=tu+sv, we obtain
(Bλw)i(x)≤g(x,ξ)⋅Dwi(x)+(tϕi+sψ)(x.ξ) for all (x,ξ)∈Tn∖N×Ξ, |
which readily implies that t(ϕ,u)+s(ψ,v)∈F(λ).
We refer the reader to [23,Lemma 2.2] for another proof of the above lemma.
We establish a representation formula for the solution of (Pλ), with λ>0, by modifying the argument in [23] (see also [18]).
For any nonnegative Borel measure ν on Tn×Ξ and ϕ∈C(Tn×Ξ), we write
⟨ν,ϕ⟩=∫Tn×Ξϕ(x,ξ)ν(dx,dξ). |
Similarly, for any collection ν=(νi)i∈I of nonnegative Borel measures on Tn×Ξ and ϕ=(ϕi)∈C(Tn×Ξ)m, we write
⟨ν,ϕ⟩=∑i∈I⟨νi,ϕi⟩∈R. |
Note that any collection ν=(νi)i∈I of nonnegative Borel measures on Tn×Ξ is regarded as a nonnegative Borel measure on Tn×Ξ×I and vice versa.
We set
ρi(x):=∑j∈Ibij(x) for i∈I. |
Note that
B1=(b11(x)⋯b1m(x)⋮⋮bm1(x)⋯bmm(x))(1⋮1)=(ρ1(x)⋮ρm(x)) and Bλ1=(λ+ρ1(x)⋮λ+ρm(x)). | (3.1) |
By assumption (M) and Lemma 3, we have ρi≥0 on Tn for all i∈I.
Given a constant λ>0, let PBλ denote the set of of nonnegative Borel measures ν=(νi)i∈I on Tn×Ξ×I such that
⟨ν,Bλ⟩=1. |
The last condition reads
∑i∈I(λ|νi|+⟨νi,ρi⟩)=1, |
where |νi| denotes the total mass of νi on Tn×Ξ. Note as well that PBλ can be identified with the space of Borel probability measures on Tn×Ξ×I by the correspondence between ν=(νi)i∈I and ∑i∈I(λ+ρi)νi⊗δi, where ⊗ indicates the product of two measures and δi denotes the Dirac measure at i. If we set μ:=∑i∈I(λ+ρi)νi⊗δi and consider μ as a collection (μi) of measures on Tn×Ξ, then νi=(λ+ρi)−1μi. We denote simply by P the space of Borel probability measures on Tn×Ξ×I.
For λ≥0 and (z,k)∈Tn×I we set
G(z,k,λ):={ϕ−uk(z)Bλ1:(ϕ,u)∈F(λ)}⊂C(Tn×Ξ)m, |
and
G′(z,k,λ)={ν=(νi)i∈I∈PBλ:⟨ν,f⟩≥0 for f=(fi)∈G(z,k,λ)}. |
Theorem 6. Assume (H), (C) and (M). Let λ>0 and (z,k)∈Tn×I. Let vλ∈C(Tn×I) be the unique solution of (Pλ). Then there exists a νz,k,λ=(νz,k,λi)i∈I∈G′(z,k,λ) such that
vλk(z)=⟨νz,k,λ,L⟩. | (3.2) |
We remark that for any ν∈G′(z,k,λ) we have ⟨ν,L⟩≥vλk(z)⟨ν,Bλ⟩=vλk(z) and, accordingly, in the theorem above, the measures νz,k,λ has the minimizing property:
vλk(z)=⟨νz,k,λ,L⟩=minν∈G′(z,k,λ)⟨ν,L⟩. | (3.3) |
We call any minimizing family (νi)i∈I∈PBλ of the optimization problem above a viscosity Green-Poisson measure for (Pλ).
Proof. Note first that (L,vλ)∈F(λ) and hence, for any ν∈G′(z,k,λ),
0≤⟨ν,L−vλk(z)Bλ⟩=⟨ν,L⟩−vλk(z)⟨ν,Bλ⟩=⟨ν,L⟩−vλk(z). | (3.4) |
Next, we show that
sup(ϕ,u)∈F(λ)infν∈PBλ⟨ν,L−ϕ+(uk(z)−vλk(z))Bλ⟩=0. | (3.5) |
Note that for z∈Tn,
sup(ϕ,u)∈F(λ)infν∈PBλ⟨ν,L−ϕ+(uk(z)−vλk(z))Bλ⟩≥infν∈PBλ⟨ν,L−ϕ+(uk(z)−vλk(z))Bλ⟩|(ϕ,u)=(L,vλ)=0. |
Hence, in order to prove (3.5), we only need to show that
sup(ϕ,u)∈F(λ)infν∈PBλ⟨ν,L−ϕ+(uk(z)−vλk(z))Bλ⟩≤0. | (3.6) |
We postpone the proof of (3.6) and, assuming temporarily that (3.5) is valid, we prove that there exists ν∈G′(z,k,λ) such that
vλk(z)=⟨ν,L⟩, | (3.7) |
which, together with (3.4), completes the proof.
To prove (3.7), we observe that PBλ and, by Lemma 5, F(λ) are convex,
PBλ∋ν↦⟨ν,L−ϕ+(uk(z)−vλk(z))Bλ⟩ |
is convex and continuous, in the topology of weak convergence of measures, for any (ϕ,u)∈F(λ) and
F(λ)∋(ϕ,u)↦⟨ν,L−ϕ+(uk(z)−vλk(z))Bλ⟩ |
is concave and continuous for any ν∈PBλ. Hence, noting moreover that Tn×Ξ×I is a compact set, we apply the minimax theorem ([34,32]), to find from (3.5) that
0=sup(ϕ,u)∈F(λ)minν∈PBλ⟨ν,L−ϕ+(uk(z)−vλk(z))Bλ⟩=minν∈PBλsup(ϕ,u)∈F(λ)⟨ν,L−ϕ+(uk(z)−vλk(z))Bλ⟩. | (3.8) |
Observe by using the cone property of F(λ) that
sup(ϕ,u)∈F(λ)⟨ν,uk(z)Bλ−ϕ⟩={0 if ν∈G′(z,k,λ),∞ if ν∈PBλ∖G′(z,k,λ). |
This and (3.8) yield
0=minν∈PBλsup(ϕ,u)∈F(λ)⟨ν,L−ϕ+(uk(z)−vλk(z)Bλ⟩=minν∈G′(z,k,λ)sup(ϕ,u)∈F(λ)⟨ν,L−vλk(z)Bλ⟩=minν∈G′(z,k,λ)⟨ν,L−vλk(z)Bλ⟩=minν∈G′(z,k,λ)(⟨ν,L⟩−vλk(z)⟨ν,Bλ⟩)=minν∈G′(z,k,λ)⟨ν,L⟩−vλk(z), |
which proves (3.7).
It remains to show (3.6). For this, we argue by contradiction and thus suppose that (3.6) does not hold. Accordingly, we have
sup(ϕ,u)∈F(λ)infν∈PBλ⟨ν,L−ϕ+(uk(z)−vλk(z))Bλ⟩>ε |
for some ε>0. We may select (ϕ,u)∈F(λ) so that
infν∈PBλ⟨ν,L−ϕ+(uk(z)−vλk(z))Bλ⟩>ε. |
That is, for any ν∈PBλ, we have
⟨ν,L−ϕ+(uk(z)−vλk(z))Bλ⟩>ε=⟨ν,εBλ⟩. |
Plugging ν=(λ+ρi)−1δ(x,ξ,i)∈PBλ, with any (x,ξ,i)∈Tn×Ξ×I, into the above, we find that
(Li−ϕi)(x,ξ)−(vλk(z)−uk(z)−ε)(Bλ1)i>0. |
Hence, we have
ϕ(x,ξ)<L(x,ξ)+(uk(z)−vλk(z)−ε)Bλ1 for (x.ξ)∈Tn×Rn. |
This ensures that u is a subsolution of
Bλu+H[u]=(uk(z)−vλk(z)−ε)Bλ1 in Tn, |
which implies that u−(uk(z)−vλk(z)−ε)1 is a subsolution of (Pλ). By comparison (Theorem 1), we get
u(x)−(uk(z)−vλk(z)−ε)≤vλ(x) for x∈Tn. |
The k-th component of the above, evaluated at x=z, yields an obvious contradiction. Thus we conclude that (3.6) holds.
We have the following characterization of G′(z,k,λ).
Proposition 7. Assume (H), (C) and (M) hold. Let ν=(νi)i∈I∈PBλ and (z,k,λ)∈Tn×I×(0,∞). Then we have ν∈G′(z,k,λ) if and only if
∑i∈I⟨νi,(Bλψ)i−gi⋅Dψi⟩=ψk(z) for ψ=(ψi)i∈I∈C1(Tn)m. | (3.9) |
Proof. Assume first that ν∈G′(z,k,λ). Fix any ψ=(ψi)i∈I∈C1(Tn)m and define ϕ=(ϕi)i∈I∈C(Tn×I)m by
ϕi(x,ξ)=(Bλψ)i(x)−gi(x,ξ)⋅Dψi(x). |
Observe that u:=±ψ satisfy, respectively,
Bλu+H±ϕ[u]=0 in Tn, |
and, hence,
±(ϕ,ψ)∈F(λ). |
Since ν∈G′(z,k,λ), we have
±ψk(z)≤⟨ν,±ϕ⟩=±⟨ν,ϕ⟩, |
respectively, which shows that (3.9) is valid.
Now, assume that (3.9) is satisfied. Fix any (u,ϕ)∈F(λ). As noted in the proof of Theorem 1, we have u∈Lip(Tn). By the standard mollification technique, given a positive constant ε>0, we can approximate u by a smooth function uε so that
maxTn|u−uε|<ε and Bλuε+Hϕ[uε]≤εBλ1 in Tn. |
The last inequality reads
Bλuεi(x)−gi(x,ξ)⋅Duεi(x)−ϕi(x,ξ)≤ε(Bλ1)i(x) for (x,ξ,i)∈Tn×Rn×I. |
Integrating the above by νi, summing up in i∈I and using (3.9), we get
uεk(z)−⟨ν,ϕ⟩≤ε⟨ν,Bλ⟩=ε. |
Sending ε→0 shows that ν∈G′(z,k,λ).
It is convenient to restate the theorem above as follows. For μ=(μi)i∈I∈P and λ>0, consider ν=(νi)i∈I∈PBλ given by
νi:=(λ+ρi)−1μi=1(Bλ1)iμi. |
(Notice by the above definition that ⟨ν,Bλ⟩=⟨μ,⟩=1.) Observe that for ϕ=(ϕi)i∈I∈C(Tn×Ξ)m,
⟨ν,ϕ⟩=∑i∈I⟨νi,ϕi⟩=∑i∈I⟨μi,(λ+ρi)−1ϕi⟩, |
and that for any (z,k)∈Tn×I, we have ν∈G′(z,k,λ) if and only if
∑i∈I⟨μi,(λ+ρi)−1ϕi⟩≥uk(z) for (ϕ,u)∈F(λ). | (3.10) |
The condition above is stated in the spirit of Proposition 7 as
∑i∈I⟨μi,(λ+ρi)−1((Bλψ)i−gi⋅Dψi)⟩=ψk(z) for ψ=(ψi)i∈I∈C1(Tn)m. |
We define
P(z,k,λ)={μ=(μi)i∈I∈P:μ satisfies (3.10)}. |
The following proposition is an immediate consequence of Theorem 6.
Corollary 8. Assume (H), (C) and (M). Let λ>0 and (z,k)∈Tn×I. Let vλ∈C(Tn×I) be the unique solution of (Pλ). Then there exists a μz,k,λ=(μz,k,λi)i∈I∈P(z,k,λ) such that
vλk(z)=∑i∈I⟨μz,k,λi,(λ+ρi)−1Li⟩=minμ=(μi)i∈I∈P(z,k,λ) ∑i∈I⟨μi,(λ+ρi)−1Li⟩. | (3.11) |
We study the asymptotic behavior of the solution vλ of (Pλ), with λ>0, as λ→0.
We make a convenient assumption on the system (P0):
problem (P0) has a solution v0∈Lip(Tn). |
If ρi>0 for all i∈I, then Theorem 1 assures that there exists a unique solution v0 of (E). In this situation, it is not difficult to show that the uniform convergence, as λ→0+, of vλ to the unique solution v0 on Tn. In general, existence and uniqueness of a solution of (P0) may fail. In fact, one can prove at least in the case when the bij are constants (see Theorem 18) that there exists c∈Rm such that
Bu+H[u]=c in Tn | (4.1) |
has a solution v0∈Lip(Tn) and possibly multiple solutions. If such a c=(ci) exists, then the introduction of a new family of Hamiltonians,
˜H=(˜Hi)i∈I, with ˜Hi(x,p)=Hi(x,p)−ci, |
allows us to view (4.1) as in the form of (P0). The link between two vanishing discount problems for the original (Pλ) and for (Pλ), with ˜H in place of H, is discussed in Sections 5 and 6.
Theorem 9. Assume (H), (C), (M) and (E). Let vλ be the unique solution of (Pλ) for λ>0. Then there exists a solution v0∈Lip(Tn)m of (P0) such that the functions vλi converge to v0i uniformly on Tn as λ→0 for all i∈I.
Lemma 10. Under the hypotheses of Theorem 9, there exists a constant C0>0 such that for any λ>0,
|vλi(x)|≤C0 for(x,i)∈Tn×I. | (4.2) |
Proof. Let v0=(v0,i)i∈I∈Lip(Tn)m be the solution of (P0). Choose a constant C0>0 so that
|v0,i(x)|≤C1 for (x,i)∈Tn×I, |
and observe by the monotonicity of B (Lemma 4) that v0+C11 and v0−C11 are a supersolution and a subsolution of (P0), respectively. Noting that v0+C11≥0 and v0−C11≤0, we deduce that v0+C11≥0 and v0−C11≤0 are a supersolution and a subsolution of (Pλ) for any λ>0, respectively. By comaprison (Theorem 1), we see that, for any λ>0, v0−C11≤vλ≤v0+C11 on Tn and, moreover, −2C11≤vλ≤2C11 on Tn. Thus, (4.2) holds with C0=2C1.
Lemma 11. Under the hypotheses of Theorem 9, the family (vλ)λ∈(0,1) is equi-Lipschitz continuous on Tn.
Indeed, the family (vλ)λ>0 is equi-Lipschitz continuous on Tn, which we do not need here.
Proof. By Lemma 10, there is a constant C0>0 such that
|(Bλvλ(x))i|≤C0 for (x,i,λ)∈Tn×I×(0,1). |
Hence, as vλ is a solution of (Pλ), we deduce by (C) that there exists a constant C1>0 such that the vλi are subsolutions of |Du|≤C1 in Tn. It is a standard fact that the vλi are Lipschitz continuous on Tn with C1 as their Lipschitz bound.
In the proof of Theorem 9, Corollary 8 has a crucial role. We need also results for λ=0 similar to the corollary.
We consider the condition for μ∈P,
⟨μ,ϕ⟩≥0 for (ϕ,u)∈F(0). | (4.3) |
We denote by P(0) the subset of P consisting of those μ which satisfy (4.3).
Theorem 12. Assume (H), (C), (M) and (E). Assume that ρi=0 on Tn for every i∈I. Then there exists a μ0=(μ0i)i∈I∈P(0) such that
0=⟨μ0,L⟩=minμ∈P(0)⟨μ,L⟩. | (4.4) |
Proof. We fix a (z,k)∈Tn×I. By Corollary 8, for each λ>0 there exists μλ=(μλi)i∈I∈P(z,k,λ) such that
λvλk(z)=∑i∈Iλ⟨μλi,λ−1Li⟩=⟨μλ,L⟩. | (4.5) |
Since (μλ)λ>0 is a family of Borel probability measures on a compact space Tn×Ξ×I, there exists a sequence (λj)j∈N⊂(0,1) converging to zero such that the sequence (μλj)j∈N converges weakly in the sense of measures to a Borel probability measure μ0 on Tn×Ξ×I. It follows from (4.5) and Lemma 10 that
0=⟨μ0,L⟩. |
Observe that if (ϕ,u)∈F(0), then, for any λ>0, u is a subsolution of
Bλu+Hϕ[u]=λu in Tn, |
and hence, (ψ,u)∈F(λ), with ψ(x,ξ)=ϕ(x,ξ)+λu(x). Hence, the inclusion μλ∈G′(z,k,λ) yields
uk(z)≤∑i∈I⟨μλi,λ−1(ϕi+λui)⟩=λ⟨μλ,ϕ⟩+⟨μλ,u⟩. |
Multiplying the above by λ and sending λ=λj→0, in view of Lemma 10, we get
0≤⟨μ0,ϕ⟩. |
This shows that μ0∈P(0). These observations together with (4.3) for μ∈P(0) guarantee that
0=⟨μ0,L⟩=minμ∈P(0)⟨μ,L⟩. |
We state a characterization of P(0) in the next, similar to Proposition 7, which we leave to the reader to verify.
Proposition 13. Assume (H), (C) and (M). Let μ=(μi)i∈I∈P. We have μ∈P(0) if and only if
∑i∈I⟨μi,(Bψ)i−gi⋅Dψi⟩=0 for ψ=(ψi)i∈I∈C1(Tn)m. |
We call any minimizer μ∈P(0) of the optimization problem (4.4) a viscosity Mather measure.
We denote by M+ the set of all Borel nonnegative measures μ=(μi)i∈I on Tn×Ξ×I. We set
M+(0)={μ∈M+:μ satisfies (4.3)}. |
Theorem 14. Let (z,k)∈Tn×I. Assume (H), (C), (M) and (E). For any λ>0, let vλ be the unique solution of (Pλ) and μλ∈P(z,k,λ) be a minimizer of (3.11). Then there exists a subsequence of (λj), which is denoted again by the same symbol, such that, as j→∞,
λjλj+ρiμλji→μ0i |
weakly in the sense of measures for some μ0=(μ0i)i∈I∈M+(0), and μ0 satisfies
⟨μ0,L⟩=0. | (4.6) |
In particular,
0=⟨μ0,L⟩=minμ∈M+(0)⟨μ,L⟩. | (4.7) |
Notice that the minimization problem (4.7) is trivial since μ0=0 is a minimizer.
Proof. The proof is similar to that of Theorem 12.
We fix a (z,k)∈Tn×I. For each λ>0, we have
λvλk(z)=∑i∈Iλ⟨μλi,(λ+ρi)−1Li⟩. | (4.8) |
Observe that
⟨λ(λ+ρi)−1μλi,⟩≤⟨μλi,⟩=∑i∈I|μλi|=1. |
Accordingly, since Tn×Ξ×I is a compact metric space, the families (λ(λ+ρi)−1μλi)λ=λj,j∈N have a common subsequence, along which all the families converge to some Borel nonnegative measures μ0i weakly in the sense of measures. We may assume by replacing the original sequence (λj) by its subsequence that
λjλj+ρiμλji→μ0i |
weakly in the sense of measures. Combining this with (4.8) yields
0=∑i∈I⟨μ0i,Li⟩=⟨μ0,L⟩. |
It is obvious to see that μ0∈M+.
Let (ϕ,u)∈F(0). As before, we have (ψ,u)∈F(λ), with ψ(x,ξ)=ϕ(x,ξ)+λu(x) and moreover
uk(z)≤∑i∈I⟨μλi,(λ+ρi)−1(ϕi+λui)⟩=⟨μλ,(λ+ρi)−1ϕ⟩+λ⟨μλ,(λ+ρi)−1u⟩. |
Multiplying the above by λ and sending λ=λj→0, we get
0≤⟨μ0,ϕ⟩. |
This shows that μ0∈M+(0).
Proof of Theorem 9. Let V denote the set of accumulation points v=(vi)∈C(Tn)m in the space C(Tn)m of vλ as λ→0. In view of the Ascoli-Arzela theorem, Lemmas 4.2 and 4.3 guarantee that the family (vλ)λ∈(0,1) is relatively compact in C(Tn)m. In particular, the set V is nonempty. Note by the stability of the viscosity property under uniform convergence that any v∈V is a solution of (P0).
If V is a singleton, then it is obvious that the whole family (vλ)λ>0 converges to the unique element of V in C(Tn)m as λ→0.
We need only to show that V is a singleton. It is enough to show that for any v,w∈V and (z,k)∈Tn×I, the inequality wk(z)≤vk(z) holds.
Fix any v,w∈V and (z,k)∈Tn×I. Select sequences (λj) and (δj) converging to zero so that
vλj→v, vδj→w in C(Tn)m as j→∞. |
By Corollary 8, there exists a sequence (μj)j∈N such that
μj∈G′(z,k,λj) and vλjk(z)=∑i∈I⟨μji,(λj+ρi)−1Li⟩ for j∈N. | (4.9) |
In view of Theorem 14, we may assume by passing to a subsequence if necessary that, as j→∞,
λjλj+ρiμji→μ0i weakly in the sense of measures |
for all i∈I and for some μ0=(μ0i)i∈I∈M+(0) and, moreover,
0=⟨μ0,L⟩. | (4.10) |
Since (L−λvλ,vλ)∈F(0) and μ0∈M+(0), in view of (4.10), we have
0≤⟨μ0,L−λvλ⟩=⟨μ0,L⟩−⟨μ0,λvλ⟩=−λ⟨μ0,vλ⟩, |
which yields after dividing by λ>0 and then sending λ→0 along λ=δj
⟨μ0,w⟩≤0. | (4.11) |
Now, note that w is a solution of
Bλw+H[w]=λw in Tn, |
and thus, (L+λw,w)∈F(λ) and infer by (4.9) that
wk(z)≤∑i∈I⟨μji,(λj+ρi)−1(Li+λjwi)⟩=vλjk(z)+λj∑i∈I⟨μji,(λj+ρi)−1wi⟩. |
Sending j→∞ now yields
wk(z)≤vk(z)+⟨μ0,w⟩. |
This together with (4.11) shows that wk(z)≤vk(z), which completes the proof.
We consider the problem of finding c=(ci)i∈I∈Rm and v=(vi)i∈I∈C(Tn)m such that v is a solution of
Bv+H[v]=c in Tn. | (5.1) |
The pair of such c and v is also called a solution of (5.1). This problem is called the ergodic problem in this paper although the term, ergodic problem, should be used only when the condition that ∑j∈Ibij(x)=0 holds for some (i,x)∈I×Tn.
Henceforth, D(x) denotes the diagonal matrix
D(x)=diag(ρ1(x),…,ρm(x)) for x∈Tn, |
where, as before, ρi(x)=∑j∈Ibij(x).
Throughout this section, we treat the case when
B(x) is irreducible. | (5.2) |
The irreducibility of B(x) is stated as follows: for any nonempty subset I of I, which is not identical to I, there exists a pair of i∈I and j∈I∖I such that bij(x)≠0.
The following result has been established in Davini-Zavidovique [11,Theorem 2.10] (see also [6,30]).
Proposition 15. Assume (H), (C), (M), (5.2), and that
∑j∈Ibij(x)=0 for all(i,x)∈I×Tn. | (5.3) |
Then there exist c0∈R and v0∈Lip(Tn)m such that the pair (c01,v0) is a solution of (5.1).
We remark that (5.3) is satisfied if and only if B(x)1=0 for all x∈Tn, which holds if and only if ρi(x)=0 for all (i,x)∈I×Tn.
The next theorem states the central result of this section.
Theorem 16. Assume (H), (C), (M), (5.2), and (5.3). Let vλ be the unique solution of (Pλ) for λ>0. Then there exists a constant c0∈R and a function v0∈Lip(Tn)m such that the functions vλ+λ−1c01 converge to v0 uniformly on Tn as λ→0. Moreover, the pair (c01,v0) is a solution of (5.1).
Proof. Thanks to Proposition 15, there exists a solution (c0,v0)∈Rm×C(Tn)m of (5.1). We set ˜H=H−c01, and note that, since B(x)1=0 for all x∈Tn, the function wλ:=vλ+λ−1c01 satisfies, in the viscosity sense,
λwλ+Bwλ+˜H[wλ]=λvλ+c01+Bvλ+H[vλ]−c01=0. |
By Theorem 9, there exists a solution v0∈Lip(Tn)m of Bv0+˜H[v0]=0 in Tn such that, as λ→0+, wλ→v0 in C(Tn)m. Noting that (c01,v0) is a solution of (5.1), we finish the proof.
The condition (5.3) in Proposition 15 can be removed and the following theorem is valid.
Theorem 17. Assume (H), (C), (M), and (5.2). Then there exist c0∈R and v0=(v0i)i∈I∈Lip(Tn)m such that the pair (c01,v0) is a solution of (5.1).
Proof. For x∈I×Tn, we set
B0(x)=(b0ij(x)):=B(x)−D(x). |
and note that B0(x) is irreducible and (5.3) holds with bij(x) replaced by b0ij(x). Note also that ρi(x)≥0 for all (i,x)∈I×Tn.
Thanks to Proposition 15, there exist c0∈R and v=(vi)∈Lip(Tn)m which solve
B0v+H[v]=c01 in Tn. |
We choose a constant C>0 so that max(i,x)∈I×Tn|vi(x)|≤C and set v±(x)=v(x)±C1, respectively. Observe that, since v+i(x)≥0 and v−i(x)≤0 for all (i,x)∈I×Tn, the functions u=v+ and u=v− are a supersolution and subsolution of
B0u+Pu+H[u]=c01 in Tn, |
that is, Bu+H[u]=c01 in Tn, respectively. In view of the Perron method, the function v0=(v0i)i∈I∈Lip(Tn) given by
v0i(x)=sup{ui(x):u=(ui)∈C(Tn)m is a subsolution of Bu+H[u]=c01 in Tn,v−≤u≤v+ in Tn}, |
is a solution of (5.1), with c=c01.
Even without the assumption (5.3), it is immediate from Theorem 9 that, under the hypotheses of Theorem 17, if c0=0, then the convergence holds for the whole family of the solutions vλ of (Pλ), with λ>0. A typical case when c0=0 is given by [6,Theorem 4.2] (see also [11,28]).
Throughout this section we assume that B is a constant matrix, that is, independent of x∈Tn.
The main results in this section are as follows.
Theorem 18. Assume (H), (C), (M), and that B is a constant matrix. Then (5.1) has a solution (c,v)∈Rm×C(Tn)m.
Theorem 19. Under the same hypotheses of Theorem 18, let (c,v0)∈Rm×C(Tn)m be a solution of (5.1) and let vλ be the unique solution of (Pλ) for λ>0. Then there exists a function v0∈C(Tn)m such that the functions vλ+(λI+B)−1c converge to v0 uniformly on Tn as λ→0. Moreover, the pair (c,v0) is a solution of (5.1).
Proof. It is well-known (and easily checked) that due to the monotonicity of B, (λI+B) is invertible for any λ>0. We set ˜H(x,p)=H(x,p)−c for (x,p)∈Tn×Rn and also wλ(x)=vλ(x)+(λI+B)−1c for x∈Tn. Observe that, in the viscosity sense,
λwλ(x)+Bwλ(x)+˜H[wλ]=λvλ+Bvλ+H[vλ]−c+λ(λI+B)−1c+B(λI+B)−1c=0 in Tn. |
It is clear that ˜H satisfies (H) and (C) and that v0 is a solution of Bu+˜H[u]=0 in Tn. By Theorem 9, we conclude that there exists a solution v0∈C(Tn)m of Bu+˜H[u]=0 in Tn such that wλ→v0 in C(Tn)m as λ→0+. Noting that (c,v0) is a solution of (5.1), we finish the proof.
For the proof of Theorem 18, we begin with a preliminary remark on the permutations.
For a given permutation π:I→I, we define the m×m matrix P by
P=(δπ(i),j)i,j∈I, | (6.1) |
where δij=δi,j:=1 if i=j and =0 otherwise. Note that P−1=(δi,π(j))i,j∈I=PT and that for any u=(ui)i∈I,
Pu=P(u1⋮um)=(uπ(1)⋮uπ(m)). |
The system of Hamilton-Jacobi equations
λu+Bu+H[u]=0 | (6.2) |
can be written component-wise as
λuπ(i)+(Bu)π(i)+Hπ(i)[uπ(i)]=0 for i∈I. |
By the use of P, the system above is expressed as
λ(Pu)i+(PBu)i+(PH)i[(Pu)i]=0, |
and furthermore, if v=Pu,
λ(v)i+(PBPTv)i+(PH)i[vi]=0. | (6.3) |
Set A=(aij)i,j∈I=PBPT and observe that if B is monotone, then
aij=∑k,l∈Iδi,π(k)bklδπ(l),j=bπ−1(i),π−1(j){≥0 if i=j,≤0 if i≠j, |
and
∑j∈Iaij=∑j∈Ibπ−1(i),π−1(j)=∑j∈Ibπ−1(i),j≥0. |
Consequently, if B is monotone, then PBPT is monotone as well, and the system (6.2), by using the permutation matrix P, is converted to (6.3).
Proof of Theorem 18. It is well-known (see for instance [35,Section 2.3]) that, given a monotone matrix B, one can find a permutation π:I→I such that
PBPT=(B(1)0⋯0∗B(2)⋱⋮⋮⋱⋱0∗⋯∗B(rp)), | (6.4) |
where, P is given by (6.1), B(1) is a diagonal matrix of order r1 and, for 1<i≤p, B(i) are irreducible matrices of order ri. In view of the preliminary remark before this proof, to seek for a solution of (5.1), we may and do assume henceforth B has the normal form of the right hand side of (6.4).
Set
sk=∑1≤i<kri and Ik={sk+1,…,sk+rk} for k∈{1,…,p}. |
Notice that s1=0. If r1≥1, then we first show that there exist an r1-vector c(1)=(c(1)i)i∈I1∈Rr1 and a function v(1)=(v(1)i)i∈I1∈C(Tn)r1 such that v(1) is a solution of
B(1)v(1)+H(1)[v(1)]=c(1) in Tn, | (6.5) |
where H(1)=(Hi)i∈I1. The system is, in fact, a collection of single equations
biiv(1)i+H(1)i[v(1)i]=c(1)i in Tn, with i∈I1, | (6.6) |
and thus the existence of a solution (c(1),v(1)) of (6.5) is a classical result. Indeed, for each i∈I1, if b(1)ii>0, then (6.6) has a (unique) solution v(1)i∈Lip(Tn) for any choice of c(1)i. If b(1)ii=0, then (6.6) has a solution (c(1)i,v(1)i)∈R×Lip(Tn) (see unpublished work by Lions PL, Papanicolaou G, and Varadhan S: Homogenization of Hamilton-Jacobi equations). If r1=m, then we are done.
Next, assume that r1<m (and equivalently, 1<p) and we show that there exist a vector c(2)=(c(2)i)i∈I2∈Rr2 and a function v(2)=(v(2)i)i∈I2∈C(Tn)r2 such that v(2) is a solution of the system
B(2)v(2)+H(2)[v(2)]=c(2) in Tn, | (6.7) |
where
H(2)i(x,p)=Hi(x,p)−∑j∈I1bi,jv(1)j(x) for i∈I2. | (6.8) |
According to Proposition 15, there exist c(2)=(c(2)i)i∈I2∈Rr2 and v(2)=(v(2)i)i∈I2∈C(Tn)r2 which satisfy (6.7). This way (by induction), we find c(1),…,c(p) and v(1),…,v(p) such that
c(k)∈Rrk and v(k)∈C(Tn)rk for k∈{1,…,p}, |
and v(k) satisfies
B(k)v(k)+H(k)[v(p)]=c(k) in Tn, for k∈{1,…,p}. | (6.9) |
where
H(k)i(x,p)=Hi(x,p)−∑1≤j<k∑q∈Ijbi,qv(j)q(x) for i∈Ik. | (6.10) |
We define c=(ci)i∈I∈Rm and v=(vi)i∈I∈C(Tn)m by setting
ci=c(k)i and vi=v(k)i for i∈Ik,k∈{1,…,p}, |
and observe that
Bv+H[v]=c in Tn. |
This completes the proof.
The author would like to thank the anonymous referee for useful and critical comments on the original version of this paper, which have helped significantly to improve the presentation. This work is partially supported by the JSPS KAKENHI #16H03948, #18H00833, #20K03688, and #20H01817.
The author declares no conflicts of interest in this paper.
[1] | Al-Aidarous ES, Alzahrani EO, Ishii H, et al. (2016) A convergence result for the ergodic problem for Hamilton-Jacobi equations with Neumann-type boundary conditions. P Roy Soc Edinb A 146: 225-242. |
[2] | Bardi M, Capuzzo-Dolcetta I (1997) Optimal Control and Viscosity Solutions of Hamilton-Jacobi-Bellman Equations, Boston: Birkhäuser Boston, Inc. |
[3] | Barles G (1993) Discontinuous viscosity solutions of first-order Hamilton-Jacobi equations: A guided visit. Nonlinear Anal 20: 1123-1134. |
[4] | Barles G (1994) Solutions de viscosité des équations de Hamilton-Jacobi, Paris: Springer-Verlag. |
[5] | Cagnetti F, Gomes D, Tran VH (2013) Adjoint methods for obstacle problems and weakly coupled systems of PDE. ESAIM Contr Optim Ca 19: 754-779. |
[6] | Camilli F, Ley O, Loreti P, et al. (2012) Large time behavior of weakly coupled systems of first-order Hamilton-Jacobi equations. NoDEA Nonlinear Diff 19: 719-749. |
[7] | Chen Q, Cheng W, Ishii H, et al. (2019) Vanishing contact structure problem and convergence of the viscosity solutions. Commun Part Diff Eq 44: 801-836. |
[8] | Crandall MG, Ishii H, Lions PL (1992) User's guide to viscosity solutions of second order partial differential equations. B Am Math Soc 27: 1-67. |
[9] | Crandall MG, Lions PL (1983) Viscosity solutions of Hamilton-Jacobi equations. T Am Math Soc 277: 1-42. |
[10] | Davini A, Fathi A, Iturriaga R, et al. (2016) Convergence of the solutions of the discounted Hamilton-Jacobi equation. Invent Math 206: 29-55. |
[11] | Davini A, Zavidovique M (2014) Aubry sets for weakly coupled systems of Hamilton-Jacobi equations. SIAM J Math Anal 46: 3361-3389. |
[12] | Davini A, Zavidovique M (2019) Convergence of the solutions of discounted Hamilton-Jacobi systems. Adv Calc Var, Available from: https://doi.org/10.1515/acv-2018-0037. |
[13] | Engler H, Lenhart SM (1991) Viscosity solutions for weakly coupled systems of Hamilton-Jacobi equations. P Lond Math Soc 63: 212-240. |
[14] | Evans LC (2004) A survey of partial differential equations methods in weak KAM theory. Commun Pure Appl Math 57: 445-480. |
[15] | Evans LC (2010) Adjoint and compensated compactness methods for Hamilton-Jacobi PDE. Arch Ration Mech Anal 197: 1053-1088. |
[16] | Fathi A (1997) Théorème KAM faible et théorie de Mather sur les systèmes lagrangiens. C R Acad Sci Paris Sér I Math 324: 1043-1046. |
[17] | Fathi A (2008) Weak KAM Theorem in Lagrangian Dynamics, Preliminary Version 10, Available from: https://www.math.u-bordeaux.fr/~pthieull/Recherche/KamFaible/Publications/Fathi2008_01.pdf. |
[18] | Gomes DA (2005) Duality principles for fully nonlinear elliptic equations. In: Trends in Partial Differential Equations of Mathematical Physics, Basel: Birkhäuser, 125-136. |
[19] | Gomes DA, Mitake H, Tran HV (2018) The selection problem for discounted Hamilton-Jacobi equations: Some non-convex cases. J Math Soc JPN 70: 345-364. |
[20] | Ishii H, Jin L (2020) The vanishing discount problem for monotone systems of Hamilton-Jacobi equations. part 2 - Nonlinear coupling. Calc Var 59: 140. |
[21] | Ishii H (1987) Perron's method for Hamilton-Jacobi equations. Duke Math J 55: 369-384. |
[22] | Ishii H, Koike S (1991) Viscosity solutions for monotone systems of second-order elliptic PDEs. Commun Part Diff Eq 16: 1095-1128. |
[23] | Ishii H, Mitake H, Tran HV (2017) The vanishing discount problem and viscosity Mather measures. Part 1: The problem on a torus. J Math Pure Appl 108: 125-149. |
[24] | Ishii H, Mitake H, Tran HV (2017) The vanishing discount problem and viscosity Mather measures. Part 2: Boundary value problems. J Math Pure Appl 108: 261-305. |
[25] | Ishii H, Siconolfi A (2020) The vanishing discount problem for Hamilton-Jacobi equations in the Euclidean space. Commun Part Diff Eq 45: 525-560. |
[26] | Mitake H, Siconolfi A, Tran HV, et al. (2016) A Lagrangian approach to weakly coupled Hamilton-Jacobi systems. SIAM J Math Anal 48: 821-846. |
[27] | Mitake H, Tran HV (2017) Selection problems for a discount degenerate viscous Hamilton-Jacobi equation. Adv Math 306: 684-703. |
[28] | Mitake H, Tran HV (2012) Remarks on the large time behavior of viscosity solutions of quasi-monotone weakly coupled systems of Hamilton-Jacobi equations. Asymptot Anal 77: 43-70. |
[29] | Mitake H, Tran HV (2014) A dynamical approach to the large-time behavior of solutions to weakly coupled systems of Hamilton-Jacobi equations. J Math Pure Appl 101: 76-93. |
[30] | Mitake H, Tran HV (2014) Homogenization of weakly coupled systems of Hamilton-Jacobi equations with fast switching rates. Arch Ration Mech Anal 211: 733-769 |
[31] | Lions PL (1982) Generalized Solutions of Hamilton-Jacobi Equations, Boston-London: Pitman. |
[32] | Sion M (1958) On general minimax theorems. Pacific J Math 8: 171-176. |
[33] | Terai K (2019) Uniqueness structure of weakly coupled systems of ergodic problems of Hamilton-Jacobi equations. NoDEA Nonlinear Diff 26: 44. |
[34] | Terkelsen F (1972) Some minimax theorems. Math Scand 31: 405-413. |
[35] | Varga RS (2000) Matrix Iterative Analysis, Berlin: Springer-Verlag. |
1. | Son N.T. Tu, Vanishing discount problem and the additive eigenvalues on changing domains, 2022, 317, 00220396, 32, 10.1016/j.jde.2022.01.055 | |
2. | Maxime Zavidovique, Convergence of solutions for some degenerate discounted Hamilton–Jacobi equations, 2022, 15, 1948-206X, 1287, 10.2140/apde.2022.15.1287 | |
3. | Son N.T. Tu, Jianlu Zhang, Generalized convergence of solutions for nonlinear Hamilton–Jacobi equations with state-constraint, 2024, 406, 00220396, 87, 10.1016/j.jde.2024.06.010 | |
4. | Qinbo Chen, Albert Fathi, Maxime Zavidovique, Jianlu Zhang, Convergence of the solutions of the nonlinear discounted Hamilton–Jacobi equation: The central role of Mather measures, 2024, 181, 00217824, 22, 10.1016/j.matpur.2023.10.006 | |
5. | Italo Capuzzo Dolcetta, Andrea Davini, On the vanishing discount approximation for compactly supported perturbations of periodic Hamiltonians: the 1d case, 2023, 48, 0360-5302, 576, 10.1080/03605302.2023.2183409 |