---
title: Contact Hamilton–Jacobi Equation Overview
url: https://www.emergentmind.com/topics/contact-hamilton-jacobi-equation
type: topic
---

# Contact Hamilton–Jacobi Equation Overview

The contact Hamilton–Jacobi equation is a first-order partial differential equation in which the Hamiltonian depends not only on the base point and momentum variables, but also on the unknown function itself. In its evolutionary form it is written as
\[
u_t(x,t)+H(x,Du(x,t),u(x,t))=0,
\]
while the stationary form is
\[
H(x,Du(x),u(x))=c
\]
or, in the critical case studied in weak KAM theory,
\[
H(x,Du(x),u(x))=0.
\]
This \(u\)-dependence is the defining contact feature: it replaces the conservative symplectic Hamilton–Jacobi framework by a contact one on \(T^*M\times \mathbb R\), introduces dissipative or nonconservative effects, and alters both the geometric and analytic structure of the theory [2101.00446]. When \(H\) is independent of \(u\), one recovers the classical autonomous Hamilton–Jacobi equation; when the \(u\)-dependence is linear, one obtains discounted or conformally symplectic models [2004.12269].

## 1. Geometric formulation and contact dynamics

A standard contact phase space is \(T^*Q\times \mathbb R\) with Darboux coordinates \((q^i,p_i,z)\) and contact form
\[
\eta = dz-p_i\,dq^i.
\]
The associated Reeb field is
\[
R=\frac{\partial}{\partial z},
\]
and the contact Hamiltonian vector field of a Hamiltonian \(H(q,p,z)\) has local form
\[
X_H = \frac{\partial H}{\partial p_i}\frac{\partial}{\partial q^i}
-\left(\frac{\partial H}{\partial q^i}+p_i\frac{\partial H}{\partial z}\right)\frac{\partial}{\partial p_i}
+\left(p_i\frac{\partial H}{\partial p_i}-H\right)\frac{\partial}{\partial z}.
\]
Its integral curves satisfy
\[
\dot q^i=\frac{\partial H}{\partial p_i},\qquad
\dot p_i=-\frac{\partial H}{\partial q^i}-p_i\frac{\partial H}{\partial z},\qquad
\dot z=p_i\frac{\partial H}{\partial p_i}-H
\]
[2103.17017]. In the notation used for manifolds \(M\), this same system is written as
\[
\dot x=\partial_p H(x,p,u),\qquad
\dot p=-\partial_x H(x,p,u)-p\,\partial_uH(x,p,u),\qquad
\dot u=\langle \partial_pH(x,p,u),p\rangle-H(x,p,u)
\]
[2004.12269].

The geometric Hamilton–Jacobi problem asks for a section \(\gamma\) whose image is compatible with the contact flow. For a Legendrian first jet \(\gamma=\mathrm J^1F\), one formulation gives
\[
d(H\circ \gamma)=0,
\]
while in the broader contact Hamiltonian formulation one obtains modified equations involving the Reeb derivative of \(H\) [2209.05922; 2103.17017]. In particular, the paper on contact Hamiltonian systems distinguishes between the contact Hamiltonian vector field \(X_H\) and the evolution vector field
\[
\mathcal E_H=X_H+HR,
\]
and these yield different Hamilton–Jacobi equations: in the Legendrian \(Q\)-section formulation, \(X_H\) leads to \(H\circ \gamma=0\), whereas \(\mathcal E_H\) leads to \(d(H\circ \gamma)=0\) [2103.17017]. This separation is specific to contact geometry and has no direct symplectic analogue.

The contact viewpoint is also compatible with generalized and implicit dynamics. One paper models implicit contact Hamiltonian systems as Legendrian submanifolds of an extended tangent contact manifold and derives an implicit contact Hamilton–Jacobi equation of the form
\[
d(E\circ \gamma)=\gamma^*(\mathcal R(E)\eta_{\mathcal Q}),
\]
with additional Morse-family constraints
\[
\left.\frac{\partial E}{\partial \lambda^a}\right|_{\operatorname{im}(\gamma)}=0
\]
[2109.14921]. This places singular Herglotz-type systems inside the same contact Hamilton–Jacobi framework.

## 2. PDE structure, viscosity solutions, and semigroup representation

For PDE purposes, a contact Hamilton–Jacobi equation is typically treated under continuity, convexity, coercivity in the momentum variable, and Lipschitz control in \(u\). A basic low-regularity setting assumes:
\[
H(x,u,p)\ \text{continuous},\qquad
H(x,u,p)\ \text{convex in }p,\qquad
H(x,u,p)\ \text{coercive in }p,
\]
together with a uniform Lipschitz condition in \(u\) [2101.00446]. Under such assumptions, the associated Lagrangian is the Legendre transform
\[
L(x,u,\dot x)=\sup_{p\in T_x^*M}\{\langle \dot x,p\rangle-H(x,u,p)\},
\]
which may take the value \(+\infty\) when only coercivity, rather than superlinearity, is assumed [2101.00446].

A central analytic object is the implicit backward Lax–Oleinik semigroup
\[
T_t^-\varphi(x)=\inf_{\substack{\gamma:[0,t]\to M\\ \gamma(t)=x}}
\left\{
\varphi(\gamma(0))+\int_0^t
L\big(\gamma(\tau),T_\tau^-\varphi(\gamma(\tau)),\dot\gamma(\tau)\big)\,d\tau
\right\}.
\]
This is the contact analogue of the classical explicit Lax–Oleinik operator, but it is implicit because the running cost depends on the evolving value \(T_\tau^-\varphi\) itself [2101.00446]. The corresponding forward semigroup \(T_t^+\) is defined dually. Under the standard assumptions, \(u(x,t)=T_t^-\varphi(x)\) is the unique continuous viscosity solution of
\[
\partial_tu+H(x,u,\partial_xu)=0,\qquad u(x,0)=\varphi
\]
[2101.00446].

The semigroup satisfies monotonicity and an exponential Lipschitz estimate:
\[
\|T_t^-\varphi-T_t^-\psi\|_\infty\le e^{\lambda t}\|\varphi-\psi\|_\infty,
\]
with the same type of bound for \(T_t^+\) [2101.00446]. Fixed points of \(T_t^-\) are precisely stationary viscosity solutions:
\[
u=T_t^-u\ \forall t>0
\quad\Longleftrightarrow\quad
H(x,u(x),\partial_xu(x))=0
\]
[2101.00446]. In the weak KAM language, backward weak KAM solutions, viscosity solutions, and fixed points of \(T_t^-\) coincide; similarly, forward weak KAM solutions correspond to fixed points of \(T_t^+\) [2101.00446; 2604.24329].

For the Cauchy problem on \(\mathbb R^n\),
\[
u_t+H(x,Du,u)=0,\qquad u(x,0)=u_0,
\]
one also obtains a solution semigroup \(S_H(t)\) with a variational formula
\[
S_H(t)[u_0](x)=\inf_{y\in\mathbb R^n} h_{y,u_0(y)}(x,t),
\]
where \(h_{x_0,u_0}\) is defined by an implicit variational principle [2603.19650]. This Euclidean representation supports semigroup methods, commutation results, and multi-time compatibility.

## 3. Weak KAM theory, critical values, and solution selection

The stationary contact Hamilton–Jacobi equation is closely linked to weak KAM theory on compact manifolds. In the generalized stationary equation
\[
H(x,Du,u)=c,
\]
one studies the set of constants \(c\) for which viscosity solutions exist. For \(u\)-independent Hamiltonians \(G(x,p)\), the critical value is characterized by
\[
c(G):=\inf\{c\in\mathbb R:\ G(x,Du)=c\ \text{admits a viscosity subsolution}\}
\]
[2604.24329]. In the contact case, especially when \(H\) is merely non-decreasing in \(u\), the solvability set
\[
\mathfrak C=\{c\in\mathbb R:\exists u\in C(M,\mathbb R)\ \text{solving }H(x,d_xu,u)=c\}
\]
is a connected interval rather than a singleton [2509.17310].

For Hamiltonians that are convex and superlinear in \(p\) and strictly increasing in \(u\), a distinguished vanishing-contact limit was established for the family
\[
H(x,Du_\varepsilon^-(x),\varepsilon u_\varepsilon^-(x))=c(H),
\]
where \(c(H)\) is the Mañé critical value of the conservative Hamiltonian \(H(x,p,0)\). The viscosity solution \(u_\varepsilon^-\) converges uniformly, as \(\varepsilon\to0_+\), to a uniquely selected critical solution \(u_0^-\) of
\[
H(x,Du_0^-(x),0)=c(H)
\]
[2004.12269]. This limit is characterized in two equivalent ways.

The first is via weighted Mather inequalities. Defining
\[
\mathcal F_-=
\left\{
u:\ u\text{ is a }c(H)\text{-subsolution of }H(x,Du,0)=c(H),\
\int_M \partial_uL(x,v(x),0)u(x)\,d\mu(x)\ge 0,\ \forall \mu\in\mathfrak M
\right\},
\]
the selected limit is
\[
u_0^-(x)=\sup_{u\in\mathcal F_-}u(x)
\]
[2004.12269]. The second uses the Peierls barrier \(h^\infty\) of the conservative Hamiltonian:
\[
u_0^-(x)=\inf_{\mu\in\mathfrak M}
\frac{\int_M h^\infty(y,x)\,\partial_uL(y,v(y),0)\,d\mu(y)}
{\int_M \partial_uL(y,v(y),0)\,d\mu(y)}.
\]
In the discounted case \(\partial_uH\equiv 1\), hence \(\partial_uL\equiv -1\), this reduces to the known formula
\[
u_0^-(x)=\inf_{\mu\in\mathfrak M}\int_M h^\infty(y,x)\,d\mu(y)
\]
[2004.12269]. This shows that the contact dependence contributes to the limit through the first \(u\)-jet \(\partial_uL(\cdot,\cdot,0)\).

When the \(u\)-dependence is only non-decreasing, classical comparison can fail. A generalized comparison principle replaces pointwise monotonicity by integral inequalities over special Mather measures. If \(u_1,u_2\) solve
\[
H(x,d_xu,u)=c
\]
and
\[
\int_{TM}u_1(x)\,d\mu(x,v)\le \int_{TM}u_2(x)\,d\mu(x,v)
\quad \forall \mu\in\mathfrak M_-(u_1),
\]
then \(u_1\le u_2\) on \(M\) [2509.17310]. Here \(\mathfrak M_-(u)\) denotes the “ordinal” Mather measures satisfying
\[
\int_{TM}\partial_uL(x,v,u(x))\,d\mu(x,v)=0.
\]
This identifies the degeneracy of \(u\)-monotonicity, rather than convexity in \(p\) alone, as the source of nonuniqueness.

## 4. Stability, long-time behavior, and perturbed equations

Lyapunov stability of stationary solutions has become a central topic for contact-type Hamilton–Jacobi equations. For the evolutionary equation
\[
u_t+H(x,Du,u)=0
\]
on a compact manifold, a stationary solution \(u_-\) is Lyapunov stable if small \(C^0\)-perturbations remain small under the backward semigroup \(T_t^-\), and asymptotically stable if they converge back to \(u_-\) as \(t\to\infty\) [2604.24329].

A recent PDE formulation treats separated Hamiltonians
\[
H(x,p,u)=G(x,p)+W(x,u),
\]
with \(G\) continuous, convex, and coercive in \(p\), and \(\partial_uW\) continuous and bounded [2604.24329]. The key object is the shifted critical value
\[
c(\varepsilon)=c\big(H(x,p,u_-(x)+\varepsilon)\big).
\]
Its one-sided derivatives satisfy
\[
D^-c(\varepsilon)\big|_{\varepsilon=0}
=
\inf_{\tilde\mu_-\in \widetilde{\mathfrak M}_-}
\int_{TM}\partial_uW(x,u_-(x))\,d\tilde\mu_-,
\]
\[
D^+c(\varepsilon)\big|_{\varepsilon=0}
=
\sup_{\tilde\mu_-\in \widetilde{\mathfrak M}_-}
\int_{TM}\partial_uW(x,u_-(x))\,d\tilde\mu_-.
\]
From this, PDE-type criteria are obtained. If one of the conditions
\[
c\big(H(x,p,u_-(x))-\zeta\,\partial_uW(x,u_-(x))\big)<0,
\]
or
\[
\exists\, w\in \mathrm{Lip}(M)\ \text{such that}\ 
H(x,Dw,u_-(x))-\zeta\,\partial_uW(x,u_-(x))<0\quad \text{a.e.},
\]
or
\[
D^-c(\varepsilon)\big|_{\varepsilon=0}>0
\]
holds, then \(u_-\) is locally asymptotically stable, with exponential decay rate in the sense
\[
\limsup_{t\to\infty}\frac1t\ln \|u(\cdot,t)-u_-\|_\infty\le -A
\]
for some \(A>0\) [2604.24329]. Dual sign conditions imply Lyapunov instability.

A one-dimensional non-monotone theory on the circle identifies a scalar averaged quantity
\[
\mu:= \frac{\int_0^1 \frac{\partial H}{\partial u}(\tau,u_0'(\tau),u_0(\tau))\,(B(\tau))^{-1}\,d\tau}
{\int_0^1 (B(\tau))^{-1}\,d\tau},
\qquad
B(x):=\frac{\partial H}{\partial p}(x,u_0'(x),u_0(x)),
\]
associated with the unique periodic Aubry orbit [2401.14679]. If \(\mu>0\), the stationary solution \(u_0\) is asymptotically stable on a local order interval and
\[
\limsup_{t\to+\infty}\frac1t\ln \|T_t\varphi-u_0\|_\infty\le -\mu.
\]
If \(\mu<0\), \(u_0\) is Lyapunov unstable, and the evolutionary equation has infinitely many nontrivial time-periodic viscosity solutions [2401.14679]. This replaces a global sign assumption on \(H_u\) by an orbit-average criterion.

Perturbation theory fits the same pattern. If \(u_-\) is a locally Lyapunov asymptotically stable viscosity solution of
\[
H(x,D_xu,u)=0,
\]
then for the perturbed stationary equation
\[
H(x,D_xu,u)+\varepsilon P(x,D_xu,u)=0
\]
there exist viscosity solutions \(u_-^\varepsilon\) with
\[
\|u_-^\varepsilon-u_-\|_\infty\to0
\qquad \text{as }\varepsilon\to 0
\]
[2501.04998]. Under the stronger positivity condition
\[
\partial_uH(x,p,u)>0\qquad \forall (x,p,u)\in \mathcal A_{u_-},
\]
the nearby perturbed solution is locally unique and remains locally asymptotically stable [2501.04998]. A key semigroup estimate is
\[
|T_t^\varepsilon\varphi(x)-T_t\varphi(x)|\le \varepsilon(e^{\lambda t}-1),
\]
which quantitatively controls the perturbation [2501.04998].

## 5. Commutation, discrete formulations, and stochastic extensions

Contact Hamilton–Jacobi theory also supports algebraic and geometric extensions beyond single-time deterministic flows. For two contact Hamiltonians \(H,F\) on \(\mathbb R^n\times \mathbb R^n\times \mathbb R\), the correct compatibility condition is not Poisson commutation but the vanishing of the contact or Jacobi bracket:
\[
D_x H\, D_pF- D_p H\, D_xF
+ p \cdot D_pF\, H_u - p \cdot D_pH\, F_u
+ F_u H-H_uF=0.
\]
Under the structural assumptions of convexity, coercivity in \(p\), and uniform Lipschitz dependence in \(u\), this implies semigroup commutation:
\[
S_H(\lambda)\circ S_F(\mu)[\phi]
=
S_F(\mu)\circ S_H(\lambda)[\phi]
\]
for all \(\lambda,\mu\ge0\) [2603.19650]. Pairwise Jacobi commutation of several Hamiltonians yields existence and uniqueness for multi-time systems
\[
\partial_{t_i}u+H_i(x,Du,u)=0,\qquad i=1,\dots,d
\]
[2603.19650].

A discrete Hamilton–Jacobi theory for contact dynamics has also been developed. Starting from a discrete contact Lagrangian \(L_d(q_\kappa,q_{\kappa+1},s_\kappa)\), one defines discrete Legendre transforms and right discrete contact Hamilton equations
\[
\mathbf q_{\kappa+1}=D_2H_d^+,\qquad
\mathbf p_\kappa=\frac{D_1H_d^+}{1-D_3H_d^+},\qquad
s_{\kappa+1}=s_\kappa+\mathbf p_{\kappa+1}\cdot D_2H_d^+-H_d^+.
\]
The associated discrete contact Hamilton–Jacobi equation is
\[
S_d^{\kappa+1}(\mathbf q_{\kappa+1})
-
S_d^{\kappa}(\mathbf q_{\kappa})
-
dS_d^{\kappa+1}(\mathbf q_{\kappa+1})\cdot \mathbf q_{\kappa+1}
+
H_d^+\big(\mathbf q_\kappa,dS_d^{\kappa+1}(\mathbf q_{\kappa+1}),S_d^\kappa(\mathbf q_\kappa)\big)
=0
\]
[2209.05922]. Its geometric meaning is the commutation relation
\[
\tilde\Phi\circ \gamma^\kappa=\gamma^{\kappa+1}\circ \tilde\Phi^\gamma
\]
between the discrete contact flow and the discrete section [2209.05922].

In the stochastic setting, a stochastic contact Hamilton–Jacobi equation is introduced for Stratonovich stochastic contact systems:
\[
dS+H_0\!\left(Q,\frac{\partial S}{\partial Q},S,t\right)dt
+
\sum_{k=1}^m
H_k\!\left(Q,\frac{\partial S}{\partial Q},S,t\right)\circ dW^k
=0.
\]
Together with
\[
P=\frac{\partial S}{\partial Q},
\]
this generates the stochastic contact Hamiltonian flow [2411.11115]. The same paper constructs contact structure-preserving numerical schemes from truncated expansions of the stochastic generating function and proves mean-square convergence order \(1.0\) for the resulting one-step approximation [2411.11115].

## 6. Applications and specialized variants

One prominent application is thermodynamics. On a thermodynamic contact manifold with coordinates \((s,q^i,p_i)\), contact Hamilton–Jacobi theory yields
\[
\frac{\partial W}{\partial t}+h\!\left(q^i,W,\frac{\partial W}{\partial q^i}\right)=0,
\]
where \(W\) is a thermodynamic potential on the equilibrium-state space [2203.03473]. In the energy representation, for an ideal gas, one has
\[
\eta = dU - T\,dS + P\,dV - \mu\,dN,
\]
and the contact HJ equation reproduces characteristic equations for isochoric and isothermal thermodynamic transformations [2203.03473]. A related geometric HJ formulation on \(T^*Q\times \mathbb R\) uses Legendrian sections \(\gamma(q)=(q,dF(q),F(q))\) and the condition
\[
d(H\circ \gamma)=0
\]
[2209.05922].

Another application arises in cosmology. For strong sustained rapid turn inflation in two-field models, after fixing the scalar potential \(V\) and the conformal class of the scalar field metric, the strong-SRRT consistency condition becomes a geometric contact Hamilton–Jacobi equation on the first jet bundle \(j^1(L_+)\):
\[
F_V(j^1(\omega))=0.
\]
In local isothermal Liouville coordinates \((x^1,x^2,u,p_1,p_2)\), this takes the form
\[
F(x_1,x_2,\phi,\partial_1\phi,\partial_2\phi)=0,
\]
where \(G=e^{2\phi}G_0\) is the unknown metric in the chosen conformal class [2407.19912]. The paper analyzes this as a proper first-order nonlinear PDE, studies its characteristic flow, and extracts asymptotic conditions near nondegenerate critical points of the potential [2407.19912].

A further specialized direction is the sign-changing discounted equation
\[
H(x,Du)+\lambda(x)u=c,
\]
which is a contact Hamilton–Jacobi equation with affine \(u\)-dependence. When \(\lambda\) changes sign, the set of admissible constants is
\[
[c_0,+\infty),
\]
and the stationary problem admits maximal and minimal viscosity solutions [2202.11315]. Large-time behavior of the evolutionary equation
\[
\partial_tu+H(x,Du)+\lambda(x)u=c
\]
is then organized around extremal weak KAM solutions and semigroup monotonicity [2202.11315]. This suggests that even the linear-in-\(u\) subclass already exhibits several distinctly contact phenomena: failure of global comparison, multiplicity of stationary solutions, and basin-dependent asymptotics.

Across these variants, the contact Hamilton–Jacobi equation retains a common structural role. It is simultaneously a geometric reduction principle, a viscosity-solution PDE with semigroup representation, a weak KAM selection mechanism, and a framework for nonconservative dynamics. What changes from problem to problem is the dominant structure: in Tonelli settings, Aubry–Mather theory and Peierls barriers are central; in low-regularity PDE settings, critical values and strict subsolutions dominate; in discrete, stochastic, and thermodynamic settings, the contact generating-function interpretation becomes primary [2004.12269; 2604.24329; 2209.05922].

Source: https://www.emergentmind.com/topics/contact-hamilton-jacobi-equation