Existence, uniqueness, and monotonicity
State a controlled additive-noise variational theorem and use coercivity and one-sided monotonicity to explain its energy and uniqueness estimates.
On this page
The question for today
What you’ll understand
- Distinguish hemicontinuity, monotonicity, coercivity and growth assumptions.
- Apply a standard additive-noise variational existence theorem.
- Derive a pathwise contraction estimate from the same-noise difference equation.
Infinite-dimensional existence proofs need uniform estimates that survive removal of a finite-dimensional cutoff. Variational monotonicity methods obtain these estimates directly from the drift's energy pairing, even when no globally Lipschitz bound in the state norm is available. Coercivity controls the solution's energy-space norm, while one-sided monotonicity controls differences of two solutions. We state a deliberately restricted additive-noise theorem so that every hypothesis can be visible, then check those hypotheses and a contraction rate in a linear Dirichlet example.
A concrete theorem has several separate hypotheses
Fix a finite horizon and a Gelfand triple with separable reflexive V continuously and densely embedded in a real separable Hilbert space H. Let U be separable, W cylindrical on U on a stochastic basis with the usual conditions, and B a fixed Hilbert–Schmidt operator from U to H. Let A be a Borel map from V to V*. Assume hemicontinuity: pairing A(u+rv) with any fixed w in V is continuous in the real parameter r.
For some p≥2 and constants c>0, C≥0, assume one-sided monotonicity , coercivity , and growth . These are sufficient conditions for this autonomous additive theorem. Multiplicative or locally monotone extensions need additional hypotheses and are not included merely by changing the noise notation.
Energy bounds survive approximation
For an H-valued, time-zero-measurable initial condition with finite second moment, the standard monotone variational theorem gives a unique adapted solution with continuous H-valued paths and finite expected time integral of its V-norm to power p. Its drift integral is interpreted in V*. Finite-dimensional Galerkin equations provide approximations, but solving those equations is only the beginning. Uniform Itô energy estimates prevent their norms from escaping as the cutoff increases.
Hemicontinuity and monotonicity then identify the weak limit of the nonlinear drift; weak convergence alone does not permit substituting a nonlinear function into a limit. This identification is a central step of the proof, not a consequence of a finite matrix being invertible. The growth condition keeps the drift bounded in the appropriate dual time-integrability space. Coercivity supplies space-time control, while stochastic martingale estimates and the initial second moment control the H-norm. Each assumption serves a different passage-to-the-limit requirement.
Uniqueness comes from comparing the same noise
Two solutions driven by the same additive noise have a difference equation with no stochastic increment. Applying the energy identity to their difference and using the one-sided monotonicity inequality yields a Gronwall bound. Equal initial states therefore give indistinguishable solutions in this framework. For multiplicative noise, a difference of diffusion coefficients remains and must appear inside the monotonicity estimate; an additive-noise cancellation cannot be reused unchanged.
A dissipative Dirichlet Laplacian illustrates a stronger conclusion. With drift nu times the Laplacian minus a nonnegative linear reaction rate, the dual pairing is negative gradient energy minus reaction energy. Poincare's inequality converts the gradient term into a state-norm contraction. This gives a quantitative same-noise stability estimate, which can later support uniqueness of invariant laws. It does not imply that every nonlinear SPDE is globally well posed: explosive growth, insufficient coercivity or singular products may fall outside the theorem entirely.
Work it through
A coercive Dirichlet drift
On (0,pi), take H=L², V=H0¹ with norm equal to the gradient norm, drift A(u)=nu Delta u−a u, nu>0 and a≥0, and fixed Hilbert–Schmidt B. Check its energy pairing and obtain a difference contraction when nu=a=1.
Integration by parts in the dual pairing gives 2<A(u),u>=−2nu||gradient u||²−2a||u||². Adding ||B||HS² gives the coercivity bound with p=2 and a constant noise-energy allowance.
The drift is bounded linear from V to V*, so hemicontinuity and linear growth hold. Its difference pairing has the same negative form, proving one-sided monotonicity. The noise is admissible because B is Hilbert–Schmidt into H.
For two solutions using the same noise and nu=a=1, Poincare on (0,pi) gives ||gradient D||²≥||D||². Hence the derivative of ||D||² is at most −4||D||², yielding ||D_t||²≤exp(−4t)||D_0||².
The squared difference contracts at rate four, while the difference norm contracts at rate two. Neither rate should be confused with the noise-driven state's energy.
The living notebook
Explore the model
Interactive lab · heat
Inspect the dissipative part of a difference estimate
Pause and Reset, then set both controls to mode n=1 and diffusivity κ=1. After two Steps to time 1, predict amplitude exp(−1) and squared amplitude exp(−2). Reset and set n=2 and κ=1; after two Steps predict amplitude exp(−4) and its square exp(−8). This scene covers the Laplacian part only. The worked reaction term adds another exp(−a t) factor, and no stochastic forcing or nonlinear existence proof is displayed.
- Current amplitude
- 1
- Spatial L² norm
- 1.772
- Decay rate · κn²
- 3.6 /s
The exact periodic heat-equation mode is u(x,t) = exp(−κn²t) sin(nx), on [0,2π]. Playback advances physical time. The initial sine is shown separately, and the unnormalized L² norm is √π exp(−κn²t). This analytic instrument does not run a finite-difference solver or test discretization stability. Parameter changes pause playback and recompute the model at the current elapsed time. Reset restores both parameters and returns time to zero. Playback starts paused. Adjust either parameter, then choose Play or Step.
Look forA norm and its square have different exponential rates; an energy estimate must preserve that factor of two.
Pause & reason
Check your understanding
Your turn
Try it yourself
For the same Dirichlet drift with nu=2 and a=0.5, use the first eigenvalue one to bound the difference norm and its square. Explain which noise assumption still matters even though the difference has no stochastic term.
Take a moment to make a prediction or work it out on paper. Open a hint when you need one.
Give me a hint
Combine the gradient and reaction rates before converting from squared norm to norm.
Compare with the solution
The squared norm derivative is at most −2(nu+a)||D||²=−5||D||². Thus ||D_t||²≤exp(−5t)||D_0||² and ||D_t||≤exp(−2.5t)||D_0||. Both solutions must first exist in the variational energy class; the fixed B still needs the stated Hilbert–Schmidt regularity and the initial conditions need their integrability. Cancellation in a comparison cannot retroactively construct an inadmissible noise integral.
Keep the idea
Take it with you
- Hemicontinuity, monotonicity, coercivity and growth play distinct roles.
- An additive variational theorem requires compatible drift, noise and initial data.
- Same-noise difference estimates yield uniqueness and quantitative stability.
Read further
Original sources and deeper explanations to follow at your own pace.
Optional progress, saved in this browser when available. Revisit any lesson at any time.
Lesson not marked complete.