Carleman estimates for time-discrete parabolic equations and applications to controllability

Franck Boyer,Víctor Hernández-Santamaría

DOI: https://doi.org/10.1051/cocv/2019072

2020-01-01

Abstract:In this paper, we prove a Carleman estimate for a time-discrete parabolic operator under some condition relating the large Carleman parameter to the time step of the discretization scheme. This estimate is then used to obtain relaxed observability estimates that yield, by duality, some controllability results for linear and semi-linear time-discrete parabolic equations. We also discuss the application of this Carleman estimate to the controllability of time-discrete coupled parabolic systems.

What problem does this paper attempt to address?

The problems that this paper attempts to solve mainly focus on the controllability and observability of time - discrete parabolic equations. Specifically, the authors prove a Carleman estimate for a time - discrete parabolic operator and use this estimate to obtain some relaxed observability estimates, and then obtain the controllability results of linear and semi - linear time - discrete parabolic equations through duality. In addition, they also discuss the application of this Carleman estimate in the controllability of time - discrete coupled parabolic systems.

### Main problems:
1. **Carleman estimate**: Prove a Carleman estimate for a time - discrete parabolic operator, which involves the relationship between a large Carleman parameter and the time - discretization step size.
2. **Controllability**: Use the above - mentioned Carleman estimate to derive the controllability results of linear and semi - linear time - discrete parabolic equations.
3. **Coupled systems**: Explore the application of the Carleman estimate in the controllability of time - discrete coupled parabolic systems.

### Specific problems:
- **Null controllability**: For a given initial data \(y_0\in L^2(\Omega)\), does there exist a control function \(v\in L^2(\omega\times(0, T))\) such that the corresponding solution satisfies \(y(\cdot, T) = 0\) at time \(T\)?
- **Relaxed observability**: Through the Carleman estimate, derive a relaxed observability inequality that contains an extra term, and this extra term decreases exponentially as the time step \(\Delta t\) approaches 0.
- **Controllability of nonlinear systems**: Study the controllability of semi - linear time - discrete parabolic equations, especially deal with the nonlinear terms through the linearization method and the fixed - point argument.

### Mathematical expressions:
- **Carleman estimate**:
\[
\tau^{-1}\iint_Q\bar{t}^-\left(e^{2\tau\theta\phi}\theta^{-1}\right)\left(|D_tq|^2 + |\Delta(\bar{t}^-q)|^2\right)+\tau\iint_Q\bar{t}^-\left(e^{2\tau\theta\phi}\theta\right)|\nabla(\bar{t}^-q)|^2+\tau^3\iint_Q\bar{t}^-\left(e^{2\tau\theta\phi}\theta^3\right)(\bar{t}^-q)^2
\]
\[
\leq C\left(\iint_Q(\bar{t}^-e^{2\tau\theta\phi})|L_Dq|^2+\tau^3\iint_{B\times(0, T)}\bar{t}^-\left(e^{2\tau\theta\phi}\theta^3\right)(\bar{t}^-q)^2\right)+C(\Delta t)^{-1}\left(\int_\Omega\left|(e^{\tau\theta\phi}q)_{1/2}\right|^2+\int_\Omega\left|(e^{\tau\theta\phi}q)_{M + 1/2}\right|^2+\int_\Omega\left|(e^{\tau\theta\phi}\nabla q)_{M+1/2}\right|^2\right)
\]

- **Relaxed observability inequality**:
\[
|q_{1/2}|_{L^2(\Omega)}