Prompt: Adjoint \(L\log L\) estimate on the Heisenberg group, Carnot step 2, and more

We used this Markdown research procedure together with the following prompt.

• mathematical_research_goal.md
• Plain text prompt

I am attaching `mathematical_research_goal.md`. Read it completely and use it as the research procedure for this task.

Resolve the exact problem below. Seek a complete, rigorous solution, not merely an approach or a literature-status answer. Preserve every hypothesis and quantifier.

EXECUTION_MODE: AUTO PROJECT_WINDOW_BUDGET: 4 hours

Use that window only through genuine long-running execution with observable timing. Otherwise do substantive work in the available execution block and save a resumable checkpoint at its actual boundary. Do not invent elapsed time or background continuation.

<TARGET_PROBLEM_BEGIN>
```latex
\paragraph{Problem (subelliptic adjoint $L\log L$ estimate).}
Let $\Omega\subset\mathbb R^n$ be open and let
$X_1,\ldots,X_m$ be fixed smooth real vector fields satisfying
\[
\operatorname{span}\{Y(x):Y\in\operatorname{Lie}(X_1,\ldots,X_m)\}
=\mathbb R^n
\qquad\text{for every }x\in\Omega.
\]
Fix $T>0$ and $0<\lambda\le\Lambda<\infty$.
For a symmetric matrix field $A=(a_{ij})$, define
\[
\mathcal L_A\varphi
:=\frac12\sum_{i,j=1}^m
a_{ij}(t,x)(X_iX_j+X_jX_i)\varphi.
\]
No additional first-order term is included.

Let $(\nu_t)_{0<t<T}$ be a weakly measurable family of
nonnegative finite Borel measures on $\Omega$, with
$\nu_t(\Omega)\le1$ for almost every $t$, and set
$d\nu(t,x)=dt\,d\nu_t(x)$.
Here weak measurability means that
$t\mapsto\int_\Omega\psi\,d\nu_t$ is measurable for every
$\psi\in C_c(\Omega)$.
Assume that $A$ is Borel measurable and
\[
\lambda I_m\le A(t,x)\le\Lambda I_m
\qquad\text{for }\nu\text{-almost every }(t,x).
\]

Let $\varepsilon\in[0,1]$ and let $\mu$ be a finite
nonnegative Borel measure on $(0,T)\times\Omega$.
Suppose that
\[
\partial_t\nu
=\mathcal L_A^*\nu+\varepsilon\Delta_x\nu-\mu
\qquad\text{in }\mathcal D'((0,T)\times\Omega),
\]
where the adjoint is defined by duality: explicitly,
\[
\int_{(0,T)\times\Omega}
\bigl(\partial_t\varphi+\mathcal L_A\varphi
      +\varepsilon\Delta_x\varphi\bigr)\,d\nu
=
\int_{(0,T)\times\Omega}\varphi\,d\mu
\]
for every $\varphi\in C_c^\infty((0,T)\times\Omega)$.

Prove or disprove that $\nu$ necessarily has a nonnegative
density $p\in L^1((0,T)\times\Omega)$ with respect to $dt\,dx$,
and that, for every compact $K\Subset\Omega$,
\[
\int_0^T\!\int_\Omega p(t,x)\,dx\,dt\le T,
\qquad
\int_0^T\!\int_K
p(t,x)\log\bigl(1+p(t,x)\bigr)\,dx\,dt\le C,
\]
where
\[
C=C(T,K,\Omega,X_1,\ldots,X_m,\lambda,\Lambda)
\]
is independent of $\varepsilon$, $A$, $\nu$, and $\mu$.
No initial density or initial entropy bound, no continuity
of $A$, and no smallness of $\Lambda/\lambda$ are assumed.
```

<TARGET_PROBLEM_END>

First preserve the exact target in `checkpoint.md` when file writing is available. Keep normalization brief, perform a concrete mathematical probe, and then develop the strongest route. Continue beyond setup while resources remain. Distinguish conditional progress from a solution, and audit a complete candidate before claiming a resolution.

Any notes or prior attempts I supply outside the target are context, not extra hypotheses unless I explicitly designate them as such.