When q=1, we obtain the following estimate by testing \log \rho to (?):
\frac{d}{dt} \int_{\Omega} \rho \log \rho \,dx + \frac{4}{m}\int_{\Omega} \left| \nabla \rho^{\frac{m}{2}} \right|^2\,dx = \underbrace{\int_{\Omega} V \cdot \nabla \rho \,dx}_{ =: \mathcal{J}_1},
because of the followings:
\begin{gathered}
\int_{\Omega} \rho_t \log \rho \,dx = \frac{d}{dt} \int_{\Omega} (\rho \log\rho - \rho) \,dx = \frac{d}{dt} \int_{\Omega} \rho \log\rho \,dx, \\
\int_{\Omega} \nabla \cdot (\nabla \rho^m - V\rho) \log \rho \,dx
= - \int_{\Omega} \left(\nabla \rho^m \nabla \log \rho - V \rho \nabla \log \rho \right) \,dx,
\end{gathered}
because of the mass conservation property and the boundary condition (?).
Now let q > 1. By testing \varphi = q \rho^{q-1} to (?), we have
\frac{d}{dt} \int_{\Omega} \rho^{q}\,dx
+ \frac{4m q (q-1)}{(q+m-1)^2} \int_{\Omega} \left| \nabla\rho^{\frac{q+m-1}{2}} \right|^2\,dx
= \underbrace{(q-1) \int_{\Omega} V \cdot \nabla \rho^{q}\,dx}_{=: \mathcal{J}_q},
because of the followings:
\begin{gathered}
\int_{\Omega} q \rho^{q-1} \rho_t\,dx = \frac{d}{dt} \int_{\Omega} \rho^q \,dx, \\
q\int_{\Omega} \nabla \cdot (\nabla \rho^m - V\rho) \rho^{q-1} \,dx
= - q\int_{\Omega} \left(\nabla \rho^m \nabla \rho^{q-1} - V \rho \nabla \rho^{q-1} \right) \,dx,
\end{gathered}
because of the boundary conditions (?).
(i) With the estimate (Proof), we observe that, for 1 - \frac m2 \geq 0 (equivalent to 1 < m \leq 2),
\mathcal{J}_{1} = \frac{2}{m} \int_{\Omega} V \rho^{1-\frac{m}{2}} \nabla \rho^{\frac m 2} \,dx \leq c \|V\|_{L^{q_1}_{x}}
\left\|\nabla \rho^{\frac{m}{2}}\right\|_{L^{2}_{x}}
\|\rho\|_{L_{x}^{\frac{q_1 m }{q_1 -2}}}^{\frac{m}{2}},
by taking Hölder inequality with \frac{1}{q_1} + \frac{1}{2} + \frac{q_1 -2 }{2q_1} = 1 for q_1 > 2. When m=2, we follow (Proof). Moreover, we denote same settings in (Proof) and (Proof) for q=1. Then the same computation in (Proof) with \varepsilon = \frac{2}{m} provides the following estimate:
\frac{d}{dt} \int_{\Omega} \rho \log \rho \,dx + \frac{2}{m}\int_{\Omega} \left| \nabla \rho^{\frac{m}{2}} \right|^2\,dx
\leq c \|V\|_{L^{q_1}_{x}}^{q_2} \left(\|\rho\|_{L_{x}^{1}} + 1\right) + c
for q_2 in (Proof) for q=1. The estimate (4.2) is obtained by taking the integration in terms of t\in [0, T] and applying that the negative part of \int_{\Omega} \rho \log \rho \,dx < \infty because \Omega is bounded and
\|\rho(\cdot, t)\|_{L^{1}_{x}} = 1 because of the mass conservation property.
(ii) In case q>1 and q \geq m-1, we observe that
\mathcal{J}_{q} = \frac{2(q-1)}{q+m-1}\int_{\Omega} V \rho^{\frac{q-m+1}{2}} \nabla \rho^{\frac{q+m-1}{2}}\,dx.
When q=m-1 (in particular, m=2 if q=1), we have
\mathcal{J}_q \leq c \int_{\mathbb{R}^d} |V| \left| \nabla \rho^{\frac{q+m-1}{2}} \right|\,dx
\leq c \|V\|_{L^{2}_{x}}
\left\|\nabla \rho^{\frac{m+q-1}{2}}\right\|_{L^{2}_{x}}
\leq \varepsilon \left\|\nabla \rho^{\frac{m+q-1}{2}}\right\|_{L^{2}_{x}}^{2}
+ c(\varepsilon ) \|V\|_{L^{2}_{x}}^{2}.
Now, let q > m-1 for all q\geq 1. Then by Hölder inequality with \frac{1}{q_1} + \frac{1}{2} + \frac{q_1 -2 }{2q_1} = 1 for q_1 > 2, we have
\mathcal{J}_q \leq c \int_{\Omega} |V| \rho^{\frac{q-m+1}{2}} \left| \nabla \rho^{\frac{q+m-1}{2}} \right|\,dx
\leq c \|V\|_{L^{q_1}_{x}}
\left\|\nabla \rho^{\frac{m+q-1}{2}}\right\|_{L^{2}_{x}}
\|\rho\|_{L_{x}^{\frac{q_1 (q-m+1)}{q_1 -2}}}^{\frac{q-m+1}{2}}.
For simplicity, let
r_1 := \frac{q_1 (q-m+1)}{q_1 -2}, \quad \text{and} \quad \theta := \frac{q}{r_1} \left( 1 - \frac{(r_1 - q)(d-2)}{2q + d(m-1)}\right).
Then by (3.2) in Lemma 3.2, we see that
\begin{cases}
2<\frac{d(q+m-1)}{q+(d-1)(m-1)}\leq q_1 \leq \frac{2q}{m-1}, & \text{ if } d>2, \vspace{1 mm}\\
2< q_1 \leq \frac{2q}{m-1}, & \text{ if } d = 2.
\end{cases}
Then by modifying interpolation inequalities in Lemma 3.2 for \Omega \Subset { \mathbb{R} }^d, we deduce
\begin{aligned}
\mathcal{J}_q
&\leq c \|V\|_{L^{q_1}_{x}}
\left\|\nabla \rho^{\frac{m+q-1}{2}}\right\|_{L^{2}_{x}}
\left(\left\|\nabla \rho^{\frac{m+q-1}{2}}\right\|_{L^{2}_{x}}^{\frac{(1-\theta)(q-m+1)}{q+m-1}}
\|\rho\|_{L_{x}^{q}}^{\frac{\theta(q-m+1)}{2}} + \|\rho\|_{L_{x}^{1}}^{\frac{q+m-1}{2}}\right) \\
&\leq \varepsilon \left\|\nabla \rho^{\frac{m+q-1}{2}}\right\|_{L^{2}_{x}}^{2}
+c(\varepsilon) \left( \|V\|_{L^{q_1}_{x}}^{q_2}\|\rho\|_{L_{x}^{q}}^{q_2 \frac{\theta(q-m+1)}{2}} + c\|V\|_{L^{q_1}_{x}}^{2}\right) \\
&\leq \varepsilon \left\|\nabla \rho^{\frac{m+q-1}{2}}\right\|_{L^{2}_{x}}^{2}
+c(\varepsilon) \|V\|_{L^{q_1}_{x}}^{q_2} \left(\|\rho\|_{L_{x}^{q}}^{q_2 \frac{\theta(q-m+1)}{2}} + 1\right) + c(\varepsilon) \\
&\leq \varepsilon \left\|\nabla \rho^{\frac{m+q-1}{2}}\right\|_{L^{2}_{x}}^{2}
+c(\varepsilon) \|V\|_{L^{q_1}_{x}}^{q_2} \left(\|\rho\|_{L_{x}^{q}}^{q} + 1\right) + c(\varepsilon)
\end{aligned}
by taking Young’s inequality (note that q_2 \geq 2). Due to V\in \mathcal{S}_{m,q}^{(q_1,q_2)}, it gives
\frac{2(q+m-1)}{(q+m-1)-(1-\theta)(q-m+1)} = \frac{q_1 (2+q_{m,d})}{q_1 (1+q_{m,d}) - d} = q_2.
The conditions (Proof) corrresponds to (4.2).
For q>1, we combine (Proof) and (Proof) with \varepsilon = \frac{4mq(q-1)}{(q+m-1)^2} that yields
\frac{d}{dt} \int_{\Omega} \rho^{q}\,dx \leq c \|V\|_{L^{q_1}_{x}}^{q_2} (\|\rho\|_{L_{x}^{q}}^{q} + 1) + c \quad \Longrightarrow \quad
\sup_{0\leq t \leq T} \int_{\Omega \times \{t\}} \rho^{q}\,dx
\leq C \int_{\Omega} \rho_{0}^{q}\,dx + C,
where C= C(\|V\|_{\mathcal{S}_{m,q}^{(q_1, q_2)}}) by Grönwall’s inequality.
When q=1, it is obvious because of the mass conservation property.
Now we take the integration in terms of t after combining estimates (Proof) and (Proof) which yields
\begin{aligned}
\sup_{0\leq t\leq T} \int_{\Omega \times \{t\}} \rho^{q}\,dx
&+ \frac{2m q (q-1)}{(q+m-1)^2} \iint_{\Omega_T} \left| \nabla\rho^{\frac{q+m-1}{2}} \right|^2\,dx\,dt \\
&\leq \int_{\Omega} \rho_{0}^{q}\,dx
+ c \|V\|_{\mathcal{S}_{m,q}^{(q_1, q_2)}}\sup_{0\leq t \leq T}\|\rho (\cdot, t)\|_{L_{x}^{q}(\Omega)}^{q_2 \frac{\theta(q-m+1)}{2}}.
\end{aligned}
Therefore, we obtain (4.2) by substituting (Proof) into (Proof).
(iii) In this case, we take the integration by parts that gives, due to V\cdot \textbf{n} = 0 on \partial \Omega,
\mathcal{J}_{1} = - \int_{\Omega} \rho \, \nabla \cdot V \,dx,
\quad \text{and} \quad
\mathcal{J}_{q} = -(q-1) \int_{\Omega} \rho^q \, \nabla \cdot V \,dx.
By taking Hölder inequality with \frac{1}{\tilde{q}_1} + \frac{\tilde{q}_1 -1}{\tilde{q}_1} = 1 for \tilde{q}_1 > 1, we have
\mathcal{J}_1 , \, \mathcal{J}_{q} \leq c \|\nabla V\|_{L_{x}^{\tilde{q}_1}} \|\rho\|_{L^{\frac{q \tilde{q}_1}{\tilde{q}_1 -1}}_{x}}^{q}.
For simplicity, let
\tilde{r}_1 := \frac{q\tilde{q}_1}{\tilde{q}_1 - 1} \quad \text{ and } \quad \theta := \frac{q}{\tilde{r}_1} \left( 1 - \frac{(\tilde{r}_1 - q)(d-2)}{2q+d(m-1)} \right).
Then by (3.2) in Lemma 3.2 for \tilde{r}_1 instead of r_1, it gives
\begin{cases}
1<\frac{d+q_{m,d}}{2+q_{m,d}}\leq \tilde{q}_1 \leq \infty, & \text{ if } d>2, \vspace{1 mm}\\
1< \tilde{q}_1 \leq \infty, & \text{ if } d= 2.
\end{cases}
Then by Lemma 3.2, we deduce
\begin{aligned}
\mathcal{J}_1 , \, \mathcal{J}_{q}
&\leq c \|\nabla V\|_{L_{x}^{\tilde{q}_1}} \left(\|\rho\|_{L^{q}_{x}}^{q\theta}
\left\|\nabla \rho^{\frac{q+m-1}{2}}\right\|_{L^{2}_{x}}^{\frac{2q(1-\theta)}{q+m-1}} + \|\rho\|^{q}_{L^1_x} \right)\\
&\leq \varepsilon \left\|\nabla \rho^{\frac{q+m-1}{2}}\right\|_{L^{2}_{x}}^{2}
+ c(\varepsilon) \|\nabla V\|_{L_{x}^{\tilde{q}_1}}^{\tilde{q}_2} \left( \|\rho\|_{L^{q}_{x}}^{q\theta \tilde{q}_2} + 1 \right) \\
&\leq \varepsilon \left\|\nabla \rho^{\frac{q+m-1}{2}}\right\|_{L^{2}_{x}}^{2}
+ c(\varepsilon) \|\nabla V\|_{L_{x}^{\tilde{q}_1}}^{\tilde{q}_2} \left( \|\rho\|_{L^{q}_{x}}^{q} + 1 \right),
\end{aligned}
by taking Young’s inequality. Because of V \in \tilde{\mathcal{S}}_{m,q}^{(\tilde{q}_1, \tilde{q}_2)}, it gives
\frac{q+m-1}{(q+m-1)-(1-\theta)q} = \frac{\tilde{q}_1 (2+q_{m,d})}{\tilde{q}_1 (2+q_{m,d}) - d} = \tilde{q}_2.
Also, we observe that (Proof) corresponds to (4.2).
Then we combine (Proof) and (Proof) with \varepsilon = \frac{4mq(q-1)}{(q-m+1)^2} which yields
\sup_{0 \leq t \leq T} \int_{\Omega} \rho^q (\cdot, t)\,dx
\leq c \|\nabla V\|_{L_{x}^{\tilde{q}_1}}^{\tilde{q}_2} \left( \|\rho\|_{L^{q}_{x}}^{q} + 1 \right)
\quad \Longrightarrow \quad
\sup_{0\leq t \leq T} \int_{\Omega \times \{t\}} \rho^{q}\,dx
\leq C \int_{\Omega} \rho_{0}^{q}\,dx + C
where C=C(\|\nabla \cdot V\|_{\tilde{\mathcal{S}}_{m,q}^{(\tilde{q}_1, \tilde{q}_2)}}) by applying Grönwall’s inequality. Now, the estimate (4.2) follows by applying (Proof) to the estimate obtained by taking integration in terms of t to the combination of (Proof) and (Proof).
(iv) In (Proof), we observe that \mathcal{J} \leq 0 for all q\geq 1 when \nabla \cdot V \geq 0. Hence we obtain (4.2) with C depending only the domain and initial data.