Let $d\in\mathbb N$, $\tau>0$, $U\subseteq\mathbb R^d$ be open and $T_t$ be a $C^1$-diffeomorphism from $U$ onto an open subset of $\mathbb R^d$ for $t\in[0,\tau)$ with $T_0=\operatorname{id}_U$. Note that $$V:=\bigcup_{t\in[0,\:\tau)}T_t(U)$$ is open.

Assume $[0,\tau)\ni t\mapsto T_t(x)$ is differentiable for all $x\in U$. Can we show (under suitable additional assumptions, if necessary) that there is a $v:[0,\tau)\times V\to\mathbb R^d$ with $$v\left(t,T_t(x)\right)=\frac{\partial T}{\partial t}(t,x)\tag1$$ for all $t\in[0,\tau)\times V$?

If $U=\mathbb R^d$ (and hence $V=\mathbb R^d$), we may simply set $$v(t,x):=\frac{\partial T}{\partial t}\left(t,T_t^{-1}(x)\right)\tag2.$$

EDIT 1: I want to choose $v$ such that it is (jointly) continuous. By assumption, $$[0,\tau)\times U\ni(t,x)\mapsto T_t(x)\tag3$$ is partially differentiable in both the first and the second variable. So, it should be differentiable and hence (jointly) continuous.

EDIT 2: I wonder whether any differentiability properties of $v$ with respect to the second variable carry over to $v$. I've found the following excerpt in a book, which seems to indicate this, but I actually don't understand how they conclude (2.76): enter image description here enter image description here


1 Answer 1


We are looking for a function $v:[0,\tau)\times V\to\mathbb R^d$ satisfying:

$v\left(t,T_t(x)\right)=\dfrac{\partial T}{\partial t}(t,x) \qquad \forall (t,x)\in[0,\tau)\times U\tag1$

Fix $\bar{t}\in [0,\tau)$. The equation: $$v\left(\bar{t},T_\bar{t}(x)\right)=\frac{\partial T}{\partial t}(\bar{t},x) \qquad \forall x \in U $$ is a condition on the value of $v(\bar{t}, \cdot)$ on the set $T_\bar{t}(U)$. It can be restated as follows: $$v\left(\bar{t},y\right)=\frac{\partial T}{\partial t}(\bar{t},T_\bar{t}^{-1}y) \qquad \forall y\in T_\bar{t}(U)$$

Consider now the set $$A := \bigcup_{t \in [0,\tau)} \big(\{t\}\times T_t(U)\big) \subseteq [0,\tau)\times V$$

and define $f: A \to \mathbb{R}^d$ as follows: $$ f(t,y):= \dfrac{\partial T}{\partial t}(t,T_t^{-1}y) \qquad \forall (t,y)\in A $$ Then $(1)$ is equivalent to the following: $$ v(t,y)=f(t,y) \qquad \forall (t,y) \in A$$

If you don't mind continuity of $v$, you can define $v$ on $A^c:=[0,\tau)\times V - A$ arbitrarily and to be equal to $f$ on $A$.


If you need $v$ continuous (or even more regular), then the problem is whether $f$ admits a continuous (or even more regular) extension to the set $[0,\tau)\times V$.

In what follows, we will consider only the simpler case in which $T_t(U)=U $ for all $ t \in [0,\tau)$.

In this case, $A=[0,\tau)\times V = [0,\tau)\times U$ and thus $v$ and $f$ must be the same function. Therefore, $v$ is necessarily defined as follows: $$ v(t,y):= \dfrac{\partial T}{\partial t}(t,T_t^{-1}y) \qquad \forall (t,y)\in A $$

If we assume

  • $(t,x)\mapsto\dfrac{\partial T}{\partial t}(t,x)$ is jointly continuous in $t,x$.
  • $(t,y)\mapsto T_t^{-1}(y)$ is jointly continuous in $t,y$.

then $v(t,y)$ is jointly continuous in $t,y$ since it is the composition of the continuous mapping $(t,y)\mapsto (t,T_t^{-1}(y))$ with the continuous mapping $\dfrac{\partial T}{\partial t}$.

Be aware that if $(t,x)\mapsto\dfrac{\partial T}{\partial t}(t,x)$ or $(t,y)\mapsto T_t^{-1}(y)$ lack of jointly continuity then it is in general unlikely for $v$ to be jointly continuous.

  • $\begingroup$ Thank you for your answer. The reason why I thought this definition of $v$ is not suitable, is that I surely want $v$ to be jointly continuous. However, the assumptions should imply that $(3)$ is differentiable (on the product space) and hence (jointly) continuous. So, your map $v$ should be jointly continuous as well, right? $\endgroup$
    – 0xbadf00d
    Jul 28, 2020 at 15:02
  • $\begingroup$ @0xbadf00d Well, this is a different question :). I'll think about it and modify the answer accordingly. $\endgroup$ Jul 28, 2020 at 15:40
  • $\begingroup$ I think the easiest solution is to assume that $T_t(U)=U$. This is actually what the reference I've added to the question seems to do. Do you see how they conclude (2.76)? $\endgroup$
    – 0xbadf00d
    Jul 28, 2020 at 16:56
  • $\begingroup$ Thank you for your edit. Do you have any thoughts on (2.76)? $\endgroup$
    – 0xbadf00d
    Jul 31, 2020 at 4:15
  • $\begingroup$ @0xbadf00d I suspect it is a matter of function spaces definition. Can you please tell me how is $C^0(0,\epsilon; C^k(\bar{D},\mathbb{R}^N))$ is defined in the book? Moreover, nowhere it is stated that $T$ can be derived with respect to $t$, nontheless $\partia T / \partial t$ appears... $\endgroup$ Jul 31, 2020 at 8:18

You must log in to answer this question.