Let $V , W_1$, and $W_2$ be finite-dimensional vector spaces, and let $W = W_1 \times W_2$. Given an open set $U \subseteq V$ and functions $f_1 :U \rightarrow W_1$ and $f_2 :U \rightarrow W_2$, the rule $f(x)=(f_1(x),f_2(x))$ defines a mapping from $U$ into $W$. Let $a \in U$.

Prove that $f$ is differentiable at $a$ if and only if $f_1$ and $f_2$ are differentiable at $a$.

$(\Rightarrow)$ Suppose $f$ is differentiable at $a$. Let $T:V\rightarrow W$ be a linear mapping. Then there exists $T_1:V\rightarrow W_1$ such that $f_1=T_1\circ f_1$. Then $f_1=T_1\circ f_1\implies$

$f_1(a)=(T_1\circ f_1)(a)\implies$

$(Df_1)(a)=D(T_1\circ f_1)(a) = (DT_1\circ Df_1)(a) = (T_1\circ Df)(a)$. Therefore $Df_1$ exits. The same argument is made for $f_2$.

$(\Leftarrow)$ Suppose $f_1$ and $f_2$ are differentiable at $a$. Let $T_1:V\rightarrow W_1$ and $T_2:V\rightarrow W_2$ be a linear mappings. Then there exists $T:V\rightarrow W$ such that $T(v)=(T_1(v), T_2(v))$ that is linear.

Is the forward part of my proof correct? I am trying to figure out the converse, but I want to be sure the first part is correct before spending too much time on it.


Actually, the forward part of your proof doesn't make much sense.

You introduce $T$ for no reason, as you never reference it again. You claim the existence of some $T_1$ from $V$ into $W_1$, that somehow satisfies $f_1 = T_1\circ f_1$ even though $f_1$ has codomain $W_1$ and so cannot be composed with $T_1$, whose domain is $V$. From the rest, I am sure you meant $f_1 = T_1 \circ f$, but this is still not right, as the codomain of $f$ is $W$, not $V$.

In fact, it is appears to me that $T_1$ is actually supposed to be $\pi_1$, the projection from $W = W_1 \times W_2 \to W_1$. $\pi_1$ is a specific known function. It maps $(w_1, w_2) \in W$ to $w_1$. So you don't introduce it by saying something like "there exists $\pi_1$ ...". Instead you say "let $\pi_1$ be the projection ...". You are referring to a specific function, not some arbitrary one. It is true that $f_1 = \pi_1 \circ f$.

However if you already have the theorem that if $r, s$ are differentiable, the so is $r\circ s$ and the derivative is given by the chain rule: $D(r\circ s) = Dr\circ Ds$, then you can use this theorem to say that $f_1$ is differentiable and has derivative $Df_1 = D\pi_1 \circ Df$. (No further simplification is needed.)

And for the reverse implication, you are already introducing unnecessary arbitrary functions again, when you need to be referring to specific ones:

If $f_1, f_2$ are differentiable at $a$, then define a map $T\ :\ V \to W \ :\ v \mapsto (D_af_1(v), D_af_2(v))$. Then $T$ is linear.

Next, use the definition of $D_af$, and show that $T$ fits it. I.e., if the definition of the derivative that your course is using is:

$D_af$ is the unique linear map $T \ :\ V \to W$ such that $$\lim_{v\to 0}\frac{\|f(a + v) - f(a) - T(v)\|}{\|v\|} = 0$$

Then you need to show that the particular $T$ defined by $T(v) = (D_af_1(v), D_af_2(v))$ satifies this definition.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.