I am struggling to solve an extension problem of smooth functions, and I would like someone to help me. The setting is as follows:

Let $X_1$, $X_2$, and $T$ be either the real lines $\mathbb R$ or the half lines $\mathbb R_+$. Suppose we have two smooth functions $f_1:X_1\times T\to\mathbb R$ and $f_2:X_2\times T\to\mathbb R$ which agree with each other on the common submanifold $0\times T$. Furthermore, we require $f_1$ and $f_2$ to be non-negative around the origin.

Problem: Find a smooth function $f:X_1\times X_2\times T\to\mathbb R$ which restricts to $f_1$ and $f_2$ on $X_1\times0\times T$ and $0\times X_2\times T$ respectively. In addition, can one require $f$ to be non-negative around the origin ?

Note that, without non-negative requirement, we have a naive extension $$f(x_1,x_2,t) = f_1(x_1,t) + f_2(x_2,t) - f_1(0,t).\tag{1}$$ This extension, however, fails to be non-negative in general. For example consider the functions $$f_1(x_1,t) = (x_1-t)^2,\ f_2(x_2,t) = (x_2-t)^2.\tag{2}$$ Then (1) is negative on the subspace $\{x_1=x_2=t\neq 0\}$. But, of course, we can take a "right" solution $$f(x_1,x_2,t)=(x_1+x_2-t)^2$$ so (2) is not a counterexample of the problem itself.

I do not know even whether the solution exists or not. Does anyone have ideas ?

Thank you.

P.S. I am also interested in higher dimensional cases; i.e. $X_1$, $X_2$, and $T$ are products of copies of $\mathbb R$ and $\mathbb R_+$.


For the half-lines, doesn't the even more naive$$f(x_1,x_2,t):=\frac{x_1}{x_1+x_2}f_1(x_1+x_2,0,t)+\frac{x_2}{x_1+x_2}f_2(0,x_1+x_2,t)$$(i.e. interpolating linearly on segments of the form $[(a,0,t),(0,a,t)]$) do the trick?

  • $\begingroup$ Great. It is actually smooth! $\endgroup$ – Jun Yoshida Feb 2 '16 at 11:20
  • $\begingroup$ Too bad this simple device doesn't work for lines. Did you find something? $\endgroup$ – Jean Duchon Feb 4 '16 at 16:39
  • $\begingroup$ No. The problem is still completely open for $X_1=\mathbb R$. I am afraid I have to consider some additional assumptions for practical use. Your answer will probably weaken them. $\endgroup$ – Jun Yoshida Feb 9 '16 at 2:43

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.