$(1+ta_1)(1+ta_2)...(1+ta_n)=1$ $\forall t\in R$ if and only if $a_1=a_2=...=a_n=0$

Is it true? If yes, how to prove it?

I already gave a proof when $n=3$.


It is equivalent to




And this gives


Is it possible to generalize it for arbitrary $n$?

$(1+ta_1)(1+ta_2)\cdots(1+ta_n)=1$, $\forall{t}$ $\in$ $\Bbb{R}$ ⟺ $a_1=a_2=\cdots=a_n=0$

  • $\begingroup$ Try to use a choose b form for coefficient to generalize to higher n $\endgroup$ – Shrey Joshi Sep 7 '18 at 2:14
  • 1
    $\begingroup$ Instead of messing with formulas, try induction on the number of $a_i$. $\endgroup$ – Jason DeVito Sep 7 '18 at 2:21
  • $\begingroup$ Are $\,a_k\,$ assumed to be real or complex? $\endgroup$ – dxiv Sep 7 '18 at 2:24
  • $\begingroup$ Hint: the corresponding formulas are those which stated in Vieta’s Theorem, I.e. it means that $a_1,a_2,a_3,\ldots ,a_n$ are the roots of the polynomial $x^n=0$. $\endgroup$ – BAI Sep 7 '18 at 15:10
  • $\begingroup$ A nice exercise! What if the equation is true only for infinite $t$? $\endgroup$ – dmtri Sep 7 '18 at 15:15

If there exist an $\,a_k \ne 0\,$ then for $\,t = -1/a_k\,$ the LHS is $\,0\,$ i.e. different from the RHS which is $\,1\,$. Therefore the equality can hold for all real $\,t\,$ iff $\,a_k = 0 \mid k=1,2,\ldots,n\,$.

Note: the above assumes that the coefficients $\,a_k\,$ are real.

  • $\begingroup$ @stressedout Added a note spelling out the underlying assumption, and also asked the OP for clarification. If the $\,a_k\,$ turn out to be complex, I will ammend my answer. $\endgroup$ – dxiv Sep 7 '18 at 2:28
  • 1
    $\begingroup$ I do assume ak are real numbers. Thanks. That's quite a simple way. $\endgroup$ – Zhenduo Cao Sep 7 '18 at 2:31
  • $\begingroup$ What if the numbers were complex? Not sure why it's different (just curious). $\endgroup$ – IntegrateThis Sep 7 '18 at 2:33
  • $\begingroup$ @IntegrateThis Then the proof would likely have to go along the line of the other posted answer, but it gets a bit more involved at that point. The simple argument above wouldn't work since $\,t=-1/a_k\,$ would not necessarily be a real number, so the given equality would not be guaranteed to hold. $\endgroup$ – dxiv Sep 7 '18 at 2:43

I've just found a new way to prove this.

It's not hard to see that the coefficients of polynomial are all zeros. When $n=4$, that is to say,





By the last equality, at least one $a_i(i=1,2,3,4)$ is zero. Let's assume $a_1=0$.

Then by the second last equality, $a_2a_3a_4=0$. Let's assume $a_2=0$.

Then by the third last equality, $a_3a_4=0$. Let's assume $a_3=0$.

Then by the first equality, $a_4=0$.

  • $\begingroup$ (+1) That works once you fill-in the blanks. That's essentially what (I think) this comment was hinting at. And btw, it works just as well if the $\,a_k\,$ are allowed to be complex. $\endgroup$ – dxiv Sep 7 '18 at 3:01
  • 1
    $\begingroup$ @dxiv: Yep, that was the gist of my comment. The polymomial $(1+ta_1)...(1+ta_n) - 1$ is zero for infinitely many $t$ (all real $t$), so must be identically 0. The coefficient on $t^n$ is the product of the $a_i$, so at least one $a_i$ must be zero. That means your really only had $n-1$ factors, so induction can take over. (And the case case is just the fact that $ta_1 = 0$ for all $t$ implies $a_1 = 0$). $\endgroup$ – Jason DeVito Sep 7 '18 at 3:46
  • $\begingroup$ You could use induction to prove your assumption, especially when the equation is supposed to be true for $t$ not in $\Bbb{R}$ but just for infinity many values. $\endgroup$ – dmtri Sep 7 '18 at 16:05

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.