-1

If $z_n\in \mathbb{C} $ and $0<\Re(z_n)<1, \ \forall n\geq 1 $

Given, $\left|\sum_{n=1}^{\infty} \frac{1}{\left(\frac{1}{2}-z_n\right)^2}\right|= \sum_{n=1}^{\infty} \frac{1}{\left|\frac{1}{2}-z_n\right|^2}$ then find the condition on $z_n$ so that the above equality holds.

Attempt

Write, $z_n=a_n+ib_n , 0<a_n<1$ $\Rightarrow|\sum_{n=1}^{\infty} \frac{1}{(\frac{1}{2}-z_n)^2}|^2= [\sum_{n=1}^{\infty} \frac{1}{|\frac{1}{2}-z_n|^2}]^2$

$\Rightarrow |\sum_{n=1}^{\infty} \frac{1}{(\frac{1}{2}-a_n-ib_n)^2}|^2= [\sum_{n=1}^{\infty} \frac{1}{(\frac{1}{2}-a_n)^2+b_n^2}]^2$ $\Rightarrow |\sum_{n=1}^{\infty} \frac{(\frac{1}{2}-a_n+ib_n)^2}{[(\frac{1}{2}-a_n)^2+b_n^2]^2}|^2= [\sum_{n=1}^{\infty} \frac{1}{(\frac{1}{2}-a_n)^2+b_n^2}]^2$ $\Rightarrow |\sum_{n=1}^{\infty} \frac{((\frac{1}{2}-a_n)^2-b_n^2)+i2b_n(\frac{1}{2}-a_n)}{[(\frac{1}{2}-a_n)^2+b_n^2]^2}|^2= [\sum_{n=1}^{\infty} \frac{1}{(\frac{1}{2}-a_n)^2+b_n^2}]^2$

$\Rightarrow [\sum_{n=1}^\infty\frac{((\frac{1}{2}-a_n)^2-b_n^2)}{ [ (\frac{1}{2}-a_n)^2+b_n^2]^2 }]^2 + [\sum_{n=1}^\infty\frac{2b_n(\frac{1}{2}-a_n)}{ [ (\frac{1}{2}-a_n)^2+b_n^2]^2 }]^2= [\sum_{n=1}^{\infty} \frac{1}{(\frac{1}{2}-a_n)^2+b_n^2}]^2$

user
  • 26,272
  • @CiaPan Thanks for the edit. Please answer. –  Mar 19 '21 at 12:24
  • You may want to apply the LaTeX/MathJax commands \left & \right to some parenthesizing characters to make them automagically adjust their height to the content, as I did to the title. :) – CiaPan Mar 19 '21 at 12:25
  • Equality certainly holds if $z_n={1\over2}+ib_n$, provided $\sum1/b_n^2$ converges. – Barry Cipra Mar 19 '21 at 12:28
  • @Barry Cipra We are given that the equality holds and we have to find condition on $z_n$. –  Mar 19 '21 at 12:30
  • Hint: If the right hand side converges, use the triangle inequality. – Tanny Sieben Mar 19 '21 at 12:31
  • @Alfred, yes, and I just gave one sufficient condition for equality to hold. Are you looking for sufficient and necessary? – Barry Cipra Mar 19 '21 at 12:32
  • @Barry Cipra yes, I am looking for necessary and sufficient condition –  Mar 19 '21 at 12:34
  • @Tanny Sieben What will we get by using triangle inequality? –  Mar 19 '21 at 12:35
  • That for any $i,j \in \mathbb{n}$ there is a real number $\lambda$ for which $\frac{1}{2}- z_i = \lambda( \frac{1}{2} - z_j)$ (or, more clearly, $\frac{\frac{1}{2} - z_i}{\frac{1}{2} - z_j} \in \mathbb{R}$) – Tanny Sieben Mar 19 '21 at 12:38
  • @Tanny Sieben Can you please write as an answer. Please. Its a humble request –  Mar 19 '21 at 12:40

1 Answers1

0

Let $w_n = \frac{1}{(\frac12-z_n)^2}$. You need that $$ \left|\sum_{n=1}^\infty w_n \right| = \sum_{n-1}^\infty |w_n|$$ First of all, for the right side of this condition to make sense, we need the series $\sum_{n-1}^\infty |w_n|$ to converge. It happens to be a sufficient condition for the series $\sum_{n=1}^\infty w_n $ to converge, so we don't need an additional condition for the convergence of the left sice of the condition.

Generaly we have $$ \left|\sum_{n=1}^\infty w_n \right| \le \sum_{n-1}^\infty |w_n|$$ (the generalized triangle inequality), with the equality happening only if all $w_n$ are proportional to each other with positive proportionaility coefficientsts, i.e. $$ \exists \omega\in\mathbb C\, \exists\lambda_n \in \mathbb R_+\, \forall n\in\mathbb N: w_n = \lambda_n\omega$$ Going back to $z_n$ this condition can be rewritten as $$ \exists \omega\in\mathbb C\, \exists\lambda_n \in \mathbb R_+\, \forall n\in\mathbb N: \frac{1}{(\frac12- z_n)^2} = \lambda_n\omega$$ $$ \exists \omega\in\mathbb C\, \exists\lambda_n \in \mathbb R_+\, \forall n\in\mathbb N : z_n -\frac12 = \pm\left(\frac{1}{\lambda_n\omega}\right)^\frac12$$ $$ \exists \zeta \in\mathbb C\, \exists\mu_n \in \mathbb R\setminus\{0\} \forall n\in\mathbb N : z_n = \frac12 + \mu_n\zeta$$

Let us assume that ${\rm Re}\, \zeta \neq 0$. From the condition $0 < {\rm Re}\,z_n < 1$ we get then $$ 0 < \frac12 + \mu_n {\rm Re}\,\zeta < 1 $$ $$ -\frac{1}{|{\rm Re}\,\zeta|} < \mu_n < \frac{1}{|{\rm Re}\,\zeta|}$$ We have then $$ |z_n-\frac12| = |\mu_n\zeta| < \frac{|\zeta|}{|{\rm Re}\,\zeta|}$$ $$ |w_n| = \frac{1}{|z_n-\frac12|^2} = \frac{1}{|\mu_n\zeta|^2}> \frac{|{\rm Re}\,\zeta|^2}{|\zeta|^2}$$ For such $|w_n|$ the series $\sum_{n=1}^\infty |w_n|$ would not be convergent. Therefore we need ${\rm Re}\, \zeta = 0$. Without a loss of generality we can choose $\zeta = i$, and we have the conditions $$ z_n = \frac12 + \mu_n i \qquad \text{for some } \mu_n\in\mathbb R\setminus\{0\}$$ $$ \text{series } \sum_{n-1}^\infty |w_n| = \sum_{n-1}^\infty \frac{1}{\mu_n^2} \text{ is convergent}$$

  • Thanks a lot.Please edit your answer. It has some typos. Please –  Mar 19 '21 at 12:41
  • 1
    There is only one consistent choice for $\zeta $. – user Mar 19 '21 at 12:52
  • @user what is the consistent choice? –  Mar 19 '21 at 12:59
  • @Alfred My comment refers to the previous redaction of the answer with the only restriction $\zeta\in\mathbb C $. – user Mar 19 '21 at 13:02
  • @user I could not get you. What are you saying? Please clarify –  Mar 19 '21 at 13:36
  • @Alfred The questions/answers on this site can be edited (you can try it yourself!). Therefore it can happen that a comment (or even an answer) refers to a previous redaction of the question/answer. To check the previous redactions you can click the button "edited" under the question/answer. – user Mar 19 '21 at 13:40
  • In your answer you used generalized traingle inequality. Please give me a reference for that. Does the triangle inequality holds when the number of terms are infinite? –  Mar 19 '21 at 15:15
  • https://math.stackexchange.com/questions/1576816/triangle-inequality-for-infinite-number-of-terms – Adam Latosiński Mar 19 '21 at 23:17