A Little Question from Math Twitter

A Little Question from Math Twitter

I normally don’t write up full solutions for the little math problems people post on Twitter (now 𝕏\mathbb{X}). This one was fun though.

=βˆ‘n=1∞1n(n+1)(n+1)!\begin{equation} = \sum_{n=1}^{\infty} \frac{1}{n\Big(n + 1)\Big(n + 1)!} \end{equation}

First, we should establish a few useful facts. Facts (1)\Big(1) and (2)\Big(2) are properties of the gamma function. Fact (3)\Big(3) uses the Maclaurin expansion of exe^x to calculate the sum of inverse factorials.

(1)Ξ“(n)=(nβˆ’1)!(2)nΞ“(n)=Ξ“(n+1)(3)βˆ‘n=1∞1n!=eβˆ’1,π‘›π‘œπ‘‘π‘–π‘π‘’:βˆ‘n=0∞xnn!=ex\begin{align} \Big(1) & \Gamma\Big(n) = \Big(n - 1)! \\ \Big(2) & n\Gamma\Big(n) = \Gamma\Big(n + 1) \\ \Big(3) & \sum\limits_{n=1}^{\infty} \frac{1}{n!} = e - 1 \qquad & ,\qquad \textit{notice:} \sum\limits_{n=0}^{\infty} \frac{x^n}{n!} = e^x \\ \end{align}

Wolphram Alpha: Partial Fraction Calc. We can quickly approximate a solution by ignoring the 1/Ξ“(n)1/\Gamma\Big(n) factor. In this case, (A)\Big(A) evaluates to the Basel Sum (Ο€2/6\pi^2/ 6), (C)\Big(C) evaluates to Ο€2/6βˆ’1\pi^2/6 - 1, and the difference in (B)\Big(B) and (D)\Big(D) is just βˆ’2-2. Overall, Ο€2/3βˆ’3β‰ˆ0.2898\pi^2/3 - 3 \approx 0.2898, not such a bad estimate!

Removing an (n+1)\Big(n + 1) and an nn from the (n+1)!\Big(n + 1)! gives the following. We can now use partial fractions to split the denominator into more managable terms.

=βˆ‘n=1∞1n2(n+1)2Ξ“(n)=βˆ‘n=1∞1Ξ“(n)(1n2⏟A+2(n+1)⏟B+1(n+1)2⏟Cβˆ’2n⏟D)\begin{align} = & \sum_{n=1}^{\infty} \frac{1}{n^2\Big(n + 1)^2\Gamma\Big(n)} \\ = & \sum_{n=1}^{\infty} \frac{1}{\Gamma\Big(n)} \left( \underbrace{\frac{1}{n^2}}_{\text{A}} + \underbrace{\frac{2}{\Big(n + 1)}}_{\text{B}} + \underbrace{\frac{1}{\Big(n + 1)^2}}_{\text{C}} - \underbrace{\frac{2}{n}}_{\text{D}} \right) \end{align}

The Β±βˆ‘1nn!\pm \sum\frac{1}{nn!} terms in (A)\Big(A) and (C)\Big(C) can be calculated just as we calculate (B)\Big(B). See: The Exponential Integral.

Taking each term one at a time we can calulate (B)\Big(B) and (D)\Big(D) quite easily. However, Attempting to calulate (A)\Big(A) and (C)\Big(C) is a bit more convoluted.

A)βˆ‘n=1∞1Ξ“(n)n2=βˆ‘n=1∞1nn!\begin{align} A) \qquad \sum_{n=1}^{\infty} \frac{1}{\Gamma\Big(n)n^2} = \sum_{n=1}^{\infty} \frac{1}{nn!} \end{align}

B)βˆ‘n=1∞2Ξ“(n)(n+1)=βˆ‘n=1∞2(nβˆ’1)!(n+1)=βˆ‘m=0∞2m!(m+2),π‘š := 𝑛 - 1=2βˆ‘m=0∞1m!∫01xm+1dx=2∫01βˆ‘m=0∞xm+1m!=2∫01xexdx=2\begin{align} & B) \qquad & \sum_{n=1}^{\infty} \frac{2}{\Gamma\Big(n)\Big(n + 1)} \\ & & = \sum_{n=1}^{\infty} \frac{2}{\Big(n - 1)!\Big(n + 1)} \\ & & = \sum_{m=0}^{\infty} \frac{2}{m!\Big(m + 2)} \; , \qquad \textit{m := n - 1} \\ & & = 2 \sum_{m=0}^{\infty} \frac{1}{m!} \int_0^1{x^{m + 1}} dx \\ & & = 2 \int_0^1 \sum_{m=0}^{\infty} \frac{x^{m + 1}}{m!} = 2 \int_0^1 xe^x dx \\ & & = 2 \\ \end{align}

Wolphram Alpha: Partial Fraction Calc for (C)\Big(C)

C)βˆ‘n=1∞1Ξ“(n)(n+1)2=βˆ’βˆ‘n=1∞1nn!+βˆ‘n=1∞1n!=βˆ’βˆ‘n=1∞1nn!+(eβˆ’1)\begin{align} & C) \qquad & \sum_{n=1}^{\infty} \frac{1}{\Gamma\Big(n)\Big(n + 1)^2} \\ & & = -\sum_{n=1}^{\infty} \frac{1}{nn!} + \sum_{n=1}^{\infty} \frac{1}{n!} \\ & & = -\sum_{n=1}^{\infty} \frac{1}{nn!} + \Big(e - 1) \end{align}

D)βˆ’2βˆ‘n=1∞1Ξ“(n)n=βˆ’2βˆ‘n=1∞1n!=βˆ’2(eβˆ’1)\begin{align} & D) \qquad & -2\sum_{n=1}^{\infty} \frac{1}{\Gamma\Big(n)n} = -2\sum_{n=1}^{\infty} \frac{1}{n!} \\ & & = -2\Big(e - 1) \end{align}

Noticing that the Β±βˆ‘1nn!\pm \sum\frac{1}{nn!} terms from (A)\Big(A) and (C)\Big(C) sum to zero, we collect the remaining terms without ever needing to evaluate those sums.

2+(eβˆ’1)βˆ’2(eβˆ’1)=3βˆ’eβ‰ˆ0.2817β—Ό\begin{equation} 2 + \Big(e - 1) - 2\Big(e - 1) = 3 - e \approx 0.2817 \qquad \qquad \blacksquare \end{equation}