Type a command or search the notes. Tab cycles completions. Up and down recall history when blank. Escape dismisses, then leaves the prompt.
MATH
[1251]

Euler's Basel sum

Combine a geometric-series integral, a Jacobian and symmetry to prove the Basel sum.

This page uses the chapter's tools to prove Euler's result ∑n≥11/n2=π2/6\sum_{n\ge1}1/n^2=\pi^2/6. The p-series test with p=2p=2 shows that the series converges; a double integral finds its value.

The geometric-series formula and Fubini's theorem write the series as an integral over the unit square. A linear change of variables turns the square into a diamond, sliced as in regions and order of integration. Symmetry and one-variable substitutions then finish the evaluation.

[THEOREM]

Change of variables

Let Ω\Omega be a region in (x,y)(x,y)-coordinates and let Φ\Phi be a region in (u,v)(u,v)-coordinates. Assume that ff is integrable on Ω\Omega, that gg is C1C^1 on Φ\Phi, that Φ\Phi can be split into finitely many pieces on which gg is one-to-one with non-overlapping interiors and that the Jacobian is nonzero away from boundary or other exceptional sets of area zero. The images of those pieces must cover Ω\Omega once, apart from boundaries of area zero; otherwise the transformed integral can omit or count part of the region more than once. If a map covers the target with multiplicity, split the parameter region into one-to-one pieces (or account for that multiplicity) before applying the formula. Then

∬Ωf(x,y) dx dy=∬Φf(x(u,v),y(u,v))∣det⁡Jg(u,v)∣ du dv.\iint_\Omega f(x,y)\,dx\,dy = \iint_\Phi f\bigl(x(u,v),y(u,v)\bigr) \left|\det J_g(u,v)\right|\,du\,dv.
[DEFINITION]

Jacobian determinant

The Jacobian determinant is

det⁡Jg(u,v)=∂(x,y)∂(u,v).\det J_g(u,v) =\frac{\partial(x,y)}{\partial(u,v)}.

Its absolute value is the local area-scaling factor:

dA=dx dy=∣det⁡Jg(u,v)∣ du dv.dA=dx\,dy =\left|\det J_g(u,v)\right|\,du\,dv.

Turn the series into an integral

For n≥0n\ge0,

∫01xn dx=1n+1.\int_0^1x^n\,dx=\frac1{n+1}.

Multiplying two copies of this identity and using Fubini on the unit square gives

1(n+1)2=∫01∫01(xy)n dx dy.\frac1{(n+1)^2} =\int_0^1\int_0^1(xy)^n\,dx\,dy.

To sum over nn, use the geometric series: for 0≤x,y<10\le x,y<1,

∑n=0∞(xy)n=11−xy.\sum_{n=0}^\infty(xy)^n=\frac1{1-xy}.

The finite geometric sums are nonnegative and increase pointwise, so monotone convergence justifies taking their limit under the integral.

The integrand 1/(1−xy)1/(1-xy) is singular only at the corner (1,1)(1,1), but it is unbounded in every neighbourhood of that corner. The integral is therefore improper, defined as the monotone limit over bounded truncations. The corner has zero area, but the blow-up near it cannot be ignored.

Summing the identities over n≥0n\ge0 therefore gives

S:=∑n=1∞1n2=∬[0,1]211−xy dx dy.S:=\sum_{n=1}^\infty\frac1{n^2} =\iint_{[0,1]^2}\frac1{1-xy}\,dx\,dy.

For the same reason, the proof applies the change-of-variables theorem on compact subregions that stay away from (1,1)(1,1) and then removes the truncation by monotone convergence. The transformed integrals are understood as the corresponding improper limits.

[THEOREM]

Euler's evaluation of the Basel sum

The reciprocal-square series satisfies

∑n=1∞1n2=π26.\sum_{n=1}^{\infty}\frac1{n^2}=\frac{\pi^2}{6}.
  1. [1]:Proof
[PROOF]

Use the linear substitution

x=u−v,y=u+v.x=u-v,\qquad y=u+v.

The inverse transformation is

u=x+y2,v=y−x2,u=\frac{x+y}{2}, \qquad v=\frac{y-x}{2},

and

∣∂(x,y)∂(u,v)∣=∣det⁡(1−111)∣=2.\left|\frac{\partial(x,y)}{\partial(u,v)}\right| = \left| \det \begin{pmatrix} 1&-1\\ 1&1 \end{pmatrix} \right| =2.

The square 0≤x,y≤10\le x,y\le1 becomes a diamond with vertices (0,0)(0,0), (12,±12)(\tfrac12,\pm\tfrac12) and (1,0)(1,0) in the uvuv-plane. Vertical slices describe it as

{0≤u≤12,−u≤v≤u,12≤u≤1,u−1≤v≤1−u.\begin{cases} 0\le u\le\frac12,\quad -u\le v\le u,\\[4pt] \frac12\le u\le1,\quad u-1\le v\le1-u. \end{cases}

In the new coordinates,

xy=(u−v)(u+v)=u2−v2,1−xy=1−u2+v2.xy=(u-v)(u+v)=u^2-v^2, \qquad 1-xy=1-u^2+v^2.

The integrand is even in vv, so we integrate over the upper half and multiply by two. The Jacobian supplies another factor of two, giving

S=4∫01/2∫0u11−u2+v2 dv du+4∫1/21∫01−u11−u2+v2 dv du.\begin{aligned} S &= 4\int_0^{1/2}\int_0^u \frac{1}{1-u^2+v^2}\,dv\,du\\ &\quad+ 4\int_{1/2}^{1}\int_0^{1-u} \frac{1}{1-u^2+v^2}\,dv\,du. \end{aligned}

For fixed uu,

∫0adv1−u2+v2=11−u2arctan⁡(a1−u2).\int_0^a\frac{dv}{1-u^2+v^2} = \frac1{\sqrt{1-u^2}} \arctan\left(\frac{a}{\sqrt{1-u^2}}\right).

Each piece now needs one trigonometric substitution. The first uses u=sin⁡θu=\sin\theta; the second uses u=cos⁡θu=\cos\theta, so that the inverse tangent simplifies by the half-angle identity.

For the first piece, let u=sin⁡θu=\sin\theta, so 0≤θ≤π/60\le\theta\le\pi/6. Then

u1−u2=tan⁡θ,\frac{u}{\sqrt{1-u^2}}=\tan\theta,

and hence

S1=4∫01/211−u2arctan⁡(u1−u2) du=4∫0π/6θ dθ=π218.\begin{aligned} S_1 &= 4\int_0^{1/2} \frac1{\sqrt{1-u^2}} \arctan\left(\frac u{\sqrt{1-u^2}}\right)\,du\\ &=4\int_0^{\pi/6}\theta\,d\theta =\frac{\pi^2}{18}. \end{aligned}

For the second piece, let u=cos⁡θu=\cos\theta, so 0≤θ≤π/30\le\theta\le\pi/3. As uu runs from 1/21/2 to 11, θ\theta runs from π/3\pi/3 to 00; flipping the limits removes the minus sign from du=−sin⁡θ dθdu=-\sin\theta\,d\theta. The ratio in the inverse tangent is

1−u1−u2=1−cos⁡θsin⁡θ=tan⁡θ2.\frac{1-u}{\sqrt{1-u^2}} =\frac{1-\cos\theta}{\sin\theta} =\tan\frac\theta2.

Therefore

S2=4∫0π/3θ2 dθ=π29.\begin{aligned} S_2 &=4\int_0^{\pi/3}\frac\theta2\,d\theta\\ &=\frac{\pi^2}{9}. \end{aligned}

Adding the two pieces,

S=S1+S2=π218+π29=π26.S=S_1+S_2 =\frac{\pi^2}{18}+\frac{\pi^2}{9} =\frac{\pi^2}{6}.

On this page