Before we start, one picture that every problem leans on: the difference between an even and an odd function, because spotting symmetry (see Even and odd functions) kills half the integrals instantly.
Alt-text / how to read this figure: Two panels share the same x-axis. Left (blue): the parabola f(x)=x2. Trace any height on the right and you find the same height at the mirrored x on the left — the yellow double-arrow marks one such matched pair. Because the two sides agree, when you multiply by the odd function sinLmπx the left and right contributions are equal-and-opposite and cancel, so every bm=0; only cosines (also even) survive. Right (red): the line g(x)=x. The two green dots sit at (1.6,1.6) and (−1.6,−1.6) — same distance from the origin but opposite sign (the yellow curved arrow is the 180∘ rotation that maps one onto the other). Multiplying by the even function cosLmπx now makes an odd product that cancels, so every am=0; only sines survive. This single mental move — "is f a mirror or a pinwheel?" — is what you apply in almost every exercise below.
Even (f(−x)=f(x), mirror across the vertical axis): only an survive, all bn=0.
Odd (f(−x)=−f(x), rotate 180∘ about the origin): only bn survive, all an=0.
a0:a0=π1∫−ππfdx=π1(∫−π01dx+∫0π0dx)=π1(π)=1. Constant term =a0/2=21 (the average of f).
am:am=π1∫−π0cos(mx)dx=π1[msinmx]−π0=π1(0−msin(−mπ))=0 (since sinmπ=0). So all am=0 for m≥1.
bm:bm=π1∫−π0sin(mx)dx=π1[−mcosmx]−π0=π1(−m1+mcos(−mπ))=mπ(−1)m−1.
This is 0 for even m and −mπ2 for odd m.
f(x)=21−π2∑k=0∞2k+11sin((2k+1)x).
Value at x=0. Here f jumps: the left limit is f(0−)=1, the right limit is f(0+)=0. By the Dirichlet convergence theorem the series converges to the midpoint21(f(0−)+f(0+))=21(1+0)=21. Check: at x=0 every sin((2k+1)⋅0)=0, so the series literally sums to the constant 21. This is why the constant term (a0/2=21) already lands on the midpoint — reassuring, not coincidence.
Recall Solution 2.2
a0:a0=π1∫−ππx2dx=π1⋅32π3=32π2. Constant term =a0/2=3π2.
am: since x2cosmx is even, am=π2∫0πx2cosmxdx.
Why integration by parts, and why twice? We have a polynomial times a trig function. There is no product rule for integrals, but integration by parts lets us trade a derivative for an integral: each time we choose u= the polynomial part, differentiating x2→2x→2lowers its degree by one. After two rounds the polynomial is a constant and the integral becomes a plain trig integral we can finish. (Choosing u= the trig part instead would raise complexity — always differentiate the polynomial.) With u=x2,dv=cosmxdx:
∫0πx2cosmxdx=[mx2sinmx]0π−m2∫0πxsinmxdx.
The bracket is 0 (sinmπ=0). Then ∫0πxsinmxdx=−mπ(−1)m (the second round of parts, as in the parent's Example 2). So
∫0πx2cosmxdx=−m2(−mπ(−1)m)=m22π(−1)m.
Thus am=π2⋅m22π(−1)m=m24(−1)m.x2=3π2+4∑m=1∞m2(−1)mcosmx.
Put x=π. By the convergence theorem the series equals f(π)=π2. Why is this point legal? The 2π-periodic extension of x2 is continuous at x=π (both sides approach π2), so there is no jump and the series hits the true value — not a midpoint. Since cosmπ=(−1)m:
π2=3π2+4∑m=1∞m2(−1)m(−1)m=3π2+4∑m=1∞m21.
Rearrange: 4∑m21=π2−3π2=32π2, so ∑m=1∞m21=6π2.✓
Recall Solution 3.2
Left side: f2=1 everywhere, so π1∫−ππ1dx=π1(2π)=2.
Right side: a0=0, all an=0, and ∑bm2=∑odd m(mπ4)2=π216∑k=0∞(2k+1)21.
Set equal: 2=π216∑k=0∞(2k+1)21, giving ∑k=0∞(2k+1)21=8π2.
No symmetry — compute all. First, why the two helper integrals below have that shape. When you integrate excosmx by parts twice, the integral reappears on the right, and you solve for it algebraically (like solving I=(stuff)−kI). The clean results are
∫excosmxdx=1+m2ex(cosmx+msinmx),∫exsinmxdx=1+m2ex(sinmx−mcosmx).
You can verify either by differentiating the right side (product rule) and watching it collapse back to the integrand — the 1+m2 in the denominator is exactly what makes the cross-terms cancel. Write sinhπ=2eπ−e−π.
a0:a0=π1∫−ππexdx=π1(eπ−e−π)=π2sinhπ.
am:am=π1[1+m2ex(cosmx+msinmx)]−ππ. At x=±π: sinmπ=0,cosmπ=(−1)m, so the bracket is 1+m2(−1)mex evaluated at the ends:
am=π1⋅1+m2(−1)m(eπ−e−π)=π(1+m2)2(−1)msinhπ.
bm:bm=π1[1+m2ex(sinmx−mcosmx)]−ππ. At x=±π the bracket is 1+m2−m(−1)mex, so
bm=π1⋅1+m2−m(−1)m(eπ−e−π)=π(1+m2)−2m(−1)msinhπ.
ex=π2sinhπ[21+∑m=1∞1+m2(−1)m(cosmx−msinmx)].
Recall Solution 4.2
Quick refresher on the complex framework. Using Euler's formula e±inx=cosnx±isinnx, the real series can be repackaged as f(x)=∑n=−∞∞cneinx where each complex coefficient is cn=2π1∫−ππf(x)e−inxdx (the e−inx is the "matching fork" being projected out, exactly as cos,sin were in the real case). The bridge back to real coefficients: grouping the +n and −n terms and using cosnx=2einx+e−inx gives an=cn+c−n and bn=i(cn−c−n). So checking cn+c−n against an tests both frameworks at once.
cn=2π1∫−ππe(1−in)xdx=2π1⋅1−ine(1−in)π−e−(1−in)π. Since e∓inπ=(−1)n (integer n: e−inπ=cosnπ−isinnπ=(−1)n, purely real), the numerator is (−1)n(eπ−e−π)=2(−1)nsinhπ. So cn=π(1−in)(−1)nsinhπ.
Then an=cn+c−n=π(−1)nsinhπ(1−in1+1+in1)=π(−1)nsinhπ⋅1+n22=π(1+n2)2(−1)nsinhπ.
For n=0: a0=π2sinhπ; for n=1: a1=2π−2sinhπ=π−sinhπ. Both match Ex 4.1. ✓
Expand x(π−x)=πx−x2. We need ∫0πxsinmxdx and ∫0πx2sinmxdx.
Why parts again? Same reason as Ex 2.2 — polynomial × trig — so differentiate the polynomial to grind its degree down.
From the parent, ∫0πxsinmxdx=−mπ(−1)m.
For the second, integrate by parts twice:
∫0πx2sinmxdx=[−mx2cosmx]0π+m2∫0πxcosmxdx.
The bracket =−mπ2(−1)m. And ∫0πxcosmxdx=[mxsinmx]0π−m1∫0πsinmxdx=0−m1⋅m1−(−1)m=m2(−1)m−1.
So ∫0πx2sinmxdx=−mπ2(−1)m+m2⋅m2(−1)m−1=−mπ2(−1)m+m32((−1)m−1).
Now combine (π⋅ first − second):
∫0π(πx−x2)sinmxdx=π(−mπ(−1)m)−(−mπ2(−1)m+m32((−1)m−1))=−m32((−1)m−1).
The π2(−1)m/m terms cancel exactly. Therefore
Bm=π2⋅(−m32((−1)m−1))=πm34(1−(−1)m)=⎩⎨⎧πm380m oddm even.
Only the odd harmonics appear — consistent with the profile being symmetric about the rod's midpoint x=π/2.
Recall Solution 5.2
Left side: ∫0πx2(π−x)2dx. Let I=∫0π(π2x2−2πx3+x4)dx=π23π3−2π4π4+5π5=π5(31−21+51)=30π5.
So LHS =π2⋅30π5=15π4.
Right side: ∑Bm2=∑odd m(πm38)2=π264∑k=0∞(2k+1)61.
Equate: 15π4=π264∑(2k+1)61, so
∑k=0∞(2k+1)61=15⋅64π6=960π6.
Recall Self-test checklist
Read each line as "Prompt ::: Answer" — cover the text after the :::, say your answer aloud, then reveal to check.
Interval [−L,L] uses which prefactor for every coefficient? ::: L1
Odd f has only which coefficients nonzero? ::: only bm (all am=0)
Even f has only which coefficients nonzero? ::: only am (all bm=0)
At a jump discontinuity the series converges to what value? ::: the midpoint of the two one-sided limits
Does a pure sine (half-range) series have a constant B0 term? ::: No — sin0=0, so there is no constant
Parseval links ∫f2 to what? ::: the sum of the squared coefficients