Weierstrass M-Test for Uniform Convergence Explained

The Weierstrass M-test is the fastest way to prove that a series of functions converges uniformly: if each term satisfies |fn(x)| ≤ Mn for all x, where the Mn are constants independent of x and Σ Mn converges, then Σ fn(x) converges absolutely and uniformly on the domain. One dominating convergent number series does all the work.

This test is beloved because uniform convergence is usually hard to check directly — computing sup-norms of partial sums is painful — while finding a simple dominating bound is often easy. It is the standard tool for showing that series defining special functions converge nicely.

Key Concepts

• M-test statement: suppose |fn(x)| ≤ Mn for every x in D and every n, with Mn ≥ 0 and Σ Mn convergent. Then Σ fn converges absolutely and uniformly on D.
• Why it works: the Weierstrass M-test is essentially the comparison test applied uniformly: the tails of Σ fn(x) are controlled by the tails of Σ Mn, independently of x.
• Payoff: if each fn is continuous and the M-test applies, the sum S(x) = Σ fn(x) is continuous, and Σ fn can be integrated term by term.
• Sufficient, not necessary: a series can converge uniformly without any dominating convergent Σ Mn existing; the test is one-directional.
• Typical Mn: constants like 1/n², rn with r < 1, or 1/n! — the usual convergent benchmarks.

Formulas

• M-test: (|fn(x)| ≤ Mn ∀x) and (Σ Mn converges) ⇒ Σ fn converges absolutely and uniformly.
• Bounded oscillation: |sin(nx)| ≤ 1 and |cos(nx)| ≤ 1 for all real x and all n, the most-used bounds in Fourier-type series.
• Termwise integration: under the M-test with continuous fn, ∫ab Σ fn = Σ ∫ab fn.
• Weierstrass function: Σ ancos(bnπx) with 0 < a < 1 passes the M-test with Mn = an, giving a continuous-everywhere, differentiable-nowhere function.

Worked Example

Prove that Σn=1∞ sin(nx)/n² converges uniformly and absolutely on the whole real line.

1. Bound each term independently of x: |sin(nx)/n²| = |sin(nx)|/n² ≤ 1/n², since |sin| ≤ 1 everywhere.
2. Set Mn = 1/n². The series Σ Mn = Σ 1/n² converges by the p-test (p = 2 > 1).
3. The Weierstrass M-test now applies directly: Σ sin(nx)/n² converges absolutely and uniformly on ℝ.
4. Bonus: each term is continuous, so the sum is a continuous function of x (whose values are not expressible in elementary closed form, but which certainly exists).

Common Misconceptions

• "The Mn may depend on x." They may not — the whole point is a bound uniform across the domain; an x-dependent bound proves nothing about uniformity.
• "Failing the M-test means the series diverges." The test is sufficient only. The alternating series Σ (−1)nxn/n converges uniformly on [0, 1/2] by other arguments even where a crude M-test is awkward.
• "Uniform convergence implies the M-test applies." No: uniformity is the conclusion, and many uniformly convergent series have no dominating convergent majorant.

Real-World Applications

• Fourier series in signal processing: M-test bounds guarantee uniform convergence of smoothed series, hence continuous reconstructed signals.
• Defining special functions (zeta function, theta functions) as series and proving they are continuous on their domains.
• Power series inside their radius of convergence: on any closed subinterval the M-test with Mn = |an|ρn gives uniform convergence.

The Weierstrass M-test reduces uniform convergence to a single easy step: dominate every term by a constant Mn, and check that Σ Mn converges like an ordinary number series. When it applies, you get absolute convergence, uniform convergence, continuity of the sum and termwise integration all at once — which is why it is the first test to try on any series of functions.