Function series and the Weierstrass M-test

Try

A function series is a sequence of partial-sum functions. For Σ from k=1 of sin(kx)/2^k, every term oscillates, but its amplitude is at most 2^(−k). The total remaining amplitude after N terms is bounded by the numerical geometric tail 2^(−N), independently of x. This is the essential mechanism of the Weierstrass M-test: replace spatially varying terms by one summable list of nonnegative bounds.

Predict

Predict whether cancellation is needed in this example. It is not. Even replacing every sine value by its largest possible absolute magnitude leaves a convergent numerical series. The resulting estimate is conservative but uniform. A series that relies on cancellation may still converge uniformly, yet the M-test might not prove it. Sufficient tests should not be turned into necessary conditions.

Experiment

Record the uniform tail enclosure.

Last retained index N	Numerical majorant tail	Guaranteed absolute function tail
1	1/2	At most 1/2
4	1/16	At most 1/16
N	2^(−N)	At most 2^(−N)

The bound controls the full infinite tail after convergence is established. First prove the corresponding estimate for every finite tail, then use the uniform Cauchy criterion. Writing an infinite-tail inequality before existence is known can conceal that logical step.

Understand

Suppose |f_k(x)|≤M_k for every x∈D, where M_k≥0 and ΣM_k converges. For m>n, the finite tail satisfies |Σ from k=n+1 to m f_k(x)|≤Σ from k=n+1 to m M_k. The numerical Cauchy criterion makes the right side arbitrarily small beyond a cutoff independent of x. Hence the partial-sum functions are uniformly Cauchy and have a uniform limit. The same bound shows absolute pointwise convergence.

If every term is continuous, every finite partial sum is continuous and the uniform-limit theorem makes the sum continuous. On a fixed compact interval, termwise integration follows from uniform convergence of partial sums. These are consequences requiring their own preservation theorems; the M-test itself establishes convergence, not every conceivable operation on the sum.

Apply

Termwise differentiation requires a different bound. For C¹ terms, convergence of the scalar series at one anchor and uniform convergence of the derivative series allow the derivative-interchange theorem to be applied to partial sums. Uniform convergence of the original series alone is not sufficient. For a series involving sin(kx), differentiation introduces a factor k, so a majorant that worked before differentiation may cease to be summable.

Repair “the M-test fails, therefore the series diverges.” Constant functions f_k(x)=(−1)^(k+1)/k give a uniformly convergent series on any nonempty domain, because its scalar alternating remainder is independent of x. But the series of supremum norms is harmonic and diverges. Cancellation proves convergence here, while absolute norm summability is absent.

Continue

Optional depth treats absolutely summable function series as a series in the complete normed space of bounded functions. The estimate Σ||f_k||∞<∞ is exactly the M-test expressed using a norm. It provides robust control under rearrangement because numerical absolute tails bound any omitted collection of indices.

Save a term bound, a finite-tail estimate and the theorem used to preserve continuity or integration as three separate steps. Next, power series will supply particularly useful majorants: strictly inside their convergence radius, geometric decay dominates the polynomial factors introduced by differentiation.