advanced29 min studyLesson 431 in Mathematics

Asymptotic sequences and Poincaré expansions

Interpret asymptotic coefficients as finite-order remainder claims, including useful expansions that diverge at every fixed nonzero parameter.

Before you begin

On this page

The question for today

What you’ll understand

Define an ordered asymptotic scale and a Poincaré expansion.
Distinguish convergence of an infinite series from finite-order asymptotic accuracy.
Derive an explicit remainder bound for a factorially divergent expansion.

An asymptotic expansion is a sequence of promises about finite approximations as a parameter approaches a limit. It need not be a convergent infinite sum. The distinction matters in applied mathematics because a divergent expansion can deliver excellent numerical accuracy if it is stopped at a suitable term. Conversely, writing many formal coefficients does not establish that a solution has the claimed expansion. The essential evidence is a remainder statement, together with its domain and the quantities held fixed while the parameter tends to its limit.

Order notation needs a declared limit

Let the positive parameter $\varepsilon$ tend to zero. The statement $f=O(g)$ means that the absolute value of their ratio is bounded for sufficiently small parameter, where the comparison is defined. The stronger statement $f=o(g)$ means that ratio tends to zero. Constants in a big-O statement may depend on a fixed truncation index or on a fixed compact interval unless uniformity is explicitly specified. Big-O is an upper bound, not an assertion of equal size, sign, or a nonzero leading coefficient.

An asymptotic sequence is ordered so that each successive comparison function is little-o of its predecessor. Powers of the parameter give one example; fractional powers, logarithms, and exponentially small functions can form other useful scales on suitable domains. The order must be checked in the stated limit. A logarithm that grows slowly is still unbounded, so grouping it with a constant can change which coefficient or remainder statement is correct. Choosing a suitable scale is part of solving a perturbation problem.

\phi_{n+1}(\varepsilon)=o(\phi_n(\varepsilon)),\qquad f(\varepsilon)\sim\sum_{n=0}^\infty a_n\phi_n(\varepsilon)\iff f-\sum_{n=0}^N a_n\phi_n=o(\phi_N)\ \text{for each fixed }N

The definition fixes the truncation index before taking the parameter limit. It does not interchange the limits N to infinity and epsilon to zero.↔ Scroll wide equations horizontally.

A divergent series can have controlled remainders

Consider the integral of $e^{-2t}/(1+\varepsilon t)$ over the positive real axis. A finite geometric identity expands the denominator into a polynomial plus an exact remainder. Integrating each polynomial term gives factorial coefficients. For every fixed positive parameter the infinite factorial series diverges, since the magnitude ratio of successive terms eventually exceeds one. Nevertheless the finite identity bounds the remainder after any fixed number of terms by the next factorial expression.

This proof does not exchange an infinite sum and an improper integral. It integrates a finite identity and estimates a nonnegative remainder. The denominator in that remainder is at least one for positive parameter, and the remaining exponential moment is elementary. Its sign also alternates with the number of retained terms, so consecutive truncations bracket the integral. These are stronger conclusions than a formal coefficient calculation and provide an actual numerical error certificate in this particular example.

I(\varepsilon)=\int_0^\infty\frac{e^{-2t}}{1+\varepsilon t}\,dt=\sum_{n=0}^{N-1}\frac{(-1)^n n!}{2^{n+1}}\varepsilon^n+R_N,\qquad |R_N|\le\frac{N!}{2^{N+1}}\varepsilon^N

The bound holds for positive epsilon and every positive integer N, by a finite geometric remainder followed by integration. It establishes the Poincaré expansion despite divergence of the infinite series.↔ Scroll wide equations horizontally.

Accuracy, uniformity, and uniqueness are different questions

If a function also depends on a spatial variable, a pointwise expansion fixes that variable first. A uniform expansion controls a supremum over a specified set with constants independent of position. An approximation may be excellent at every fixed positive position yet fail near an endpoint whose distance shrinks with the parameter. That failure is the central motivation for boundary-layer analysis. A list of coefficients without a stated spatial domain cannot settle uniform accuracy.

Coefficients in a prescribed ordered scale are uniquely determined when the expansion exists: subtract two candidate expansions and divide by the first scale at which their coefficients differ. The limiting remainder cannot cancel a nonzero constant. The function itself need not be determined by all those coefficients. Adding a term smaller than every power leaves an algebraic expansion unchanged. Later lessons examine how exponentially small information and sector-dependent continuation recover distinctions invisible to the power series.

Work it through

Certify a three-term approximation

Use the first three terms of $I(\varepsilon)$ at $\varepsilon=0.1$ , and bound the absolute error using the exact finite remainder.

For $n=0,1,2$ the terms are $1/2$ , $-\varepsilon/4$ , and $\varepsilon^2/4$ . At the chosen parameter they are $0.5$ , $-0.025$ , and $0.0025$ .
Their sum is 0.4775. Three retained terms mean N = 3 in the remainder bound, not N = 2.
The bound is $3!(0.1)^3/2^4=0.000375$ . The remainder has negative sign, so the integral lies between $0.477125$ and $0.4775$ .

The result

The finite approximation has certified absolute error at most 0.000375 even though its associated infinite series diverges at epsilon = 0.1.

The living notebook

Explore the model

Interactive lab

Interactive lab · function

A remainder can be smaller than the leading term

Reset and set both coefficient $a=1$ and position $x=0.2$ ; predict the quadratic value $0.04$ . Reset and set both $a=1,x=0.1$ ; predict $0.01$ , a factor-four decrease while the input halves. This illustrates power ordering and the ratio $x^2/x=x$ . The parabola is not the factorial integral and supplies no error bound for it; that bound comes from the finite identity.

y = ax²TangentSigned integral

y at chosen x: 0.608
Tangent slope · 2ax: 1.35
Signed area · ax³/3: 0.182

Coefficient a0.75Position x0.9

t = 0 s

This lab uses only y = ax². Playback traces the curve within the fixed integration interval from 0 to x; it does not estimate numerical error or change the chosen point. Parameter changes pause playback and recompute the model at the current elapsed time. Reset restores both parameters and returns time to zero. Playback starts paused. Adjust either parameter, then choose Play or Step.

Look forAn asymptotic scale compares rates of decay, while a proof must connect the scale to the actual remainder.

Pause & reason

Check your understanding

Interactive checks need JavaScript. You can still read the explanation.

Read the checkpoint explanation

The expansion concerns a family of finite-order claims with truncation fixed first. Convergence of the infinite series is a separate property, and the order of limits cannot be silently reversed.

Your turn

Try it yourself

For positive $\varepsilon$ , compare $f=1/(1+\varepsilon)$ and $g=f+e^{-1/\varepsilon}$ . Do they have the same expansion in nonnegative integer powers of $\varepsilon$ ? Explain why that does not make them equal.

Take a moment to make a prediction or work it out on paper. Open a hint when you need one.

Give me a hint

Set r = 1/epsilon and compare exponential decay with every fixed power of r.

Compare with the solution

For every fixed nonnegative integer $N$ , $e^{-1/\varepsilon}/\varepsilon^N=r^Ne^{-r}\to0$ . Thus the added exponential is smaller than every algebraic order and both functions have coefficients $1,-1,1,-1,\ldots$ . They differ by a strictly positive term at every positive parameter, so equality of all power-expansion coefficients does not imply equality of functions.

Keep the idea

Take it with you

An asymptotic expansion is a finite-truncation remainder statement.
Divergent factorial series can still yield rigorously accurate approximations.
Domains, uniformity, and beyond-all-orders differences must be stated separately.

Read further

Original sources and deeper explanations to follow at your own pace.

Optional progress, saved in this browser when available. Revisit any lesson at any time.

Lesson not marked complete.