Series: Convergence, Asymptotics, and Divergence in Physics

Welcome to the first lesson of our module on Asymptotic and Semiclassical Methods. In our last discussion, we examined how to assess the validity of an approximation by comparing the physics we keep with the physics we discard. A central theme was that approximations break down when a neglected parameter becomes large. This naturally leads to a profound question: what happens when our approximation, often a power series in some small parameter, doesn't just become inaccurate, but fails to converge at all?

This lesson confronts that issue directly. We will investigate a class of series that are mathematically divergent yet are cornerstones of theoretical physics. You will learn to distinguish between the familiar concept of a convergent series and the more subtle, powerful idea of an asymptotic series. More importantly, we'll explore why these divergent series are not a mathematical pathology to be avoided, but a fundamental and unavoidable feature of many physical problems, from simple algebraic equations to the core of quantum field theory.

1. A Tale of Two Perturbations

Often, the way we introduce a small parameter into a problem dramatically changes the nature of the solution we find. A simple algebraic equation, unsolvable in closed form, provides a perfect illustration of how two different perturbative approaches can lead to starkly different results: one beautifully convergent, the other stubbornly divergent.

Professor Carl Bender, a leading expert in this field, gives a masterclass on this exact issue. His approach, prioritizing powerful calculational techniques and physical insight, should align well with your goal of mastering the craft of theoretical physics.

Mathematical Physics 01 - Carl Bender

In this lecture, "Mathematical Physics 01," Carl Bender introduces perturbation theory by tackling a quintic equation, x^5 + x = 1, which cannot be solved exactly. He demonstrates two different ways to apply perturbation theory.

Please watch from this section. First, follow his "strong coupling" approach, where he inserts the parameter \epsilon as x^5 + \epsilon x = 1. Observe how this leads to a well-behaved, convergent series that gives a good approximation. Next, pay close attention starting at the "weak coupling" approach, where the equation is instead written as \epsilon x^5 + x = 1. Notice how this seemingly innocent change leads to a perturbation series with a radius of convergence less than 1, meaning it diverges for the physical problem where \epsilon=1. This contrast sets up the central puzzle of our lesson.

This example leaves us with a critical question: Why did the second approach yield a divergent series? And if a series diverges, is it useless? To answer this, we first need to be precise about what we mean by "convergent" and introduce its more powerful sibling, the "asymptotic" series.

2. Convergence vs. Asymptoticity: A Formal Distinction

The concepts of convergent and asymptotic series are defined by the order in which we take limits. This subtle difference in definition leads to a profound difference in their properties and utility.

A convergent power series is likely familiar from your analysis courses. For a series to converge to a function $f (x)$ , we fix the point $x$ and demand that the remainder term goes to zero as we include an infinite number of terms.

In contrast, an asymptotic power series provides a good approximation in a limiting regime. We fix the number of terms $N$ and demand that the remainder is "smaller" than the last term kept, as $x$ approaches a specific point. The series itself might diverge if we try to sum up all its terms.

The following reading provides the formal definitions.

[PDF] Asymptotic Power Series

This note by Howard Haber provides a concise and clear definition of both convergent and asymptotic power series, highlighting the different limiting procedures involved.

Please read the first page of the document. Focus on the formal definitions presented in equations (1) through (4). Compare the condition for a convergent series (taking N \to \infty) with the condition for an asymptotic series (taking x \to a).

The key conceptual distinction, which Professor Bender also emphasizes, is this:

Convergence is an absolute property. A series either converges or it doesn't, based solely on its coefficients. We can ask "Is this series convergent?" without reference to any function.
Asymptoticity is a relative property. A series is asymptotic to a specific function. The question "Is this series asymptotic?" is meaningless. The correct question is "Is this series asymptotic to function $f (x)$ as $x \to x_{0}$ ?"

This has a surprising consequence: a function has a unique asymptotic series, but a given asymptotic series can correspond to infinitely many functions (they would differ by terms, like $e^{- 1/ x}$ at $x = 0$ , that are "beyond all orders" of the power series).

Now that we have the vocabulary, let's tackle the main question: why are divergent asymptotic series so common and essential in physics?

3. Why Physics Demands Divergent Series

Divergent series are not a nuisance; they are often the only way to describe the solution to a physical problem. They typically arise when our "small parameter" fundamentally alters the nature of the equations, a situation known as a singular perturbation.

Reason 1: Singular Perturbations and "Lost" Solutions

Let's return to Bender's quintic equation, $ϵ x^{5} + x - 1 = 0$ . When we set $ϵ = 0$ , the equation becomes $x - 1 = 0$ , a linear equation with one root. The original quintic equation, however, must have five roots in the complex plane. In the limit $ϵ \to 0$ , four roots have vanished! This abrupt change in the problem's character is the hallmark of a singular perturbation. The perturbation series for the one "surviving" root tries to account for the complex behavior of the four "lost" roots, and the price it pays is divergence.

Mathematical Physics 01 - Carl Bender

In this segment of his lecture, Bender uses asymptotic analysis to find out what happened to the four "lost" roots.

Please watch from this part. He employs a powerful technique called the method of dominant balance to analyze the asymptotic behavior of the roots as \epsilon \to 0. The key insight is that the lost roots didn't disappear; they ran off to infinity. Focus on how assuming a balance between different terms of the equation allows him to find the asymptotic scaling of these runaway roots.

This is a general feature. The semiclassical limit of quantum mechanics, where $ℏ \to 0$ , is a singular perturbation. The Schrödinger equation, a second-order differential equation, degenerates into an algebraic equation of classical mechanics. The oscillatory, wavelike nature of the solution is "lost" in this limit, and as a result, perturbation series in powers of $ℏ$ are almost always divergent asymptotic series.

Reason 2: Irregular Singular Points

Sometimes, the mathematical structure of a problem guarantees that no convergent series solution can possibly exist. This occurs frequently when solving differential equations that have "irregular singular points."

Your background in mathematics will be useful here. For a linear differential equation, solutions near an "ordinary point" are analytic and have convergent Taylor series. Near a "regular singular point," we can find a convergent solution called a Frobenius series. But near an "irregular singular point," all bets are off.

Mathematical Physics 07 - Carl Bender

In his seventh lecture on Mathematical Physics, Carl Bender demonstrates why some problems have no convergent series solution.

Watch the segment from this point. Bender introduces the differential equation x^3 y'' = y. He shows that the point x=0 is an irregular singular point and proves that any attempt to find a convergent Taylor or Frobenius series solution leads to a contradiction. Then, from this point on, he shows what kind of solution does exist. Inspired by simpler examples, he tries an ansatz of the form y(x) = e^{S(x)}, where S(x) is singular at the origin. This is the foundation of the WKB approximation we saw in the previous lesson. Finally, watch the conclusion of this analysis from the conclusion. He reveals that the only way to represent the solution is via a structure involving an exponential "controlling factor" multiplied by a divergent asymptotic series.

This is a profound result. For a vast class of problems in physics described by such equations, divergent asymptotic series are not an artifact of a poor approximation method; they are the mathematically required, fundamental language for describing the solution. We are forced to use them.

4. The Practical Utility of Divergence

If we are forced to use divergent series, how can we get meaningful answers? The magic of an asymptotic series is that even though the infinite sum diverges, its partial sums can provide extraordinarily accurate approximations. The trick is to know when to stop summing.

For a typical divergent asymptotic series encountered in physics, the terms $a_{n} x^{n}$ will initially decrease in magnitude, reach a minimum, and then grow to infinity. The best possible approximation is obtained by truncating the series just before its smallest term. This is known as optimal truncation.

The following plot, which arises from calculating the chemical potential $μ$ of a Fermi gas at finite temperature $T$ (a classic condensed matter problem), provides a perfect visual. The calculation yields an asymptotic series in powers of $T$ .

A plot of the chemical potential $\mu$ versus temperature $T$ for a non-interacting Fermi gas. The solid blue line is the exact answer. The dashed lines show the approximations from truncating the divergent asymptotic series at different orders ($O(T^2)$, $O(T^4)$, etc.). Each truncated series provides a better approximation at low $T$, but inevitably diverges at higher $T$.

This plot beautifully illustrates the properties of an asymptotic series:

For a fixed, small temperature (e.g., $k T / ϵ_{F} = 0.2$ ), adding more terms (going from $O (T^{2})$ to $O (T^{4})$ ) improves the accuracy.
Each truncated series is only a good approximation in a limited range. The $O (T^{10})$ approximation is incredibly accurate for very small $T$ but is the first to diverge wildly.
By stopping at the right order for a given $T$ , we can obtain a highly accurate result.

In fact, it can be shown that for many quantum mechanical systems, the energy levels, when expanded in a perturbation parameter $ϵ$ , are functions that have singularities in the complex $ϵ$ -plane. A function with singularities cannot be represented by a power series that converges everywhere. This is the deep reason why perturbation theory in quantum mechanics and quantum field theory almost always yields divergent asymptotic series.

Conclusion

Today, we've ventured beyond the familiar territory of convergent series and discovered the power and ubiquity of their divergent, asymptotic counterparts.

Key Takeaways:

Convergent vs. Asymptotic: A convergent series is defined by taking the number of terms $N \to \infty$ for a fixed $x$ . An asymptotic series is defined by taking $x$ to its limit for a fixed $N$ . Convergence is an absolute property of a series; asymptoticity is a relative property between a series and a function.
Divergence is Unavoidable: Singular perturbations and irregular singular points in differential equations—common features of physical models—force solutions to be represented by divergent asymptotic series. Convergent series simply do not exist for these problems.
Divergence is Useful: Despite diverging, an asymptotic series can provide highly accurate numerical approximations when optimally truncated. The error of the approximation is typically on the order of the first neglected term.

You now understand what asymptotic series are and why they are a fundamental part of the theorist's toolkit. They are not a sign of failure but a signal of rich mathematical structure and singular limits.

In our next lesson, we will begin learning the practical techniques for generating these series. We will explore the saddle-point method (or method of steepest descent), a powerful tool for finding the asymptotic behavior of integrals that frequently appear in statistical mechanics and quantum field theory.

Recommended resources