Lacunary Series

SciencePedia

Key Takeaways

Lacunary series are power series defined by large, systematically growing gaps between their exponents.
These gaps often cause the series to have a "natural boundary" on its circle of convergence, a barrier beyond which the function cannot be analytically continued.
The Hadamard Gap Theorem provides a formal condition for this behavior, linking it to the rapid growth rate of the exponents.
The concept of a natural boundary appears in other domains, explaining the "jagged" nature of fractals and creating unique stability challenges in signal processing.
Contrary to intuition, functions with natural boundaries are the norm rather than the exception in the space of all possible analytic functions.

Introduction

In the world of complex analysis, power series are a fundamental tool, allowing us to build intricate functions from simple polynomial blocks. One of their most powerful features is analytic continuation, a process that lets us extend a function's definition far beyond its initial domain, like discovering a vast continent attached to a small island. But what if a function's domain is an island with no bridge, surrounded on all sides by an impassable sea? This is the fascinating problem posed by lacunary series—power series with deliberate, ever-widening gaps—which often generate functions confined within a "natural boundary." This article explores these remarkable mathematical objects.

The following chapters will guide you through the strange and beautiful world of lacunary series. In Principles and Mechanisms, we will dissect the structure of these gapped series, using intuitive examples and the powerful Hadamard Gap Theorem to understand exactly how they construct an impenetrable wall of singularities. Then, in Applications and Interdisciplinary Connections, we will see that this is no mere mathematical curiosity, discovering how the principle of natural boundaries provides deep insights into the jagged geometry of fractals, the stability of engineered systems, the secrets of prime numbers, and our very definition of a "typical" function.

Principles and Mechanisms

Imagine you are walking along a path defined by a mathematical formula. For some functions, the path is smooth and extends nearly forever, perhaps with a single pothole or a cliff at one specific point you can easily avoid. For others, you find yourself on a small island, and the moment you step off in any direction, you fall into an abyss. This island is the function's home, and its shoreline is a natural boundary. Lacunary series are the architects of these strange and beautiful islands.

But what is it about these series that creates such an impassable frontier? The secret lies not in the complexity of their terms, but in the emptiness between them.

The Anatomy of a Gap

Let's start with a familiar friend, the geometric series:

f(z) = \sum_{n=0}^{\infty} z^n = 1 + z + z^2 + z^3 + z^4 + \dots

This is a "full" or "dense" series; every integer power of $z$ is present. Its domain of convergence is the open unit disk, $|z| < 1$ .

Now, let's create a lacunary series by systematically punching holes in it. We'll keep only the terms whose powers are powers of two:

g(z) = \sum_{n=0}^{\infty} z^{2^n} = z + z^2 + z^4 + z^8 + z^{16} + \dots $$. The gaps, or ​**​lacunae​**​, between the powers aren't just present; they grow wider at an astonishing rate. The gap between $z^4$ and $z^8$ has 3 missing terms, but the gap between $z^8$ and $z^{16}$ has 7 missing terms, and so on. You might wonder if these gaps affect the function's [fundamental domain](/sciencepedia/feynman/keyword/fundamental_domain). Let's find the [radius of convergence](/sciencepedia/feynman/keyword/radius_of_convergence). For both the full geometric series and our gappy series, a quick check shows that they both converge when $|z| < 1$ and diverge when $|z| > 1$. So, both functions live happily inside the same home: the [unit disk](/sciencepedia/feynman/keyword/unit_disk). The profound difference between them is not about where they live, but whether they can ever leave. The defining feature of a lacunary series is this gapped structure of its exponents. The coefficients can be anything they like—they might grow, like in the series $\sum n (2z)^{n^2}$, or shrink, as in $\sum x^{m!}/m^2$. What defines their character are the yawning chasms between the powers. In fact, one of the most striking features of these series is that most of their Taylor coefficients at the origin are zero—they are built from an incredibly sparse set of "notes". ### The Wall of Singularities For our friendly [geometric series](/sciencepedia/feynman/keyword/geometric_series), $f(z) = \sum z^n$, we have a secret identity: inside the [unit disk](/sciencepedia/feynman/keyword/unit_disk), it is exactly equal to the function $\frac{1}{1-z}$. This new expression makes sense for *all* complex numbers except for the single point $z=1$, where it has a [simple pole](/sciencepedia/feynman/keyword/simple_pole). We have successfully performed an ​**​[analytic continuation](/sciencepedia/feynman/keyword/analytic_continuation)​**​: we found a function that matches our original series on its home turf but extends its definition far beyond its initial boundaries. The original boundary, the unit circle $|z|=1$, was like a fence with a single locked gate at $z=1$. We could sneak past it everywhere else. Now, let's try to do the same for our lacunary series, $g(z) = \sum z^{2^n}$. We search for a similar, more general formula. And we find... nothing. There is no way to extend this function beyond the unit disk. The unit circle is not a fence; it is a solid, infinitely high wall. Every single point on the circle is a ​**​singular point​**​, a place where the function's analytic nature breaks down. This is the essence of a ​**​[natural boundary](/sciencepedia/feynman/keyword/natural_boundary)​**​,. This has a beautiful and intuitive consequence. Imagine you are an analyst living inside the unit disk at the point $a = \frac{1}{2}$. You want to represent the function using a new power series centered at your home. How far will this new series converge? The rule is that a power series converges in a disk that extends to the nearest singularity. For the [geometric series](/sciencepedia/feynman/keyword/geometric_series), with its single singularity at $z=1$, you can draw a circle of convergence around $a=\frac{1}{2}$ that stretches all the way to $z=1$. But for our lacunary function, the entire unit circle is a wall of singularities. The closest point on this wall to you at $a=\frac{1}{2}$ is the point $z=1$. The distance is $|1 - \frac{1}{2}| = \frac{1}{2}$. That's your radius of convergence. You are hemmed in on all sides by the boundary. ### The Whispering Gallery of Pathologies Why? What is the physical mechanism behind this dramatic behavior? How do mere gaps in a series conspire to build an impenetrable wall? We can find a stunningly simple clue in the function's structure. Let's look at $g(z) = z + z^2 + z^4 + z^8 + \dots$ again. Notice a remarkable pattern of [self-similarity](/sciencepedia/feynman/keyword/self_similarity):

g(z) = z + (z^2) + (z^2)^2 + (z^2)^4 + \dots

The part in the parenthesis is just the original function, but with $z^2$ plugged in instead of $z$! This gives us a simple and powerful [functional equation](/sciencepedia/feynman/keyword/functional_equation):

g(z) = z + g(z^2) $$,. This equation is a propagator of pathology. It tells us that if the function misbehaves at some point $z_0$ , it must also misbehave at $z_0^2$ . But more importantly, if it misbehaves at $z_0$ , then $g(w)$ must also misbehave if $w^2 = z_0$ .

Let's start with an easy-to-spot trouble point: $z=1$ . The series $\sum 1^{2^n}$ is $1+1+1+\dots$ , which clearly diverges. So, $z=1$ is a singular point. According to our functional equation, this "sickness" must propagate. A singularity at $z=1$ implies singularities at its square roots, which are $z=1$ and $z=-1$ . Now we have two sick points. What about their square roots? The square roots of $1$ are $1$ and $-1$ ; the square roots of $-1$ are $i$ and $-i$ . So now $1, -1, i, -i$ must all be singular points.

We can continue this forever. The set of all points of the form $e^{i\pi m/2^k}$ for integers $m$ and $k$ must be singular. This set of points—the roots of unity of order $2^k$ —is dense on the unit circle. Like dust motes in a sunbeam, they are everywhere. No matter how tiny an arc of the circle you examine, you will find one of these singular points. And if you cannot find a single clean arc, you cannot perform an analytic continuation. The boundary is truly natural.

This beautiful intuition is formalized by the Hadamard Gap Theorem. It states that for a power series $\sum a_k z^{n_k}$ , if the exponents grow sufficiently fast—specifically, if the ratio of consecutive exponents $\frac{n_{k+1}}{n_k}$ is always greater than some number $q > 1$ —then the circle of convergence is a natural boundary. Our series $\sum z^{2^n}$ has a ratio of $2$ . A series like $\sum z^{k!}$ has a ratio of $k+1$ , which grows to infinity. They are all quintessential examples of Hadamard's powerful result.

Echoes in Other Fields

This principle of gaps creating barriers is not just a quirk of power series. It is a fundamental pattern in mathematics. It appears, for instance, in the heart of number theory, in the study of Dirichlet series. These are series of the form $\sum a_n n^{-s}$ . The most famous is the Riemann Zeta Function, $\zeta(s) = \sum n^{-s}$ , which can be analytically continued across its boundary of convergence ( $\operatorname{Re}(s)=1$ ) to almost the entire complex plane, save for a single pole at $s=1$ .

But we can construct a Dirichlet series with a natural boundary. Consider:

F(s) = \sum_{k=1}^{\infty} 2^{-2^k s} = \sum_{k=1}^{\infty} (2^{-s})^{2^k} $$. If we make the substitution $z = 2^{-s}$, this series transforms into our old friend $g(z) = \sum z^{2^k}$! The [domain of convergence](/sciencepedia/feynman/keyword/domain_of_convergence) for $F(s)$, the right half-plane $\operatorname{Re}(s) > 0$, maps perfectly onto the unit disk $|z|1$. The boundary of convergence, the imaginary axis $\operatorname{Re}(s)=0$, maps onto the unit circle $|z|=1$. Because the power series has a [natural boundary](/sciencepedia/feynman/keyword/natural_boundary) on the unit circle, the Dirichlet series must have a [natural boundary](/sciencepedia/feynman/keyword/natural_boundary) on the [imaginary axis](/sciencepedia/feynman/keyword/imaginary_axis). The dense set of [singular points](/sciencepedia/feynman/keyword/singular_points) on the circle corresponds to a [dense set](/sciencepedia/feynman/keyword/dense_set) of [singular points](/sciencepedia/feynman/keyword/singular_points) all along the vertical line $\operatorname{Re}(s)=0$. This reveals a deep and beautiful unity, where the same underlying principle governs the behavior of functions in seemingly disparate fields. One final, important clarification. A "singularity" on a [natural boundary](/sciencepedia/feynman/keyword/natural_boundary) does not necessarily mean the function's value blows up to infinity, or even that the series itself diverges. It is a more subtle kind of breakdown. Consider the series $f(x) = \sum_{n=1}^{\infty} \frac{x^{2^n}}{n^2}$. Due to its lacunary nature, it has a [natural boundary](/sciencepedia/feynman/keyword/natural_boundary) on the interval $[-1, 1]$. Yet, if we test the boundary points, we find that at both $x=1$ and $x=-1$, the series becomes $\sum \frac{1}{n^2}$, which converges to the finite value $\frac{\pi^2}{6}$. The function is perfectly well-behaved at these points. The "singularity" means that the property of being analytic—of being locally representable by a convergent [power series](/sciencepedia/feynman/keyword/power_series)—is lost. The function becomes, in a sense, infinitely "wrinkly" at every point on the boundary, preventing any smooth extension. It is this breakdown of local smoothness, not a simple explosion in value, that erects the impenetrable wall of a [natural boundary](/sciencepedia/feynman/keyword/natural_boundary).

Applications and Interdisciplinary Connections

In our previous discussion, we uncovered a remarkable property of lacunary series: their tendency to erect an "impenetrable wall" at the edge of their convergence disk, a barrier known to mathematicians as a natural boundary. At first glance, this might seem like a mere curiosity, a pathological quirk confined to the abstract realm of complex numbers. But nothing in mathematics lives in isolation. This strange behavior is, in fact, a deep and recurring theme that echoes across an astonishing range of scientific disciplines. It is a thread that connects the jagged edges of fractals, the stability of electronic systems, the mysteries of prime numbers, and even our understanding of what a "typical" function truly looks like. Let us now embark on a journey to trace these connections and witness the surprising unity that a simple concept—a series with gaps—can reveal.

The Wall and Its Guardians: Deeper into Function Theory

The Hadamard Gap Theorem gives us a clear rule: if the exponents $n_k$ in a power series $\sum c_k z^{n_k}$ grow fast enough (satisfying $\frac{n_{k+1}}{n_k} \ge q > 1$ ), then the circle of convergence is a natural boundary. For a series like $f(z) = \sum_{k=1}^{\infty} \frac{1}{k} ( \frac{z}{3} )^{2^k}$ , this theorem immediately tells us that the circle $|z|=3$ is an impassable barrier.

You might wonder if this wall is a fragile construction. Could we perhaps tear it down with a simple mathematical operation, like differentiation? Let's try. If we take a function defined by a lacunary series, say with exponents growing as fast as $k!$ , and compute its derivative, we find that the new series for the derivative is also a lacunary series. The wall remains standing, unmoved and just as impenetrable for the derivative as it was for the original function. This robustness is a clue that we are dealing with a fundamental property, not an accidental one.

What does it feel like to be a function living inside such a walled domain? As you approach the boundary, things can get incredibly wild. Consider the seemingly simple series $f(z) = \sum_{k=0}^{\infty} z^{2^k}$ . As $z$ approaches any point on the unit circle, the terms of the series stop canceling each other out and begin to pile up, causing the function's magnitude to spiral out of control. We can make this idea precise. The "Hardy space" $H^1$ is a club for "well-behaved" analytic functions whose average value remains finite as you approach the boundary. Our lacunary function $f(z)$ is so unruly that it is denied entry; its average value explodes, a direct consequence of its gappy nature.

Yet, in a beautiful paradox, this very gappiness can also enforce a certain kind of "honesty." A famous result known as a Tauberian theorem states that if a lacunary series is summable in a weak sense (like Abel summability), then it must also converge in the ordinary, stronger sense. The gaps prevent the kind of subtle cancellations that allow a non-convergent series to have a well-defined weak limit. The wall, in this sense, not only confines the function but also disciplines its behavior.

Echoes of the Wall in the Physical World

The abstract wall in the complex plane casts a very real shadow on the world we can see and build. Its echoes appear in the intricate geometry of nature and the practical design of engineering systems.

The Jagged Edge of Reality: Fractals

Let's walk along the boundary of our disk, substituting $z = e^{ix}$ into a lacunary series. What kind of real-valued function do we get? We get a function that wiggles and oscillates, but in a very peculiar way. The gaps in the frequencies, $\lambda_k$ , mean that we are adding up sine waves whose frequencies increase at an exponential rate. The result is a function that never smooths out. No matter how closely you zoom in on its graph, you find more wiggles, a hallmark of self-similar, fractal geometry.

The famous Weierstrass function—one of the first-discovered examples of a function that is continuous everywhere but differentiable nowhere—is precisely of this type. It's a "monster" from the perspective of classical calculus, but its construction is an elegant application of lacunary Fourier series. The degree of its "jaggedness" can be quantified by a number called the Hölder exponent. In a stunningly direct link between algebra and geometry, this exponent is determined by the balance between the rate at which the coefficients decay and the rate at which the frequencies grow. The faster the frequencies grow (the larger the gaps), the rougher the function becomes.

The Ghost in the Machine: Signal Processing

Can we build a physical system that embodies a natural boundary? In digital signal processing, the behavior of a system is described by its impulse response h[n], a sequence of numbers representing its output to a single input pulse at time $n=0$ . The system's properties are encoded in its transfer function $H(z)$ , which is the $z$ -transform of h[n]—nothing more than a power series in the variable $z^{-1}$ .

Imagine we construct a system that only responds at lacunary time intervals—say, at $n = 1, 2, 4, 8, \dots$ . The impulse response would be $h[n]=1$ if $n=2^k$ and $0$ otherwise. The transfer function for this system is $H(z) = \sum_{k=0}^{\infty} z^{-2^k}$ . This is our canonical lacunary series! Its region of convergence is $|z| 1$ , and the unit circle $|z|=1$ is a natural boundary.

For an engineer, this is deeply significant. A system is considered stable if its transfer function is well-behaved on this unit circle. For a typical system with a few isolated "poles" (singularities), an engineer can analyze the behavior near these points to understand and control instabilities. But for our lacunary system, every single point on the unit circle is part of an impenetrable wall of singularities. There is no "nearby" region to analyze; the instability is smeared across the entire boundary. The very concept of analytic continuation, a tool engineers might use to assess stability, fails completely. This isn't just an unstable system; it's a pathologically fragile one, a ghost in the machine whose instability is an intrinsic, holistic feature, not a localized flaw. Amazingly, one can even construct stable systems whose stability boundary is a natural boundary, creating a scenario where the system is stable, but lives right on the edge of this singular, impassable wall.

Abstract Vistas and Deeper Truths

The journey now takes us to the more abstract, yet profoundly insightful, realms of modern mathematics, where the concept of lacunarity provides both a powerful tool and a source of deep perspective.

Gaps in the Numbers

In analytic number theory, the central objects of study are not power series but Dirichlet series, of the form $F(s) = \sum a_n n^{-s}$ . The most famous of these is the Riemann zeta function, $\zeta(s) = \sum_{n=1}^{\infty} n^{-s}$ , which is deeply connected to the distribution of prime numbers. The series for $\zeta(s)$ converges for $\operatorname{Re}(s) > 1$ . Crucially, it can be analytically continued to the entire complex plane (with a single pole at $s=1$ ). This continuation is not just a technical trick; the properties of the zeta function in these new regions hold the key to understanding the primes.

Now, let's contrast this with a Dirichlet series built with lacunary exponents, such as $f(s) = \sum_{n=0}^{\infty} e^{-2^n s}$ . Here, the "frequencies" are $\lambda_n = 2^n$ . The Hadamard Gap Theorem applies just as before, and it tells us that the line of convergence, $\operatorname{Re}(s)=0$ , is a natural boundary. The difference is profound. The ability of the zeta function to be continued is tied to the "non-gappy," arithmetic structure of its frequencies, $\log n$ . The lacunary series, by its very nature, is trapped. This stark contrast highlights how the analytic properties of these functions, dictated by their frequency structure, are inextricably linked to the arithmetic secrets they encode.

The Futility of Approximation

If we can't analytically continue a function past its natural boundary, what happens if we try to force it? A powerful method for extending a function is to approximate it with a rational function (a ratio of polynomials), a technique known as Padé approximation. For an ordinary function, its Padé approximants are remarkably clever: the poles of the approximants often converge to the singularities of the function, revealing its hidden structure.

But when we apply this technique to a function with a natural boundary, like $f(z) = \sum z^{2^k}$ , the approximation scheme is met with a formidable challenge. It cannot create a finite number of poles to mimic the singularities, because there are infinitely many, smeared all along the boundary. Instead, the poles of the successive rational approximants begin to accumulate and spread themselves out along the unit circle, desperately trying to construct a "fence" to replicate the impenetrable barrier they are trying to model. It is a beautiful picture of failure, where the very way the approximation breaks down tells us about the true nature of the wall.

The Rule, Not the Exception

After seeing all these strange and wonderful examples, you might be left with the impression that lacunary series, and the functions they define, are rare beasts—carefully constructed oddities. The final twist in our story is perhaps the most profound. According to the Baire Category Theorem, a powerful result from topology, the opposite is true. If you consider the vast space of all possible analytic functions on the unit disk, the functions that can be analytically continued beyond their initial domain of convergence are the rare, exceptional cases. The "typical" function, in a precise topological sense, has a natural boundary.

Our lacunary series, therefore, are not pathological freaks. They are our simplest and most concrete examples of what is, in fact, the universal standard. The well-behaved, infinitely differentiable functions of an introductory calculus course are the true curiosities! We can even use the advanced tools of functional analysis, like Sobolev norms, to precisely measure the "roughness" of the boundary values generated by these typical functions, revealing a rich structure in their supposedly chaotic behavior.

From a simple algebraic property—gaps between exponents—we have journeyed through jagged fractal coastlines, explored the stability of engineered systems, touched upon the secrets of prime numbers, and ultimately, reshaped our very notion of what a "normal" function looks like. The story of the lacunary series is a perfect testament to the interconnectedness of mathematical ideas, showing how a single, simple concept can serve as a unifying thread, weaving a rich tapestry of insight across the entire landscape of science.