Badly Approximable Numbers

The quest to understand the nature of numbers often begins with a simple question: how well can we approximate irrational numbers using simple fractions? While some irrationals are surprisingly easy to pin down, others seem to elude our fractional messengers with remarkable persistence. This disparity gives rise to a fascinating class of numbers known as ​​badly approximable numbers​​—the masters of social distancing on the number line. This article tackles the mystery of these "most irrational" numbers, which, far from being a mathematical curiosity, represent a fundamental principle of order and stability found across the scientific world.

In the chapters that follow, we will first delve into the "Principles and Mechanisms," exploring the mathematical foundation of number approximation, from Dirichlet's universal rule to the unique properties that define badly approximable numbers. We will uncover how the elegant tool of continued fractions serves as a fingerprint to identify them, with the golden ratio emerging as the archetypal example. Subsequently, in "Applications and Interdisciplinary Connections," we will journey beyond pure mathematics to witness how this abstract property manifests in the real world, dictating the stability of solar systems, the efficiency of plant growth, and the future of frictionless technologies.

Imagine you're standing on one side of a vast canyon, and on the other side is an irrational number, say $\pi$. You want to send a message to it, but you can only use rational numbers—fractions like $p/q$—as your messengers. How close can your messenger get to $\pi$? You can use a bigger denominator, $q$, to get a more precise fraction, but that's like using a more expensive rocket. The real question is, for a given cost (the size of $q$), what's the best accuracy you can achieve?

It turns out there's a beautiful, universal rule of thumb. In the 19th century, the mathematician Peter Gustav Lejeune Dirichlet discovered something remarkable. He proved that for any irrational number $x$, you can always find infinitely many rational messengers $p/q$ that get startlingly close, satisfying the inequality:

Think about what this means. If you use a denominator of $100$, you're guaranteed to find a fraction that's within $1/10000$ of your target. If you use a denominator of a million, you'll get within one-trillionth. This isn't just a possibility; it's a guarantee, for every single irrational number on the line. Dirichlet's theorem sets a kind of universal speed limit for approximation; it tells us the baseline quality of approximation we can always expect to achieve.

This naturally sparks a new question. Is this the end of the story? Is everyone equally well-approximable? Or are there outliers? Are there numbers that are far, far easier to approximate, and others that stubbornly resist, doing the bare minimum to satisfy Dirichlet's law and nothing more?

The answer is a resounding yes. The universe of numbers is not so uniform. It has its superstars and its recluses.

On one extreme, we have numbers that are almost begging to be caught by fractions. These are the ​​Liouville numbers​​, named after Joseph Liouville. A number $x$ is a Liouville number if you can approximate it absurdly well. For any power $n$ you can dream of, no matter how large, you can find a fraction $p/q$ such that:

These numbers are so exceptionally close to rationals that Liouville was able to use this property to prove they are transcendental (meaning they are not the root of any polynomial equation with integer coefficients).

But what about the other extreme? What about numbers that are as "un-rational" as possible? These are the heroes of our story: the ​​badly approximable numbers​​. A number $x$ is badly approximable if it puts up a fight. It keeps its distance from all rational numbers. Formally, there exists some constant $c(x) > 0$ (depending on the number $x$ itself) such that for all rational numbers $p/q$, the following holds:

These numbers are the masters of social distancing on the number line. While Dirichlet's theorem guarantees they can be approached within $1/q^2$, they refuse to get much closer. They are the polar opposite of the easily-snared Liouville numbers. Interestingly, many familiar numbers, like $e$, are neither Liouville numbers nor badly approximable. The irrationality measure of $e$ is exactly $2$, meaning it sits right on the fence defined by Dirichlet's theorem, a fact that required a completely different method, developed by Charles Hermite, to prove its transcendence.

So, we have these elusive, "badly approximable" numbers. But how do we find them? Do they even exist? The key to unlocking this mystery lies in one of the most elegant tools in all of mathematics: the ​​continued fraction​​.

Any irrational number can be written as a unique, infinite nested fraction of the form:

This is often written in a compact notation as $[a_0; a_1, a_2, a_3, \dots]$. The integers $a_n$ are called the ​​partial quotients​​. You can think of this sequence of integers as a unique fingerprint or DNA sequence for the number $x$.

Here's the beautiful connection: the size of the partial quotients tells you everything about how well the number can be approximated. Each time you encounter a large partial quotient $a_n$, it signals that you've found an exceptionally good rational approximation. So, for a number to be badly approximable—to avoid having any exceptionally good approximations—its partial quotients must not get too large. In fact, the condition is stunningly simple:

A number is badly approximable if and only if its sequence of partial quotients is bounded.

Suddenly, we have a way to construct these numbers! The most famous example is the ​​golden ratio​​, $\phi = (1+\sqrt{5})/2 \approx 1.618\dots$. Its continued fraction is the simplest one imaginable:

All of its partial quotients are $1$, the smallest possible value. They are certainly bounded! This makes $\phi$ the "baddest" of them all—the most irrational of the irrationals, in a sense. This isn't just a curiosity. This property of $\phi$ is the deep reason behind the constant in Hurwitz's improved version of Dirichlet's theorem, which states that for any irrational $x$ there is an approximation with $|x - p/q| 1/(\sqrt{5}q^2)$. The constant $\sqrt{5}$ is optimal; you can't make it any larger, precisely because the golden ratio and its relatives would defy such a stronger statement.

This connection also gives us a vast, concrete family of badly approximable numbers: all ​​quadratic irrationals​​ (irrational numbers that are solutions to quadratic equations, like $\sqrt{3}$ or the golden ratio). A famous theorem by Lagrange states that a number is a quadratic irrational if and only if its continued fraction is eventually periodic. A periodic sequence is always bounded, so all quadratic irrationals are badly approximable. In contrast, algebraic numbers of higher degree, like $\sqrt[3]{2}$, are not badly approximable. Roth's theorem tells us they behave more like "typical" numbers in this regard.

We now know what badly approximable numbers are and how to spot them. Let's call the set of all such numbers $\mathrm{Bad}$. A natural question arises: how many of them are there? Are they common or rare? This is where the story takes a wonderfully paradoxical turn, revealing that the answer depends entirely on how you choose to measure "bigness".

Perspective 1: The View from Measure Theory

Our first tool is ​​Lebesgue measure​​, the standard mathematical way of defining the "length" of a set of points on the real line. The interval $[0, 1]$ has measure 1. The set of all integers has measure 0. What is the measure of $\mathrm{Bad}$?

The metric theory of continued fractions, using tools like the Gauss map, tells us something profound: for "almost every" real number, the sequence of partial quotients is unbounded. This means that a typical number, chosen at random, will have arbitrarily large partial quotients popping up in its continued fraction expansion, leading to exceptionally good rational approximations from time to time.

Since the set $\mathrm{Bad}$ consists of numbers whose partial quotients are bounded, it is the exception, not the rule. The conclusion is stark: the set of badly approximable numbers has Lebesgue measure zero. From this perspective, $\mathrm{Bad}$ is infinitesimally small, a mere dusting on the number line.

Perspective 2: The View from Topology

Let's try another tool. ​​Baire category theory​​ provides a topological way of thinking about the size of sets. A set is "meager" (or of the first category) if it's a countable union of "nowhere dense" sets—think of it as being topologically insignificant. A set is "comeager" if its complement is meager; it is topologically huge.

How does $\mathrm{Bad}$ fare? It turns out that $\mathrm{Bad}$ is a ​​meager set​​. We can write $\mathrm{Bad}$ as a union of sets $E_N$, where each $E_N$ contains numbers whose partial quotients are all less than or equal to $N$. Each of these sets $E_N$ can be shown to be "nowhere dense"—a kind of porous, filament-like structure. Since $\mathrm{Bad}$ is a countable union of these topologically flimsy sets, it is itself flimsy.

So, from two different and powerful perspectives, the set of badly approximable numbers seems vanishingly small. Its complement, the set of numbers that are not badly approximable, has full measure and is comeager. Case closed? Not quite.

Perspective 3: The View from a Game

Let's look at our set through a third lens, that of ​​fractal geometry​​ and a clever idea called ​​Schmidt's game​​. Imagine two players, Alice and Bob, playing a game on the number line. Bob picks an interval. Alice must then pick a smaller interval inside Bob's. Bob picks an even smaller one inside Alice's, and so on. The intervals shrink down to a single point. Alice wins if this final point lands in a pre-decided target set, say, $\mathrm{Bad}$. A set is called a ​​winning set​​ if Alice has a strategy to win no matter how cleverly Bob plays.

A winning set is robust. It can't be easily "pinned down" or avoided. It must be, in a sense, everywhere dense and complex. In a landmark result, Wolfgang Schmidt proved that the set $\mathrm{Bad}$ is a ​​winning set​​.

This changes everything! Winning sets have remarkable properties. One is that they have ​​full Hausdorff dimension​​. This is a concept from fractal geometry that generalizes our notion of dimension. A line has dimension 1, a plane has dimension 2. What's the dimension of $\mathrm{Bad}$? Even though its Lebesgue measure (length) is zero, its Hausdorff dimension is 1—the same as the entire real line!

This means that while the set is "thin" in terms of length, it is so intricately wrinkled and folded that its "complexity" or "richness" fills up a full dimension. Furthermore, $\mathrm{Bad}$ is what's known as a ​​thick set​​. This means if you take any open interval on the real line, no matter how tiny, and look at the part of $\mathrm{Bad}$ that lies inside it, that little piece still has Hausdorff dimension 1. The set $\mathrm{Bad}$ is not just a sprinkle of dust; it's a fractal web of infinite complexity, woven densely throughout the entire number line.

So, is the set of badly approximable numbers big or small? The beautiful answer is that it's both. It depends on your perspective.

• From the viewpoint of measure and classical topology, it's a "small" set of measure zero and meager category.
• From the viewpoint of dimensional complexity and game theory, it's a "large" set, a thick, winning set of full dimension.

There is no contradiction here. Instead, we find a deeper unity. The set of badly approximable numbers is a perfect illustration of the richness hidden in the structure of the real numbers. It forces us to appreciate that simple questions about "size" can have wonderfully complex answers, revealing that different mathematical tools can illuminate different, equally valid facets of the same profound truth. Its intricate structure, classified in the Borel hierarchy as a $G_{\delta\sigma}$ set, further hints at the deep complexity lurking just beneath the surface of our number system. These "bad" numbers, far from being a mere curiosity, are fundamental to our understanding of the very fabric of the number line.

We have seen that some numbers, like the golden ratio $\phi$, are special. They are "badly approximable," meaning they stubbornly resist being pinned down by simple fractions. One might be tempted to dismiss this as a mere curiosity, a niche obsession for number theorists. But to do so would be to miss one of the most beautiful and unifying stories in science. The property of being "badly approximable" is not a bug; it's a feature—a fundamental design principle that nature employs to create stability, efficiency, and order. Let us now take a journey through the disciplines and see how the echoes of these strange numbers resonate in the most unexpected places.

Imagine a miniature solar system, with planets orbiting a central star. The motion seems regular, almost clockwork. Each planet has its own orbital period, its own "frequency." What happens if the ratio of two planets' frequencies is a simple fraction, like $\frac{2}{1}$ or $\frac{3}{5}$? The planets will periodically align in the same configuration, giving each other repeated gravitational tugs. Just as a series of well-timed pushes can send a child on a swing higher and higher, these periodic tugs can amplify, destabilizing the orbits and eventually throwing the system into chaos. This is the specter of resonance, the great destroyer of celestial harmony.

To avoid this fate, the frequency ratio must be an irrational number. But what if it's an irrational number that is very close to a simple fraction, like $0.500000001$? The resonance may not be perfect, but its effects can still be devastatingly strong. In the mathematics of dynamics, this leads to the infamous "small denominator problem," where the calculations that predict the system's evolution threaten to explode.

The key to survival, as illuminated by the celebrated Kolmogorov-Arnold-Moser (KAM) theory, lies in having frequency ratios that are as far from rational as possible. The system's stability is a direct measure of how "badly approximable" its frequency ratio is. When a planetary system or any similar dynamical system is perturbed, the orbits corresponding to rational frequencies are the first to be destroyed. The last bastion of order, the most robust structure to survive the encroaching chaos, is the one whose frequency ratio is "the most irrational" of all. And which number holds this title? The golden ratio, $\phi = \frac{1+\sqrt{5}}{2}$, and its relatives, whose continued fraction expansions consist of the smallest possible integers. An orbit tuned to the golden ratio is nature's most resilient masterpiece, the "last surviving torus" in a sea of chaos. We can even imagine, as a pedagogical exercise, a 'fragility index' based on the size of the numbers in a frequency's continued fraction; the golden ratio would have the lowest fragility, the highest score for robustness.

Let's shift our perspective from dynamics in time to patterns in space. Suppose you have a task: place points, one by one, on a circle. Your goal is to keep the points as evenly spread out as possible at every stage. If you choose to place each new point at a fixed angular distance of, say, $\frac{2}{5}$ of a full circle from the last, you will quickly find yourself in a rut. After just five points, you will start placing new points directly on top of old ones. Your distribution is clumpy and inefficient.

The solution, once again, is to choose an angular step that is an irrational fraction of the circle. This guarantees you will never land on the same spot twice. But which irrational is best? Some irrationals, being "close" to rationals, will create near-collisions and temporary clumps before evening out. The most "fair" distribution—the one that fills the space most uniformly at every step—is generated by the badly approximable numbers.

We can quantify this "evenness" with a concept called discrepancy. A low-discrepancy sequence is one that is highly uniform. It has been proven that sequences generated by rotating by a badly approximable number, like the golden angle, exhibit the lowest possible discrepancy. They are the champions of equitable spacing, a principle that, as we shall see, is not just mathematically elegant but biologically essential.

Walk through a garden and look closely at a sunflower, a pinecone, or the arrangement of leaves on a stem (a pattern known as phyllotaxis). You will often find spiral patterns, and if you count them, the numbers of spirals will almost always be consecutive Fibonacci numbers: 8 and 13, or 21 and 34. This is no coincidence. It is the visible manifestation of our badly approximable hero, the golden ratio, at work.

At the growing tip of a plant, the shoot apical meristem, new leaves or seeds (called primordia) emerge one by one. Each new primordium needs space to grow and access to resources. The plant's simple, local rule is to place the next primordium in the largest available gap, the spot where it will face the least competition from its neighbors. Biochemically, this is the location where growth-inhibiting hormones, like auxin, are at their lowest concentration.

This is precisely the "fair distribution" problem we just discussed! Nature's solution, honed over eons of evolution, is to separate successive primordia by the golden angle, approximately $137.5^\circ$. This angle divides the circle in the golden ratio. Because it's the "most irrational" angle, it ensures that each new leaf is placed in a way that minimizes crowding and maximizes its access to light and air, not just with respect to its immediate neighbors, but with respect to all predecessors. This simple, iterative process gives rise to the complex and beautiful spiral patterns we admire.

From a more modern perspective, this strategy can be viewed as one of information maximization. By distributing its parts as uniformly as possible, the plant is maximizing the "spatial entropy" of its structure, ensuring there is no wasted space or redundant positioning. The plant isn't solving complex equations; it's simply following a local rule that, thanks to the magic of number theory, leads to a globally optimal, information-rich configuration.

Our final stop is at the frontier of modern technology: the world of nanotechnology and the fundamental nature of friction. Imagine two perfectly crystalline surfaces sliding past one another. We can model this as two interlocked atomic "combs." If the spacing of the teeth on both combs is the same, or if their ratio is a simple fraction (a commensurate interface), the atoms of one surface will lock neatly into the valleys of the other. To initiate sliding, you must lift the entire atomic layer "uphill" over the potential barrier. This results in significant static friction.

But what if the ratio of the atomic spacings is irrational? This is an incommensurate interface. The atoms can no longer all sit in the potential valleys simultaneously. For every atom that is in a low-energy position, another is forced to sit on a high-energy peak. Across an infinite surface, these energy contributions average out, and the energy barrier to sliding can vanish entirely. This remarkable phenomenon, where friction disappears due to atomic mismatch, is called superlubricity.

The Aubry transition marks the point where, as the substrate potential gets stronger, an incommensurate system suddenly goes from being free-sliding (unpinned) to being locked (pinned). And when is the system most resistant to being pinned? When is superlubricity most robust? You may have guessed it: when the ratio of the lattice spacings is a badly approximable number, such as the golden ratio. These numbers describe the interfaces that are "maximally incommensurate," making them the most difficult to lock into a periodic potential. The principle that stabilizes planets and grows sunflowers also provides the blueprint for the most slippery surfaces imaginable.

From the macrocosm of the planets to the microcosm of atoms, from the living architecture of a plant to the abstract realm of pure mathematics, a common thread appears. The humble property of being "badly approximable" provides a universal strategy for stability and efficiency.

What, then, is this set of magical numbers? It is a strange beast. The set of badly approximable numbers, $\mathrm{Bad}$, has a total "length" (or Lebesgue measure) of zero. If you were to pick a real number at random, the probability of it being badly approximable is nil. Yet, this set is not "small" in every sense. In the language of fractal geometry, its box-counting dimension is 1—the same as the entire line segment it lives in. It is an infinitely intricate, ghost-like structure, woven throughout the number line, weightless yet profoundly influential.

The discovery of this single, elegant principle at work in so many disparate fields is a testament to the deep unity of the mathematical and physical worlds. It is a beautiful reminder that by listening carefully to the abstract music of numbers, we can often hear the very tune to which the universe dances.