The Hidden Pattern That Splits Every Integer Exactly Once

Added: 2026-05-31

[00:00:00]Pick any irrational number bigger than one. Call it alpha, square root of two, 1.414213, pi divided by three, 1.047, the golden ratio 1.61803.

[00:00:16]Whatever you like, as long as it's irrational and bigger than one. Now compute 1 over 1 minus 1 over alpha.

[00:00:24]Call this beta.

[00:00:25]For square root of two, beta works out to 3.41421.

[00:00:31]For the golden ratio, beta is exactly the golden ratio squared, which is 2.61803.

[00:00:37]For pi over three, beta is something close to 3.89.

[00:00:42]So we have two irrational numbers, alpha and beta, derived from any irrational we picked.

[00:00:48]They satisfy a clean relationship. 1 divided by alpha plus 1 divided by beta equals exactly 1.

[00:00:56]Now make two sequences. The first sequence is the floor of n times alpha for n equals 1, 2, 3, and so on.

[00:01:04]The floor function just rounds down to the nearest integer. So for alpha equals square root of two, the sequence is floor of 1.414, floor of 2.828, floor of 4.243, and so on. That gives 1, 2, 4, 5, 7, 8, 9, 11, 12, 14. The second sequence is the floor of n times beta, also for n equals 1, 2, 3, and so on.

[00:01:31]For beta corresponding to square root of two, that's 3, 6, 10, 13, 16, 20, 23, 27.

[00:01:38]Look at these two sequences together. 1, 2, 4, 5, 7, 8, 9, 11, 12, 14.

[00:01:45]3, 6, 10, 13, 16, 20, 23, 27.

[00:01:49]Every positive integer appears in exactly one of these two sequences. Not zero, not two, exactly one.

[00:01:58]This is Beatty's theorem. It works for every positive irrational alpha greater than one and the corresponding beta.

[00:02:05]The two sequences together partition the positive integers perfectly with no overlaps and no gaps forever. Let's be precise about what's happening.

[00:02:16]Given an irrational number alpha bigger than one, we define beta by the equation 1 over alpha plus 1 over beta equals 1.

[00:02:25]This means beta equals alpha divided by alpha minus 1.

[00:02:30]If alpha is irrational, beta is also irrational. If alpha is between 1 and infinity, beta is also between 1 and infinity.

[00:02:39]The relationship is symmetric. If you start with beta and compute the corresponding partner, you get back alpha. Now we form two sequences, one indexed by alpha and one by beta. The alpha sequence is S sub alpha equals the set of floor of n times alpha for n in the positive integers.

[00:02:59]The beta sequence is S sub beta, the set of floor of n times beta for n in the positive integers. Floor just means round down.

[00:03:08]Floor of 2.7 equals 2. Floor of 1.1 equals 1.

[00:03:13]Floor of an integer equals itself, but since alpha and beta are irrational, n times alpha is never an integer for n positive, so we don't worry about that case. Beatty's theorem asserts two things simultaneously.

[00:03:28]First, the two sequences are disjoint.

[00:03:31]No positive integer appears in both S sub alpha and S sub beta.

[00:03:36]Second, the two sequences cover everything.

[00:03:40]Every positive integer appears in at least one of them. Together these two facts say that S sub alpha and S sub beta partition the positive integers exactly. Each positive integer is in exactly one sequence.

[00:03:53]The partition is precise.

[00:03:56]For example, take alpha equals the golden ratio, 1.61803.

[00:04:01]Then beta equals the golden ratio squared, 2.61803.

[00:04:06]S sub alpha for the golden ratio is 1 3 4 6 8 9 11 12 14 16 17 19.

[00:04:15]S sub beta is 2 5 7 10 13 15 18 20 23 26. Combine them in order. 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 and so on. Every positive integer, exactly once.

[00:04:34]This is a partition algorithm encoded in the irrationality of alpha. The geometric picture is striking. Imagine a number line stretching out to the right with every positive integer marked as a small notch.

[00:04:48]Now imagine two combs sitting above the number line. The first comb has teeth at positions 1 2 4 5 7 8 9. The second comb has teeth at 3 6 10.

[00:05:02]The teeth of the first comb cover certain positions. The teeth of the second comb cover the remaining positions. Together they cover every integer. The two combs interlock perfectly with no missed positions and no overlaps. If you shifted either comb slightly, you would break the partition.

[00:05:20]If you stretched either comb slightly, you would also break the partition.

[00:05:25]The structure is rigid. It only works for the precise irrational alpha you started with. Change alpha to a different irrational and you get a different partition.

[00:05:35]But the partition is always a clean split into two combs. The combs themselves change. The two combs structure does not. This raises an obvious question. Can we use three combs instead of two? Three irrationals alpha, beta, gamma satisfying 1 over alpha plus 1 over beta plus 1 over gamma equals 1.

[00:05:56]With three Beatty style sequences partitioning the positive integers, the answer is no.

[00:06:02]For three or more irrationals, Beatty's theorem fails. There is no irrational triple such that the three Beatty sequences partition the positive integers.

[00:06:12]This is a theorem due to Aviezri Fraenkel from the 1970s.

[00:06:16]He conjectured a more general result called Fraenkel's conjecture about which combinations of sequences can cover the integers exactly, but for irrational Beatty sequences specifically, two is the magic number. The two combs partition is structurally unique. You cannot do it with three combs. The proof of Beatty's theorem is surprisingly short.

[00:06:37]The key insight is a counting argument.

[00:06:40]How many elements of S sub alpha are less than or equal to some integer n?

[00:06:46]Well, S sub alpha consists of floor of n times alpha for n in the positive integers. The condition floor of n times alpha less than or equal to n is equivalent to n times alpha less than n plus 1, which is equivalent to n less than n plus 1 divided by alpha. So, the number of elements of S sub alpha less than or equal to n is the floor of n plus 1 divided by alpha. Let's call that quantity A of n.

[00:07:12]Similarly, the number of elements of S sub beta less than or equal to n is the floor of n plus 1 divided by beta. Call this B of n.

[00:07:22]Beatty's theorem says A of n plus B of n equals n exactly. The total count of elements from both sequences below or equal to n is n. Misha, every integer below or equal to n is in exactly one sequence.

[00:07:36]The verification of A of n plus B of n equals n uses the relationship 1 over alpha plus 1 over beta equals 1.

[00:07:45]Multiplying through by n + 1 gives n + 1 / alpha + n + 1 / beta = n + 1.

[00:07:53]Floor of n + 1 / alpha + floor of n + 1 / beta is essentially n + 1 - the fractional parts. Since alpha and beta are irrational, the fractional parts are non-zero and their sum works out to exactly 1.

[00:08:10]So, the two floors sum to n + 1 - 1, which is n.

[00:08:15]That's the proof, a counting argument and the magic of 1/alpha + 1/beta = 1.

[00:08:22]The cleanliness of this argument is part of why Beatty's theorem feels almost too good to be true.

[00:08:28]A simple algebraic identity controls the partition of the integers into two interlocked sequences for any irrational, but there are subtleties hidden in the irrationality assumption.

[00:08:39]The If alpha were rational, the theorem would fail.

[00:08:44]Suppose alpha = 3/2, then beta = 3.

[00:08:48]Both rational. The sequence S sub alpha would be floor of 1 * 1.5, floor of 2 * 1.5, floor of 3 * 1.5, and so on. 1 3 4 6 7 9. The sequence S sub beta would be 3 6 9 12.

[00:09:07]But 3 is in both, so is 6, so is 9. The two sequences overlap at every multiple of 3.

[00:09:14]They don't partition the integers, they double count.

[00:09:18]The reason is that when alpha is rational, n * alpha can land exactly on an integer for certain n. The floor function does nothing in that case, and the overlap can occur. If alpha is [clears throat] irrational, n * alpha is never an integer for any positive n.

[00:09:35]The fractional part of n * alpha is always strictly positive. This strict positivity is exactly what makes the floors of n * alpha and m * beta disjoint and complete. So, irrationality is not a technical assumption. It is the structural reason the partition works.

[00:09:54]The randomness of the fractional parts, guaranteed by irrationality, makes the two sequences avoid each other and cover everything. This is a beautiful instance of irrational numbers doing something that rational numbers cannot.

[00:10:07]The integers, the most discrete object in mathematics, are partitioned cleanly by sequences indexed by an irrational number. The irrationality is required.

[00:10:17]The rationals cannot do it. It's one of the surprising places where the continuum and the integers meet in a way that produces a clean discrete structure that depends essentially on the continuous. The most famous example uses the golden ratio.

[00:10:31]Let alpha equal phi, the golden ratio.

[00:10:34]Phi is approximately 1.61803.

[00:10:38]It's irrational. Phi satisfies the equation phi squared equals phi plus one.

[00:10:43]The corresponding beta is one over one minus one over phi. After simplification, beta equals phi squared, which is approximately 2.61803.

[00:10:54]Notice that alpha plus beta equals phi plus phi squared, which by the defining relation of the golden ratio equals 2 phi plus one, approximately 4.236.

[00:11:05]And alpha times beta equals phi times phi squared equals phi cubed, approximately 4.236.

[00:11:13]They're equal. So, in this particular case, alpha plus beta equals alpha times beta. The two Beatty sequences for the golden ratio are S sub phi 1 3 4 6 8 9 11 12 14 16 17 19 21 22 24 25 27 29 30 S sub phi squared 2 5 7 10 13 15 18 20 23 26 28 31. Together they partition the positive integers. These two sequences have a name. They're called the Wythoff sequences. They appear in the analysis of a game called Wythoff's game, a two-pile Nim variant from 1907.

[00:11:54]In the game, the losing positions, the positions from which the player to move will lose, given optimal play, are exactly the pairs of numbers from the Wythoff sequences. Phi shows up in game theory because Phi shows up in number theory because Phi is the simplest irrational. The Wythoff sequences also appear in the structure of Fibonacci numbers, in the explicit formulas for continued fractions of the golden ratio, in the theory of substitution dynamical systems, and in the recursive structure of the Stern-Brocot tree. The golden ratio's Beatty pair is a hub. Many connections in number theory, game theory, and combinatorics root through it. Other choices of alpha give other special partition pairs.

[00:12:35]Alpha equals square root of 2 gives a sequence 1 2 4 5 7 8 9 11 12 14. The partner sequence is 3 6 10 13 16 20 23.

[00:12:48]Each appears in number theory contexts that I won't list exhaustively, but they are studied. Alpha equals e, e is the base of natural logarithms, gives the Beatty sequence 2 5 8 10 13 16 19 21, and the partner 1 3 4 6 7 9 11 12.

[00:13:06]These sequences are related to the distribution of e's continued fraction coefficients.

[00:13:12]Alpha equals pi minus 2 gives a Beatty pair with no particular fame, but the partition is just as clean. The space of all Beatty partitions is parameterized by irrationals greater than 1. There is one Beatty partition for each irrational.

[00:13:28]The space is uncountably infinite.

[00:13:30]We have in this single theorem an uncountable family of partitions of the positive integers, all sharing the same simple structure. Compare this to the space of partitions of the positive integers in general.

[00:13:43]That space is also uncountable.

[00:13:45]But most partitions in it have no closed-form description. They are abstract objects.

[00:13:52]The Beatty partitions are different.

[00:13:54]Each one has a simple closed-form description in terms of a single irrational number.

[00:13:59]You could imagine a function from irrationals bigger than one to partitions of the positive integers into two sequences.

[00:14:06]The function is injective. Different irrationals produce different partitions. So, the Beatty partitions embed the irrationals into the space of integer partitions in a way that respects the algebraic structure of both.

[00:14:20]This is a deep but quiet result. It is not famous, but it is one of the cleanest connections between continuous mathematics, the irrationals, and discrete mathematics, the integers, that mathematics has. Sam Beatty published his theorem in 1926 as a problem in the American Mathematical Monthly.

[00:14:38]He was a Canadian mathematician working at the University of Toronto.

[00:14:43]The problem was simple to state, and his proof, essentially the counting argument I sketched earlier, was elegant. The problem and its proof became one of the most discussed problems in the early history of the Monthly.

[00:14:55]Within a few years, it had been generalized, restated, and connected to other parts of number theory. Lord Rayleigh, the same Rayleigh of Rayleigh scattering in physics, had stated a closely related result in 1894 in his book on the theory of sound.

[00:15:12]Whether Rayleigh's version or Beatty's came first is a matter of academic interest. The name Beatty's theorem stuck because Beatty's 20th century proof was cleaner, and the monthly's problem framing made it widely accessible.

[00:15:26]In the century since, Beatty sequences have appeared in combinatorial game theory, Wythoff's game, the theory of continued fractions, dynamical systems on the interval, substitution sequences, the structure of the Stern-Brocot tree, the analysis of low discrepancy sequences in number theory, and the design of certain pseudo-random number generators. The theorem itself is short, its consequences are wide. The most beautiful consequence, in some sense, is the geometric picture. Two combs, indexed by an irrational, partitioning the integers perfectly forever.

[00:16:02]The fact that this can happen at all is a small miracle of arithmetic. The fact that it happens for every irrational greater than one is the larger miracle.

[00:16:13]You pick the number, the integers split themselves accordingly, every time, every irrational.

[00:16:20]No exceptions, no errors, forever. The integers have many such hidden structures, patterns waiting to be discovered, partitions waiting to be defined, sequences waiting to be enumerated. Beatty's theorem is one of the simplest. It is also one of the most beautiful. What other patterns hide in the integers waiting for an irrational to reveal them?

[00:16:43]Drop your favorite in the comments. We have several queued up.