Sum of Three Squares Theorem

What determines if a whole number can be written as the sum of three squares? This seemingly simple question, first explored by ancient mathematicians, leads to one of the most elegant results in number theory. While numbers like 3 ($1^2+1^2+1^2$) and 6 ($2^2+1^2+1^2$) are easily formed, others like 7 and 15 stubbornly resist. This isn't a lack of effort; it's a sign of a deep, hidden law governing the very structure of integers. This article unravels this mathematical mystery, addressing the knowledge gap between simple observation and profound theoretical understanding.

Across the following chapters, you will embark on a journey to understand this beautiful piece of mathematics. In "Principles and Mechanisms," we will uncover the complete rule, known as Legendre's three-square theorem, by exploring the worlds of modular arithmetic and the powerful local-global principle. Following this, the "Applications and Interdisciplinary Connections" chapter will reveal the theorem's surprising and far-reaching impact, demonstrating how a pure number theory concept provides a blueprint for computer algorithms, dictates the laws of quantum physics, and opens a gateway to the unified landscape of modern mathematics.

So, we've been introduced to a seemingly simple question: which whole numbers can be written as the sum of three squares? Some numbers seem to jump at the chance. For example, $1=1^2+0^2+0^2$, $2=1^2+1^2+0^2$, and $3=1^2+1^2+1^2$. It feels like we might be able to get all of them this way. But let's be good physicists—or in this case, mathematicians—and test the idea. Let's try to make the number 7. We can use $1^2=1$ and $2^2=4$. We could try $1+1+1=3$, or $4+1+1=6$, or $4+4+1=9$. No matter how we arrange them, we can't hit 7. Is this just because we haven't tried hard enough, or is there a deeper reason? Is there some hidden law preventing 7 from being the sum of three squares?

It turns out there is a hidden law, a remarkably simple and elegant one. To see it, we don't need to look at the numbers themselves, but at their "flavor" or character when we divide them by 8. This is the world of modular arithmetic, a powerful lens for uncovering hidden patterns in numbers.

Let's take any integer, square it, and see what remainder it leaves when we divide by 8.

• If the number is even, say $2k$: Its square is $4k^2$. If $k$ is also even ($k=2m$), the square is $4(4m^2) = 16m^2$, which leaves a remainder of $0$. If $k$ is odd ($k=2m+1$), the square is $4(2m+1)^2 = 4(4m^2+4m+1) = 16(m^2+m)+4$, which leaves a remainder of $4$.
• If the number is odd, say $2k+1$: Its square is $(2k+1)^2 = 4k^2+4k+1 = 4k(k+1)+1$. Now, a wonderful little fact is that the product of any two consecutive integers, $k(k+1)$, is always even. So, $4k(k+1)$ must be a multiple of 8. This means any odd square leaves a remainder of exactly $1$.

So, here is the first rule of our new game: ​​any perfect square, when divided by 8, must leave a remainder of 0, 1, or 4​​. There are no other possibilities.

Now, what happens when we add three of these squares together? We are just adding three numbers, where each number must be a 0, a 1, or a 4 in this "world of 8". What possible totals can we get?

• $0+0+0 = 0$
• $1+0+0 = 1$
• $1+1+0 = 2$
• $1+1+1 = 3$
• $4+0+0 = 4$
• $4+1+0 = 5$
• $4+1+1 = 6$
• $4+4+0 = 8 \equiv 0$
• $4+4+1 = 9 \equiv 1$
• $4+4+4 = 12 \equiv 4$

Listing them all out, we find that the sum of three squares, modulo 8, can only be $0, 1, 2, 3, 4, 5,$ or $6$.

Look at what's missing! The number 7. It's impossible. We have discovered a fundamental obstruction: ​​a sum of three integer squares can never be congruent to 7 modulo 8​​. Our failed attempt to write 7 as a sum of three squares wasn't a failure of imagination; it was an impossibility dictated by the very structure of numbers. The same goes for 15, 23, 31, and any other number of the form $8b+7$. They are all filtered out by this "Sieve of Eight."

So, we have a powerful rule: if a number is of the form $8b+7$, it's out. Is that the whole story? If a number is not of the form $8b+7$, is it always a sum of three squares?

Let's test this hypothesis. Consider the number $n=28$. When we divide 28 by 8, we get a remainder of 4 ($28 = 3 \times 8 + 4$). So, 28 passes our Sieve of Eight. Our rule suggests it should be a sum of three squares. But try as you might, you won't find one. What have we missed?

The culprit is the number 4. Let's suppose a number $n = x^2+y^2+z^2$ is divisible by 4. This means $x^2+y^2+z^2 \equiv 0 \pmod 4$. Remember how squares behave, this time modulo 4: even squares are congruent to 0, and odd squares are congruent to 1. The only way for three numbers, each being 0 or 1, to sum to a multiple of 4 is if they are all 0.

So, $x^2 \equiv 0 \pmod 4$, $y^2 \equiv 0 \pmod 4$, and $z^2 \equiv 0 \pmod 4$. This in turn means that $x, y,$ and $z$ must all be even integers. We can write them as $x=2x'$, $y=2y'$, and $z=2z'$. Let's substitute this back into our equation:

$n = (2x')^2 + (2y')^2 + (2z')^2 = 4(x'^2+y'^2+z'^2)$

This reveals something beautiful. If a multiple of 4, $n$, is a sum of three squares, then the number $n/4$ must also be a sum of three squares. It's as if the property of "being a sum of three squares" is inherited downwards when we divide by 4. This gives us a powerful descent mechanism.

Now we can solve the mystery of 28. If 28 were a sum of three squares, then $28/4 = 7$ would also have to be. But we already know that 7 is forbidden by our Sieve of Eight! The trail goes cold. Therefore, 28 cannot be a sum of three squares.

The French mathematician Adrien-Marie Legendre combined these two insights—the Sieve of Eight and the Descent of Four—into one of the most elegant classification theorems in number theory.

​​Legendre's Three-Square Theorem:​​ A positive integer $n$ can be written as the sum of three squares if and only if it is not of the form $4^a(8b+7)$ for any non-negative integers $a$ and $b$.

This compact formula is the complete law. It tells us to first "peel away" all the factors of 4 from our number $n$. If the core that's left over is of the forbidden form $8b+7$, then the original number $n$ cannot be a sum of three squares. Otherwise, it can.

Let's see this in action with a more complex number, $n=320$. We can write $320 = 64 \times 5 = 4^3 \times 5$. Here, $a=3$. The "core" of the number is 5. Is 5 of the form $8b+7$? No, $5 \equiv 5 \pmod 8$. So, Legendre's law predicts that 320 can be written as a sum of three squares. (Indeed, $320 = 16^2 + 8^2 + 0^2$). Furthermore, the descent argument tells us something deeper. Since we had to peel away $4^3$, any solution $(x,y,z)$ for 320 must come from a solution for 5. This forces $x, y,$ and $z$ to all be multiples of $2^3=8$, meaning that $\gcd(x,y,z)$ must be a multiple of 8. For $n=320$, the $\gcd$ is always exactly 8, a stunning consequence of this structure.

This might all seem a bit complicated. Why are there forbidden numbers for three squares, when Lagrange's Four-Square Theorem famously states that every positive integer is a sum of four squares? And Fermat's theorem on sums of two squares has its own, different-looking rule. Why is three the difficult number?

The reason lies in a broken symmetry. For sums of two squares and sums of four squares, there exist beautiful algebraic identities that show that the product of two such numbers is again a number of the same type. For two squares, the identity is $(a^2+b^2)(c^2+d^2) = (ac-bd)^2+(ad+bc)^2$. For four squares, a more complex identity was discovered by Euler. These identities mean that the sets of numbers representable as sums of two or four squares are closed under multiplication.

But for three squares, this is not true. The set is not closed under multiplication. Consider the numbers 3 and 5. $3 = 1^2 + 1^2 + 1^2$ $5 = 2^2 + 1^2 + 0^2$ Both are sums of three squares. What about their product, $15$? As we know, $15 = 8(1)+7$, so $15 \equiv 7 \pmod 8$. It falls into the forbidden category and cannot be written as a sum of three squares. This lack of a multiplicative identity is the deep reason for the complexity and beauty of the three-square problem.

There is one final, modern way to view this entire story, which puts Legendre's remarkable discovery in an even grander context. It's called the ​​local-global principle​​.

The idea is to ask whether an equation has integer solutions (a "global" property) by checking if it has solutions in various "local" number systems. These local systems include the familiar real numbers, $\mathbb{R}$, but also the strange and powerful $p$-adic numbers, $\mathbb{Z}_p$, one for each prime $p$. If we can't find a solution in even one of these local systems, then we certainly can't find an integer solution.

What does this mean for our problem, $n = x^2+y^2+z^2$?

• ​​The real numbers ($\mathbb{R}$):​​ For a solution to exist, $n$ must be non-negative. This is an obvious local condition.
• ​​The $p$-adic numbers for odd primes $p$:​​ It turns out that for any odd prime $p$, a solution always exists. There are no obstructions here.
• ​​The 2-adic numbers ($\mathbb{Z}_2$):​​ This is where the action is. The condition that a solution exists in the 2-adic integers is exactly that $n$ is not of the form $4^a(8b+7)$. Our Sieve of Eight and Descent of Four are just the elementary shadows of this deeper 2-adic structure.

So, the only places an obstruction can arise are in the real numbers (if $n 0$) and in the 2-adic numbers. For the sum of three squares, a magnificent thing happens: this is the whole story. If a number $n$ has a representation locally everywhere (i.e., in $\mathbb{R}$ and in every $\mathbb{Z}_p$), then it is guaranteed to have a global integer representation. This local-to-global principle does not hold for all equations, but its truth here is what makes Legendre's theorem so complete. It shows that the simple congruence rules we found are not just happy coincidences; they are the complete and only barriers to a number being a sum of three squares. This is why a purely geometric argument, like one from Minkowski's theorem, cannot produce a solution for $n=7$; it operates in the "global" world of $\mathbb{R}^3$ and is blind to the "local" obstruction that lives in the 2-adic world. The arithmetic is absolute.

Now that we have grappled with the principles behind which numbers can be written as a sum of three squares, we might be tempted to ask, "So what?" Is this just a charming little puzzle for mathematicians, a curious property of numbers with no bearing on the world at large? The answer, you will be delighted to find, is a resounding no. The story of three squares does not end in the quiet halls of number theory. Instead, it bursts forth, its tendrils reaching into the practical realm of computation, the fundamental structure of our physical reality, and the deepest, most abstract vistas of modern mathematics. This is where the journey gets truly exciting, for we are about to witness the unexpected unity of seemingly disparate ideas.

Let's begin with the most direct application. Imagine you have a computer, and you want to teach it how to answer our question: is a given number $n$ a sum of three squares? You could, of course, tell the machine to start trying all possible combinations of integers $x, y, z$. But this is a terribly inefficient, brute-force approach. For a large number $n$, the search would be immense.

This is where the beauty of Legendre's theorem shines. As we saw in the previous chapter, a number fails to be a sum of three squares if and only if it falls into the very specific family of numbers of the form $4^a(8b+7)$. This isn't just a theoretical curiosity; it's a blueprint for an astonishingly efficient algorithm. To check any number $n$, no matter how large, we don't need to search at all. We simply perform a couple of quick arithmetic operations: first, we repeatedly divide $n$ by $4$ until we can't anymore, and then we check if the remainder of the result when divided by $8$ is $7$. If it is, the answer is no. Otherwise, the answer is yes. This is a perfect example of how deep theoretical insight leads to practical, powerful computation.

Of course, this decision algorithm doesn't tell us what the squares are, only that they exist. If we need to find an actual representation, we might need to search. But even here, number theory provides clever shortcuts. For instance, we could search for just one square, say $z^2$, and then check if the remainder, $n-z^2$, can be written as a sum of two squares—a question governed by its own beautiful set of rules, Fermat's two-square theorem. We can even design clever "sieves" to quickly discard unpromising candidates, dramatically speeding up the search. The abstract structure of numbers provides a map, guiding our computational journey through what would otherwise be a vast, dark wilderness.

The connections now take a leap from the logical to the physical. One of the simplest, yet most important, systems in all of quantum mechanics is the "particle in a box." Imagine a tiny particle, like an electron, trapped inside a perfect cube. According to the laws of quantum mechanics, the particle cannot have just any energy it pleases. Its energy is "quantized"—restricted to a discrete set of allowed levels, like the notes on a piano. The formula for these allowed energies, derived from the Schrödinger equation, contains a wonderful surprise. The energy of a state is proportional to a simple quantity:

$E \propto n_x^2 + n_y^2 + n_z^2$

Here, $n_x$, $n_y$, and $n_z$ are positive integers that label the quantum state. And there it is, right at the heart of a fundamental physical model: the sum of three squares! The allowed energy levels of a particle in a box are not random; their spectrum is governed by the arithmetic of our very problem.

This has immediate and profound consequences. Since the energy is determined by the integer $N = n_x^2+n_y^2+n_z^2$, we must ask: can $N$ be any integer? The answer is no! We know from Legendre's theorem that integers of the form $4^a(8b+7)$ can never be written as a sum of three squares. Therefore, a particle in a cubic box can never have an energy corresponding to $N=7$, $N=15$, or $N=28$, among infinitely many others. These energy levels are fundamentally forbidden, not by some physical force, but by a simple truth of arithmetic. There are silent gaps in the "music" of the cubic box, gaps that sing of the structure of numbers.

The connection deepens when we consider ​​degeneracy​​. It's possible for different combinations of quantum numbers $(n_x, n_y, n_z)$ to give the same total energy. For example, the states $(1,1,2)$, $(1,2,1)$, and $(2,1,1)$ all have $N=1^2+1^2+2^2=6$. They are distinct quantum states, but they have the exact same energy. The number of such states for a given energy level $N$ is called its degeneracy, $D(N)$. This physical degeneracy is precisely the number of ways the integer $N$ can be written as an ordered sum of three positive squares. So, a question about the spectrum of a physical system becomes a question for number theorists: how many ways can you write $N$ as a sum of three squares?

Let's zoom out. Instead of looking at individual energy levels, what if we look at the big picture? What are the statistical properties of this spectrum, especially at very high energies? This is the domain of ​​quantum chaos​​, which studies the fingerprints of classical chaos in quantum systems. The particle in a box is a simple, "integrable" (non-chaotic) system, and its spectrum has a tell-tale signature: the energy levels tend to cluster together.

Why do they cluster? Because of the high degeneracies we just discussed! Many states are crowded into a single energy level. We can even ask, what is the average degeneracy as the energy gets very large? Combining the physics of counting quantum states with deep results from number theory, we find a remarkable answer. The mean degeneracy $\langle d(E) \rangle$ grows with energy, and it does so in a very specific way:

$\langle d(E) \rangle \propto \sqrt{E}$

This result arises from a beautiful argument. The total number of states up to a certain energy can be approximated by the volume of a sphere in "momentum space," a continuous, geometric concept. But the number of distinct levels depends on counting integers that are sums of three squares, a discrete, arithmetic problem. The ratio of these two quantities gives the average degeneracy.

Furthermore, number theory tells us something about the global structure of these levels. Asymptotically, as we consider larger and larger numbers, only about $5/6$ of all integers can be written as a sum of three squares. This means that, on a grand statistical scale, a full $1/6$ of the energy levels we might naively expect to see are systematically absent due to the ancient arithmetic rule of Legendre. The large-scale structure of the quantum spectrum is dictated by number theory.

Our journey ends where it began, in the world of pure mathematics, but at a far greater height. The quest to understand the number of representations $r_3(n)$ has led mathematicians to create some of the most profound and beautiful theories.

One powerful tool is the concept of a ​​generating function​​. Imagine a magical machine that encodes an entire infinite sequence of numbers into a single function. For our sequence $r_3(n)$, this function is the Jacobi theta function, raised to the third power:

$\vartheta_3(q)^3 = \left( \sum_{k=-\infty}^{\infty} q^{k^2} \right)^3 = \sum_{n=0}^{\infty} r_3(n) q^n$

This function, $\vartheta_3(q)^3$, is not just any function; it belongs to a special class called ​​modular forms​​. These are functions of a complex variable that possess an almost unbelievable amount of symmetry. They are central objects in modern number theory, acting as bridges between different areas of mathematics. Using the theory of modular forms, one can derive exact formulas for $r_3(n)$ that involve other deep arithmetic quantities called "class numbers," which measure the structure of number systems.

But the most breathtaking vista lies ahead. The theta function $\vartheta_3(q)^3$ is a modular form of "half-integral weight" ($3/2$). There is a miraculous connection, known as the ​​Shimura correspondence​​, that acts like a secret passageway, linking these half-integral weight forms to a completely different family of modular forms of integral weight (like weight 2). This correspondence is a profound duality, revealing that the arithmetic information encoded in the coefficients $r_3(n)$ is a "shadow" or an "echo" of information contained in a seemingly unrelated mathematical world.

And so, we see the full tapestry. A simple question, "Which numbers are sums of three squares?", leads us to design computer algorithms. It appears as the organizing principle for the energy levels of a quantum particle. It governs the statistical properties of chaotic systems. And finally, it serves as a gateway to the spectacular, unified world of modular forms and the deepest dualities in modern mathematics. It is a perfect illustration of the surprising, profound, and beautiful interconnectedness of all things mathematical and physical.