Newton's sums

When faced with a set of significant numbers—such as the eigenvalues of a physical system or the roots of an equation—how do we capture their collective essence? We can describe them through their relational structure, using elementary symmetric polynomials which form the coefficients of their parent polynomial. Alternatively, we can measure their individual contributions through power sums, like the sum of their squares or cubes. This raises a crucial question: are these two distinct descriptions—one of structure, the other of power—connected? Is there a way to translate from one to the other without the laborious task of finding every single root?

This article unveils the "secret passage" that provides a resounding "yes" to this question: a set of elegant and powerful relations known as Newton's sums. These identities form a bridge between the world of polynomial coefficients and the world of power sums, offering profound theoretical insights and remarkable computational shortcuts. We will first delve into the core "Principles and Mechanisms" of these identities, deriving them and revealing their surprising connection to linear algebra. Following this, under "Applications and Interdisciplinary Connections," we will embark on a tour to witness how this single algebraic concept resonates across an astonishing range of fields, from engineering and physics to pure mathematics, demonstrating its status as a truly fundamental pattern in science.

Imagine you are given a collection of numbers. Perhaps they are the resonant frequencies of a violin string, the energy levels of an atom, or simply the roots of some polynomial equation that a physicist has scrawled on a blackboard. How would you describe the character of this collection as a whole? You might try two very different approaches.

One approach is to look at how the numbers interact with each other in a collective, democratic way. You could sum them all up. You could sum up all possible products of pairs. You could sum up all products of triplets, and so on, until you get to the final grand product of all the numbers together. If our numbers are $\alpha_1, \alpha_2, \ldots, \alpha_n$, these sums are what mathematicians call the ​​elementary symmetric polynomials​​, denoted by $e_k$.

$e_1 = \sum_i \alpha_i$ $e_2 = \sum_{i \lt j} \alpha_i \alpha_j$ ... $e_n = \alpha_1 \alpha_2 \cdots \alpha_n$

These $e_k$ building blocks are profoundly important because they are precisely the coefficients of the polynomial that has our numbers as its roots (up to a plus or minus sign). The polynomial $P(x) = (x-\alpha_1)(x-\alpha_2)\cdots(x-\alpha_n)$ expands to $P(x) = x^n - e_1 x^{n-1} + e_2 x^{n-2} - \cdots + (-1)^n e_n$. So, the $e_k$'s give us a "social" or "relational" view of our numbers.

But there is another, more direct approach. Instead of looking at their combinations, we could look at their individual "powers" and sum those. We can sum the numbers themselves. We can sum their squares. We can sum their cubes. These are called the ​​power sums​​, denoted by $p_k$.

$p_1 = \sum_i \alpha_i$ $p_2 = \sum_i \alpha_i^2$ $p_3 = \sum_i \alpha_i^3$ ... and so on.

The power sums give us a different kind of information, focusing on the distribution of "magnitude" or "strength" within our set of numbers.

Now, a physicist, or any curious person, should immediately ask: are these two descriptions related? They seem to be measuring different aspects of the same collection of numbers. The elementary polynomials ($e_k$) are about combination and structure, while the power sums ($p_k$) are about raw power. Is there a secret passage between them? If I know the polynomial's coefficients, can I figure out the sum of the cubes of its roots? Conversely, if I have measured the first few power sums, can I reconstruct the polynomial's coefficients?

The answer is a resounding yes, and the secret passage is a set of wonderfully elegant and powerful relations known as ​​Newton's sums​​ or ​​Newton's identities​​.

These identities are not some arcane bolt from the blue; they arise from a simple, beautiful observation. Let's see if we can discover the first few ourselves.

The very first power sum, $p_1 = \sum \alpha_i$, is, by definition, identical to the first elementary symmetric polynomial, $e_1$. So, our first identity is trivial:

$p_1 = e_1$

Now for the second one. How can we relate $p_2 = \sum \alpha_i^2$ to $e_1$ and $e_2$? Let's try playing with what we know. What happens if we square $e_1$?

$e_1^2 = (\alpha_1 + \alpha_2 + \cdots + \alpha_n)^2$

When you expand this, you get two kinds of terms: terms where a root is multiplied by itself ($\alpha_i^2$), and terms where two different roots are multiplied ($\alpha_i \alpha_j$ for $i \neq j$).

$e_1^2 = (\alpha_1^2 + \alpha_2^2 + \cdots + \alpha_n^2) + 2(\alpha_1 \alpha_2 + \alpha_1 \alpha_3 + \cdots)$

Look closely! The first part is just our power sum $p_2$. The second part is exactly twice the second elementary symmetric polynomial, $2e_2$. So we have found:

$e_1^2 = p_2 + 2e_2$

Rearranging this gives us the second Newton's identity: $p_2 - e_1 p_1 + 2e_2 = 0$. We've built a bridge! If you know $e_1$ and $e_2$ (from a polynomial's coefficients), you can find $p_2$. If you know $p_1$ and $p_2$ (from some measurements), you can find $e_2$.

This pattern continues. One can derive a whole ladder of these identities, each one connecting the next power sum $p_k$ to the previous ones and the elementary polynomials:

$p_1 - e_1 = 0$ $p_2 - e_1 p_1 + 2e_2 = 0$ $p_3 - e_1 p_2 + e_2 p_1 - 3e_3 = 0$ ...and so on.

The remarkable feature is their ​​recursive​​ nature. You can use them to climb a ladder, computing higher and higher power sums one step at a time, armed only with the polynomial's coefficients. It allows for astonishingly quick calculations of quantities like the sum of the sixth powers of a polynomial's roots, a task that would be nightmarish to do by finding the roots themselves.

This is all very neat as a piece of algebra, but where does it show up in the real world? Here, we stumble upon one of those moments of delightful surprise that make science so rewarding. The abstract world of polynomial roots has a stunningly direct incarnation in the very concrete world of ​​linear algebra​​ and physics.

Consider an $n \times n$ matrix, $A$. In quantum mechanics, such matrices might represent physical observables like energy or momentum. The fundamental numbers that characterize such a matrix are its ​​eigenvalues​​, let's call them $\lambda_1, \lambda_2, \ldots, \lambda_n$. These eigenvalues are, in essence, the "roots" of the matrix. They answer the question: in which directions does the matrix act simply by stretching or shrinking vectors?

The elementary symmetric polynomials $e_k$ of these eigenvalues appear as the coefficients of the matrix's ​​characteristic polynomial​​, $P(\lambda) = \det(\lambda I - A)$. This polynomial is fundamental; its roots are the eigenvalues.

What about the power sums, $p_k = \sum_i \lambda_i^k$? It turns out that they correspond to something very easy to calculate: the ​​trace​​ of the matrix powers! The trace of a matrix is the sum of its diagonal elements, a simple piece of bookkeeping. It's a fundamental fact of linear algebra that:

$\text{tr}(A^k) = \sum_{i=1}^n \lambda_i^k = p_k$

Suddenly, our abstract algebraic identities are transformed into a powerful computational tool. The two sides of our coin are now:

• ​​Elementary Polynomials ($e_k$)​​: The coefficients of the characteristic polynomial. Hard to compute in general because they involve a determinant.
• ​​Power Sums ($p_k$)​​: The traces of the powers of the matrix, $\text{tr}(A^k)$. Easy to compute; just matrix multiplication and summing a few numbers.

Newton's identities are the bridge! They tell us that you can determine the characteristic polynomial of a matrix—a deep fact about its geometry—simply by calculating the traces of its first few powers. Conversely, if a theorist gives you the characteristic polynomial of some system, you can immediately tell them the trace of any power of the matrix, say $A^4$, without ever writing down the matrix itself! This is an incredible shortcut, a testament to the hidden unity between different mathematical concepts.

The story doesn't end here. The robustness of Newton's identities hints at an even deeper, more general structure.

For one, these identities are not picky about their numbers. They work just as beautifully for complex numbers as they do for real numbers. More surprisingly, they even hold in the strange, finite number systems known as ​​finite fields​​, which form the bedrock of modern cryptography and coding theory. This shows that the relationship between $e_k$ and $p_k$ is a fundamental law of algebra, not just a property of our familiar number line.

Furthermore, the elementary polynomials are not the only game in town. There are other families of symmetric polynomials, like the ​​complete homogeneous symmetric functions​​ ($h_k$), which are sums of all monomials of a certain degree. It turns out that they, too, are connected to the power sums by a nearly identical set of Newton-like identities. It's as if we've discovered a family of related species, all sharing a common ancestor. This hints at the existence of a vast, interconnected continent of mathematical objects—the ring of symmetric functions—governed by these elegant laws.

This web of relationships is so rigid and predictive that it's possible to write down a single, explicit formula for any $e_k$ in terms of the power sums $p_1, \ldots, p_k$ using a determinant. We don't need to go through the recursive steps; we can, in principle, write down the answer in one fell swoop. This ​​determinantal formula​​ is a bit like Newton's law of universal gravitation—a single, compact statement from which all the messy, step-by-step calculations of planetary orbits can be derived. It's the ultimate proof that beneath the surface of complex calculations lies a simple, powerful, and beautiful organizing principle.

We have played with these remarkable identities, the Newton's sums, and seen how they work their magic, connecting the sums of the powers of a set of numbers to the symmetric combinations of those same numbers. At first glance, this might seem like a clever but niche algebraic trick. A cute puzzle for mathematicians. But the truth is something far more wonderful. This relationship is not just a curiosity; it is a deep pattern that echoes throughout science and mathematics. It's as if we've found a universal tuning fork, and by striking it, we can hear its resonance in the most unexpected corners of human knowledge. Let's go on a tour and listen for these echoes.

Our first stop is the world of the tangible, of systems that change and materials that stretch. In physics and engineering, we often describe the state of a system using a collection of numbers arranged in a square grid called a matrix. This matrix can represent anything from the connections in a circuit to the stresses inside a steel beam. A key question is always: what are the fundamental modes of this system? These are its "eigenvalues," a set of special numbers that act as the system's DNA. The coefficients of a special polynomial, the "characteristic polynomial," are built from simple combinations of these eigenvalues—like the sum, the sum of products in pairs, and so on. These are the elementary symmetric polynomials, the $e_k$ from our previous discussion.

Now, suppose we want to know something about the system's behavior over time, or its response to energy. This often involves calculating the "trace" of the matrix raised to some power, like $\operatorname{tr}(A^2)$ or $\operatorname{tr}(A^3)$. The trace is simply the sum of the elements on the matrix's main diagonal, but it has a deeper meaning: it's also the sum of all the eigenvalues. So, $\operatorname{tr}(A^k)$ is the sum of the $k$-th powers of the eigenvalues—our old friend, the power sum $p_k$. And right there, the connection clicks into place! Newton's sums provide a direct, elegant bridge. If we know the system's characteristic polynomial (the $e_k$), we can instantly calculate the trace of any power of its matrix (the $p_k$) without the brutish labor of actually multiplying the matrix by itself over and over again. It’s a spectacular shortcut, a piece of mathematical magic that exchanges sweat for insight.

This isn't just for abstract matrices. Let's get our hands dirty. Imagine you're an engineer studying how a piece of rubber deforms when you stretch it. The deformation is described by a tensor—a sort of matrix for physicists—called the Cauchy-Green tensor, $\boldsymbol{C}$. Its eigenvalues tell you the squared amount of stretch in three principal directions. The fundamental properties of the material's response to deformation are captured by "invariants," quantities that don't change no matter how you rotate the material. Two of these, $I_1$ and $I_2$, are none other than the elementary symmetric polynomials of the eigenvalues. Other important physical quantities, however, might depend on the sum of the squares of the eigenvalues, which is the trace of the tensor squared, $\operatorname{tr}(\boldsymbol{C}^2)$. How do these different measures of strain relate? You guessed it. Newton's sums provide the Rosetta Stone, giving a simple formula like $\operatorname{tr}(\boldsymbol{C}^2) = I_1^2 - 2I_2$, directly linking these fundamental physical quantities. The abstract algebra of polynomial roots is, it turns out, written into the very laws of elasticity.

Let's leave the physical world for a moment and wander into the purer, crystalline world of mathematics. Surely this pattern must break down? On the contrary, it gets louder. Consider the famous Fibonacci sequence: 1, 1, 2, 3, 5, 8... A related sequence is the Lucas numbers: 1, 3, 4, 7, 11... which follows the same "add the last two" rule but starts differently. These sequences seem to be defined by their history, each number born from its parents. But there's another, hidden way to see them. It turns out the Lucas numbers are exactly the power sums ($p_k$) of the roots of the simple polynomial $x^2 - x - 1 = 0$. How could we possibly know that? By running Newton's identities in reverse! Given the sequence of power sums (the Lucas numbers $p_1=1, p_2=3, \dots$), the identities allow us to solve for the coefficients of the parent polynomial. It’s like finding the genetic code of the sequence.

The tune continues in other fields. Take graph theory, the study of networks. A network of friends, a molecule, or the internet can be drawn as a graph of nodes and edges. We can encode this drawing in an adjacency matrix, $A$. The power sums $p_k = \operatorname{tr}(A^k)$ have a wonderfully intuitive meaning here: they count the number of ways you can take a walk of $k$ steps along the network's edges and end up back where you started. Newton's sums then build a bridge between this tangible act of "walking on a graph" and the graph's fundamental algebraic properties, expressed by the elementary symmetric polynomials of its eigenvalues.

This universal theme also plays out in the study of special functions, the celebrity functions of mathematical physics. Polynomials named after titans like Legendre and Chebyshev pop up when solving equations for gravity, electromagnetism, and wave motion. The locations of their roots are not random; they have a deep internal structure. For example, the roots of Legendre polynomials are the optimal points to use for a powerful numerical integration technique called Gaussian quadrature. To understand the collective properties of these important roots, must we calculate each one? No. We can use Newton's sums to compute the sum of their squares, cubes, or any other power, directly from the polynomial's coefficients, giving us profound insight into their distribution and average behavior without a fuss.

By now, you might be sensing the power of this idea. But we have only been walking in the foothills. The true mountains lie ahead, in the most abstract and fundamental realms of modern science. In quantum mechanics and particle physics, symmetry is everything. The language of symmetry is a beautiful subject called representation theory. It's about how to represent abstract symmetries as matrices. The "character" of a representation is its essential fingerprint, and it's calculated by taking the trace of these matrices. If you have the character for a basic symmetry operation $g$, what is the character for a more complex symmetry built from it, say its "third exterior power"? This sounds impossibly abstract, but it turns out the answer is handed to us on a silver platter by Newton's identities. The characters of powers of an operation ($g, g^2, g^3, \dots$) are the power sums $p_k$. The characters of its exterior powers are the elementary symmetric functions $e_k$. The identities provide a direct formula to get one from the other, forming a core part of the toolbox for physicists and mathematicians building models of our universe.

Let's make one final ascent. At the pinnacle of modern geometry, mathematicians study strange and beautiful objects called "vector bundles." Think of the surface of the Earth. At every point, there is a "tangent plane," the flat space of possible directions you can travel. A vector bundle is this idea generalized to more abstract spaces. A fundamental question is: how "twisted" is this bundle? Is it like the flat planes stacked over a tabletop, or is it like the twisted directions on a Möbius strip? To measure this twistedness, geometers invented "Chern classes," which are the $e_k$ in this new language. They also invented the "Chern character," which turns out to be based on the power sums $p_k$. And the universal, dictionary-like relationship between them? Once again, it's our friend, Newton's sums. The same simple algebraic rules that govern matrix traces and polynomial roots also describe the fundamental shape of these abstract geometric spaces. Even in algebraic number theory, these identities are indispensable, linking the coefficients of the polynomial that defines a number to its 'trace', a key concept in the study of number fields.

So, we have journeyed from the stretching of rubber to the structure of spacetime. In every field, we heard the same song. The relationship captured by Newton's sums is a profound truth about the nature of a whole and its parts. It shows us that if we know the elementary building blocks of a system, we can deduce a great deal about its collective behavior, and vice versa. It is a stunning example of the unity of mathematics, a simple, elegant melody that resonates through the grand orchestra of science.