Subtracted dispersion relations

In physics, the principle of causality—that an effect cannot precede its cause—is not just a philosophical concept but a powerful predictive tool. When translated into the language of complex analysis, it gives rise to dispersion relations, which remarkably connect a system's absorptive properties (the imaginary part of a response function) to its reactive properties (the real part). However, this powerful tool seems to fail for many real-world systems where the response does not vanish at infinite energies, leading to meaningless divergent integrals.

This article delves into the elegant solution: subtracted dispersion relations. It addresses this critical knowledge gap by demonstrating how a simple modification not only restores predictive power but also enriches our physical understanding. The first chapter, ​​"Principles and Mechanisms,"​​ will explore the mathematical foundation of this technique, explaining why subtractions are necessary and revealing the profound physical meaning of the "subtraction constants" that arise. Following this, the second chapter, ​​"Applications and Interdisciplinary Connections,"​​ will showcase the vast utility of this tool across diverse fields, from reconstructing particle scattering amplitudes and deriving fundamental sum rules to understanding the optical properties of materials and even probing the nature of gravity.

Imagine you had an oracle. An oracle that, if you told it just one aspect of a physical process—say, how much energy it absorbs—it could tell you everything else about it. How it responds, how it scatters, how it bends light. It sounds like magic, but in physics, we have something astonishingly close. This oracle is built upon one of the most fundamental and intuitive principles of our universe: ​​causality​​. The effect cannot come before the cause. This simple truth, when translated into the precise language of mathematics, gives birth to a tool of immense power and beauty: the ​​dispersion relation​​.

Let's think about what causality means. If you strike a bell, it rings after you strike it, not before. If light passes through a piece of glass, the wave that emerges is a delayed and altered version of the wave that entered. The material simply cannot react to the light wave before it arrives. In technical terms, the response function of a system, let's call it $\chi(t)$, must be zero for any time $t 0$ if the stimulus happens at $t=0$.

This is where the magic begins. A fantastic piece of mathematics known as Titchmarsh's theorem tells us something profound. If a function is zero for all negative times, then its Fourier transform—the representation of that function in the frequency domain, let's call it $\chi(\omega)$—must have a very special property. It must be ​​analytic​​ in the entire upper half of the complex frequency plane. What does "analytic" mean? For our purposes, it's a very strong form of "smoothness." It means the function has no sharp spikes, no sudden jumps, no singularities of any kind in that region. The function is well-behaved and predictable.

So, causality in the time domain implies analyticity in the upper-half complex frequency domain. This is a cornerstone of physics. And because the function is analytic there, we can bring in the powerful machinery of complex analysis, specifically ​​Cauchy's Integral Theorem​​. This theorem allows us to relate the value of the function at any point to an integral of the function along a closed path. By choosing a clever path along the real axis and around a large semicircle in the upper half-plane, we can derive the celebrated ​​Kramers-Kronig relations​​, or more generally, dispersion relations.

In their simplest form, they look like this:

Here, $\mathcal{P}$ stands for the Cauchy Principal Value, a prescription for carefully handling the point where $\omega' = \omega$. This equation is the voice of our oracle. It tells us that the real part of our response function, $\text{Re}[\chi(\omega)]$, can be completely determined if we just know the imaginary part, $\text{Im}[\chi(\omega)]$, at all frequencies!

The two parts of a complex response function have deep physical meaning. The real part typically describes the reactive or elastic response (like the refractive index of a material), while the imaginary part describes the absorptive or dissipative response (like the absorption of light). In the world of particle scattering, this connection is made explicit by the ​​Optical Theorem​​, which states that the imaginary part of the forward scattering amplitude $f(\omega)$ is directly proportional to the total scattering cross-section $\sigma_{\text{tot}}(\omega)$—something we can go out and measure! The dispersion relation then connects this measurable absorption to the elastic part of the scattering. You measure how much stuff gets scattered away, and causality tells you how the remaining wave is bent. It's a breathtaking unification of two seemingly separate phenomena.

Furthermore, this framework is robust. Many physical systems contain resonances or damped excitations, which correspond to poles in the response function. As long as the system is stable, these poles lie in the lower half-plane, safely outside the region of analyticity required by causality. Their influence is perfectly and implicitly captured by the integral along the real axis.

There is, however, a crucial fine print in the derivation of this simple dispersion relation. The integral along the large semicircle in the complex plane must vanish. This is only guaranteed if the function $\chi(\omega)$ falls off to zero as the frequency $|\omega|$ approaches infinity.

But what if it doesn't? What if the universe is more stubborn?

Consider a piece of a polymer, like plexiglass. At low frequencies of vibration, it's flexible. But if you probe it at extremely high frequencies—equivalent to a very sharp, quick tap—it doesn't have time to flow, and it behaves like a rigid, glassy solid. Its "storage modulus" $E'(\omega)$, the real part of its response, doesn't go to zero at high frequency; it approaches a finite, non-zero "glassy modulus," $E_{\infty}$. Similarly, in some particle scattering processes, the amplitude doesn't vanish at infinite energy but instead approaches a constant value.

In these cases, our beautiful integral formula breaks down. The integral diverges, and the oracle falls silent. It seems our powerful tool has a fatal flaw. Did we make a mistake? Is causality not as powerful as we thought? The answer is no. Causality is fine, and analyticity still holds. We just need to be a little more clever.

The solution is an elegant piece of mathematical jujitsu. If our function, let's call it $F(\omega)$, doesn't go to zero at infinity but instead approaches a constant $C$, we can't apply the dispersion relation to $F(\omega)$. But what about the function $G(\omega) = F(\omega) - C$? This new function, by its very construction, does go to zero at infinity!

We can apply our trustworthy dispersion relation to $G(\omega)$:

Now, let's substitute back what $G(\omega)$ is. Since $C$ is a real constant, $\text{Re}[G(\omega)] = \text{Re}[F(\omega)] - C$ and $\text{Im}[G(\omega)] = \text{Im}[F(\omega)]$. The result is a modified, but equally powerful, formula:

This is a ​​once-subtracted dispersion relation​​. The structure is almost the same, but now there's an extra term, $C$, which is called a ​​subtraction constant​​. This constant is the value that $F(\omega)$ approaches at infinite energy. In essence, the oracle now requires a bit more information from us. Before, it could divine the entire real part from the imaginary part alone. Now, it says, "Tell me the imaginary part, and tell me the value of the real part at one point (infinity), and I will tell you the rest."

This idea can be taken even further. What if an amplitude doesn't just approach a constant, but actually grows with energy? For instance, what if $|F(s)|$ grows proportionally to the energy-squared variable $s$? Then even subtracting a constant won't help; the function still blows up at infinity.

The strategy is the same, but we apply it more aggressively. If the function grows like $s$, we might need to subtract off its behavior at low energy. For example, consider the new function $G(s) = F(s) - F(0) - sF'(0)$, which has its value and scavenger-hunt first derivative at $s=0$ subtracted off. If this function $G(s)$ now vanishes fast enough at infinity, we can write a dispersion relation for it. This results in a ​​twice-subtracted dispersion relation​​. The price we pay is that we now need to provide two pieces of information—the subtraction constants $F(0)$ and $F'(0)$—to get our prediction.

A fascinating question then arises: how do we know how many subtractions are needed? The answer comes from a high-energy detective story. The high-energy growth of a scattering amplitude is not arbitrary; it is rigorously constrained by the fundamental principles of quantum field theory. One of the most famous constraints is the ​​Froissart-Martin Bound​​, which limits how fast the total cross-section $\sigma_{\text{tot}}(s)$ can grow with energy. It states that $\sigma_{\text{tot}}(s)$ can grow no faster than $(\ln s)^2$.

Let's follow the clues. Using the Optical Theorem, a cross-section growing like $(\ln s)^2$ implies a forward scattering amplitude $|F(s)|$ that grows like $s (\ln s)^2$. To write a convergent dispersion relation, we need to divide $F(s)$ by $s^n$ and have the result vanish at infinity. For an amplitude growing like $s (\ln s)^2$, the quantity $\frac{|s (\ln s)^2|}{|s|^n}$ vanishes only if $n-1 > 0$. The smallest integer $n$ that satisfies this is $n=2$.

This is a stunning conclusion! The most extreme behavior allowed by our fundamental theory of nature corresponds precisely to needing ​​two subtractions​​. Causality and quantum theory work hand-in-hand to tell us exactly what mathematical structure to use.

At this point, you might be thinking that these subtraction constants are a bit of a nuisance, arbitrary parameters we have to plug in to make our formulas work. But the truth is far more beautiful. Subtraction constants are not bugs; they are essential features of the physics. They represent the low-energy properties of a system that, together with the high-energy absorptive behavior integrated over in the dispersion relation, determine the system's response.

Let's look at the example of an elementary particle's ​​form factor​​, $F(s)$, which describes its spatial distribution of charge. For many particles, this requires a twice-subtracted dispersion relation, with subtraction constants $F(0)$ and $F'(0)$. These are not just abstract numbers. $F(0)$ is literally the particle's total charge. And $F'(0)$ is proportional to its mean square charge radius, $\langle r^2 \rangle$. These are fundamental, measurable properties of the particle! The dispersion relation becomes a profound equation linking a particle's static properties (charge, radius) to the dynamics of its interactions across all energy scales.

In other cases, these constants can be determined by other physical principles. A hidden symmetry in a theory might force the scattering amplitude to be zero at a specific energy, which can be used to fix the value of a subtraction constant. Or, demanding that causality holds in a particularly strong way, even at infinite spacelike momentum, can provide an equation that constrains them.

Subtracted dispersion relations, therefore, represent a beautiful dialogue between low and high energies. They are a quantitative expression of how the fine details of low-energy, long-distance physics are intertwined with the broad strokes of high-energy, short-distance interactions. Far from being a mere mathematical patch, the need for subtractions reveals a deeper structure of physical reality, forcing us to recognize that a complete description of nature requires stitching together information from different energy scales into a single, cohesive, and causal tapestry.

Now that we have grappled with the mathematical machinery of subtracted dispersion relations, you might be asking, "What is all this for?" It is a fair question. It is one thing to appreciate the elegance of a mathematical tool, but it is quite another to see it in action, carving out new paths of understanding in the messy, real world. The truth is, these relations are not just a curiosity of complex analysis; they are a profound expression of causality—the simple idea that an effect cannot precede its cause. And this single principle, when dressed in the language of analytic functions, becomes one of the most powerful and versatile tools in the physicist's arsenal.

It allows us to do something that feels almost like magic: to know the whole by seeing only a part. In the context of scattering, the "whole" is the full, complex scattering amplitude $\mathcal{M}(s)$, which contains all information about an interaction. The "part" is its imaginary component, $\text{Im}\mathcal{M}(s)$, which, through the optical theorem, is related to the total cross-section—a quantity we can actually go out and measure!

So, let's take a tour through the world of physics and see how this remarkable tool helps us connect seemingly disparate phenomena, from the structure of a proton to the optical properties of glass, and even to the nature of gravity itself.

Let's start in the native land of scattering theory: particle physics. When we calculate a scattering amplitude in quantum field theory, say for two particles interacting and flying apart, the full calculation can be a technical nightmare. However, the imaginary part of the amplitude is often much simpler to compute. It represents the processes that can happen "for real"—where intermediate particles in the interaction can be created and exist, however fleetingly. A subtracted dispersion relation is then our bridge to reconstruct the full amplitude. We can take the simpler imaginary part, feed it into a dispersion integral, and out comes the real part. The subtraction constants we need are not a nuisance; they are an essential part of the story, representing the pieces of the theory we fix through renormalization, setting our baseline from known physical measurements.

This technique is especially powerful when we talk about ​​form factors​​. You can think of a form factor as a kind of "shape profile" of a particle as it interacts. A proton, for example, isn't a simple point charge; it has a rich internal structure of quarks and gluons. This structure affects how it scatters other particles. A form factor, $G(s)$, is a function of the energy-momentum transfer that describes this momentum-dependent structure. By measuring how a particle scatters at various energies, we are probing its form factor. Dispersion relations provide the theoretical glue that holds these measurements together. Knowing the absorptive part and a value of the form factor at one particular energy (a normalization point), we can use a subtracted dispersion relation to predict its value at any other energy, building a complete and consistent picture of the particle's "shape". This is how we move from a collection of experimental data points to a deep understanding of the fundamental structure of matter.

You might think that this is all well and good for high-energy physicists working at giant colliders. But the same beautiful principle of causality shows up in a place you might not expect: the science of light traveling through a material, like a piece of glass or water.

When light passes through a medium, its properties are described by a complex dielectric function, $\varepsilon(\omega) = \varepsilon'(\omega) + i\varepsilon''(\omega)$. The real part, $\varepsilon'(\omega)$, tells us about the speed of light in the material (it's related to the square of the refractive index), which governs how light bends. The imaginary part, $\varepsilon''(\omega)$, tells us how much light is absorbed by the material at a given frequency $\omega$. It's what makes a material colored or opaque.

Now, which of these two do you think is easier to measure? It's often the absorption, $\varepsilon''(\omega)$. You shine a light source of varying frequencies through a sample and simply measure how much of it gets absorbed. Measuring the refractive index $\varepsilon'(\omega)$ over a wide range of frequencies can be much more challenging.

Here is where causality steps in. The response of the material's electrons to the incoming light wave must be causal—the electrons can't jiggle before the wave hits them! This principle means that the dielectric function $\varepsilon(\omega)$ must be analytic in the upper half-plane of complex frequencies. And so, its real and imaginary parts must be linked by a dispersion relation. These are the famous ​​Kramers-Kronig relations​​. They state that if you know the absorption spectrum $\varepsilon''(\Omega)$ at all frequencies $\Omega$, you can calculate the refractive index $\varepsilon'(\omega)$ at any frequency $\omega$ you choose! For instance, a common form of this relation looks like: $\varepsilon'(\omega) - \varepsilon_{\infty} = \frac{2}{\pi} \mathcal{P} \int_{0}^{\infty} \frac{\Omega \, \varepsilon''(\Omega)}{\Omega^2 - \omega^2} d\Omega$ This equation is a direct cousin of the ones we use in particle physics. That a single principle unites the description of a proton's structure and the optical properties of a diamond is a testament to the profound unity of physics.

Perhaps the most profound application of dispersion relations is in the derivation of ​​sum rules​​. A sum rule is a remarkable type of equation that relates a static, low-energy property of a particle to an integral over its dynamic behavior at all energies.

Consider a particle's polarizabilities, such as the electric polarizability $\alpha_E$ and magnetic polarizability $\beta_M$. These numbers tell you how "squishy" the particle is—how easily its internal charge and current distributions are distorted by static electric and magnetic fields. They are low-energy, classical-sounding properties. You would think you could figure them out just by looking at the particle's structure at rest.

But a twice-subtracted dispersion relation for photon-particle scattering reveals something astonishing. It leads to sum rules that express these properties in terms of an integral over the total photo-absorption cross-section, $\sigma_{\text{tot}}(\omega')$, over all possible photon energies $\omega'$. A well-known example is the Baldin sum rule: $\alpha_E + \beta_M = \frac{1}{2\pi^2} \int_0^\infty \frac{\sigma_{\text{tot}}(\omega')}{\omega'^2} d\omega'$ This is stunning!. The very "squishiness" of a particle at rest is dictated by the total probability of it absorbing a photon of any energy, from radio waves to gamma rays and beyond. The low-energy world is not isolated; it "knows" about all the high-energy physics that can happen. The subtractions performed on the scattering amplitude to derive this rule are crucial, as they separate out the static charge an object has from its dynamical response, reminding us that the story is truly complete only when all scales are accounted for.

Because they connect all energy scales, dispersion relations are the perfect tool to "police" our theories and check them for consistency, especially in the strange world of infinite energy.

A famous historical example is the ​​Pomeranchuk theorem​​. For decades, physicists wondered: if we collide a particle (like a proton) with a target, and then collide its antiparticle (an antiproton) with the same target at ever-increasing energies, must their total interaction probabilities, or cross-sections, become the same? It’s not obvious. But dispersion relations provide the answer. By analyzing the forward scattering amplitude, one can show that if the particle and antiparticle cross-sections were to approach different constant values at infinite energy, the real part of the amplitude would have to grow logarithmically forever. This behavior is considered "unphysical" as it would violate more general theoretical bounds. Thus, causality demands that $\sigma_{\text{particle}}(E) = \sigma_{\text{antiparticle}}(E)$ as $E \to \infty$.

This policing role has taken on a new life in the modern era of Effective Field Theory (EFT). We often don't know the ultimate high-energy theory of nature, so we write down effective theories that are valid at the low energies we can access, with unknown coefficients. For example, a low-energy theory of light might include a tiny self-interaction term, described by a coefficient $c$. Dispersion relations give us a window into the high-energy physics that determines these coefficients. The optical theorem tells us that the imaginary part of an amplitude is related to a probability, and must therefore be positive. By plugging a positive imaginary part into a subtracted dispersion relation, we can prove that certain low-energy coefficients, like $c$, must also be positive. These "positivity bounds" are powerful, model-independent constraints on our theories, derived purely from the fundamental principles of causality and unitarity. The very number of subtractions needed also gives us a clue about how the theory behaves at inaccessible energies.

We have journeyed from the subatomic to the macroscopic. But can these principles reach even further, into the realm of general relativity and gravity? The answer is a resounding yes.

Imagine we want to calculate the tiny corrections to Newton's gravitational potential around a massive object, like a black hole. This is a classical, low-energy problem in gravity. Where would you start? You might try to solve Einstein's equations to a higher order. But dispersion relations offer a completely different, and arguably more profound, route.

Consider the scattering of a massless particle, like a photon or graviton, off the black hole. We can write a sum rule, analogous to the one for polarizability, that relates a coefficient in the low-energy scattering amplitude—which turns out to be exactly the post-Newtonian correction we are looking for—to an integral over the black hole's total absorption cross-section at all energies.

Let that sink in. A classical correction to Newton's law of gravity is determined by the quantum-mechanical probability that the black hole absorbs particles of all possible energies. It is a breathtaking connection between general relativity, quantum mechanics, and thermodynamics (since the black hole's absorption is related to its entropy). It suggests that the principles of causality and analyticity are so fundamental that they form a bridge between our most cherished, and hitherto separate theories of the universe.

From the heart of the proton to the event horizon of a black hole, subtracted dispersion relations are more than just a formula. They are a golden thread of logic, a manifestation of causality that runs through the entire tapestry of physics, binding it together into a single, coherent, and beautiful whole.