Hubbard Bands: Understanding Electron Correlation in Solids

In the world of materials, standard band theory provides a powerful framework, successfully explaining why copper conducts electricity and diamond does not. It paints a picture of electrons moving freely through a crystal lattice, with a material's conductivity determined simply by whether its energy bands are partially or completely full. However, this elegant picture shatters when confronted with a class of materials known as 'strongly correlated systems.' Here, materials that should be metals according to band theory are found to be staunch insulators. This discrepancy highlights a fundamental gap in the simple, non-interacting electron model and points to a deeper, more complex reality governed by the fierce repulsion between electrons.

This article provides the key to understanding this puzzle: the concept of Hubbard bands. We will journey from foundational principles to real-world applications across two chapters. In the first chapter, "Principles and Mechanisms," we will explore the fundamental battle between an electron's urge to move and the high energy cost of sharing a space, revealing how this conflict tears a single band apart to form the distinct Lower and Upper Hubbard bands. Following this, the "Applications and Interdisciplinary Connections" chapter will demonstrate how these theoretical ideas explain the metal-insulator transitions in real materials, how scientists experimentally 'see' these bands, and how doping these systems paves the way for exotic phenomena like high-temperature superconductivity. Our exploration begins with the core quantum mechanics that set the stage for this entire drama.

Imagine you are trying to understand the flow of traffic in a city. A simple model might just count the number of cars and the number of roads. If there are plenty of open lanes, cars move freely. This, in a nutshell, is the traditional picture of electrons in a simple metal. The electrons are the cars, and the crystal lattice provides the roads. As long as the "energy bands" (the highways) are not completely full, electrons can move and conduct electricity. This standard ​​band theory​​ is a monumental achievement, explaining why materials like copper are shiny metals and diamond is a transparent insulator.

But what if the drivers deeply disliked sharing a stretch of road with another car? What if this "social distancing" for cars was so strong that it completely changed the flow of traffic, causing gridlock even on a half-empty highway? This is precisely the kind of strange and wonderful physics we encounter in "strongly correlated" materials. The simple counting of electrons and states is not enough. We must face the consequences of their mutual repulsion, and when we do, a whole new landscape emerges, dominated by what we call ​​Hubbard bands​​.

At the heart of this entire story are two competing energies. Think of a string of atoms in a crystal, and for simplicity, let's say each atom contributes one electron to the collective.

First, there's the quantum mechanical urge of an electron to not be tied down. An electron on one atom can "hop" to a neighboring atom. This delocalization lowers its kinetic energy. The characteristic energy scale for this process is called the ​​hopping integral​​, denoted by $t$. A large $t$ means electrons are restless and mobile, zipping across the lattice—the hallmark of a metal.

Second, there is the formidable electrostatic repulsion between two electrons. While electrons are spread out in a crystal, they still feel each other's presence. There is a particularly large energy penalty if two electrons try to occupy the same atomic site at the same time. This energy cost is known as the ​​on-site Coulomb repulsion​​, or simply $U$. A large $U$ means electrons will do everything they can to avoid each other, staying one to a site.

The entire drama—from simple metal to bizarre insulator to exotic superconductor—is staged by the battle between $t$ and $U$. It is the ratio of these two numbers that dictates the fate of the material.

To understand the power of $U$, let us perform a thought experiment. Let's crank up the repulsion until it is enormous, and dial down the hopping until it is zero ($U \gg t$, and eventually $t=0$). What happens?

With no hopping, the atoms are completely isolated islands. The electrons are prisoners on their respective atomic sites. We have a crystal where each site has exactly one electron. Now, for an electric current to flow, electrons must move. Imagine trying to move an electron from site A to a neighboring site B, which is already occupied. This move would create two very special sites: site A would become empty (we call this a ​​holon​​), and site B would become doubly occupied (we call this a ​​doublon​​).

But creating this doublon comes at a steep price: the energy cost is precisely $U$. So, to get even a single electron to move and create a current, the system must pay an energy toll of $U$. If this cost is too high, no current flows. The system, which according to simple band theory should be a metal (its single energy band is only half-full), is completely stuck. It has become an insulator!

This is a ​​Mott insulator​​—a material that is insulating not because it lacks available energy states, but because the strong repulsion between electrons localizes them and forbids them from moving. This mechanism is profoundly different from that of a conventional ​​band insulator​​, like silicon, where insulation arises because all energy bands are either completely full or completely empty even without any repulsion. The Mott gap is born from interaction, not from a pre-existing band structure.

In the language of quantum mechanics, the single, continuous band of available energy states that we would have in a simple metal has been torn apart by the colossal force of $U$. It splits into two distinct, separate collections of states:

1. A lower energy set of states corresponding to the process of removing an electron (creating a holon). This is the ​​Lower Hubbard Band (LHB)​​.
2. A higher energy set corresponding to adding an electron (creating a doublon). This is the ​​Upper Hubbard Band (UHB)​​.

In our atomic limit ($t=0$), these "bands" are just single, sharp energy levels separated by a gap of energy $U$.

Of course, in a real material, hopping is never truly zero. So let's turn on a small but finite $t$ ($U \gg t > 0$). Our atomic prison now has tiny cracks in the walls. An electron has a small chance to "tunnel" to a neighbor. What does this do to our two sharp energy levels?

Consider the state where we have created one doublon. That extra electron on the doubly-occupied site can now hop to an adjacent, singly-occupied site. In doing so, it has effectively moved the doublon! Similarly, if we have a holon (an empty site), an electron from a neighboring site can hop into it, effectively moving the holon.

This mobility is a quantum mechanical reality. And just as a single electron hopping in a crystal gives rise to an energy band with a certain width, the motion of our holon and doublon "quasiparticles" also broadens their respective energy levels into bands. The width of these brand-new Hubbard bands is determined by the ease of hopping—that is, by the hopping integral $t$. Remarkably, in the limit of small $t$, the bandwidth of the LHB (describing the holon's motion) and the UHB (describing the doublon's motion) is proportional to the original non-interacting bandwidth that would have existed if $U$ were zero.

So now our picture is more complete. For $U \gg t$, the electronic spectrum consists of two Hubbard bands, separated by a large ​​Mott-Hubbard gap​​. The lower band is full, the upper band is empty, and the system is an insulator. The size of this gap is dominated by $U$, but it is slightly reduced by the kinetic energy gained from hopping. A good approximation for the gap is $\Delta_{gap} = \sqrt{U^2+W^2}-W$, where $W$ is a measure of the non-interacting bandwidth (which is proportional to $t$). You can see that if hopping is impossible ($W=0$), the gap is exactly $U$. As hopping becomes more significant (as $W$ increases), the gap shrinks. Eventually, it will collapse entirely.

What happens when $U$ and $t$ (or $W$) become comparable? The gap between the Hubbard bands shrinks and, at a critical value of $U/W$, vanishes. The insulator gives way to a metal. But this is no ordinary metal. This is a ​​correlated metal​​, a state of matter teetering on the edge of localization, and its properties are wonderfully strange.

To see this strangeness, we must look at the material's ​​single-particle spectral function​​, $A(\mathbf{k}, \omega)$. Think of this as a detailed energy map, telling us for a given momentum $\mathbf{k}$, what are the allowed energies $\omega$ for an electron. For any valid quantum state, the total probability of finding an electron must be one, which leads to the fundamental sum rule $\int_{-\infty}^{\infty} d\omega A(\mathbf{k}, \omega) = 1$.

In the Mott insulator, the map is simple: it shows two massive continents of states (the Hubbard bands) separated by a vast ocean (the Mott gap). But as we cross over into the correlated metal phase, a new, needle-like island appears right in the middle of the ocean, at the Fermi energy ($\omega=0$).

This sharp central feature is called the ​​quasiparticle peak​​. It represents a miraculous accommodation by the system. Despite the intense repulsion, the system conspires to create an excitation that behaves almost like a free electron. This "quasiparticle" is, however, heavily "dressed." It moves through the lattice dragging a complex cloud of interactions with other electrons, making it much heavier than a bare electron.

The total spectral weight of this peak, known as the ​​quasiparticle residue​​ $Z$, quantifies what fraction of the original electron survives as a coherent, mobile entity. The rest of the electron's identity, the weight $1-Z$, is smeared across the two large, incoherent Hubbard bands which still loom at high energies. The weight $Z$ can be calculated from the electron's ​​self-energy​​ $\Sigma(\omega)$, a term that encapsulates all the complex interaction effects, via the relation $Z = [1 - \frac{\partial \Sigma'(\omega)}{\partial \omega}|_{\omega=0}]^{-1}$.

As one approaches the Mott insulating state by increasing $U$, the quasiparticle gets heavier and heavier, and its coherent fraction $Z$ dwindles toward zero. At the transition point, $Z$ vanishes, the central peak disappears entirely, and all electronic states are exiled to the Hubbard bands. The miracle is over, and the system is locked back into an insulating state.

This quasiparticle state is a delicate, low-temperature quantum phenomenon. It relies on the coherent, wave-like nature of electrons to cleverly navigate the correlated landscape. What happens if we turn up the heat?

Temperature is the great enemy of quantum coherence. Thermal jiggling introduces randomness and scattering, which disrupts the fragile phase relationships that allow the quasiparticle to exist. There is a characteristic ​​coherence temperature​​, $T_{coh}$, below which the quasiparticle is well-defined. This temperature scale is itself proportional to the quasiparticle's weight, $T_{coh} \propto Z$. A more robust quasiparticle (larger $Z$) can survive to higher temperatures.

As the temperature rises above $T_{coh}$, the sharp quasiparticle peak in the spectral function gets broader and shorter, ultimately melting away into the incoherent background. The spectral weight that was concentrated in the peak is redistributed back into the Hubbard bands. The material transforms from a low-temperature, (somewhat) well-behaved correlated metal into a high-temperature "bad metal," a state with very poor conductivity where no simple particle-like description of charge carriers is valid.

This beautiful, intricate dance—between hopping and repulsion, between coherence and incoherence, between low and high energy, and between zero and finite temperature—is the essence of Hubbard band physics. It shows how simple rules, when combined in the quantum realm, can give birth to a stunning complexity that continues to challenge and inspire physicists today.

In the previous chapter, we uncovered a remarkable piece of quantum drama. We saw that a collection of electrons in a crystal, which band theory naively predicts should be a bustling metallic city, can instead become a frozen, insulating grid. The reason? A powerful social force among electrons: their mutual repulsion. When the cost of two electrons sharing the same atomic "room," the on-site repulsion $U$, becomes too high, they choose to self-isolate, each confined to their own site. This collective lockdown splits a once-unified electronic band into two—the occupied Lower Hubbard Band and the empty Upper Hubbard Band. But this is far more than a theoretical curiosity. This single idea provides the key to unlocking the behavior of a vast and technologically important class of materials, known as strongly correlated systems. Now, let's take a journey to see where this concept leads, from explaining the identity of materials to probing the frontiers of modern physics, including the mystery of high-temperature superconductivity.

The most direct consequence of the battle between electron delocalization and repulsion is the metal-insulator transition. Imagine the electrons as tenants in a building (the crystal lattice). The hopping energy, $t$, related to the overall bandwidth $W$, represents their desire to move freely between rooms, socializing and spreading out. The repulsion energy, $U$, is a steep "double occupancy" fee for any two electrons trying to share the same room. Which force wins?

A simple and powerful heuristic argument gives us the answer. If the energy an electron gains by delocalizing across the entire crystal (an energy on the order of the bandwidth, $W$) is less than the penalty $U$ for sitting on an already-occupied site, the electrons will give up on a nomadic lifestyle. They will lock into place, one per site. The system grinds to a halt, becoming an insulator. The transition, a line in the sand between metal and insulator, happens roughly when the repulsion equals the bandwidth: $U_c \approx W$. For $U \gt W$, a gap opens up between the Hubbard bands, and the material is a ​​Mott insulator​​. For $U \lt W$, the bands overlap, and the material remains a metal. This simple criterion is the first great success of the Hubbard model, explaining why many transition metal oxides with partially filled $d$-shells are, contrary to simple band theory, excellent insulators.

Of course, nature's bookkeeping is always a bit more subtle than our first, back-of-the-envelope calculations. More sophisticated theoretical treatments, like the Hubbard-I approximation or the modern Dynamical Mean-Field Theory (DMFT), refine this picture. They show that the size of the Mott gap, $\Delta_{gap}$, isn't just a simple subtraction, but depends on the interplay between $U$ and the hopping $t$ in more intricate ways, often yielding expressions like $\Delta_{gap} = \sqrt{U^2 + (c t)^2} - c t$ for some constant $c$ that depends on the lattice geometry. This progression from a simple rule of thumb to a detailed, quantitative formula is the very essence of how physics advances.

How can we be sure these Hubbard bands are not just figments of a theorist's imagination? The answer is that we can "see" them, not with our eyes, but with the power of spectroscopy.

One of the most direct methods is ​​optical spectroscopy​​. If you shine light on a Mott insulator, a photon can be absorbed, but only if it has enough energy to kick an electron across the forbidden energy gap. This means promoting an electron from the filled Lower Hubbard Band into the empty Upper Hubbard Band. At the microscopic level, this corresponds to forcing an electron to hop from its site onto a neighbor that is already occupied. This process creates a mobile pair of charge defects: an empty site, or ​​holon​​, and a doubly-occupied site, or ​​doublon​​. The minimum photon energy required to do this directly measures the Mott gap. Below this energy, the material is transparent; at this energy, it begins to absorb light. A crucial signature of a Mott insulator is the complete absence of a "Drude peak" at zero frequency in the conductivity—the hallmark of mobile charges in a conventional metal. This starkly confirms the frozen, insulating nature of the ground state.

Another powerful tool is ​​photoemission spectroscopy​​, which is essentially the photoelectric effect applied to solids with surgical precision. By striking the material with high-energy photons (typically X-rays), we can knock electrons clean out. We then measure the kinetic energy of these escaping electrons. By subtracting this kinetic energy from the photon's known initial energy, we can work backward to determine the energy the electron had when it was inside the material. This technique allows us to map out the entire energy landscape of the occupied states. For a Mott insulator, photoemission experiments directly image the Lower Hubbard Band. The minimum photon energy required to eject an electron, the photoemission threshold $\hbar\omega_{\text{th}}$, is determined by the energy needed to lift an electron from the very top of this Lower Hubbard Band all the way to the vacuum, outside the material. This threshold is directly related to the material's work function $\Phi$ and the Hubbard repulsion $U$, often expressed as $\hbar\omega_{\text{th}} = \Phi + U/2$ in a simplified model. These experiments provide incontrovertible evidence for the existence and structure of Hubbard bands.

Here is where the story takes a truly fascinating turn. What happens if we aren't at perfect half-filling? We can chemically alter the material to have slightly fewer (hole-doping) or slightly more (electron-doping) electrons per site. This seemingly small change can have dramatic consequences, transforming the insulator into a metal, and not just any metal, but often a very strange one, a "bad metal" that defies many conventional descriptions. In the case of cuprate materials, this process can even lead to high-temperature superconductivity.

Doping a Mott insulator doesn't just put a few carriers into otherwise rigid, static bands. The strong correlations that created the Mott state are still a dominant force. Instead, the entire electronic structure reorganizes itself. As you introduce holes by removing electrons, the system reshuffles its available states. Spectral weight—a term physicists use to describe the "amount" of electron-like character at a given energy—is transferred from the high-energy Hubbard bands into new states that emerge right inside the original Mott gap. A sharp, coherent ​​quasiparticle peak​​ appears at the Fermi level, growing in intensity as the doping level, $\delta$, increases. In a beautiful demonstration of the power of theoretical physics, one can show using exact sum rules that in the limit of very large $U$, the total spectral weight of these new in-gap states is precisely proportional to the doping: $W_{\text{gap}}(\delta) = 2\delta$. This means that for every hole you introduce, the system creates two available states in the gap! This is a profound consequence of many-body physics, utterly absent in non-interacting pictures.

Even more remarkably, these doped Mott insulators, while strange, must still obey some fundamental laws of quantum mechanics. One such law is Luttinger's theorem, which states that the volume of the Fermi surface—the boundary in momentum space separating occupied and unoccupied states—is determined by the total number of electrons, $1-\delta$, not just the small number of doped holes, $\delta$. This results in a "large" Fermi surface, a surprising feature that has been confirmed in many correlated materials and places strong constraints on any theory attempting to describe them.

So far, our discussion has centered on a simplified model with just one orbital per atom. But real materials, especially the transition metal oxides that are the canonical examples of Mott physics, have a richer structure. In copper oxides (cuprates), for instance, the low-energy physics involves not only the copper $3d$ orbitals but also the oxygen $2p$ orbitals that surround them.

This introduces a new character to our drama: the charge-transfer energy, $\Delta$. This is the energy cost to move an electron from an oxygen atom to a copper atom. Now, the system has a choice for its lowest-energy charge excitation. Is it cheaper to move an electron between two copper sites (a $d \to d$ excitation, costing energy $U$)? Or is it cheaper to move an electron from oxygen to copper (a $p \to d$ excitation, costing energy $\Delta$)?

The ​​Zaanen-Sawatzky-Allen (ZSA)​​ classification scheme organizes materials based on the answer to this question.

• If $U < \Delta$, the system is a ​​Mott-Hubbard insulator​​. The gap is determined by $U$, and the physics is dominated by the copper $d$-orbitals.
• If $\Delta < U$, the system is a ​​charge-transfer insulator​​. The gap is set by the smaller energy $\Delta$. In this scenario, the oxygen $p$-band is wedged energetically between the Lower and Upper Hubbard bands of the copper sites.

Crucially, the parent compounds of the high-temperature cuprate superconductors fall into this second category: they are charge-transfer insulators. This has a profound implication: when these materials are hole-doped, the holes predominantly reside on the oxygen sites, not the copper sites. This insight is a cornerstone of virtually every theory of high-temperature superconductivity. Physicists have even created detailed phase diagrams based on parameters like $U$, $\Delta$, and various hopping integrals, mapping the territory that separates these different classes of insulators.

Our journey, which began with the simple idea of electrons avoiding one another, has led us to the heart of modern condensed matter physics. The concept of Hubbard bands has given us a language to understand metal-insulator transitions, interpret complex spectroscopic data, and classify the diverse family of transition metal oxides. It forms the essential backdrop for understanding the emergence of high-temperature superconductivity in doped Mott insulators.

And yet, for all its successes, the story is far from over. The precise mechanism of superconductivity in the cuprates remains one of the greatest unsolved problems in physics. The "strange metal" phase that appears upon doping exhibits properties that defy our standard theories of metals. To tackle these challenges, physicists employ ever more powerful theoretical tools, such as Dynamical Mean-Field Theory (DMFT), and cutting-edge experiments. The simple Hubbard model, born of a thought experiment, has become a vast and fertile ground for discovery, reminding us that sometimes the most profound secrets of the universe are hidden within the intricate, collective dance of its most fundamental particles.