Natural Variables and Thermodynamic Potentials

Thermodynamics offers a powerful framework for understanding energy, but its true utility lies in perspective. The state of a system can be described in various "languages," each defined by a specific set of variables. While the fundamental description of a system's energy is given by its internal energy (U), its natural variables—entropy and volume—are often difficult to control in a real-world laboratory setting. This creates a critical gap between fundamental theory and experimental practice, where we typically manipulate temperature and pressure.

This article bridges that gap by exploring the concept of natural variables and the mathematical tools used to navigate between different thermodynamic perspectives. Across two chapters, you will learn how to translate the language of thermodynamics to fit the problem at hand. The "Principles and Mechanisms" chapter will introduce the Legendre transform, the mathematical engine that systematically generates a toolbox of thermodynamic potentials, such as Enthalpy, Helmholtz free energy, and Gibbs free energy, each suited for specific experimental conditions. Following this, the "Applications and Interdisciplinary Connections" chapter will demonstrate how this versatile framework is applied to solve real-world problems in chemistry, biology, and materials engineering, from predicting chemical reactions to designing smart materials.

In our journey to understand the world, we often find that the most profound truths are not about discovering new things, but about finding a better way to look at the things we already know. Thermodynamics is a masterclass in this art of perspective. At its heart lies a single, powerful concept—energy. But as we'll see, the story of thermodynamics is the story of learning how to speak about energy in different "languages," each one perfectly suited to the question we are asking.

Imagine a completely isolated box floating in space, filled with a gas. It's a universe unto itself. No heat can get in or out, its volume is fixed, and no particles can escape. The total energy inside, which we call the ​​internal energy​​ $U$, is constant. The laws of thermodynamics tell us something remarkable: left to its own devices, this system will wiggle and jiggle its way to the most probable, most disordered state. We have a name for this measure of disorder: ​​entropy​​, denoted by $S$.

There is an equivalent, and for us, more useful way to state this. If we were to magically hold the total entropy $S$, volume $V$, and particle number $N$ constant, the system would settle into a state that minimizes its internal energy $U$. This isn't an arbitrary choice of variables. It's a fundamental principle of nature. Because $U$ finds its minimum when $S$, $V$, and $N$ are held fixed, we call these variables the ​​natural variables​​ of the internal energy. We write this relationship as $U(S, V, N)$.

This is the native tongue of internal energy. The fundamental equation of thermodynamics, a compact statement of the first and second laws combined, is written in this language:

$dU = TdS - PdV + \sum_i \mu_i dN_i$

Here, $T$ is temperature, $P$ is pressure, and $\mu_i$ is the chemical potential of component $i$. Notice how each natural variable ($S$, $V$, $N_i$) is paired with a partner ($T$, $-P$, $\mu_i$). These are called ​​conjugate variables​​. The temperature $T$ is the answer to "How much does the energy change if I add a bit of entropy?" The pressure $P$ is related to "How much does the energy change if I squeeze the volume a little?"

This is all very elegant. But there's a practical problem. In a real-world laboratory, we don't control entropy. Who has an "entropy knob"? We control temperature by putting our experiment in a water bath. We often control pressure by leaving our beaker open to the atmosphere. Our chosen variables, the things we actually control, are $(T, P, N)$, not $(S, V, N)$. To make sense of our experiments, we need to translate the language of thermodynamics from its "natural" basis to our "practical" basis.

How do we change our perspective from a function of $S$ to a function of its conjugate partner, $T$? The answer lies in a beautiful mathematical tool called the ​​Legendre transform​​. It sounds intimidating, but the idea is wonderfully simple. It's a systematic procedure for swapping a variable for its conjugate partner in the description of a function.

Think of it this way. You have a curve, $f(x)$. At any point on the curve, there is a value, $x$, and a slope, $p = df/dx$. The original function, $f(x)$, thinks of the curve as a collection of points $(x, f(x))$. The Legendre transform creates a new function, $g(p)$, that describes the same curve but thinks of it as a collection of tangent lines, each defined by its slope $p$ and its y-intercept. The transform is simply the value of that y-intercept: $g(p) = f(x) - px$.

In thermodynamics, this is exactly what we need! We have $U(S,V)$ and its "slope" with respect to entropy is temperature, $T = (\partial U / \partial S)_V$. To switch from a description that depends on $S$ to one that depends on $T$, we perform a Legendre transform.

By applying this transform, we can build a whole toolbox of new functions, called ​​thermodynamic potentials​​, each tailored for a specific job. Let's build the most important ones, starting from $U(S,V)$ for a simple, closed system ($dN=0$).

1. ​​For Constant Temperature and Volume: The Helmholtz Free Energy ($A$)​​ Imagine your experiment is in a rigid, sealed container (constant $V$) held at a constant temperature $T$. We need a potential whose natural variables are $(T,V)$. To get this, we transform $U$ with respect to the entropy $S$. Our new potential is the ​​Helmholtz free energy​​, $A$:

   $A = U - TS$

   Let's see what happens to its differential: $dA = dU - d(TS) = (TdS - PdV) - (TdS + SdT) = -SdT - PdV$

   Look at that! The differential of $A$ depends only on $dT$ and $dV$. This tells us that $(T,V)$ are indeed its natural variables. This equation means that for a system at constant temperature and volume, nature will act to minimize the Helmholtz free energy, $A$.

2. ​​For Constant Entropy and Pressure: The Enthalpy ($H$)​​ Sometimes, we might be studying a process that is very fast and well-insulated (constant $S$), but occurs at a constant external pressure $P$ (like in a cylinder with a movable piston). We need a potential for $(S,P)$. This time, we transform $U$ with respect to volume $V$. The conjugate variable to $V$ is $-P$, so the transform is $U - V(-P) = U+PV$. We call this the ​​enthalpy​​, $H$:

   $H = U + PV$

   Its differential is: $dH = dU + d(PV) = (TdS - PdV) + (PdV + VdP) = TdS + VdP$

   Perfect. The natural variables are $(S,P)$, just as we wanted. Enthalpy is the workhorse of chemistry, as most reactions are studied in open flasks at constant atmospheric pressure.

3. ​​For Constant Temperature and Pressure: The Gibbs Free Energy ($G$)​​ This is the most common scenario in chemistry and biology. A reaction occurs in a test tube at a constant temperature and exposed to the atmosphere's constant pressure. We need a potential for $(T,P)$. To get it, we have to transform away both $S$ and $V$ from the internal energy. We can do this in one step:

   $G = U - TS + PV$

   You might also see this as starting from Enthalpy and transforming away entropy ($G=H-TS$), or starting from Helmholtz energy and transforming away volume ($G=A+PV$). All paths lead to the same destination. Let's find its differential:

   $dG = dU - d(TS) + d(PV) = (TdS - PdV) - (TdS + SdT) + (PdV + VdP) = -SdT + VdP$

   And there it is. The change in the ​​Gibbs free energy​​ $G$ depends only on changes in temperature and pressure. For any process happening at constant $T$ and $P$, the system will evolve to the state with the minimum possible Gibbs free energy. This single principle governs everything from whether a chemical reaction will proceed to whether a protein will fold.

It might be tempting to think we can mix and match variables however we please. Can we make a potential that depends on both entropy $S$ and temperature $T$? Let's try. The fundamental relationship $T = (\partial U / \partial S)_V$ tells us that for a given system, $T$ and $S$ are not independent. If you specify $S$ (and $V$), the temperature $T$ is already determined. The Legendre transform is specifically designed to swap one for the other in the list of independent variables; you can't have both. Having them both as independent variables is like trying to describe a car's motion by specifying both its position and its velocity independently at every moment—it's a contradiction in terms.

What if we try to be clever and invent a new potential, say by combining variables in a non-standard way? For example, one might propose a potential $\chi = H - VS$. At first glance, why not? But when we examine it, the whole elegant structure falls apart. Its differential becomes a messy combination of three differentials, $d\chi = (T - V) dS - S dV + V dP$, so it has no clean pair of natural variables. Even worse, $H$, $V$, and $S$ are all extensive—they double if you double the system size. This means $H$ scales with the size of the system, but the term $VS$ scales with the size squared. The resulting "potential" isn't extensive, a fatal flaw for a useful thermodynamic quantity. This shows us that the four main potentials, $U, H, A,$ and $G$, are not just random combinations; they are the unique, well-behaved functions that arise from the systematic logic of the Legendre transform.

So, we have our toolbox of potentials. What's the payoff? The payoff is immense. First, once you have a potential as a function of its natural variables, you can find other thermodynamic properties just by taking derivatives. From $A(T,V)$, for instance, we saw that $dA = -SdT - PdV$. By simple inspection, we see that:

$\left(\frac{\partial A}{\partial T}\right)_V = -S \quad \text{and} \quad \left(\frac{\partial A}{\partial V}\right)_T = -P$

This is incredible. If a physicist could, through theory or experiment, determine the formula for the Helmholtz energy of a substance as a function of temperature and volume, they could then calculate the substance's entropy and pressure at any state with simple calculus.

But the real magic comes next. For any well-behaved function of two variables, like our potential $A(T,V)$, the order of differentiation doesn't matter. The mixed second derivatives are equal:

$\frac{\partial^2 A}{\partial V \partial T} = \frac{\partial^2 A}{\partial T \partial V}$

Let's apply this. Differentiating $(\partial A / \partial T)_V = -S$ with respect to $V$, and $(\partial A / \partial V)_T = -P$ with respect to $T$:

$\frac{\partial}{\partial V}\left(\frac{\partial A}{\partial T}\right) = -\left(\frac{\partial S}{\partial V}\right)_T$ $\frac{\partial}{\partial T}\left(\frac{\partial A}{\partial V}\right) = -\left(\frac{\partial P}{\partial T}\right)_V$

Setting them equal gives us: $\left(\frac{\partial S}{\partial V}\right)_T = \left(\frac{\partial P}{\partial T}\right)_V$

This is a ​​Maxwell relation​​, one of several such identities. Stop and appreciate what this says. It connects a change in entropy with volume (something very difficult to measure) to a change in pressure with temperature (something you can measure with a thermometer and a pressure gauge). It's a gift from mathematics to physics—a "cheat sheet" that allows us to calculate elusive quantities from easy-to-measure ones.

But this power comes with a crucial warning. These beautiful relationships only work if you are taking derivatives of a potential with respect to its natural variables. If you get sloppy and try to apply the logic to a potential expressed in other variables, the entire framework collapses. For instance, if you (incorrectly) thought you could find a Maxwell relation from $U(T,V)$, you would end up deriving nonsense, like proving that $0$ equals a non-zero number for an ideal gas. The choice of potential is not a matter of taste; it is a matter of correctness. You must use the potential whose natural variables match the variables you are holding constant or treating as independent.

Perhaps the greatest beauty of this formalism is its universality. We've talked about gases with pressure-volume work, but the world is more varied. What about a magnetic material? The work done on it is not $-PdV$, but $\mu_0 H dM_{tot}$, where $H$ is the magnetic field and $M_{tot}$ is its total magnetization. The fundamental equation simply gains a new term:

$dU = TdS - PdV + \mu_0 H dM_{tot}$

Does this break our system? Not at all! It expands it. Now we can define new potentials for magnetic systems. Suppose we want to study a material at constant temperature, pressure, and magnetic field. We just need to perform the appropriate Legendre transforms on $U$:

$G^*(T,P,H) = U - TS + PV - \mu_0 H M_{tot}$

This new "magnetic Gibbs free energy" will be minimized under these exact experimental conditions, and we can use it to predict the material's behavior. The same logic applies to elastic solids, where work is described by stress and strain tensors, or to multicomponent chemical mixtures, where we can exchange particles of one type but not another.

The structure remains the same. Identify the work terms. Write down the fundamental energy equation. Use Legendre transforms to switch to the variables you can control. The resulting potential is your key to understanding the system. This is the true power of thermodynamics: it provides a universal, logical framework for describing change, a language that speaks of energy, entropy, and equilibrium, and can be translated to describe any system we encounter, from a simple gas to the complex machinery of life itself.

Having mastered the art of the Legendre transform, we might be tempted to view it as a clever mathematical reshuffling of variables. But to do so would be to miss the forest for the trees. This mathematical tool is not merely a formal exercise; it is a powerful conceptual lens that allows us to reframe physical questions in their most natural language. The universe, it seems, presents itself differently depending on the questions we ask of it. By choosing the right thermodynamic potential, we are choosing the right question to ask, and in doing so, we unlock a profound and unified understanding of phenomena stretching from heat engines and chemical reactions to the very stability of matter and the design of advanced technologies.

Let's begin our journey with one of the cornerstones of thermodynamics: the Carnot cycle. We learned to describe it as a sequence of four steps: two isothermal (constant temperature) and two adiabatic (constant entropy). If we describe the system's state using internal energy, $U(S,V)$, the adiabatic steps are beautifully simple—they are just processes where one of the natural variables, the entropy $S$, is held constant. But the isothermal steps look more complicated. What if we switch our perspective? By performing a Legendre transform to the Helmholtz free energy, $A(T,V)$, the isothermal steps suddenly become the simple ones—processes where the natural variable, temperature $T$, is held constant. The choice of potential is a choice of convenience; we pick the one whose natural variables match the constraints of the process we care about.

This principle is far more general than just pressure-volume systems. Consider stretching a simple rubber band. The work done on it isn't $-P dV$, but rather $f dL$, where $f$ is the tension and $L$ is the length. The internal energy is naturally a function $U(S, L)$. But what if we are conducting an experiment where we hang a constant weight from the band, thereby fixing the force $f$? It becomes incredibly convenient to define an "elastic enthalpy," $H_L = U - fL$, whose natural variables are $(S, f)$. This new potential behaves for the rubber band exactly as the standard enthalpy $H = U+PV$ behaves for a gas under constant pressure. The formalism is universal. We can generalize this idea to any system by identifying the appropriate "work" terms. This flexibility is the key to extending thermodynamics into other disciplines.

Nowhere is this flexibility more apparent than in chemistry and biology. Here, systems are often "open," meaning they can exchange particles with their surroundings, and the identity of substances can change through chemical reactions.

Let's imagine a chemical reaction proceeding in a beaker. We can describe its progress using an "extent of reaction," $\xi$. The Gibbs free energy $G(T,P,\xi)$ tells us the state of the system. But chemists often prefer to think in terms of the "driving force" of the reaction, a quantity called the chemical affinity, $\mathcal{A}$. Is it possible to have a potential where this driving force is the fundamental variable? Absolutely. A simple Legendre transform gives us a new potential $J(T,P,\mathcal{A}) = G - \mathcal{A}\xi$, which is minimized when the affinity is held constant. This allows us to frame questions about chemical equilibrium in terms of the forces causing the reaction.

This concept becomes even more powerful when we consider mixtures, which are central to biology. Imagine a cell separated from its environment by a semi-permeable membrane—a barrier that allows water to pass but not salt. This is the setup for osmosis. To analyze this, we might want to control the temperature, the volume, and the chemical potential $\mu_1$ of the water (by connecting it to a large reservoir of water), while the amount of salt, $N_2$, inside is fixed. The perfect tool for this job is a potential whose natural variables are precisely $(T, V, \mu_1, N_2)$. We can construct this potential, often called a grand canonical potential, via a Legendre transform: $\Omega = U - TS - \mu_1 N_1$. The state of equilibrium under these conditions—the very balance that prevents a cell from bursting or collapsing—corresponds to the minimum of this specific, custom-built potential.

The true power of natural variables is unleashed when we venture into the world of materials science and engineering. Here, we deal with complex solids and "smart materials" where mechanical, thermal, electrical, and magnetic properties are intricately coupled.

A simple block of steel doesn't just experience uniform pressure; it can be stretched, sheared, and twisted. To describe its deformation, we need the language of tensors—the strain tensor $\boldsymbol{\varepsilon}$ and the stress tensor $\boldsymbol{\sigma}$. Yet, the thermodynamic framework holds. We can define an internal energy density $u(\boldsymbol{\varepsilon}, s)$ and perform a series of Legendre transforms to generate a full suite of potentials—like the Helmholtz density $\psi(\boldsymbol{\varepsilon}, T)$ or the Gibbs density $g(\boldsymbol{\sigma}, T)$—that are perfectly suited for describing a solid under different thermal and mechanical loads.

The beauty of this is that we can keep adding work terms for any physical effect we can imagine.

• For an electrically charged, elastic filament, the energy includes terms for both mechanical tension ($\tau dL$) and electrical charging ($\Phi dq$). If we want to analyze an experiment at constant temperature, tension, and voltage, we simply construct the potential $\Omega = U - TS - \tau L - \Phi q$.
• For a magnetic fluid that also has a surface, the energy contains terms for volume changes ($-PdV$), magnetization changes ($\mu_0 H dM$), and surface area changes ($\gamma dA$). We can construct a potential for any combination of controlled variables, such as a "Gibbs-like" potential $G_H(T, P, A, H)$ for an experiment at constant temperature, pressure, surface area, and external magnetic field, or a "grand" potential $\Omega(T, P, H, \gamma)$ for an experiment where the surface tension is controlled instead of the area. Each potential is a fit-for-purpose theoretical tool.

This framework is not just descriptive; it is the predictive engine behind modern engineering. Consider a piezoelectric material, like quartz, which generates a voltage when squeezed. These materials are the hearts of sensors, actuators, and resonators. To model such a device, an engineer needs to know the relationships between stress, strain, electric field, and electric displacement. These relationships are found in a thermodynamic potential. For a typical simulation where strain $\boldsymbol{\varepsilon}$ and electric field $\boldsymbol{E}$ are the primary variables, the Gibbs-like potential $G(\varepsilon, T, \boldsymbol{E})$ is the perfect choice. Why? Because its second derivatives—the curvature of the potential's energy landscape—give the material's constitutive properties: the elastic stiffness tensor, the dielectric tensor, and, most importantly, the piezoelectric coupling tensor that links the mechanical and electrical worlds. These are the numbers that go into finite element (FE) simulation software to design our next generation of devices.

Finally, the choice of potential and its natural variables touches upon two fundamental aspects of the physical world: why is matter stable, and how can we reliably simulate it?

The stability of any substance—the reason a glass of water remains liquid water and doesn't spontaneously boil and freeze at the same time—is guaranteed by the shape of its thermodynamic potentials. For a material to be stable, its Helmholtz free energy $A(T,V)$ must be convex with respect to volume. This mathematical condition translates directly into a physical requirement: the isothermal compressibility $K_T$ must be positive. If it weren't, an increase in pressure would cause the material to expand, leading to a catastrophic collapse. The boundary of this stability is called the spinodal, a point where the compressibility diverges to infinity. Approaching this boundary, where the potential flattens out before turning unstable, also causes other properties, like the thermal expansion coefficient $\alpha_P$, to diverge. The very structure of matter is written in the geometry of these potentials.

This has profound implications for how we simulate the real world. In a finite element simulation of a thermoelastic structure, the choice of thermodynamic potential is a critical decision that depends on the problem's boundary conditions.

• If we are simulating a block of material where the displacements are prescribed on its boundaries, we are essentially controlling the strain field. The most "natural" choice for a variational principle is one based on a potential whose natural variables include strain, such as the internal energy $u(\boldsymbol{\varepsilon}, s)$ for an adiabatic process or the Helmholtz free energy $\psi(\boldsymbol{\varepsilon}, T)$ for an isothermal one.
• If, instead, we prescribe the forces (tractions) on the boundary, we can still use a displacement-based formulation with $\psi$, but the forces now contribute as an external work term in our total potential functional.

The abstract choice of variables, made so elegantly by the Legendre transform, has direct and practical consequences for the formulation of our most advanced computational tools. It is the bridge connecting the fundamental laws of thermodynamics to the predictive power of modern engineering. From the simple piston to the smart sensor, the language of natural variables provides a single, coherent, and breathtakingly elegant framework for understanding and manipulating the physical world.