Non-Orthogonal Coordinates

In a world often simplified by neat, rectangular grids, the Cartesian coordinate system has long been our trusted guide. However, the true geometry of the universe—from the arrangement of atoms in a crystal to the curvature of spacetime—rarely conforms to such rigid structure. Attempting to force these complex, curved systems into a flat, right-angled framework leads to distortions and unnecessary complexity, obscuring the elegant physical laws that govern them. This article addresses this fundamental limitation by introducing the powerful world of non-orthogonal, or curvilinear, coordinates.

This guide will equip you with the language needed to navigate these skewed and curved spaces. In the first chapter, ​​"Principles and Mechanisms"​​, we will deconstruct the mathematical machinery, exploring local basis vectors, the all-important metric tensor, the dual nature of vector components, and the new rules of calculus required for a world where our rulers bend and stretch. Subsequently, in ​​"Applications and Interdisciplinary Connections"​​, we will see this framework in action, discovering how it provides clarity and deep insights across a vast range of fields, from solid mechanics and general relativity to computational fluid dynamics and theoretical chemistry.

Imagine you’re trying to give directions. On a neat, rectangular city grid, it’s a breeze: "Go three blocks east and two blocks north." This is the world of René Descartes, a world of perpendicular lines and constant distances. The basis of this world is the familiar Cartesian coordinate system $(x,y,z)$, where every step in the $x$ direction is independent of any step in the $y$ or $z$ direction. It's clean, it’s simple, and for a vast number of problems, it’s perfect.

But the universe, in all its magnificent complexity, is rarely so polite. What if you're a crystallographer studying the arrangement of atoms in a skewed lattice? Or a geophysicist mapping gravitational fields on our spherical Earth? Or an engineer analyzing fluid flow around a curved airfoil? Forcing these problems onto a rigid rectangular grid is like trying to wrap a basketball in a flat sheet of paper—you end up with awkward folds and distortions. We need a language, a mathematical framework, that embraces the natural shape of the problem. This is the world of non-orthogonal, or curvilinear, coordinates. It’s a world where our grid lines can bend, stretch, and meet at peculiar angles, and learning its rules is like uncovering a deeper grammar of the physical world.

In the Cartesian world, the signposts—the basis vectors $\mathbf{\hat{i}}$, $\mathbf{\hat{j}}$, and $\mathbf{\hat{k}}$—are steadfast and loyal. They point in the same direction with the same length, no matter where you are. But in a curvilinear system, our signposts become local guides, changing their orientation and length from one point to the next.

How do we find these new guides? We let the coordinate system itself define them. If our position in space, $\mathbf{r}$, is described by some new coordinates, say $(u^1, u^2, u^3)$, then the most natural way to define a basis vector is to see how our position changes as we take a tiny step along one of these coordinate lines. Mathematically, this is just a partial derivative. The ​​covariant basis vectors​​ are defined as:

These vectors are tangent to the coordinate curves at every point. For instance, if you were using parabolic cylindrical coordinates $(\sigma, \tau, z)$, the basis vector $\mathbf{e}_\sigma$ points in the direction you'd move if you changed $\sigma$ slightly while keeping $\tau$ and $z$ constant. Unlike our old Cartesian friends, these basis vectors are, in general, neither perpendicular to each other nor of unit length. They are living, breathing functions of position. And this is where all the fun begins.

With our familiar right-angled grid gone, the Pythagorean theorem, in its simple form $ds^2 = dx^2 + dy^2$, is no longer up to the task of measuring distances. We need a new, more powerful tool. Enter the star of our show: the ​​metric tensor​​, $g_{ij}$.

The metric tensor is the fundamental object that encodes all the geometric information of our coordinate system—lengths, angles, and volumes. Its components are surprisingly easy to define: they are simply the dot products of our new basis vectors with each other.

Let's unpack this. The diagonal components, like $g_{11} = \mathbf{g}_1 \cdot \mathbf{g}_1 = |\mathbf{g}_1|^2$, tell us the squared length of our basis vectors. They are the "scale factors" along each coordinate axis. But the real magic lies in the off-diagonal components, like $g_{12} = \mathbf{g}_1 \cdot \mathbf{g}_2$. If the coordinate axes are orthogonal, their dot product is zero, and the off-diagonal components vanish. If they are not orthogonal—if our grid is skewed—these components will be non-zero. They directly measure the "skewness" of our system.

Consider a simple 2D system where the axes meet at an angle $\alpha$. The distance $d$ from the origin to a point with coordinates $(U, V)$ isn't $\sqrt{U^2 + V^2}$. A bit of geometry shows that it's given by the Law of Cosines:

Now, let's look at this through the lens of the metric tensor. The infinitesimal squared distance, our new "Pythagorean theorem," is written as:

For the oblique system with angle $\alpha = \frac{\pi}{3}$, the metric tensor turns out to be $\begin{pmatrix} 1 \cos(\pi/3) \\ \cos(\pi/3) 1 \end{pmatrix}$. That off-diagonal term, $\cos(\pi/3)$, is the signature of non-orthogonality. In fact, the metric tensor gives us a universal formula for the angle $\theta$ between any two basis vectors $\mathbf{g}_i$ and $\mathbf{g}_j$:

This little equation is incredibly powerful. It means that if someone hands you the metric tensor for a space, like $ds^2 = du^2 + dv^2 + du\,dv$, you can immediately tell them the angle between their coordinate axes at every point—in this case, $60^\circ$. The metric tensor is the geometric DNA of the coordinate system.

In the orderly Cartesian world, representing a vector is simple: we break it down into components along the axes. But in our skewed world, a subtle and profound duality emerges. Because our basis vectors $(\mathbf{g}_1, \mathbf{g}_2, ...)$ are no longer orthogonal, there exists a second, equally valid set of basis vectors, called the ​​dual basis​​ or ​​reciprocal basis​​, denoted $(\mathbf{g}^1, \mathbf{g}^2, ...)$.

These two sets are linked by a beautiful relationship of biorthogonality:

Here, $\delta^i_j$ is the Kronecker delta, which is $1$ if $i=j$ and $0$ otherwise. What does this mean? It means that the dual basis vector $\mathbf{g}^1$ is perpendicular to all the original basis vectors except $\mathbf{g}_1$. Think of the original basis vectors as forming the ribs of a skewed canopy. The dual vectors are like poles that are perpendicular to the fabric surfaces between the ribs. For any given covariant basis, there is a unique dual basis that satisfies this condition.

Now, here's the kicker: any vector $\mathbf{v}$ can be expressed as a combination of either set of basis vectors.

The numbers $v^i$ are the ​​contravariant components​​, and the numbers $v_i$ are the ​​covariant components​​. They are not the same! They are two different ways of describing the exact same physical vector. The contravariant components tell you "how many steps to take along the original basis vectors" (like a parallelogram rule), while the covariant components are the dot products of the vector with the original basis vectors.

And what connects these two descriptions? Our old friend, the metric tensor. It acts as the universal translator:

(Here, $g^{ij}$ are the components of the inverse metric tensor, and we use the Einstein summation convention where a repeated index implies summation). This is why index position—up or down—is so critical in this language. It tells you which type of component and which type of basis vector you're working with.

This machinery isn't just for show. It has real computational power. For example, the dot product of two vectors $\mathbf{u}$ and $\mathbf{v}$, a fundamental physical operation, is no longer a simple sum of products of components. Its true, coordinate-independent form is revealed:

This single, elegant formula works in any coordinate system, whether it’s Cartesian, skewed, polar, or something far more exotic. It automatically incorporates all the geometric information—the lengths and angles of the basis vectors—to give the correct physical result.

Now for the ultimate challenge: how do we do calculus? How do we talk about rates of change, like gradients and divergences, when our very coordinate grid is changing from point to point?

If we take a simple partial derivative of a vector's components, $\frac{\partial A^i}{\partial x^j}$, the result tragically fails to transform like a proper tensor. It's not a coordinate-independent object. The reason is that this simple derivative ignores a crucial fact: the basis vectors themselves are changing.

To fix this, we must introduce the ​​covariant derivative​​, often denoted with a semicolon ($;$) or the nabla symbol $\nabla$. The covariant derivative of a vector component looks like this:

Those new symbols, $\Gamma^i_{kj}$, are the ​​Christoffel symbols​​. You can think of them as "correction terms." They are derived from the metric tensor and account for how the basis vectors twist and turn as you move through space. They are the price we pay for the freedom of using any coordinate system we like.

With the covariant derivative, all the familiar operators from vector calculus—gradient, divergence, curl—can be generalized into forms that are truly universal. Physical laws, like the equilibrium equations in solid mechanics, must be written using these operators. The statement that "net force is zero" cannot depend on whether you're using Cartesian or polar coordinates. The equation for static equilibrium, $\text{div}(\boldsymbol{\sigma}) + \mathbf{b} = \mathbf{0}$, thus takes the tensorial form $\sigma^{ij}{}_{;j} + b^i = 0$. This is not just a mathematical curiosity; it is a profound statement about the objectivity of physical law.

After building this vast and seemingly complex machinery, one might wonder if anything simple is left. The answer is a resounding yes. The whole point of this framework is not to create complexity, but to manage it, in order to reveal the simple, underlying truths that are independent of our descriptive choices.

Consider the famous vector identity: the divergence of the curl of any vector field is always zero, $\nabla \cdot (\nabla \times \mathbf{A}) = 0$. In Cartesian coordinates, this is a straightforward, if tedious, calculation. Does this beautiful identity survive in the wild world of curvilinear coordinates, with all its Christoffel symbols and shifting basis vectors?

Yes, it does. In a flat Euclidean space, even when described by the most contorted coordinates, the identity holds perfectly. The covariant derivatives and Levi-Civita tensors (the proper tensorial version of the permutation symbol $\varepsilon_{ijk}$ are designed so that all the correction terms conspire to cancel out perfectly. The identity is a statement about the fundamental nature of space itself, not about the grid we lay on top of it.

This is the ultimate payoff. By learning the language of tensors and non-orthogonal coordinates, we are not just learning a new calculational tool. We are learning to distinguish between the superficial features of our description and the deep, invariant truths of nature. We can bend and twist our perspective, yet the fundamental laws of physics remain unchanged, their inherent beauty and unity shining through.

After our journey through the machinery of non-orthogonal coordinates—the basis vectors, the metric tensor, and the new rules of calculus—you might be wondering, "Why go to all this trouble?" Life on a grid of perfect squares is so comfortable and familiar. Why would we ever willfully abandon the safety of our Cartesian graph paper?

The answer is simple and profound: the universe is not laid out on graph paper. The problems we want to solve, from the structure of a crystal to the flow of air over a wing, come with their own inherent geometries. The most elegant, efficient, and sometimes the only way to understand these problems is to adopt a point of view, a coordinate system, that respects the "grain" of the situation. By choosing coordinates that are tailor-made for the problem, we often find that overwhelming complexity melts away, revealing a beautiful and simple underlying structure. Let's see how this powerful idea plays out across a spectacular range of scientific and engineering disciplines.

Physics is often the study of fields—electric, magnetic, gravitational, velocity—that permeate space. These fields don't care about our chosen axes; they follow their own rules. To describe them properly, we need to speak their language.

Consider the physics of crystalline solids. A crystal is a beautiful, repeating lattice of atoms, but these lattices are not always cubic. Many important materials have skewed, non-orthogonal crystal structures. Their properties, like electrical conductivity or how light passes through them, are anisotropic; that is, they depend on direction. To describe an electric potential inside such a material, it is far more natural to use a coordinate system aligned with the crystal's own axes. In this skewed frame, calculating the electric field, which is just the gradient of the potential, becomes a straightforward exercise. The non-orthogonal coordinates are not a mathematical abstraction; they are a direct reflection of the material's physical reality, simplifying our description of the forces at play. The basic act of finding a gradient, which we have learned how to do in any system, becomes the key to unlocking the electrodynamics of these complex materials.

This principle extends beautifully to the mechanics of continuous media, like fluids and elastic solids. Imagine a simple shear flow, where layers of fluid slide over one another like a deck of cards being pushed from the side. You could describe this with standard $(x, y)$ coordinates, but it's a bit awkward. The fluid particles are all moving horizontally, but the velocity depends on the vertical position. A more insightful approach is to use a coordinate system that shears with the flow. In these sheared coordinates, the description of the fluid's deformation—quantified by a mathematical object called the rate-of-strain tensor—can become much clearer and more intuitive.

Now, think about a solid object, perhaps a complexly shaped engine component, held in static equilibrium. Every tiny piece of this object is in balance, with the internal stresses and any external body forces canceling out. Writing down this condition of force balance, which involves the covariant divergence of the stress tensor, can be a nightmare in Cartesian coordinates if the object has curved or slanted boundaries. However, by using a curvilinear coordinate system that conforms to the shape of the body, we can express these fundamental laws of equilibrium in a manageable way. The Christoffel symbols, which seemed so abstract, become essential tools that tell us how the basis vectors twist and turn, allowing us to correctly state Newton's laws in a curved and skewed world. This is the heart of modern solid mechanics and structural engineering.

Much of modern science and engineering relies on computers to solve complex equations. Here, the choice of coordinates is not just a matter of elegance, but of accuracy and feasibility.

Imagine trying to simulate the flow of air over a smooth, curved airplane wing using a grid of tiny squares. At the boundary of the wing, the grid can only form a crude, "stair-stepped" approximation of the true shape. This jagged representation introduces significant errors into the calculation, polluting the entire solution. The cure is to use a "body-fitted" coordinate system, a grid that gracefully wraps around the wing. Away from the wing, the grid might look nearly Cartesian, but near the surface, it will be distorted, stretched, and invariably non-orthogonal.

All the tensor calculus we've developed is precisely what's needed to translate physical laws, like the equations of heat transfer or fluid dynamics, onto these distorted computational grids. While more complex to set up, the payoff is enormous. Numerical simulations on body-fitted grids are vastly more accurate than their Cartesian counterparts for the same number of grid points. As a typical (though hypothetical) example might show, the error in a quantity like the total heat transfer might decrease in proportion to $1/\sqrt{N}$ for a crude Cartesian grid of $N$ cells, but as $1/N$ for a well-designed curvilinear grid. For a simulation with $10,000$ cells, this can mean a hundred-fold improvement in accuracy. This is why non-orthogonal coordinates are an indispensable tool in computational fluid dynamics, weather forecasting, and countless other fields.

And what is the ultimate "body-fitted" coordinate system? It is spacetime itself. In Einstein's theory of General Relativity, gravity is not a force but the curvature of spacetime. The laws of physics must be written in a way that is true no matter what bizarre coordinate system an observer uses. This principle of "general covariance" is the philosophical heart of the theory, and the mathematical engine that drives it is the tensor calculus of non-orthogonal coordinate systems. The metric tensor, our tool for measuring distances, becomes the star of the show; its components, in fact, are the gravitational field.

The utility of choosing the right coordinates extends all the way down to the quantum world of molecules. A molecule is not a static object; its atoms are in constant vibrational motion. A chemist, thinking about the structure of a water molecule, doesn't instinctively reach for a Cartesian grid. Instead, they think in terms of two O-H bond lengths and one H-O-H bond angle. These are the "natural" coordinates of the molecule.

For small vibrations, treating the atoms as moving in straight lines (rectilinear coordinates) is often a decent approximation. But for many crucial chemical processes, this picture fails dramatically. Consider a reaction involving a hydrogen atom transfer, where the transfer is coupled to a large-amplitude twisting motion (a torsion) of the molecular backbone. Describing this floppy, twisting motion with straight-line Cartesian axes is as unnatural as describing a merry-go-round from a fixed spot on the ground—it leads to needlessly complex equations and, worse, physically incorrect results.

The solution is to work from the start in the molecule's natural, curvilinear internal coordinates. Doing so properly separates the large-amplitude torsion from other vibrations and from the overall rotation of the molecule. This leads to far more accurate calculations of the molecule's vibrational frequencies and zero-point energies. For theoretical chemists trying to predict reaction rates and kinetic isotope effects (the change in rate when an atom is replaced by a heavier isotope), this is not a minor correction. It is the difference between a prediction that agrees with experiment and one that is qualitatively wrong. Curvilinear coordinates are essential for accurately modeling the tunneling of atoms through energy barriers and a host of other quantum phenomena that govern the chemical world.

This powerful idea of letting the problem choose its own axes is not just a modern invention. We can find its roots in the work of the ancient Greek geometers. Apollonius of Perga, in his masterwork Conics written over two millennia ago, gave a stunningly elegant definition of the hyperbola.

He essentially defined it using its two asymptotes as axes of a coordinate system. In this oblique frame of reference, the beautifully simple relationship $uv = \text{constant}$ describes every point on the curve. He had no concept of Cartesian coordinates or modern algebra, yet he intuited that the most natural way to understand the hyperbola was through the lens of its own intrinsic symmetries—the asymptotes.

From the timeless beauty of conic sections to the quantum dance of molecules, from the stresses in a bridge to the fabric of the cosmos, the principle remains the same. The courage to abandon the familiar comfort of right angles and adopt the coordinate system that nature provides is one of the most powerful tools we have for understanding the world. It is the art of finding the right point of view.