Coordinate Basis

In our everyday experience with geometry, we rely on a simple grid of perpendicular lines, where "up" and "right" are constant and universal. But how do we describe motion and geometry on a curved surface like a sphere, or within a specialized system like polar coordinates? Our standard notions of basis vectors fail us in these more complex worlds. This article bridges that gap, introducing the ​​coordinate basis​​ as a powerful, flexible concept essential for modern physics and geometry. It redefines a basis vector not as a static pointer, but as a dynamic, local direction of motion. In the following chapters, we will first delve into the principles and mechanisms of coordinate bases, exploring how to define these vectors, use the metric tensor to measure distances, and transform between different descriptive frameworks. We will then journey through a wide range of applications and interdisciplinary connections to see how this single idea brings clarity to fields as diverse as computer graphics, crystallography, and Einstein's theory of relativity, revealing the deep structure of our world.

Imagine you are an ant on a vast, gently curved sheet of paper. How would you describe your world? You could lay down a simple Cartesian grid of perpendicular lines, and your basis for movement would be simple: "one step east" and "one step north." These steps are always the same length and always at right angles to each other. This is the world of high school geometry, comfortable and familiar. But what if your world isn't a flat sheet? What if it's a sphere, or a saddle shape, or some complex, warped surface? Or what if, even on a flat plane, you decide to use a different system, like the circles and radial lines of polar coordinates? Suddenly, "one step east" isn't a universal concept. Your directions and the very meaning of "a step" change depending on where you are. This is the world of differential geometry, and at its heart is the wonderfully flexible idea of a ​​coordinate basis​​.

Let's get rid of a common misconception right away. In a curvy or non-standard coordinate system, a basis vector is not a pointer from some central origin to your current location. Instead, it's something much more dynamic and local.

Imagine yourself at a point $(r, \phi)$ in a polar coordinate system. What is the basis vector associated with the radius, $\partial_r$? The most fundamental way to think about it is as a recipe for motion: ​​it is the velocity vector you would have if you decided to increase the coordinate $r$, and only the coordinate $r$, while keeping $\phi$ perfectly constant​​. If you are at $(r_P, \phi_P)$, moving along this direction means walking straight out along the radial line passing through your position. Similarly, the basis vector $\partial_\phi$ is the velocity you'd have if you kept your distance $r$ fixed and started to walk along the circle of that radius.

These basis vectors, $\{\partial_r, \partial_\phi\}$, define a "local grid" at every single point. They form a set of directions that are tangent to the coordinate lines passing through that point. Formally, we define these basis vectors as the partial derivatives of the position vector $\vec{p}$ with respect to each coordinate. For a general coordinate system with coordinates $q^i$, the $i$-th basis vector is:

For example, in polar coordinates where the position is $\vec{p} = (r\cos\theta)\hat{i} + (r\sin\theta)\hat{j}$, the basis vectors are indeed the tangent vectors we pictured:

Notice something interesting? The basis vector $\vec{e}_r$ has a length of 1, but the length of $\vec{e}_\theta$ is $|\vec{e}_\theta| = \sqrt{(-r\sin\theta)^2 + (r\cos\theta)^2} = r$. The "step" you take in the $\theta$ direction depends on how far you are from the origin! This is a general feature of coordinate bases: they are typically not of unit length, and their lengths can change from place to place. The area of the tiny parallelogram formed by these basis vectors tells us how much surface area a small change $d\theta$ and $d\phi$ covers on a sphere. This area turns out to be $r^2 \sin\theta$, which explains why the area element in spherical coordinates is $dA = r^2\sin\theta\,d\theta\,d\phi$. At the origin ($r=0$) or the poles ($\sin\theta=0$), this area vanishes, a sign that the coordinate system itself is becoming degenerate or ill-defined at those special points.

So far, we've pictured basis vectors as little arrows tangent to coordinate lines. This is a perfectly good picture. But there is a deeper, more powerful way to view them: as ​​operators​​. A basis vector like $\frac{\partial}{\partial \phi}$ can be thought of as a machine that takes in any scalar field (a function that assigns a number to every point in space, like temperature) and tells you how fast that function is changing as you move purely in the $\phi$ direction.

Let's see this in action. Consider the function $g(x,y,z) = x$, which simply reports the Cartesian $x$-coordinate of any point. In cylindrical coordinates $(\rho, \phi, z)$, this function is written as $g = \rho\cos\phi$. Now, let's "act" on this function with the basis vector operator $\frac{\partial}{\partial \phi}$:

What does this result, $-\rho\sin\phi$, mean? It tells us that if we are at some point and take a small step purely in the direction of increasing $\phi$ (moving along a circle), the rate at which our $x$-coordinate changes is $-\rho\sin\phi$ (which happens to be just $-y$). This abstract-seeming definition is incredibly powerful because it frees us from having to visualize arrows in some background Euclidean space. It defines the basis vectors intrinsically, by what they do to functions living on the space. This is the language of modern geometry and physics.

We've seen that coordinate basis vectors can have varying lengths and might not be perpendicular. This seems like a disaster! How can we measure distances or angles if our fundamental rulers and protractors are constantly changing? The hero of this story is the ​​metric tensor​​, denoted $g_{ij}$.

The metric tensor is a collection of functions that provides the complete geometric information of the space. It's the "universal dot product machine." Given any two basis vectors $\vec{e}_i$ and $\vec{e}_j$ at a point, the component $g_{ij}$ of the metric tensor is their dot product:

This simple definition is the key to everything.

• ​​Lengths of Basis Vectors:​​ The length (or magnitude) of a basis vector $\vec{e}_i$ is $\sqrt{\vec{e}_i \cdot \vec{e}_i} = \sqrt{g_{ii}}$. The diagonal components of the metric tell you the squared lengths of your basis vectors.
• ​​Angles Between Basis Vectors:​​ The angle $\theta_{ij}$ between two basis vectors $\vec{e}_i$ and $\vec{e}_j$ is given by the familiar dot product formula: $\cos(\theta_{ij}) = \frac{\vec{e}_i \cdot \vec{e}_j}{|\vec{e}_i| |\vec{e}_j|} = \frac{g_{ij}}{\sqrt{g_{ii}g_{jj}}}$. The off-diagonal components tell you how non-orthogonal your system is.

If a coordinate system is ​​orthogonal​​, like standard spherical or cylindrical coordinates, then all its basis vectors are mutually perpendicular. This means their dot products are zero for $i \neq j$, so all the off-diagonal components of the metric, $g_{ij}$, are zero. The metric tensor matrix becomes beautifully simple: a ​​diagonal matrix​​.

What if the metric is not diagonal? That's just nature's way of telling you that your chosen coordinate lines do not meet at right angles. We can calculate the exact angle between them. For instance, given a metric with a component $g_{uv} = \cos(\frac{\pi u}{2})$, we know immediately that the basis vectors $\partial_u$ and $\partial_v$ are not orthogonal. We can use the formula above to find the precise angle between them at any point, revealing the local "skew" of our grid.

Once we have the metric, we can find the magnitude of any vector. If a vector $\mathbf{u}$ has components $u^i$ in our coordinate basis (so $\mathbf{u} = u^i \vec{e}_i$), its squared magnitude is no longer a simple sum of squares. It is a weighted sum, where the metric components are the weights:

(using the Einstein summation convention where repeated indices are summed over). This formula is fundamental. It lets us calculate the speed of a particle moving on a helical path in cylindrical coordinates, for example, directly from its coordinate-velocity components and the cylindrical metric $ds^2 = d\rho^2 + \rho^2 d\phi^2 + dz^2$.

A vector, like the velocity of a swirling fluid, is a physical, geometric object. It exists independent of any coordinate system we might invent. But its components—the numbers we use to describe it—depend entirely on the basis we choose. It's like describing a person: the person is the same whether you list their height in feet or meters. The numbers change, but the physical reality does not.

Switching from one basis to another is a matter of applying the chain rule. For instance, we can express the spherical basis vector $\frac{\partial}{\partial \phi}$ in terms of the Cartesian basis $\{\frac{\partial}{\partial x}, \frac{\partial}{\partial y}, \frac{\partial}{\partial z}\}$:

This tells us exactly how to "translate" the direction of pure azimuthal motion into the language of Cartesian directions. This allows us to take a vector field described in one coordinate system, like a fluid flow $\vec{V} = k(-y\hat{i} + x\hat{j})$, and find its components in another, like polar coordinates. We find that this specific flow is purely in the angular direction, with $\vec{V} = k \vec{e}_\theta$, meaning its contravariant component $V^\theta$ is just the constant $k$.

This raises an important distinction. The ​​coordinate basis​​ is natural and fundamental to the coordinates themselves, but it can be awkward (non-orthogonal, non-unit-length). For practical physics, we often want a local ​​orthonormal frame​​—a set of perfectly perpendicular, unit-length basis vectors $\{\vec{e}_{(1)}, \vec{e}_{(2)}, \dots\}$ at each point. This is like a physicist carrying around a personal set of gyroscopes that always define "up," "north," and "east" locally. How do we get such a frame? We can build it directly from the coordinate basis using the trusty Gram-Schmidt procedure, using our metric tensor $g_{ij}$ as the dot product at every step. This gives us the best of both worlds: the global structure of coordinates and the local convenience of an orthonormal frame.

The most profound consequence of all this is that the coordinate basis vectors themselves are not constant. They form vector fields. The direction and length of $\partial_r$ and $\partial_\theta$ in polar coordinates depend on where you are. This seems obvious, but it has a deep implication.

What happens if you try to carry a vector, keeping it "pointing in the same direction," as you move along a curve? If your basis vectors are twisting and turning underneath you, the components of your "constant" vector must change to compensate. The study of how the basis vectors themselves change from point to point leads to the concept of the ​​covariant derivative​​ and ​​Christoffel symbols​​.

Even in a perfectly flat plane, a "weird" coordinate system like parabolic coordinates will have basis vectors that appear to rotate as you move along a coordinate line. This change is part geometry, part artifact of our coordinate choice. The magic of a concept like the Riemann curvature tensor (the subject for another day!) is that it is the tool that can distinguish the change that comes from our quirky coordinate system from the change that comes from the space actually being curved. And with that, we have moved from simply mapping a space to understanding its very essence. The humble coordinate basis vector, understood as a local direction of motion, is the first step on this grand journey.

We have spent some time getting to know the machinery of coordinate bases, how to define them, and how they relate to the underlying structure of a space. You might be tempted to think this is just a formal exercise for mathematicians—a way of tidying up our definitions. But nothing could be further from the truth! The idea of a basis, and more importantly, the freedom to change your basis, is one of the most powerful and practical tools in the entire arsenal of science and engineering. It's the secret to taking a problem that looks messy and complicated from one point of view and making it look beautifully simple from another. It’s like discovering that a seemingly chaotic musical piece is actually just a few simple melodies played on top of each other. The trick is to find the right "basis" of melodies.

Let's take a journey through some of the surprising places where this single idea brings clarity and power.

Our first stop is the vibrant, dynamic world inside our computers. Every time you play a video game, use a design program, or watch an animated movie, you are witnessing a symphony of basis transformations.

Imagine a rover exploring a digital landscape in a video game. The rover itself has a natural sense of "forward" and "right". A command like "move 5 units forward and 2 units right" is perfectly clear in its local coordinate system. But the game world has its own master coordinate system—perhaps "North" and "East". To show the rover moving on the screen, the game engine must constantly translate the rover's local coordinates into the world's coordinates. This is a change of basis in action! The "local basis" vectors (forward, right) are expressed in terms of the "world basis" vectors (North, East), and a simple matrix multiplication does the trick, allowing the rover to turn and move freely while the game keeps track of where it is in the grand scheme of things.

This isn't just for moving objects. The very concept of "local" versus "world" space is fundamental to all modern 3D graphics. An artist designs a character model in its own local coordinate system, centered on itself. When that character is placed in a scene, its coordinates are transformed into the world system. When the camera moves, all world coordinates are transformed into the camera's coordinate system for rendering. It's a chain of basis changes, from model to world to camera to screen.

The power of choosing a basis extends beyond just geometry. Think about how information itself is stored. A smooth curve, for instance, might be represented as a polynomial. We could store the coefficients of $1, t, t^2$, which forms a standard basis. But for certain hardware, it might be much more efficient to use a different basis, like one made of polynomials such as $1, (t-1), (t-1)^2$. Converting from the storage format to the rendering format is, once again, nothing more than a change of basis in the abstract vector space of polynomials. Or consider analyzing a picture of woven fabric. The standard horizontal and vertical pixel grid might be a poor choice of coordinates if the threads are woven at an angle. By defining new basis vectors that align with the weave directions, an algorithm can analyze the pattern much more efficiently and naturally. In all these cases, we are not changing the object or the data itself; we are changing our description of it to suit our purpose.

The idea of a basis isn't just an abstract convenience; it's written into the very structure of the physical world. Let's zoom in, way down to the atomic scale.

Consider a perfect crystal, like a grain of salt or a diamond. At first glance, it's a bewilderingly complex arrangement of countless atoms. But the secret to this complexity is a profound simplicity. A crystal structure is nothing more than a repeating grid of points in space, called a lattice, and a "basis" of one or more atoms that is placed, identically, at every single one of those lattice points. For example, the common Body-Centered Cubic (BCC) structure can be described as a simple cubic lattice, with a two-atom basis: one atom at the lattice point itself (coordinates $(0,0,0)$ relative to the point) and another atom in the very center of the cube (coordinates $(\frac{1}{2}, \frac{1}{2}, \frac{1}{2})$). The entire, seemingly infinite crystal is generated by just two things: the simple repeating lattice and this small cluster of atoms that serves as the basis. It’s an astonishing piece of natural economy.

Now, let's look at a single molecule, like sulfur dioxide ($\text{SO}_2$). It can vibrate and wiggle in all sorts of ways. We could try to describe this motion by tracking the stretching of each of the two sulfur-oxygen bonds. But these motions are coupled and complicated. The real magic happens when we change our basis. Instead of "stretch of bond 1" and "stretch of bond 2," we can define a new basis: a "symmetric stretch" (where both bonds lengthen and shorten together) and an "antisymmetric stretch" (where one bond lengthens as the other shortens). In this new basis of "normal modes," the complex wiggling motion decomposes into two simple, independent harmonic oscillations. The laws of quantum mechanics and group theory provide a precise mathematical tool, the projection operator, to find exactly these right basis vectors that simplify the physics.

So far, our basis vectors have been straight arrows living in a flat space. But what if the space itself is curved? The concept of a coordinate basis not only survives but becomes even more essential, allowing us to navigate the warped landscapes of geometry and Einstein's theory of relativity.

Imagine a surface like a catenoid—the shape a soap film makes between two rings. We can lay down a coordinate grid on this surface, say with coordinates $u$ and $v$. At any point, we can ask: what direction do I move in if I change only $u$? That defines a tangent basis vector, $\sigma_u$. What if I change only $v$? That defines another, $\sigma_v$. This pair of vectors, $\sigma_u$ and $\sigma_v$, forms a coordinate basis for the tangent space at that point. Unlike in flat space, these basis vectors change their direction and length as we move from point to point on the surface. The geometry of the curved surface is entirely encoded in how these basis vectors behave—specifically, in their dot products. For the catenoid, it turns out that these natural basis vectors are beautifully orthogonal everywhere on the surface, which reveals a deep geometric property of this shape.

This idea is the heart of differential geometry. Any coordinate system you choose, even a "skewed" one where the basis vectors are not orthogonal, gives you a valid way to describe the geometry. The key is the metric tensor, a collection of all the inner products between your chosen basis vectors. The metric tells you how to measure distances and angles in your specific coordinate system.

This brings us to a profound connection: symmetries. If you find a coordinate system where the components of the metric tensor don't depend on one of the coordinates (say, the $x$ coordinate), it means you can move along that coordinate direction without the geometry changing at all. You've found a symmetry of the space! The coordinate basis vector corresponding to that direction, $\partial_x$, is called a Killing vector field, and it is a mathematical manifestation of this continuous symmetry.

The ultimate playground for these ideas is Einstein's theory of relativity. In the four-dimensional world of spacetime, our choice of coordinates is our choice of reference frame. The physics must be the same regardless of the basis we choose. But some bases are more insightful than others. In the flat spacetime of special relativity, we usually use time and space coordinates ($t, x, y, z$). But it's incredibly useful to switch to a basis that is aligned with the paths of light rays. These are "null" vectors—vectors whose length in spacetime is zero. It's possible to define a coordinate system where the basis vectors themselves are null vectors! This "light-cone coordinate" system simplifies the equations of relativity and makes the causal structure of spacetime—what events can influence what other events—crystal clear.

Finally, let's return to the world of engineering, to the abstract spaces of control theory. The behavior of a complex LTI (Linear Time-Invariant) system—like a cruise control circuit or a chemical process regulator—can be described by a mathematical object called a transfer function. It turns out that all possible transfer functions for a given system form a vector space.

Just as with polynomials, we can choose different bases for this space. A straightforward choice is a "monomial basis," which is easy to write down but doesn't tell you much about the system's behavior. A much more insightful choice is a "partial fraction basis." Each basis vector in this new system corresponds to a fundamental response mode of the system—how it naturally tends to oscillate or decay over time. Changing from the monomial basis to the partial fraction basis is a standard technique that allows an engineer to immediately see the system's core characteristics. The coordinates of the transfer function in this new, physically-motivated basis tell the engineer the "strength" of each mode, which is crucial for analyzing stability and designing controllers.

From the pixels on a screen to the atoms in a crystal, from the vibrations of a molecule to the symmetries of spacetime itself, the humble coordinate basis is a key that unlocks a deeper understanding. It teaches us that how we choose to describe the world can be just as important as the world itself. It is a testament to the power of finding the right point of view.