Barycentric Coordinates

While Cartesian coordinates define a point's position using an external grid, what if we could describe it from within a shape itself? This is the essence of barycentric coordinates, an elegant system that represents any point as a "recipe" or weighted mixture of a shape's corners. This approach moves beyond simple location-finding to provide a powerful language for geometry and physics, bridging the intuitive concept of a center of mass with sophisticated computational methods. This article explores this powerful framework. The first section, "Principles and Mechanisms," demystifies the system using the physical analogy of a center of mass, explains the rules for calculating coordinates, and reveals the geometric story told by their values. The following section, "Applications and Interdisciplinary Connections," demonstrates how this mathematical tool becomes a workhorse in fields like computer graphics and engineering, enabling everything from smooth 3D shading to complex physical simulations.

How do we describe the location of a point? We usually think in terms of a grid—a Cartesian coordinate system. We say, "go three blocks east and two blocks north." This is a wonderful system, but it relies on an external frame of reference. What if we wanted to describe a point's location relative to a shape itself, from the inside? What if we could describe any point within a triangle as a "recipe" or a "mixture" of its corners? This is the central idea behind barycentric coordinates, a system as elegant as it is powerful.

Let's begin with an idea from physics that we can all grasp: the center of mass. Imagine a perfectly flat, weightless triangle with vertices we'll call $v_0$, $v_1$, and $v_2$. Now, let's place little weights at each corner. Suppose we put a 2-unit mass at $v_0$, a 3-unit mass at $v_1$, and a 5-unit mass at $v_2$. Where would this triangle balance? The balance point is the system's ​​center of mass​​.

Instinctively, we know this point will be pulled closer to $v_2$, where the heaviest mass is. The total mass of our system is $M = 2 + 3 + 5 = 10$ units. The influence of each vertex on the balance point is proportional to the fraction of the total mass it holds. Vertex $v_0$ accounts for $\frac{2}{10}$ of the mass, $v_1$ for $\frac{3}{10}$, and $v_2$ for $\frac{5}{10}$.

Just like that, we have just discovered barycentric coordinates. The barycentric coordinates of the center of mass are precisely these fractions: $(\frac{2}{10}, \frac{3}{10}, \frac{5}{10})$, or $(\frac{1}{5}, \frac{3}{10}, \frac{1}{2})$. The position of the center of mass, let's call it $p$, can be written as a weighted average of the vertex positions:

Notice something crucial about these coordinates: they add up to 1. This is no accident. It’s a consequence of them representing fractions of a whole. This simple physical analogy is the heart of barycentric coordinates. They are a "recipe" for any point, telling us how much of each vertex to "mix" together. This principle is wonderfully scalable. If we had a more complex system, say a uniform triangular plate (whose own center of mass is at its geometric center, or ​​barycenter​​, with coordinates $(\frac{1}{3}, \frac{1}{3}, \frac{1}{3})$) plus additional weights at the vertices, we could find the new combined center of mass simply by taking a weighted average of the barycentric coordinates of each component.

Now, let's abstract this away from physics. For any triangle (or more generally, any ​​simplex​​, like a tetrahedron in 3D), we can define a coordinate system for any point $p$ in its vicinity. We say that $p$ has barycentric coordinates $(\lambda_0, \lambda_1, \lambda_2)$ with respect to vertices $(v_0, v_1, v_2)$ if two conditions are met:

1. $p = \lambda_0 v_0 + \lambda_1 v_1 + \lambda_2 v_2$
2. $\lambda_0 + \lambda_1 + \lambda_2 = 1$

The first rule is our recipe for mixing the vertices. The second rule, the ​​normalization condition​​, ensures that the recipe is consistent and, as we will see, that the coordinates are unique.

Let's see this in action. Imagine a triangle with vertices $v_0 = (0, 0)$, $v_1 = (3, 0)$, and $v_2 = (0, 3)$. Where is the point $p = (1, 2)$ in this system? We need to find $(\lambda_0, \lambda_1, \lambda_2)$ such that:

This immediately tells us that $3\lambda_1 = 1$ and $3\lambda_2 = 2$, so $\lambda_1 = \frac{1}{3}$ and $\lambda_2 = \frac{2}{3}$. Using our second rule, $\lambda_0 = 1 - \lambda_1 - \lambda_2 = 1 - \frac{1}{3} - \frac{2}{3} = 0$. So, the barycentric coordinates for the point $(1, 2)$ are $(0, \frac{1}{3}, \frac{2}{3})$. This isn't just a mathematical trick; it's a new way of seeing.

The true beauty of this system is how the numbers themselves tell a geometric story. The values of $(\lambda_0, \lambda_1, \lambda_2)$ act as a map to the triangle and its surroundings.

• ​​At the Vertices:​​ What are the coordinates of vertex $v_1$? To get $p = v_1$, our recipe is simple: take 100% of $v_1$ and 0% of the others. So its coordinates are $(0, 1, 0)$. In general, vertex $v_i$ has coordinates where $\lambda_i=1$ and all other $\lambda_j=0$.

• ​​On the Edges:​​ Our calculated point $(0, \frac{1}{3}, \frac{2}{3})$ had $\lambda_0 = 0$. This means it has no "ingredient" $v_0$ in its recipe. It is made entirely from $v_1$ and $v_2$. Geometrically, this means the point must lie on the line segment connecting $v_1$ and $v_2$. In general, if any single coordinate $\lambda_i = 0$, the point lies on the face (an edge in a triangle, a triangular face in a tetrahedron) opposite to the vertex $v_i$. This extends to higher dimensions; the barycenter of a triangular face spanned by $\{v_1, v_2, v_3\}$ within a 4-dimensional simplex will have coordinates like $(0, \frac{1}{3}, \frac{1}{3}, \frac{1}{3}, 0)$ with respect to the full set of five vertices $\{v_0, \dots, v_4\}$.

• ​​Inside the Triangle:​​ If all coordinates $\lambda_i$ are positive (and sum to 1), the point $p$ is strictly inside the triangle. This is our "center of mass" intuition at play.

• ​​Outside the Triangle:​​ What happens if a coordinate is negative? This is where things get really interesting. A negative coordinate means the point is outside the triangle. Imagine the line forming the edge $v_1v_2$. This line acts as a "zero-contour" for the coordinate $\lambda_0$. On one side of the line (inside the triangle), $\lambda_0$ is positive. On the other side, it must be negative. For example, if we take the barycenter of a triangle, $P$, which is nicely inside, and reflect it across the edge $v_1v_2$, the new point $P'$ lands outside. Its $\lambda_0$ coordinate flips from positive to negative, while $\lambda_1$ and $\lambda_2$ adjust to keep the sum equal to 1. This property is incredibly useful in computer graphics for determining which side of a boundary a point lies on.

Barycentric coordinates don't just describe static positions; they behave with remarkable grace when things move. Suppose a point starts at time $t=0$ at the midpoint of edge $v_0v_1$ and moves in a straight line to arrive at vertex $v_2$ at time $t=1$. The starting point has coordinates $(\frac{1}{2}, \frac{1}{2}, 0)$ and the endpoint has coordinates $(0, 0, 1)$.

What are the coordinates of the point along its journey? Since the path is a straight line (an ​​affine​​ path), the barycentric coordinates also change linearly! The coordinates at any time $t$ are simply a linear interpolation of the start and end coordinates: $p(t) = (1-t)p(0) + t p(1)$. This gives coordinates of $(\frac{1}{2}(1-t), \frac{1}{2}(1-t), t)$. This predictable, linear behavior is a cornerstone of why barycentric coordinates are indispensable for smoothly interpolating color, texture, and other properties across triangular meshes in computer graphics.

This affine nature also provides a beautifully intuitive proof that a simplex is a ​​convex set​​—meaning that the line segment between any two points inside the simplex is also entirely inside. If we take two points, $P_1$ and $P_2$, that are inside a triangle, any point $Q$ on the line segment between them is a "mixture" of $P_1$ and $P_2$. It turns out the barycentric coordinates of $Q$ are the exact same mixture of the coordinates of $P_1$ and $P_2$. Since the coordinates of $P_1$ and $P_2$ were all non-negative, the mixed coordinates for $Q$ will also be non-negative, proving that $Q$ is also inside the triangle.

The connection to standard vector algebra is just as elegant. If we fix one vertex, say $A$, as our origin, the vectors to the other two vertices, $\vec{AB}$ and $\vec{AC}$, form a basis for the plane. Any vector $\vec{AP}$ from $A$ to a point $P$ can be written in terms of this basis. If $P$ has barycentric coordinates $(u, v, w)$, then the vector is simply $\vec{AP} = v\,\vec{AB} + w\,\vec{AC}$. The barycentric coordinates $v$ and $w$ directly tell you how far to "walk" along the edge directions to get to your destination!

Throughout this journey, we've relied on a hidden assumption: that our vertices are well-behaved. For a triangle, this means the three vertices are not on the same line. For a tetrahedron, it means the four vertices are not on the same plane. We call this property ​​affine independence​​.

What happens if we break this rule? Imagine our three vertices $v_0, v_1, v_2$ lie on a single straight line. They no longer define a plane, but merely a line. Can we still describe any point in the plane with them? No. The set of all points we can "cook up" from these three vertices is just the line they all lie on. Our coordinate system has collapsed from 2D to 1D. We can no longer describe any point that is off that line.

Worse, for the points that are on the line, our coordinates are no longer unique. Because the vertices are themselves related (e.g., $v_1$ is a mix of $v_0$ and $v_2$), there are now infinitely many "recipes" to make the same point. It’s like having a redundant ingredient. This ambiguity makes the system useless for unique identification.

Therefore, the requirement of affine independence is not just a mathematical footnote; it is the fundamental rule that guarantees our barycentric coordinate system is both comprehensive (for the space spanned) and unique. It ensures that our map is reliable, a true and faithful guide to the geometry of the simplex.

Now that we have a grasp of what barycentric coordinates are and how they work, you might be asking a fair question: "So what?" Is this just a clever mathematical curiosity, a neat trick for solving geometry puzzles? The answer, which I hope you will find delightful, is a resounding no. The true power and beauty of this idea lie in how it unexpectedly bridges disciplines, connecting intuitive physics, elegant geometry, and the brute-force power of modern computation. It’s a journey from a child's balancing act to the simulation of a jet engine.

The name itself, barycentric, comes from the Greek barys, meaning "heavy." The coordinates' most intuitive origin is in physics, specifically in finding the center of mass, or barycenter. Imagine a weightless triangular plate with vertices $A$, $B$, and $C$. If you place masses $m_A$, $m_B$, and $m_C$ at these vertices, the triangle will balance perfectly at a single point $P$. The location of this point is the weighted average of the vertex positions. If we normalize these masses such that their sum is one, these normalized weights are precisely the barycentric coordinates of the balance point.

This physical picture is more than just an analogy. Consider a particle trapped in a plane, tethered by three springs to points $A$, $B$, and $C$. Each spring pulls with a force proportional to its extension, with "spring constants" $k_A$, $k_B$, and $k_C$. Where will the particle come to rest? It will settle at the unique equilibrium point where the three forces cancel out. A little algebra reveals something wonderful: the equilibrium position is given by $\vec{p} = u\vec{A} + v\vec{B} + w\vec{C}$, where the barycentric coordinates $(u, v, w)$ are simply the normalized spring constants: $u = k_A / (k_A+k_B+k_C)$, and so on. The stiffer the spring at a vertex, the closer the equilibrium point is pulled towards it. The physics of static equilibrium is written in the language of barycentric coordinates.

This is where the magic begins. We can flip the script. Instead of using masses to find a point, we can use a point to define a set of unique "virtual masses" or weights—its barycentric coordinates. This simple shift in perspective transforms a physical concept into a powerful and universal language for geometry.

Suddenly, problems that would cause headaches in a traditional Cartesian $(x,y)$ system become astonishingly simple. Want to find the midpoint of an edge, say between $B$ and $C$? In this new language, it's just the point where the "masses" are split equally between those two vertices and zero at the third: $(0, \frac{1}{2}, \frac{1}{2})$. Lines and ratios, the bread and butter of classical geometry, are handled with an algebraic elegance that would make Euclid weep. For instance, if you draw a line (a cevian) from vertex $A$ through an interior point $P$ with coordinates $(\lambda_A, \lambda_B, \lambda_C)$, where does it hit the opposite side $BC$? The ratio in which it divides that side is simply $\lambda_C / \lambda_B$. No angles, no sine laws, just a simple ratio of coordinates. Finding the intersection of two such cevians? It's no longer a geometric construction but a straightforward algebraic calculation.

This language is not disconnected from the familiar world of Cartesian coordinates. Any straight line, which we know as $ax+by+c=0$, can be translated directly into a barycentric equation of the form $k_A \lambda_A + k_B \lambda_B + k_C \lambda_C = 0$. The coefficients $k_A, k_B, k_C$ are found by simply evaluating the Cartesian line function at each of the triangle's vertices. We can even describe more complex geometric objects. The locus of points equidistant from two vertices, $A$ and $B$—what we know as the perpendicular bisector—is also a straight line with a simple linear equation in barycentric coordinates. A geometric property (collinearity of three points) becomes a simple algebraic test: three points are on the same line if and only if the determinant of the matrix formed by their barycentric coordinates is zero. This transforms geometric intuition into computable arithmetic.

If barycentric coordinates were only a tool for elegant geometry, they would be a beautiful footnote in mathematics. But their true role in the modern world is far grander. They are a workhorse of computational science.

Have you ever wondered how a 3D video game renders a smoothly shaded triangle? The artists define colors at the vertices, and the graphics card fills in the millions of pixels inside. How? For each pixel, it calculates its barycentric coordinates relative to the on-screen triangle. These coordinates then act as weights to interpolate the vertex colors, normals, and texture coordinates, resulting in a smooth, continuous image. What you are seeing is a direct, real-time application of barycentric interpolation.

The most profound application, however, is in the Finite Element Method (FEM), a cornerstone of modern engineering. How do you predict the stress in a bridge support, the heat flow in a nuclear reactor, or the airflow over an airplane wing? These are objects with complex shapes governed by complex differential equations. The FEM's strategy is one of "divide and conquer": break the complex shape down into a mesh of millions of simple, tiny elements—often triangles (in 2D) or tetrahedra (in 3D).

Within each tiny triangular element, barycentric coordinates are not just a coordinate system; they are the solution. They are what engineers call ​​shape functions​​. Any physical quantity—temperature, pressure, displacement—is approximated within that element as a weighted sum of the values at the vertices, and the weights are precisely the barycentric coordinates. The property that the coordinates sum to one (the "partition of unity") is crucial, ensuring that the approximation is consistent and well-behaved.

Furthermore, this system has a property that can only be described as a superpower: ​​affine invariance​​. This means that the barycentric coordinates of a point inside a triangle do not change if you translate, rotate, scale, or shear the triangle. The physical laws we formulate using this system are inherently independent of the specific shape or orientation of our little triangular elements. This robustness is what allows engineers to mesh complex, curved geometries and still trust that their physical simulations are correct.

When we move beyond simple scalar fields like temperature and need to describe vector quantities like velocity or force, barycentric coordinates provide the framework for a local, non-orthogonal coordinate system. We can determine how the components of a vector in the global Cartesian system transform into local "contravariant" components within an element. And if we need to perform calculus—for example, to integrate the total energy over an element—we need a way to relate an infinitesimal area $dx\,dy$ to its corresponding area in barycentric space. The "exchange rate" for this transformation is the Jacobian determinant, which, beautifully, is a constant related to the area of the triangle itself.

From balancing weights on a platter to describing the fundamental fields of physics inside a supercomputer, the journey of barycentric coordinates is a testament to the unity of scientific thought. An idea born from simple physical intuition provides a new language for geometry, and this language, in turn, becomes the bedrock for some of the most sophisticated computational tools humanity has ever devised. It is a perfect example of how the abstract and the applied are not separate worlds, but two sides of the same, elegant coin.