Complex numbers, with their one real and one imaginary part, are great for describing rotations in a 2D plane. But what if we want to work in three dimensions? In the 19th century, Irish mathematician William Rowan Hamilton wrestled with this problem. His solution was a new number system: quaternions.

A quaternion is an extension of complex numbers. It has one real part and three distinct imaginary parts, denoted by $i$, $j$, and $k$. The general form of a quaternion $q$ is:

$q = a + bi + cj + dk$

Here, $a$, $b$, $c$, and $d$ are real numbers. The term $a$ is called the scalar part, and $bi + cj + dk$ is the vector part.

The magic of quaternions comes from the rules governing how the imaginary units multiply. While $i^2 = -1$ is familiar from complex numbers, quaternions add a few more rules for $j$ and $k$.

From this single identity, we can figure out the other multiplication rules. For instance, if $ijk = -1$, we can multiply both sides by $k$ on the right:

$ijk(k) = -1(k) \implies ij(k^2) = -k \implies ij(-1) = -k \implies ij = k$

But what about $ji$? It turns out that quaternion multiplication is not commutative. The order matters. Reversing the order of multiplication flips the sign.

Working with quaternions involves a few basic operations. Addition is straightforward, while multiplication requires careful attention to the rules we just covered.

Addition To add two quaternions, you simply add their corresponding parts. It's just like adding vectors or complex numbers. If you have two quaternions:

$q_1 = a_1 + b_1i + c_1j + d_1k$ $q_2 = a_2 + b_2i + c_2j + d_2k$

Their sum is:

Multiplication Multiplying quaternions is more involved because you have to use the distributive property and apply the multiplication rules for $i, j,$ and $k$. This process reveals the non-commutative nature of quaternions. The full formula is quite long, but the key is to multiply each term in the first quaternion by each term in the second and then simplify.

Conjugation The conjugate of a quaternion, denoted as $q^*$, is found by negating its vector part. If $q = a + bi + cj + dk$, its conjugate is:

Conjugation is useful for finding the magnitude of a quaternion and for calculating its inverse, which are important steps in using quaternions for rotations.

The most significant application of quaternions is representing rotations in 3D space. While rotation matrices can also do this, quaternions are more compact and computationally efficient. They avoid a problem known as "gimbal lock," a loss of one degree of rotational freedom that can plague other rotation systems.

To represent a rotation, we use a special type of quaternion called a unit quaternion, which has a magnitude of 1. A rotation of angle $\theta$ around a unit vector axis $\vec{u} = (u_x, u_y, u_z)$ is represented by the quaternion $q$:

To apply this rotation to a point $P$ in space, you represent the point as a "pure" quaternion $p = 0 + P_x i + P_y j + P_z k$. The rotated point $P'$ is then found by the following quaternion multiplication:

This calculation, known as a sandwich product, neatly applies the rotation. The result, $p'$, is another pure quaternion whose vector part gives the new coordinates of the rotated point. This elegant method is why quaternions are fundamental to 3D computer graphics, robotics, and aerospace engineering.

Quaternions offer a powerful way to think about numbers and geometry, connecting algebra directly to the rotations we see in the physical world.