21.9.2. Orthonormal Set

Two vectors u and v are orthogonal if: $u � v = 0$

Two vectors are considered orthogonal when their dot product equals zero. This means that they meet at a right angle, or 90 degrees, when you visualize them in a geometric space such as 2D or 3D. Orthogonality is a key concept in linear algebra as it indicates that the vectors have no component in the direction of each other.

Think of orthogonal vectors like the x-axis and y-axis on a graph. The x and y axes intersect at a right angle and do not influence each other’s direction. If you were to project a point onto the x-axis from the y-axis, it would have no movement along the y-axis, which exemplifies the idea of orthogonality.

An orthonormal set is a set of vectors that are both orthogonal and unit vectors.

An orthonormal set of vectors is not only made up of orthogonal vectors, but each vector also has a length (or magnitude) of one. This means that if you take any vector from this set and calculate its length, you will find it equals one. This property simplifies many calculations in linear algebra, particularly in areas like Fourier transforms and other vector space analyses.

Imagine a team of players in a game where each player has a specific role but also needs to reach specific positions on the field without interfering with each other. Like the positions occupied by players on a soccer field, where each occupies their own space (orthogonal) and needs to maintain their position effectively (unit vectors), an orthonormal set of vectors has each vector in its own space with a set 'unit' position.

A method to convert a set of linearly independent vectors into an orthonormal set.

The Gram-Schmidt process is an algorithm used to take a collection of linearly independent vectors and convert them into an orthonormal set. This involves systematically orthogonalizing them and then normalizing each vector so its length is 1. The result is a new set of vectors that maintain the same span as the original set but are now orthonormal.

Think of making a pizza where each topping represents a vector. You may start with different toppings (linearly independent vectors) and want to ensure that they balance evenly on the pizza (being orthonormal). The Gram-Schmidt process is like adjusting the toppings and their positioning so that they complement each other perfectly while also fitting neatly into the available space without overwhelming each other. The finished pizza is well-balanced (orthonormal set) and still keeps the essence of the original flavors (span of the original vectors).

Applications: Numerical solutions of partial differential equations, finite element methods in structural analysis.

Orthonormal sets are highly useful in various numerical methods and engineering applications. For example, they simplify the representation of functions in numerical solutions for partial differential equations, helping to reduce computational complexity. In structural analysis, finite element methods use orthonormal bases to describe displacement fields, enabling precise calculations of stress and strain in structures.

Consider the use of orthonormal sets like having a toolbox with specific tools designed for specific tasks. Each tool is 'calibrated' (orthonormal) to fit its function perfectly (like discrete points in a numerical solution) so that the overall task (like constructing a structure) can be completed efficiently and accurately without interferences that could lead to errors.

--