Geometric Interpretration of the Cauchy–Schwarz Inequality

The Cauchy–Schwarz inequality is usually introduced as a deep structural fact about inner product spaces. In standard courses, the inequality is proved algebraically from the axioms of an inner product. However, it is a more fundamental geometric fact.

In this blog post, I will show that the Cauchy–Schwarz inequality simply tells us projecting a vector onto another cannot increase its length. To make this work, we need introduce two properties to a vector space:

1. Symmetric bilinear form (enables orthogonal decomposition)
2. Positive definiteness (ensures all vectors have non-negative squared length)

The bilinear form $\langle \cdot, \cdot \rangle$ is not limited to the standard dot product. For example, for continuous, integrable functions $f,g : [0,1] \to [0, 1]$, we can define $\langle f, g \rangle = \int_0^1 f(x)g(x)\,dx$, which gives a weighted inner product on a function space. The Cauchy–Schwarz inequality holds for any such inner product.

From these properties alone, we can show Cauchy–Schwarz represents the cosine of an angle between two vectors.

Setup

Let $V$ be a vector space with a symmetric, positive definite bilinear form $\langle \cdot, \cdot \rangle : V \times V \to \mathbb{R}$. Define the induced norm $\|u\| = \sqrt{\langle u, u \rangle}$.

Positive definiteness means:

• $\langle u, u \rangle \geq 0$ for all $u \in V$
• $\langle u, u \rangle = 0$ if and only if $u = 0$

Symmetry and positive definiteness together imply Pythagoras’ theorem: if $\langle u, v \rangle = 0$ (orthogonality), then

This follows from expanding $\langle u+v, u+v \rangle = \langle u,u \rangle + \langle u,v \rangle + \langle v,u \rangle + \langle v,v \rangle$. By symmetry, $\langle u,v \rangle = \langle v,u \rangle$, so when $\langle u,v \rangle = 0$, the middle terms vanish and we get $\|u+v\|^2 = \|u\|^2 + \|v\|^2$.

Orthogonal Projection and Pythagoras

For non-zero $v \in V$, define the orthogonal projection of $u$ onto $v$:

This is the unique vector parallel to $v$ such that $u - \text{proj}_v(u)$ is orthogonal to $v$:

We can decompose $u$ into parallel and perpendicular components:

By Pythagoras:

Deriving Cauchy–Schwarz

By positive definiteness, $\|u - \text{proj}_v(u)\|^2 \geq 0$, so:

Compute the norm of the projection:

Therefore:

Rearranging:

This is the Cauchy–Schwarz inequality:

Equality holds if and only if $u - \text{proj}_v(u) = 0$, meaning $u$ is parallel to $v$.

The Cosine Interpretation

The projection naturally defines an angle between $u$ and $v$. Since the projection has a certain length and $u$ has length, define:

This is the ratio of the projection’s length to the original vector’s length. From our earlier calculation:

so:

Therefore:

The Cauchy–Schwarz inequality is now transparent: it simply says the cosine is at most 1, i.e., the projection cannot be longer than the original vector.

Accounting for sign (parallel vs anti-parallel), we get:

This is the law of cosines in its inner product form. The bilinear form measures the “alignment” between vectors where the angle can be 0 (parallel), $\pi/2$ (orthogonal), or $\pi$ (anti-parallel).

Implications

• If $u$ and $v$ have zero mean, then: \(\cos \theta = \frac{\langle u, v \rangle}{\|u\| \|v\|}\) is the Pearson correlation coefficient between $u$ and $v$.

• The Cauchy–Schwarz inequality becomes an equality when $u - \text{proj}_v(u) = 0$, i.e., when $u$ is parallel to $v$ (or anti-parallel). This happens when:
• $\cos \theta \approx 1$ (high positive correlation), the vectors are nearly parallel and the bound is nearly saturated

• $\cos \theta \approx -1$ (high negative correlation), the vectors are nearly anti-parallel and the bound is nearly saturated

• When $\cos \theta = 0$ (orthogonality/no correlation), the bound becomes loose: the left side is zero while the right side can be arbitrarily large.

Conclusion

We have shown that the Cauchy–Schwarz inequality is the geometric statement that projecting a vector onto another cannot increase its length.