Cauchy–Schwarz and Geometry of Hilbert Space
~5 min read · Tier toggle changes depth (T1 plain · T2 standard · T3 rigour)
You should be comfortable defining the inner product \(\langle u, v \rangle\) for real and complex vector spaces, understanding linearity in the first argument, and calculating the induced norm \(\|v\| = \sqrt{\langle v, v \rangle}\).
Why this matters
Cauchy–Schwarz is the bridge between algebraic manipulation and geometric intuition in abstract spaces.
In finite-dimensional Euclidean space, we visualize angles and lengths using dot products. In the infinite-dimensional Hilbert spaces of quantum mechanics, we cannot "see" vectors. The Cauchy–Schwarz inequality provides the rigorous algebraic foundation that ensures our abstract definitions of "angle" and "distance" behave exactly like their geometric counterparts. Without it, the concept of an inner product space would lack the structural integrity required for spectral theory and probability amplitudes.
The Geometric Meaning
Imagine two vectors in a plane. The inner product \(\langle u, v \rangle\) measures how much \(v\) "projects" onto \(u\). If \(v\) is perpendicular to \(u\), the projection is zero. If they point in the same direction, the projection is maximal. The Cauchy–Schwarz inequality simply states that the projection of one vector onto another can never exceed the length of the vector being projected. It is a statement about the limits of alignment. In complex spaces, where "direction" is subtler, it bounds the magnitude of the overlap between two states, ensuring that probabilities (which are squared magnitudes) remain physically sensible and bounded by unity.
The Inequality and Equality Conditions
The core result, derived in D-076, is that for any vectors \(u, v\) in an inner product space:
This inequality is sharp. The equality condition is crucial for identifying linear dependence. If \(\left|\langle u, v \rangle\right| = \|u\| \, \|v\|\), then \(u\) and \(v\) are linearly dependent. In the context of quantum states, this means that if two normalized states achieve the maximum possible overlap, they represent the same physical state (up to a global phase). Conversely, if the inner product is zero, the vectors are orthogonal, a condition that will be central to the next lecture on Gram–Schmidt orthogonalization.
Connection to the Triangle Inequality
A norm must satisfy the triangle inequality \(\|u + v\| \le \|u\| + \|v\|\) to define a valid metric space. The Cauchy–Schwarz inequality is the algebraic engine that proves this. By expanding \(\|u+v\|^2 = \langle u+v, u+v \rangle\), we obtain terms involving \(\langle u, v \rangle\). Applying Cauchy–Schwarz to these cross-terms allows us to bound the expansion by \((\|u\| + \|v\|)^2\). Taking the square root yields the triangle inequality. This confirms that the inner product induces a genuine geometric structure where the "straight line" distance between two points is always the shortest path.
Complex Spaces and Phase
In complex Hilbert spaces, the inner product is not symmetric but conjugate-symmetric: \(\langle u, v \rangle = \overline{\langle v, u \rangle}\). The Cauchy–Schwarz inequality applies to the modulus \(\left|\langle u, v \rangle\right|\). This is vital because the inner product itself can be complex. The physical interpretation relies on the magnitude of the overlap. For example, in quantum mechanics, the probability of finding a system in state \(v\) given it is in state \(u\) is \(|\langle u, v \rangle|^2\). Cauchy–Schwarz guarantees that if \(u\) and \(v\) are normalized, this probability is always less than or equal to 1, preventing unphysical results such as probabilities exceeding unity.
The derivation
Cauchy–Schwarz and the Triangle Inequality
In any real or complex inner-product space \((V,\langle\cdot,\cdot\rangle)\) with induced norm \(\|x\|=\sqrt{\langle x,x\rangle}\), every pair of vectors \(u,v\in V\) satisfies the Cauchy–Schwarz inequality \(\left|\langle u,v\rangle\right|\le \|u\|\,\|v\|\), with equality if and only if \(u\) and \(v\) are linearly dependent. From this the norm obeys the triangle inequality \(\|u+v\|\le\|u\|+\|v\|\), so \(\|\cdot\|\) is a genuine norm.
Students often confuse the condition for equality in Cauchy–Schwarz with orthogonality. Remember: equality in \(\left|\langle u, v \rangle\right| \le \|u\| \, \|v\|\) implies linear dependence (parallel or anti-parallel), whereas \(\langle u, v \rangle = 0\) implies orthogonality. These are distinct geometric relationships. Confusing them leads to errors in constructing orthogonal bases. C
Worked example
Problem. Consider the complex vector space \(\mathbb{C}^2\) with the standard inner product. Let \(u = \begin{pmatrix} 1 \\ i \end{pmatrix}\) and \(v = \begin{pmatrix} 2 \\ 1 \end{pmatrix}\). Calculate \(\|u\|\), \(\|v\|\), and \(\left|\langle u, v \rangle\right|\). Verify the Cauchy–Schwarz inequality and determine if equality holds.
- Calculate the norm of \(u\): \(\|u\|^2 = \langle u, u \rangle = 1^2 + |i|^2 = 1 + 1 = 2\). Thus, \(\|u\| = \sqrt{2}\).
- Calculate the norm of \(v\): \(\|v\|^2 = \langle v, v \rangle = 2^2 + 1^2 = 4 + 1 = 5\). Thus, \(\|v\| = \sqrt{5}\).
- Calculate the inner product \(\langle u, v \rangle\). Using linearity in the first argument: \(\langle u, v \rangle = 1 \cdot \overline{2} + i \cdot \overline{1} = 2 + i\).
- Calculate the modulus: \(\left|\langle u, v \rangle\right| = \sqrt{2^2 + 1^2} = \sqrt{5}\).
- Compute the product of norms: \(\|u\| \, \|v\| = \sqrt{2} \cdot \sqrt{5} = \sqrt{10}\).
- Compare: \(\sqrt{5} \approx 2.236\) and \(\sqrt{10} \approx 3.162\). Since \(2.236 \le 3.162\), the inequality holds. Since \(\sqrt{5} \neq \sqrt{10}\), equality does not hold, implying \(u\) and \(v\) are linearly independent.
Units check. The vectors are dimensionless in this abstract space, so the inequality compares dimensionless quantities. The arithmetic is consistent: \(5 \le 10\).
Three quick tests
- Does the Cauchy–Schwarz inequality hold for the zero vector?
Answer
Yes. If \(u=0\), both sides are zero, so \(0 \le 0\) holds. The equality condition is satisfied as the zero vector is linearly dependent with any vector. - If \(\left|\langle u, v \rangle\right| = \|u\| \, \|v\|\) and \(u \neq 0\), what is the relationship between \(u\) and \(v\)?
Answer
They are linearly dependent. Specifically, \(v = \lambda u\) for some scalar \(\lambda\). In complex spaces, \(\lambda\) can be any complex number. - Why is the modulus \(\left|\langle u, v \rangle\right|\) used in the inequality rather than the inner product itself?
Answer
In complex spaces, the inner product can be a complex number, which does not have a natural ordering. The modulus is a non-negative real number, allowing for a meaningful inequality comparison with the positive real norms \(\|u\|\) and \(\|v\|\).
Problem set for PU-104 →
Each derivation in this unit ends with five graded problems with full solutions.
Lecture notes drafted with a local model and checked line by line by the editor; every number in the worked example was recomputed.