Chapter 34 Quiz — Inner Product Spaces

Twelve conceptual checks. Try each before opening the answer. These test understanding — whether you see that geometry comes from the axioms, not from the components.


1. Which three properties define a (real) inner product, and which one is the one you must actually check when proposing a new candidate (because symmetry and linearity are usually obvious)?

Answer Symmetry ($\langle\mathbf{u},\mathbf{v}\rangle=\langle\mathbf{v},\mathbf{u}\rangle$), linearity in each argument (sums split, scalars pull out), and positive-definiteness ($\langle\mathbf{v},\mathbf{v}\rangle\ge 0$, with equality iff $\mathbf{v}=\mathbf{0}$). **Positive-definiteness** is the fragile one to verify — symmetric, bilinear forms are common, but only the positive-definite ones are inner products. (Example of a symmetric bilinear form that fails: the Minkowski form of relativity.)

2. Once a vector space has an inner product, how are length, angle, and orthogonality defined — and why are these not new formulas to memorize?

Answer $\lVert\mathbf{v}\rVert=\sqrt{\langle\mathbf{v},\mathbf{v}\rangle}$ (induced norm), $\cos\theta=\langle\mathbf{u},\mathbf{v}\rangle/(\lVert\mathbf{u}\rVert\lVert\mathbf{v}\rVert)$ (angle), and $\mathbf{u}\perp\mathbf{v}$ iff $\langle\mathbf{u},\mathbf{v}\rangle=0$ (orthogonality). They are the *exact Chapter 18 definitions* with $\langle\cdot,\cdot\rangle$ replacing $\cdot$. The dot product was one example; the definitions were always written in terms of the operation, so they transfer unchanged.

3. What single guarantee makes the angle formula $\theta=\arccos(\dots)$ legitimate in any inner product space, and where does it come from?

Answer The general **Cauchy–Schwarz inequality** $|\langle\mathbf{u},\mathbf{v}\rangle|\le\lVert\mathbf{u}\rVert\lVert\mathbf{v}\rVert$, which forces the fraction into $[-1,1]$, the domain of $\arccos$. It is proved from the three axioms alone (the discriminant-of-a-nonnegative-quadratic argument of §34.6), so it holds in every inner product space at once — that is why "the angle between two functions" or "the angle in $\mathbb{C}^2$" is meaningful.

4. Why must every weight in a weighted inner product $\langle\mathbf{u},\mathbf{v}\rangle_w=\sum_i w_i u_iv_i$ be strictly positive?

Answer Positive-definiteness. If some $w_j=0$, then the nonzero vector $\mathbf{e}_j$ has $\langle\mathbf{e}_j,\mathbf{e}_j\rangle_w = w_j = 0$, violating "zero length only for the zero vector." If some $w_j<0$, a vector concentrated in coordinate $j$ has *negative* squared length, which destroys the induced norm entirely. Strict positivity of every weight is exactly the condition that keeps the form an inner product.

5. Two functions $f$ and $g$ are orthogonal under $\langle f,g\rangle=\int fg\,dx$. Does that mean their graphs cross at right angles somewhere?

Answer No. Orthogonality means the *integral* $\int fg\,dx=0$ — the signed area under the product is zero, so positive and negative contributions cancel. It says nothing about the visual appearance of the curves. $\sin x$ and $\cos x$ are orthogonal on $[-\pi,\pi]$ yet their graphs do not "look perpendicular" anywhere. Orthogonality lives in the inner product, never in the picture of the graphs.

6. Over the complex numbers, the naive $\sum u_iv_i$ fails to be an inner product. What goes wrong, and what is the fix?

Answer It violates positive-definiteness: for $\mathbf{v}=(1,i)$, $\sum v_i^2 = 1+i^2 = 0$, a nonzero vector with zero "squared length" (and other vectors get negative values). The fix is to **conjugate one argument**: $\langle\mathbf{u},\mathbf{v}\rangle=\sum\overline{u_i}v_i$. Then $\langle\mathbf{v},\mathbf{v}\rangle=\sum|v_i|^2\ge 0$ is real and non-negative, because $\overline{z}z=|z|^2$. The conjugate is forced, not cosmetic.

7. Conjugating one argument replaces symmetry with what property, and what is the name for "linear in one slot, conjugate-linear in the other"?

Answer **Conjugate symmetry** (Hermitian symmetry): $\langle\mathbf{u},\mathbf{v}\rangle=\overline{\langle\mathbf{v},\mathbf{u}\rangle}$. This implies $\langle\mathbf{v},\mathbf{v}\rangle$ is real (it equals its own conjugate). An operation linear in one argument and conjugate-linear in the other is called **sesquilinear** ("one-and-a-half-linear"). With the physics convention, the inner product is linear in the *second* argument and conjugate-linear in the *first*.

8. What is a Hilbert space, and why is "completeness" invisible in finite dimensions but essential in infinite ones?

Answer A **Hilbert space** is a *complete* inner product space — one in which every Cauchy sequence (terms getting arbitrarily close together) converges to a limit *inside the space*. Finite-dimensional inner product spaces ($\mathbb{R}^n$, $\mathbb{C}^n$) are automatically complete. Infinite-dimensional spaces can have "holes" (a Cauchy sequence converging to something outside the space, like continuous functions ramping toward a discontinuous step). Completeness fills the holes, which is what lets infinite basis expansions — Fourier series, quantum eigenstate expansions — converge to genuine vectors.

9. In an orthonormal basis $\{\mathbf{e}_i\}$, how do you find the coordinate $c_j$ of a vector $\mathbf{v}=\sum_i c_i\mathbf{e}_i$, and what is this formula called?

Answer $c_j=\langle\mathbf{e}_j,\mathbf{v}\rangle$ — each coordinate is just the inner product of the vector with that basis vector (take $\langle\mathbf{e}_j,\cdot\rangle$ of both sides; orthonormality kills every term but $i=j$). These are the **generalized Fourier coefficients**, and they unify three things: the entries of a vector in $\mathbb{R}^n$, a Fourier coefficient (Chapter 22), and the amplitudes $\alpha,\beta$ of a qubit. No linear system to solve — orthogonality decouples the coordinates.

10. A qubit is a unit vector $|\psi\rangle=\alpha|0\rangle+\beta|1\rangle$ in $\mathbb{C}^2$. What does the Born rule say, and how does the unit-norm condition relate to probability?

Answer The Born rule: $P(0)=|\langle 0|\psi\rangle|^2=|\alpha|^2$ and $P(1)=|\langle 1|\psi\rangle|^2=|\beta|^2$ — the probability of an outcome is the squared overlap (squared projection) of the state onto that outcome's direction. The normalization $\lVert\psi\rVert^2=|\alpha|^2+|\beta|^2=1$ is, by the Pythagorean/Parseval identity (§34.7), exactly $P(0)+P(1)=1$: the qubit yields *some* outcome with certainty. Geometry (unit length) = physics (total probability one).

11. When are two quantum states distinguishable with certainty by a single measurement, in inner-product-space terms?

Answer When they are **orthogonal**: $\langle\psi|\phi\rangle=0$. Orthogonal states share no overlap, so a measurement can always tell them apart. States that are nearly parallel (overlap near $1$) are nearly impossible to distinguish. The probability of "confusing" two states is governed by $|\langle\psi|\phi\rangle|^2$, the squared overlap — pure inner-product geometry deciding a physical question.

12. Why does the same Gram–Schmidt implementation orthogonalize arrows in $\mathbb{R}^3$, polynomials, and quantum states, with no change to the algorithm?

Answer Because the Gram–Schmidt formula and its correctness proof (Chapter 20) refer only to the inner product, vector addition, and scaling — never to components or to the dot product specifically. Each subtracted "shadow" cancels by bilinearity, an argument valid under the three axioms in any inner product space. In code, you pass the inner product in as an argument; the only difference between geometries is *which* inner product you supply. This is the chapter's thesis: geometry lives in the operation, not in the arrows.