Abstract Linear Algebra 10 | Inner Products

General inner products are the mechanism that turn a purely algebraic vector space into a geometric one—enabling notions of length and angle even when no coordinates or standard geometry exist. The core move is defining an inner product as a map ⟨·,·⟩: V×V→F (with F being the real numbers or the complex numbers) that satisfies three constraints: positive definiteness, linearity in one argument, and conjugate symmetry. Together, these properties ensure that ⟨x,x⟩ is always a nonnegative real number, equals 0 only when x is the zero vector, and that the inner product behaves predictably under addition and scalar multiplication.

The transcript emphasizes how the definition must be adapted to complex vector spaces. Scalars and matrix entries use complex conjugation, denoted with a bar, and the conjugate-symmetric condition becomes the key difference from the real case. In real spaces, swapping the two inputs leaves the value unchanged (symmetry). In complex spaces, swapping inputs conjugates the result: ⟨x,y⟩ relates to ⟨y,x⟩ through complex conjugation. Linearity is taken in the second argument (with the note that some conventions place it in the first), meaning ⟨x, y+z⟩=⟨x,y⟩+⟨x,z⟩ and ⟨x,λy⟩=λ⟨x,y⟩ for all vectors and scalars.

To ground the abstraction, the standard inner product on F^n is given as ⟨u,v⟩ = Σ_i (conjugate(u_i)) v_i. The same quantity can be written compactly using matrices: if u and v are column vectors, then ⟨u,v⟩ equals u* v, where u* is the conjugate transpose (transpose plus entrywise complex conjugation). For real vectors, the conjugation has no effect, so u* reduces to the transpose.

A contrasting example on F^2 illustrates what can go wrong. A proposed bilinear rule that “mixes” components—pairing u1 with v2 and u2 with v1—can satisfy linearity and conjugate symmetry, but it fails positive definiteness. Plugging the same vector into both slots (using the specific choice (1,−1)) yields a negative value for ⟨x,x⟩, which violates the requirement that lengths come from nonnegative real numbers. The takeaway is blunt: all three inner-product properties must hold simultaneously.

Finally, the transcript extends the idea beyond finite-dimensional spaces by moving to polynomial spaces on the unit interval. For polynomials P and Q, an inner product is defined via an integral of the form ⟨P,Q⟩ = ∫_0^1 (conjugate(P(x)) Q(x)) dx, mirroring the finite-dimensional “sum of conjugate times” structure but replacing the sum over components with an integral over x. Using P(x)=x as an example, the computation reduces to integrating x^2 from 0 to 1, producing 1/3, a positive real number as required. The discussion also notes that with different polynomials, the output need not be positive or even real, which is consistent with the inner product’s general definition.

In short: inner products are the axiomatic bridge from algebra to geometry, and the transcript shows how conjugation, positivity, and linearity determine whether a proposed formula truly measures lengths and angles.