Questions: L^p Norm and Metric Structure

Question 1

Two functions f and g have small L¹ distance but large L^∞ distance. What does this tell you about their relationship?

• A. The functions are close everywhere, but L^∞ is unreliable for measuring closeness
• B. The functions have large total area between them but agree on the essential supremum
• C. The functions differ dramatically on a very small set but agree closely almost everywhere else, keeping total integral area small
• D. L¹ and L^∞ are measuring the same thing, so this combination is impossible

Answer: (C)

Small L¹ distance means ∫|f - g| dμ is small — the total area between them is small — which is compatible with f and g differing enormously on a tiny set. Large L^∞ distance means the essential supremum of |f - g| is large — somewhere f and g differ dramatically. Together these mean: large spike on a small set, but agreement almost everywhere. L¹ averages out spikes; L^∞ is worst-case sensitive. They measure fundamentally different things.

Question 2

The most critical step in proving that ‖·‖_p is a genuine norm on L^p (for 1 ≤ p < ∞) is:

• A. Showing that ‖f‖_p = 0 if and only if f = 0 almost everywhere
• B. Showing that ‖cf‖_p = |c|‖f‖_p for all scalars c
• C. Establishing the triangle inequality via Minkowski's inequality, which requires Hölder's inequality as a lemma
• D. Showing that L^p contains all square-integrable functions

Answer: (C)

The first two properties (positivity and homogeneity) follow almost immediately from the definition of the integral. The triangle inequality — ‖f + g‖_p ≤ ‖f‖_p + ‖g‖_p — is Minkowski's inequality, and its proof is genuinely non-trivial, requiring Hölder's inequality as a preliminary result. Without the triangle inequality, d(f,g) = ‖f-g‖_p would not satisfy the metric triangle inequality d(f,h) ≤ d(f,g) + d(g,h), and L^p would not be a metric space at all.

Question 3

For p = ∞, the L^∞ norm ‖f‖_∞ equals the maximum value of |f| over the domain.

• T. True
• F. False

Answer: False

The L^∞ norm is the *essential* supremum of |f| — the smallest value M such that |f| ≤ M almost everywhere — not the maximum. The essential supremum ignores sets of measure zero. If |f| exceeds M only on a null set, the essential supremum is still M even though the maximum may be larger. This distinction matters in measure theory where individual points carry no weight.

Question 4

As p increases, the L^p norm becomes increasingly sensitive to large local deviations in a function, since high values of |f| are raised to a higher power before integrating.

• T. True
• F. False

Answer: True

When computing (∫|f|^p dμ)^(1/p), raising |f| to a higher power amplifies regions where |f| is large relative to where it is small. In the limit p → ∞, the norm is dominated entirely by the essential supremum — the largest value of |f| anywhere. L¹ gives equal weight to all deviations by total area; L^∞ is purely worst-case. Higher p interpolates toward worst-case sensitivity, making L^p norms with large p progressively less tolerant of spikes.

Question 5

Why does establishing the L^p metric require Minkowski's inequality, and what would fail if the triangle inequality were violated?

Model answer: The triangle inequality ‖f + g‖_p ≤ ‖f‖_p + ‖g‖_p is one of the three axioms a norm must satisfy. If it failed, d(f,g) = ‖f-g‖_p would also violate the metric triangle inequality d(f,h) ≤ d(f,g) + d(g,h), since that metric inequality follows directly from the norm inequality. Without a valid metric, L^p would not be a metric space — we could not define convergence of sequences, Cauchy sequences, or completeness. The entire functional-analytic structure (Banach space theory, density results, approximation theorems) depends on L^p being a complete metric space, which requires the norm triangle inequality as its foundation.

Minkowski's inequality is not a technical formality — it is the result that distinguishes a genuine geometric space (with a meaningful notion of distance) from a mere set of functions equipped with an arbitrary non-negative function. Every downstream result that uses the metric structure of L^p — from Hölder's inequality to the Riesz representation theorem — ultimately rests on this.