A student wants to bound ∫|fg| dμ for functions f ∈ L³ and g ∈ L². She attempts to apply Hölder's inequality with p = 3 and q = 2. What is wrong with this approach?
AHölder's inequality only applies when f and g are in the same Lᵖ space
BHölder's inequality requires conjugate exponents satisfying 1/p + 1/q = 1; since 1/3 + 1/2 = 5/6 ≠ 1, the pair (3, 2) is not conjugate and the standard Hölder bound does not apply
CThe bound ‖f‖₃ ‖g‖₂ is valid whenever both exponents exceed 1
DHölder's inequality only applies to functions on probability spaces where the total measure is 1
The conjugate exponent condition 1/p + 1/q = 1 is not a technicality — it is precisely what makes the bound tight and what emerges from the proof via Young's inequality. For p = 3, the conjugate is q = 3/2 (since 1/3 + 2/3 = 1), not q = 2. The pair (3, 2) gives 1/3 + 1/2 = 5/6 < 1, and applying Hölder directly would be invalid. If g ∈ L² but not L^(3/2), a direct Hölder bound is not available; one would need to use a different inequality or interpolation.
Question 2 Multiple Choice
When p = q = 2, Hölder's inequality reduces to a well-known classical inequality. Which one?
AThe triangle inequality for L² norms (Minkowski's inequality)
BThe Cauchy–Schwarz inequality: ∫|fg| ≤ ‖f‖₂ ‖g‖₂
CYoung's inequality for products: ab ≤ a²/2 + b²/2
DJensen's inequality for convex functions
When p = q = 2, the conjugate condition 1/2 + 1/2 = 1 is satisfied, and Hölder's inequality becomes ∫|fg| ≤ ‖f‖₂ ‖g‖₂ — precisely the Cauchy–Schwarz inequality in L². This reveals Cauchy–Schwarz as a special case of a more general phenomenon. Hölder's inequality generalizes it to all conjugate pairs (p, q) with 1/p + 1/q = 1 and p, q ∈ (1, ∞), showing that the bounding of ∫|fg| by separate norms is a feature of the conjugate structure, not special to L².
Question 3 True / False
The conjugate condition 1/p + 1/q = 1 in Hölder's inequality is an arbitrary technical restriction; with additional work, the inequality could be extended to non-conjugate pairs.
TTrue
FFalse
Answer: False
The conjugate condition is not arbitrary — it is precisely what emerges from the proof. The standard argument normalizes f and g to have unit Lᵖ and Lq norms, then applies Young's inequality pointwise: |f(x)g(x)| ≤ |f(x)|ᵖ/p + |g(x)|^q/q. Integrating both sides gives ∫|fg| ≤ 1/p + 1/q. For this to equal 1 (recovering the desired bound of ‖f‖_p ‖g‖_q after un-normalizing), we need exactly 1/p + 1/q = 1. The conjugate condition is what makes the proof work, not an afterthought.
Question 4 True / False
Hölder's inequality is essential to Lᵖ duality theory because it provides the bound showing that integration against a function in Lq defines a bounded linear functional on Lᵖ — a fact central to the Riesz representation theorem.
TTrue
FFalse
Answer: True
This is the deep significance of Hölder's inequality beyond computation. Every bounded linear functional on Lᵖ can be represented as Λ(f) = ∫fg dμ for some g ∈ Lq (where 1/p + 1/q = 1). Hölder's inequality is what makes this functional bounded: |Λ(f)| = |∫fg| ≤ ‖f‖_p ‖g‖_q. Without this bound, the correspondence (Lᵖ)* ≅ Lq would not be well-defined as a bounded isomorphism. Hölder is also used directly in proving Minkowski's inequality (the triangle inequality for Lᵖ norms), so it is load-bearing for the entire structure of Lᵖ space theory.
Question 5 Short Answer
Explain the role of Hölder's inequality in establishing the duality between Lᵖ and Lq spaces, and why the conjugate exponent condition is essential to this duality.
Think about your answer, then reveal below.
Model answer: The Riesz representation theorem for Lᵖ spaces states that every bounded linear functional Λ on Lᵖ(μ) can be represented as Λ(f) = ∫fg dμ for a unique g ∈ Lq, where 1/p + 1/q = 1. Hölder's inequality provides the bound |∫fg| ≤ ‖f‖_p ‖g‖_q that proves this functional is bounded with norm at most ‖g‖_q. The conjugate condition is essential because it is precisely the condition under which Hölder holds — pairing Lᵖ with any other space does not yield a controlled bound. The duality (Lᵖ)* ≅ Lq is therefore built directly on the structure of Hölder's inequality: conjugate exponents are not just a technical convenience but the algebraic expression of the duality itself.
The conjugate condition 1/p + 1/q = 1 can be read as the statement that p and q are 'complementary' in a specific sense: their reciprocals sum to 1. This is precisely the algebraic condition that makes the Lᵖ-Lq pairing work, and Hölder's inequality is the analytic theorem that realizes that algebraic structure.