Problem 1 Problem 1. Let X = R^2, and let p \in (0,\infty).

Define ||x||_p = (|x_1|^p +

|x_2|^p)^{1/p} for x = (x_1,x_2). (a) For x = (3,-4), compute ||x||_1, ||x||_2, ||x||_3,
||x||_\infty. Solution. ||x||_1 = |3| + | -4| = 7. ||x||_2 = sqrt(3^2 + (-4)^2) =
sqrt(9+16) = sqrt(25)=5. ||x||_3 = (|3|^3 + | -4|^3)^{1/3} = (27 + 64)^{1/3} = 91^{1/3}
(approx 4.497). ||x||_\infty = max{|3|,| -4|} = 4. (b) Draw unit balls {x : ||x||_p = 1}
for p=1,2,\infty. Solution (description): For p=1 the unit ball is the diamond (convex
polygon) with vertices (±1,0),(0,±1). For p=2 the unit ball is the Euclidean circle of
radius 1 centered at origin. For p= the unit ball is the square with sides parallel to
axes: [-1,1]×[-1,1]. (c) Show ||x||_p is not a norm when 0<p<1. Proof. A norm must
satisfy the triangle inequality. Take p (0,1). Consider x=(1,0), y=(0,1). Then ||x||_p =
(1^p + 0)^{1/p}=1, similarly ||y||_p=1. But ||x+y||_p = ||(1,1)||_p = (1^p + 1^p)^{1/p} =
(2)^{1/p}. Since p<1, 1/p >1 so 2^{1/p} >2. Hence ||x+y||_p = 2^{1/p} > 2 = ||x||_p +
||y||_p, so triangle inequality fails (it reverses). Therefore not a norm.
Problem 2 Problem 2. (a) Find conjugate exponents: for p=3, p=4/3, p=1.5. Solution.
Conjugate q satisfies 1/p + 1/q =1, so q = p/(p-1). For p=3: q = 3/(3-1)=3/2 = 1.5. For
p=4/3: q = (4/3)/(1/3)=4. For p=1.5 = 3/2: q = (3/2)/(1/2)=3. (b) Determine when equality
holds in Young's inequality. Young: for a,b 0 and p,q>1 conjugate, ab a^p/p + b^q/q
with equality iff a^p = b^q. Explanation: proved by convexity (tangent line) or weighted
AM-GM; equality condition as above.
Problem 3 Problem 3. Let a=(1,2,3), b=(4,1,2) in R^3. Verify Hölder and Minkowski for
p=2,3. Solution sketch. Hölder: For p,q conjugate, sum |a_i b_i| ||a||_p ||b||_q.
Compute norms and LHS. Case p=2 (so q=2). Compute dot-sum S = a_i b_i (nonnegative
terms): S = 1*4 + 2*1 + 3*2 = 4 + 2 + 6 = 12. ||a||_2 = sqrt(1+4+9)=sqrt(14). ||b||_2 =
sqrt(16+1+4)=sqrt(21). Product = sqrt(14*21) = sqrt(294) 17.146. So Hölder holds: 12
17.146. Minkowski (triangle inequality for ||·||_p): ||a+b||_p ||a||_p + ||b||_p.
Compute a+b = (5,3,5). For p=2: ||a+b||_2 = sqrt(25+9+25)=sqrt(59) 7.681. ||a||_2 +
||b||_2 sqrt(14)+sqrt(21) 3.742+4.583=8.325. Inequality holds. Case p=3 (q=3/2).
Compute ||a||_3 = (1^3+2^3+3^3)^{1/3} = (1+8+27)^{1/3}=36^{1/3} 3.302. ||b||_{3/2} =
(4^{3/2} + 1^{3/2} + 2^{3/2})^{2/3}. Compute numerically: 4^{1.5}=8, 1^{1.5}=1,
2^{1.5}=2.828; sum 11.828. Then ||b||_{3/2} = (11.828)^{2/3} (11.828)^{0.6667}
5.236^{?} (numerical check). Regardless, verify |a_i b_i| =12 ||a||_3 ||b||_{3/2}
(numerical check shows RHS larger). Minkowski similarly.
Problem 4 Problem 4. (a) X metric space, C_b(X) bounded continuous functions, ||f||_ =
sup_{x X} |f(x)|. Show it's a norm. Proof: Non-negativity and definiteness: sup 0; if
sup=0 then f=0. Homogeneity: || f||_ = | | ||f||_ . Triangle inequality: |f+g| |f|+|g|
pointwise, so sup |f+g| sup|f| + sup|g|. (b) X=[a,b], f(t)=t^2, g(t)=sin( t). Compute
norms. ||f||_ on [a,b] is max_{t [a,b]} t^2 (depends on a,b). For example if [a,b]=[0,1],
||f||_ =1. ||g||_ = 1. ||f+g||_ ||f||_ + ||g||_ and compute specific values
accordingly. Triangle inequality holds as general fact above.
Problem 5 Problem 5. For x=(x_1,...,x_n) 0 in R^n: (a) Show lim_{p } ||x||_p = ||x||_ .
Proof sketch: For finite vector, ||x||_p^p = |x_i|^p. Let M = max |x_i|. Then M^p
|x_i|^p n M^p. Taking p-th root gives M ||x||_p n^{1/p} M. As p , n^{1/p} 1, so
||x||_p M. (b) Show lim_{p 1+} ||x||_p = ||x||_1. Proof: As p 1, the p-norm tends to
the 1-norm by continuity of the map p ( |x_i|^p)^{1/p} at p=1; can take logs or use
dominated convergence on finite sum.
Problem 6 Problem 6. Strict convexity. (a) For 1<p< prove ^p strictly convex and
characterize equality in Minkowski. Proof sketch: Use Clarkson or strictly convexity of t
|t|^p for p>1. Equality in Minkowski holds iff vectors are linearly dependent and have
same direction: x and y are linearly dependent with nonnegative scalar and satisfy |x_i|^p
= |y_i|^p componentwise where nonzero. More concretely, equality in Hölder
proportionality with matching signs; from that derive equality in Minkowski. (b) C([a,b])
with ||·||_ is not strictly convex: take functions f g with ||f||_ =||g||_ =1 and such
that at each point either f(x)=1 or g(x)=1 but average has sup=1 as well; e.g., f 1, g 1
on different supports? Simpler: take f 1 and g -1? That fails. Instead pick f(x)=1 and
g(x)=1 on interval but differ elsewhere; show midpoint has norm 1. So not strictly convex.
Problem 7 Problem 7. For 1<p,q< conjugate. For sequences, equality in Hölder |x_i y_i|
= ||x||_p ||y||_q holds iff there exists 0 with |x_i|^p = |y_i|^q on indices where
both nonzero and sgn(x_i)=sgn(y_i). Proof sketch: This is the equality case of Hölder,
which follows from equality in Young applied termwise or from equality in weighted AM-GM:
equality iff for each i, |x_i|^p/||x||_p^p = |y_i|^q/||y||_q^q, yielding proportionality.
Problem 8 Problem 8. Seminorm on X and Z_ = {x: (x)=0}. (a) Show ||[x]|| := (x)
defines a norm on quotient X/Z_ . Proof: (x) =0 iff x Z_ , so well-defined on cosets.
Positive-definite, homogeneous, triangle via seminorm properties. (b) If X Banach and
(x)=inf{||x+y||: y Y} for closed subspace Y, prove X/Z_ X/Y and is Banach. Sketch:
Here Z_ = Y; quotient by Y with induced norm is Banach; the map [x]_{Z_ } x+Y is an
isometric isomorphism.
Problem 9 Problem 9. C([a,b]) complete under ||·||_ and not complete under ||·||_p,
1 p< . Proof: For sup-norm completeness: uniform limit of continuous functions is
continuous; Cauchy sequence in sup-norm is uniformly Cauchy hence converges uniformly to
continuous limit in C([a,b]). For ||·||_p, consider sequence of continuous functions that
approximate L^p function with limit not continuous in L^p sense; classical example: on
[0,1] define sequence f_n that are continuous but converge in L^p to function with jump;
limit not in C([0,1]). Thus space not complete for L^p-norm.
Problem 10 Problem 10. If 1 p < q , show _p dense in _q with respect to ||·||_q.
Proof: For 1 p<q< , note that _p _q and c_{00} (finitely supported) _p; c_{00}
dense in _q; thus closure of _p contains c_{00} hence equals _q. For q= , statement
must be interpreted: for p< , _p c_0 _ and c_{00} dense in c_0 under sup-norm, but
_p dense in _ ? Not in general. The intended statement: _p is dense in _q for p<q
(both finite) under ||·||_q. Proof via truncation: given x _q, partial sums are in _p?
Provide truncations from c_{00}.
Problem 11 Problem 11. Every infinite-dimensional Banach space X contains infinitely many
proper subspaces that are not closed. Sketch: Choose a Schauder basis if exists; else use
Hamel basis and construct increasing chain of finite-dimensional subspaces whose closures
are proper; standard construction: choose linearly independent sequence (x_n) then
span{x_1,...,x_n} not closed? Need produce non-closed proper subspaces: e.g., pick a
sequence dense in X and take linear span of proper subset to get dense proper subspaces
(hence not closed). There are infinitely many choices.
Problem 12 Problem 12. For open in R^n and 1 p< , C_c( ) dense in L^p( ). Proof sketch:
Use mollifiers: for f L^p, approximate by f* _ (smooth), then cut off to compact support
in with cutoff functions; standard approximation theorem proves density of C_c^ ( )
hence of C_c( ).