QUALIFYING EXAMINATION

Harvard University
Department of Mathematics
Tuesday September 3, 2024 (Day 1)

1. (AG) Let C be a smooth complex projective curve of genus g. For any P ∈ C,

prove the following statements:

1. For any k ≥ 2g, there is always a nonconstant rational function on C

which is regular everywhere except for a pole of order k at P .
2. (Weierstrass gaps) There are exactly g numbers 0 < k1 < k2 < · · · <
kg < 2g such that for each 1 ≤ i ≤ g, there is no nonconstant rational
function on C which is regular everywhere except for a pole of order ki
at P .

Solution:

1. Recall that for a divisor D on C with deg(D) ≥ 2g −1, we have deg(KC −

D) = 2g − 2 − deg(D) < 0, whence dim H 0 (C, OC (KC − D)) = 0. Then
Riemann-Roch says dim H 0 (C, OC (D)) = deg(D) − g + 1. In particular,
for k ≥ 2g − 1, dim H 0 (C, OC (kP )) = k − g + 1 increases by 1 each
time k increases by 1. Therefore, for each k ≥ 2g, there exists some
f ∈ H 0 (C, OC (kP )) ∖ H 0 (OC ((k − 1)P )), as desired.
2. We have

1 = dim H 0 (C, OC ) ≤ dim H 0 (C, OC (P )) ≤ · · · ≤ dim H 0 (C, OC ((2g−1)P )) = g.

Given that there are 2g − 1 inequalities above, and given that

dim H 0 (C, OC (kP ))/H 0 (OC ((k − 1)P )) ≤ 1

for every k ≥ 1, we deduce the conclusion.

2. (AT) Suppose that X is a finite connected CW complex such that π1 (X) is

finite and nontrivial. Prove that the universal covering X
e of X cannot be
contractible.

Solution: Since X is a finite CW complex and π1 (X) is finite, X

e is also a
finite CW complex, so the Lefschetz fixed point theorem applies. Since X e
e ∼
is contractible, we have H0 (X) e ∼
= Z and Hi (X) = 0, so that the Lefschetz
 number of any self-map is 1. In particular, any self-map of X e must have a
fixed point.
This is a contradiction, since any nontrivial deck transformation of X
e is a self-
map without fixed points, and the group of deck transformations is isomorphic
to π1 (X).
Alternatively, you could argue that the Euler characteristic χ(X)e is |π1 (X)| ·
χ(X), so that if π1 (X) is nontrivial we can’t have χ(X)e = 1.

3. (CA) Suppose R1 , R2 are bounded, simply-connected open subsets of C. Let

z1 ∈ R1 and z2 ∈ R2 .
i) Prove that there is a holomorphic bijective function f : R1 → R2 such that
f (z1 ) = z2 .
ii) Suppose that g : R1 → R2 is a holomorphic function such that g(z1 ) = z2 .
Prove that g ′ (z1 ) ≤ f ′ (z1 ) .

Solution. Let ∆ be the open disc {w ∈ C : |w| = 1}.

i) By the Riemann mapping theorem there exist holomorphic bijections ϕ1 :
R1 → ∆ and ϕ2 : R2 → ∆ such that ϕ1 (z1 ) = ϕ2 (z2 ) = 0. Take f = ϕ−1 2 ◦ ϕ1 .
ii) With ϕ1 , ϕ2 , f as above, define ψ : ∆ → ∆ by ψ = ϕ2 ◦ g ◦ ϕ−1 1 . Then
ψ(0) = H0; and By Cauchy’s integral formula |ψ ′ (0)| ≤ 1: indeed ψ ′ (0) =
(2πi)−1 |w|=r ψ(w) dw/w2 for each r < 1, whence |ψ ′ (0)| ≤ (2πr)/(2πr2 ) =
1/r. By the Chain Rule, f ′ (z1 ) = ϕ′1 (z1 )/ϕ′2 (z2 ), and ψ ′ (0) = ϕ′2 (z2 )g ′ (z1 )/ϕ′1 (z1 ).
Thus g ′ (z1 ) = ψ ′ (0)f ′ (z1 ); the desired inequality g1′ (z1 ) ≤ f1′ (z1 ) follows.

4. (A) Let k0 be the field Z/pZ, and let k be an algebraically closed field contain-
2
ing k0 . Fix a ∈ k, and let P ⊂ k[X] be the polynomial P (X) = X p +aX p +X.
Finally let K be the field k(y) of rational functions in an indeterminate y.
(a) Prove that V := {x ∈ k : P (x) = 0} is a 2-dimensional vector space
over k0 .
(b) Let F be the extension of K obtained by adjoining a root x of P (x) =
y. Prove that F/K is a Galois extension with Galois group isomorphic
with V .
(c) How many fields E are there such that K ⊂ E ⊂ F , other than K and F
themselves?

Solution
i) The map x → xp is k0 -linear. So is multiplication by a, and thus so is
x 7→ P (x). Hence V = ker(P ) is a k0 -vector space. Since k is algebraically
 closed, it will follow that |V | = deg P = p2 , and thus that dimk0 V = 2, once
we check that P has distinct roots. This follows from the observation that
P ′ (x) = 1 ̸= 0 for all x (since p = 0 in K), and thus in particular for all x ∈ V .
ii) We first show F/K is Galois by checking that F contains all the roots of
P (x) = y in an algebraic closure of K, and that these roots are distinct. Since
x 7→ P (x) is linear, P (x) = y implies P (x + v) = y for all v ∈ V . Hsnce P has
at least p2 distinct roots x + v in F ; since p2 = deg(P ) we are done.
This also gives a bijection V → Gal(F/K), v 7→ gv where gv (x) = x + v.
So it remains to check that this is a group homomorphism from (V, +) to
Gal(F/K): for v, v ′ ∈ V we compute

gv′ gv x = (x + v) + v ′ = x + (v + v ′ ) = gv′ +v (x),

using in the first step the fact that gv is a field automorphism acting trivially
on K and thus on k ⊂ K.
iii) By the Galois correspondence, intermediate fields E correspond to sub-
groups H ⊂ V other than {1} and V itself (with H corresponding to the fixed
subfield E = F H ). Such a subgroup has order p (by Lagrange, for example),
and these groups are in (p − 1)-to-one correspondence with nonzero elements
of V , so the count is (p2 − 1)/(p − 1) = p + 1.

5. (DG) Let Ω be the 2-form on R3 − {0} defined by

1
Ω= (x dy ∧ dz + y dz ∧ dx + z dx ∧ dy).
(x2 + y2+ z 2 )3/2

(a) Prove that Ω is closed

(b) Let f : R3 − {0} → S 2 be the map (x, y, z) 7→ ( x2 +y12 +z 2 )1/2 (x, y, z).
Prove that Ω is the pullback along f of a form on S 2 .
(c) Prove that Ω is not exact.

Solution:
Let us use spherical coordinates (r, θ, ϕ), defined via x = r sin θ cos ϕ, y =
r sin θ sin ϕ and z = r cos θ. In spherical coordinates, we have

Ω = sin θ dθ ∧ dϕ.

To prove (a), we compute

dΩ = cos θ dθ ∧ dθ ∧ dϕ + sin θ(ddθ ∧ dϕ − dθ ∧ ddϕ) = 0.

 In spherical coordinates, the map f appearing in part (b) takes the form
(r, θ, ϕ) 7→ (1, θ, ϕ). It therefore follows from the lack of r in the formula for
Ω that it is pulled back along f .
To prove that Ω is not exact, it suffices to prove that its restriction to S 2 is
not exact. But the integral of Ω over S 2 is
Z Z 2π Z π
sin θdθ ∧ dϕ = sin θdθdϕ = 2π(− cos(π) + cos(0)) = 4π ̸= 0,
S2 0 0

so it cannot be exact.

6. (RA) Let L2 ([1, 2])

R 2 be the Hilbert space of real-valued functions with inner
product ⟨f, g⟩ = 1 f (x)g(x)dx.
1. Consider the linear map T : L2 ([1, 2]) → L2 ([1, 2]) given by T f (x) =
xf (x) for x ∈ [1, 2]. Show that this map is continuous and invertible
(with continuous inverse), and show that T = T ∗ .
2. Show that T : L2 ([1, 2]) → L2 ([1, 2]) has no (non-zero) eigenvectors.
3. Fix any λ ∈ [1, 2]. Find a sequence {fn }∞ 2
n=1 in L ([1, 2]) such that
∥fn ∥L2 ([1,2]) = 1 for all n and ∥(T − λI)fn ∥L2 ([1,2]) → 0, where I is
the identity.

Solution
R2
1. Continuity follows from boundedness, i.e. ∥T f ∥2 = 1 |T f (x)|2 dx ≤
R2 2 2 2
1 4|f (x)| dx = 4∥f ∥ for all f ∈ L ([1, 2]). To show invertibility, we
−1 −1
note that T f (x) = x f (x), and that R 2 this map is bounded
R2 from
L2 ([1, 2]) → L2 ([1, 2]), since ∥T −1 f ∥2 = 1 |T −1 f (x)|2 dx = 1 x−2 |f (x)|2 dx ≤
R2 2 2 ∗
1 |f (x)| dx = ∥f ∥ . To show T = T , it suffices to note that for any
f, g ∈ L2 ([1, 2]), we have
Z 2 Z 2 Z 2
⟨T f, g⟩ = T f (x)g(x)dx = xf (x)g(x)dx = f (x)T g(x)dx = ⟨f, T g⟩.
1 1 1

2. Suppose T f (x) = λf (x) for some f ∈ L2 ([1, 2]) and λ ∈ C and all
x ∈ [1, 2]. We have xf (x) = λf (x) for all x ∈ [1, 2]. So, for all λ ̸= x, we
have f (x) = 0, i.e. f vanishes almost everywhere in [1, 2] and is thus the
zero function.
3. Suppose λ ̸= 2. Let fn = n1/2 χ[λ,λ+1/n] ; this is well-defined in L2 ([1, 2])
R λ+1/n
if n is large enough since λ ̸= 2. Note ∥fn ∥2 = λ ndx = 1. Now,
note (T − λI)fn (x) = xfn (x) − λfn (x) = (x − λ)fn (x), and that
Z λ+ 1 Z 1+ 1
n n
2
∥(T − I)fn ∥ = 2
|x − λ| ndx ≤ n−1 dx → 0.
λ 1
If λ = 2, then replace [λ, λ + 1/n] by [λ − 1/n, λ], and the same argument
works.
 QUALIFYING EXAMINATION
Harvard University
Department of Mathematics
Wednesday September 4, 2024 (Day 2)

1. (CA) Suppose u ̸∈ Z. Prove that

∞
X 1 π2
2
= 2
.
n=−∞
(u + n) (sin πu)

(Hint: consider the function f (z) = π(u+z)

cot πz
2 ; you can use the fact that |sin(πw)|

is bounded away from 0 if the distance between w and Z is bounded away from
0.)

Solution For any N ≥ 1, let KN be the circle of radius N + 21 ; we will assume

N ≥ |u| + 1. We will compute
Z
1 π cot πz
dz
2πi KN (u + z)2
and send N → ∞ by using residues.
Note that inside this circle, we have simple poles of cot(πz) at every integer
n ≤ N ; these are also simple poles of f (z) = π(u+z)
cot(πz)
2 since u is not an integer.
−1 −1
Since cot(πz) = cos(πz)/ sin(πz) ∼ π (z − n) for every n if z ≈ n, each of
these poles contributes (u + n)−2 .
Now, note that (u + z)−2 = (z − (−u))−2 has a pole of order 2 at z = −u.
d
Thus, this pole contributes π dz cot(πz) = −π 2 / sin(πz)2 . In particular, we
deduce
∞
π2
Z
1 π cot πz X 1
lim dz = − .
N →∞ 2πi KN (u + z)2 (u + n)2 (sin πu)2
n=−∞

However, we also know that |cot(πz)| is bounded on KN uniformly in N per

the hint, so that
Z Z
1
|f (z)|dz ≥ c 2
dz
KN KN N

R c > 0. Since the length of KN is ≤ CN for some finite C,

if N is large, where
we deduce that KN f (z)dz → 0 as N → ∞, so that
∞
X 1 π2
0= − ,
n=−∞
(u + n)2 (sin πu)2
 which finishes the proof.

2. (A) Let G be a finite group, H be a normal subgroup of G, and P be a Sylow

p-subgroup of H.

1. (Frattini’s Argument) Show that G = HNG (P ), where NG (P ) is the

normalizer subgroup of P in G. (Hint: Use one of Sylow’s theorems.)
2. Show that if every maximal proper subgroup of G is normal in G, then
so is any Sylow p-subgroup of G.

Solution:

1. Let g ∈ G. Then gP g −1 is a Sylow p-subgroup of gHg −1 = H, so by

Sylow’s theorem, there exists h ∈ H such that gP g −1 = hP h−1 . Thus,
h−1 gP (h−1 g)−1 = P , which means h−1 g ∈ NG (P ). Since g ∈ G was
arbitrary, we are done.
2. Let P be a Sylow p-subgroup of G. If P is not normal in G, let M be
a maximal proper subgroup of G containing NG (P ). By hypothesis, M
is normal in G, but then by Frattini’s Argument, G = M NG (P ) = M , a
contradiction.

3. (DG) Let   
 x 0 0 
G :=  0 y z  : x, y ∈ R+ , y ∈ R .
0 0 1
 

(a) Show that G is a Lie subgroup of GL3 (R).

(b) Prove that  
∂ ∂ ∂
x ,y ,y
∂x ∂y ∂z
forms a basis for the left-invariant vector fields on G.
(c) Determine the structure of the Lie algebra g of G in terms of the basis
of (b).

Solution:
 Part (a): to check that G is a subgroup of GL3 (R), we need to check that it
is closed under multiplication and inversion, which follows from the following
computations:
    
x1 0 0 x2 0 0 x1 x2 0 0
 0 y1 z1   0 y2 z2  =  0 y1 y2 z1 + y1 z2 
0 0 1 0 0 1 0 0 1
 −1  −1 
x 0 0 x 0 0
 0 y z  =  0 y −1 −zy −1 
0 0 1 0 0 1
To check that it is a Lie subgroup, it suffices to check that G is a submanifold
of GL3 (R). Using the open immersion GL3 (R) ⊂ M3 (R) ∼ = R9 , it suffices to
9
check that G is a submanifold of R . Up to translation and reordering of the
coordinates, G ⊂ R9 is equal to the standard inclusion R+ × R+ × R ⊂ R9 ,
which is a submanifold since it is locally cut out by the equations x4 = . . . x9 =
0.
Part (b): in coordinates, left-translation is given by L(a,b,c) (x, y, z) = (ax, by, bz+
c). It follows that the induced map on tangent spaces at every point is given
by
∂ ∂
L(a,b,c)∗ ( ) = a
∂x ∂x
∂ ∂
L(a,b,c)∗ ( ) = b
∂y ∂y
∂ ∂
L(a,b,c)∗ ( )=b .
∂z ∂z
Using this, we can verify from the definitions that the listed vector fields are
left-invariant. For example,
    
∂ ∂ ∂
L(a2 ,b2 ,c2 )∗ y |(a1 ,b1 ,c1 ) = L(a2 ,b2 ,c2 )∗ b1 = b1 b2
∂z ∂z ∂z
is equal to
   
∂ ∂ ∂
y |(a1 ,b1 ,c1 )·(a2 ,b2 ,c2 ) = y |(a1 a2 ,b1 b2 ,c1 +b1 c2 ) = b1 b2 .
∂z ∂z ∂z
The computation for the other elements is similar. They restrict to the basis
∂ ∂ ∂
∂x , ∂y , ∂z for the tangent space at the unit (1, 1, 0), hence are a basis for the
left-invariant vector fields.
∂ ∂ ∂ ∂ ∂ ∂ ∂
Part (c): We compute [x ∂x , y ∂y ] = 0, [x ∂x , y ∂z ] = 0, [y ∂y , y ∂z ] = y ∂z .

4. (RA) Let {fn }∞ p

n=1 be a sequence of functions in L (R) with p ≥ 1 finite.
 1. Suppose supn ∥fn ∥Lp (R) < ∞ and fn → f almost everywhere. Show that
∥f ∥Lp (R) < ∞.
2. Give an example of {fn }∞
n=1 converging almost everywhere to f such that
∥fn − f ∥Lp (R) ̸→ 0.
3. Give an example of {fn }∞
n=1 such that ∥fn − f ∥Lp (R) → 0 but fn − f → 0
holds nowhere.

Solution

1. Fatou implies ∥f ∥pLp (Rd ) ≤ lim inf n→∞ ∥fn ∥pLp (Rd ) , so we are done.
2. Let fn (x) = χ[n,n+1] (x). We have ∥fn ∥Lp (Rd ) = 1 for all n, but fn → f =
0 pointwise, so ∥fn − f ∥Lp (Rd ) = ∥fn ∥Lp (Rd ) = 1.
3. For any n, k, let fn,k = χ[k2−n ,(k+1)2−n ] . We know that ∥fn,k ∥Lp (R) =
2−n/p → 0, but any x ∈ R is contained infinitely many intervals of the
form [k2−n , (k + 1)2−n ] and thus fn,k → 0 holds nowhere.

5. (AG)

(a) Let X ⊂ Pn be a Zariski closed subset. Define the Hilbert function hX (m)
and the Hilbert polynomial pX (m).
(b) Suppose X = {p1 , . . . , pd } ⊂ Pn . Show that hX (d − 1) = d.
(c) Again, suppose X = {p1 , . . . , pd } ⊂ Pn . Show that hX (d − 2) = d unless
X is contained in a line.

Solution Let S = K[Z0 , . . . , Zn ] be the homogeneous coordinate ring of Pn ,

and let I(X) ⊂ S be the homogeneous ideal of X. The Hilbert function hX (m)
is defined to be the codimension of the mth graded piece I(X)m of I(X) in
Sm (equivalently, the dimension of the mth graded piece of the homogeneous
coordinate ring of X). For large values of m this is a polynomial in m, called
the Hilbert polynomial of X.
For the second part, we have to exhibit for each i = 1, . . . , d a polynomial
P (Z0 , . . . , Zn ) of degree d − 1 vanishing at pj for j ̸= i but nonzero at pi ; for
this, just choose for each j a linear form Q Lj vanishing at pj but nonzero at all
the other points of X and take P = j̸=i Lj .
For the third part, we use a variant of the construction above. We have to
show that if X is not contained in a line, for each i = 1, . . . , d a polynomial
P (Z0 , . . . , Zn ) of degree d − 2 vanishing at pj for j ̸= i but nonzero at pi .
To do this, note that for any i we can find pj , pk ∈ X not collinear with
 pi . Let MQbe a linear form vanishing at pj and pk but not at pi , and take
P := M · l̸=i,j,k Ll .

6. (AT) Let Σg denote a closed oriented surface of genus g. Prove that there is
a degree 1 map Σg → Σh if and only if g ≥ h.

Solution:
If g ≥ h, then Σg ∼= Σh #Σg−h , so there is a map Σg → Σh obtained by
crushing Σg−h to a point. Since this map is an orientation-preserving homeo-
morphism over an open subset of Σh , it is degree 1.
If g < h, let f : Σg → Σh be degree 1. Then the induced map f ∗ :
H 1 (Σh ; Q) → H 1 (Σg ; Q) must have nontrivial kernel for dimension reasons.
Let x ̸= 0 be an element of this kernel. By Poincaré duality, we know that
there is some y ∈ H 1 (Σh ; Q) such that x ∪ y is generator of H 2 (Σh ; Q). But
this is a contradiction, since f induces an isomorphism on H 2 by assumption
and f ∗ (x ∪ y) = f ∗ (x) ∪ f ∗ (y) = 0.
 QUALIFYING EXAMINATION
Harvard University
Department of Mathematics
Thursday September 5, 2024 (Day 3)

1. (DG) Let U be a coordinate ball in R2 with coordinates u1 , u2 and Φ be a C ∞

map from U to an open subset Ω of R3 with coordinates x1 , x2 , x3 such that
the image of Φ is a closed smooth surface X in Ω which is diffeomorphic to U
under Φ. Let ds2U be the Riemannian metric Φ∗ (dx21 + dx22 + dx23 ) on U which
is the Φ-pullback of the Euclidean metric dx21 + dx22 + dx23 on R3 . Denote by ∇
the covariant differentiation on U with respect to the Levi-Civita connection
of the Riemannian metric ds2U . Recall that the Levi-Civita connection is char-
acterized by the two properties that (i) parallel transport preserves the metric
and (ii) the connection is torsion-free. In a general Riemannian manifold this
characterization of the Levi-Civita connection is known as the fundamental
theorem of Riemannian geometry. For any vector fields
∂ ∂ ∂ ∂
ξ = ξ1 + ξ2 and η = η1 + η2
∂u1 ∂u2 ∂u1 ∂u2
on U , denote by ⃗vξ (respectively ⃗vη ) the vector field on X which is the Φ-image
of ξ (respectively η). Denote by ∂⃗vη ⃗vξ the (Euclidean) partial differentiation
of ⃗vξ in R3 with respect to the vector field ⃗vη in R3 . Prove that the Φ-image
of the covariant derivative ∇η ξ of the vector field ξ with respect to the vector
field η is equal to the projection of ∂⃗vη ⃗vξ onto the tangent space of X.

Hint: Use the fundamental theorem of Riemannian geometry which charac-

terizes the Levi-Civita connection by its two properties.

Solution. Let ⃗n be the unit tangent vector to X in R3 . We use the two char-
acteristic properties of the Levi-Civita connection to check that the covariant
ˆ defined by
differentiation ∇
ˆ ⃗v ⃗vξ = ∂⃗v ⃗vξ − ⃗n · ∂⃗v ⃗vξ ⃗n



∇ η η η

(which is the Euclidean partial differentiation followed by projection onto the

tangent space of X) is compatible
with the metric and has no torsion. Here
the scalar coefficient ⃗n · ∂⃗vη ⃗vξ of the second term on the right is the Euclidean
inner product of the two vectors in R3 . The first property means
ˆ ⃗v ⃗vξ ) · ⃗vξ ,
⃗vη (⃗vξ · ⃗vξ ) = 2(∇ η

which follows from

⃗vη (⃗vξ · ⃗vξ ) = 2(∂⃗vη ⃗vξ ) · ⃗vξ ,
 because ⃗vξ is tangential to X and orthogonal to ⃗n so that the inner product
ˆ ⃗v ⃗vξ and ⃗vξ .
between ∂⃗vη ⃗vξ and ⃗vξ is the same as the inner product between ∇ η

The torsion-free property of ∇ ˆ means that

ˆ ⃗v ⃗vη − ∇
∇ ˆ ⃗v ⃗vξ
ξ η

is equal to the Lie bracket [⃗vξ , ⃗vη ]. Since

∂⃗vξ ⃗vη − ∂⃗vη ⃗vξ = [⃗vξ , ⃗vη ]

and [⃗vξ , ⃗vη ] is tangential to X, it follows that

⃗n · ∂⃗vξ ⃗vη − ⃗n · ∂⃗vη ⃗vξ = 0

and
ˆ ⃗v ⃗vη − ∇
ˆ ⃗v ⃗vξ = ∂⃗v ⃗vη − ⃗n · ∂⃗v ⃗vη ⃗n − ∂⃗v ⃗vξ − ⃗n · ∂⃗v ⃗vξ ⃗n







∇ ξ η ξ ξ η η




= ∂⃗vξ ⃗vη − ∂⃗vη ⃗vξ − ⃗n · ∂⃗vξ ⃗vη − ⃗n · ∂⃗vη ⃗vξ ⃗n
= ∂⃗vξ ⃗vη − ∂⃗vη ⃗vξ = [⃗vξ , ⃗vη ] ,

which means that the torsion-free property of ∇ˆ is also verified. We conclude

ˆ
that the Φ-image of ∇η ξ is equal to ∇⃗vη ⃗vξ which is the projection of ∂⃗vη ⃗vξ
onto the tangent space of X.

2. (RA)
(a) Let n ≥ 2 and p1 , . . . , pn be real numbers > 1 such that
1 1
+ ··· + = 1.
p1 pn
Let −∞ ≤ a < b ≤ ∞ and fj be an Lpj function on (a, b) for 1 ≤ j ≤ n.
Denote by ∥fj ∥Lpj ((a,b)) the Lpj norm of fj on (a, b). Prove that

∥f1 · · · fn ∥L1 ((a,b)) ≤ ∥f1 ∥Lp1 ((a,b)) · · · ∥fn ∥Lpn ((a,b)) .

(b) Let p, q, r ∈ [1, ∞] such that

1 1 1
+ = + 1.
p q r
Let f be an Lp function on R and g be an Lq function on R. Let f ∗ g
be the convolution of f and g in the sense that
Z
(f ∗ g)(x) = f (y)g(x − y)dy.
R

Prove that
∥f ∗ g∥Lr (R) ≤ ∥f ∥Lp (R) ∥g∥Lq (R) .
 Hint: Show that the desired inequality is equivalent to
Z
f (y)g(x − y)h(x)dxdy ≤ ∥f ∥Lp (R) ∥g∥Lq (R) ∥h∥Ls (R)
(x,y)∈R2

for h ∈ Ls (R) with

1 1
+ = 1.
r s
Consider Hölder’s inequality for the following three functions on R2
p−1 q−1 s−1
(|g(x − y)|q |h(x)|s ) p , (|f (y)|p |h(x)|s ) q , and (|f (y)|p |g(x − y)|q ) s

and use
1 1 1
p + q + s = 1.
p−1 q−1 s−1

Solution. In the arithmetic-geometric mean inequality

ξ1p1 ξnpn
ξ1 · · · ξn ≤ + ··· +
p1 pn
for ξ1 ≥ 0, · · · , ξn ≥ 0, set
|fj (x)|
ξj =
∥fj ∥Lpj (R)
for x ∈ R. The desired inequality follows from integrating over R with respect
to x.

For the second part, the reformulation of the desired inequality in the hint is
simply the appropriately-interpreted duality between Lr space and Ls space.
From
1 1 1
+ + =2
p q s
we have
p(q − 1) p(s − 1) q(s − 1) q(p − 1) s(p − 1) s(q − 1)
+ = 1, + = 1, and + =1
q s s p p q
so that
p−1 q−1 s−1
(|g(x − y)|q |h(x)|s ) p (|f (y)|p |h(x)|s ) q (|f (y)|p |g(x − y)|q ) s = |f (y)g(x−y)h(x)|.

With the use of

1 1 1
p + q + s = 1,
p−1 q−1 s−1
 the application of Hölder’s inequality for three functions (on R2 instead R)
yields
Z Z
f (y)g(x − y)h(x)dxdy ≤ |f (y)g(x − y)h(x)| dxdy
(x,y)∈R2 (x,y)∈R2
p−1 q−1
≤ (|g(x − y)|q |h(x)|s ) p p (|f (y)|p |h(x)|s ) q q ·
L p−1 (R2 ) L q−1 (R2 )
s−1
· (|f (y)|p |g(x − y)|q ) s s
L s−1 (R2 )
! p−1 ! q−1
Z p Z q

= |g(x − y)|q |h(x)|s dxdy |f (y)|p |h(x)|s dxdy ·

(x,y)∈R2 (x,y)∈R2
! s−1
Z s

· |f (y)|p |g(x − y)|q dxdy

(x,y)∈R2
Z  q−1 + s−1 Z  p−1 + s−1 Z  p−1 + q−1
q s p s p q
p q s
= |f (y)| dy |g(y)| dy |h(x)| dy
y∈R y∈R x∈R
′
(from Fubini s theorem of integrating with respect to x or y first)
Z  1 Z  1 Z 1
p q s
p q s
= |f (y)| dy |g(y)| dy |h(x)| dy ,
y∈R y∈R x∈R

which is the inequality we would like to prove, where the last step comes from
q−1 s−1 1 s−1 p−1 1 p−1 q−1 1
+ = , + = , and + = .
q s p s p q p q s

3. (AG) Let X ⊂ Pn be a variety of dimension k, let G(1, n) be the Grassmannian

parametrizing lines in Pn and let

F1 (X) := {L ∈ G(1, n) | L ⊂ X}

be the locus of lines contained in X. Show that

dim F1 (X) ≤ 2k − 2

with equality holding only if X is a k-plane in Pn .

Solution Let ∆ ⊂ X × X be the diagonal. We have a surjective map

(X × X) \ ∆ → G(1, n)

sending (p, q) to p, q, and this map has fiber dimension 2 over F1 (X) ⊂ G(1, n),
establishing the inequality. Moreover, equality can hold only if the image of ϕ
 is equal to F1 (X), which is to say X contains the line joining any two points
of X, in which case X must be a linear subspace of Pn .

4. (AT) What are the homology groups of the 5-manifold RP2 × RP3 ,

(a) with coefficients in Z?

(b) with coefficients in Z/2?
(c) with coefficients in Z/3?

Solution. RP2 and RP3 have cell complexes with sequences

0→Z→Z→Z→0 and 0→Z→Z→Z→Z→0

where the maps are alternately 0 and multiplication by 2; from this the ho-
mology groups of RP2 and RP3 can be calculated as Z, Z/2, 0 and Z, Z/2, 0, Z
respectively. The rest is just Künneth and the universal coefficient theorem;
the answers are
(a): Z, (Z/2)2 , (Z/2)2 , Z, Z/2, 0;
(b): Z/2, (Z/2)2 , (Z/2)3 , (Z/2)3 , (Z/2)2 , Z/2,
(c): Z/3, 0, 0, Z/3, 0, 0

5. (CA) Suppose f, g are holomorphic on a neighborhood of the closed unit disc

in C. Suppose that f (z) has a simple zero at z = 0 and vanishes nowhere else
on the closed unit disc. For any ε > 0, define fε (z) = f (z) + εg(z).

1. Show that if ε > 0 is small enough, then fε has a unique zero in the
closed unit disc.
2. Show that if ε > 0 is small enough and the unique zero from part (1) is
denoted by zε , then zε → 0 as ε → 0 from above.

Solution.

1. Since f (z) ̸= 0 for any |z| = 1, |f (z)| must attain a strictly positive
minimum on the unit circle. Also, |g(z)| must attain a finite maxi-
mum on the unit circle. So, we can choose ε > 0 small enough so that
ε|g(z)| ≤ |f (z)| for all |z| = 1. By Rouche’s theorem, we deduce that
f (z) and fε (z) = f (z) + εg(z) have the same number of zeros (counting
multiplicity) on the unit disc.
 2. Fix any δ > 0; we must show for ε > 0 small enough, we have |zε | ≤ δ.
Take the closed disc around 0 of radius δ. Because f (z) ̸= 0 for any z on
the boundary of said disc, |f (z)| must attain a strictly positive minimum,
and |g(z)| must attain a finite maximum, both on said circle. Thus, if
ε > 0 is small enough, we have ε|g(z)| ≤ |f (z)| on said circle, and thus f
and fε must have the same number of zeros on the closed disc of radius
δ. But this zero must be zε by part (1), so |zε | ≤ δ.

6. (A) Let G be a finite group, and let χ1 , χ2 , . . . , χr be the irreducible characters

of G. Let H be a subgroup of G, and ψ be an irreducible character of H. Show
that the integers d1 , d2 , . . . , dr for which ψ G = d1 χ1 + d2 χ2 + · · · + dr χr satisfy
r
X
d2i ≤ [G : H].
i=1

Solution. We have [G : H]ψ(1) = ψ G (1) = d1 χ1 (1) + d2 χ2 (1) + · · · + dr χr (1),

where
di = ⟨ψ G , χi ⟩G = ⟨ψ, χi |H ⟩H
by the Frobenius Reciprocity Theorem. Since ψ is irreducible, this means that
for every i, χi |H = di ψ + α, where α is a character of H or α = 0. Thus,
χi (1) ≥ di ψ(1) for every i, so

[G : H]ψ(1) ≥ (d21 + d22 + · · · + d2r )ψ(1)

whence the desired inequality.