QUALIFYING EXAMINATION

Harvard University
Department of Mathematics
Tuesday September 3, 2019 (Day 1)

1. (AT) Suppose that M is a compact connected manifold of dimension 3, and

that the abelianization (π1 M )ab is trivial. Determine the homology and co-
homology groups of M (with integer coefficients).

Solution: Since the obstruction to orienting M is a homomorphism from π1 M

to the abelian group {±1} we know that M is orientable, and we may avail
ourselves of Poincaré duality. By Poincaré’s theorem H1 (M ) = π1 (M )ab = 0.
By Poincaré duality this gives us H 3 (M ) = Z and H 2 (M ) = 0. We now use
the universal coefficient sequence

0 → Ext(Hi−1 M, Z) → H i (M ) → Hom(Hi M ; Z) → 0

to conclude that H 0 (M ) = Z and H 1 (M ) = 0, from which, say, by Poincaré

duality you can conclude that H3 (M ) = Z and H2 (M ) = 0. In short M has
the homology and cohomology of S 3 .

2. (A) Prove that for every finite group G the number of groups homomorphisms
h : Z2 → G is n|G| where n is the number of conjugacy classes of G.

Solution. Z2 = ha, b | ab = bai, so the homomorphisms Z2 → G are in

bijection with pairs (g, h) ∈ G such that gh = hg.
Given g0 ∈ G, the group elements h such that g0 h = hg0 constitute the
stabilizer of g0 under the action of G on itself by conjugation. The orbit of g0
is its conjugacy class [g0 ], so the number of solutions (g0 , h) is |G|/|[g0 ]| (orbit-
stabilizer theorem). Thus for each conjugacy class c the number of solutions
with g ∈ c is |c| · |G|/|c| = |G|. Summing over the n conjugacy classes gives
n|G|, Q.E.D.

3. (AG) Let X ⊂ Pn be a projective variety over a field K, with ideal I(X) ⊂

K[Z0 , . . . , Zn ] and homogeneous coordinate ring S(X) = K[Z0 , . . . , Zn ]/I(X).
The Hilbert function hX (m) is defined to be the dimension of the mth graded
piece of S(X) as a vector space over K.

a. Define the Hilbert polynomial pX (m) of X.

 b. Prove that the degree of pX is equal to the dimension of X.
c. For each m, give an example of a variety X ⊂ Pn such that hX (m) 6=
pX (m).

Solution:

a. The Hilbert polynomial pX (m) is the unique polynomial such that pX (m) =
hX (m) for all sufficiently large integers m.
b. By induction on the dimension of X: if X is a finite set of d points,
then the Hilbert polynomial is the constant d; and in general the Hilbert
polynomial pX∩H of a general hyperplane section of a variety X is the
first difference of the Hilbert polynomial pX
c. Let X consist of any k distinct points of Pn . Then X is a variety of
dimension 0 and degree k, so by the previous part pX (m) = k. But
hX (m) is at most the dimension of the space of homogeneous degree m
polynomials in n + 1 variables, so for sufficiently large k, hX (m) < k =
pX (m).

4. (CA) Use contour integration to prove that for real numbers a and b with
a > b > 0, Z π
dθ π
=√ .
0 a − b cos θ a − b2
2

Solution: Since the integrand is symmetric

R 2π about θ = π, we can extend the
integration interval to [0, 2π], I = 21 0 a−bdθcos θ , which can be written as an
integral around an unit circle in the complex plane
I I
1 1 dz 1 dz
I= b 1
= 2
2 a − 2 (z + z ) iz 2az − bz − b i
I
1 1
=− 2 2a dz.
bi z − b z+1
2a
Then the singular points are the zeros of the denominator, z 2 − b z + 1 = 0.
The roots z1 , z2 are
1 p
1 p

z1 = a − a2 − b2 , z2 = a + a2 − b2 .
b b
 Note that only z1 lies inside the unit circle. Thus
I
dz 1
= 2πiRes|z=z1 [ ]
(z − z1 )(z − z2 ) (z − z1 )(z − z2 )
1
= 2πi lim (z − z1 )
z→z1 (z − z1 )(z − z2 )
1
= 2πi
(z1 − z2 )
b
= −2πi √ .
2 a − b2
2

Therefore
1 b
π
I=− · − 2πi √ =√ .
bi 2 a2 − b2 a2 − b2

5. (RA) Dirichlet’s function D is the function on [0, 1] ⊂ R that equals 1 at

every rational number and equals 0 at every irrational number. Thomae’s
function T is the function on [0, 1] whose value at irrational numbers is 0 and
whose value at any given rational number r is 1/q, where r = p/q with p and
q relatively prime integers, q > 0.

1. Prove that D is nowhere continuous.

2. Show that T is continuous at the irrational numbers and discontinuous
at the rational numbers.
3. Show that T is nowhere differentiable.

Solution. For the first part: for any rational number α we can find a sequence
of irrational numbers αn converging to α; since lim D(αn ) 6= D(α), D cannot
be continuous at α. Similarly, if α is irrational, we can find a sequence of
rational numbers αn converging to α to conclude that D cannot be continuous
at α.
For the second, exactly the same argument shows that T is discontinuous
at rational numbers. But now if α is irrational and αn = pn /qn is a se-
quence of rational numbers converging to α then we must have lim qn = ∞,
so lim T (αn ) = T (α) and T is continuous at α.
Finally, suppose that α is irrational. We can certainly find a sequence of
irrational numbers βn 6= α converging to α, from which we see that if the
limit T 0 (α) exists it must be 0. But for any n we can also find a rational
number αn = pn /2n with denominator 2n such that |α − αn | ≤ 1/2n , from
which we see that if the limit T 0 (α) exists it must be ≥ 1.
6. (DG)
Consider the Riemannian manifold (D, g) with D the unit disk in R2 and
1
g= (dx2 + dy 2 )
1 − x2 − y 2

Find the Riemann curvature tensor of (D, g). Use this to read off the Gaussian
curvature of (D, g).
Solution. This is a straightforward computation. We have

1 − x2 − y 2
 
−1 0
g =
0 1 − x2 − y 2

We set x1 = x and x2 = y and compute the Christoffel symbols

x
Γ111 = Γ212 = Γ221 = −Γ122 =
1 − x2 − y 2
y
− Γ211 = Γ112 = Γ121 = Γ222 = .
1 − x2 − y 2
With these, we can compute the only independent component of the Riemann
curvature tensor:
     
∂ ∂ ∂ ∂ ∂ ∂ ∂ ∂ −2
R , , , =g R , , =
∂x ∂y ∂x ∂y ∂x ∂y ∂y ∂x (1 − x2 − y 2 )2

(Indeed, by antisymmetry, all other components of the Riemann curvature

tensor either vanish or are equal to this up to a sign.)
The Gaussian curvature K satisfies

Rabcd = K(gac gbd − gad gbc )

which implies K = −2.

 QUALIFYING EXAMINATION
Harvard University
Department of Mathematics
Wednesday September 4, 2019 (Day 2)

1. (CA) Fix a ∈ C and an integer n ≥ 2. Show that the equation az n + z + 1 = 0

has a solution with |z| ≤ 2.

Solution: There are two cases. First, assume that |a| < 2−n . Let D denote
the disk where |z| ≤ 2 and let ∂D denote the circle |z| = 2. Let f (z) =
az n + z + 1 and let g(z) = z + 1. On ∂D, the function g − f obeys the
inequality |g(z) − f (z)| = |a||z|n < 1. Since this is less than |g(z)| for each
z ∈ ∂D, and since g has no zeros on ∂D, none of the members of the 1-
parameter family of functions {ft = f + t(g − f )}t∈[0,1] has a zero on ∂D.
Therefore f (which is ft = 0) and g (which is ft = 1) have the same number
of zeros (counting multiplicity) in D and the number is 1.
Now assume that |a| ≥ 2−n . By the fundamental Q theorem of algebra, the
function f (z) = az n + z + 1 factors as f (z) = a nk=1 (z − αk ), where αk ∈ C.
In particular,
Yn
(−1)n a αk = 1.
k=1

2n .
Q
Thus, |αk | ≤ This implies at least one of |αk | ≤ 2.

2. (AG) Let PN be the space of nonzero homogeneous polynomials of degree d

in n + 1 variables over a field K, modulo multiplication by nonzero scalars,
and let U ⊂ PN be the subset of irreducible polynomials F such that the zero
locus V (F ) ⊂ Pn is smooth.

(a) Show that U is a Zariski open subset of PN .

(b) What is the dimension of the complement D = PN \ U ?
(c) Show that D is irreducible.

Solution: In the product PN × Pn , consider the incidence correspondence

∂F
Φ = {(F, p) | F (p) = (p) = 0 ∀i}.
∂Xi

This is a closed subvariety of PN × Pn and hence projective, so that its image

D = π1 (Φ) ⊂ PN is closed; hence U is open. Moreover, since the projection
map π2 : Φ → Pn is a projective bundle with fiber PN −n−1 , we see that Φ is
 irreducible of dimension N − 1. Finally, since there exist hypersurfaces with
just one singular point (e.g., cones), the general fiber of Φ over a point in
its image D ⊂ PN is 0-dimensional; it follows that D is again irreducible of
dimension N − 1.

3. (RA) Let B denote the Banach space of continuous, real valued functions on
[0, 1] ⊂ R with the sup norm.

1. State the Arzela-Ascoli theorem in the context of B.

2. Define what is meant by a compact operator between two Banach spaces.
3. Prove that the operator T : B → B defined by
Z x
(T f )(x) = f (y) dy
0

is compact.

Solution. If B, B 0 are Banach spaces, a linear operator T : B → B 0 is said to

be compact if the closure of {T v : v ∈ B, kvk ≤ 1} (that is, the closure of the
image of the closed unit ball) is compact in B 0 .
For our T , the image of the closed unit ball is an equicontinuous family of
functions on [0, 1]. Indeed if f ∈ B with kf k ≤ 1 then
Z x0
0
(T f )(x ) − (T f )(x) = f (x) dx ≤ |x0 − x|
x

so that given  > 0 the same δ (namely δ = ) works uniformly for all such
T f . Moreover this image is uniformly bounded: (T f )(0) = 0 for all f , so
|(T f )(x)| ≤ x ≤ 1 for all x ∈ [0, 1]. Hence the closure of {T v : v ∈ B, kvk ≤ 1}
is compact by the Arzelà–Ascoli theorem.

4. (A) Let Fq be the finite field with q elements. Show that the number of 3 × 3
nilpotent matrices over Fq is q 6 .

Solution: Although Fq is not algebraically closed, there is still a Jordan normal

form for nilpotent matrices; by Cayley-Hamilton, for example, we know T 3 is
identically zero if T denotes our endomorphism of our vector space V , and we
may consider 0 ⊂ ker T ⊂ ker T 2 ⊂ ker T 3 = V . There are only three different
possibilities for the list of nontrivial dimensions, namely (3, 3), (2, 3), (1, 2),
 and choosing bases appropriately, we find that all nilpotent matrices are con-
jugate, over Fq , to exactly one of
   
0 1 0 0 1 0
0, 0 0
 0 , 0 0 1 .
0 0 0 0 0 0

There is exactly one matrix in the first case. For the latter two cases, we
have only to compute the stabilizer subgroup under the conjugation action
for each matrix above to find the size of the conjugacy orbit. First, recall
|GL3 (Fq )| = (q 3 − 1)(q 3 − q)(q 3 − q 2 ) by successively
 picking
 the images of
a11 a12 a13
the standard basis vectors; that a matrix a21 a22 a33  commutes with
a31 a32 a33
the second matrix above is equivalent to the conditions that a11 = a22 , a21 =
a31 = a23 = 0 and that the determinant vanish is that a11 = a22 , a33 be units,
and so the size of this stabilizer group is (q − 1)2 q 3 . Similarly, finding the
subgroup of matrices that commutes with the third matrix above is explicitly
equivalent to a11 = a22 = a33 , a12 = a23 , a21 = a31 = a32 = 0 of order (q−1)q 2 .
Hence, the total number of nilpotent matrices is

(q 3 − 1)(q 3 − q)(q 3 − q 2 ) (q 3 − 1)(q 3 − q)(q 3 − q 2 )

|N3 (Fq )| = 1 + +
(q − 1)2 q 3 (q − 1)q 2
= 1 + (q 3 − 1)(q + 1) + (q 3 − 1)(q 3 − q)
= q6.

5. (AT) Let Symn X denote the nth symmetric power of a CW complex X, i.e.
X n /Sn , where the symmetric group Sn acts by permuting coordinates. Show
that for all n ≥ 2, the fundamental group of Symn X is abelian.
Solution: This is a version of the Eckmann-Hilton argument. If one takes
as base-point some point in the (thin) diagonal X ⊂ Symn X, then every
based loop S 1 → Symn X lifts to a loop S 1 → X n , and so one directly has
π1 X n  π1 Symn X, but π1 X n ' (π1 X)n , and

(γ1 , · · · , γn ) = (γ1 , 1, · · · 1) ◦ (1, γ2 , 1, · · · , 1) ◦ · · · ◦ (1, · · · , 1, γn )

= (γ1 , 1, · · · , 1) ◦ (γ2 , 1, · · · , 1) ◦ · · · ◦ (γn , 1, · · · , 1)
= (γ1 · · · γn , 1, · · · , 1)

in π1 Symn X by simply “waiting” to do each loop γi in turn and then using

that as we’re in the symmetric product, it doesn’t matter in which factor we’re
doing the loop. So it suffices to show (γ, 1, · · · , 1) and (σ, 1, · · · , 1) commute
 but

(γ, 1, · · · , 1)(σ, 1, · · · , 1) = (γ, 1, · · · , 1)(1, σ, 1, · · · , 1)

= (γ, σ, 1, · · · , 1)
= (σ, γ, 1, · · · , 1)
= (σ, 1, · · · , 1)(γ, 1, · · · , 1)

by using the above reasoning a few more times.

6. (DG) Let S 2 ⊂ R3 be the unit 2-sphere, with its usual orientation. Let X be
the vector field generating the flow given by
   
cos(t) − sin(t) 0 x
 sin(t) cos(t) 0 · y  ,
0 0 1 z

and let ω be the volume form induced by the embedding in R3 (so the total
“volume” is 4π). Find a function f : S 2 → R satisfying

df = ιX ω

where ιX ω is the contraction of ω by X.

Solution: The outward unit normal to S 2 is

x∂x + y∂y + z∂z

so the volume form of S 2 is

ω = ιx∂x +y∂y +z∂z dx ∧ dy ∧ dz = x dy ∧ dz − y dx ∧ dz + z dx ∧ dy.

The vector field X is −y ∂x + x ∂y. So the contraction ιX ω is

y 2 dz − yz dy + x2 dz − xz dx = (x2 + y 2 ) dz − z(y dy + x dx).

Using the fact that on S 2 one has

x dx + y dy + z dz = 0,

this becomes
(x2 + y 2 + z 2 ) dz = dz.
We may therefore take f (x, y, z) = z.
 QUALIFYING EXAMINATION
Harvard University
Department of Mathematics
Thursday September 5, 2019 (Day 3)

1. (RA) Let f : [0, 1] → R be in the Sobolev space H 1 ([0, 1]); that is, functions
f such that both f and its derivative are L2 -integrable. Prove that
 Z 1 
−2πinx
lim n f (x)e dx = 0.
n→∞ 0

Solution. The quantity

Z 1
f (x)e−2πinx dx = fˆ(n)
0

is the n-th Fourier coefficient of f . Since f ∈ H 1 ([0, 1]), it holds that nfˆ(n)
is square-summable and, in particular, forms a null-sequence.

2. (CA) Given that the sum

X 1
(z − n)2
n∈Z

converges uniformly on compact subsets of C \ Z to a meromorphic function

on the entire complex plane, prove the identity

π2 X 1
2 = .
sin πz (z − n)2
n∈Z

Solution: Both sides of the desired equality are entire meromorphic functions
with double poles at the integers. Moreover, they have the same polar part
1/(z − n)2 at each n ∈ Z, so that the difference

π2 X 1
g(z) = 2 −
sin πz n∈Z (z − n)2

is an entire holomorphic function.

We claim now that g(z) is bounded, and hence by Liouville’s theorem constant.
By the periodicity g(z + n) = g(z) ∀n ∈ Z, it suffices to prove that it’s
bounded in the strip 0 ≤ <(z) ≤ 1, and hence that it’s bounded in the region
 0 ≤ <(z) ≤ 1, |=(z)| > N . But each of the two terms in the expression for
g(z) above has limit 0 as |=(z)| → ∞; it follows that g(z) is constant and
hence 0.

3. (AG) Let C ⊂ P3 be a smooth curve of degree 5 and genus 2.

(a) By considering the restriction map ρ : H 0 (OP3 (2)) → H 0 (OC (2)), show
that C must lie on a quadric surface Q.
(b) Show that the quadric surface Q is unique.
(c) Similarly, show that C must lie on at least one cubic surface S not con-
taining Q.
(d) Finally, deduce that there exists a line L ⊂ P3 such that the union C ∪ L
is a complete intersection of a quadric and a cubic.

Solution: For the first part, we know that h0 (OP3 (2)) = 10, while by Riemann-
Roch we have h0 (OC (2)) = 9; thus the map ρ must have a kernel. For the
second, observe that Q must be irreducible, since C cannot lie in a plane
(there are no smooth plane curves of degree 5). If there were a second quadric
Q0 6= Q containing C, then, the intersection Q ∩ Q0 would be one-dimensional
and so by Bezout of degree at most 4.
Similarly, we have h0 (OP3 (3)) = 20, while by Riemann-Roch we have h0 (OC (3)) =
14, so there is at least a 6-dimensional vector space of cubic polynomials van-
ishing on C; only a four-dimensional subspace of these can be multiples of the
quadratic polynomial defining Q. Finally, if S is any cubic surface containing
C but not containing Q, by Bezout the intersection S ∩ Q will have to consist
of the union of C and a line L.

4. (A) Show that if p, q are distinct primes then the polynomial (xp −1)/(x−1) is
irreducible mod q if an only if q is a primitive residue of p (i.e. if every integer
that is not a multiple of p is congruent to q e mod p for some integer e).
ii) Prove that x6 + x5 + x4 + x3 + x2 + x + 1 factors mod 23 as the product
of two irreducible cubics.

Solution. Since p is not 0 mod q, the zeros of xp −1 mod q are distinct (a multi-
ple root would have pxp−1 = 0). So the zeros of fp (x) := (xp − 1)/(x − 1) mod q
are the p − 1 nontrivial p-th roots of unity in the splitting field, call it F , of fp
over k := Z/qZ. Now fp is irreducible if and only if the Gal(F/k) permutes
those zeros transitively. For an extension of finite fields, Gal(F/k) is gener-
ated by the Frobenius automorphism φ : x 7→ xq . If ζ is a root of fp then the
 e
other roots are ζ n with n ranging over (Z/pZ)∗ , and φe takes ζ to ζ q , so the
Galois orbit contains all the roots if and only if q is a generator of (Z/pZ)∗ ,
Q.E.D.
ii) Here (p, q) = (7, 23) and 23 ≡ 2 mod 7. The smallest e > 0 such that
2e ≡ 1 mod 7 is e = 3, so the Galois orbits on nontrivial 7th roots of unity
have size 3, whence each Galois-stable factor of f7 has degree 3, Q.E.D. [The
factorization is (x3 + 10x2 + 9x − 1)(x3 − 9x2 − 10x − 1).]

5. (DG) Suppose that G is a Lie group.

(a) Consider the map ι : G → G defined by ι(g) = g −1 . Show that the

derivative of ι at the identity element is multiplication by −1.
(b) For g ∈ G define maps Lg , Rg : G → G by

Lg (x) = gx
Rg (x) = xg.

Show that if ω is a k-form which is bi-invariant in the sense that L∗g ω =

Rg∗ ω then ι∗ ω = (−1)k ω.
(c) Show that bi-invariant forms on G are closed.

Solution: For the first part note that if g(t) = exp(λt) then g(t)−1 = exp(−λt).
The claim follows by taking the derivative with respect to t at t = 0. For the
second part note that

ι∗ ωg = ι∗ L∗g−1 ωe = Rg∗ ι∗ ωe

so it suffices to show that ι∗ ωe = (−1)k ωe . But if X1 , . . . , Xk are tangent

vectors at e then

ι∗ (ω)(X1 , . . . , Xk ) = ω(dι(X1 ), . . . , dι Xk )
= ω(−X1 , . . . , −Xk )
= (−1)k ω(X1 , . . . , Xk )

by the first part. The third part follows from the equations

(−1)(k+1) dω = ι∗ dω = dι∗ ω = (−1)k dω.

6. (AT) Suppose that m is odd. Show that if n is odd there is a fixed point free
action of Z/m on S n . What happens if n is even?

Solution: For the first part write n = 2k − 1 and regard S n as the unit sphere
in Cn . The formula

(z1 , . . . , zn ) 7→ e2πi/m (z1 , . . . , zn )

defines a free action of Z/m on S n . (This part does not require m to be odd).
There can be no free action on S 2k . This follows from the Lefschetz fixed
point formula. Since the automorphism group of Z is cyclic of order 2 and
m is odd, there are no non-trivial actions of Z/m on Z. It follows that the
Lefschetz number of any action is 2, so there must be a fixed point.