Qualifying Examination

HARVARD UNIVERSITY
Department of Mathematics
Tuesday, January 19, 2016 (Day 1)

PROBLEM 1 (DG)
Let S denote the surface in R3 where the coordinates (x, y, z) obey x2 + y2 = 1 + z2. This
surface can be parametrized by coordinates t ∈ R and θ ∈ R/(2πZ) by the map

(t, θ) → ψ(t, θ) = ( 1+ t 2 cos θ, 1+ t 2 sin θ, t).

a) Compute the induced inner product on the tangent space to S using these coordinates.
b) Compute the Gaussian curvature of the metric that you computed in Part a).
c) Compute the parallel transport around the circle in S where z = 0 for the Levi-Civita
connection of the metric that you computed in Part a).

PROBLEM 2 (T)
Let X be path-connected and locally path-connected, and let Y be a finite Cartesian
product of circles. Show that if π1(X) is finite, then every continuous map from X to Y is
null-homotopic. (Hint: recall that there is a fiber bundle Z → R → S1.)

PROBLEM 3 (AN)
Let K be the field C(z) of rational functions in an indeterminate z, and let F ⊂ K be the
subfield C(u) where u = (z6 + 1)/z3.
a) Show that the field extension K/F is normal, and determine its Galois group.
b) Find all fields E, other than F and K themselves, such that F ⊂ E ⊂ K. For each E,
determine whether the extensions E/F and K/E are normal.
PROBLEM 4 (AG)
The nodal cubic is the curve in CP2 (denoted by X) given in homogeneous coordinates
(x, y, z) by the locus {z y2 =x2 (x + z)}.
a) Give a definition of a rational map between algebraic varieties.
b) Show that there is a birational map from X to CP1.
c) Explain how to resolve the singularity of X by blowing up a point in CP2.

PROBLEM 5 (RA)
Let B and L denote Banach spaces, and let || · ||B and || · ||L denote their norms.
a) Let L: B → L denote a continuous, invertible linear map and let m: B ⊗ B → L denote
a linear map such that || m(φ ⊗ ψ) ||L ≤ || φ ||B ||ψ ||B for all φ, ψ ∈ B. Prove the following
assertions:
• There exists a number κ > 1 depending only on L such that if a ∈ B has norm less
than κ -2, then there is a unique solution to the equation Lφ + m(φ ⊗ φ) = a with
|| φ ||B < κ -1.
• The norm of the solution from the previous bullet is at most κ || a ||L.
b) Recall that a Banach space is defined to be a complete, normed vector space. Is the
assertion of Part a) of the first bullet always true if B is normed but not complete? If
not, explain where the assumption that B is complete enters your proof of Part a).

PROBLEM 6 (CA)
Fix a ∈ C and an integer n ≥ 2. Show that the equation a zn + z + 1 = 0 for a complex
number z necessarily has a solution with |z| ≤ 2.
PROBLEM 1 SOLUTION:

! and
Answer to a): The vector fields !t !
along S are
!!

!
!t = 1+tt2 (x !x
! ! )+ !
+ y !y !z and !
!!
!
= -y !x ! .
+ x !y

! , ! 〉 = t 2 + 1 〈 ! , ! 〉 = 0 and 〈 ! , ! 〉 = (1 + t2), it
Since their inner product is 〈 !t !t 1+t 2 !t !! !! !!
follows that the square of the line element for the induced metric is

2
ds2 = 1+2t2 dt ⊗ dt + (1 + t2) dθ ⊗ dθ.
1+t

2
Answer to b): The 1-forms e0 = ( 1+2 t2 )1/2 dt and e1 = (1 + t2)1/2dθ are orthonormal. Write
1+t
" 0 ! %
The connection matrix of 1-forms is A = $ with the 1-form Γ obeying
# -! 0 '&

de0 = -Γ ∧ e1 and de1 = Γ ∧ e0.

t
The unique solution is Γ = - dθ. The Gauss curvature is denoted by κ and it is
1+2t 2
1
defined by writing dΓ as κ e0 ∧ e1. Thus, κ = -( 1+2t 2
2 ) .

Answer to c): Since Γ = 0 on the z = 0 circle, the parallel transport is given by the
identity matrix when written using the orthonormal frame { !t ! , ! } for TS at (1, 0, 0).
!!

PROBLEM 2 SOLUTION:

Here are two solutions:

Solution 1: Let Y denote the space ×n S1. It is enough to prove that the map from X to Y
factors as a map

To prove this factorization, note that a map ƒ: X → Y lifts through a cover p: Y ! → Y if

! )) (they are both subgroups of π1(Y)). (See,
and only if ƒ*(π1(X)) is a subgroup of p∗(π1( Y
for example Proposition 1.33 in Hatcher’s book on algebraic topology.) Since π1(Rn) = 0
and ƒ∗ in this case must be the zero homomorphism, this condition is satisfied and so ƒ
lifts to some ƒ! . Because Rn is contractible, this lift is null-homotopic and any null-
homotopy pushes forward to give a null-homotopy of ƒ.

Solution 2: Recall that S1 (which is K(Z, 1)) classifies integral cohomology classes of
degree 1. As a consequence, a map X → Y is (up to homotopy) determined by an n-tupel
of elements in H1(X; Z). The universal coefficient short exact sequence in this degree is

The two end groups are zero: The right most group is zero because H1(X; Z) is the
Abelianization of π1(X) and thus it is a finite group; and finite groups have no non-trivial
homomorphisms to Z. The left most group is zero because H0(X: Z) = Z and Ext(Z; Z) is
trivial since Z is a free group. Thus H1(X; Z) = 0 and so all maps from X to Y are
homotopic to the constant map.

PROBLEM 3 SOLUTION:

Answer to a) One has [K: F] = 6 because the extension K/F is generated by the solution z
of the polynomial equation z6 − uz3 + 1 = 0 which has degree 6. The Galois group
contains the automorphisms α : z → 1/z and β : z → ρz, where ρ = e i2! /3 = (−1 + √−3)/2.
Since α and β have orders 2 and 3 respectively, the group G generated by α and β has
order at least 6. However, |Gal(K/F)| ≤ [K: F] = 6 with equality iff K/F is normal, so K/F
must be normal with Galois group G of order 6, which is readily identified with the
symmetric group the symmetric group S3 (for instance, via its permutation action on the
set {1, ρ, ρ2}).

Answer to b) By the fundamental theorem of Galois theory, the intermediate fields E of

the Galois extension K/F correspond to subgroups H ⊂ G by E = KH (fixed subfield); K/E
is always normal with Gal(K/E) = H, while E/F is normal iff H ⊴ G. Since F and K are
excluded, one need not consider H = G and H = {1}. The remaining subgroups are
A3 = 〈β〉, which yields the normal extension C(z3) of F, and three two-element subgroups
which yield non-normal extensions C(z + 1/z), C(z + ρz), C(z + ρ2z). (The fact that each
of these is indeed the corresponding KE can be confirmed by computing its degree as in
Part a).)

PROBLEM 4 SOLUTION:
Answer to a) A rational map from X to Y is an equivalence class of pairs (U, f ) where
U ⊂ X is a Zariski dense open subset and f :U → Y is a regular map. Two pairs (U, f)
and (V, g) are equivalent if f = g on the intersection U ∩ V.

Answer to b) The projection from the point (0,0,1) ∈ CP2 to the line where z=0 restricts
to a rational map p: X = {z y2 = x2(x + z)} → CP1. An inverse is given by the map given
in homogeneous coordinates by the rule (u, v) → (x = (v2−u2)u, y = (v2−u2)v, z = u3).
This is an inverse since x3 = (y2 − x2)z on X. It follows that p is a birational map.

Answer to c) Away from the line z = 0 the blowup of CP2 at (0,0,1) is given by the locus
{xt = ys} ⊂ {((x, y),(s, t))}=C2 × CP1. Consider the chart in CP1 where s ≠ 0. The blow
up of X is defined here by the equations xt = y and y2 = x2(x + 1). Substituting for y
gives the equation x2(t2 − x − 1) = 0 which has one irreducible component being the
locus x = y = 0 (which is the exceptional curve), and the other being the locus where both
t2 = x+1 and xt = y. This is the blow-up of X. In the chart where t ≠ 0, the blow up of X
is defined by the locus where x = ys and 1 = s2(sy + 1). By the Jacobian criterion the
curve defined by these equations is nonsingular.

PROBLEM 5 SOLUTION:

Answer to a) Since L is invertible, its inverse defines a bounded linear map from L to B
to be denoted by L-1. Using L-1, one can define a map T: B → B by the rule

T(φ) = L-1(a - m(φ, φ)).

This is relevant because φ is a fixed point of T (it obeys T(φ) = φ) if and only if φ obeys
the equation Lφ + m(φ ⊗ φ) = a. Let c denote the norm of the operator L-1. Then the
following are computations:

• || T(φ) ||B ≤ c (||a ||L + || φ ||B2).

• || T(φ) - T(φ´)|| ≤ 4 c (|| φ ||B + || φ´||B) || φ - φ´||B.

Given δ > 0, let B(δ) denote the ball of radius δ about the origin in B. If E > 0 and if
|| a ||L ≤ E then the top bullet implies that T maps B(δ) to B(cE + c δ2). Thus, if δ < (2c)-1
and if E < (2c)-1δ, then T maps B(δ) to itself. Meanwhile, if δ < (8c)-1 then the lower
bullet implies that || T(φ) - T(φ´)|| ≤ γ || φ - φ´||B for fixed γ < 1 when φ, φ´ ∈ B(δ). This
implies in turn that T is a contraction mapping of B(δ) to itself. The contraction mapping
theorem supplies a unique fixed point of T in B(δ) under these circumstances. Noting
again that an element φ ∈ B is a fixed point of T if and only if φ obeys Lφ + m(φ ⊗ φ) = a,
the top bullet follows if || a ||L ≤ (16 c)-1. Take κ to be the maximum of 4c1/2 and 8c to
obtaine the answer to the first bullet of Part a). The second bullet of Part a) follows
directly from the fact that φ = T(φ) and || φ ||B2 ≤ 12 || φ ||B because these and the inequality
|| T(φ) ||B ≤ c (||a ||L + || φ ||B2) imply that 1
2 || φ ||B ≤ c ||a ||L.

Answer to b) The completeness of B is required. Here is an example: Take B and L to

be the span of the polynomials functions on [-1, 1] with the norms || ƒ ||B = || ƒ ||L =
supt |ƒ(t)|. Take the equation φ + φ2 = δ t with δ being a small, non-zero number. A
solution, must be either φ = - 12 + 12 (1 + 4δ2 t2)1/2 or φ = - 12 - 12 (1 + 4δ2 t2)1/2; but neither is
in B. Note that the contraction mapping theorem does not hold if the Banach space in
question is not complete because the contraction mapping theorem constructs the desired
solution as a limit of a Cauchy sequence in B.

PROBLEM 6 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) = azn + 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 {fτ = f + τ (g - f)}τ∈[0,1] has a zero on ∂D. Therefore, f (which is fτ=0)
and g (which is fτ=1) have the same number of zeros (counting multiplicity) in D. This
number is 1 (This is Rouche’s theorem). Now assume that |a| ≥ 2-n. By the fundamental
theorem of algebra, the function f(z) = a zn + z + 1 factors as

n
f(z) = a " (z -
k =1
k )

where the {αk}k=1….,n are complex numbers. This implies in particular the identity

n
(-1)n a " ! k = 1.
k =1

n
hence " |! k | ≤ 2n. This can happen only if one or more roots αk are in D.
k =1
 Qualifying Examination
HARVARD UNIVERSITY
Department of Mathematics
Wednesday, January 20, 2016 (Day 2)

PROBLEM 1 (DG)
Let k denote a positive integer. A non-optimal version of the Whitney embedding
theorem states that any k-dimensional manifold can be embedded into R2k+1. Using this,
show that any k-dimensional manifold can be immersed in R2k. (Hint: Compose the
embedding with a projection onto an appropriate subspace.)

PROBLEM 2 (T)
Let X be a CW-complex with a single cell in each of the dimensions 0, 1, 2, 3, and 5 and
no other cells.
a) What are the possible values of H∗(X; Z)? (Note: it is not sufficient to consider
Hn(X; Z) for each n independently. The value of H1(X; Z) may constrain the value of
H2(X; Z), for instance.)
b) Now suppose in addition that X is its own universal cover. What extra information
does this provide about H∗(X; Z)?

PROBLEM 3 (AN)
Let k be a finite field of characteristic p, and n a positive integer. Let G be the group of
invertible linear transformations of the k-vector space kn. Identify G with the group of
invertible n × n matrices with entries in k (acting from the left on column vectors).
n-1
a) Prove that the order of G is ! (qn - qm ) where q is the number of elements of k.
m=0

b) Let U be the subgroup of G consisting of upper-triangular matrices with all diagonal

entries equal 1. Prove that U is a p-Sylow subgroup of G.
c) Suppose H ⊂ G is a subgroup whose order is a power of p. Prove that there is a basis
(v1, v2,..., vn) of kn such that for every h ∈H and every m ∈{1, 2, 3,..., n}, the vector
h(vm) - vm is in the span of {vd: d < m}.
PROBLEM 4 (AG)
Let X be a complete intersection of surfaces of degrees a and b in CP3. Compute the
Hilbert polynomial of X.

PROBLEM 5 (RA)
Let C 0 denote the vector space of continuous functions on the interval [0, 1]. Define a
norm on C 0 as follows: If ƒ ∈ C 0, then its norm (denoted by || ƒ ||) is

|| ƒ || = supt∈[0,1] |ƒ(t)| .

Let C ∞ denote the space of smooth functions on [0, 1]. View C ∞ as a normed, linear space
with the norm defined as follows: If ƒ ∈ C ∞, then its norm (denoted by || ƒ ||∗) is

|| ƒ ||∗ = !
[0,1]
d ƒ| + | ƒ|) dt .
(| dt

a) Prove that C 0 is Banach space with respect to the norm || · ||. In particular, prove that it
is complete.
b) Let ψ denote the ‘forgetful’ map from C ∞ to C 0 that sends ƒ to ƒ. Prove that ψ is a
bounded map from C ∞ to C 0, but not a compact map from C ∞ to C 0.

PROBLEM 6 (CA)
Let D denote the closed disk in C where |z| ≤ 1. Fix R > 0 and let ϕ: D → C denote a
continuous map with the following properties:
i) ϕ is holomorphic on the interior of D.
ii) ϕ(0) = 0 and its z-derivative, ϕ´, obeys ϕ´(0) = 1.
iii) |ϕ| ≤ R for all z ∈D.
Since ϕ´(0) = 1, there exists δ > 0 such that ϕ maps the |z| < δ disk diffeomorphically
onto its image. Prove the following:
a) There is a unique solution in [0, 1] to the equation 2Rδ = (1 - δ)3.
b) Let δ∗ denote the unique solution to this equation. If If 0 < δ < δ∗, then ϕ maps
the |z| < δ disk diffeomorphically onto its image.
PROBLEM 1 SOLUTION:

The desired immersion will come from a projection onto the orthogonal complement of a
suitably chosen, nonzero vector in R2k. To find this vector, let M denote the manifold in
question and let ƒ denote the embedding of M into R2k. Let g denote the map from TM to
R2k+1 that is defined as follows: Supposing that x ∈ M and v ∈ TM|x set g(x, v) = ƒ∗|x·v
where ƒ∗ denotes the differential of ƒ. Sard’s theorem can be invoked to see that g is not
surjective. Let a denote a point that is not in the image of g. (Note that a is necessarily
nonzero.) Use π to denote the projection onto the orthogonal complement of a. To see
that π ! ƒ is an immersion, let x denote a point in M and let v denote a nonzero vector in
TM|x. Suppose for the sake of argument that (π ! ƒ)∗v is zero. If this is so, then the chain
rule and the fact that π is linear implies that ƒ∗|x·v = ta for some nonzero t ∈ R. This
implies in turn that ƒ∗|x(t-1v) = a which is nonsense because a is in the complement of the
image of ƒ∗.

PROBLEM 2 SOLUTION:

Answer to a) The cellular chain complex for X must be of the form

Since X is connected, it must have H0(X; Z) = Z, so the map c must be zero. The only
other restriction is that the sequence form a complex, so b ! a = 0; but since b ! a is
multiplication by some integer, either a = 0 or b = 0. In the case a = 0 and b ≠ 0, the
homology groups take the form

In the case a ≠ 0 and b = 0, the homology groups take the form

In the remaing a = 0 = b case, they take the form

Answer to b) The assertion that X is its own universal cover is the same as the assertion
π1(X) = 0. But, since H1(X) = π1(X)ab, this means H1(X) = 0. The only case where this is
possible is when a = 0 and b ≠ 0. Moreover, since Z/b = 0 in this case, b must be a
multiplicative unit: b = ±1.

PROBLEM 3 SOLUTION:

Answer to a) The elements of G are in bijection with ordered bases (v1,...,vn) of kn (the
map takes each matrix to its columns). For each j ∈ {0, 1, 2,..., n−1}, once vi for all i ≤ j
has been chosen, then there are qn - qj choices for the index (j + 1) basis element because
any of the qn elements of kn except the qm linear combinations of v1, . . . , vj will do.
n-1
Hence the number of possible bases is ! (qn - qm ) .
m=0

Answer to b) Each factor qn - qj is qj times an integer not divisible by p because it is

congruent to -1 modulo q, and q is a multiple of p. Hence the number of elements in G is
n-1
qd times some integer not divisible by p, where d = ! j . But qd is the order of U because
j= 0

there are d entries above the diagonal, and a power of p. Hence U is a p-Sylow
subgroup of G.

Answer to c) U consists of the matrices h that satisfy the desired property with respect to
the standard basis of unit vectors. Hence the matrices h that satisfy this property for the
basis (v1,...,vn) constitute the subgroup of G obtained by conjugating U by the matrix with
columns v1, . . . , vn. But by Sylow’s second theorem H is contained in a conjugate of U.
PROBLEM 4 SOLUTION:

Let S = C[x0, x1, x2, x3] be the homogeneous coordinate ring of CP3. The coordinate ring
of X is of the form S/(f, g) for some irreducible polynomials f and g of degrees a, b
respectively. There is a four-term exact sequence of graded modules

with maps given by multiplication with f and g. Hence the Hilbert polynomial of X is

PROBLEM 5 SOLUTION:

Answer to a) One has to show that a Cauchy sequence {ƒn}n=1,2,… in C 0 converges to a

continuous function. To do this, note that for each t ∈ [0, 1], the sequence {ƒn(t)}n∈{1,2,…}
is a Cauchy sequence in R so it converges. Let ƒ(t) denote the limit. The assignment
t → ƒ(t) defines a function on [0, 1]. The task is to prove that this function is continuous.
This means the following: Given ε > 0, there exists δ > 0 such that |ƒ(t) - ƒ(t´)| < ε when
|t - t´| < δ. To find δ, first fix N so that |ƒn(t) - ƒm(t)| < 13 ε for all t ∈ [0, 1] and all pairs n,
m > N. This implies that |ƒn(t) - ƒ(t)| ≤ 13 ε for all t. Such N exists because {ƒn}n∈{1,2,…} is
a Cauchy sequence in C 0. To continue, take n > N and fix δ so that |ƒn(t) - ƒn(t´)| < 13 ε
when |t´ - t| < δ. It then follows by the triangle inequality that

|ƒ(t) - ƒ(t´)| ≤ |ƒ(t) - ƒn(t)| + |ƒ(t´) - ƒn(t´)| + |ƒn(t´) - ƒn(t)| < ε.

Answer to b) The map ψ is bounded because for all t, one has the identity

t
1
ƒ(t) = !0 ( ! dsd ƒ(s)ds + ƒ(r)) dr ,
r

and thus |ƒ(t)| ≤ || ƒ ||∗ for all t. It is not a compact map. To prove this, fix a smooth
function on [0, ∞) that is equal to 1 near t = 0 and equal to 0 for t > 12 . Call this function
ƒ. Define ƒn(t) = ƒ(n t). This function is smooth on [0, 1]. The sequence {ƒn(t)} has
bounded || · ||∗ norm but it has no convergent sequence in C 0.

PROBLEM 6 SOLUTION:

Answer to a) The function ƒ(δ) = 2Rδ/(1 - δ)3 has strictly positive derivative and
therefore defines a diffeomorphism from [0, 1) to [0, ∞). It follows from this that there
is a single point where ƒ is equal to 1.

Answer to b) To obtain the asserted lower bound for δ, note that ϕ maps the disk where
|z| < δ diffeomorphically to its image if it is 1-1 on this disk and if |ϕ´| > 0 on this disk.
The Cauchy integral formula is used to see when this happens. Here is Cauchy’s
formula:

ϕ(z) = 1
2!i " z 1- w !(w)dw .
!D

Differentiating this, one sees that |ϕ´´| on the |z| < δ disk is bounded by 2R(1 - δ)-3. This
implies that

|ϕ´ - 1| < 2 R δ (1 - δ)-3 where |z| < δ.

If ϕ´ > 0, then ϕ is a local diffeomorphism. This is the case when δ < δ∗ with δ∗ being the
solution in (0, 1) to the equation 2Rδ∗(1 - δ∗)-3 = 1. Meanwhile, if z, z´ have norm less
than δ, then |ϕ(z) - ϕ(z´)| ≥ (1 - 2 R δ (1 - δ)-3) |z - z´| which is a positive multiple of |z - z´|
precisely when δ < δ∗.
 Qualifying Examination
HARVARD UNIVERSITY
Department of Mathematics
Thursday, January 21, 2016 (Day 3)

PROBLEM 1 (DG)
Recall that a symplectic manifold is a pair (M, ω), where M is a smooth manifold and ω
is a closed nondegenerate differential 2-form on M. (The 2-form ω is called the
symplectic form.)
a) Show that if H: M → R is a smooth function, then there exists a unique vector field, to
be denoted by XH, satisfying ! XH " = dH. (Here, ι denotes the contraction operation.)
b) Supposing that t > 0 is given, suppose in what follows that the flow of XH is defined
for time t, and let φt denote the resulting diffeomorphism of M. Show that φt*ω = ω.
c) Denote the Euclidean coordinates on R4 by (x1, y1, x2, y2) and use these to define the
symplectic form ω0 = dx1 ∧ dy1 + dx2 ∧ dy2. Find a function H: R4 → R such that the
diffeomorphism φt=1 that is defined by the time t = 1 flow of XH fixes the half space
where x1 ≤ 0 and moves each point in the half space where x1 ≥ 1 by 1 in the y2
direction.

PROBLEM 2 (T)
Let X denote a finite CW complex and let ƒ: X → X be a self-map of X. Recall that the
Lefschetz trace of ƒ, denoted by τ(ƒ), is defined by the rule

"
τ(ƒ) = # (-1) n
tr(ƒ n : H n (X; Q) ! H n (X; Q))
n=0

with ƒn denoting the induced homomorphism. Use τ(·) to answer the following:
a) Does there exist a continuous map from RP2 to itself with no fixed points? If so, give
an example; and if not, give a proof.
b) Does there exist a continuous map from RP3 to itself with no fixed points? If so, give
an example; and if not, give a proof.

PROBLEM 3 (AN)
Let A be the ring Z[ 5 2016 ] = Z[X]/(X5 - 2016). Given that 2017 is prime in Z,
determine the factorization of 2017·A into prime ideals of A.
PROBLEM 4 (AG)
a) State a version of the Riemann–Roch theorem.
b) Apply this theorem to show that if X is a complete nonsingular curve and P ∈ X is any
point, there is a rational function on X which has a pole at P and is regular on X−{P}.

PROBLEM 5 (RA)
Let ℘ denote a probability measure for a real valued random variable with mean 0.
Denote this random variable by x. Suppose that the random variable |x| has mean equal
to 2.
a) Given R > 2, state a non-trivial upper bound for event that x ≥ R. (The trivial upper
bound is 1.)
b) Give a non-zero lower bound for the standard deviation of x.
c) A function ƒ on R is Lipshitz when there exists a number c ≥ 0 such that

|ƒ(p) - ƒ(p´)| ≤ c |p - p´| for any pair p, p´ ∈ R.

ˆ denote the function on R whose value at a given p ∈ R is the expectation of the

Let !
random variable e ip x . (This is the characteristic function of ℘.) Give a rigorous
ˆ is Lipshitz and give an upper bound for c in this case.
proof that !
d) Suppose that the standard deviation of x is equal to 4. Let N denote an integer greater
than 1, and let {x1, …, xN} denote a set of independent random variables each with
1
probabilities given by ℘. Use SN to denote the random variable N (x1 + ···· + xN).
The central limit theorem gives an integral that approximates the probability of the
event where SN ∈ [-1, 1] when N is large. Write this integral.

PROBLEM 6 (CA)
Let H ⊂ C denote the open right half plane, thus H = {z = x + iy: x > 0}. Suppose that
ƒ: H → C is a bounded, analytic function such that ƒ(1/n) = 0 for each positive integer n.
Prove that ƒ(z) = 0 for all z.
N
(Hint: Consider the behavior of the sequence of functions {hN(z) = ! zz +- 11/n/n }
n =1
N=1.2… on H

and, in particular, on the positive real axis.}

PROBLEM 1 SOLUTION:

Answer to a) To say that ω is non-degenerate is to say that the contraction operation

defines a vector bundle isomorphism between TM and T*M.

!
Answer to b) The definition of the Lie derivative is such that !t (φt*ω) = φt*( L XH ! ) with
L XH ! denoting the Lie derivative of ω along the vector field XH. Cartan’s formula for
L XH ! is L XH ! = d( ! XH " ) + ! XH d" and both of these terms are zero. Thus, φt*ω is
independent of t and thus equal to its value at t = 0 which is ω.

Answer to c) Choose a smooth function ƒ: R → [0, 1] so that ƒ(s) = 0 for s ≤ 0 and

ƒ(s) = 1 for s ≥ 1. The function sending (x1, y1, x2, y2) → H(x1, y1, x2, y2) = -ƒ(x1) x2 has
the desired properties because XH = 0 for x1 ≤ 0 and XH = !y! for x1 ≥ 1.
2

PROBLEM 2 SOLUTION:

Answer to a) The Lefschetz trace theorem states that if τ(f) ≠ 0, then f must have a fixed
point. To see that τ(ƒ) is never zero, note first that the rational homology of RP2 is zero
except for H0(RP2; Q), which is Q. Since ƒ0 is multiplication by 1, it τ(ƒ) is never zero.

Answer to b) In this case, the non-zero rational homology is in dimensions 0 and 3, each
being isomorphic to Q. As a consequence, the argument used for RP2 can not be used
here. In fact, there is a self-map with no fixed points and it is constructed momentarily.
It is instructive to consider first the case of RP1 which is S1, where a rotation by angle π
has no fixed points. Now viewing RP1 as (R2−0)/R∗, then this rotation through angle π is
depicted using homogeneous coordinates [x1, x2] as the map [x1, x2] → [x2, -x1] which
can’t have a fixed point because there is no non-zero real number λ and (x1, x2) ∈ R2−0
with x2 = λx1 and x1 = -λx2. To mimick this for RP3, write RP3 as (R4−0)/R* and then
define the desired self map using homogeneous coordinates [x1, x2, x3, x4] by the rule
whereby [x1, x2, x3, x4] → [x2, -x1, x4, -x3]. This has no fixed points because there is no
non-zero real number λ and (x1, x2, x3, x4) ∈ R4−0 such that x2 = λx1, x1 = -λx2, x4 = λx3
and x3 = -λx4.

PROBLEM 3 SOLUTION:
2017A is the product of the prime ideals (2017, X + 1) and (2017, X4 - X3 + X2 - X + 1).
In general, if the polynomial P(X) factors modulo a prime p into distinct irreducibles {Pi}
then the ideal pZ[X]/(P(X)) is the product of ideals (p, Pi). In our case, p = 2017 and
P = X5 − 2016 ! X5 + 1 mod p. The roots of X5 + 1 in an algebraic closure of Z/pZ are
the set {-1, -w, -w2, -w3, -w4} where w is a nontrivial 5th root of unity. The irreducible
factors correspond to orbits of the permutation x → xp of those roots. Clearly −1 is a
fixed point, and since p ! 2 mod 5 the remaining roots fall in to a single orbit

-w → -w2 → -w4 → -w3 → -w.

Hence the irreducible factors of X5 +1 mod p are X + 1 and (X5 +1)/(X +1) which is the
polynomial X4 −X3 +X2 −X +1.

PROBLEM 4 SOLUTION:

Answer to a) Let X be a complete non-singular curve of genus g. Let K denote the

canonical divisor. If D is any divisor on X, let !(D) = dim(H0(X, OX(D))). The
Riemann-Roch theorem asserts that !(D) - !(K - D) = deg(D) + 1 - g.

Answer to b) Fix a point Q ≠ P and let D denote the divisor 2P - Q. Choose a positive
integer n such that n > max{2g − 2, 0}. Noting that n = deg(nD) and that deg(K) = 2g - 2,
it follows that deg(K - nD) < 0. This implies that !(K - D) = 0. Therefore, the Riemann–
Roch theorem applied to nD implies that !(nD) = n + 1 - g which is greater than 1. This
means that there is an effective divisor (to be denoted by D´) and a rational function on X
(to be denoted by ƒ) such that n D + (ƒ) = D´. Rewriting this gives (ƒ) = D´ - 2n P + n Q so
ƒ has poles only at P.

PROBLEM 5 SOLUTION:

Answer to a) The event in question is " !.

x!R
This is no smaller than R1 " |x|! which
x!R

in turn is no greater than R2 .

Answer to b) The square of the standard deviation is the square root of the expectation of
the random variable x2. Since
 " |x|! ≤ ( "! ) ( " x ! )
1/2 2 1/2
(∗)
! ! !

(which is proved momentarily), and since "! = 1, it follows that ( " x 2 ! )1/2 ≥ 2. To
! !
-1 2
prove (∗), note that for any t ∈ (0, ∞), the expectation of (t - t x) is the sum

t 2 "! - 2 " |x|! + t -2 " x 2 ! .

! ! !

This is non-negative for any t ∈ (0, 1) since it is the expectation of a positive random
variable. The assertion that it is non-negative for the case t = ( " x 2 ! )1/4 ( "! )-1/4 is (∗).
! !

Answer to c) Supposing that p, p´ ∈ R, then

ˆ
!(p) ˆ
- !(p´) = " (e ix p - e ix p´ )! . (∗∗)
!

p´

!e
ixp ixp´ ixq
Noting that e -e = ix dq by the fundamental theorem of calculus, it follows
p

that |eixp - eixp´| ≤ |x| |p - p´|. This understood, then (∗∗) leads to the bound

ˆ
|!(p) ˆ
- !(p´)| ≤ ( " |x|! ) |p - p´| = 2 |p - p´| .
!

Answer to d) The random variable SN has mean 0 and standard deviation equal to N-1/2
times the standard deviation of x, thus 4 N-1/2. (The expecation of SN2 is the that of N-
2
∑i,k=1,…,N xixk. Only the i = k terms are non-zero (because x has mean zero), there are N
of them and each is the expectation of x2 which is 16.) Denote this standard deviation of
SN by σN for the moment. The central limit theorem approximates the probability in
1
2
/ 2 " 2N
question by #
-1
1
2! " N
e-x dx where σN again denotes 4 N-1/2 .

PROBLEM 6 SOLUTION
This is a form of Jensen’s inequality. To elaborate, fix B so that |ƒ(z)| ≤ B for all z ∈ H.
For each integer N, define

N
FN(z) = ƒ(z)/hN(z) = ƒ(z) ! zz +- 11/n
/n
.
n =1

This function is analytic on H because the poles at z = 1, 2, 3,..., N are matched by zeros
of ƒ. Moreover, the absolute value of each of the factors (z + 1/n)/(z − 1/n) approaches 1
as Re(z) → 0 (uniformly in Im(z)), and also approaches 1 as |z| → ∞. Hence |Fn(z)| ≤ B
for all z ∈ H by virtue of the maximum modulus principle (the norm of an analytic
function can not take on a local maximum). With the preceding understood, note that for
N
any fixed, positive real z, the factor ! zz +- 11/n/n
n =1
becomes unbounded as N → ∞. Hence its

product with ƒ(z) cannot remain bounded unless ƒ(z) = 0 on the real axis. But a
holomorphic function on any domain has discrete zeros, so ƒ(z) must be everywhere 0.