4 Phase functions
In this section we introduce the phase functions used to count the eigenvalues.
4.1 The eigenvalue equations
Let s
j
= p
n − j − 12 and p
j
= p
m − j − 12. Conjugating the matrix ˜
A
n,m
6 with a 2n × 2n
diagonal matrix D = D
n
with diagonal elements D
1,1
= 1, D
2i,2i
= ˜
χ
βm−i−1
p β p
i i
−1
Y
ℓ=1
˜ χ
βm−ℓ
χ
βn−ℓ
β p
ℓ
s
ℓ
, D
2i+1,2i+1
=
i
Y
ℓ=1
˜ χ
βm−ℓ
χ
βn−ℓ
β p
ℓ
s
ℓ
we get the tridiagonal matrix ˜ A
D n,m
= D
−1
˜ A
n,m
D:
˜ A
D n,m
=
p
+ X p
1
s + Y
s
1
p
1
+ X
1
... ...
... p
n −1
s
n −2
+ Y
n −2
s
n −1
p
n −1
+ X
n −1
p
n
29
where X
ℓ
= ˜
χ
2 βm−ℓ−1
β p
ℓ+1
− p
ℓ
, ≤ ℓ ≤ n − 1,
Y
ℓ
= χ
2 βn−ℓ−1
βs
ℓ+1
− s
ℓ
, ≤ ℓ ≤ n − 2.
The first couple of moments of these random variables are explicitly computable using the moment generating function of the
χ
2
-distribution and we get the following asymptotics: EX
ℓ
= O m − ℓ
−32
, EX
2 ℓ
= 2β + O m − ℓ
−1
, EX
4 ℓ
= O 1, EY
ℓ
= O n − ℓ
−32
, EY
2 ℓ
= 2β + O n − ℓ
−1
, EY
4 ℓ
= O 1, 30
where the constants in the error terms only depend on β.
We consider the eigenvalue equation for ˜ A
D n,m
with a given Λ ∈ R and denote a nontrivial solution
of the first 2n − 1 components by u
1
, v
1
, u
2
, v
2
, . . . , u
n
, v
n
. Then we have s
ℓ
v
ℓ
+ p
ℓ
+ X
ℓ
v
ℓ+1
= Λu
ℓ+1
, ≤ ℓ ≤ n − 1,
p
ℓ+1
u
ℓ+1
+ s
ℓ
+ Y
ℓ
u
ℓ+2
= Λv
ℓ+1
, ≤ ℓ ≤ n − 2,
where we set v = 0 and we can assume u
1
= 1 by linearity. We set r
ℓ
= r
ℓ,Λ
= u
ℓ+1
v
ℓ
, 0 ≤ ℓ ≤ n−1
and ˆr
ℓ
= ˆr
ℓ,Λ
= v
ℓ
u
ℓ
, 1 ≤ ℓ ≤ n. These are elements of R ∪ {∞} satisfying the recursion
ˆr
ℓ+1
= −
1 r
ℓ
· s
ℓ
p
ℓ
+ Λ
p
ℓ
1 + X
ℓ
p
ℓ −1
, ≤ ℓ ≤ n − 1
31 r
ℓ+1
= −
1 ˆr
ℓ+1
· p
ℓ+1
s
ℓ
+ Λ
s
ℓ
1 + Y
ℓ
s
ℓ −1
, ≤ ℓ ≤ n − 2,
32 with initial condition r
= ∞. We can set Y
n
= 0 and define r
n
via 32 with ℓ = n − 1, then Λ is an
eigenvalue if and only if r
n
= 0. 324
4.2 The hyperbolic point of view
We use the hyperbolic geometric approach of [13] to study the evolution of r and ˆr. We will view R ∪ {∞} as the boundary of the hyperbolic plane H = {ℑz 0 : z ∈ C} in the Poincaré half-plane
model. We denote the group of linear fractional transformations preserving H by PSL2, R. The recursions for both r and ˆr evolve by elements of this group of the form x
7→ b − ax with a 0. The Poincaré half-plane model is equivalent to the Poincaré disk model U =
{|z| 1} via the con- formal bijection
Uz =
iz+1 z+i
which is also a bijection between the boundaries ∂ H = R ∪ {∞} and
∂ U = {|z| = 1, z ∈ C }. Thus elements of PSL2, R also act naturally on the unit circle ∂ U. By lifting these maps to R, the universal cover of
∂ U, each element T in PSL2, R becomes an R → R
function. The lifted versions are uniquely determined up to shifts by 2 π and will also form a group
which we denote by UPSL2, R. For any T
∈ UPSL2, R we can look at T as a function acting on
∂ H , ∂ U or R. We will denote these actions by:
∂ H → ∂ H : z 7→ z.T, ∂ U → ∂ U : z 7→ z
◦
T, ∂ R → ∂ R : z 7→ z
∗
T.
For every T
∈ UPSL2, R the function x 7→ f x = x
∗
T is monotone, analytic and quasiperiodic modulo 2
π: f x + 2π = f x + 2π. It is clear from the definitions that e
i x
◦
T = e
i f x
and
2 tanx.T = 2 tan f x.
Now we will introduce a couple of simple elements of UPSL2, R. For a given α ∈ R we will denote
by Q
α the rotation by α in U about 0. More precisely, ϕ
∗
Q α = ϕ + α. For a 0, b ∈ R we
denote by Aa, b the affine map z
→ az + b in H . This is an element of PSL2, R which fixes ∞ in H and
−1 in ∂ U. We specify its lifted version in UPSL2, R by making it fix π, this will uniquely determines it as a R
→ R function. Given
T ∈ UPSL2, R, x, y ∈ R we define the angular shift
ash T, x, y = y
∗
T
− x
∗
T
− y − x which gives the change in the signed distance of x, y under
T. This only depends on v = e
i x
, w = e
i y
and the effect of T on
∂ U, so we can also view ashT, ·, · as a function on ∂ U ×∂ U and the following
identity holds: ash
T, v, w = arg
[0,2π
w
◦
T v
◦
T − arg
[0,2π
wv. The following lemma appeared as Lemma 16 in [13], it provides a useful estimate for the angular
shift.
Lemma 12. Suppose that for a T ∈ UPSL2, R we have i + z.T = i with |z| ≤
1 3
. Then ash
T, v, w = ℜ
h ¯
w − ¯v
−z −
i2+¯ v+ ¯
w 4
z
2
i + ǫ
3
= −ℜ [ ¯
w − ¯vz] + ǫ
2
= ǫ
1
, 33
where for d = 1, 2, 3 and an absolute constant c we have |ǫ
d
| ≤ c|w − v||z|
d
≤ 2c|z|
d
. 34
If v = −1 then the previous bounds hold even in the case |z|
1 3
.
325
4.3 Regularized phase functions
Because of the scaling in 11 we will set Λ = µ
n
+ λ
4n
1 2
. We introduce the following operators
J
ℓ
= QπAs
ℓ
p
ℓ
, µ
n
s
ℓ
, M
ℓ
= A1 + X
ℓ
p
ℓ −1
, λ4n
1 2
p
ℓ
A
p
ℓ
p
ℓ+1
, 0, ˆ
J
ℓ
= QπAp
ℓ
s
ℓ
, µ
n
p
ℓ
, ˆ
M
ℓ
= A1 + Y
ℓ
s
ℓ −1
, λ4n
1 2
s
ℓ
. Then 31 and 32 can be rewritten as
r
ℓ+1
= r
ℓ
. J
ℓ
M
ℓ
ˆ J
ℓ
ˆ M
ℓ
, r
= ∞. We suppressed the
λ dependence in r and the operators M, ˆ M. Lifting these recursions from
∂ H to R we get the evolution of the corresponding phase angle which we denote by
φ
ℓ
= φ
ℓ,λ
. φ
ℓ+1
= φ
ℓ∗
J
ℓ
M
ℓ
ˆ J
ℓ
ˆ M
ℓ
, φ
= −π. 35
Solving the recursion from the other end, with end condition 0 we get the target phase function φ
⊙ ℓ,λ
: φ
⊙ ℓ
= φ
⊙ ℓ+1∗
ˆ M
−1 ℓ
ˆ J
−1 ℓ
M
−1 ℓ
J
−1 ℓ
, φ
⊙ n
= 0. 36
It is clear that φ
ℓ,λ
and φ
⊙ ℓ,λ
are independent for a fixed ℓ as functions in λ, they are monotone
and analytic in λ and we can count eigenvalues using the formula 21.
In our case both M
ℓ
and ˆ M
ℓ
will be small perturbations of the identity so J
ℓ
ˆ J
ℓ
will be the main part of the evolution. This is a rotation in the hyperbolic plane if it only has one fixed point in H. The
fixed point equation ρ
ℓ
= ρ
ℓ
. J
ℓ
ˆ J
ℓ
can be rewritten as ρ
ℓ
= p
ℓ
s
ℓ
µ
n
p
ℓ
− p
ℓ µ
n
s
ℓ
−
1 ρ
ℓ
=
ρ
ℓ
µ
2 n
− p
2 ℓ
− µ
n
s
ℓ
ρ
ℓ
µ
n
s
ℓ
− s
2 ℓ
. This can be solved explicitly, and one gets the following unique solution in the upper half plane if
ℓ n + 12:
ρ
ℓ
= µ
2 n
− m + n 2
µ
n
s
ℓ
+ i s
1 −
µ
2 n
− m + n
2
4 µ
2 n
s
2 ℓ
. 37
One also needs to use the identity p
2 ℓ
− s
2 ℓ
= m − n and 12. This shows that if ℓ n then
J
ℓ
ˆ J
ℓ
is a rotation in the hyperbolic plane. We can move the center of rotation to 0 in U by conjugating it
with an appropriate affine transformation:
J
ℓ
ˆ J
ℓ
= Q−2 argρ
ℓ
ˆ ρ
ℓ T
−1 ℓ
. Here
T
ℓ
= Aℑρ
ℓ −1
, −ℜρ
ℓ
, X
Y
= Y
−1
XY and
ˆ ρ
ℓ
= µ
2 n
+ m − n 2
µ
n
p
ℓ
+ i s
1 −
µ
2 n
+ m − n
2
4 µ
2 n
p
2 ℓ
. 38
326
In order to regularize the evolution of the phase function we introduce ϕ
ℓ,λ
:= φ
ℓ,λ∗
T
ℓ
Q
ℓ−1
, ≤ ℓ n
where Q
ℓ
= Q2 argρ ˆ
ρ · · · Q2 argρ
ℓ
ˆ ρ
ℓ
and Q
−1
is the identity. It is easy to check that the initial condition remains
ϕ
0, λ
= π. Then ϕ
ℓ+1
= ϕ
ℓ∗
Q
−1 ℓ−1
T
−1 ℓ
J
ℓ
M
ℓ
ˆ J
ℓ
ˆ M
ℓ
T
ℓ+1
Q
ℓ
= ϕ
ℓ∗
Q
−1 ℓ−1
T
−1 ℓ
Q −2 argρ
ℓ T
−1 ℓ
M
ˆ J
ℓ
ℓ
ˆ M
ℓ
T
ℓ
T
−1 ℓ
T
ℓ+1
Q
ℓ
= ϕ
ℓ∗
M
ˆ J
ℓ
ℓ T
ℓ
Q
ℓ
ˆ
M
ℓ T
ℓ
Q
ℓ
T
−1 ℓ
T
ℓ+1
Q
ℓ
Note that the evolution operator is now infinitesimal: M
ℓ
, ˆ M
ℓ
and T
−1 ℓ
T
ℓ+1
will all be asymptotically small, and the various conjugations will not change this.
We can also introduce the corresponding target phase function ϕ
⊙ ℓ,λ
:= φ
⊙ ℓ,λ∗
T
ℓ
Q
ℓ−1
, ≤ ℓ n
. 39
The new, regularized phase functions ϕ
ℓ,λ
and ϕ
⊙ ℓ,λ
have the same properties as φ, φ
⊙
, i.e.: they are independent for a fixed
ℓ as functions in λ, they are monotone and analytic in λ and we can count eigenvalues using the formula 24. See 22 and the discussion before it.
We will further simplify the evolution using the following identities: −
a r
+ b = b
2
+ 1 a
r − b
Q
arg b
− i b + i
, r.ˆ
J
ℓ
T
ℓ
= − 1
r p
ℓ
s
ℓ
ℑρ
ℓ
+ µ
n
s
ℓ
ℑρ
ℓ
− ℜρ
ℓ
ℑρ
ℓ
. From this we get
ˆ J
ℓ
T
ℓ
= ˆ T
ℓ
Q
ℓ
−2 arg ˆ ρ
ℓ
where r.ˆ
T
ℓ
=
s
ℓ
ℑρ
ℓ
p
ℓ
− 2 ℜρ
ℓ
ℑρ
ℓ
µ
n
+ µ
2 n
ℑρ
ℓ
p
ℓ
s
ℓ
r
− µ
n
s
ℓ
ℑρ
ℓ
+ ℜρ
ℓ
ℑρ
ℓ
= 1
ℑ ˆ ρ
ℓ
r −
ℜ ˆ ρ
ℓ
ℑ ˆ ρ
ℓ
. This allows us to write
M
ˆ J
ℓ
ℓ T
ℓ
Q
ℓ
= M
ˆ T
ℓ
ℓ Q
−2 arg ˆ ρ
ℓ
Q
ℓ
= M
ˆ T
ℓ
ℓ ˆ
Q
ℓ
. 40
where ˆ
Q
ℓ
= Q
ℓ
Q
−2 arg ˆ ρ
ℓ
. Thus
ϕ
ℓ+1
= ϕ
ℓ∗
M
ˆ T
ℓ
ℓ ˆ
Q
ℓ
ˆ M
T
ℓ
ℓ Q
ℓ
T
−1 ℓ
T
ℓ+1
Q
ℓ
.
327
We will introduce the following operators to break up the evolution into smaller pieces:
L
ℓ,λ
= A1, λ4n
1 2
p
ℓ
, ˆ
L
ℓ,λ
= A1, λ4n
1 2
s
ℓ
, S
ℓ,λ
= L
ˆ T
ℓ
ℓ,λ
ˆ T
−1 ℓ
A
p
ℓ
p
ℓ+1
1 + X
ℓ
p
ℓ −1
, 0 ˆ T
ℓ
, 41
ˆ S
ℓ,λ
= ˆ L
T
ℓ
ℓ,λ
T
ℓ −1
A1 + Y
ℓ
s
ℓ −1
, 0 T
ℓ+1
.
Then ϕ
ℓ+1
= ϕ
ℓ∗
L
ˆ T
ℓ
ℓ ˆ
Q
ℓ
S
ℓ,0
ˆ Q
ℓ
ˆ L
T
ℓ
ℓ Q
ℓ
ˆ S
ℓ,0
Q
ℓ
= ϕ
ℓ∗
S
ℓ,λ
ˆ Q
ℓ
ˆ S
ℓ,λ
Q
ℓ
. 42
We also introduce the relative regularized phase function and target phase function: α
ℓ,λ
:= ϕ
ℓ,λ
− ϕ
ℓ,0
, α
⊙ ℓ,λ
:= ϕ
⊙ ℓ,λ
− ϕ
⊙ ℓ,0
. 43
5 SDE limit for the phase function
Let F
ℓ
denote the σ-field generated by ϕ
j, λ
, j ≤ ℓ − 1. Then ϕ
ℓ,λ
is a Markov chain in ℓ with
respect to F
ℓ
. Indeed, the relation 42 shows that ϕ
ℓ+1,λ
= h
ℓ,λ
ϕ
ℓ+1,λ
, X
ℓ
, Y
ℓ
where h
ℓ,λ
is a deterministic function depending on
ℓ and λ. Since X
ℓ
, Y
ℓ
are independent of F
ℓ
it follows that E
ϕ
ℓ+1,λ
|F
ℓ
= E
ϕ
ℓ+1,λ
|ϕ
ℓ,λ
. We will show that this Markov chain converges to a diffusion
limit after proper normalization. In order to do this we will estimate E
ϕ
ℓ+1,λ
− ϕ
ℓ,λ
|F
ℓ
and
E
ϕ
ℓ+1,λ
− ϕ
ℓ,λ
ϕ
ℓ+1,λ
′
− ϕ
ℓ,λ
′
|F
ℓ
using the angular shift lemma, Lemma 12.
To simplify the computations we introduce ‘intermediate’ values for the process ϕ
ℓ,λ
by breaking the evolution operator in 42 into two parts:
ϕ
ℓ+
1 2
, λ
= ϕ
ℓ∗
S
ℓ,λ
ˆ Q
ℓ
, F
ℓ+
1 2
= σF
ℓ
∪ {ϕ
ℓ+
1 2
, λ
}. Note that
ϕ
ℓ,λ
is still a Markov chain if we consider it as a process on the half integers.
Remark 13. We would like to note that the ‘half-step’ evolution rules ϕ
ℓ,λ
→ ϕ
ℓ+
1 2
, λ
, ϕ
ℓ+
1 2
, λ
→ ϕ
ℓ+1,λ
are very similar to the one-step evolution of the phase function ϕ in [13]. Indeed, in [13],
the evolution of ϕ is of the type ϕ
ℓ+1
= ϕ
ℓ∗
˜ S
ℓ,λ
˜ Q
ℓ
where ˜ S is an affine transformation and ˜
Q is a
rotation similar to our S, ˆ
S and Q, ˆ Q. In our case the evolution of
ϕ
ℓ+1,λ
is the composition of two transformations with similar structure. The main difficulties in our computations are caused by the
fact that
Q and ˆ Q are rather different which makes the oscillating terms more complicated.
5.1 Single step estimates