Remez inequality and propagation of smallness
for solutions of second order elliptic PDEs
Part IV. Smallness propagation from sets
of positive measure
Eugenia Malinnikova
NTNU
March 2018
E. Malinnikova Propagation of smallness for elliptic PDEs
Not all doubling indices are large
Suppose that u is a function om a compact manifold, N2,u (q)
is the doubling index for the L2 -norm. We partition M into
cubes on approximately the same size. Then there are cubes
with small doubling index. One may estimate the number of
such cubes from the estimates on kuk∞ /kuk2 .
Now let Lf = 0 in CQ, consider the doubling index Ñf and a
partition of Q into Ad small cubes. Then if Ñ(q) > N0 for
each small cube q then Ñ(Q) > AN0 /2
Iterating this result we obtain: If Ñ(Q) > N0 and Q is divided
into B d small cubes then the number of cube where
Ñ(q) ≥ Ñ(Q)/2 is ≤ B d−γ .
E. Malinnikova Propagation of smallness for elliptic PDEs
A new result on quantitative uniqueness, non-analytic
coefficients
Theorem (E.M., A. Logunov, 2017)
Let f be a solution of Lf = 0. Assume that
|f | ≤ on E ⊂ Ω,
where |E | > 0. Let K be a compact subset of Ω then
max |f | ≤ C sup |f |1−α α ,
K Ω
where C , α depend on L, |E |, dist(E , ∂Ω) and K (but not on
E and f ).
E. Malinnikova Propagation of smallness for elliptic PDEs
Remez inequality for solutions
Let Q be a unit cube. Assume f is a solution to div(A∇f ) = 0
sup |u|
in Cd and define the doubling index N = log sup2Q |u| . Then
Q
CN
|Q|
sup |u| ≤ C sup |u| C
Q E |E |
where C depends on A only, E is any subset of Q of a positive
measure.
E. Malinnikova Propagation of smallness for elliptic PDEs
Reformulation
The Remez inequality is equivalent to the following estimate of
the sub-level set.
Lemma
Suppose that div(A∇u) = 0 in Cd Q and supQ |u| = 1. Let
N = N(u, Q) ≥ 1. Set
Ea = {x ∈ Q : |u(x)| < e −a }.
Then
|(Ea )| < Ce −βa/N |s(Q)|d ,
for some C , β > 0.
This lemma implies the propagation of smallness result.
E. Malinnikova Propagation of smallness for elliptic PDEs
Induction
We will prove the estimate
|(Ea )| < Ce −βa/N |s(Q)|d
using double induction in (a, N).
E. Malinnikova Propagation of smallness for elliptic PDEs
Induction
We will prove the estimate
|(Ea )| < Ce −βa/N |s(Q)|d
using double induction in (a, N).
• Induction base 1: a < N the inequality trivially holds for
C large enough.
E. Malinnikova Propagation of smallness for elliptic PDEs
Induction
We will prove the estimate
|(Ea )| < Ce −βa/N |s(Q)|d
using double induction in (a, N).
• Induction base 1: a < N the inequality trivially holds for
C large enough.
• Induction base 2: N < N0 and all a, we will prove it next.
E. Malinnikova Propagation of smallness for elliptic PDEs
Induction
We will prove the estimate
|(Ea )| < Ce −βa/N |s(Q)|d
using double induction in (a, N).
• Induction base 1: a < N the inequality trivially holds for
C large enough.
• Induction base 2: N < N0 and all a, we will prove it next.
• Induction step: from N/2 and all a to N and all a using
induction on a and base 1.
E. Malinnikova Propagation of smallness for elliptic PDEs
Induction base 2: estimate of the zero set
The set Ea is concentrated near the zero set. The following
estimates for the size of the zero set are used:
E. Malinnikova Propagation of smallness for elliptic PDEs
Induction base 2: estimate of the zero set
The set Ea is concentrated near the zero set. The following
estimates for the size of the zero set are used:
For any N > 0 there exist cN and CN such that for any
solution of Lf = 0 satisfying N(u, Q) ≤ N we have
H d−1 ({f = 0} ∩ Q) ≤ CN s(Q)d−1
E. Malinnikova Propagation of smallness for elliptic PDEs
Induction base 2: estimate of the zero set
The set Ea is concentrated near the zero set. The following
estimates for the size of the zero set are used:
For any N > 0 there exist cN and CN such that for any
solution of Lf = 0 satisfying N(u, Q) ≤ N we have
H d−1 ({f = 0} ∩ Q) ≤ CN s(Q)d−1
and if {f = 0} ∩ 1/2Q 6= ∅ then
H d−1 ({f = 0} ∩ Q) ≥ cN s(Q)d−1 .
E. Malinnikova Propagation of smallness for elliptic PDEs
Induction base 2
Partition Q into small cubes q with side-length Ce −a/N s(Q).
We count how many of cubes 2q intersect the zero set Zf .
Denote this number by L.
E. Malinnikova Propagation of smallness for elliptic PDEs
Induction base 2
Partition Q into small cubes q with side-length Ce −a/N s(Q).
We count how many of cubes 2q intersect the zero set Zf .
Denote this number by L.
For any q we have that sup2q |f | ≥ Ce −a supQ |f | from the
estimate on the doubling index. If f is positive in 2q then by
the Harnack inequality q does not intersect Ea .
E. Malinnikova Propagation of smallness for elliptic PDEs
Induction base 2
Partition Q into small cubes q with side-length Ce −a/N s(Q).
We count how many of cubes 2q intersect the zero set Zf .
Denote this number by L.
For any q we have that sup2q |f | ≥ Ce −a supQ |f | from the
estimate on the doubling index. If f is positive in 2q then by
the Harnack inequality q does not intersect Ea .
Then
|Ea | ≤ Ls(q)d
E. Malinnikova Propagation of smallness for elliptic PDEs
Number of cubes intersecting the zero set
It remains to estimate L,
L
X
d−1 d−1
CN s(Q) ≥H (Zf ∩Q) = H d−1 (Zf ∩qj ) ≥ LcN s(q)d−1
j=1
E. Malinnikova Propagation of smallness for elliptic PDEs
Number of cubes intersecting the zero set
It remains to estimate L,
L
X
d−1 d−1
CN s(Q) ≥H (Zf ∩Q) = H d−1 (Zf ∩qj ) ≥ LcN s(q)d−1
j=1
We have also s(q) = Ce −a/N s(Q) and then
Ls(q)d ≤ CN (cN )−1 s(q)s(Q)d−1 = bN e −a/N s(Q)d .
Combining with the previous estimate we get the statement of
induction base 2.
E. Malinnikova Propagation of smallness for elliptic PDEs
Choosing the right notation
By relieving the brain of all unnecessary work, a good notation
sets it free to concentrate on more advanced problems, and, in
effect, increases the mental power of the race.
Alfred North Whitehead, "‘An Introduction to Mathematics"’
E. Malinnikova Propagation of smallness for elliptic PDEs
Induction step
Let Q0 be the unit cube in Rd and let
m(f , a) = |{x ∈ Q0 : |u(x)| < e −a sup |f |}|,
Q0
and
M(N, a) = sup m(f , a),
∗
where the supremum is taken over all elliptic operators
div(A∇·) and functions f satisfying the following conditions in
Cd Q 0 :
(i) A(x) = [aij (x)]1≤i,j≤n is a symmetric uniformly elliptic
matrix with Lipschitz entries.
(ii) f is a solution to div(A∇f ) = 0 in Cd Q0 ,
(iii) N(f , Q0 ) ≤ N.
E. Malinnikova Propagation of smallness for elliptic PDEs
Partition lemma in action
We want to show by induction that M(N, a) ≤ Ce −βa/N .
E. Malinnikova Propagation of smallness for elliptic PDEs
Partition lemma in action
We want to show by induction that M(N, a) ≤ Ce −βa/N .
We divide Q into B d small cubes q and deduce a recursive
inequality for M(N, a).
M(N, a) ≤ B d (M(N/2, a1 )B −d ) + B d−γ (M(N, a1 )B −d )
E. Malinnikova Propagation of smallness for elliptic PDEs
Partition lemma in action
We want to show by induction that M(N, a) ≤ Ce −βa/N .
We divide Q into B d small cubes q and deduce a recursive
inequality for M(N, a).
M(N, a) ≤ B d (M(N/2, a1 )B −d ) + B d−γ (M(N, a1 )B −d )
Here a1 = a − cN log B it appears since supQ0 |f | and supq |f |
differs, but we have supq |f | ≥ B −cN supQ |f |.
E. Malinnikova Propagation of smallness for elliptic PDEs
Partition lemma in action
We want to show by induction that M(N, a) ≤ Ce −βa/N .
We divide Q into B d small cubes q and deduce a recursive
inequality for M(N, a).
M(N, a) ≤ B d (M(N/2, a1 )B −d ) + B d−γ (M(N, a1 )B −d )
Here a1 = a − cN log B it appears since supQ0 |f | and supq |f |
differs, but we have supq |f | ≥ B −cN supQ |f |.
M(N, a) ≤ M(N/2, a − cN log B) + B −γ M(N, a − cN log B)
E. Malinnikova Propagation of smallness for elliptic PDEs
Partition lemma in action
We want to show by induction that M(N, a) ≤ Ce −βa/N .
We divide Q into B d small cubes q and deduce a recursive
inequality for M(N, a).
M(N, a) ≤ B d (M(N/2, a1 )B −d ) + B d−γ (M(N, a1 )B −d )
Here a1 = a − cN log B it appears since supQ0 |f | and supq |f |
differs, but we have supq |f | ≥ B −cN supQ |f |.
M(N, a) ≤ M(N/2, a − cN log B) + B −γ M(N, a − cN log B)
≤ Ce −βa/N e −βa/N B cβ + B −γ B cβ
E. Malinnikova Propagation of smallness for elliptic PDEs
