Exponentials rarely maximize Fourier adjoint restriction estimates on cones
by Giuseppe Negro
Instituto Superior Técnico, Lisbon, Portugal
Modern Trends in Harmonic Analysis, ICTS Bengaluru 26 June 2023
The adjoint restriction estimates on 2–Cone and 1–Cone
ξ1
τ
ξd
τ2“|ξ|2
ξ1
τ
ξd
τ “|ξ|
µ“δpτ2´|ξ|2q µ“δpτ2´|ξ|2q1τą0
Restriction Conjecture: for all d ě2,
∥fxµ∥LqpRd`1q
∥f∥Lppµq
ďCp,d ă 8, for
$
&
%
q “ d`1d´1p1,
1
q ă d´12d .
We are interested inmaximizers of this ratio(provided the estimate holds).
The Strichartz case: p “ 2, q “ 2
d`1d´1Is f‹pτq “expp´|τ|q a maximizer?
Spatial dim. d 2–Cone 1–Cone
2 NO YES
3 YES YES
4, 6, 8, . . . NO Local 5, 7, 9, . . . Local Local Meaning that:
∥fxµ∥LqpRd`1q
∥f∥L2pµq
ď
∥fx‹µ∥LqpRd`1q
∥f‹∥L2pµq
for
#
all f PL2pµq (YES)
∥f ´f‹∥L2pµqďCd (Local) f‹ is not a critical point (see next slide). (NO) Due to:
YES: Foschi 2007 NO: G.N. 2018 Local: G.N. & F.Gonçalves 2019
The general p case, on the 1–Cone
Recall: we want to maximize ∥fxµ∥∥fLqpR∥ d`1q
Lppµq aroundf‹pτq “expp´|τ|q.
Def.: f‹ is a critical point iff B
Bϵ
∥pf‹{`ϵfqµ∥LqpRd`1q
∥f‹`ϵf∥Lppµq
ˇ ˇ ˇ ˇ ˇϵ“0
“0, @f PLppµq.
Theorem. (G.N.& D.Oliveira e Silva & B.Stovall & J.Tautges 2023)
(i) Maximizers exist for all p (for which the estimate holds).
(ii)f‹ is a critical point ðñ p “2.
τ
We say thatf‹ rarely maximizesthe Fourier extension inequality.
Exactly the same happens on the paraboloid with µ“δpτ ´|ξ|2q (i) Stovall 2020, (ii) Christ–Quilodrán 2012.
Digression: Foschi 2007. First appearance of f
‹“ e
´|τ|.
Let µ“δpτ2´|ξ|2q1τą0,p“2,d“3. Writeζ “ pτ, ξq.
τ
∥xfµpt,xq∥4L4pR4q “ ż
fpζ1qfpζ2qfpζ3qfpζ4qeipt,xq¨pζ1`ζ2´ζ3´ζ4qdµb4dtdx
“ ż
fpζ1qfpζ2qfpζ3qfpζ4qδpζ1`ζ2´ζ3´ζ4qdµb4
(Cauchy–Schwarz)ď ż
|fpζ1qfpζ2q|2δpζ1`ζ2´ζ3´ζ4qdµb4
“ ż
|fpζ1qfpζ2q|2µ˚µpζ1`ζ2qdµb2
pMiracleq “ p2πq12∥f∥4L2pµq.
Miracle: µ˚µ“ p2πq12 on its support.
Digression: Foschi 2007. First appearance of f
‹“ e
´|τ|.
Let µ“δpτ2´|ξ|2q1τą0,p“2,d“3. Writeζ “ pτ, ξq.
τ
∥fxµpt,xq∥4L4pR4q“ ż
fpζ1qfpζ2qfpζ3qfpζ4qδpζ1`ζ2´ζ3´ζ4qdµb4
(C–S)ď ż
|fpζ1qfpζ2q|2δpζ1`ζ2´ζ3´ζ4qdµb4.
Why f‹? Because (C–S) is an identity when f “f‹. Proof:
e´τ1´τ2´τ3´τ4δpζ1`ζ2´ζ3´ζ4q “e´2τ1´2τ2δpζ1`ζ2´ζ3´ζ4q.
There must be a less technical reason for the appearance of f‹!
The Penrose map R
1`dÑ r´ π, π s ˆ S
dr “|x| t
R1`d
θ T
´π
π π
r´π, πs ˆSd
Equations of Penrose map (and definition of θ)
T “arctanpt`rq `arctanpt´rq, θ“arctanpt`rq ´arctanpt´rq.
θ
Sd
The Penrose map R
1`dÑ r´ π, π s ˆ S
dWe call D1`d the range of the Penrose map - the Penrose diamond.
´π
π π
θ T
θ
Figure: The Penrose diamond (left) is a submanifold of the cylinderRˆSd.
(Right: Taken from S. Carroll,Notes on general relativity, arXiv:gr-qc/9712019.)
Where does f
‹p τ q “ exp p´ |τ | q come from?
For exposition purposes, let us focus on the 2–Cone only.
upt,xq “fxµ solves B2tu “∆u. Recall: f is defined on τ
The Penrose map...
...yields a functional transform pU0,U90q ÐÑf such that ż
Rd`1
|fxµ|q“ ż
Rd`1
|u|q“ ż
Dd`1
|U|q¨ psome weightq
d`1 2pp´1qp2´pq
, (*)
where
!
B2TU “∆SdU´pd´1q
2
4 U, pU,BTUq|T“0 “ pU0,U90q )
. Constant initial data p1,0q correspond to f‹.
p “2: RHS of (*) appears to be rotationally invariant: we can expect a maximizer to be rotational invariant, i.e. constant.
p ‰2: the weight breaks the rotational symmetry.
Back to the case p “ 2, q “ 2
d`1d´1Recall: the estimate reads ∥xfµ∥LqpRd`1qďCd∥f∥L2pµq. Is f‹pτq “expp´|τ|q a maximizer?
Spatial dim. d 2–Cone 1–Cone
2 NO YES
3 YES YES
4, 6, 8, . . . NO Local 5, 7, 9, . . . Local Local Let d “2,3. We need to maximize
ż
Dd`1
|U|q,whereU solves
#
B2TU “∆S2U´14U, d “2,
B2TU “∆S3U´U, d “3, onRˆSd. Theorem
f‹ is a critical point for d “3, but not for d “2.
Proof. The nice case d “ 3, p “ 2: f
‹is critical
Let X denote the generic point on S3. Key Symmetry Lemma
Suppose that
BT2U “∆S3U ´U, onRˆS3. Then UpT `π,´Xq “ ´UpT,Xq.
Proof. We haveU “cospTa
1´∆S3qU0`sinpT
?1´∆
S3q
?1´∆
S3
U90. Expand
U0 “
8
ÿ
ℓ“0
Yℓ, with ´∆S3Yℓ “ℓpℓ`2qYℓ; so Yℓp´Xq “ p´1qℓYℓ.
Then UpT,Xq “ ÿ8
ℓ“0
cospTpℓ`1qqYℓ`. . . l
We used: cosppT`πqpℓ`1qq “ p´1qℓ`1cospTpℓ`1qq.
Proof. The nice case d “ 3, p “ 2: f
‹is critical (cont.)
Key Symmetry Lemma
BT2U “∆S3U ´U onRˆS3.Then UpT `π, π´θq “ ´UpT, θq.
Corollary.
ż
D1`3
|U|q“ 1 2
żπ
´π
ż
S3
|U|q.
Proof. D1`3 “AYB. By the lemma, ż
A
|U|q “ ż
A˜
|U|q and ż
B
|U|q “ ż
B˜
|U|q.l
Remark
For d “2 the symmetry lemma fails. For example, B2TU‹ “∆S2U‹´ 14U‹, pU0,U90q “ p1,0q yields U‹ “cos12T, so |U‹pT `πq|‰|U‹pTq|.
´π
π π
θ T
A B
A˜
B˜
θ
Proof. The nice case d “ 3, p “ 2: f
‹is critical (end)
Recall: q “2d`1d´1 “4. We want: maxt∥xfµ∥L4 : ∥f∥L2pµq “1u.
Claim I “
ż
R3`1
pf‹{`ϵfKqµ4´ ż
R1`3
fx‹µ4 “opϵq for xf‹|fKyL2pµq“0.
Proof. By Penrose, fK ” pU0K,U90Kq andf‹ ” p1,0q.
Note B2TU‹ “∆S3U‹´U‹,pU0,U90q “ p1,0q yieldsU‹“cosT.
2I“ żπ
´π
ż
S3
|U‹`ϵUK|4´ żπ
´π
ż
S3
|U‹|4
“4ϵ żπ
´π
ż
S3
U‹3UK`opϵq “4ϵ żπ
´π
pcosTq3 ˆż
S3
UKpT,¨q
˙
`opϵq.
Now xf‹|fKy “0ñ xUK|1y “0 at T “0. Therefore xUK|1y “0 at all times T. Henceş
S3UKpT,¨q ”0 and we concludeI “0.l
Final summary
We want to maximize ż
R1`d
|fxµ|q, which equals:
$
’’
’’
’’
’’
’’
&
’’
’’
’’
’’
’’
% ż
Dd`1
|U|qpcosT `cosθq
d`1 2pp´1qp2´pq
, p ‰2, ż
Dd`1
|U|q, p “2,d even,
ż1
´1
ż
Sd
|U|q, p “2,d odd.
Only in the latter case,
f‹ “ expp´|τ|q is critical (and so has a chance of being a maximizer).
´π
π π
θ T
Dd`1
θ