In Memory of Professor Michel Marias (1953-2020) : Bstract

BLOW-UPS OF CALORIC MEASURE IN TIME VARYING DOMAINS AND
APPLICATIONS TO TWO-PHASE PROBLEMS

arXiv:2008.06968v3 [math.AP] 13 Apr 2021
MIHALIS MOURGOGLOU AND CARMELO PULIATTI
In memory of Professor Michel Marias (1953–2020)
A BSTRACT. We develop a method to study the structure of the common part of the bound-
aries of disjoint and possibly non-complementary time-varying domains in Rn+1 , n ≥ 2,
at the points of mutual absolute continuity of their respective caloric measures. Our set of
techniques, which is based on parabolic tangent measures, allows us to tackle the following
problems:
(1) Let Ω1 and Ω2 be disjoint domains in Rn+1 , n ≥ 2, which are quasi-regular for the
heat equation and regular for the adjoint heat equation, and their complements satisfy
a mild non-degeneracy hypothesis on the set E ⊂ ∂Ω1 ∩ ∂Ω2 of mutual absolute
continuity of the associated caloric measures ωi with poles at p̄i = (pi , ti ) ∈ Ωi ,
i = 1, 2. Then, we obtain a parabolic analogue of the results of Kenig, Preiss, and
Toro, i.e., we show that the parabolic Hausdorff dimension of ω1 |E is n + 1 and
the tangent measures of ω1 at ω1 -a.e. point of E are equal to a constant multiple of
the parabolic (n + 1)-Hausdorff measure restricted to hyperplanes containing a line
parallel to the time-axis.
dω2 |E
(2) If, additionally, ω1 |E and ω2 |E are doubling, log dω 1 |E
∈ VMO(ω1 |E ), and E is
relatively open in the support of ω1 , then their tangent measures at every point of
E are caloric measures associated with adjoint caloric polynomials. As a corollary
we obtain that if Ω1 is a δ-Reifenberg flat domain for δ small enough and Ω2 =
dω2
Rn+1 \ Ω1 , and log dω 1
∈ VMO(ω1 ), then Ω1 ∩ {t < t2 } is vanishing Reifenberg
flat. This generalizes results of Kenig and Toro for the Laplacian.
(3) Finally, we establish a parabolic version of a theorem of Tsirelson about triple-points
for harmonic measure. Assuming that Ωi , 1 ≤ i ≤ 3, are quasi-regular domains for
both the heat and the adjoint heat equations, the set of points on ∩3i=1 ∂Ωi , where the
three caloric measures are mutually absolutely continuous has null caloric measure.
In the course of proving our main theorems, we obtain new results on heat potential theory,
parabolic geometric measure theory, nodal sets of caloric functions, and Optimal Transport,
that may be of independent interest.
C ONTENTS
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2
Discussion of the proofs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 11
2010 Mathematics Subject Classification. 35K05, 35B44, 28A75, 35B05, 31C45, 28A78, 28A33, 49Q22.
Key words and phrases. Heat equation, caloric measure, Green function, time-varying domains, free bound-
ary problems, capacity density condition, tangent measures, absolute continuity, triple points, heat potential
theory, parabolic Reifenberg flat domains, optimal transport, Kantorovich-Rubinshtein norm.
M.M. was supported by IKERBASQUE and partially supported by the grant MTM-2017-82160-C2-2-P of
the Ministerio de Economı́a y Competitividad (Spain), and by IT-1247-19 (Basque Government). C.P. was
supported by IT-1247-19 (Basque Government).
1
2 MIHALIS MOURGOGLOU AND CARMELO PULIATTI
Acknowledgements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
2. Preliminaries and notation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 12
3. Heat Potential Theory and PDE estimates . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
4. Hausdorff and tangent measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 30
5. Nodal set of caloric functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 35
6. Caloric measure associated with a caloric function . . . . . . . . . . . . . . . . . . . . . . . . . . 36
7. Caloric polynomial measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 39
8. Elements of the theory of Optimal Transport . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
9. Blow-ups in time varying domains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
10. Proofs of the main theorems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
10.1. Proofs of Theorems I and II . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
10.2. Proofs of Theorems III and IV . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65
References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 67
1. I NTRODUCTION
In this paper, we study two-phase problems for harmonic measure associated with the
heat equation, which is traditionally called caloric measure. We consider two disjoint open
sets Ω+ and Ω− in Rn+1 , and ω + and ω − the associated caloric measures, so that their
common boundary ∂Ω+ ∩ ∂Ω− has positive ω + measure. Our main goal is to study how
mutual absolute continuity of the respective caloric measures provides information about
the infinitesimal structure of ∂Ω+ ∩ ∂Ω− .
In the elliptic case, Bishop, Carleson, Garnett, and Jones studied in [BCGJ89] a two-
phase problem for simply connected planar disjoint domains: they showed that the two
harmonic measures are mutually singular if and only if the intersection of the respective
sets of inner tangent points has zero one-dimensional Hausdorff measure. More recently, in
[KPT09], Kenig, Preiss, and Toro studied the higher-dimensional case, assuming that Ω+
and Ω− = Rn+1 \ Ω+ are both NTA-domains (as defined by Jerison and Kenig in [JK82]).
Combining the blow-up analysis at points of mutual absolute continuity in [KT06] with
the theory of tangent measures along with the monotonicity formula of Alt, Caffarelli and
Friedman (see [ACF84]), they showed that mutual absolute continuity of the interior and
exterior harmonic measure ω + and ω − implies that the harmonic measure has Hausdorff
dimension n and that as we zoom in at ω + -a.e. point of the common boundary, ∂Ω±
looks flatter and flatter. This blow-up technique was further improved by Azzam, the first
named author, and Tolsa [AMT17] and the same authors along with Volberg in [AMTV19],
who eventually showed that, without any further assumption on the domains, the harmonic
measure restricted to the set of mutual absolute continuity is n-rectifiable. The connection
between the Riesz transform and n-rectifiability was an element of major importance in the
proofs of [AMT17] and [AMTV19] that allowed to improve on the results of [KPT09] from
this perspective as well (since n-rectifiable measures have Hausdorff dimension n).
All the aforementioned results describe a.e. phenomena, while it is interesting to inves-
tigate conditions that ensure some nice limiting behavior of our blow-ups at every point.
In [KT06], Kenig and Toro considered Ω+ and Ω− to be complementary 2-sided NTA do-
−
mains and log dω +
dω + ∈ V M O(dω ), and they showed that, in a sequence of arbitrarily small
BLOW-UPS OF CALORIC MEASURE 3
scales, the boundary around any point starts resembling the zero set of a harmonic polyno-
mial (instead of a hyperplane). If the domains are δ-Reifenberg flat, for δ small enough,
then the conclusion improves to a hyperplane implying that the domains are Reifenberg
flat with vanishing constant. They also proved that the same conclusion holds without the
δ-Reifenberg flatness assumption if we assume that Ω+ is two-sided NTA, ∂Ω+ is Ahlfors-
David regular and the logarithms of the interior and exterior Poisson kernels are in VMO
with respect to the surface measure on ∂Ω+ . Badger went a step further in [Bad11], show-
ing that those harmonic polynomials are always homogeneous, while later, in [Bad13], he
explored the topological properties of sets where the boundary is approximated by zero
sets of harmonic polynomials in this way. Badger, Engelstein and Toro investigated in
[BET17] two-phase problems for the harmonic measure below the continuous threshold
using [KPT09], [Bad11], [Bad13], and an interesting structure theorem for sets that are lo-
cally bilaterally well approximated by zero sets of harmonic polynomials. More recently,
the geometry of the singular set of the boundary ∂Ω± of complementary NTA domains such
their caloric measures ω ± are mutually absolutely continuous and log(dω − /dω + ) is Hölder
continuous has been studied in the work of McCurdy [McC19]. In [AM19], Azzam and the
first named author, among others, extended the results in [KPT09], [KT06], and [Bad11]
to general domains and elliptic measures associated with uniformly elliptic operators in
divergence form with merely bounded coefficients that are also in the Sarason’s space of
vanishing mean oscillation with respect to ω + . For relevant results for elliptic operators
with W 1,1 -coefficients see [TZ17].
It is interesting to understand if similar phenomena occur also in the context of parabolic
PDEs and, in particular, the heat equation. The implementation of blow-up methods in
the context of one-phase free boundary problems for the heat equation was initiated by
Hofmann, Lewis, and Nyström in [HLN04]. They extended the work of Kenig and Toro
showing that, in parabolic chord arc domains with vanishing constant, the logarithm of the
density of caloric measure with respect to a certain projective Lebesgue measure (or else
the Poisson kernel) is of vanishing mean oscillation, and also obtained a partial converse,
which amounts to a one-phase problem in Reifenberg flat domains with parabolic uniformly
rectifiable boundaries. The full converse was proved by Engelstein in [Eng17A], where a
key fact of his proof was a delicate classification of “flat” blow-ups, which was an open
problem in the parabolic setting. He also examined free boundary problems with conditions
above the continuous threshold. This problem had already been solved by Nyström (see
[Nys06b] and [Nys12]) in graph domains under the assumption that the Green function is
comparable with the distance function from the boundary.
Results analogous to those in [KT06] for the heat equation were considered by Nyström
in [Nys06a]. He proved that, if Ω and Rn+1 \ Ω are parabolic NTA domains with parabolic
Ahlfors regular boundary and the logarithms of the associated Poisson kernels are of van-
ishing mean oscillation, then a portion of ∂Ω suitably away from the poles of the caloric
measures is Reifenberg flat with vanishing constant. Furthermore, it was shown in [Nys06c]
that if one drops the Ahlfors regularity hypothesis and, instead of parabolic NTA, they ask
for the domains to be δ-Reifenberg flat for δ > 0 small enough, then the same conclu-
sion holds if one substitutes the assumption on the Poisson kernels with a vanishing mean
oscillation hypothesis on the logarithm of the density dω − /dω + of the caloric measures as-
sociated with the two domains. We remark that the proof uses the construction of the Green
function and caloric measure with pole at infinity. Finally, we refer to [Eng17B] for an
interesting geometric result for planar NTA domains with an application to the previously
discussed one-phase problem.
To the best of our knowledge, the present paper is the first one that studies free boundary
problems in so general domains.
A proficient way to address the kind of questions we referred above is to analyze the
fine structure of the boundary by “zooming-in” on boundary points. There are two ways
to do that. The first one is by looking at the Hausdorff convergence of rescaled copies of
the support of a measure as we zoom in, for which we follow the framework of Badger and
Lewis [BL15].
We define the ball with respect to the parabolic norm kȳk := max{|y|, |s|1/2 } centered
at x̄ ∈ Rn+1 with radius r > 0 as
Cr (x̄) = {ȳ ∈ Rn+1 : kȳ − x̄k < r}.
Note that Cr (x̄) is a euclidean right circular cylinder centered at x̄ of height 2r 2 and radius
r. If x̄ = 0̄, we simply write Cr (x̄) = Cr unless stated otherwise. We also need the
following time-backwards and time-forwards versions of the parabolic ball:
Cr− (x, t) = {(y, s) ∈ Rn+1 : y ∈ Br (x), t − r 2 < s < t}
Cr+ (x, t) = {(y, s) ∈ Rn+1 : y ∈ Br (x), t < s < t + r 2 }.
A point x̄ ∈ Rn+1 is denoted by x̄ = (x, t) = (x′ , xn , t) := (x1 , . . . , xn , t) and we also
use the notation 0̄ = (0, . . . , 0). For r > 0 and x̄, ȳ ∈ Rn+1 , we set
δr (x̄) := (rx, r 2 t), and Tȳ,r (x̄) := δ1/r (x̄ − ȳ).
Let A ⊂ Rn+1 be a set and let S be a collection of subsets of Rn+1 . Given x̄ ∈ A and
r > 0 we define

S dist(ā, x̄ + S) dist(z̄, A)
ΘA (x̄, r) = inf max sup , sup ,
S∈S ā∈A∩Cr (x̄) r z̄∈(x̄+S)∩Cr (x̄) r
where “dist” denotes the distance with respect to the parabolic norm k · k. Let S be a local
appoximation class, i.e. let us assume that for all S ∈ S and λ > 0, we have that δλ S ∈ S .
We say that A is locally bilaterally well approximated by S , also abbreviated LBW A(S ),
if for all ε > 0 and all K ⊂ A compact, there is rε,K > 0 such that ΘS A (x̄, r) < ε fro all
x̄ ∈ K and 0 < r < rε,K . We say that x̄ ∈ A is an S -point of A if limr→0 ΘS A (x̄, r) = 0.
In other words, for x̄ to be a S point means that, as we zoom in on A at the point x̄, the set
A resembles more and more an element of S . Remark that this element may change as we
move from one scale to the next.
The second way is by investigating the weak convergence of rescaled copies of the mea-
sure itself. In [Pr87], Preiss developed the theory of tangent measures, which turned out to
be vital in the study of two-phase problems for harmonic and elliptic measure; his results
are at the core of [KPT09], [Bad11], [AMT17], [AMTV19], and [AM19]. The definition of
a tangent measure can be readily adapted to the parabolic context, in which Rn+1 is natu-
rally endowed with a set of non-isotropic dilations. Furthermore, we remark that although
the theory of local set approximation was originally developed by Badger and Lewis in the
euclidean setting, it can be generalized to Rn+1 endowed with the parabolic metric (see
[BL15, Remark 1.8] and the references therein).
We say that ν is a tangent measure of µ at a point x̄ ∈ Rn+1 and write ν ∈ Tan(µ, x̄) if
ν is a non-zero Radon measure on Rn+1 and there are sequences ci > 0 and ri ց 0 so that
ci Tx̄,ri [µ] converges weakly to ν as i → ∞.
Let us record here that a parabolic version of Besicovitch covering theorem is available
for parabolic balls (see [It18]), which allows for the basic properties of tangent measures to
hold in the parabolic setting by the same proofs as in the Euclidean case. For further details,
we refer to Section 4.
The geometry of the blow-ups at a point of mutual absolute continuity of caloric measures
is intimately related with that of nodal sets of caloric functions. We define the heat and
adjoint heat equations by
(1.1) Hu := ∆u − ∂t u = 0, and H ∗ u := ∆u + ∂t u = 0,
and a C 2,1 function satisfying Hu = 0 (resp. H ∗ u = 0) is called (resp. adjoint) caloric
function.
As in the harmonic case, polynomials play a special role in the theory of caloric func-
tions. Indeed, as harmonic functions of polynomial growth are polynomials, it has been
long known that the so-called ancient solutions of the heat equation of polynomial growth
are polynomials. For a more in-depth discussion, we refer to [LZ19] and the references
therein. Let Θ denote the set of caloric functions vanishing at 0̄, P (d) the set of the caloric
polynomials of degree d vanishing at 0̄, and F (d) the set of homogeneous caloric polyno-
mials of degree d. In Lemma 6.1, we show that, for any caloric function h on Rn+1 , there
exists a unique Radon measure ωh such that
Z Z
1
ϕ dωh = |h|(∆ + ∂t )ϕ, for all ϕ ∈ Cc∞ (Rn+1 ).
2
The measure ωh is called caloric measure ωh associated to h. In fact, this is the unique
adjoint caloric measure with pole at infinity in Ω± = {h± 6= 0} and h± is the Green
function with pole at infinity. If h is an adjoint caloric function in Rn+1 , we define the
adjoint caloric measure analogously.
We say that a function f (x, t) is (parabolic) homogeneous of degree m ∈ R if for any
λ > 0 and (x, t) ∈ Rn+1 \ {0̄}, we have

f δλ (x, t) = f (λx, λ2 t) = λm f (x, t).
We remark that every caloric polynomial can bePwritten as the sum of homogeneous caloric
polynomials. Indeed, assuming that h(x, t) = α,ℓ cα,ℓ xα tℓ for cα,ℓ ∈ R and that cα,ℓ = 0
if |α| + 2ℓ > d, we can write
d X
X d
X
h(x, t) = cα,ℓ xα tℓ =: hj (X, t).
j=0 |α|+2ℓ=j j=0
Observe that hj is a homogeneous polynomial of degree j, which implies that Hhj is a

homogeneous polynomial of degree j − 2. So
X
Hh = Hhj = 0
j
if and only if Hhj = 0 for every j = 0, . . . , d. Hence, it is consistent to say that a caloric
function h is a caloric polynomial of degree d if it can be written as the sum of homogeneous
caloric polynomials of degree at most d. Thus, we define

HΘ := {ωh : h ∈ Θ}, P(d) := {ωh : h ∈ P (d)} and F (d) := {ωh : h ∈ F (d)}.
Given a caloric function h, we use the notation Σh := {h = 0}. Furthermore, we indicate
HΣ = {Σh : h ∈ Θ},
PΣ (d) = {Σh : h ∈ P (1) ∪ · · · ∪ P (d)} and FΣ (d) = {Σh : h ∈ F (d)}.
The families Θ∗ , P ∗ (d), F ∗ (d),T ∗ , P ∗ (d), F ∗ (d), TΣ∗ , PΣ∗ (d), and FΣ∗ (d) are defined
analogously for adjoint caloric functions, polynomials, and homogeneous polynomials. Set,
moreover,
FΣ := {V ⊂ Rn+1 : V is an n-plane through 0̄ containing a line parallel to the t-axis}
and
F := {Hpn+1 |V : V ∈ FΣ },
and observe that F = F (1) = F ∗ (1) and FΣ = FΣ (1) = FΣ∗ (1).
Potential theory is also very important in problems relating the metric properties of
caloric measure to the geometry of the boundary but, in order to take full advantage of
it, we need to ensure that the boundary is regular enough so that the Perron-Wiener-Brelot
with continuous boundary data is continuous all the way to the boundary in that domain
(sometimes up to a polar set of the boundary). One can find several geometric conditions in
the literature that imply the desired regularity of the boundary; e.g., the parabolic Reifenberg
flatness (see the definition before Corollary 1.3), the non-tangential accessibility (see e.g.
[Nys06a, pp. 263-264]), and the time-backwards Ahlfors-David regularity (see [GH20]).
A well-known condition which implies regularity of the boundary in the Euclidean setting
and has been used in the harmonic measure framework is the so-called Capacity-Density
Condition (see e.g. [AMT17]). The CDC ensures that the capacity of the complement of Ω
in each ball centered at the boundary is large in a scale invariant way. We will now introduce
two similar notions of thickness that, in the parabolic context, serve the same purpose as the
CDC.
Given ρ > 0 and x̄ = (x, t) ∈ Rn+1 we define the heat ball centered at x̄ with radius ρ
to be the set
n r ρ o
n+1
E(x̄; ρ) := (y, s) ∈ R : |x − y| < 2n(t − s) log , t−ρ<s<t
(1.2) t−s
n+1 −n/2

= ȳ ∈ R : Γ(x̄ − ȳ) > (4πρ) ,
where Γ(·, ·) stands for the fundamental solution of the heat equation (see Section 3). We
indicate by E ∗ (x̄; r) := {(ȳ, s) ∈ Rn+1 : (ȳ, t − s) ∈ E(x̄, r)} the adjoint heat ball.
We say that x̄ ∈ ∂Ω belongs to the parabolic boundary of Ω (resp. adjoint parabolic
boundary) and write x̄ ∈ PΩ (resp. P ∗ Ω), if Cr− (x̄) ∩ Ωc 6= ∅ (resp. Cr+ (x̄) ∩ Ωc 6= ∅
) for every r > 0. If x̄ ∈ PΩ (resp. P ∗ Ω), we say that x̄ is in the bottom boundary of
Ω (resp. adjoint bottom boundary) and write x̄ ∈ BΩ (resp. B ∗ Ω) if there exists ε > 0
such that Cε+ (x̄) ⊂ Ω (resp. Cε− (x̄) ⊂ Ω). Moreover, we define the lateral boundary (resp.
adjoint lateral boundary) as S ′ Ω = PΩ \ BΩ (resp. S ′ ∗ Ω = P ∗ Ω \ B ∗ Ω) of Ω and denote
by SΩ the quasi-lateral boundary of Ω (for the precise definition see Section 2).
We also refer to Section 3 for the definition of Cap(·), the thermal capacity of a set.
Definition 1.1. Let Ω ⊂ Rn+1 be an open set and F ⊂ PΩ be compact. We say that a Ω
satisfies the Time-Backwards Capacity Porosity Condition (TBCPC) along F if there exists
a constant c > 0 and two sequences {ξ̄j }j≥1 ⊂ F and rj → 0, such that

(1.3) Cap E(ξ̄j ; rj2 ) ∩ Ωc ≥ c rjn .
If F = {ξ̄}, then we say that Ω satisfies the TBCPC at ξ̄. We remark that c, ξ¯j , and rj may
depend on F .
Let Ω+ and Ω− be two disjoint open sets in Rn+1 with S ′ Ω+ ∩ S ′ Ω− 6= ∅ and assume
that F ⊂ S ′ Ω+ ∩ S ′ Ω− is compact. Then we say that Ω+ and Ω− satisfy the joint TBCPC
along F if there exists a sequence (ξ̄j , rj )j≥1 ⊂ F ×(0, 1) so that rj → 0 and (1.3) holds for
Ω± , then there exists a subsequence of (ξ̄j , rj )j≥1 so that (1.3) holds for Ω∓ (with possibly
different constant c). We may define the time-forwards CPC (TFCPC) by replacing the heat
balls by adjoint heat balls in the definitions above.
A subset E of Rn+1 is (geometrically) porous if for any ball B centered on E with radius
small enough, there exists a ball of comparable radius that is contained in B ∩ E c . Porosity
of a domain Ω ⊂ Rn+1 is directly related to boundary regularity of solutions of elliptic and
prabolic PDE defined in Ω as it quantifies the non-degeneracy of the exterior of the domain.
There are different notions of scale-invariant non-degeracy of the exterior of Ω such as the
geometric porosity of Ω defined above, (Lebesgue) measure porosity or capacitary porosity
of Ω. This is the reason why we use the term “porosity”. For instance, we refer to [BB11,
Corollary 11.25].
Note that since Ω+ and Ω− are disjoint, it holds that PΩ± ∩ BΩ∓ = BΩ+ ∩ BΩ− = ∅
and thus, we could have assumed PΩ+ ∩ PΩ− 6= ∅ in the definition above. For Ω ⊂ Rn+1 ,
we denote
Tmin = inf{T ∈ R : Ω ∩ {t = T } =
6 ∅},
Tmax = sup{T ∈ R : Ω ∩ {t = T } =
6 ∅}.
Note that it is possible that Tmin = −∞ and Tmax = +∞. We also define
Es = {(x, t) ∈ E : t = s}
to be the time-slice of E for t = s.
Definition 1.2. We say that Ω satisfies the Time-Backwards Capacity Density Condition
(TBCDC) at ξ̄0 = (ξ0 , t0 ) ∈ SΩ if there exists e
c > 0 such that
p
(1.4) Cap E(ξ̄0 ; r 2 ) ∩ Ωc ≥ ec r n , for all 0 < r < t0 − Tmin /2.
Analogously, we say that Ω satisfies the Time-Forwards Capacity Density Condition (TFCDC)
if, for sime e
c > 0,
p
(1.5) Cap E ∗ (ξ̄0 ; r 2 ) ∩ Ωc ≥ e
c r n , for all 0 < r < Tmax − t0 /2.
The joint TBCPC guarantees regularity of the parabolic boundary for the heat equation
and is a much weaker condition than TBCDC, which, in turn, is a natural setting for our
problems (see e.g. [AMT17]). The joint TBCPC assumption is a sufficient condition for
our blow-up arguments to work and, at the moment, it is not clear how to remove it from our
hypotheses. In the case of harmonic and elliptic measures, it has been proved in [AMTV19,
Lemma 5.3] and [AM19, Lemma 4.13] that mutual absolute continuity of the interior and
the exterior measures implies an even stronger version of (1.3) in terms of the (n + 1)-
dimensional Lebesgue measure of euclidean balls. This method, however, is based on a
dichotomy argument and cannot be generalized directly to the caloric measure because it is
not enough to obtain that the exterior of the domain is “large” in a sequence of parabolic
balls; we need to know that this is true in a sequence of time-backwards cylinders, which
complicates things significantly.
Criteria for parabolic Wiener regularity have been extensively studied in the literature
(see Section 3 for the definition and more references). This property is particularly impor-
tant to our purposes because we need to extend the Green function by zero to the comple-
ment of the domain in a continuous way. In order to work in more generals domains, the
intrinsic difficulties of heat potential theory constitute a challenging obstacle to overcome.
For instance, we remark that Harnack inequality is time-directed (see e.g. [Wa12, Corollary
1.33]) and that it is possible to construct domains so that the capacity of the set of the ir-
regular points for the heat equation is positive (see [TW85, p.336]). The most general class
of domains in which (some of) our arguments work seem to be the so-called quasi-regular
domains for the heat operator (resp. adjoint heat operator), which amounts to domains for
which the set of irregular points of the essential boundary for the heat equation (resp. adjoint
heat equation) is polar (see Section 3). This is in accordance to the elliptic theory where,
indeed, the set of irregular points is always polar (which is not the case in the parabolic
context).
Before we state our first main theorem, which is a generalization of [KPT09] and [AM19,
Theorem I], we need the notion of parabolic Hausdorff dimension of a Borel measure ω.
This is defined as

dimHp (ω) := inf dimHp (Z) : ω(Rn+1 \ Z) = 0 ,
where “dimHp ” stands for the parabolic Hausdorff dimension (see Section 2), and it can be
linked to a pointwise notion (see Section 4). Moreover, given A ⊂ ∂Ω we denote by A◦ the
relative interior A with respect to the topology of ∂Ω.
Theorem I. Let Ω+ and Ω− be two disjoint n+1

◦ domains in R ± which are quasi-regular
∗
for H and regular for H , such that P Ω ∗ ± 6= ∅, and let ω be the caloric measures
◦ ◦
associated with Ω with poles p̄± ∈ Ω . Let also E ⊂ P ∗ Ω+ ∩ P ∗ Ω− ∩ supp ω +
± ±
be such that ω + (E) > 0 and ω + |E ≪ ω − |E ≪ ω + |E , and assume that Ω+ and Ω− satisfy
the joint TBCPC at all points of E. Then, Tan(ω ± , ξ̄) ⊂ F for ω + -a.e. ξ̄ ∈ E and
dimHp ω ± |E = n + 1.
Moreover, if Ω± also satisfy the TFCDC,
(1.6) lim ΘFΣ± (ξ̄, r) = 0 for ω + -a.e. ξ¯ ∈ E.
∂Ω
r→0
We remark that if ∂a∗ Ω± ∩ S ∗ Ω± = ∅, then (P ∗ Ω± )0 = P ∗ Ω± . Any domain that

satisfies the TFCDC has this property (see Remark 3.13).
The theorem above (and similarly all the theorems we prove) can be formulated for ad-
joint caloric measures if we assume the joint◦ TFCPC and the TBCDC in place of the joint
◦
TBCPC and TFCDC and let E be in PΩ+ ∩ PΩ− . We point out that caloric measure
is not necessarily doubling in domains such as the ones considered in Theorem I and also
that it is natural to study mutual absolute continuity of ω + and ω − on the lateral part of the
boundary because it cannot occur elsewhere. For more details on this matter, we refer to
Section 10.
Secondly, we prove the caloric equivalent of Tsirelson’s theorem [Ts97] about triple-
points for harmonic measure. Tsirelson’s method is based on the fine analysis of filtrations
for Brownian and Walsh-Brownian motions, although, more recently, Tolsa and Volberg
[TV18] obtained the same result using analytical tools and, in particular, blow-up argu-
ments, which is exactly the method we follow as well.
Theorem II. Let Ωi ⊂ Rn+1 , 1 ≤ i ≤ 3, be three disjoint open sets which are quasi-
regular for both H and H ∗ and such that P ∗ Ωi )◦ 6= ∅, i = 1, 2, 3. Let also ω i be the
◦
caloric measure in Ωi with pole at p̄i ∈ Ωi and assume that E ⊂ ∩3i=1 P ∗ Ωi ∩ supp ω 1
is such that ω i |E ≪ ω j |E ≪ ω i |E , for 1 ≤ i, j ≤ 3. Then ω i (E) = 0 for i = 1, 2, 3.
Note that Theorem II needs neither the joint TBCPC assumption on the domains nor reg-
ularity for H ∗ . Quasi-regularity is all we need to assume since, by an interior approximation
argument that we prove in Section 3, we show that it suffices to study the problem in regular
domains.
Under a pointwise VMO-type condition on dω − /dω + on a particular subset of the
boundary E, we prove a local analogue of [AM19, Theorem II].
Theorem III. Let Ω+ and Ω− be two disjoint domains in Rn+1 which are quasi-regular
for H and regular for H ∗ , such that P ∗ Ω± )◦ 6= ∅, and let ω ± be the caloric measures
◦ ◦
associated with Ω± with poles p̄± ∈ Ω± . Let also E ⊂ P ∗ Ω+ ∩ P ∗ Ω− ∩ supp ω +
be a relatively open set in supp ω + such that ω + (E) > 0 and ω − |E ≪ ω + |E . Assume that
−|
Ω+ and Ω− satisfy the joint TBCPC at all points of E and, if we set f = dω E
dω + |E to be the
Radon-Nikodym derivative of ω − |E with respect to ω + |E , we assume that for ξ̄ ∈ E,
Z Z
+ +
(1.7) lim − f dω |E exp −− log f dω |E = 1,
r→0 Cr (ξ̄) Cr (ξ̄)
¯ 6= ∅, then there is k ≥ 1 such that Tan(ω + , ξ̄) ⊂ F (k) and
and Tan(ω + , ξ)
ω + (C2r (ξ̄))
lim sup < ∞.
r→0 ω + (Cr (ξ̄))
Furthermore, if Ω+ and Ω− also satisfy the TFCDC, then
F (k)
(1.8) Θ∂ΩΣ± (ξ̄, r) → 0, as r → 0.
Before we go any further, let us introduce the space of Vanishing Mean Oscillation.
Given a Radon measure ω in Rn+1 , we denote
Z Z
1
fCr (x̄) := − f dω := f dω, x̄ ∈ supp ω,
Cr (x̄) ω(Cr (x̄)) Cr (x̄)
and we say that f ∈ VMO(ω) if f ∈ L1loc (ω), and
Z

lim sup − f (ȳ) − fC (x̄) dω(ȳ) = 0.
r
r→0 x̄∈supp ω Cr (x̄)
The condition (1.7) implies a vanishing mean oscillation assumption on dω − /dω + on E.

However, in general, these assumptions are not equivalent. For more details, we refer to
[AM19, Section 7].
The next theorem is a global version of Theorem III under the additional assumption that
ω± is doubling. This is the analogue of [AM19, Theorem III], which, in turn generalized,
the main results of [KT06] and [Bad11]. We prove that at sufficiently small scales, the
support of ω + resembles the zero set of an adjoint caloric polynomial uniformly on compact
subsets of the set of mutual absolute continuity. This assertion can be formulated both in
P (d)
terms of Θ∂ΩΣ± and the functional d1 (·, P(d)), which is essentially a distance between
measures and the set P(d) (see (4.4) for the exact definition).
Theorem IV. Let Ω+ , Ω− , ω + , ω − , f and E be as in the statement of Theorem III. If
log f ∈ VMO(ω + |E ) and ω + |E is C-doubling, i.e.
ω + |E (C2r (x̄)) ≤ Cω + |E (Cr (x̄)), x̄ ∈ Rn+1 , r > 0,
then, there exists an integer d = d(n, C) > 0, so that for every compact set F ⊂ E, it holds

(1.9) lim sup d1 Tξ̄,r [ω + ], P(d) = 0.
r→0 ξ̄∈F
If Ω+ and Ω− also satisfy the TFCDC, then

P (d)
(1.10) lim sup Θ∂ΩΣ± (ξ̄, r) = 0.
r→0 ξ̄∈F
That is, E ∈ LBW A(PΣ (d)).
Theorem IV also applies directly to the study of (parabolic) Reifenberg flat domains,
and gives an alternative proof of a result that Nyström proved with different techniques in
[Nys06c]. We recall that Ω ⊂ Rn+1 is δ-Reifenberg flat, δ > 0, if for R > 0 and ξ̄ ∈ ∂Ω
there exists a n-plane Lξ̄,R through ξ̄ containing a line parallel to the time-axis and such
that

ȳ + rn̂ ∈ CR (ξ̄) : r > δR, ȳ ∈ Lξ̄,R ⊂ Ω,

ȳ − rn̂ ∈ CR (ξ̄) : r > δR, ȳ ∈ Lξ̄,R ⊂ Rn+1 \ Ω,
where n̂ is the normal vector to Lξ̄,R pointing into Ω at ξ̄.
For t0 ∈ R, we denote Ωt0 := Ω ∩ {t < t0 }. The domain Ω (resp. Ωt0 ) is vanishing
Reifenberg flat if it is δ-Reifenberg flat for some δ > 0 and, for any compact set K (resp.
K ⊂⊂ {t < t0 }),
lim sup ΘF Σ
∂Ω (ξ̄, r) = 0.
r→0 ξ̄∈K∩∂Ω
Let us also observe that, if Ω is δ-Reifenberg flat, it readily follows that ∂a Ω = ∂a∗ Ω = ∅.
Additionally, it is easy to see that a δ-Reifenberg flat domain satisfies the TBCDC and the
TFCDC (with parameters depending on δ) and, thus, is regular for H and H ∗ .
Corollary 1.3. Let also Ω+ ⊂ Rn+1 be a δ-Reifenberg flat domain and set Ω− = Rn+1 \
Ω+ . Let ω ± be the caloric measures of Ω± with poles at p̄± = (p± , t± ) ∈ Ω± . If ω − ≪ ω + ,
−
log f = log dω +
dω + ∈ VMO(ω ), and δ is small enough (depending on n), then (Ω )
+ t− is
vanishing Reifenberg flat.

The analogous result for harmonic measure can be found in [KT06, Corollary 4.1], while
for the original theorem for caloric measure we refer to [Nys06c, Theorem 1].
Discussion of the proofs. In the current paper, we follow the same strategy as in [AM19],
although there are many significant challenges along the way that we overcome to adapt this
method successfully to the parabolic setting. In Section 3, we exhibit various properties of
thermal capacity, the most important of which is the backwards in time self-improvement
of a pointwise time-backwards capacity density condition on a particular scale. This is
the building block of the proofs of several PDE estimates around the boundary that are
absolutely necessary for our blow-up arguments in Section 9. Those results are new and
interesting in their own right, and, judging from their elliptic analogues, we expect them
to be used widely in future works. In Section 4, we develop the required parabolic GMT
framework and confirm that, due to a parabolic Besicovitch covering theorem, the theory
of tangent measures translates almost unchanged to the parabolic setting. Moreover, we
show that the blow-ups of the “parabolic surface measure” on euclidean rectifiable sets are
parabolic flat, meaning that there is a plane that contains a line parallel to the time axis such
that the blow-up measure is the parabolic Hausdorff measure restricted to that plane. Han
and Lin [HL94] had proven that if h is caloric, then in any ball centered on Σh = {h = 0}, it
holds that Σh can be written as the union of its regular set, which is a smooth n-submanifold,
and its singular set, which is an (n − 1)-rectifiable set. Unlike the harmonic case, this is not
enough for our purposes and so, in Section 5, we investigate the finer structure of the regular
set separating space and time-singularities. We demonstrate that any x̄ ∈ Rx = Σh ∩
{|∇h| = 6 0} has a neighborhood so that Σh is given by an admissible n-dimensional smooth
graph, while any x̄ ∈ Rt = Σh ∩{|∇h| = 0}∩{|∂t h| = 6 0}, has a neighborhood in which Rt
is contained in a smooth (n − 1)-graph. Those results are fundamental to several measure-
theoretic arguments and are used repeatedly in Sections 6-9. To the best of our knowledge,
this description of the time-regular set is novel in the literature. In Section 6, we prove the
existence and uniqueness of the caloric measure associated with a caloric function of the
form dωh = −∂νt hdσh , where σh is the “surface measure” on Σh . In the elliptic setting,
this is a straightforward application of the Gauss-Green formula on sets of locally finite
perimeter, which is far from being the case here. The aim of Section 7 is to prove that, if
the parabolic tangent measures Tan(ω, ξ) ¯ to a given Radon measure ω are caloric measures
associated with caloric polynomials and there is a measure corresponding to a homogeneous
caloric polynomial of degree k, then all the elements of Tan(ω, ξ) ¯ are caloric measures
associated with a homogeneous caloric polynomial of the same degree. The connectivity
arguments for tangent measures translate unchanged from [AM19], although, this is not true
for the analogues of the main lemmas from [Bad11]. In fact, it is not clear to us how to adapt
Badger’s proofs directly to our case, although, his ideas inspired us to come up with new
ones, which we find simpler even for harmonic functions. In Section 8, we introduce some
basic notions of Optimal Transport. We consider the space of signed Borel measures in a
bounded open subset of Rn equipped with the adapted Kantorovich-Rubinshtein norm and
prove that its Banach dual can be identified with the space Lip01 of 1-Lispchitz functions
on Ω vanishing at the boundary. We also identify its completion as a subspace of the space
of first order distributions in the dual of Lip01 and show that L2 functions are embedded
continuously in that completion. As we deal with signed measures so that µ(Ω) is not
necessarily zero, the results are less standard and not broadly used in the literature and
so we have to adapt the existing proofs to our setting. Although, those modifications are
not trivial and require some care. This functional analytic approach is novel in our area
of research and it is an essential component of a compactness argument for the blow-ups
of the Green’s function. Section 9 groups the main one- and two-phase blow-up lemmas,
for which we follow the approach in [AMT17], [AMTV19], and [AM19]. However, there
are substantial obstacles such as the compact embedding of a “parabolic” Sobolev space
in L2 , the lack of analyticity in time, the unique continuation arguments, the global Hölder
continuity of the rescaled Green functions, and the locality of Lemma 9.8. We highlight that
for the blow-up arguments to work, we prove a novel boundary Caccioppoli-type inequality
for the t-derivative of the square of a positive caloric function vanishing on a “boundary
ball” and introduce a new mixed-norm Sobolev space based on the aforementioned space of
finite signed measures with the Kantorovich-Rubinshtein norm. Finally, in Section 10 we
provide the proofs of the four main theorems using the results from the previous sections.
Acknowledgements. We would like to thank Jonas Azzam, Max Engelstein, Luis Escau-
riaza, Steve Hofmann, Tuomo Kuusi, Andrea Merlo, and Xavier Tolsa for several fruitful
discussions on topics related to the current paper. The first named author would like to
dedicate this paper to the memory of his teacher, mentor, and dear friend, Professor Michel
Marias.
2. P RELIMINARIES AND NOTATION
We denote points in Rn+1 by x̄ = (x, t) = (x′ , xn , t) := (x1 , . . . , xn , t) and define the

(parabolic) norm
kx̄k := max{|x|, |t|1/2 },
p
where |x̄| = x21 + · · · + x2n stands for the standard euclidean norm. We also use the
notation 0̄ = (0, . . . , 0).
If E ⊂ Rn+1 , we define the parabolic distance to E as distp (x̄, E) = inf{kx̄ − ȳk : ȳ ∈
E} and the parabolic diameter as diamp (E) = sup{kx̄ − ȳk : (x̄, ȳ) ∈ E × E}.
Given an open set Ω, and a point ξ¯ ∈ Ω, we denote by Λ(ξ̄, Ω) (resp. Λ∗ (ξ̄, Ω)) the set
of points x̄ ∈ Ω that are lower (resp. higher) than ξ̄ relative to ∂Ω, in the sense that there is
a polygonal path γ ⊂ Ω joining ξ̄ to x̄, along which the time variable is strictly decreasing
(resp. increasing). By a polygonal path, we mean a path which is a union of finitely many
line segments.
Following [Wa12, Definition 8.1], we define the normal boundary
(
PΩ , if Ω is bounded
∂n Ω =
PΩ ∪ {∞} , if Ω is unbounded
and the abnormal boundary ∂a Ω = {x̄ ∈ ∂Ω : ∃ ε > 0 such that Cε− (x̄) ⊂ Ω} . The ab-
normal boundary is further decomposed into ∂a Ω = ∂s Ω ∪ ∂ss Ω, where ∂s Ω stands for the
singular boundary and ∂ss Ω for the semi-singular boundary, which are defined respectively
by

∂s Ω = x̄ ∈ ∂a Ω : ∃ ε > 0 such that Cε+ (x̄) ∩ Ω = ∅
and

∂ss Ω = x̄ ∈ ∂a Ω : Cr+ (x̄) ∩ Ω 6= ∅, for all r > 0 .
By [Wa12, Theorem 8.40], ∂a Ω is contained in a sequence of hyperplanes of the form
Rn × {t}. The essential boundary is defined as ∂e Ω = ∂n Ω ∪ ∂ss Ω = ∂Ω \ ∂s Ω, replacing
∂Ω by ∂Ω ∪ {∞} if Ω is unbounded. Finally, following [GH20], we define the quasi-lateral
boundary to be


 ∂Ω , if Tmin = −∞ and Tmax = ∞

∂Ω \ (BΩ)
Tmin , if Tmin > −∞ and Tmax = ∞
SΩ =

 ∂Ω \ (∂s Ω)Tmax , if Tmin = −∞ and Tmax < ∞


∂Ω \ ((BΩ)Tmin ∪ (∂s Ω)Tmax ) , if − ∞ < Tmin < Tmax < ∞,
where (BΩ)Tmin is the time-slice of BΩ with t = Tmin and (∂s Ω)Tmax is the time-slice
of ∂s Ω with t = Tmax . Observe that for a cylindrical domain U × (Tmin , Tmax ), where
U ⊂ Rn is a domain in the spatial variables, the quasi-lateral boundary coincides with the
lateral boundary. By [GH20, Lemma 1.14], both ∂e Ω and SΩ are closed sets.
Given f : Rn+1 → R we write Df = (∇f, ∂t f ) for the (full) gradient of the function
f and, D α,ℓ f = ∂xα11 · · · ∂xαnn ∂tℓ f for higher order derivatives, where α ∈ Zn+ and ℓ ∈ Z+ .
If E ⊂ Rn+1 and f is a continuous function with compact support in E ⊂ Rn+1 , then we
m
write f ∈ Cc (E). If Ω is an open set, we denote by C m, 2 (Ω) the class of functions such
m
that f (·, t) ∈ C m (Ωt ) for any fixed t ∈ (Tmin , Tmax ) and f (x, ·) ∈ C 2 (Tmin , Tmax ) for
any fixed x̄ ∈ Rn+1 such that x̄ ∈ Ω. If f is C m in both space and time variables, we will
m, m
simply write that f ∈ C m (Ω). Finally, we say that f ∈ Cc 2 (E) if f has compact support
m
in E ⊂ Rn+1 and f ∈ C m, 2 (Rn+1 ).
We write a . b if there is C > 0 so that a ≤ Cb and a .t b if the constant C depends on
the parameter t. We writeRa ≈ b to mean a . b . a and define
R a ≈t b similarly. Sometimes
we also use the notation −F dµ for the average µ(F )−1 F dµ over a set F ⊂ Rn+1 with
respect to the measure µ.
If E ⊂ Rn+1 is a Borel set and Hd stands for the Euclidean d-Hausdorff measure in
Rn+1 for d ≤ n − 1, we define the d-“surface measure” on E as the measure dσ = dσt dt,
where σt = Hd |Et .
If s ∈ [2, n + 2] and 0 < δ < ∞, we set
nX [ o
s
Hp,δ (E) = inf diamp (Ej )s : E ⊂ Ej , diamp (Ei ) ≤ δ ,
j
for E ⊂ Rn+1 , and, as in the Euclidean case, define the parabolic s-Hausdorff measure by
s
Hps (A) = lim Hp,δ (E),
δ→0
which is a Borel measure by the Carathéodory criterion. We also define the parabolic
Hausdorff content of E by
nX [ o
s
Hp,∞ (E) = inf diamp (Ej )s : E ⊂ Ej .
j
We say that the parabolic Hausdorff dimension of E is

dimHp (E) = sup{s : Hps (E) > 0} = inf{t : Hpt (E) = 0}.
3. H EAT P OTENTIAL T HEORY AND PDE ESTIMATES
Let Ω ⊂ Rn+1 be an open set. Define the parabolic operator

(3.1) Ha u := a∆ − ∂t , and Ha∗ u := a∆ + ∂t , for a > 0.
When a = 1, we simply write H1 = H and H1∗ = H ∗ for the heat and the adjoint heat
operator respectively.
For x̄ = (x, t) ∈ Rn+1 , we denote by

Γa (x̄) = (4πat)−n/2 exp −|x|2 /4at χ{t>0} (t)
the Gaussian kernel. Note that Γa is the fundamental solution associated with Ha , i.e., it
satisfies
(1) Ha Γa (x, t) = δ0̄ (x, t), in the distributional sense, for {t > 0}, where δ0̄ stands for
the Dirac mass at 0̄;
(2) Γa (x, t) = δ0 (x), for t = 0;
Moreover, Γa ∈ C ∞ (Rn+1 \ {0̄}) and
(3.2) Γa (x̄) ≤ Ch π −n/2 kx̄k−n ,
for some constant Ch > 0 depending on n and a.
We say that a function u ∈ C 2,1 (Ω) is caloric (resp. adjoint caloric) if it satisfies the
(adjoint) heat equation Hu = 0 (resp. H ∗ u = 0) in a pointwise sense. If u ∈ C 2,1 (Ω) and
Hu ≥ 0 (resp. Hu ≤ 0) in Ω, we say that u is subcaloric (resp. supercaloric).
A distribution u ∈ D ′ (Ω) is called weakly caloric (resp weakly subcaloric or weakly
supercaloric), if it satisfies
Z
u H ∗ ϕ = 0, (resp. ≥ 0 or ≤ 0),
Ω
for all ϕ ∈ C 2,1 (Ω)

(resp. 0 ≤ ϕ ∈ C 2,1 (Ω)). If u is a supertemperature (resp. subtem-
perature), then it is weakly supercaloric (resp. weakly subcaloric) (see [Wa12, Definition
3.7, Theorem 6.28]). Moreover, by [Wa12, Theorem 6.28], if u is a subcaloric in Ω, then
there exists a unique non-negative Radon measure µu , which is called the Riesz measure
associated with u, such that
Z Z
(3.3) u H ∗ϕ = ϕ dµu , for all ϕ ∈ C 2,1 (Ω).
Ω Ω
By making the obvious adjustments we obtain the dual definitions for H ∗ .
Lemma 3.1. If Ω ⊂ Rn+1 is an open set, then the following are equivalent:
(1) u ∈ C 2,1 (Ω) is caloric in Ω,
(2) u ∈ D ′ (Ω) is weakly caloric in Ω,
(3) u ∈ C ∞ (Ω) and Hu = 0 in Ω.
Proof. That (1) implies (2) follows from integration by parts, while (3) implies (1) is trivial.
Remark that the heat operator is hypoelliptic (see [Vl02, 15.6, p. 210]) since the heat kernel
is the fundamental solution for the heat equation. Thus, the remaining implication is true
because of [Vl02, Theorem 1, p. 210].
Let f be an extended real-valued function defined in ∂e Ω. The upper class Uf of f

consists of lower bounded hypertemperatures w in Ω (see [Wa12, Definition 3.10]) such
that w ≥ f on ∂e Ω in the sense that it holds
lim inf w(y, s) ≥ f (ξ, t), for all (ξ, t) ∈ ∂n Ω,
(y,s)→(ξ,t)
and
lim inf w(y, s) ≥ f (ξ, t), for all (ξ, t) ∈ ∂ss Ω.
(y,s)→(ξ,t+ )
The lower class Lf of f consists of upper bounded hypotemperatures v in Ω such that v ≤ f
on ∂e Ω in the sense that it holds
lim sup v(y, s) ≤ f (ξ, t), for all (ξ, t) ∈ ∂n Ω,
(y,s)→(ξ,t)
and
lim sup v(y, s) ≤ f (ξ, t), for all (ξ, t) ∈ ∂ss Ω.
(y,s)→(ξ,t+ )
The function Uf (x) = inf{w ∈ Ω : w ∈ Uf } is called the upper solution for f in Ω,
and Lf (x) = sup{v ∈ Ω : v ∈ Lf } is called the lower solution for f in Ω. We say that f
is resolutive if Uf = Lf =: Sf and Sf is caloric in Ω. In this case, Sf is called the PWB
solution for f in Ω. By [Wa12, Theorem 8.26], any f ∈ C(∂e Ω) is resolutive. Therefore,
for any x̄ ∈ Ω, the map f 7→ uf (x̄) is linear and, by the Riesz representation theorem, there
exists a unique probability measure ω x̄ on ∂e Ω, which is called caloric measure, such that
Z
uf (x̄) = f dω x̄ .
∂e Ω
R
More generally, if f is an extended real-valued function and x̄ ∈ Ω and ∂e Ω f dω x̄ exists,
then
Z
(3.4) Uf (x̄) = Lf (x̄) = f dω x̄ .
∂e Ω
Conversely, if Uf (x̄) = Lf (x̄) and is finite, then f is ω x̄ -integrable and (3.4) holds (see
[Wa12, Theorem 8.32]). In the last statement, if we also assume that f is Borel, then we
obtain that f is resolutive. In particular, if E ⊂ ∂e Ω is ω x̄ -measurable for all x̄ ∈ Ω,
then χE is resolutive and SχE (x̄) = ω x̄ (E) (see [Wa12, Corollary 8.33]). Therefore, if
ω x̄ (E) = 0 for some x̄ ∈ Ω, then, by minimum principle, ω ȳ (E) = 0 for all ȳ ∈ Λ(x̄, Ω).
If f ∈ C(∂e Ω), we say that u is a solution of the classical Dirichlet problem with data f
if u is caloric and it holds that
(3.5) lim u(y, s) = f (ξ, t), for all (ξ, t) ∈ ∂n Ω,
(y,s)→(ξ,t)
and
(3.6) lim u(y, s) = f (ξ, t), for all (ξ, t) ∈ ∂ss Ω.
(y,s)→(ξ,t+ )
If there exists a solution to the classical Dirichlet problem in Ω with data f , then f is
resolutive and uf is the PWB solution for f in Ω (see [Wa12, Theorem 8.26]).
Definition 3.2. A point ξ¯0 ∈ ∂n Ω (resp. ∂ss Ω) is called regular for H or H-regular if for
any f ∈ C(∂e Ω), the PWB solution Sf satisfies (3.5) (resp. (3.6)). A point ξ̄0 ∈ ∂e Ω that
is not H-regular is called H-irregular.
Let GΩ (·, ·) be the non-negative real-valued function defined in Ω × Ω as

GΩ (x̄, p̄) = Γ(x̄ − p̄) − hΩ (x̄, p̄),
where hΩ (·, p̄) is the greatest thermic minorant of Γ(· − p̄), which is non-negative (see
[Wa12, Definition 3.65] and the paragraph before this definition). We call G(·, p̄) the Green
function in Ω with pole at p̄ ∈ Ω and, by [Wa12, Theorem 8.53], if ψp̄ = Γ(· − p̄)|∂e Ω , we
also have the representation
Z
GΩ (x̄, p̄) = Γ(x̄ − p̄) − Sψp̄ (x̄) = Γ(x̄ − p̄) − Γ(ȳ − p̄) dω x̄ (ȳ),
∂e Ω
where the last equality follows from the fact that Γ(·, p̄) is continuous on ∂e Ω and thus
resolutive. It is straightforward to see that
(3.7) GΩ (x̄, ȳ) ≤ 2Ch π −n/2 kx̄ − ȳk−n .
In fact, more is true: if Cr := Br × (0, r) is a cylinder of radius r > 0, then there exists
c1 > 0 and c2 > 0 (independent of r) such that
(3.8) c1 m(x̄, ȳ) Γc2 (x̄ − ȳ) ≤ GCr (x̄, ȳ) ≤ c−1
1 m(x̄, ȳ) Γ 1 (x̄ − ȳ), for x̄ 6= ȳ ∈ Cr ,
c2
where
dist(x, Br ) dist(y, Br )
m(x̄, ȳ) := min 1, √ min 1, √ .
t−s t−s
See e.g. [Zh02, Theorem 1.2]1 and [Cho06, Theorem 3.3].
Note that GΩ (·, ·) is lower semi-continuous on the diagonal {(p̄, p̄) : p ∈ Ω} and con-
tinuous everywhere else in Ω. In fact, it is a supertemperature in Ω and a temperature in
Ω \ {p̄}. Recall that Λ∗ (p̄, Ω) is the set of points x̄ ∈ Ω for which there is a polygonal path
γ ⊂ Ω joining p̄ to x̄, along which the time variable is strictly increasing. By [Wa12, Theo-
rem 6.7], it holds that GΩ (·, p̄) > 0 in Λ∗ (p̄, Ω) and GΩ (·, p̄) = 0 in Ω \ Λ∗ (p̄, Ω), for any
p̄ ∈ Ω. If ξ¯ ∈ ∂n Ω (resp. ∂ss Ω) is a regular point for H, we have that limx̄→ξ̄ GΩ (x̄, p̄) = 0
(resp. limx̄→(ξ,t+ ) GΩ (x̄, p̄) = 0). When the domain Ω where the Green function is defined
is clear from the context, we will drop the subscript and just write G(·, ·).
Let us remark that the Riesz measure associated with the Green function GΩ (x̄, ȳ) in Ω
is the caloric measure ωΩ x̄ in Ω (see [Do84]), i.e.,
Z Z
(3.9) GΩ (x̄, ȳ)Hϕ(ȳ) dȳ = ϕ dω x̄ , for any ϕ ∈ Cc∞ (Rn+1 ).
Given a Borel measure µ on Rn+1 and an open set Ω ⊂ Rn+1 ,

Z
GΩ µ(x̄) := GΩ (x̄, ȳ) dµ(ȳ)
1The constants depend on the geometry but not the diameter of the domain. For cylinders, the constants are
only dimensional.
defines a non-negative supertemperature in Ω provided the integral is finite in a dense subset

of Ω and is called the heat potential of µ in Ω. If Ω = Rn+1 , we replace GΩ by Γ.
Given a compact set K ⊂ Rn+1 , we define its thermal capacity as

(3.10) Cap(K, Ω) := sup µ(K) : GΩ µ ≤ 1 in Ω, µ ≥ 0 and supp µ ⊂ K .
Equivalently, if u = Rb1 (K, Ω) is the smoothed reduction 1 over K in Ω (see [Wa12, Defini-
tion 7.23]) and µu is the associated Riesz measure, then Cap(K, Ω) = µu (K) (see [Wa12,
Definition 7.33]). By [Wa12, Theorem 7.28],
b1 (K, Ω) = GΩ µu .
u=R
It is easy to see that if K ⊂ Ω1 ⊂ Ω2 ⊂ Rn+1 are open sets, then
(3.11) Cap(K, Rn+1 ) ≤ Cap(K, Ω2 ) ≤ Cap(K, Ω1 ).
This definition can be extended to arbitrary sets S ⊂ Rn+1 : the inner thermal capacity
of S is given by
(3.12) Cap− (S, Ω) = sup{Cap(K, Ω) : S ⊃ K compact},
and its outer thermal capacity by
(3.13) Cap+ (S, Ω) = inf{Cap− (D, Ω) : S ⊂ D open}.
If the inner and the outer thermal capacities of S coincide, we say that S is capacitable
and we denote Cap(S, Ω) := Cap− (S, Ω) = Cap+ (S, Ω). In the case Ω = Rn+1 , we
simply write Cap(S). If E is a Borel set, then, by [Wa12, Theorems 7.15 and 7.49], it is
capacitable.
A set Z ⊂ Rn+1 is called polar if there is an open set Ω ⊃ Z and a supertemperature
w on Ω such that Z ⊂ {x̄ ∈ Ω : w(x̄) = +∞}. Observe that if Z1 ⊂ Z and Z is polar,
then Z1 is polar itself, and a set S ⊂ Rn+1 is polar if and only if Cap(S) = 0 (see [Wa12,
Theorem 7.46]). Moreover, if a set is contained in a horizontal hyperplane its capacity
is equal to its n-dimensional Lebesgue measure (see [Wa12, Theorem 7.55]). For more
properties of polar sets we refer the reader to [Wa12, Chapter 7].
An open set Ω is quasi-regular (resp. regular) if the set of irregular points of ∂e Ω is polar
(resp. empty). We remark here that there are domains so that the capacity of the set of the
irregular points of ∂e Ω for the heat equation is positive (see [TW85, p. 336]). This comes
in contrast to the corresponding result in elliptic potential theory, where the set of irregular
boundary points has zero capacity and thus, is polar.
Lemma 3.3. If Cr is a cylinder of radius r, there exists a dimensional constant c1 > 0 such
that
(3.14) Cap(Cr , C2r ) ≤ c1 r n .
Proof. Let ϕ ∈ Cc∞ (C2r ) such that ϕ = 1 in Cr , ϕ ≥ 0, and |H ∗ ϕ| ≤ c′n r −2 . Then, if µu
b1 (Cr , C2r ), since 0 ≤ u ≤ 1, it holds
is the Riesz measure associated with u = R
Z Z
Cap(Cr , C2r ) = µu (Cr ) ≤ ϕ dµu = u H ∗ ϕ ≤ c′n r −2 |C2r | = cn r n .
Lemma 3.4. Let Cr be a cylinder of radius r centered at ξ¯0 ∈ Rn+1 and let K ⊂ Cr be
a compact set such that ξ̄0 ∈ K. If s ∈ (0, 2], then there exists c2 > 0 (independent of K)
such that
n+s (K)
Hp,∞
(3.15) Cap(K, C2r ) ≥ c2 .
min(diam(K), r)s
Proof. If we set ρ = min(diam(K), r), it holds that K ⊂ Cρ ∩ Cr . Without loss of
generality we assume Hp,∞ n+s (K) > 0, otherwise (3.15) is trivially verified. By the parabolic
version of Frostman’s lemma, whose proof is analogous to the Euclidean one and is omitted
(see e.g. [Mat95, Lemma 8.8]), we can find a Borel measure µ′ supported on K, such that
• µ′ (Cr (ȳ)) . r n+s , for any r > 0 and ȳ ∈ Rn+1 , and
n+s (K) ≥ µ′ (K) ≥ c Hn+s (K),
• Hp,∞ p,∞
where c1 depends on n.
Let G2r be the Green function in C2r and define µ = µ′ ρ−s . We claim that the corre-
sponding heat potential G2r µ(x̄) . 1 for any x̄ ∈ C2r . Indeed, notice that
Z
Gµ(x̄) = G2r (x̄, ȳ) dµ(ȳ)
kȳ−x̄k≤ρ
Z
+ G2r (x̄, ȳ) dµ(ȳ) =: I1 + I2 .
kȳ−x̄k>ρ
Using (3.8) in conjunction with the fact that µ′ is supported on K and has (n + s)-growth,
we have that
Z
I2 . kȳ − x̄k−n dµ(ȳ)
kȳ−x̄k>ρ
. ρ−n µ′ (K)ρ−s ≤ µ′ (Cρ )ρ−n−s . 1.
If Ak (x̄) := C2−k ρ (x̄) \ C2−k−1 ρ (x̄), arguing as above, we infer that
XZ X
I1 . (2−k−1 ρ)−n dµ(ȳ) . (2−k−1 ρ)−n ρ−s (2−k ρ)n+s . 1,
k≥0 Ak (x̄) k≥0
concluding the proof of the claim.

If we normalize µ so that G2r µ(x̄) ≤ 1, we deduce that µ is an admissible measure for
the definition of thermal capacity on compact sets. Therefore,
Cap(K, C2r ) ≥ µ(K) & ρ−s Hp,∞
n+s
(K),
and the proof of the lemma is now complete.
Corollary 3.5. Let 0 < b < a ≤ 1 and r > 0, and for ξ¯0 = (ξ0 , t0 ) ∈ Rn+1 we set
−

Ra,b (ξ̄0 ; r) = B(ξ0 , r) × t0 − (ar)2 , t0 − (br)2 .
Then, we have that
−

(3.16) Cap Ra,b (ξ̄0 ; r), C2r (ξ̄0 ) ≈ r n ,
where the implicit constant depends on n, a, and b.
Proof. The upper bound follows from Lemma 3.3. For the lower bound, if s = n + 2
−
and K = Ra,b (ξ̄0 ; r), the proof of Lemma 3.4 goes through unchanged if instead of the
Frostman’s measure we use Hpn+2 , which, in turn, is equal to a constant multiple of the
(n + 1)-dimensional Lebesgue measure Ln+1 , showing that
(3.17) Cap(K, C2r (ξ̄0 )) & Ln+1 (K)/r 2 ≈a,b,n r n .
−
As a and b are arbitrary, we may approximate Ra,b (ξ̄0 ; r) by a sequence of compact subsets
of the form Ra−k ,bk (ξ̄0 ; r) and obtain our result.
For x̄ = (x, t) ∈ Rn+1 , we define the heat ball centered at x̄ with radius ρ > 0 to be the
set

(3.18) E(x, t; ρ) := ȳ ∈ Rn+1 : Γ(x̄ − ȳ) > (4πρ)−n/2
n r ρ o
n+1
= (y, s) ∈ R : |x − y| < 2n(t − s) log , t−ρ<s<t .
t−s
It is not hard to see that E(x, t; ρ) is a convex body, axially symmetric about the line {x}×R,
and
r 2nρ
E(x̄; ρ) ⊂ B x, × (t − ρ, t).
e
Let us set
A(x̄; ρ/2, ρ) = E(x̄; ρ) \ E(x̄; ρ/2)
to be the closed heat annulus of radius ρ. It is clear that if ȳ ∈ A(ξ̄0 ; ρ/2, ρ) satisfies
0 < t − s < ρ/2, then it holds that

ρ 2 ρ
(3.19) 2n(t − s) log − [2n log 2](t − s) ≤ |x − y| ≤ 2n(t − s) log .
t−s t−s
Remark 3.6. Note that if ρ = r 2 , α ∈ (0, 1), and t − s = αr 2 , then the region given
′
√ spatial variable which can be covered by at most cn many
by (3.19) is an annulus in the
(spatial) cubes of sidelength αr.
There are parabolic analogues of the Wiener criterion that determine whether a point
ξ̄ ∈ ∂n Ω is regular in terms of the capacity of the complement of Ω that lies either in
− (ξ̄ ). To be precise,
heat balls centered at ξ̄0 or in time-backwards parabolic cylinders Rar 0
ξ̄0 ∈ ∂n Ω is regular if and only if
Z 1
Cap(A(ξ̄0 ; ρ/2, ρ) ∩ Ωc ) dρ
(3.20) = ∞.
0 ρn/2 ρ
The necessity was proved by Evans and Gariepy [EG82, Theorem 1] and the sufficiency by
Lanconelli [Lanc73]2 Equivalently, ξ̄0 ∈ ∂n Ω is regular if and only if
Z 1
Cap(E(ξ̄0 ; ρ) ∩ Ωc ) dρ
(3.21) = ∞.
0 ρn/2 ρ
2A different necessary and sufficent condition is given by [Land69].

This follows from (3.20) and [Br90, Theorem 1.13]. Alternatively, ξ¯0 is regular if and only
if
X
(3.22) λ−kn Cap(A(ξ̄0 ; λ−k−1 , λ−k ) ∩ Ωc ) = ∞, for some λ > 1.
k≥0
In fact if ξ̄0 is regular, then (3.22) holds for all λ > 1.

Remark 3.7. In [EG82] and [Lanc73], the authors do not specify that this criterion only
works for points on ∂n Ω, although it is clear this is the case. Observe that, by definition, if
ξ̄0 ∈ ∂a Ω, there exists ρ > 0 small enough so that E(ξ̄0 ; ρ) ∩ Ωc = ∅ and thus, the integral
in (3.21) clearly converges. Nevertheless, Watson [Wa14, Theorem 4.1] proved that a point
ξ̄0 ∈ ∂ss Ω is regular if and only if there exists r0 > 0 such that H(ξ̄0 , r0 ) is a connected
component of Ω ∩ B(ξ̄0 , r0 ), where B(ξ̄0 , r0 ) = {(x, t) : |x − ξ0 |2 + |t − t0 |2 < r02 } and
H(ξ̄0 , r0 ) = B(ξ̄0 , r0 ) ∩ {t < t0 }.
Lemma 3.8. Let Ω ⊂ Rn+1 be an open set and ξ̄0 = (ξ0 , t0 ) ∈ ∂n Ω, If

Cap(E(ξ̄0 ; r 2 ) ∩ Ωc ) ≥ 2cr n ,
then there exists a ∈ (0, 1/2) depending on n such that

(3.23) Cap E(ξ̄0 ; r 2 ) ∩ Ωc ∩ {t < t0 − (ar)2 } ≥ cr n .
Proof. If we set
n o
Ek := x̄ = (x, t) ∈ E(ξ̄0 ; r 2 ) : 2−2(k+1) r 2 ≤ t0 − t ≤ 2−2k r 2 ,
then, for any x̄ ∈ Ek , it holds
r √
−k r2 p k
|x − ξ0 | < 2 r 2n log ≤ 2n log 2 r.
2−2(k+1) r 2 2k
Thus, we can cover Ek by at most c(n)kn/2 number of cylinders Cj (k) of radius 2−k r,
which, in view of the subadditivity of capacity and Lemma 3.3, infers that
X
Cap(Ek ) ≤ Cap(Cj (k)) ≤ c(n)c1 kn/2 2−kn r n .
j≥1
S
Therefore, if we set EM := Ek , we have that
k≥M
X X
Cap(EM ) ≤ Cap(Ek ) ≤ c̃(n) kn/2 2−kn r n .
k≥M k≥M
P
Since k≥1 kn/2 2−kn converges, we can find M > 0 depending on n so that Cap(EM ) ≤
cr n , which implies (3.23) for a = 2−M .
Similarly we can prove the following lemma.
Lemma 3.9. If Ω ∈ Rn+1 is an open set and ξ̄0 = (ξ0 , t0 ) ∈ ∂n Ω is a regular point, then
there exists a sequence Mk > 1 such that
X
(3.24) 2−kn Cap A(x̄0 ; 2−2(k+1) , 2−2k ) ∩ {t < t0 − (2−Mk 2−k )2 ∩ Ωc } = ∞.
k≥0
Given a ∈ (0, 1), we define the translated time-backwards cylinder

(3.25) Ra− (x̄; r) := Cr− (x, t − (ar)2 ) = B(x, r) × t − r 2 , t − (ar)2 .
Lemma 3.10. If a ∈ (0, 1) and ξ̄0 = (ξ0 , t0 ) ∈ Rn+1 , it holds that

1
(3.26) Ra− (ξ̄0 ; r) ⊂ E(ξ̄0 ; ρ), for ρ ≥ e 2n r 2 .
Proof. Let R′ := B(ξ0 , r) × {t0 − (ar)2 } and R′′ := B(ξ0 , r) × {t0 − r 2 }. By the definition
(3.18), we have that R′ ⊂ E(ξ̄0 ; ρ) if and only if
1
ρ ≥ a2 exp r 2 =: c̃1 r 2 ,
2na2
and R′′ ⊂ E(ξ̄0 ; ρ) if and only if
1
ρ ≥ exp r 2 =: c̃2 r 2 .
2n
Remark that, for any fixed λ > 0, the function yeλy is increasing for y > −λ−1 , and thus,
c̃2 > c̃1 . Hence, if ρ ≥ c̃2 r 2 , we have that R′ ∪ R′′ ⊂ E(ξ̄0 ; ρ) and, by the convexity of
E(ξ̄0 ; ρ), we get (3.26) for c̃ = c̃2 .
As backwards cylinders Ra− (ξ̄0 ; r) appear more naturally in applications, it is interesting
to know a version of the Wiener’s criterion with Ra− (ξ̄0 ; r) instead of heat balls. This can
be obtained using the following lemma (compare it with [GZ82, Theorem 3.1]).
Lemma 3.11. Let Ω be an open set and ξ̄0 ∈ ∂n Ω. If there exists a constant a > 0 such
that
Z 1
Cap(Ra− (ξ̄0 ; r) ∩ Ωc ) dr
(3.27) = ∞,
0 Cap(Ra− (ξ̄0 ; r)) r
then ξ̄0 is regular. Conversely, if ξ̄0 ∈ ∂n Ω is regular, then there exists a function a :
−
(0, 1) → (0, 21 ) so that (3.27) holds for Ra(r) (ξ̄0 ; r).
Proof. For ξ¯0 ∈ ∂n Ω, we have that

Z c̃ Z 1
Cap(E(ξ̄0 ; ρ) ∩ Ωc ) dρ 2 Cap(E(ξ̄0 ; c̃r 2 ) ∩ Ωc ) dr
=
0 ρn/2 ρ c̃n/2 0 rn r
Z 1 − c
(3.26) Cap(Ra (ξ̄0 ; r) ∩ Ω ) dr
≥ = ∞,
0 rn r
where c̃ is given in the proof of Lemma 3.10. The converse direction follows from Lemma
3.9.
Lemma 3.12. If Ω has the TBCPC at ξ¯0 ∈ ∂n Ω, then (3.27) is satisfied and ξ̄0 is regular.
Proof. If Ω satisfies the TBCPC at ξ¯0 , by Lemma 3.8, there exists a ∈ (0, 1/2) so that
a ≈n 1 and
Z 1 Z
Cap(Ra− (ξ̄0 ; r) ∩ Ωc ) dr X 2rj Cap(Ra− (ξ̄0 ; r) ∩ Ωc ) X
≥ & 1 = ∞,
0 rn r rj rn
j≥1 j≥1
where rj → 0 is the sequence in the definition of TBCPC. Thus, (3.27) holds and ξ̄0 is
regular.
We will now introduce a class of regular domains that has played an important role in
(free) boundary value problems for harmonic and elliptic measure.
Let ξ¯0 ∈ SΩ. If there exists a ∈ (0, 1] and c > 0 such that
Cap(Ra− (ξ̄0 ; r) ∩ Ωc ) p
(3.28) ≥ c, for all 0 < r < (t0 − Tmin )/2,
Cap(Ra− (ξ̄0 ; r))
then, we say that Ω has the time backwards cylindrical capacity density condition at the
point ξ̄0 . Although the TBCDC looks more general, because of Lemmas 3.10 and 3.8, we
can see that the two conditions are in fact equivalent. Note that, by Lemma 3.11 and Remark
3.13, if Ω has the TBCDC at the point ξ¯0 ∈ SΩ, then ξ¯0 is a regular point and belongs to
∂n Ω.
Remark 3.13. If Ω satisfies the TBCDC, then it clearly holds that ∂s Ω∩SΩ = ∂ss Ω∩SΩ =
∅ and
(3.29) SΩ = ∂n Ω ∩ SΩ = ∂e Ω ∩ SΩ.
Moreover, the range of r in (3.28) is chosen so that, if x̄ = (x, t) ∈ SΩ and ȳ ∈ Cr (x̄) ∩ Ω,
then
(3.30) distp (ȳ, SΩ) = distp (ȳ, ∂e Ω).
√
Indeed, if we choose r > 0 so that r < t − Tmin − r 2 , we have
p
distp (ȳ, SΩ) ≤ distp (ȳ, x̄) < r < t − Tmin − r 2
= distp (BCr (x̄), BΩ|Tmin ) ≤ distp (ȳ, BΩ|Tmin ).
− (ξ̄ ) = C (ξ , t − (ar)2 ), we obtain the range of r in (3.28).
Recalling that Rar 0 r 0 0
A crucial property of our definitions of TBCDC is their invariance under parabolic scal-
ing.
Lemma 3.14. Let Ω be a domain that satisfies the TBCDC with constants a, c as in (3.28).
Let ξ̄ = (ξ, τ ) ∈ SΩ, ρ > 0 and denote Ω̃ := Tξ,ρ [Ω]. Then Ω̃ satisfies the TBCDC with the
same parameters.
e and every r > 0 we have that
Proof. Let us observe that for ζ̄ ∈ S Ω

−
Cr/ρ e c = Tξ,ρ C − (T −1 ζ̄) ∩ Ωc
(ζ̄) ∩ Ω r ξ,ρ
and Γ(δρ x̄) = ρ−n Γ(x̄). We claim that
(3.31) −
Cap(Cr/ρ e c ) = ρ−n Cap(Cr− (T −1 ζ̄) ∩ Ωc ).
(ζ̄) ∩ Ω ξ̄,ρ
Set K = Cr− (Tξ̄,ρ

−1
ζ̄) ∩ Ωc and K e c , and let µ be the unique Radon measure
e = C − (ζ̄) ∩ Ω
r/ρ
e such that µ(K)
supported in K e = Cap(K).
e If µe = T −1 µ, then it is clear that supp µ
e ⊂ K.
ξ̄,ρ ♯
Moreover, since Gµ ≤ 1 in Rn+1 , it holds that

Z Z
−1

Ge µ(x̄) = Γ x̄ − Tξ̄,ρ ȳ dµ(ȳ) = ρ−n Γ Tξ̄,ρ x̄ − ȳ dµ(ȳ) ≤ ρ−n .
Therefore, the measure ρn µ

e is an admissible measure for Cap(K) and thus,
e = ρn µ(K)
ρn Cap(K) e = ρn µ
e(K) ≤ Cap(K).
The proof of the converse inequality is similar and we omit it. This proves (3.31), which,
−
in turn, is valid if we replace the backwards in time cylinders Cr/ρ by Ra− (ζ̄; r/ρ), the
truncated cylinders defined in (3.25). Thus, the TBCDC for Ω e readily follows from the
TBCDC for Ω.
The next lemma was proved in [GH20, Lemma 2.2] in the particular case of domains
with Ahlfors-David regular boundaries that, in addition, satisfy the time-backwards Ahlfors-
David regular condition. In light of (3.15), those domains satisfy the TBCDC.
Lemma 3.15. Let ω x̄ be the caloric measure in Ω with pole at x̄ = (x, t) ∈ Ω. If ξ̄0 =
(ξ0 , t0 ) ∈ PΩ ∩ SΩ, r > 0, and a ∈ (0, 1/2), we define

(3.32) Rb+ (ξ̄0 ; r) = B(ξ0 , r) × t0 − (ar)2 /2, t0 + r 2 .
a
ba+ (ξ̄0 ; r) ∩ Ω,
Then there exists M0 > 1 depending on n and a, such that for any x̄ ∈ R
Cap(R− (ξ̄ ; r) ∩ Ωc )
x̄ 2 a 0
(3.33) ω CM0 r (ξ̄0 ) ∩ {|t − t0 | < r } & ,
rn
where the implicit constant depends only on n and a. If Ω satisfies the TBCDC at ξ¯0 , then
there exists a ∈ (0, 1/2) and c = c(n, a) > 0 such that
(3.34)
p
ba+ (ξ̄0 ; r/M0 ) ∩ Ω, and 0 < r < M0 (t0 − Tmin )/2.
ω x̄ (Cr (ξ̄0 )) ≥ c, for x̄ ∈ R
Proof. Without loss of generality, we may assume that ξ¯0 = 0̄ ∈ SΩ and set K =
Ra− (0̄; r) ∩ Ωc . For notational convenience, we will just write Ra,r ba,r
− and R + . We also
−
assume that Cap(Ra,r ∩ Ωc ) > 0 since otherwise (3.33) is trivially true.
By the definition of capacity on compact sets, there exists a unique positive Radon mea-
+
sure µ supported on K such that kGµkL∞ ≤ 1 and kµk = Cap(K). Since for any x̄ ∈ Rar
−
and ȳ = (y, s) ∈ Ra,r , it holds that
a2 r 2 /2 ≤ t − s ≤ 2r 2 and |x − y| ≤ 2r,
we infer that
2
Γ(x̄ − ȳ) ≥ (8πr 2 )−n/2 e−2/a =: c1 r −n ,
which, in turn, implies that
Gµ(x̄) ≥ c1 kµkr −n , ba,r
for any x̄ ∈ R +
.
Moreover, if x̄ ∈ Rn+1 \ CM0 r and (y, s) ∈ Ra,r
− , we have that
kx̄ − ȳk ≥ kx̄k − kȳk ≥ M0 r − r =: (M0 − 1)r,

and thus, by (3.2),

√
Gµ(x̄) ≤ ((M0 − 1) π)−n kµkr −n , for any x̄ ∈ Rn+1 \ CM0 r .
√ √
Define u = Gµ−((M0 −1) π)−n kµkr −n and choose M0 so that 2((M0 −1) π)−n ≤ c1 .
Then, since kGµkL∞ ≤ 1, the following hold:
(1) u is continuous in Rn+1 and Hu = 0 in Ω,
(2) u ≤ 1 in Rn+1 ,
(3) u ≤ 0 in [Rn+1 \ CM0 r ] ∪ [Rn × {t ≤ −r 2 }],
(4) u ≥ c1 kµkr −n in Rba,r
+
.
2
Since F := CM0 r ∩ [Rn × {t > −r 2 }] ∩ ∂e Ω is Borel, we have that χF is resolutive,
u ∈ LχF , and ω x̄ (F ) = SχF (x̄). Thus, u(x̄) ≤ ω x̄ (F ) for all x̄ ∈ Ω which, by the item (4)
above and the fact that supp ω x̄ ⊂ ∂e Ω ∩ [Rn × {t < r 2 }] for any x̄ ∈ R ba,r
+
, proves (3.33).
It is straightforward to see that, by the definition of TBCDC, (3.34) follows from the latter
estimate.
Remark 3.16. Lemma 3.15 holds for the adjoint caloric measure if we replace the cylinders
ba+ (ξ̄0 , r) with their reflections across the hyperplane passing through their
Ra− (ξ̄0 , r) and R
centers orthogonal to the time axis.
If Ω is regular there is the above lemma has a much easier proof based on the reduction
function. Also, the dilation factor M0 need not be taken large enough and just M0 = 2 does
the job.
Lemma 3.17. Let Ω ⊂ Rn+1 be open, ξ¯0 ∈ PΩ ∩ SΩ, a ∈ (0, 1/2), β ≥ 2, and K =
b1 (K, Cβr (ξ̄0 )), then for any x̄ ∈ R
Ra− (ξ̄0 ; r) ∩ Ωc . If we set u = R ba+ (ξ̄0 ; r) ∩ Ω,

(3.35) u(x̄) & Cap K, Cβr (ξ̄0 ) /r n & Cap(K)/r n .
If w is a non-negative subtemperature in Cβr (ξ̄0 ) so that w = 0 on K and w ≤ 1 in
Cβr (ξ̄0 ), there exists κ ∈ (0, 1) such that

(3.36) w(x̄) ≤ 1−κ Cap K, Cβr (ξ̄0 ) /r n ≤ 1−κCap(K)/r n , for x̄ ∈ R ba+ (ξ̄0 ; r)∩Ω.
If SΩ ∩ Cβr (ξ̄0 ) ⊂ PΩ consists of regular points, then (3.34) holds with M0 = 2.

Proof. As in the proof of the previous lemma, we assume that ξ¯0 = 0̄ and use the notation
− and R
Ra,r b+ . Moreover, we assume that Cap(K) > 0 since otherwise (3.35) is trivial. We
a,r
also denote
e− = B(0, (1 + 0.25a)r) × (−(1 + 0.25a2 )r 2 , −0.75a2 r 2 )
R a,r
ea,r
and let ϕ ∈ Cc∞ (R − ) such that ϕ = 1 in R− and |H ∗ ϕ| .
a,r
−2
a,n r . Then, if µ is the
unique Radon measure such that µ(K) = Cap(K), it holds that
Z Z Z
−n −n −n −n ∗ −(n+2)
r Cap(K) = r µ(K) ≤ r ϕ dµ = r uH ϕ . r u .a,n inf u,
e−
Ra,r b+
Ra,r
where in the last step we used the weak Harnack inequality since u is a non-negative su-
pertemperature and thus, a non-negative supercaloric function (see [Lieb96, Corollary 6.24,
p. 128]).
By the definition of u, we have that for any non-negative supertemperature v in Cβr (ξ̄0 )
such that v ≥ 1 on K, it holds that v ≥ u and thus, (3.35) is true for v as well. Now, if w is
a subtemperature in Cβr (ξ̄0 ) so that w = 0 on K and w ≤ 1 in C2r (ξ̄0 ), then 1 − w is a non-
negative supertemperature in Cβr (ξ̄0 ) that is identically 1 on K and (3.36) readily follows.
Finally, by the regularity of ∂Ω ∩ Cβr (ξ̄0 ), we may extend ω x̄ (Cβr (ξ̄0 )) by 1 in Cβr (ξ̄0 ) \ Ω
and use [Wa12, Theorem 7.20] to show that it is a non-negative supertemperature in Cβr (ξ̄0 )
that is 1 on K and non-neagtive in Cβr (ξ̄0 ) and use (3.35) to show (3.34) for M0 = 2.
Remark 3.18. One can show that (3.36) holds for upper semicontinuous weakly subcaloric
functions as well. Indeed, if we follow the proof of Lemma 3.15 substituting Γ(·, ·) with the
Green function GCβr (ξ̄0 ) (·, ·) and using (3.8) together with the weak minimum principle3
(instead of the resolutivity of characteristic functions of Borel sets) we can show that if v is
a non-negative supercaloric function in Ω ∩ CβM r (ξ̄0 ) such that lim inf x̄→ξ̄ v ≥ 1 for every
ξ̄ ∈ ∂n Ω ∩ CβM r (ξ̄0 ), then v satisfies (3.35). The rest of the proof is the same as before and
we skip the details.
√
Lemma 3.19. Let Ω ⊂ Rn+1 be open, ξ¯0 ∈ PΩ ∩ SΩ, a ∈ (0, 1/2), a1 = a/ 2,
and β = a−1 1 . Let w be a non-negative subtemperature in Cβr (ξ̄0 ) so that w = 0 on
−
Ra (ξ̄0 ; r) ∩ Ωc and w ≤ M1 in Cβr (ξ̄0 ). If x̄ ∈ Cρ (ξ̄0 ) ∩ Ω for some 0 < ρ ≤ r, then
Z r !
Cap(Ra− (ξ̄0 ; s) ∩ Ωc ) ds
(3.37) w(x̄) ≤ exp −c sup w,
ρ sn s Ω∩Cβr (ξ̄0 )
for some constant c > 0 depending on n and a.

Proof. We assume that Cap Ra− (ξ̄0 ; ρ) ∩ Ωc > 0, since otherwise (3.38) is trivial. We set
ak = ak1 for k ≥ 0, and
Cap(Ra− (ξ̄0 ; a2k r) ∩ Ωc )
θk := ,
(βa2k r)n
and note that ak−1 = βak . By (3.36), it holds that
w(x̄) ≤ (1 − κθ1 ) sup w ≤ exp(−κθ1 ) sup w, ba+ (ξ̄0 ; r).
for x̄ ∈ R
Cβr (ξ̄0 ) Cβr (ξ̄0 )
In particular, the latter inequality holds in Ca1 r (ξ̄0 ). We apply (3.36) once again in Ca2 r (ξ̄0 )
and get that
w(x̄) ≤ exp(−κθ2 ) sup w ≤ exp(−κ(θ1 + θ2 )) sup w, ba+ (ξ̄0 ; a2 r).
for x̄ ∈ R
Ca1 r (ξ̄0 ) Cβr (ξ̄0 )
The latter inequality holds in Ca3 r (ξ̄0 ) and we may apply (3.36) in Ca4 r (ξ̄0 ). By iteration,
if M is an integer such that a2M +2 ≤ ρ ≤ a2M , we get that
XM
w(x̄) ≤ exp − κ θk sup w, for x̄ ∈ R ba+ (ξ̄0 ; a2M r).
k=0 Cβr (ξ̄0 )
3The weak minimum principle still holds for lower semicontinuous (instead of continuous in Ω) supercaloric
functions in the form lim inf x̄→P(Ω) v ≤ inf Ω v. See e.g. the proof of [Lieb96, Corollary 6.26].
Therefore, since
Z r M Z
Cap(Ra− (ξ̄0 ; s) ∩ Ωc ) ds X a2k Cap(Ra− (ξ̄0 ; s) ∩ Ωc ) ds
≤
ρ sn s a2k+2 sn s
k=0
M Z a2k
X X M
Cap(Ra− (ξ̄0 ; a2k r) ∩ Ωc ) ds 3n+2
≤ ≤ β θk ,
a2k+2 an2k+2 a2k+2
k=0 k=0
if c = a13n+2 κ,
we obtain that
Z !
r
Cap(Ra− (ξ̄0 ; s) ∩ Ωc ) ds
(3.38) w(x̄) ≤ exp −c sup w, for x̄ ∈ Cρ (ξ̄0 ).
ρ sn s Cβr (ξ̄0 )
Given T > 0, we define E(T ) := {(x, t) ∈ Rn+1 : t < T }. Moreover, for an open set
Ω, x̄ ∈ ∂Ω and r > 0 we define Ωr := Ωr (x̄) := Ω ∩ Cr (x̄) and Ωr (T ) := Ωr ∩ E(T ).
n+1 be an open set satisfying the TBCDC, and let ξ̄ ∈ SΩ and
Lemma 3.20.p Let Ω ⊂ R 0
0 < r < (t0 − Tmin )/2. Then for any non-negative function u, which is either weakly
subcaloric or subtemperature in Ω2r (T1 ) vanishing continuously on C2r (ξ̄0 )∩∂e Ω∩E(T1 ),
it holds that
(3.39) u(ȳ) . (dist(ȳ, ∂e Ω)/r)α sup u, for every ȳ ∈ Ωr (T1 ),
Ω3r/2 (T1 )
where T1 = Tmax (Cr (ξ̄0 )) = t0 + r 2 .

Proof. For subtemperatures, this is a direct consequence of Lemma 3.19 (with a slightly
larger cylinder on the right hand-side of (3.39)), while for weakly subcaloric functions one
can follow the proof of [GH20, Lemma B.2] (which still works for TBCDC domains).
Lemma 3.21. Let Ω ⊂ Rn+1 be a quasi-regular open set for H and let M > 1. If ξ̄0 ∈ SΩ
and r > 0, for any x̄ ∈ Ω \ Cr (ξ̄0 ) and ȳ ∈ Ω ∩ Cr/4 (ξ̄0 ) we have that
(3.40) inf ω z̄ (CM r (ξ̄0 )) G(x̄, ȳ)r n . ω x̄ (CM r (ξ̄0 )).
z̄∈Cr (ξ̄0 )∩Ω
Moreover, if Ω satisfies either the TBCDC at ξ̄0 , then if M0 > 1 is the constant from Lemma
3.15, we have that for x̄ ∈ Ω \ Cr (ξ̄0 ) and ȳ ∈ Ω ∩ Cr/4 (ξ̄0 ),
p
(3.41) G(x̄, ȳ)r n . ω x̄ (CM0 r (ξ̄0 )), for 0 < r < M0 (t0 − Tmin )/2.
Proof. Fix ȳ = (y, s) ∈ Ω ∩ Cr/4 (ξ̄0 ) and recall that G(x̄, ȳ) = 0 for any x̄ 6∈ Λ∗ (ȳ, Ω). If
x̄ = (x, t) ∈ ∂e Cr (ξ̄0 ) ∩ Λ∗ (ȳ, Ω), then there exists c > 0 such that
rn
G(x̄, ȳ)r n . ≤ c,
kx̄ − ȳkn
and so, for any x̄ ∈ ∂e Cr (ξ̄0 ) ∩ Λ∗ (ȳ, Ω),
(3.42) inf ω z̄ (CM r (ξ̄0 )) G(x̄, ȳ)r n ≤ c inf ω z̄ (CM r (ξ̄0 ))
z̄∈Cr (ξ̄0 )∩Ω z̄∈Cr (ξ̄0 )∩Ω
≤ c ω x̄ (CM r (ξ̄0 )).

Define now
u(x̄) := c−1 inf ω z̄ (CM r (ξ̄0 ))G(x̄, ȳ)r n , v(x̄) := ω x̄ (CM r (ξ̄0 )).
z̄∈Cr (ξ̄0 )
If we set w(x̄) = u(x̄) − v(x̄), then it is clear that w is a caloric function, hence continuous,
in Ω \ Cr/2 (ξ̄0 ). Let E(ξ̄0 , r/2) be the ellipsoid that circumscribes Cr/2 (ξ̄0 ) lying inside
Cr (ξ̄0 ). If z̄∗ = (z∗ , t∗ ) is the point on the boundary of the ellipsoid such that t∗ < t for
every z̄ = (z, t) ∈ ∂E(ξ̄0 , r/2) \ {z̄∗ }, then
∂E(ξ̄0 , r/2) \ {z̄∗ } ⊂ ∂n (Rn+1 \ E(ξ̄0 , r/2)) and z̄∗ ∈ ∂ss (Rn+1 \ E(ξ̄0 , r/2)),
while it is clear by (3.21) and Remark 3.7, all the points of ∂E(ξ̄0 , r/2) are regular points
of ∂e (Rn+1 \ E(ξ̄0 , r/2)) for both H and H ∗ . Using (3.42) we obtain
lim w(ȳ) = w(ζ̄) ≤ 0,
Ω\E(ξ̄0 ,r/2)∋x̄→ζ̄
for any ζ̄ ∈ ∂E(ξ̄0 , r/2) ∩ Ω.

We set Z to be the set of irregular points of ∂e Ω, which is polar because of the quasi-
regularity assumption on Ω. Thus, if ζ̄ ∈ ∂n Ω \ Z (resp. ∂ss Ω \ Z), it holds that w(x̄) → 0
as Ω ∋ x̄ → ζ̄ (resp. Ω ∋ (x, t) → (ζ, τ + )). Moreover, it is clear that
lim sup w(x̄) < ∞.
x̄→ζ̄∈Z
Hence, by the maximum principle [Wa12, Theorem 8.2] in Ω \ E(ξ̄0 , r/2), we infer that
w ≤ 0 in Ω \ E(ξ̄0 , r/2) concluding the proof of the lemma.
If we assume the TBCDC, (3.41) readily follows from (3.40) using (3.34).
A pretty standard result in elliptic theory is that polar sets have zero harmonic measure.
The same is true for caloric measure as well. As we were not able to find an appropriate
reference, we will write the proof for completeness.
Proposition 3.22. Let Ω be an open set, x̄ ∈ Ω, and ω x̄ be the caloric measure in Ω. If
E ⊂ Rn+1 is polar, then ω x̄ (E) = 0.
Proof. Fix x̄ ∈ Ω. Since E ∩ ∂Ω is polar as a subset of the polar set E, without loss
of generality, we may assume that E ⊂ ∂Ω. As we have that Cap(E) = 0, by (3.12),
there exists a sequence of open sets Sj ⊃ E such that limj→∞ Cap− (Sj ) = 0. Set now
Ee := T Sj , and note that E e is Borel, E ⊂ E,
e and Cap(E)
e = 0. Therefore, without loss of
j
generality, we may assume that E is Borel.
Let us assume that Cap(Ωc ) > 0. Then, there exists an open set Ωe such that Ω ∪ E ⊂ Ω e
e c c
and Cap(Ω ) > 0. Indeed, as Ω \ E is Borel, by (3.12), there exists a compact set K ⊂
Ωc \ E so that Cap(K) > Cap(Ωc \ E)/2. We set Ω e = K c , which is an open set such that
e
Ω ∪ E ⊂ Ω, and, since Cap(E) = 0, it clearly satisfies Cap(Ω e c ) > 0. If Cap(Ωc ) = 0, we
simply assume that Ωe = Rn+1 .
Now, we apply [Wa12, Theorem 7.3] to find u ≥ 0, a supertemperature in Ω, e such that
u = ∞ on E and u(x̄) < +∞. Hence, λu|Ω belongs to the upper class UE for every λ > 0,
which implies that 0 ≤ ω x̄ (E) ≤ λu(x̄). We conclude the proof by taking λ → 0.
Lemma 3.23. Let Ω1 and Ω2 be two disjoint quasi-regular open sets in Rn+1 for H and
p̄i
H ∗ so that (PΩi )0 6= ∅, i = 1, 2, with caloric measures ωi = ωΩ i
for some p̄i ∈ Ωi and
p̄
suppose ω1 ≪ ω2 ≪ ω1 on a Borel set E ⊂ (PΩ1 )0 ∩ (PΩ2 )0 and ωΩ (E) > 0. Then there
are open subsets Ω e i ⊂ Ωi which are regular for H and H and contain p̄i for i = 1, 2 so
∗
e 1 )0 ∩ (P Ω
that there exists G0 ⊂ E ∩ (P Ω e 2 )0 with ω
ei (G0 ) > 0 upon which ω e2 ≪ ωe1 ≪ ωe2 .
Proof. We follow closely the proof of [AMTV19, Lemma 2.3]. Let Fi and Fi∗ be the sets
of irregular points for Ωi for H and H ∗ respectively, which are polar and co-polar sets
in Rn+1 . Although, since by [Wa12, Theorem 7.46] polar and co-polar sets coincide, by
[Wa12, Definition 7.1], there is a positive supertemperature vi on Ωi so that
lim vi (ȳ) = ∞ for all x̄ ∈ F := F1 ∪ F2 ∪ F1∗ ∪ F2∗ .
ȳ→x̄
ȳ∈Ωi
Let λ > 0. Since vi is supertemperature on Ωi , it is lower semicontinuous, and remains

so when we extend it by zero to Ωci . Thus, for each x̄ ∈ F there is a closed cylinder Ci′ (x)
centered at x̄ (not containing either p̄i ) such that vi ≥ λ on Ci′ (x̄) ∩ Ωi . For each Ci′ (x)
let Ci (x) be the cylinder of the same center and half the radius of Ci′ (x). Let {Cj } be a
Besicovitch subcovering (see Lemma 4.8) and if Ej is the closed ellipsoid of revolution
around the axis of the cylinder Cj′ that is inscribed in Cj′ , we define
[
e i = Ωi \ H ⊂ Ωi , where H =
Ω Ej .
j≥1
e i is open. Indeed, to show that Ω

Note that Ω e c is closed consider x̄k ∈ Ω e c , k ≥ 1 and
i i
x̄k → x̄. Then we need to show x̄ ∈ Ω e c . If there is a subsequence contained in Ωc , we are
i i
done. Otherwise, assume that x̄k ∈ H\Ωci = H ∩ Ωi . If x̄k ∈ Ej for infinitely many k,
then x̄ ∈ Ej and we are done since Ej is closed and Ej ⊂ Ω e c . Otherwise, suppose x̄k is not
i
in any Ej more than finitely many times. By the bounded overlap property, if j(x̄k ) is such
that x̄k ∈ Ej(x̄k ) , then diam(Ej(x̄k ) ) ↓ 0 as k → ∞, and since the ellipsoids are centered
on F ⊂ Ωci , we have that x̄ ∈ Ωci ⊂ Ω e c , and we are done. Thus, Ω e i is open.
i
Set Vj = R n+1 1 2
\ Ej and note that if z̄j = (z1 , t1 ) and z̄j = (z2 , t2 ), are the points
on ∂Ej ∩ ℓj , where ℓj is the line containing the axis of Cj , so that t1 < t2 , it holds that
z̄1 ∈ ∂ss Vj and every x̄ ∈ ∂Ej \ {z̄j1 } ⊂ PVj is in ∂n Vj . Moreover, it is clear by (3.21) and
Remark 3.7 that every x̄ ∈ ∂Ej is regular for Vj . Therefore, by [Wa12, Corollary 8.47],
it holds that ∂Ej ∩ Ωi consists of regular points of ∂e Ω e i . As the points of ∂e Ωi ∩ ∂e Ω ei
are regular for Ωi then, by another application of [Wa12, Corollary 8.47], they are regular
for Ωe i as well. Therefore, Ω e i is regular for H and H ∗ , and E ⊂ ∂e Ω e i . In fact, since
E ⊂ (PΩ1 )0 ∩ (PΩ2 )0 and ∂a Ω e i ∩ ∂Ωi ⊂ ∂a Ωi , then it holds that
ei ⊂ PΩ
E ∩ ∂e Ω e1 ∩ PΩ
e 2 ∩ (PΩ1 )0 ∩ (PΩ2 )0 .
ei = ω p̄ei and G = E\H. As the cylinders are closed, it is not hard to see that
Let ω
Ωi
G ⊂ (P Ω e 1 )0 ∩ (P Ω
e 2 )0 . Since vi is positive in Ωi and ω x (H) ≤ 1 ≤ vi (x̄)/λ for all
i
x̄ ∈ H, by the maximum principle on Ω e i , we have that ωi (H) ≤ λ−1 vi (p̄i ). Picking
1
λ−1 < min {ωi (E)/vi (p̄i )},
2 i=1,2
this gives ωi (H) ≤ 21 ωi (E) and hence ωi (G) > 0. Similarly, by the maximum principle,
since vi /λ ≥ 1 on H, ωei (H) ≤ λ−1 vi (p̄i ). Picking
1
λ−1 < min {ωi (G)/vi (p̄i )},
2 i=1,2
this gives
1
ei (H) ≤ ωi (G).
ω
2
Moreover, by the maximum principle, and since Ω e i is a regular domain,
ei (H c ∩ Gc ) ≤ ωi (H c ∩ Gc ).
ω
Thus,
ei (Gc ) = 1 − ω
ei (G) = 1 − ω
ω ei (H ∩ Gc ) − ω
ei (H c ∩ Gc )
1 1 1
≥ 1 − ωi (G) − ωi (H c ∩ Gc ) ≥ ωi (G) − ωi (G) = ωi (G) > 0.
2 2 2
e1 ≪ ω1 on G by the maximum principle and since ω
Note that ω e1 (G) > 0, it is not hard
to show using the Lebesgue decomposition theorem that there is G1 ⊂ G of full ω e1 -measure
upon which we also have ω1 ≪ ω e1 . Hence ω1 (G1 ) > 0, which implies ω2 (G1 ) > 0. The
same reasoning gives us a set G2 ⊂ G1 upon which ω e 2 ≪ ω2 ≪ ω e2 ≪ ω
e2 . Thus, ω e1 ≪ ω e2
on G2 .

The notion of halving metric space was first introduced by Korey [Kor98] and plays an
important role in the proofs of our theorems.
Definition 3.24. A probability space (X, ω) is called halving if for every subset E ⊂ X
such that ω(E) > 0, there exists F ⊂ E such that ω(F ) = ω(E)/2.
We conclude this section by proving the caloric analogue of [AM19, Lemma 7.9].
Lemma 3.25. Let Ω ⊂ Rn+1 be an open set and x̄ ∈ Ω. Then, the probability space
(∂e Ω, ω x̄ ) is halving.
Proof. Set ω := ωΩ x̄ and let us assume that there exists E ⊂ ∂Ω such that ω(E) > 0 and
ω(F ) 6= ω(E)/2 for every F ⊂ E. Given s ∈ R and v ∈ Sn , we define the half-space

Hs,v := {ξ̄ ∈ Rn+1 : ξ̄ · v ≥ s}. By the mean value theorem and our assumption, the map
s 7→ ω(Hs,v ∩ E) is not continuous for any v ∈ Sn . In particular, for any v ∈ Sn , there
exists sv , such that ω(∂Hsv ,v ∩ E) > 0. Set now Vv := ∂Hsv ,v , which is an n-dimensional
plane. Define
S := {(y ′ , yn , τ ) ∈ Sn : y ′ = 0},
and observe that, since S is uncountable, there is ε > 0 such that ω(Vv ∩ E) > ε for all v in
some uncountable subset A of S. Let us consider A′ ⊂ A countable. For every v, v ′ ∈ A′
we have that Vv ∩Vv′ is an (n−1)-plane orthogonal to the t axis, which is a polar set since it
is contained in a horizontal hyperplane and has n-dimensional Lebesgue measure zero (see
[Wa12, Theorem 7.55]). Therefore, by Proposition 3.22, it holds that ω(Vv ∩ Vv′ ) = 0. Set
now [
Wu := Vu \ Vv ,
v∈A′ ,v6=u
and notice that ω(Wu ∩ E) = ω(Vu ∩ E) > ε and Wu ∩ Wu′ = ∅ for any u 6= u′ ∈ A′ .
Thus, since A′ is countable,
X
1 ≥ ω(E) ≥ ω(Wu ∩ E) = +∞,
u∈A′
which is a contradiction.
Every result related to the heat equation we have stated so far has a dual for the adjoint
heat equation H ∗ u = 0 obtained by reversing the sign of the time variable. Therefore,
we can define the associated fundamental solution Γ∗ , the Green function G∗Ω , the adjoint
caloric measure denoted by ω∗p̄ , the associated parabolic capacity Cap∗ and so forth. A
solution of H ∗ u = 0 is called adjoint caloric. Remark that, by [Wa12, Theorem 6.10], we
have that
Γ(x̄, ȳ) = Γ∗ (ȳ, x̄) and GΩ (x̄, ȳ) = G∗Ω (ȳ, x̄).
For the regularity of the ∂e∗ Ω for H ∗ and the corresponding capacity density conditions, it is
pretty clear that we should just take time-forwards cylinders and the so-called co-heat balls,
which are defined as the heat balls using the adjoint heat kernel.
4. H AUSDORFF AND TANGENT MEASURES
If d ≤ n + 1 is an integrer and Hd (E) = 0 (resp. Hd (E) < ∞), then Hpd+1 (E) = 0
(resp, Hpd+1 (E) < ∞). For a proof see [He17, Lemma 3.2]. Note that this was originally
stated under the additional hypothesis that E is Euclidean d-rectifiable but it is easy to see it
is redundant. Moreover, by [He17, Lemma 3.8], it holds that there exists c1 > 0 and c2 > 0
depending only on d, such that on Rd−1 × R ⊂ Rn × R,
Hpd+1 = c2 Ld = c1 Hd−1 × Hp2 .
We recall that the Hausdorff dimension of a Borel measure ω is defined by

dimHp (ω) = inf dimHp (Z) : ω(Rn+1 \ Z) = 0 .
This definition is related with the concept of pointwise dimension of ω at x̄ ∈ supp ω. More
specifically, if
log µ(Cr (x̄)) log µ(Cr (x̄))
dµ (x̄) = lim inf and dµ (x̄) = lim sup
r→0 log r r→0 log r
denote the lower and upper pointwise dimension respectively, we can argue as in [BW06]
to show that
dimHp (ω) = ess sup{dµ (x̄) : x̄ ∈ supp µ}.
If dµ (x̄) = dµ (x̄), we denote the common value by dµ (x̄).
Definition 4.1. If ℓ ∈ [ 21 , 1], we say that a function ψ : Rd → R is (1, ℓ)-Lipschitz and write
f ∈ Lip1,ℓ (Rd ) if there exists a constant L > 0 such that

(4.1) |ψ(x′ , t)−ψ(y ′ , s)| ≤ L max |x′ −y ′ |, |t−s|ℓ , for (x′ , t) 6= (y ′ , s) ∈ Rd−1 ×R.
We call L the Lipschitz constant of f and write Lip(f ) = L.
Observe that for ℓ = 1/2 these functions correspond to Lipschitz functions with respect
to the parabolic norm k · k or the equivalent norm |x′ − y ′ | + |t − s|1/2 . For ℓ = 1, those
are just the usual Lipschitz functions with respect to the Euclidean norm.
Definition 4.2. We say that Γ ⊂ Rn+1 is an admissible d-dimensional graph if there exists a
vector field f : Rd → Rn−d+1 such that, possibly after a rotation in space and a translation,

Γ = Γf = x′ , f (x′ , t), t) : x ∈ Rd−1 , t ∈ R .
If f ∈ C m (Rd ; Rn−d+1 ), 1 ≤ m ≤ ∞, we say that Γf is an admissible d-dimensional C m -
graph. If f is an affine map, then Γf = V is a plane that contains a line parallel to the time-
axis and we call it an admissible d-dimensional plane. In fact, after rotation in space and a
translation, we can always assume V = Rd−1 × {~0} × R, where ~0 = (0, . . . , 0) ∈ Rn−d+1 .
Definition 4.3. A closed set E ⊂ Rn+1 is Euclidean d-rectifiable if there exists Ei ⊂ R
d
and Lip(1,1) -functions fi : Ei → R n+1 d ∞

so that H E \ ∪i=1 fi (Ei ) = 0. Equivalently, E
is Euclidean d-rectifiable if there exists a countable collection of d-dimensional Lip(1,1) -

graphs (or d-dimensional C 1 -manifolds) {Γi }i≥1 such that Hd E \ ∪∞ i=1 Γi = 0.
By the co-area formula [He17, Theorem D], if E is Euclidean d-rectifiable, Hd |E is
locally finite, and σ indicates its associated surface measure as defined in Section 2, then
(4.2) σ = cHpd+1 |E .
In fact, using the Rademacher theorem in [Or19], one can show that this theorem is true
for parabollically rectifiable sets as well, that is, for sets that are exhausted, up to a set of
Hpd+1 -measure zero, by admissible Lip1, 1 -graphs with 21 -derivative in time in BMO.4
2
If µ and ν are Radon measures on Rn+1 , following [Pr87] we define the distance between
µ and ν in the parabolic ball Cr by
Z
(4.3) dCr (µ, ν) = sup f d(µ − ν),
f
where the supremum is taken over all functions f ∈ Lip(1, 1 ) (Rn+1 ) which are supported
2
in Cr and satisfy Lip(f ) ≤ 1. If a sequence of Radon measures µj converges weakly to a
Radon measure µ, we use the notation µj ⇀ µ.
For r > 0 and x̄, ȳ ∈ Rn+1 , set
δr (x̄) := (rx, r 2 t), and Tȳ,r (x̄) := δ1/r (x̄ − ȳ).
If µ and ν are Radon measures in Rn+1 , we define
−1

Tȳ,r [µ](A) := µ(δr (A) + ȳ) = µ Tȳ,r (A) , A ⊂ Rn+1 .
and Z
Fr (µ, ν) := sup f d(µ − ν), for r > 0,
f
where the supremum is taken over all functions f ∈ Lip(1, 1 ) (Rn+1 ) which are supported in
2
Cr and satisfy Lip(f ) ≤ 1. By density, it is enough to consider the supremum in the class
of Cc∞ (Cr ) functions such that Lip(f ) ≤ 1.
4We omit the detailed definitions and the proof since we will not be dealing with such general sets in the
present paper.
We also define Fr (µ) := Fr (µ, 0̄). A standard argument shows that

Z Z r
n+1
Fr (µ) = distp (x̄, R \ Cr ) dµ(x̄) = µ(Cs ) ds.
0
As it is easy to see that

distp x̄, Rn+1 \ Cr = r − kx̄k + ,
where (·)+ stands for the positive part of a function, we infer that
Z

Fr (µ) = r − kx̄k + dµ(x̄).
Definition 4.4 (d-cone). A set of Radon measures M is a d-cone if c T0̄,r [µ] ∈ M for all
µ ∈ M, c, r > 0. Given a d-cone M, the set {µ ∈ M : F1 (µ) = 1} is referred to as its
basis. We say that M has closed (resp. compact) basis if its basis is closed (resp. compact)
with respect to the weak topology of the space of Radon measures.
For a d-cone M, r > 0, and a Radon measure µ such that Fr (µ) ∈ (0, ∞), we define
n µ o
(4.4) dr (µ, M) := inf Fr , ν : ν ∈ M, Fr (ν) = 1 .
Fr (µ)
The next lemma collects some of the relevant properties of Fr and dr (·, M). For more
details, see [KPT09, Section 2] and the references therein.
Lemma 4.5. Let µ, ν be Radon measures in Rn+1 , ξ¯ ∈ Rn+1 and r > 0. The following
properties hold:

(1) Fr (µ) = rF1 T0̄,r [µ] .
(2) 2r µ(Cr/2 ) ≤ Fr (µ) ≤ rµ(Cr ).
(3) µj ⇀ µ if and only if Fr (µj , µ) → 0 for all r > 0.
(4) dr (µ, M) ≤ 1 and dr (µ, M) = d1 (T0̄,r [µ], M).
(5) if µj ⇀ µ and Fr (µ) > 0, then dr (µj , M) → dr (µ, M).
Definition 4.6. We say that ν is a tangent measure of µ at a point x̄ ∈ Rn+1 if ν is a non-
zero Radon measure on Rn+1 and there are sequences ci > 0 and ri ց 0 so that ci Tx̄,ri [µ]
converges weakly to ν as i → ∞ and write ν ∈ Tan(µ, x̄).
Remark 4.7. A Besicovitch covering theorem for parabolic balls in Rn+1 was proved in
[It18, Theorem 1.1]. This is an important tool for parabolic geometric measure theory. In
particular, one can show that Radon measures in Rn+1 satisfy the Lebesgue density theo-
rems and the Lebesgue differentiation theorems with respect to parabolic balls, as reported
in the next lemma.
Lemma 4.8. Let µ be a Radon measure on Rn+1 . If f : Rn+1 → R ∪ {∞} is locally
µ-integrable, then
Z
1
f (x̄) = lim f dµ for µ-a.e x̄ ∈ Rn+1 .
r→0 µ(Cr (x̄)) Cr (x̄)
Furthermore, if E ⊂ Rn+1 is µ-measurable, then the limit

µ(E ∩ Cr (x̄))
lim
r→0 µ(Cr (x̄))
exists and equals 1 for µ-a.e. x̄ ∈ E and 0 for µ-a.e. x̄ ∈ Rn+1 \ E.
Proof. The proof follows from the argument in [Mat95, Corollary 2.14] using the Besicov-
itch theorem for parabolic balls in [It18].
Remark 4.9. Once we have made the appropriate modifications in the definitions of the
blow-up mappings to reflect the parabolic dilation, all the theorems related to tangent mea-
sures that are required to obtain our results hold with the same proofs as in the Euclidean
setting.
If µ is a Radon measure, s ∈ [0, ∞), and x̄ ∈ Rn+1 , we define the lower and upper
s-density of µ at x̄ as
µ(Cr (x̄)) µ(Cr (x̄))
Θsµ,∗ (x̄) := lim inf and Θs,∗
µ (x̄) := lim sup .
r→0 rs r→0 rs
A measure µ is asymptotically doubling at x̄ ∈ Rn+1 if
µ(C2r (x̄))
lim sup < ∞.
r→0 µ(Cr (x̄))
s,∗
Remark that if 0 < Θsµ,∗ (x̄) ≤ Θµ (x̄) < ∞, then µ is asymptotically doubling at x̄ since
µ(C2r (x̄)) Θs,∗
µ (x̄)
lim sup = 2s s < ∞.
r→0 µ(Cr (x̄)) Θµ,∗ (x̄)
Lemma 4.10. If µ is a Radon measure on Rn+1 and x̄ ∈ Rn+1 , then the following hold:
(1) If ν ∈ Tan(µ, x̄), there exists {ri }i≥1 decreasing to 0 and ρ, c > 0 so that
Tx̄,ri [µ]
⇀ c T0̄,ρ [ν] and c T0̄,ρ [ν](C1 (0̄)) > 0.
µ(B(x̄, ri ))
(2) If, additionally, µ is asymptotically doubling at x̄ ∈ Rn+1 , then for any ν ∈
Tan(µ, x̄) there exists c1 > 0 and a sequence {ri }ı≥1 decreasing to 0 such that
Tx̄,ri [µ]
c1 ⇀ ν.
µ(B(x̄, ri ))
In this case, 0̄ ∈ supp ν for all ν ∈ Tan(µ, x̄).
(3) If, additionally, 0 < Θsµ,∗ (x̄) ≤ Θs,∗
µ (x̄) < ∞, then for any ν ∈ Tan(µ, x̄) there
exists c1 > 0 and a sequence {ri }ı≥1 decreasing to 0 such that
Tx̄,ri [µ]
c2 ⇀ ν.
ris
Proof. For a proof see [Mat95, Theorem 14.3] and [Mat95, Remarks 14.4 (1)-(4)].
As a result of Lemma 4.8 we obtain the following localization property of tangent mea-
sures.
Lemma 4.11. Let µ be a Radon measure in Rn+1 and f ∈ L1 (µ) a non-negative Borel
function. Then, for µ-a.e. x̄ ∈ Rn+1 , it holds that Tan(f µ, x̄) = f (x̄) Tan(µ, x̄).
Proof. With Lemma 4.8 at our disposal, we just follow the proof of [DeL08, Lemma 3.12].

Lemma 4.12 (see [Mat95], Theorem 14.16). Let µ be a Radon measure on Rn+1 . For
µ-a.e. x̄ ∈ Rn+1 , if ν ∈ Tan(µ, x̄), then the following properties hold:
(1) Tȳ,r [ν] ∈ Tan(µ, x̄) for all ȳ ∈ supp ν and r > 0.
(2) Tan(ν, ȳ) ⊂ Tan(µ, x̄) for all ȳ ∈ supp ν.
Lemma 4.13. Let Γ ⊂ Rn+1 be an admissible d-dimensional Lip(1,1) -graph5 and let µ =
Hpd+1 |Γ . Then, for µ-a.e. x̄ ∈ Γ, there exists a positive constant cx̄ and an admissible
d-plane Vx̄ passing through the origin such that
(4.5) r −d−1 Tx̄,r [µ] ⇀ Hpd+1 |Vx̄ as r → 0.
Proof. By the parabolic Rademacher theorem in [Or19], one can show that a parabolic
Lipschitz vector field ψ : Rd → Rn+1−d satisfies
|ψ(ȳ) − ψ(x̄) − Ax̄ (y − x)|
(4.6) = ǫx̄ (|ȳ − x̄|),
kȳ − x̄k
where ǫx̄ (r) → 0 as r → 0. In fact, that theorem is only stated for d = n and globally
parabolic Lipschitz functions (see e.g. [Or19, Definition 3.1]) but a local hypothesis is
enough. To be precise, it is sufficient for the functions to be locally in Lip(1,1/2) and to
satisfy [Or19, display (3.4)] locally (see [Or19, Remark 3.6]). Moreover, one can apply
(4.6) to each component of ψ and obtain the result above. Note that Ax̄ : Rd → Rn+1−d is
a horizontal linear map. In fact, it is the horizontal differential for the Lipschitz vector field
ψ(·, t) : Rd−1 → Rn+1−d at x̄. Using this map we can construct the approximating admis-
sible d-plane passing through the origin. Namely, since any admissible Lip(1,1) -function
satisfies the aforementioned conditions, if we set Vx̄ = (y ′ , Ax̄ (y ′ ), s) and follow the proof
in [DeL08, pp. 38-39] using the parabolic area formula [He17, Theorem 3.64], we can
prove (4.5). We skip the details.
S
Corollary 4.14. Let E ⊂ Rn+1 be such that Hd |E is locally finite and σ(E\ j≥1 Γj ) = 0,
where Γj are admissible d-dimensional Lip(1,1) -graphs and σ is the surface measure on
E. Then, for σ-a.e. x̄ ∈ Rn+1 , there exists an admissible d-dimensional plane Vx̄ passing
through the origin, such that
Tan(σ, x̄) = {cHpd+1 |Vx̄ : c > 0}.
Proof. By Lemma 4.11, we have that for Hpd+1 -a.e. x̄ ∈ Γj ∩ E, j ≥ 1, it holds that
Tan(Hpd+1 |E , x̄) = Tan(Hpd+1 |Γj ∩E , x̄) = Tan(Hpd+1 |Γj , x̄).
Observe that Hpd+1 |Γj (Cr (x̄)) ≈ r d+1 for any x̄ ∈ Γj and r > 0 (the implicit con-
stant depends on the Lipschitz character of Γj ). Thus, Lemma 4.13 implies that if ν ∈
Tan(Hpd+1 |Γj , x̄), there exists a positive constant cx̄ and an admissible d-dimensional plane
Vx̄ passing through the origin, such that ν = cx̄ Hpd+1 |Vx̄ . Thus, as E is Euclidean d-
rectifiable and Hd |E is locally finite, it holds that σ = cHpd+1 |E and the result follows.
5Lemma 4.13 can be proved for parabolic Lipschitz graphs as well. Although, since we will not deal with
such general sets in the present manuscript and the proof is more involved, we skip it.
5. N ODAL SET OF CALORIC FUNCTIONS
Given a caloric function h, we denote by Σh := {h = 0} the nodal set of h and by σh

the associated surface measure
(5.1) dσh = dHn−1 |Σt dt.
In most of the paper the function h is clear from the context, in which case we simply
understand σ = σh .
If h is a non-zero caloric function on Rn+1 , by unique continuation [Po96, Theorem 1.2]
it cannot vanish in an open set. Hence, we may apply [HL94, Theorem 1.1] and [HL94,
Proposition 1.2] to deduce that Σh has locally finite Hn -measure and Σh ∩ |Dh|−1 (0) is
a (euclidean) (n − 1)-rectifiable set. In particular, for any cylinder Cr centered on Σh ,
the set Σh ∩ Cr can be decomposed into a union of an n-dimensional C 1 -submanifold
Cr ∩ Σh ∩ {|Dh| > 0} with finite n-dimensional Hausdorff measure and a closed set
Cr ∩ Σh ∩ {|Dh| = 0} of Hausdorff dimension not larger than n − 1.
For our purposes, we need a finer study of this decomposition in terms of admissible
graphs and so we define the regular and the singular set of Σh by

R := x̄ ∈ Σh : |∂t h| + |∇h| > 0

S := x̄ ∈ Σh : |∂t h| + |∇h| = 0 .
Additionally, we set

Rx := x̄ ∈ R : |∇h| > 0

Rt := x̄ ∈ R : |∂t h| > 0 and |∇h| = 0 = R \ Rx ,
to be the space-regular and the time-regular sets respectively.
Lemma 5.1 (Structure of the space-regular set). If h : Rn+1 → R is a non-zero caloric

function, then for every ȳ ∈ Rx there exists ρ = ρ(ȳ) > 0 and an admissible C ∞ -graph
Σ′ȳ such that Σh ∩ Cρ (ȳ) = Σ′ȳ ∩ Cρ (ȳ).
Proof. The proof is an easy application of the implicit function theorem. Indeed, if |∇h(ȳ)| >
0, we can assume without loss of generality that ∂n h(ȳ) 6= 0. Thus, we can find a cylinder
Cρ (ȳ) centered at ȳ in which ∂n h 6= 0 and a smooth function ϕ : Rn → R such that

x̄ = x′ , ϕ(x′ , t), t , for any x̄ ∈ Cρ (ȳ) ∩ Σh .
We remark that, despite the implicit function theorem defines ϕ just locally, we can extend
it to a C ∞ -function defined in the whole Rn (see e.g. [St70, Theorem 5, p. 181]).
In general, we cannot expect to express Rt locally as an admissible graphs (see also the
examples after the next lemma). However, we can still make some general consideration
about its geometry and we show that its dimension is lower than that of Rx .
Lemma 5.2 (Structure of the time-regular set). If h : Rn+1 → R is a non-zero caloric
function, then for every ȳ ∈ Rt there exists ρ̃ = ρ̃(ȳ) > 0 and a smooth (n−1)-dimensional

C ∞ -graph Σe ȳ such that Rt ∩ Cρ̃ (ȳ) ⊂ Σ
e ȳ ∩ Cρ̃ (ȳ). In particular, σ Rt = 0.
Proof. If Rt = ∅ there is nothing to prove. Fix ȳ = (y, s) ∈ Rt , and note that since
∂t h(ȳ) 6= 0, by the implicit function theorem, we can find ρ = ρ(ȳ) > 0 such that Cρ (ȳ) ∩
Σh agrees (possibly after a rotation in space) with the graph (x, ϕ(x)) of a C ∞ -function
ϕ : Rn → R inside Cρ (ȳ). Moreover, since Hh = 0 the latter implies that ∆h(ȳ) 6= 0.
Without loss of generality, we assume that ∂n2 h(ȳ) 6= 0, where ∂n stands for the partial
derivative in xn variable.
If we denote g := ∂n h and Σg = {g = 0}, then g is smooth, ȳ = (y ′ , yn , s) ∈ Σg
and ∂n g(ȳ) 6= 0. Thus, by the implicit function theorem, there exists ρ′ ≤ ρ such that
Cρ′ (ȳ)∩Σg is the (possibly rotated) graph (x′ , ψ(x′ , t), t) of a smooth function ψ : Rn → R
and ȳ = (y ′ , ψ(y ′ , s), s). Let us define
e
h(x′ , t) := h(x′ , ψ(x′ , t), t)
and note that e
h(y ′ , s) = h(y, s) = 0. So by the chain rule and the fact that (y, s) ∈ Rt
(hence ∂n h(y, s) = 0),
∂te
h(y ′ , s) = ∂n h(y, s)∂t ψ(y ′ , s) + ∂t h(y, s) = ∂t h(y, s) 6= 0.
Hence, as e h is clearly a smooth function, we can apply the implicit function theorem at
(y ′ , s) and obtain that there exists a neighborhood of (y ′ , s) and a C ∞ function ϕ : Rn−1 →
R such that {e h = 0} coincides with the graph (·, ϕ(·)) in that neighborhood. More specifi-
cally, there exists ρ′′ ≤ ρ′ such that Σ e ρ′′ := Σh ∩ Σg ∩ Cρ′′ (y, s) admits the parametrization

x′ , ψ x′ , ϕ(x′ ) , ϕ(x′ ) , x′ ∈ Bρ′′ (y ′ ).

In particular, Σ e ρ′′ is euclidean (n − 1)-rectifiable, so Hn+1 Σ e ρ′′ = Hn Σe ρ′′ = 0 and, by
p
the co-area formula [He17, Theorem D],
Z

Hn−1 Σ e ρ′′ |t dH2 (t) = 0,
p

e ρ′′ = 0. It is clear that Rt ∩ Cρ′′ (ȳ) ⊂ Σ
which, since Hp2 ≈ L1 , proves that σ Σ e ρ′′ and,
by a covering argument, we get that σ Rt ) = 0, as wished.
Example 5.3. a) Let us consider the caloric polynomial h1 (x1 , x2 , t) = x21 + x22 + 4t.
The set Σh1 is a rotational paraboloid around the time-axis. Its time-regular set Rt is the
singleton {0̄} and its space-regular set is Σh1 \ {0̄}.
b) Let h2 (x1 , x2 , t) = x21 + x22 − 2x1 x2 + 4t, whose nodal set is a parabolic cylinder. We
have that S = ∅, which gives that Σh2 = Rx ∪ Rt and Rt = {(x, x, 0) : x ∈ R}. So Rt is
a 1-dimensional manifold (in particular, a line), in contrast to what we have for the function
h1 of the previous example.
6. C ALORIC MEASURE ASSOCIATED WITH A CALORIC FUNCTION
Lemma 6.1. Let h be a caloric function in Rn+1 and let h+ and h− indicate the positive
and negative parts of h respectively. There exists a unique Radon measure ωh supported on
{h = 0} such that
Z Z Z Z
1
(6.1) ϕ dωh = h+ H ∗ ϕ = h− H ∗ ϕ = |h|H ∗ ϕ, for ϕ ∈ Cc∞ (Rn+1 ).
2
Proof. Let h̃± be the extension by zero of h± in the complement of {h± > 0}. Then, since
h is continuous in Rn+1 , h± → 0 continuously on {h = 0}. Thus, as h± is caloric in
{h± > 0}, we have that h̃± is a subcaloric function in Rn+1 and by [Wa12, Theorem 6.28],
there exists a unique Radon measure ωh̃± such that
Z Z Z
± ∗
ϕ dωh̃± = h̃ H ϕ = h± H ∗ ϕ, for ϕ ∈ Cc∞ (Rn+1 ).
Rn+1 {h± >0}
R
Now, since h̃± is caloric in {h± > 0} and zero in {h∓ > 0}, we have that ϕ dωh± = 0
for ϕ ∈ Cc∞ ({h± > 0}) and ϕ ∈ Cc∞ ({h∓ > 0}), implying that supp ωh± ⊂ {h = 0}.
Therefore, since Ln+1 (Σh ) = 0, for any ϕ ∈ Cc∞ (Rn+1 ), we have that
Z Z Z Z Z
+ ∗ − ∗
ϕ dωh̃+ − ϕ dωh̃− = h̃ H ϕ − h̃ H ϕ = h H ∗ ϕ = 0,
Rn+1 Rn+1 Rn+1
and so ωh̃+ = ωh̃− . Thus, the first two equalities in (6.1) hold, while the last one follows by
adding instead of subtracting the two identities above.
Given a caloric function h, by the discussion in Section 5, Σh is smooth away from a

euclidean (n − 1)-rectifiable set, and so the set Ω± = {h± > 0} is a set of locally finite
perimeter in Rn+1 (see Definition 5.1 and Theorem 5.23 in [EG15]). Hence, by [Mag12,
Theorem 18.11], for a.e. t ∈ R, its horizontal section Ω± t is a set of locally finite perimeter
in Rn . In fact, more is true. By Lemma 5.1, around σ-a.e. any point, we have that Σh
is given by an admissible Lipschitz graph, while the rest of the points lie on an (n − 1)-
rectifiable set. So, for a.e. t, Σht is also locally Lipschitz and thus, its measure theoretic
boundary (see [EG15, Definition 5.7]) coincides with its topological boundary. Therefore,
for a.e. t there is a unique measure theoretic outward unit normal νt± to ∂Ω± t such that we
have the generalized Gauss-Green Theorem
Z Z
(6.2) div ϕ dx = ϕ · νt± dHn−1 , for all ϕ ∈ Cc1 (Rn ; Rn ).
Ω±
t Σh
t
Moreover, νt± coincides with the usual (geometric) outward unit normal on ∂ ∗ Ω± (see
Definitions 5.4 and 5.6, Theorems 5.15 and 5.16, and Lemma 5.5 in [EG15]) .
Let us recall the parabolic Cauchy estimates for caloric functions.
Proposition 6.2 (see e.g. [HL94], Proposition 2.1). Let R > 0 and let h be a caloric
function in CR . For r < R and α ∈ Zn+ and any positive integer ℓ with |α| + 2ℓ = m, we
have
(6.3) |D α,ℓ h(x, t)| .n,m (R − r)−m khkL∞ (CR ) , for all (x, t) ∈ Cr .
Lemma 6.3. If h is a caloric function in Rn+1 and let σ be the surface measure on Σh as
defined in (5.1), then ∂t |h| ∈ L∞ n+1 ) and for any ϕ ∈ C 2,1 (Rn+1 ),
loc (R c
Z Z Z
∂h
(6.4) ϕ dωh = ∂t |h| ϕ dxdt − + ϕ dσ,
Rn+1 Σh ∂νt
where ∂h
∂νt+
:= νt+ · ∇h 6= 0 σ-a.e.
Proof. If we apply (6.2) first to h(·, t)∇ϕ(·, t) ∈ Cc1 (Rn ; Rn ) and then to ϕ(·, t)∇h(·, t) ∈
Cc1 (Rn ; Rn ) in Ω± t , for fixed t, and integrate in t, we obtain
Z Z Z
∗
hH ϕ dxdt − h ∂t ϕ dxdt = h∆ϕ dxdt
Ω± Ω± Ω±
Z Z Z Z
∂ϕ ∂h
(6.5) =− ∇h∇ϕ dxdt + h ± dσ = ∆h ϕ dxdt − ± ϕ dσ
Ω± Σh ∂νt Ω± Σh ∂νt
Z Z
∂h
= ∂t h ϕ dxdt − ± ϕ dσ,
Ω ± Σ ∂νt
h
where we used that h is caloric and h = 0 on Σh . As f (·) := | · | is smooth away from 0

and Lipschitz in R with Lip(f ) ≤ 1 and h ∈ C 2,1 (Rn+1 ), by Rademacher’s theorem, we
have that for x̄ ∈ Rn+1 \ Σh ,
(6.6) ∇x,t f (h)(x̄) = f ′ (h(x̄))∇x,t h(x̄),
which, by (6.3), implies that |h| is locally Lipschitz (in the euclidean norm) and so ∂t |h| ∈
L∞
loc (R
n+1 ). Thus, since hχ ± = ±h± χ n+1 and ∂ hχ ± = ±∂ h± χ n+1
Ω R t Ω t R \Σh , we get that
Z Z Z
(6.1)
2 ϕ dωh = hH ∗ ϕ dxdt − hH ∗ ϕ dxdt
Ω + −
Z Z Ω Z Z
(6.5) ∂h ∂h
= − |h| ∂t ϕ dxdt + ∂t |h| ϕ dxdt − + ϕ dσ + − ϕ dσ
Rn+1 Rn+1 \Σh Σh ∂νt Σh ∂νt
Z Z
∂h
=2 ∂t |h| ϕ dxdt − 2 + ϕ dσ,
Rn+1 Σh ∂νt
where in the last equality we used νt+ = −νt− and integrated by parts in t using that
Ln+1 (Σh ) = 0 and |h| is locally Lipschitz in Rn+1 . Recall that the points of Σh where
∇h = 0 are contained in an (n − 1)-rectifiable set and thus, have σ-measure zero. As the
tangential component of ∇h on each slice Σht is zero, we have that ∇h = ∂ν + h and so
t
∂ν + h 6= 0 σ-almost everywhere.
t
Lemma 6.4. If h is a caloric function in Rn+1 and σ is the surface measure on Σh as

defined in (5.1), then we have that
Z
∂h
(6.7) ωh (A) = − + dσ, for any Borel A ⊂ Σh ,
A ∂ν t
∂h ∞
where the associated Poisson kernel kh = − ∂ν + is positive σ-a.e. and in Lloc (σ). In
t
particular, ωh ≪ σ.
Proof. Let A be a compact subset of Σh . Note that since σ and ωh are Radon measures,
it holds that ωh (A) < ∞ and σ(A) < ∞. By Urysohn’s lemma, we can find a sequence
of decreasing functions {ϕj }∞ ∞
j=1 ⊂ Cc (R
n+1 ), 0 ≤ ϕ ≤ 1, so that ϕ = 1 on A and
j j
ϕj → χA pointwisely. Let us denote Kj := supp ϕj and observe that Kj+1 ⊂ Kj for all
j ≥ 1. Then, by (6.4), we have
Z Z Z
∂h
ϕj dωh = − + ϕj dσ + ∂t |h| ϕj .
Σh ∂νt Rn+1
As σ(A) < ∞ and Σh is Euclidean n-rectifiable, (4.2) entails Hpn+1 (A) < ∞ and conse-
quently Ln+1 (A) ≈ Hpn+2 (A) = 0. Therefore, since supp ϕj ⊂ Kj ⊂ K1 for any j ≥ 1
and ∂t |h| ∈ L∞ (K1 ), then, by dominated convergence, we have
Z
∂h
(6.8) ωh (A) = − lim ϕj dσ.
j→∞ Σh ∂ν +
t
If Cρ is a cylinder of radius ρ ≈ diam K1 which is centered at Σh and satisfies K1 ⊂ Cρ/2 ,
by (6.3) we have that for x̄ ∈ Rn+1 ,

∂h

sup + (x̄) ϕj (x̄) .n χK1 (x̄) sup |∇h| .ρ χK1 (x̄) khkL∞ (Cρ ) ∈ L1 (σ),
j≥1 ∂ν
t Cρ/2
since h ∈ L∞ (Cρ ) and σ is locally finite. Thus, by (6.8) and the dominated convergence
theorem, we conclude (6.7) for compact subsets of Σh . Recall that ωh and σ are Radon mea-
sures and so the result for Borel sets follows from inner regularity (see [Mat95, Definition
1.5]).
Remark 6.5. In [Bad11], [AMT17], [AMTV19], and [AM19], it is stated that if h is har-
monic, then the “Poisson kernel” of ωh is ∂ν + h+ . Although, since h+ is not differentiable
on Σh , a direct application of (6.2) is not allowed and one should obtain ∂ν + h instead. Nev-
ertheless, this detail does not create any issues in the proofs of the aforementioned papers.
It is only in the proof of [Bad11, Lemma 4.2] one has to notice that h± agrees with h in Ω±
and so the conclusion of the lemma is still true.
7. C ALORIC POLYNOMIAL MEASURES
We recall that we write Θ to denote the set of caloric functions on Rn+1 vanishing at 0̄.
Lemma 7.1. If h ∈ Θ and ωh the associated caloric measure, then
(7.1) T0̄,r [ωh ] = r n ωh◦T −1 .
0̄,r
Moreover, if h ∈ F (k),
(7.2) T0̄,r [ωh ] = r n+k ωh .
Proof. Let ϕ ∈ Cc∞ (Rn+1 ). An application of the chain rule gives
1
(7.3) H ∗ ϕ ◦ T0̄,r (x̄) = 2 (H ∗ ϕ) ◦ T0̄,r (x̄).
r
So, given h ∈ Θ, we have
Z Z
ϕ dT0̄,r [ωh ] = ϕ ◦ T0̄,r dωh
Z Z
1 ∗
(7.3) 1
= |h|H ϕ ◦ T0̄,r = 2
|h| (H ∗ ϕ) ◦ T0̄,r
2 2r
Z Z
r n
−1 ∗ n
= h ◦ T0̄,r H ϕ = r ϕ dωh◦T −1 ,
2 0̄,r
which proves the first part of the statement. If we further assume that h ∈ F (k), we obtain
−1

h ◦ T0̄,r (x̄) = h δr (x̄) = r k h(x̄), x̄ ∈ Rn+1 ,
and (7.2) follows.
We recall that we use the notation Cr := Cr (0̄) for cylinders centered at the origin. A
consequence of the previous lemma is the following corollary.
Corollary 7.2. If h ∈ F (k) and r1 , r2 > 0 it holds that
n+k+1
r1
(7.4) Fr1 (ωh ) = Fr2 ωh .
r2
In particular, for any r > 0 and M ≥ 1,
M r n+k+1 r n+k+1
(7.5) ωh (CM/2 ) ≤ Fr ωh ≤ M ωh (CM ).
2 M M
Proof. By Lemma 4.5 and Lemma 7.1 we have that

Fr (ωh ) = rF1 T0̄,r [ωh ] = r n+1 F1 ωh◦T −1 = r n+k+1 F1 ωh ,
0̄,r
and so (7.4) readily follows. By Lemma 4.5-(2) and (7.4), it is straightforward to see that
(7.5) holds.
Lemma 7.3. Let hj be a sequence in Θ which converges uniformly on compact sets to some
h ∈ Θ. Then ωhj ⇀ ωh .
Proof. The proof is a minor variant of the one of [AM19, Lemma 5.4] and we omit it.
Lemma 7.4. Let h ∈ Θ and let ωh be the associated caloric measure. Then 0̄ ∈ supp ωh .
Proof. Let us recall that Σh = Rx ∪ Rt ∪ S (see section 5) and that the Poisson kernel is
given by −∂νt h. Therefore, by Lemma 5.1, for any x̄ ∈ Rx there is a sufficiently small
neighborhood of x̄ in which Σh agrees with an admissible smooth graph. Since h vanishes
on Σh , the component of ∇h which is tangential to Σht is the zero vector and we have
∇h = ∂νt h. So ∂νt h(x̄) 6= 0 for any x̄ ∈ Rx and thus x̄ ∈ supp ωh showing that Rx ⊂
Rx ⊂ supp ωh ⊂ Σh .
In light of [HL94, Theorem
1.1] and Lemma 5.2, for ρ > 0 small enough, it holds that
H n−1 (Rt ∪ S) ∩ Cρ < ∞, and so Rt ∪ S has empty interior in the relative topology of
Σh . Hence,
0̄ ∈ Σh ∩ Cρ = Σh \ (Rt ∪ S) ∩ Cρ = Rx ∩ Cρ ⊂ supp ωh ,
which finishes the proof.
Lemma 7.5. The d-cones F (k) and P(k) have compact basis for all k. Moreover, for
h ∈ P(k) and r > 0 we have
khkL∞ (Cr ) ≈k r −n−1 Fr (ωh ).
Proof. In light of Lemma 7.4, the proof is a routine adaptation of Lemmas 5.5 and 5.6, and
Corollary 5.7 in [AM19].
In our argument we need the following formula for the expansion of caloric functions by
Han and Lin, which we report for the reader’s convenience.
Theorem 7.6 (see [HL94], Theorem 2.2). Let R > 0 and let h be a caloric function in CR .
Then, for any positive integer d, 0 < r < R/2, and x̄ ∈ Cr , we have that
d
X X X d
D α,ℓ h(0̄) α ℓ
(7.6) h(x, t) = x t + Rd (x, t) =: hj + Rd (x, t),
α!ℓ!
j=1 |α|+2ℓ=j j=1
where
r d+1
(7.7) |Rd (x, t)| .n,d khkL∞ (CR ) , for (x, t) ∈ Cr .
(R − r)d+1
Lemma 7.7. Let h ∈ Θ and hj be as in (7.6). If m ≥ 1 is the smallest integer such that
hm 6≡ 0, then there exists ρ0 > 0 and c0 > 0 such that
(7.8) c−1
0 Fr (ωhm ) ≤ Fr (ωh ) ≤ c0 Fr (ωhm ) for all r ≤ ρ0 .
Proof. Let r < 1/100. By Theorem 7.6 and the fact that hj = 0 for j < m, there exists a
function Rm : Rn+1 → R such that
h(x̄) = hm (x̄) + Rm (x̄), x̄ ∈ C2r ,
and
r m+1
(7.9) |Rm (x̄)| .n,m khkL∞ (C1 ) , x̄ ∈ C2r .
(1 − 2r)m+1
Let ψr ∈ C 2,1 (Cr ) be so that ψr = 1 in Cr/2 and 0 ≤ ψr ≤ 1, satisfying
|∇ψr | ≤ C/r, and |∂t ψr | + |∂x2i xj ψr | ≤ C/r 2 , for 1 ≤ i, j ≤ n.
One way to construct such a cut-off is the following: let η ∈ C ∞ ([0, ∞)) be the standard
cut-off so that η = 1 in [0, 1/2], η = 0 in (1, ∞), 0 ≤ η ≤ 1, |η ′ | ≤ C and |η ′′ | ≤ C. If
ζr (x) = η(|x|/r) and ξr (t) := η(|t|/r 2 ), we define ψr (x, t) = ζr (x)ξr (t). It is easy to see
that the function ϕ = (2C)−1 r ψr is an admissible function for Fr . With this choice of ϕ
we have
Z Z Z 1Z
∗
ϕ dω h − ϕ dω hm = (|h| − |hm |)H ϕ ≤ |Rm (x, t)|
Rn+1 r C2r
and by (7.14),
Z
1 r m+n+2
|Rm (x, t)| .m,n khkL∞ (C1 ) .
r C2r (1 − 2r)m+1
Hence, since r < 1/100, there is c(n, m) > 0 such that
Z Z

(7.10) ϕ dωh − ϕ dωhm ≤ c(n, m)r n+m+2 khkL∞ (C1 )
and so, if for a fixed ε > 0 we further assume

r ≤ ρ0 := εF1 (ωhm )c(n, m)−1 khk−1
L∞ (C1 ) ,
the inequality (7.10) reads

Z Z

(7.11) ϕ dωh − ϕ dωhm ≤ εr n+m+1 F1 (ωhm ) = εFr (ωhm ),
where in the last step we used (7.4). Let us observe that, by Lemma 7.4, 0̄ ∈ supp ωhm ,
which, in turn, by (7.5), implies that 0 < Fr (ωhm ) < ∞ for all r > 0.
Therefore, since ϕ is admissible for Fr , we obtain
Z
1 r (7.11)
(7.12) Fr (ω h m ) ≤ ω h m (C r/2 ) ≤ ϕ dω h m ≤ Fr (ωh ) + εFr (ωhm )
C2m+n+3 4C
and
Z
−1 r (7.11)
(4C) Fr/2 (ωh ) ≤ ωh (Cr/2 ) ≤ ϕ dωh ≤ (1 + ε)Fr (ωhm )
(7.13) 4C
= 2m+n+1 (1 + ε)Fr/2 (ωhm ),
which conclude (7.7) by choosing ε = C −1 2−n−m−4 .
We apply the previous lemma to show that the first non-zero term in the expansion hm of
h determines the density of ωh at 0̄.
Lemma 7.8. Let h, hm and c0 be as in Lemma 7.7. Then
ωh (Cr ) ωh (Cr )
lim inf n+m
≈ lim sup n+m ≈ ωhm (C1 ) ∈ (0, ∞),
r→0 r r→0 r
where the implicit constants depend on n, m and c0 .
Proof. Let ρ0 and c0 be as in Lemma 7.7 and let r < ρ0 /2. We apply Lemma 4.5-(2), (7.8),
and (7.5), and we have that
ωh (Cr )
≤ r −n−m−1 F2r (ωh ) ≤ c0 r −n−m−1 F2r (ωhm ) = c0 2n+m+1 F1 (ωhm ) . ωhm (C1 ).
r n+m
Arguing similarly but using the converse inequalities of the ones we used above, we infer
that
ωh (Cr )
≥ r −n−m−1 Fr (ωh ) ≥ c−1
0 r
−n−m−1
Fr (ωhm )
r n+m
= c−1
0 2
n+m+1
F1 (ωhm ) & ωhm (C1 ) > 0,
where the last inequality holds because 0̄ ∈ supp ωhm .
Lemma 7.9. Let h ∈ Θ and hj be as in (7.6). If m ≥ 1 is the smallest integer such that
hm 6≡ 0, then Tan(ωh , 0̄) = {cωhm : c > 0}.
Proof. Let R > 0 and r < 1/2. By Theorem 7.6 and hj = 0 for j < m, we have that
h(x̄) = hm (x̄) + Rm (x̄), x̄ ∈ CrR ,
where
r m+1
(7.14) |Rm (x̄)| .n,m khkL∞ (CR ) , x̄ ∈ CrR .
Rm+1 (1 − r)m+1
Given ȳ ∈ CR , we have that δr (ȳ) ∈ CrR , so we combine (7.14) with the homogeneity of
hm in order to get
−1
|r −m h ◦ T0̄,r (ȳ) − hm (ȳ)|
= |r −m h(ry, r 2 s) − hm (y, s)| = r −m |h(ry, r 2 s) − hm (ry, r 2 s)|
r r
.n,m m+1 m+1
khkL∞ (CR ) .m m+1 khkL∞ (CR ) ,
R (1 − r) R
−1
where the last term converges to 0 as r → 0. In particular, this shows that r −m h ◦ T0̄,r
converges to hm uniformly on compact subsets of Rn+1 .
The definition of caloric polynomial measure together with Lemma (7.1) gives us that
(7.15) ωr−m h◦T −1 = r −m ωh◦T −1 = r −n−mT0̄,r [ωh ].
0̄,r 0̄,r
In order to finish the proof of the lemma, it suffices to use (7.15) and the fact that ωr−m h◦T −1
0̄,r
converges weakly to ωhm by Lemma 7.3. Indeed, ωh has positive lower and finite upper
(n + m)-density at 0̄ by Lemma 7.8, so we can apply Lemma 4.10 to conclude that every
measure in Tan(ωh , 0̄) is of the form cωhm for some c > 0.
The next result is an important application of the previous lemma.

Lemma 7.10. Let ω be a Radon measure on Rn+1 , x̄ ∈ supp ω and let d be the minimal
integer such that Tan(ω, x̄) ∩ P(d) 6= ∅. Then Tan(ω, x̄) ∩ P(d) ⊂ F (d).
Proof. It is enough to argue as in [AM19, Lemma 5.9]6 and invoke Lemma 7.9.
Lemma 7.11. Let h ∈ P (d) and assume that

d
X d
X X
h(x, t) = hj (x, t) := akℓ xk tℓ .
j=1 j=1 |k|+2ℓ=j
Then, there exists r0 > 0 and C0 ≥ 1 such that

(7.16) C0−1 Fr (ωhd ) ≤ Fr (ωh ) ≤ C0 Fr (ωhd ), for all r ≥ r0 ,
P
where r0 depends on d, n, F1 (ωhd ), max1≤j≤d−1 |k|+2ℓ=j |akℓ |, and C0 depends on d
and n.
Proof. Let ψr be a cut-off function as in Lemma 7.7. In particular, ψr ∈ C 2,1 (Cr ), 0 ≤
ψr ≤ 1, ψr = 1 in Cr/2 and
|∇ψr | ≤ C/r, and |∂t ψr | + |∂x2i xj ψr | ≤ C/r 2 , for 1 ≤ i, j ≤ n.
6The proof of the cited lemma refers to [Mat95, Theorem 14.16] in order to guarantee that tangent measures
of tangent measures are tangent measures themselves. However, that result holds at almost every point. The
proof can be fixed invoking [Bad11, Lemma 2.6].
Hence, the function ϕ := (2C)−1 rψr is admissible for the functional Fr , |H ∗ ϕ| ≤ r −1 ,

and we have
Z Z Z

ϕ dωh − ϕ dωh = ∗
d n+1 (|h| − |hd |)H ϕ
R
Z d−1 Z
1 1X
≤
|h| − |hd | ≤ |hj |.
r C2r r C2r j=1
For any x̄ ∈ C2r , it holds that

X
|hj (x̄)| ≤ (2r)j |akℓ | =: cj r j ,
|k|+2ℓ=j
which, in turn, implies that

d−1 Z
X X
d−1
−1 r n+d+1
r |hj | ≤ r −1 (2r)n+2 max cj r j ≤ 2n+2 max cj .
C2r 1≤j≤d−1 r−1 1≤j≤d−1
j=1 j=1
Therefore, if we let
r0 > 1 + (εF1 (ωhd ))−1 2n+2 max cj ,
1≤j≤d−1
for some ε > 0 small enough, we infer that

Z Z

ϕ dωh − ϕ dωh ≤ εr d+n+1 F1 (ωd ) = εFr (ωd ),
d
where in the last equality we used (7.4). Note that 0̄ ∈ supp(ωhd ) and thus, by (7.5), we
have that 0 < F1 (ωhd ) < ∞. We conclude the proof arguing as in (7.12) and (7.13).
Lemma 7.12. Let h, r0 and C0 be as in Lemma 7.11. There exists ε0 > 0 and r1 > 0 such
that if dr ωh , F (k) < ε0 for all r ≥ r1 , then k = d.

Proof. Fix τ ≥ 2 to be chosen and pick r ≥ r1 such that dτ r ωh , F (k) < ε0 . In particular,
there exists ψ ∈ F (k) such that Fτ r (ψ) = 1 and
ω ω
h
Fr , ψ ≤ Fτ r , ψ < ε0 .
Fτ r (ωh ) Fτ r (ω)
Hence
Fr (ωh )
(7.17) Fr (ψ) − ε0 < < Fr (ψ) + ε0 .
Fτ r (ωh )
Also, since ψ is homogeneous, (7.4) gives
(7.18) Fr (ψ) = r n+k+1 F1 (ψ) = τ −n−k−1 Fτ r (ψ) = τ −n−k−1 .
Assuming that r > r2 := max{r0 , r1 }, by Lemma 7.11 and (7.4) we have
Fr (ωh ) Fr (ωhd )
(7.19) ≤ C02 ≤ C02 τ −n−d−1 .
Fτ r (ωh ) Fτ r (ωhd )
Therefore, (7.17), (7.18), and (7.19) infer that
τ −n−k−1 − ε0 < C02 τ −n−d−1 ,
or equivalently,
τ d−k (1 − ε0 τ n+d+1 ) < C02 .
If we pick τ /2 = C02 and ε0 = τ −n−d−2 , the latter inequality implies that
τ2
τ d−k (τ − 1) < ,
2
which can only hold if d = k, since τ ≥ 2.
The next lemma is crucial to our purposes: it provides a connectivity result for parabolic
cones of Radon measures. The proof translates unchanged to the parabolic setting and we
skip it.
Lemma 7.13 (see Lemma 3.10, [AM19]). Let F and M be parabolic d-cones and assume
that F has compact basis. Moreover, suppose that there is ε0 > 0 such that the following
property holds: if µ ∈ M and there exists r0 > 0 such that dr (µ, F ) ≤ ε0 for all r ≥ r0 ,
then µ ∈ F . If η is a Radon measure and x̄ ∈ supp η are such that Tan(η, x̄) ⊂ M and
Tan(η, x̄) ∩ F 6= ∅, then Tan(η, x̄) ⊂ F .
The following proposition gathers all the results of this section. After proving all the
previous lemmas, the proof is analogous to that of [AM19, Proposition II]. We report it
anyways, in order to give the reader the precise references inside this section.
Proposition 7.14. Let ω be a Radon measure in Rn+1 and let x̄ ∈ Rn+1 be such that
Tan(ω, x̄) ⊂ P(k) for some k. If Tan(ω, x̄) ∩ F (k) 6= ∅ for some integer k, then
Tan(ω, x̄) ⊂ F (k).
Proof. Let us assume that Tan(ω, x̄) ⊂ P(k) and let m ≤ k be the smallest integer for
which Tan(ω, x̄) ∩ P(m) 6= ∅. In particular Tan(ω, x̄) ∩ P(m) ⊂ F (m) by Lemma
7.10, which gives that Tan(ω, x̄) ∩ F (m) 6= ∅. Furthermore, F (k) has compact basis by
Lemma 7.5 and, in light of Lemma 7.12, we can apply Lemma 7.13 with M = P(k),
F = F (k) and η = ω. Thus, Tan(ω, x̄) ⊂ F (k).
8. E LEMENTS OF THE THEORY OF O PTIMAL T RANSPORT

In this section, we record and develop some basic theory of Optimal Transport. In par-
ticular, we are interested in the following:
(1) The relationship between the Kantorovich norm k·kKR on the space of signed mea-
sures M(Ω) with not necessarily zero mass and the Wasserstein distance between
finite non-negative measures on a bounded open set Ω ⊂ Rn whose masses are not
necessarily equal.
(2) Kantorovich duality in the class of finite non-negative measures on Ω whose masses
are not necessarily equal.
(3) Identifying the space of 1-Lipschitz functions defined on Ω that vanish on ∂Ω as
the dual of (M(Ω), k · kKR ).
(4) The identification of the completionof M(Ω) or, equivalently, the closure of M(Ω)
as a subspace of Lip(Ω)∗ , k · kKR .
Let Ω ⊂ Rn be a bounded open set, M(Ω) be the set of signed Borel measures on Ω
such that |µ|(Ω) < ∞, and let M+ (Ω) be the set of non-negative measures in M(Ω). Set

Lip01 (Ω) := ϕ : Ω → R : ϕ ∈ Lip(Ω), kϕkLip ≤ 1 and ϕ(x) = 0 on ∂Ω
and note that dist(·, Rn \ Ω) ∈ Lip01 (Ω) and supx∈Ω |ϕ(x)| ≤ diam(Ω), for any ϕ ∈
Lip01 (Ω). We define the Kantorovich-Rubinshtein norm on M(Ω) by
Z
(8.1) kµkKR(Ω) := sup ϕ dµ : ϕ ∈ Lip01 (Ω)
and, for µ, ν ∈ M+ (Ω) the modified 1-Wasserstein distance

Z
(8.2) W b1 (µ, ν) := inf |x − y| dγ(x, y) : γ ∈ Πb(µ, ν) ,
Ω×Ω
where

Πb(µ, ν) := γ ∈ M+ (Ω × Ω) : π♯1 (γ)|Ω = µ and π♯2 (γ)|Ω = ν .
Here we used the notation π i : Ω1 × Ω2 → Ωi for the canonical projection from Ω1 × Ω2
onto Ωi , i = 1, 2. The distance W b1 differs from the classical Wasserstein distance because
we allow mass to be “transported” also on ∂Ω: in our case the transport plan γ takes values
in M+ (Ω × Ω) instead of M+ (Ω × Ω). For σ ∈ M(Ω), we set

Γb(σ) := γ ∈ M+ (Ω × Ω) : π♯1 (γ)|Ω , π♯2 (γ)|Ω ∈ M+ (Ω) and π♯1 (γ)|Ω − π♯2 (γ)|Ω = σ
and define the Wasserstein functional
Z
(8.3) kσkW := inf |x − y| dγ(x, y) : γ ∈ Γb(σ) .
Ω×Ω
It was proved in [FG10, Proposition 2.9] that the space (M+ (Ω), W b1 (·, ·)) is a geodesic
metric space. In fact, the result was stated for the W b2 distance but the same proof works
for W b1 as well. We also refer to [BBS97] and [BB01]. Moreover, as stated in [San15, p.
313], the infimum in (8.2) is attained and the following duality formula holds:
Z
(8.4) W b1 (µ, ν) = min |x − y| dγ(x, y) : γ ∈ Πb(µ, ν) = kµ − νkKR(Ω) .
Ω×Ω
Another important result is the Kantorovich duality, which holds in this setting as well.
Let C0 (Ω) be the set of all continuous functions on Ω vanishing on ∂Ω and Φc (Ω) be the set
of functions (ϕ, ψ) ∈ C0 (Ω) × C0 (Ω) satisfying ϕ(x) + ψ(y) ≤ |x − y| for µ-a.e. x ∈ Ω
and ν-a.e. y ∈ Ω. Then
Z Z
(8.5) W b1 (µ, ν) = sup ϕ dµ + ψ dν .
(ϕ,ψ)∈Φc (Ω) Ω Ω
We refer to the proof of [Vil03, Theorem 1.3] and in particular, [Vil03, pp. 26-27, case 1].
The only two modifications that are required are the following:
• One should use the class C0 instead of the one of bounded continuous functions
denoted by Cb .7
7This comment should be used every time we modify a proof and will not be repeated.
• At the top of p. 27, where it is demonstrated that the functional on the right-hand
e one should
side of (8.5) is well-defined, in order to deduce that ϕ = ϕ̃ and ψ = ψ,
utilize that the functions in Φc (Ω) vanish on ∂Ω instead of the fact that µ(Ω) and
ν(Ω) are equal (which does not hold in our case).
Let us emphasize that it is not necessary to invoke [FG10] in order to show that W b1
is a metric since the case p = 1 is easier. Indeed, we can follow the proof of [Ed11,
Theorem 4.1] (making some easy modifications) and use (8.5) to conclude (8.4). Then,
combining [Ed11, Lemma 4.3] and a standard argument that can be found for instance in
[Sle16, Lemma 1.13], we can show directly that k · kKR is a norm using the fact that for a
closed set C ⊂ Ω, the function fn (x) = n−1 (1 − n dist(x, C))+ ∈ Lip(Ω) and vanishes
on ∂Ω for n ≥ [dist(C, ∂Ω)]−1 . Thus, W b1 becomes a metric and by (the proof of) [Ed11,
Lemma 4.5], it holds that k · kW = k · kKR(Ω) on M(Ω).
For any ϕ ∈ Lip01 (Ω), we define
Z
(8.6) ϕ(µ)
b := ϕ dµ.
Ω
Then the following theorem holds:
Theorem 8.1. The map ϕ 7→ ϕ b is a linear bijection of Lip01 (Ω) onto the Banach dual of
b = kϕkLip(Ω) . Thus, the Banach dual of (M(Ω), k · kKR(Ω) )
(M(Ω), k · kKR(Ω) ) and kϕk
is indeed Lip01 (Ω).
Proof. We follow the proof of [Ed11, Theorem 7.3]. One only has to check the results in
sections 2, 3, 4, 6, and 7 up to Theorem 7.3. Since most of the arguments require minor
changes we will provide only the roadmap for the proof apart from Lemma 7.2 and Theorem
7.3, that require some care.
• One can deal with Section 3 invoking the Kantorovich duality (8.5).
• In Section 4, one only has to check Theorem 4.5, since Theorems 4.1 and 4.4 are
taken care of by (8.4) and [FG10]. The proof is exactly the same. For future refer-
ence, we get that k · kW = k · kKR(Ω) .
• From Section 6, we only need Theorem 6.1, which is already stated in the desired
generality.
• Lemma 7.1 should be stated just for points in Ω and thus, the second part of the
proof of Lemma 7.2 works as it is for points in Ω. To extend it to the boundary
points it is enough to use that Lip01 (Ω) functions vanish on the boundary. Indeed,
let εz stand for the Dirac mass at z ∈ Ω, and let x ∈ Ω and y ∈ ∂Ω. Then, for any
ϕ ∈ Lip01 (Ω) we have
|ϕ(x) − ϕ(y)| |ϕ(x)| |ϕ(x)| |ϕ(ε
b x )|
= ≤ = = |ϕ(δ
b x )|,
|x − y| |x − y| dist(x, ∂Ω) dist(x, ∂Ω)
where δx := dist(x, ∂Ω)−1 εx . Note that if d(x) ∈ ∂Ω is such that |x − d(x)| =
dist(x, ∂Ω),
kεx kKR(Ω) = sup ϕ(x) = sup (ϕ(x) − ϕ(d(x))) ≤ |x − d(x)| = dist(x, ∂Ω),
ϕ∈Lip01 (Ω) ϕ∈Lip01 (Ω)
and thus, kδx kKR(Ω) ≤ 1. Hence,

|ϕ(x) − ϕ(y)|
sup ≤ sup |ϕ(ν)|
b : ν ∈ M(Ω), kνkKR(Ω) ≤ 1 .
x∈Ω,y∈∂Ω |x − y|
Since k · kW = k · kKR , the result follows by the latter estimate and the proof of
Lemma 7.2.
• To show Lemma 7.3 we should also deal with the boundary points. For f ∈ M(Ω)∗ ,
we define u(x) = f (εx ), if x ∈ Ω, and notice that u(x) − u(y) = f (εx ) − f (εy ) =
f (εx − εy ), which is 1-Lipschitz by Lemma 7.1. To extend the definition of u to the
boundary, for x ∈ ∂Ω, we define u(x) = limk→∞ u(xk ), where xk is a sequence
of points in Ω so that xk → x. This is well-defined, since if yk ∈ Ω is another
sequence so that yk → x, then
|u(xk ) − u(yk )| ≤ kf k|xk − yk | ≤ kf k(|xk − x| + |yk − x|)
and so limk→∞ u(xk ) = limk→∞ u(yk ). It is easy to show that the extension of u,
that we still denote by u, is 1-Lipschitz on Ω and vanishes on ∂Ω, i.e., u ∈ Lip01 (Ω).
We will only deal with the latter. Fix x ∈ ∂Ω and pick Ω ∋ xk → x as k → ∞.
Then, in light of the identity k · kW = k · kKR , we obtain |u(xk )| = |f (εxk )| ≤
kf kkεxk kKR(Ω) and since
kεxk kKR(Ω) = sup ϕ(xk ) = sup (ϕ(xk ) − ϕ(x)) ≤ |xk − x|,
ϕ∈Lip01 (Ω) ϕ∈Lip01 (Ω)
by taking limits as k → ∞, we infer that u(x) = 0. Remark that

b(εx ) = u(x) = f (εx ), for x ∈ Ω,
u
and so u
b and f are bounded linear functionals that agree on a dense subspace of
M(Ω) (namely, the space of finite Borel measures whose support is a finite set).
The theorem readily follows.
We move to the last part of this section, which is the identification of the completion of
M(Ω) (or, equivalently, the closure of M(Ω)) as a subspace of (Lip01 (Ω)∗ , k · kKR(Ω) ).
Following [BCJ05], we denote this completion by M(Ω) f and emphasize that it is a strict
∗
subspace of (Lip01 (Ω) , k · kKR(Ω) ) since otherwise Lip01 (Ω) would be a reflexive Banach
f
space. If u ∈ M(Ω) ⊂ Lip01 (Ω)∗ , the pairing hu, ϕi is well-defined for any ϕ ∈ Lip01 (Ω).
We define the first order distribution
Tu (ϕ) = hu, ϕi, for ϕ ∈ C0∞ (Ω),
where C0∞ (Ω) are the smooth functions in Ω vanishing on ∂Ω. Then the following theorem
holds:
Theorem 8.2. Let Ω ⊂ Rn be a bounded open set. Then
n o
(8.7) f
Tu : u ∈ M(Ω) = − div σ : σ ∈ L1 (Ω, Rn )
as subsets of D ′ (Ω). Moreover, if V0 is the closed subspace {σ ∈ L1 (Ω, Rn ) : div σ = 0},

the linear map σ ∈ L1 (Ω, Rn )/V0 7→ − div σ ∈ M(Ω) f is an isometry, i.e.,
kσkL1 (Ω,Rn )/V0 = k − div σkKR(Ω) .
Proof. The proof is almost the same as the one of [BCJ05, Theorem 2.4] and so we shall
only make a few comments:
• In the statement of Lemma 2.1, ϕn ∈ C0∞ (Ω) and in display (2.1), ϕn → 0 (i.e.,
c = 0).
• In the proof of Lemma 2.1, one should define Sδ : Lip01 (Ω) → C0∞ (Ω) to be the
approximation vδ given in the proof of [BBDP03, Theorem 3.1, Steps 1 and 2]
setting Σ = ∂Ω. It is clear from the proof of [BBDP03, Theorem 3.1] that if
u ∈ Lip01 (Ω), then vδ ∈ C0∞ (Ω) satisfying the conditions (1)-(3) in the proof
of Lemma 2.1. The rest of the argument is identical. Let us point out that this
approximation works in arbitrary open sets.
• The proof of Lemma 2.3 is the same. In particular, it is even simpler as we do not
need to rule out the constants.
• In Lemma 2.5, for p > n, one should define ap over functions u ∈ W01,p (Ω) (see
[BBDP03, p. 15, display (21)]). Then
R the infimum is attained at a unique point
up ∈ C0 (Ω) without the condition Ω up = 0 since this is just the unique solution
T
of the Dirichlet problem ∆p u = µ in Ω for u ∈ q< n(p−1) W01,q (Ω). For p large
n−1
enough, since the boundary values are zero in the Sobolev sense, we may extend up
by zero and use Morrey’s inequality to obtain a Hölder continuous representative
on Ω that vanishes on ∂Ω. The rest of the proof is almost the same and is based on
the one of [BBDP03, Theorem 4.1].
• Using Theorem 8.1 and following the proof of Theorem 2.4 mutatis mutandis, we
arrive to the desired conclusion. We record that the reason why we can extend
the arguments to arbitrary domains is the fact that we require for our functions to
vanish on the boundary. Thus we can extend them by zero to the exterior without
the requirement that the domain is such that we can construct continuous or Sobolev
extensions to its complement.
For the applications in the next section we need the following two lemmas.
f
Lemma 8.3. If Ω is a bounded open set, then L2 (Ω)∗ ⊂ M(Ω) and the inclusion map
2 ∗ f
i : L (Ω) → M(Ω) is continuous with constant depending only on the Lebesgue measure
of Ω.
Proof. Let f ∈ L2 (Ω) and let us define
Z
uf := ∆−1
Ω f (x) = GΩ (x, y)f (y) dy,
Ω
and recall that −∆uf = f in the weak sense in Ω and uf ∈ W01,2 (Ω) satisfying kuf kW 1,2 (Ω) ≤
kf kL2 (Ω) . Define σf = ∇uf and note that − div σf = f weakly in Ω. By Cauchy-Schwarz,
Z
|σf | ≤ |Ω|1/2 kσf kL2 (Ω) = |Ω|1/2 k∇uf kL2 (Ω) ≤ |Ω|1/2 kf kL2 (Ω) .
Ω
We conclude by Theorem 8.2.
f
Remark 8.4. It is clear that i : W −1,2 (Ω) → M(Ω) is an embedding.
Definition 8.5. Let Ω ⊂ Rn be a bounded open set and I ⊂ R be an open interval. If

W 1,2 (Ω) stands for the inhomogeneous Sobolev space with norm kukL2 (Ω) + k∇ukL2 (Ω) ,
we define the parabolic Sobolev space
n o
f
W(Ω × I) := u ∈ L2 (I; W 1,2 (Ω)) : ∂t u ∈ L1 (I; M(Ω))
and the semi-norm
Z
kukẆ (Ω×I) := k∇ukL2 (Ω×I) + k∂t u(·, t)kKR(Ω) dt.
I
If we equip W(Ω × I) with the norm
kukW(Ω×I) := kukL2 (Ω×I) + kukẆ (Ω×I)
it becomes a Banach space.
Lemma 8.6. Let Ω ⊂ Rn be a bounded domain so that W 1,2 (Ω) ֒→ L2 (Ω) compactly
and let I ⊂ R be an open interval. Then the parabolic Sobolev space W(Ω × I) embeds
compactly in L2 (Ω × I).
f
Proof. Set X = W 1,2 (Ω), B = L2 (Ω), Y = M(Ω). By hypothesis, the inclusion map
i : X → B is compact and, by Lemma 8.3, the inclusion map i : B∗ → Y is continuous.
Consequently, if the sequence {uj }j≥1 ⊂ W(Ω×I) is uniformly bounded in the W(Ω×I)-
norm, we can apply [Sim87, Corollary 4] to deduce that {uj }j≥1 has a subsequence that
converges in L2 (Ω × I)-norm.
9. B LOW- UPS IN TIME VARYING DOMAINS
Lemma 9.1. Let Ω ⊂ Rn+1 be an open set and Cr := Br × Ir be a cylinder of radius r

centered on ∂Ω. Set Ωr := Ω ∩ Cr and let u be a caloric function in Ω4r that vanishes
continuously on ∂Ω ∩ C4r . If ū is its extension by zero in C4r \ Ω, then the following hold:
(1) ū ∈ W 1,2 (C2r ).
(2) ∂t ū χΩ2r = ∂t u χΩ2r and ∇ū χΩ2r = ∇u χΩ2r .
(3) For all t ∈ Ir , it holds that ϕu(·, t) ∈ W01,2 ((Ωr )t ), for any ϕ ∈ Lip01 (Br ), and
ū(·, t) ∈ W 1,2 (Br ).
Proof. The proof is standard and follows from arguments similar to those in the proof of
[AMT17, Lemma 4.12]. See also Theorem 9.17 and Proposition 9.18 in [Bre11]. We omit
the details.
The following lemma is a boundary Caccioppoli-type inequality for the time derivative
of the square of a non-negative solution that is crucial for the blow-up lemmas.
Lemma 9.2. Under the same assumptions as in Lemma 9.1, if we set v := ū2 /2, it holds
that
Z
1
(9.1) k∂t v(·, t)kKR(Br ) dt .n kvkL1 (C2r ) .
Ir r
Proof. In view of Lemma 9.1, ū ∈ W 1,2 (C2r ), and, by [Wa12, Theorem 7.20], ū is a
subtemperature of H and so weakly subcaloric in C4r (ξ̄). Applying the standard parabolic
Caccioppoli’s inequality for non-negative subcaloric functions we get
Z Z
1
(9.2) |∇ū(x, t)|2 dxdt . 2 ū(x, t)2 dxdt.
Cr r C2r
Since ∂t u ∈ L2 (C2r ), then for a.e. t ∈ I2r , it holds that ∂t u(·, t) ∈ L2 (B2r ). Fix such
t ∈ Ir and set D := (Ωr )t , which is an open subset of Rn × {t}. If ϕ ∈ Lip01 (Br ), since
ϕu(·, t) ∈ W01,2 (D), for fixed ε > 0, there exists ψ ∈ Cc∞ (D) such that
kw(·, t)kW 1,2 (D) := kϕu(·, t) − ψkW 1,2 (D) < ε.
Thus,
Z Z
∂t v(z, t)ϕ(z) dz = ∂t ū(z, t)ū(z, t)ϕ(z) dz
Z Z
= ∂t ū(z, t)w(z, t) dz + ∂t ū(z, t)ψ(z) dz
Z
< εk∂t ū(z, t)kL2 (Br ) + ∂t u(z, t)ψ(z) dz
ZD
= εk∂t ū(z, t)kL2 (Br ) + ∆u(z, t)ψ(z) dz =: II1 (ε) + II2 ,
D
where we used that ∂t u = ∆u in D (pointwisely). In light of the fact that ∆u(·, t) ∈

L1loc (D) (as ∆u is continuous in Ω4r ), it is clear that
Z
II2 = − ∇z u(z, t)∇z ψ(z) dz
Z D Z
= ∇z u(z, t)∇z w(·, t) dz − ∇z u(z, t)∇z (ϕu(·, t))(z) dz
D D
Z
< εk∇ū(·, t)kL2 (Ba/16 ) − ∇z u(z, t)(∇z ϕ(z)u(z, t) + ϕ(z)∇z u(z, t)) dz
D
=: II21 (ε) + II22 .
Recalling that |ϕ| ≤ r and |∇ϕ| ≤ 1 in Br , we obtain
Z Z
2
II22 ≤ r |∇ū(z, t)| dz + ū(z, t)|∇ū(z, t)| dz.
Br Br
Therefore, taking limits as ε → 0, we infer that

Z Z Z
2
∂t v(z, t)ϕ(z) dz ≤ r |∇ū(z, t)| dz + ū(z, t)|∇ū(z, t)| dz,
Br Br
which, by Cauchy-Schwarz’s inequality and (9.2), implies that

Z
k∂t v(·, t)kKR(Br ) dt ≤ r −1 kūk2L2 (C2r ) ,
Ir
concluding the proof of (9.1).

Remark 9.3. It is important to work with v instead of ū since we want to exploit the
caloricity of u in D as subcaloricity in the whole cylinder is not enough. Thus, taking the
t-derivative of v gives us ū∂t ū, and as ū vanishes in Cr \ D, we have that ūϕ is supported
in D and vanishes continuously on ∂D. This way the domain of integration is restricted
to the set D where u is caloric, which wouldn’t have been achieved if we had just used
∂t ū. Of course, the same proof works for u in D and ϕt ∈ Lip01 (D) (ϕ depends on t
now), but it is not helpful for the blow-up lemmas. Indeed, in that case, we would have
a uniformly bounded sequence of functions in a sequence of Sobolev spaces and thus, the
compact embedding in L2 would not be applicable.
Remark R9.4. The need for the introduction of the Sobolev space W(Br × Ir ) comes from
the term Br |∇ū|2 ϕ we obtain at the end. If we did not have an L∞ bound for ϕ but rather
an Lp bound for large p, one would hope to apply Hölder’s inequality and then use higher
integrability estimates for the gradient of ū and Caccioppoli. Although, as ū is not caloric
but merely subcaloric in the whole cylinder, such estimates are out of the question.
Now, we are ready to proceed to the first blow-up lemma.

Lemma 9.5. Let Ω ⊂ Rn+1 be an open set which is quasi-regular for H and regular for
p̄
H ∗ . Let also ω = ωΩ be the associated caloric measure with pole at p̄ ∈ Ω. Assume that
F ⊂ P ∗ Ω ∩ S ′ Ω is compact, ξ̄j is a sequence in F , and there exists rj → 0 and cj > 0
such that C(ξ̄j , arj /4) ∩ ∂Ω = C(ξ̄j , arj /4) ∩ P ∗ Ω, where a ∈ (0, 1/2) is as in Lemma
3.8, and
ωj := cj Tξ̄j ,rj [ω] ⇀ ω∞
for some non-zero Radon measure ω∞ . Let us also assume that there exists a subsequence
of rj and a constant c > 0 so that

(9.3) Cap E(ξ̄j ; rj2 ) \ Ω ≥ c rjn .
If u = GΩ (p̄, ·) in Ω and u = 0 in Rn+1 \ Ω, let us denote

uj (x̄) := cj u δrj x̄ + ξ̄j rjn .
and
Cr := Cr (0̄) = Br (0) × (−r 2 , r 2 ) =: Br × Ir , r > 0.
Then, if α = a/16 ∈ (0, 1/32), there exists R > 0, a subsequence of {rj }j and a non-
negative function u∞ ∈ L2 (CαR ) such that uj → u∞ in L2 (CαR )-norm. Moreover u∞ is
adjoint caloric in CαR ∩ {u∞ > 0},

(9.4) ku∞ kL2 (CαR ) .a,R ω∞ CM R
and Z Z
ϕ dω∞ = u∞ Hϕ, for all ϕ ∈ Cc∞ (CαR ).
Rn+1
Proof. For simplicity, we will only prove the lemma assuming ξ¯j ≡ ξ̄ for some ξ̄ ∈ P ∗ Ω ∩
SΩ, since the proof of the general case is analogous. Let us recall that for ξ¯0 ∈ Rn+1 ,
r > 0, and a > 0 we denote
Rb+ (ξ̄0 ; r) = Br (ξ0 ) × (t0 − (ar)2 /2, t0 + r 2 ).
a
ba+ (ξ̄0 ; r).
and it holds that Car/2 (ξ̄0 ) ⊂ R
As (9.3) is satisfied, by Lemma 3.8 we have that there exists a ∈ (0, 1/2] such that

Cap Ra− (ξ̄; rj ) \ Ω & crj2 ,
and so, by (3.33), it holds that

ω z̄ CM rj (ξ̄) ≥ c > 0,
for all z̄ ∈ Carj (ξ̄) ⊂ R̂a+ (ξ̄; rj ). Since ω∞ 6= 0, there exists R > 0 such that ω∞ (CR ) > 0.
Without loss of generality we may assume that R = a/8. By passing to a subsequence, if
necessary, we may assume that p̄ ∈ Ω\Carj (ξ̄) and combining the latter bound with Lemma
3.21 for arj /2 and M ′ = 2M/a, we obtain

(9.5) ω p̄ CM rj (ξ̄) = ω p̄ CM ′ arj (ξ̄) & (arj )n u(ȳ), for ȳ ∈ Carj /8 (ξ̄).
In particular,

(9.6) lim sup sup uj (ȳ) . a−n lim sup ωj CM (0̄) . ω∞ CM (0̄) .
j→∞ Ca/8 (0̄) j→∞
Recall that G(p̄, ·) is adjoint caloric in Ω\{p̄} and Ω is regular for H ∗ , and so G(p̄, ·) is in
C 2,1 (Ω\{p̄})∩C(Ω∪∂e∗ Ω\{p̄}) vanishing on ∂e∗ Ω. Thus, for j large, it holds that G(p̄, ·) is
adjoint caloric in Ω∩Carj /4 (ξ̄) and, by assumption, ∂Ω∩Carj /4 (ξ̄) = P ∗ Ω∩Carj /4 (ξ̄). By
Lemmas 9.1 and 9.2, and the standard parabolic Caccioppoli’s inequality for non-negative
subcaloric functions, along with a simple rescaling argument and (9.6), we infer that there
exists j0 > 1 large such that for j ≥ j0 , if vj = u2j /2,
Z Z
2
|∇vj (x, t)| dxdt + k∂t vj (·, t)kKR(Ba/16 ) dt
Ca/16 Ia/16
2 Z
(9.7) .a,n ω∞ CM (0̄) + 1 uj (x, t)2 dxdt
Ca/8
2 2
.a,n ω∞ CM (0̄) 1 + ω∞ CM (0̄) .
We apply Lemma 8.6 to find a subsequence of {vj }j≥j0 and a non-negative function
v∞ ∈ L2 (Ca/16 ) such that
vj → v∞ strongly in L2 (Ca/16 ) and a.e. in Ca/16 .
Note that if z̄ ∈ Ca/16 is so that v∞ (z̄) = 0, then trivially, uj (z̄) → 0. If z̄ ∈ Ca/16 is so
√
that v∞ (z̄) > 0, then, if we set u∞ = 2v∞ , we have that
|u2j (z̄) − 2v∞ (z̄)| 2|vj (z̄) − v∞ (z̄)|
|uj (z̄) − u∞ (z̄)| = ≤ → 0, as j → ∞.
uj (z̄) + u∞ (z̄) u∞ (z̄)
Thus, uj → u∞ a.e. in Ca/16 and using (9.6) to get a uniform L2 bound on Ca/16 ,
we deduce that uj → u∞ weakly in L2 (Ca/16 ) and u∞ satisfies (9.4). Moreover, by
Cauchy–Schwarz, it holds that
Z Z Z

2 2
u − u∞ ≤ |u2j − u2∞ | .a,n kvj − v∞ kL2 (Ca/16 ) → 0, as j → ∞.
Ca/16 j Ca/16 Ca/16
Hence, kuj kL2 (Ca/16 ) → ku∞ kL2 (Ca/16 ) and since uj → u∞ weakly in L2 (Ca/16 ), by the
Radon-Riesz Theorem, we conclude that uj → u∞ in L2 (Ca/16 )-norm.
Now, a change of variables gives that

Z Z
ϕ dωj = uj Hϕ, ϕ ∈ Cc∞ (Ca/16 )
which, if we take limits as j → ∞, entails

Z Z
ϕ dω∞ = u∞ Hϕ, ϕ ∈ Cc∞ (Ca/16 ).
In order to complete the proof of the lemma, it suffices to choose α = a/16.
The previous lemma applies to the following “two-phase” blow-up lemma.

Lemma 9.6. Let Ω± be two disjoint open sets in Rn+1 which are quasi-regular for H and
regular for H ∗ and let ω ± be the respective caloric measures with poles p̄± ∈ Ω± . Assume
that F ⊂ P ∗ Ω+ ∩ P ∗ Ω− ∩ S ′ Ω+ ∩ S ′ Ω− is compact, ξ̄j is a sequence in F and that Ω+
and Ω− satisfy the joint TBCPC. Let us also assume that there exists rj → 0, cj > 0, and a
constant c > 0, such that C(ξ̄j , arj /4) ∩ ∂Ω± = C(ξ̄j , arj /4) ∩ P ∗ Ω± , where a ∈ (0, 1/2)
is as in Lemma 3.8, and
ωj+ := cj Tξ̄j ,rj [ω + ] ⇀ ω∞
+
and
(9.8) ωj− := cj Tξ̄j ,rj [ω − ] ⇀ c ω∞
+ −
=: ω∞ .
Let u± := GΩ± (p̄± , ·) on Ω± , u± ≡ 0 on Rn+1 \ Ω and denote

u±j (x̄) := cj u
±
δrj x̄ + ξ̄j rjn .
The following properties hold:
(a) There exists α ∈ (0, 1/16), R > 0, a subsequence of rj , u± 2
∞ ∈ L (CαR ), and
u∞ ∈ L2 (CαR ), a non-zero adjoint caloric function in CαR , such that u± j converge
2 ± + −1 − 2
in L (CαR )-norm to u∞ and uj := uj − c uj converges in L (CαR )-norm to
u∞ = u+ −1 −
∞ − c u∞ .
(b) The function u∞ extends to an adjoint caloric function in Rn+1 and it holds that

(9.9) ku∞ kL2 (Cr ) .a r −n ω∞
+
CM r/a for all r > 0,
and for any ϕ ∈ Cc∞ (Rn+1 ),
Z Z Z
+ + 1
(9.10) ϕ dω∞ = u∞ Hϕ = |u∞ | Hϕ.
2 Rn+1
(c) We have that
+
supp ω∞ = ∂Ω± ±
∞ := ∂{u∞ > 0} = {u∞ = 0} =: Σ∞ ,
+ = −∂ + u dσ
and dω∞ νt ∞ Σ∞ , where σΣ∞ stands for the parabolic surface measure on
+
Σ∞ and νt is the measure-theoretic outer unit normal on the time-slice (Ω+
∞ )t .
Proof. As in the proof of the previous lemma, we assume for simplicity that ξ̄j ≡ ξ¯ ∈
P ∗ Ω+ ∩ P ∗ Ω− for all j and ω∞+ (C + −
a/16 ) > 0. Since the domains Ω and Ω are disjoint,
the subadditivity property for the thermal capacity [Wa12, Theorem 7.45-(b)] entails

Cap E(ξ̄; rj2 ) \ Ω+ + Cap E(ξ̄; rj2 ) \ Ω− ≥ Cap E(ξ̄; rj2 ) , j ≥ 0,
and so, possibly after passing to a subsequence, without loss of generality, we can assume
that
2)

+
Cap E( ξ̄; rj
Cap E(ξ̄; rj2 ) \ Ω ≥ ≈ rjn , j ≥ 0.
2
Moreover, as Ω+ and Ω− satisfy the joint TBCPC condition, there exists q subsequence
such that

Cap E(ξ̄; rj2 ) \ Ω− & rjn j ≥ 0.
providing us with a common sequence rj such that Lemma 9.5 applies to both ω + and ω − .
Thus, for α = a/16 ∈ (0, 1/32) there exist two adjoint subcaloric functions u±
∞ such that
u±
j → u ± in L2 (C ) and
∞ α
Z Z
±
(9.11) ϕ dω∞ = u± ∞ Hϕ, ϕ ∈ Cc∞ (Cα ).
In particular, uj converges in L2 (Cα )-norm to the function u∞ := u+ −1 −

∞ − c u∞ and, for
∞
ϕ ∈ Cc (Cα ),
Z Z Z Z Z
+ −1 − (9.11) + −1 − (9.8)
(9.12) u∞ Hϕ = u∞ Hϕ − c u∞ Hϕ = ϕ dω∞ − c ϕ dω∞ = 0.
which proves that u∞ is adjoint caloric in Cα .

Furthermore, as supp u+ −
j ∩ supp uj = ∅ for every j, it holds that
Z Z
+ −
u∞ u∞ = lim u+ −
j uj = 0,
Cα j→∞ Cα
which implies
(9.13) Ln+1 (supp u+ −
∞ ∩ supp u∞ ∩ Cα ) = 0.
As a result of (9.11), {u+

∞ > 0} ∩ Cα is a set of positive L
n+1 -measure and, thus, u
∞ 6≡ 0.
Another consequence is that
(9.14) u+
∞ = u∞ χ{u∞ >0} and u−
∞ = −c
−1
u∞ χ{u∞ <0} .
In particular, u±
∞ are continuous in Cα because u∞ is an adjoint caloric function.
Let us prove (b). The same argument as the one in the proof Lemma 9.5 shows that,
given K ≥ 1,

u± ±
j (ȳ) .a ωj CKM (0̄) , ȳ ∈ CKa/8 (0̄)
with the implicit constant independent of j. Moreover

lim sup ωj± CKM (0̄) ≤ ω∞ ±
CKM (0̄) , for all K ≥ 1.
j
So arguing again as we did in the proof of Lemma 9.5, we get that for any fixed K =
1, 2, . . ., the sequence uj admits a converging subsequence in L2 (CaK/4 ). By a standard
diagonalization argument we can show that u∞ extends to a caloric function in Rn+1 satis-
fying (9.9). As it is easy to see that (9.10) holds, the proof of (b) is concluded.
Let us turn our attention to the proof of (c). Let x̄ ∈ supp ω∞ + and assume that there
exists a cylinder Cr (ȳ) such that x̄ ∈ Cr (ȳ) ⊂ {u+∞ > 0}. In particular, (9.14) implies that
Cr (ȳ) ⊂ {u− ∞
∞ = 0}. Then, if ϕ ∈ Cc (Cr (ȳ)) such that ϕ(x̄) > 0, since u∞ is globally
adjoint caloric, we have
Z Z Z
+ +
(9.15) 0 < ϕ dω∞ = u∞ Hϕ = u∞ Hϕ = 0,
so we get a contradiction. A similar argument shows that Cr (ȳ) 6⊂ {u+ ∞ = 0} for any
cylinder such that x̄ ∈ Cr (ȳ). So Cr (ȳ) ∩ {u+
∞ > 0} 6
= ∅ and C r (ȳ) ∩ {u+ 6 ∅,
∞ = 0} =
which shows that
+
(9.16) supp ω∞ ⊂ ∂{u±
∞ > 0}.
We now claim that

(9.17) ∂{u± +
∞ > 0} ⊂ supp ω∞ .
To prove this, we take x̄ = (x, t) ∈ ∂{u+ ∞ > 0} and assume, by contradiction, that x̄ ∈
Cr (ȳ) ⊂ (Rn+1 \ supp ω∞ + ) for some ȳ ∈ (supp ω + )c and r > 0. In particular,
∞
Z Z
(9.10)
u+∞ Hϕ = ϕ dω∞+
=0 for all ϕ ∈ Cc∞ Cr (ȳ) ,
which gives that u+ +

∞ is adjoint caloric in Cr (ȳ). Since x̄ ∈ ∂{u > 0} ∩ Cr (ȳ), we have
+
u∞ (x̄) = 0. Hence, the strong minimum principle for adjoint caloric functions (which
readily follows from [Wa12, Theorem 3.11]) gives that u+ ∞ ≡ 0 on {(ζ, θ) ∈ Cr (ȳ) :
+
θ < t}. In particular, u∞ = u∞ = 0 in an open cylinder Cθ (ζ̄), where the first equality
follows from (9.10). So the fact that u∞ extends to a globally adjoint caloric function and
the unique continuation principle [Po96, Theorem 1.2] entail u∞ ≡ 0 on the whole Rn+1 .
This contradicts (a) and proves the claim (9.17).
An immediate consequence of (9.16) is that
+
supp ω∞ ⊂ {u+ −
∞ = 0} ∩ {u∞ = 0} ⊂ {u∞ = 0}.
In order to prove the converse inclusion, we observe that for x̄ ∈ Rn+1 such that u∞ (x̄) =
0, (9.14) implies that u+ −
∞ (x̄) = u∞ (x̄) = 0. Moreover, by the unique continuation, u∞
cannot vanish in any open cylinder containing x̄, so either u+ −
∞ or u∞ must be positive in
that cylinder. This in turn implies
(9.17)
{u∞ = 0} ⊂ ∂{u+ − +
∞ > 0} ∪ ∂{u∞ > 0} ⊂ supp ω∞ .
Lemma 6.4 concludes the proof.
Before presenting the blow-up result that we need in order to prove the second part of
Theorem I, we recall that the TFCDC condition is invariant under parabolic scaling (see
Section 3). Moreover, if Ω is a domain that satisfies the TFCDC as in 3.28, ξ̄ ∈ SΩ, and
ρ > 0, then, if we denote Ω̃ := Tξ,ρ [Ω], we have that S Ω̃ ⊂ ∂e∗ Ω̃.
The next lemma lists the main properties of the blow-ups of caloric measure and the
adjoint Green’s function in a domain satisfying the assumptions of Lemma 9.5 that has the
TFCDC instead of just being regular for H ∗ . Our proof is inspired from the one of the
corresponding result for harmonic measure in [AMT17] and although it is the essentially
the same for items (a)-(c), the proof of (d) is completely different.
Lemma 9.7. Let Ω ⊂ Rn+1 be an open set which is quasi-regular for H and satisfies
the TFCDC, and let F ⊂ P ∗ Ω ∩ SΩ be compact and {ξ̄j }j ⊂ F . We denote by ω the
caloric measure for Ω with pole at p̄ ∈ Ω, and assume that there exists cj > 0 and rj → 0
such that ωj = cj Tξ̄j ,rj [ω] ⇀ ω∞ for some non-zero Radon measure ω∞ . If we denote
Ωj := Tξ̄j ,rj (Ω) and assume that there is c > 0 so that

(9.18) Cap E(ξ̄j ; rj2 ) \ Ω ≥ c rjn , j = 1, 2, . . .
then there is a subsequence of {rj }j≥1 and a closed set Σ ⊂ Rn+1 such that
(a) For all R > 0 sufficiently large, ∂Ωj ∩ CR (0̄) → Σ ∩ CR (0̄) in the Hausdorff metric.
(b) Rn+1 \ Σ = Ω∞ ∪ Ω b ∞ where Ω∞ is a nonempty open set and Ω b ∞ is also open but
possibly empty. Moreover, there exists λ > 1 so that, if Cr (x̄) is a cylinder such that
Cr (x̄) ⊂ Ω∞ , then Cλr (x̄) is contained in Ωj for all j large enough.
(c) supp ω∞ ⊂ Σ.
(d) If we set u(x̄) = GΩ (p̄+ , x̄) on Ω and u ≡ 0 on (Ω)c , and define

uj (x̄) = cj rjn u δrj (x̄) + ξ̄j ,
then the sequence uj converges uniformly on compact subsets of Rn+1 to a non-zero
function u∞ which is continuous in Rn+1 , adjoint caloric in Ω∞ , and vanishes in
(Ω∞ )c . Moreover if a ∈ (0, 1/2) and M > 1 are the constants obtained in Lemmas 3.8
and 3.15 respectively, then for x̄ ∈ Σ and r > 0, it holds that

(9.19) u∞ (ȳ) .a r −n ω∞ C4a−1 M r (x̄) , ȳ ∈ Cr (x̄) ∩ Ω+
∞,
and
Z Z
(9.20) ϕ dω∞ = u∞ Hϕ, for any ϕ ∈ Cc∞ (Rn+1 ).
Rn+1
Proof. For simplicity, we assume K = {ξ̄}, as the general case can be proved similarly.
Proof of (a): Let R > 0 and observe that 0̄ ∈ ∂Ωj for all j ≥ 1, which entails ∂Ωj ∩

CR (0̄) 6= ∅. The Hausdorff distance is a metric on the collection C CR (0̄) of all closed

subsets of CR (0̄). Since CR (0̄) is compact, C CR (0̄) , dH is compact too. Thus, after
passing to a subsequence, CR (0̄) ∩ ∂Ωj → CR (0̄) ∩ Σ in the Hausdorff distance sense for

some Σ ∈ C CR (0̄) . A standard diagonalization argument concludes the proof of (a).
Proof of (b): We will first prove that there exists a parabolic cylinder Cρ (x̄′ ) which
is contained in all Ωj , for j large enough. Arguing by R contradiction we assume not. Let
∞
ϕ ∈ Cc (R n+1 +
) be a non-negative function such that ϕ dω∞ 6= 0 and supp ϕ ⊂ CK (0̄)
for some K > 0. Hence, there exists x̄0 belonging to CK (0̄) ∩ supp ω∞ , and our coun-
terassumption implies that

(9.21) ρj := sup distp (x̄, (Ωj )c ) : x̄ ∈ C2K (0̄) → 0, as j → ∞.
We denote by ζ̄j (x̄) ∈ (Ωj )c a point which realizes distp (x̄, (Ωj )c ). In particular, since
ρj → 0 and 0 ∈ ∂Ωj for all j, we have that kx̄ − ζ̄j (x̄)k ≤ ρj < 2K, for j large enough
and x̄ ∈ C2K (0̄). Moreover, kx̄ − x̄0 k ≤ kx̄k + kx̄0 k < 3K.
Observe that Ωj satisfies the TFCDC with the same parameters as Ω because of Lemma
3.14. Moreover, ζ̄j (x̄) ∈ SΩj for j big enough because Tmin (Ωj ) = (Tmin (Ω) − τ )/rj2 →
−∞ and Tmax (Ωj ) = (Tmax (Ω) − τ )/rj2 → +∞ as j → ∞. In fact, ζ̄j (x̄) ∈ ∂e∗ Ωj (see
the discussion before the statement of this lemma).
Note that for x̄ ∈ C2K (0̄), we have C2K ζ̄j (x̄) ⊂ C4K (0̄) ∩ C5K (x̄0 ). If we consider
j large enough so that p̄ 6∈ C16K (0̄), then if we combine (3.40), (3.33), and (9.18), we have
that for ȳ ∈ C4K (0̄),

uj (ȳ) = cj rjn u δrj ȳ + ξ¯
(9.22)
. cj rjn (4Krj )−n ω δrj C4KM (0̄) + ξ̄ ≈K ωj C4KM (0̄) .
Remark that M depends on a and is chosen so that (3.34) holds. Hence, using the γ-
Hölder continuity at the boundary for uj in Ωj (that holds with constants not depending on
j because of Lemma 3.14), we infer that
Z Z Z
(3.39) kx̄ − ζ̄j (x̄)k γ
0 < ϕ dωj = uj Hϕ . |Hϕ| sup u+
j dx̄
Ωj Ωj C3K (ζ̄j (x̄))∩Ωj 2K
Z
(9.22) ρj γ
.K ωj C4KM (0̄) |Hϕ| .K,ϕ ωj C4KM (0̄) ρj γ .
2K
So taking the limit in the previous inequalities as j → ∞ we obtain
Z
(9.21)
0 < ϕ dω∞ .a,K,ϕ lim sup ργj ωj C4KM (0̄) ≤ ω∞ C4KM (0̄) lim ργ = 0,
j→∞ j→0
which is a contradiction. Thus, after passing to a subsequence, if necessary, there exists a

cylinder Cρ (x̄′ ) ⊂ Ωj for j big enough.
For the the construction of Ω∞ and, once we define Ω b ∞ := Rn+1 \Ω∞ , for the remaining
part of the proof of (b), we refer to [AMT17, p. 2143]. For future reference, we remark that
Ω∞ is built as
[
(9.23) Ω∞ = Cr (x̄),
Cr (x̄)∈D
where D is the collection of all open cylinders Cr (x̄) such that r is rational, x̄ has rational
coordinates, Cr (x̄) ⊂ Rn+1 \ Σ and Cr (x̄) is contained in all but finitely many Ωj , for a
proper subsequence.
Proof of (c): If Cr (x̄) ⊂ Cr (x̄) ⊂ Rn+1 \ Σ, we have that

ω∞ Cr (x̄) ≤ lim inf ωj Cr (x̄) ≤ lim inf ωj Rn+1 \ ∂Ωj = 0,
j→∞ j→∞
which implies x̄ 6∈ supp ω∞ .

Proof of (d): Let us show that uj converges uniformly on compact subsets of Rn+1 to
a continuous function u∞ . Recall that 0̄ ∈ ∂Ωj for j ≥ 1, and let j large enough so that
p̄ ∈ Ω \ C400Krj (ξ̄). We can argue as in (9.22) and, for K > 0 and 4 ≤ λ < 100, we get
that for j large,
(9.24)
uj (ȳ) . (λK)−n ωj CλKM (0̄) . (λK)−n ω∞ CλKM (0̄) , if ȳ ∈ CλK (0̄) ∩ Ωj ,

where the last inequality holds because lim supj ωj (CλKM (0̄)) ≤ ω∞ CλKM (0̄) . In par-
ticular, {uj }j is uniformly bounded on CλK (0̄) for j large.
We will show that it is equicontinuous on CK (0̄) for j large. To do so, we split into cases
depending on the positions of x̄, ȳ ∈ CK (0̄).
Case 1. If x̄, ȳ ∈ CK (0̄) \ Ωj , then trivially |uj (x̄) − uj (ȳ)| = 0.
Case 2. Let us study the case x̄ = (x, t) ∈ Ωj and ȳ = (y, s) ∈ CK (0̄) \ Ωj . If t ≤ s,
we denote by γ(x̄, ȳ) the (possibly degenerate) parabola with vertex at x̄, whose axis is
parallel to the time-axis so that ȳ ∈ γ(x̄, ȳ). If z̄ = (z, u) ∈ γ(x̄, ȳ), then, by construction,
|z − x| ≤ |y − x| and |u − t| ≤ |s − t|. Hence
(9.25) kz̄ − x̄k = max{|z − x|, |u − t|1/2 } ≤ max{|y − x|, |s − t|1/2 } = kȳ − x̄k.
Similarly, one can show that kȳ − z̄k ≤ kȳ − x̄k.
If s < t, we set γ̃(x̄, ȳ) ≡ γ(ȳ, x̄), where γ(·, ·) is defined above. The parabola γ̃(x̄, ȳ)
meets ∂Ωj at a point, say z̄. Since we choose j big enough, (3.39) applies (with constant
independent on j because the TFCDC condition holds with uniform constants for Ωj ) and
we have
|uj (x̄) − uj (ȳ)| = uj (x) .K kx̄ − z̄kα sup uj ≤ kx̄ − z̄kα sup uj
C2K (z̄) C3K (0̄)
(9.24) (9.25)
.K kx̄ − z̄kα ω∞ C3KM (0̄) .Ke kx̄ − ȳkα .
n+1 \ Ω
Case 3. Let us assume that x̄, ȳ ∈ Ωj ∩ CK (0̄) and that there exists a point z̄ ∈ R j
such that dist z̄, γ(x̄, ȳ) ≤ kx̄ − ȳk/3. Without loss of generality, we assume uj (x̄) ≤
uj (ȳ). If ζ̄ ∈ γ(x̄, ȳ) denotes a point such that kz̄ − ζ̄k = dist z̄, γ(x̄, ȳ) , we have
1 4 8
(9.26) kȳ − z̄k ≤ kȳ − ζ̄k + kζ̄ − z̄k ≤ kx̄ − ȳk + kx̄ − yk = kx̄ − ȳk ≤ K.
3 3 3
Observe that ȳ ∈ C3K (z̄) ⊂ C5K (0̄). Since we have p̄ ∈ Ω \ C400Krj (ξ̄), we can apply
(3.39) and get

|uj (x̄) − uj (ȳ)| ≤ 2uj (ȳ) . sup uj kȳ − z̄kα ≤ sup uj kȳ − z̄kα
C3K (z̄) C5K (0̄)
(9.24) (9.26)
.K ω∞ C5KM kȳ − z̄kα .K kx̄ − ȳkα .

Case 4. Suppose that x̄, ȳ ∈ Ωj ∩ CK and that dist z̄, γ(x̄, ȳ) > kx̄ − ȳk/3 for all
z̄ ∈ Rn+1 \ Ωj . We denote δ := kx̄ − ȳk/100 and we cover γ(x̄, ȳ) with N cylinders
of the form Cδ (ȳk ), for ȳk ∈ γ(x̄, ȳ), k = 1, . . . , N . The cylinders are chosen so that
Cδ (ȳk+1 ) ∩ Cδ (ȳk ) 6= ∅, x̄ ∈ Cδ (ȳ1 ) and ȳ ∈ Cδ (ȳN ). For i = 1, . . . , N − 1 we choose
x̄i ∈ Cδ (ȳi+1 ) ∩ Cδ (ȳi ) and define x̄0 = x̄ and x̄N = ȳ. Remark that N is independent of
δ and j. Since uj is caloric in Ωj ∩ CK (0̄), we can use the interior Hölder continuity (see
e.g. [Lieb96, Theorem 6.29]) and write
N
X N
X
|uj (x̄) − uj (ȳ)| ≤ |uj (x̄i+1 ) − uj (x̄i )| . kuj kL∞ (C4δ (ȳi )) kx̄k+1 − x̄k kβ
i=0 k=0
(9.24)
. kuj kL∞ (C2K (0̄)) kx̄ − ȳkβ .K ω∞ C2KM (0̄) kx̄ − ȳkβ ,
where β ∈ (0, 1) and the implicit constants are independent of j.
Combining the four cases above, after passing to a subsequence, we obtain that uj is
equicontinuous in CK for each K ≥ 1. Hence, we can apply Ascoli-Arzelà’s theorem
and a standard diagonalization argument to get that uj has a subsequence that converges
uniformly on compact subsets of Rn+1 to a continuous function u∞ . Moreover, for x̄ ∈ Σ,
r > 0, and ȳ ∈ Cr (x̄) ∩ Ω∞ , we have
u∞ (ȳ) = lim uj (ȳ) .a lim sup r −n ωj (C4M a−1 (x̄)) ≤ r −n ω∞ (C4M a−1 (x̄)) < ∞,
j→∞ j
which gives (9.19). It remains to prove (9.20). To this end, for ϕ ∈ Cc∞ (Rn+1 ) it holds
Z Z Z Z
−2

ϕ dωj = cj ϕ ◦ Tξ̄,rj dω = cj H ϕ ◦ Tξ̄,rj u = cj rj (Hϕ) ◦ Tξ̄,rj u
Z Z

(9.27) n
= cj rj ¯
Hϕ(ξ̄) u δrj (ȳ) + ξ dȳ = uj Hϕ.
Taking limits as j → ∞ in (9.27), in light of the fact that supp ω∞ ∩ Ω∞ ⊂ Σ ∩ Ω∞

by (c) and that uj → u∞ locally uniformly in Rn+1 , we have that (9.20) holds. It is
straightforward to see that u∞ is adjoint caloric in Ω∞ .
The function Ru∞ is not identically zero. Indeed, if ϕ ∈ Cc∞ (Rn+1 ) is a non-negative
function so that ϕ dω∞ + > 0 and plug it in (9.20), we obtain
Z Z
0 < ϕ dω∞ = u∞ Hϕ.
This readily implies that u∞ 6≡ 0 in Rn+1 .

In order to finish the proof of (d), we claim that u∞ ≡ 0 on Rn+1 \ Ω∞ . Assume that
there is x̄ ∈ Rn+1 \ Ω∞ so that u∞ (x̄) > 0. By the continuity of u∞ , there exists δ > 0
such that u∞ (ȳ) > u∞ (x̄)/2 for all ȳ ∈ Cδ (x̄). The local uniform convergence of uj to u
implies that uj > u∞ (x̄)/4 on Cδ (x̄) for all j big enough. Thus, since uj vanishes outside
of Ωj , we have that Cδ (x̄) ⊂ Ωj for j large. In particular, it holds that Cδ/2 (x̄) ⊂ Rn+1 \ Σ,
so Cδ/2 (x̄) ⊂ Ω∞ by the construction (9.23) and we can conclude that x̄ ∈ Ω∞ , which is a
contradiction. This proves our claim and finishes the proof of the lemma.
The next lemma studies the two-phase version of the previous result for domains that
satisfy the joint TBCPC. Although this proof is inspired by the one of [AMT17, Lemma
5.9], there are several differences as well. Let us remark that an important feature of our
proof is that we do not assume the domains to be complementary (unlike the corresponding
lemma in [AMT17]), which gives us significant freedom in the applications in Section 10
but also requires additional care in the proof.
Lemma 9.8. Let Ω± ⊂ Rn+1 be two disjoint open sets that are quasi-regular for H and
satisfy the joint TBCPC and the TFCDC. Let F ⊂ P ∗ Ω+ ∩P ∗ Ω− ∩SΩ+ ∩SΩ− be compact
and {ξ̄j }j ⊂ F . We denote by ω ± the caloric measure for Ω± with pole at p̄± ∈ Ω± , and
assume that there exists cj > 0, c > 0 and rj → 0 such that
ωj+ := cj Tξ̄j ,rj [ω + ] ⇀ ω∞
+
and
ωj− := cj Tξ̄j ,rj [ω + ] ⇀ cω∞
+ −
=: ω∞ .
We set u± (x̄) = GΩ± (p̄± , x̄) on Ω± and u± ≡ 0 on (Ω± )c , and define

u± (x̄) = cj r n u± δr (x̄) + ξ¯j .
j j j
Then there exists a subsequence of {rj }j≥1 , u± ± b± ±

∞ , Ω∞ , Ω∞ , and Σ such that the properties
(a)-(d) in Lemma 9.7 hold; moreover, if uj := u+ −1 − +
j − c uj , then uj → u∞ := u∞ − c u∞
−1 −
uniformly on compact subsets of Rn+1 and the following properties hold:

(a) u∞ is an adjoint caloric function on Rn+1 .
(b) Ωb± ∓ + −
∞ = Ω∞ and Σ = Σ =: Σ. Moreover Σ = {u∞ = 0}, with u∞ > 0 on Ω∞ and
+
u∞ < 0 on Ω∞ . −
(c) Σ is a set of Euclidean Hausdorff dimension n and

+
(9.28) dω∞ = −∂νt u∞ dσ∂Ω+
∞
,
where σ∂Ω+ ∞
stands for the parabolic surface measure on ∂Ω+
∞ and νt is the measure-
theoretic outer unit normal on the time-slice (Ω+ )
∞ t .
Proof. Let us assume for brevity that ξ̄j ≡ ξ.¯ Given Ω± as in the statement, by the same
pigeonholing argument as in Lemma 9.6 and in light of the joint TBCPC property, we can
find a subsequence of rj such that

(9.29) Cap E(ξ̄; rj2 ) \ Ω± & rjn , j ≥ 0.
So (9.18) holds for both Ω+ and Ω− and we can apply Lemma 9.7 to find u± ± b±
∞ , Ω∞ , Ω∞ ,
±
and Σ satisfying the properties (a)-(d) of Lemma 9.7.
Proof of (a): Let ϕ ∈ Cc∞ (Rn+1 ). By hypothesis we have that ωj+ ⇀ ω∞ + and ω − ⇀
j
cω∞+ , and thus, by the local uniform convergence of u± on Rn+1 , we can find a subsequence
j
of {rj }j≥1 such that
Z Z Z Z
+ −1 − + −1 −
u∞ Hϕ = lim (uj − c uj ) Hϕ = lim ϕ dωj − c ϕ dωj
j→∞ j→∞
Z Z Z Z
+
= ϕ dω∞ − c−1 ϕ dω∞ −
= ϕ dω∞ +
− ϕ dω∞ +
= 0.
So u∞ is a weakly adjoint caloric function in Rn+1 and Lemma 3.1 applies.

Proof of (b): Let us first prove that Rn+1 \(Ω+ −
∞ ∪ Ω∞ ) = {u∞ = 0}. By Lemma 9.7-(d)
we have that the functions u± + c − c
∞ simultaneously vanish on (Ω∞ ) ∩ (Ω∞ ) = R
n+1 \ (Ω+ ∪
∞
−
Ω∞ ). Thus, R n+1 + − + −
\ (Ω∞ ∪ Ω∞ ) ⊂ {u∞ = 0} ∩ {u∞ = 0} ⊂ {u∞ = 0}.
In order to prove the converse inclusion, we first observe that u∞ is non-zero since by
Lemma 9.7-(d) it holds that u± ±
∞ 6≡ 0 and vanish outside Ω∞ . By construction, we have that
u± ± + +
∞ ≥ 0 on Ω∞ . Let us assume by contradiction that u (x̄) = 0 for some x̄ ∈ Ω∞ and let
E ∗ (x̄; r) := {(y, s) ∈ Rn+1 : (y, t − s) ∈ E(x̄, r)}

be a reflected heat ball so that E ∗ (x̄; r) ⊂ Ω+ ∞ . The mean value theorem for adjoint caloric
functions (that readily follows from the mean value theorem for caloric functions [Wa12,
Theorem 1.16]) gives that
Z
|x − y|2
0 = u∞ (x̄) = (4πr)−n/2 u (y, s) dy ds
2 ∞
E ∗ (x̄;r) 4(t − s)
Z
−n/2 |x − y|2 +
= (4πr) u (y, s) dy ds,
2 ∞
E ∗ (x̄;r) 4(t − s)
which implies that u∞ = u+ ∗

∞ ≡ 0 on E (x̄; r). In particular, u∞ has infinite order of
vanishing at all the points of E ∗ (x̄; r) in the sense of [Po96, p. 522]. Hence, by the unique
continuation principle for globally adjoint caloric functions [Po96, Theorem 1.2], we de-
duce that u∞ ≡ 0 on Rn+1 , which is a contradiction. The same argument can be applied to
show that u− − ±
∞ > 0 on Ω∞ . So since u∞ = 0 in R
n+1 \ Ω± , we get u
∞
+
∞ 6= 0 in Ω∞ ∪ Ω∞
−
or equivalently,
{u∞ = 0} ⊂ (Rn+1 \ Ω+
∞ ) ∩ (R
n+1
\ Ω−
∞) = R
n+1
\ (Ω+ −
∞ ∪ Ω∞ ).
Let us remark that, by construction, Rn+1 = Ω± b± ±

∞ ∪ Ω∞ ∪ Σ , where the unions are
disjoint. Hence

(9.30) b− = Ω
Ω b − ∩ Ω+ ∪ Ω b− ∩ Ω
b+ ∪ Ω b − ∩ Σ+ .
∞ ∞ ∞ ∞ ∞ ∞
We claim that

b−
Ω b+ b−
∞ ∩ Ω∞ ∪ Ω∞ ∩ Σ
+
= ∅.
b+
Firstly, let us show that Ω b− b+ b−
∞ ∩ Ω∞ = ∅. Indeed, since both sets are open, if x̄ ∈ Ω∞ ∩ Ω∞ ,
there exists ρ > 0 such that Cρ (x̄) ⊂ Ω b ∩Ω
+ b ⊂ (R
− n+1 +
\ Ω∞ ) ∩ (R n+1 \ Ω−
∞ ∞ ∞) ⊂
R n+1 + −
\ (Ω∞ ∪ Ω∞ ) ⊂ {u∞ = 0}. Hence, by [Po96, Theorem 1.2] we have that u∞ ≡ 0,
which is a contradiction.
Secondly, we prove that Ω b− + b−
∞ ∩ Σ = ∅. If x̄ ∈ Ω∞ ⊂ R
n+1 \ Ω− , Lemma 9.7-(d) gives
∞
−
that u∞ (x̄) = 0. Analogously, if x̄ ∈ Σ ⊂ R + n+1 \ Ω∞ , we have u+
+
∞ (x̄) = 0. Hence, if
b − + b −
x̄ ∈ Ω∞ ∩ Σ we get u∞ (x̄) = 0. Moreover, Ω∞ is an open set, hence Cε (x̄) ⊂ Ω b − for ε
∞
−
small enough. In particular, u∞ = 0 on Cε (x̄), which entails that u∞ ≥ 0 in this cylinder.
So since u∞ is globally adjoint caloric and u∞ (x̄) = 0, the strong minimum principle
[Wa12, Theorem 3.11] implies that u∞ = 0 on Cε+ (x̄). Thus, by the unique continuation
we get u∞ ≡ 0 on Rn+1 , which is again a contradiction.
Therefore, by (9.30), we obtain Ω b− b− + + +
∞ = Ω∞ ∩ Ω∞ ⊂ Ω∞ . Since Ω and Ω are dis-
−
joint, we also have Ω+ ∞ ⊂ R

n+1 \ Ω− = Ω
∞
b− b− +
∞ , concluding that Ω∞ = Ω∞ as desired.
Symmetrically, we can show that Ω− ∞ = Ω∞ .
b+
In order to finish the proof of (b), it is enough to observe that the construction of Σ± and
the equality Ω b ± = Ω∓ infer
Σ+ = Rn+1 \ Ω+ b+ n+1
\ Ω+ −
∞ ∪ Ω∞ ) = R ∞ ∪ Ω∞ )
and
Σ− = Rn+1 \ Ω− b−
∞ ∪ Ω∞ ) = R
n+1
\ Ω+ −
∞ ∪ Ω∞ ).
Hence, the set Σ := Σ+ = Σ− is well-defined and it coincides with Rn+1 \ (Ω+ −

∞ ∪ Ω∞ ).
Since we have already proved that R n+1 + −
\ (Ω∞ ∪ Ω∞ ) = {u∞ = 0}, we obtain that
Σ = {u∞ = 0} as wished.
Proof of (c): The fact that dimH Σ ≤ n is given by [HL94, Theorem 1.1]. To show that
dimH Σ = n, consider two cylinders C+ and C− contained in Ω+ −
∞ and Ω∞ respectively. By
(b), ±u∞ > 0 on C± and, by continuity of u∞ , any line connecting a point in C+ to a point
in C− has to cross Σ. Finally, the formula (9.28) holds because of Lemma 6.4.
10. P ROOFS OF THE MAIN THEOREMS
10.1. Proofs of Theorems I and II. Let us recall that we use the notation

F := cHpn+1 |Σh : c > 0, h ∈ F ∗ (1)

= cHpn+1 |V : c > 0, V admissible n-plane passing through the origin ,
which is a d-cone of Radon measures.
Given ω + , ω − , and a set E as in the statement of Theorem I, we denote
n ω + (Cr (ξ̄) ∩ E) ω − (Cr (ξ̄) ∩ E) o
E ∗ := ξ̄ ∈ E : lim = lim = 1 ,
r→0 ω + (Cr (ξ̄)) r→0 ω − (Cr (ξ̄))
which, by Lemma 4.8 and the mutual absolute continuity assumption for ω ± , satisfies
ω ± (E \ E ∗ ) = 0.
For ξ̄ ∈ E ∗ , we define the Radon-Nikodym derivative
dω + ω + (Cr (ξ̄)) ω + (Cr (ξ̄) ∩ E)
h(ξ̄) = (ξ̄) = lim = lim ,
dω − r→0 ω − (Cr (ξ̄)) r→0 ω − (Cr (ξ̄) ∩ E)
the set
Λ := {ξ̄ ∈ E ∗ : 0 < h(ξ̄) < ∞}
and
Γ := {ξ̄ ∈ Λ : ξ̄ is a Lebesgue point for h with respect to ω + }.
Observe that ω ± (E \ Γ) = 0 (see Lemma 4.8 and [Mat95, Remark 2.15 (3)]).
Lemma 10.1. Let ξ¯ ∈ Γ, cj > 0, and rj → 0 be such that ωj+ = cj Tξ̄,rj [ω + ] ⇀ ω∞
+ . Then
ωj− = cj Tξ̄,rj [ω − ] ⇀ h(ξ̄)ω∞

+.
Proof. The proof is identical to the one of [AMT17, Lemma 5.8] and we omit it.
Lemma 10.2. Let ω + be as in Theorem I. For ω + -a.e. ξ̄ ∈ Γ we have that

Tan(ω + , ξ̄) ∩ F 6= ∅.
Proof. Under the hypotheses of Theorem I, we have that E ⊂ supp ω + ∩ supp ω − ⊂
∂e Ω+ ∩ ∂e Ω− and that Ω+ is disjoint from Ω− . By the classification of boundary points
in Section 2, this implies that E ∩ (BΩ± ∪ ∂s Ω± ) = ∅ and thus, E ⊂ S ′ Ω+ ∩ S ′ Ω− .
◦
Furthermore, since E ⊂ P ∗ Ω± by assumption, C(ξ̄, r) ∩ ∂Ω = C(ξ̄, r) ∩ P ∗ Ω for all r
sufficiently small.
It is not difficult to see that [Pr87, Theorem 2.5] translates to our context. In particular,
¯ is nonempty for ω + -a.e. ξ¯ ∈ Γ. Let ξ̄ ∈ Γ be such that Tan(ω + , ξ)
the set Tan(ω + , ξ) ¯ 6= ∅
+ +
and consider cj > 0 and rj → 0 such that cj Tξ̄,rj [ω ] ⇀ ω∞ for some non-zero Radon
measure ω∞ + ∈ Tan(ω + , ξ). ¯ Lemma 10.1 implies that cj T [ω − ] ⇀ h(ξ̄)ω + and so we
ξ̄,rj ∞
may apply Lemma 9.6 with c = h(ξ̄). Thus, we obtain an adjoint caloric function u∞ in
Rn+1 such that
Z Z Z
+ 1
ϕ dω∞ = u+ ∞ Hϕ = |u∞ | Hϕ, for all ϕ ∈ Cc∞ (Rn+1 ),
2
and supp ω∞+ = {u +

∞ = 0} =: Σ. By Lemma 6.4 we get that ω∞ is absolutely continuous
with respect to the surface measure of Σ = ∂Ω+
∞ and
+
dω∞ = −∂ν + u∞ dσ|Σ ,
t
where ∂ν + is the derivative along the measure theoretic outer unit normal to the time-slice
t
Ω+t . The set Σ is the nodal set of an adjoint caloric function so, by Lemmas 5.1 and 5.2,
for σ|E -a.e. ζ̄ ∈ Σ, there exists ε > 0 such that Σ agrees with a smooth admissible graph
in Cε (ζ̄). Hence, if we combine Lemmas 4.11 and 4.12 with Corollary 4.14, we can find a
non-zero Radon measure µ ∈ F ∗ (1) = F such that µ ∈ Tan(ω + , ξ). ¯
With the results of Section 7 at our disposal, the following lemma can be proved follow-
ing the strategy of [AM19, Lemma 6.1]. Its proof relies on a corollary of a “connectivity”
lemma (see [AM19, Corollary 3.12]), which also holds in the parabolic setting as it does
not use the Euclidean structure.
p̄
Lemma 10.3. Let Ω be an open set in Rn+1 and, given p̄ ∈ Ω, let ω := ωΩ be its associated
caloric measure. Let ξ¯ ∈ supp ω, suppose that Tan(ω, ξ̄) ⊂ HΘ∗ and that there exists
λ > 0 such that for all ωh ∈ Tan(ω, ξ̄) we have

khkL∞ (Cr (0̄)) .λ r −n ωh Cλr (0̄) , for r > 0.
¯ ∩ F ∗ (k) 6= ∅, then Tan(ω, ξ)
If k is the smallest integer for which Tan(ω, ξ) ¯ ⊂ F ∗ (k).
Moreover,
log ω(Cr (ξ̄))
(10.1) lim = n + k.
r→0 log r
We are now ready to gather all the results we have proved so far in order to prove the first
main theorem of the paper.
Proof of Theorem I. We recall that, arguing as in Lemma 10.2, we can assume that E ⊂
¯ ∩ F 6= ∅ for ω + -a.e. ξ¯ ∈ Γ,
S ′ Ω+ ∩ S ′ Ω− . Moreover, Lemma 10.2 infers that Tan(ω + , ξ)
+ ¯
which, by Lemma 9.6, gives that Tan(ω , ξ) ⊂ HΘ∗ . Therefore, we can apply Lemma
10.3 with k = 1 and obtain
log ω(Cr (ξ̄))
lim = n + 1, for ω + -a.e. ξ¯ ∈ Γ.
r→0 log r
Since ω + (E \ Γ) = 0, the previous formula implies that dimHp ω + |E = dimHp ω − |E =
n + 1. If, in addition, Ω± satisfy the TFCDC assumption, we can apply Lemma 9.8 and use
Lemma 9.7-(b) in order to conclude that limr→0 ΘF Σ
∂Ω±
(ξ̄, r) = 0 for ω ± -a.e. ξ̄ ∈ E.
We will now show Theorem II following the approach in [TV18].
Proof of Theorem II. The proof follows from the ones of Lemmas 9.5 and 9.6, and so we
will only sketch it. Let us assume that ω 1 (E) > 0, which by mutual absolute continuity
implies that ω i (E) > 0 for i = 2, 3 as well. By Lemma 3.23 (or rather its proof), we can
find open sets Ωe i ⊂ Ωi , i = 1, 2, 3, which are regular for H and H ∗ , and a set

e ⊂ E ∩ ∩3i=1 (P ∗ Ω
E e i )0 ∩ supp ω e 1 ,
Ω
e > 0 for i = 1, 2, 3, and ω e 1 , ω e 2 , and ω e 3 are mutually absolutely continuous

so that ω i (E) Ω Ω Ω
e So it is enough to prove the result assuming that Ωi is regular for H and H ∗ for
on E.
i = 1, 2, 3.
To this end, by [Pr87, Theorem 2.5], for ω 1 -a.e. ξ¯ ∈ E, we have that there exists cj > 0
and rj → 0 such that ωj1 ⇀ ω∞ 1 , for some non-zero Radon measure ω 1 . Let Γ and Γ be
∞ 2 3
defined as Γ with ω + = ω 1 and ω − = ω i , for i = 2, 3. By Lemma 10.1 and mutual absolute
continuity, we obtain that ωj2 ⇀ h2 (ξ̄)ω∞ 1 and ω 3 ⇀ h3 (ξ̄)ω 1 for ξ¯ ∈ Γ e := Γ2 ∩ Γ3 . Note
j ∞
1 e = 0 and so ω (Γ)
that ω (Γ) i e = 0, for i = 2, 3. By a similar pigeonholing argument as
in the proof of Lemma 9.6, after passing to a subsequence, we may assume without loss of
generality that
(3.15) n+1
rj2 Cap E(ξ̄; rj2 ) \ Ωi & H E(ξ̄; rj2 ) \ Ωi & rjn+2 , j ≥ 0 and i = 1, 2.

In particular, there exists Fj ⊂ E(ξ̄; rj2 ) \ (Ω1 ∩ Ω2 ) such that Hn+1 Fj & rjn+2 . If we

set Gj = Tξ̄,rj (Fj ) ⊂ E(0̄; 1) \ (Ω1j ∩ Ω2j ), then it is clear that Hn+1 Gj & 1 and we have
that
Z
(10.2) |uj | = 0, for all j ≥ 1.
Gj

Since Gj is compact and Hn+1 Gj & 1 for all j ≥ 1, there is a compact set G such
that, after passing to a subsequence, Gj → G in
the Hausdorff
metric and Hn+1 G & 1.
It is easy to see that χGj → g weakly in L2 E(0̄; 1) for some non-negative function

g ∈ L2 E(0̄; 1) , and thus, we must have that g = χG . Repeating the proof of Lemma 9.6
for + = 1 and − = 2, we obtain that uj → u∞ in L2loc (Rn+1 ) for some globally adjoint
caloric function u∞ 6≡ 0 which vanishes only on a set of zero Hn+1 -measure. Taking limits
as j → ∞ in (10.2), we obtain
Z Z
|u∞ | = lim |uj | = 0,
G j→∞ Gj
which, in turn, implies that u∞ = 0 on a set of positive Lebesgue measure reaching a

contradiction.
10.2. Proofs of Theorems III and IV.

Lemma 10.4. Let µ± be two halving Radon measures such that µ− ≪ µ+ and supp µ+ =
−
supp µ− ⊂ ∂e Ω+ . Let us define f := log dµ dµ+ and assume that there is rj → 0 and

+ C (ξ̄ ) −1 T
ξ̄j ∈ ∂e Ω+ so that µ+
j := µ rj j
+
ξ̄j ,rj [µ ] ⇀ µ for some Radon measure µ such
that µ(C1 (0̄)) > 0. Moreover, we assume that µ+ is so that for all M > 0
Z ! Z !
(10.3) lim − f dµ+ exp − − log f dµ+ = 1.
j CM rj (ξ̄j ) CM rj (ξ̄j )
−1
Then µ−
j := µ
− Crj (ξ̄j ) Tξ̄j ,rj [µ− ] ⇀ µ.
Proof. The proof of [AM19, Lemma 8.1] does not use the Euclidean structure and can be
repeated in the parabolic setting.
Proof of Theorem III. Since E is nonempty and relatively open in supp ω + , we have that
ω ± (E) > 0 and
ω + (Cr (ζ̄) \ E)
lim =0 for all ζ̄ ∈ E,
r→0 ω + (Cr (ζ̄))
so [Mat95, Lemma 14.5] gives Tan(ω + , ξ) ¯ = Tan(ω + |E , ξ̄). If ω ∈ Tan(ω + |E , ξ)
¯ =
+ ¯
Tan(ω , ξ), then by Lemma 4.10-(1), ω = c T0̄,r [µ], for some constants c > 0 and r > 0,
and some mesaure µ of the form
1 1
µ = lim + Tξ̄,rj ω + = lim + Tξ̄,rj ω + |E ,
j→∞ ω (Crj (ξ̄)) j→∞ ω |E (Crj (ξ̄))
for some rj → 0, is such that µ(C1 (0̄)) > 0 and the second equality holds because E is
relatively open in supp ω + . Let us observe that the hypothesis of Lemma 10.4 is satisfied
for µ± = ω ± (E)−1 ω ± |E (the normalization is chosen so that µ± are probability measures)
and for ξ̄j = ξ̄ for all j. Hence, we can conclude that
1 1
µ = lim − Tξ̄,rj ω − |E = lim − Tξ̄,rj ω − .
j→∞ ω |E (Crj (ξ̄)) j→∞ ω (Crj (ξ̄))
By Lemma 9.6 there exists g ∈ Θ∗ such that µ = ωg and since the caloric measures
associated with adjoint caloric functions form a d-cone and ω = cT0̄,r [µ], we have that
ω ∈ HΘ∗ too. More specifically there exists h ∈ Θ∗ such that ω = ωh . Let us define
X D α,ℓ h(0̄)
hj (x̄) := xα t ℓ , j > 0,
α!ℓ!
|α|+2ℓ=k
and let k be the smallest number for which we have hk 6≡ 0. By Lemmas 4.12 and 7.9, it
follows that Tan(ωh , 0̄) ⊂ Tan(ω + , ξ̄) and
Tan(ωh , 0̄) = {cωhk : c > 0} ⊂ F ∗ (k).
¯ ∩ F ∗ (k) 6= ∅, which, in turn, by Proposition 7.14 implies
This shows that Tan(ω + , ξ)
+ ¯
Tan(ω , ξ) ⊂ F (k). If we additionally assume that Ω± have TFCDC, then we can argue
∗
as in the proof of Theorem I to show (1.8).
Proof of Theorem IV. Since F is compact, there exists ρ > 0 such that Cρ (ξ̄) ∩ supp ω + ⊂
E for all ξ̄ ∈ F Arguing by contradiction, we assume that for every d ∈ N we can find
sequences {ξ̄j }j≥1 ⊂ F and rj → 0 such that

(10.4) d1 Tξ̄j ,rj [ω + ], P ∗ (d) ≥ ε > 0.
By the doubling assumption on ω + , after passing to a subsequence, we have that
1 1
lim Tξ̄ ,r [ω + ] = lim + Tξ̄ ,r [ω + |E ] = ω
j→∞ ω + (Crj (ξ̄j )) j j j→∞ ω |E (Crj (ξ̄j )) j j
for some measure ω, where the equality holds because E is relatively open in supp ω + .
We claim that ω − (Crj (ξ̄j ))−1 Tξ̄j ,rj [ω − ] ⇀ ω as well. Indeed, since log f ∈ VMO(ω + |E )
and ω + |E is doubling by assumption, f is an Ap -weight on small enough cylinders by the

John-Nirenberg theorem (see [Gar07, Chapter 6.2] for the proof of this implication in the
Euclidean setting), which still holds for spaces of homogeneous type. Hence, ω − |E ∈
V A(ω + |E ) by [AM19, Corollary 7.8] and the hypothesis of Lemma 10.4 is satisfied, so
1 1
ω = lim − Tξ̄j ,rj [ω − |E ] = lim − Tξ̄ ,r [ω − ].
j→∞ ω |E (Crj (ξ̄j )) j→∞ ω (Crj (ξ̄j )) j j
Lemma 9.6 gives that ω = ωh for some h ∈ Θ∗ , i.e.

Z Z
ϕ dω = h Hϕ, ϕ ∈ Cc∞ (Rn+1 ).
The measure ωh is doubling because ω + |E is. Thus, there exists λ ∈ N such that for every
k ∈ N and N ∈ N,

(10.5) ωh C2k+N (0̄) . 2kλ ωh C2N (0̄) .
Let a and M be as in Lemma 9.6 and denote by N a positive integer such that 4a−1 M ≤
2N . Inequality (10.5) together with the Cauchy estimates (6.3), gives that, for α ∈ Nn ,
k ∈ N and ℓ > 0 such that |α| + 2ℓ = m,
|D α,ℓ h(0̄)| . 2−mk khk
L∞ C2k (0̄)
(10.5)
.a,M 2−k(m+n) ωh C2k+N (0̄) . 2−k(m+n−λ) ωh C2N (0̄) ,
where the second inequality follows from the proof of Lemma 9.6. Thus, for m+n−λ > 0
we have that the right hand side in the previous expression converges to 0 as k → ∞, which
implies that D α,ℓ h(0̄) = 0. In particular, h is a polynomial of degree at most d = m+n−λ,
which contradicts (10.4). In order to prove (1.10), it suffices to use Lemma 9.7 and argue as
in the proof of Theorem I.
Proof of Corollary 1.3. One can modify the proof above according to the original proof in
[KT06, pp.34-35] to conclude the corollary bearing in mind that ω − ≪ ω + implies that
t− ≤ t+ . We leave the details to the interested reader.
R EFERENCES
[ACF84] H. W. Alt, L. A. Caffarelli, and A. Friedman. Variational problems with two phases and their free
boundaries. Trans. Amer. Math. Soc., 282(2):431–461, 1984. 2
[AM19] J. Azzam and M. Mourgoglou. Tangent measures of elliptic measure and applications Anal. PDE,
12(8):1891–1941, 2019. 3, 4, 7, 8, 9, 10, 11, 12, 29, 39, 40, 43, 45, 64, 66, 67
[AMT17] J. Azzam, M. Mourgoglou, and X. Tolsa. Mutual absolute continuity of interior and exterior har-
monic measure implies rectifiability. Comm. Pure Appl. Math., 71(11):2121–2163, 2017. 2, 4, 6,
7, 12, 39, 50, 56, 58, 60, 63
[AMTV19] J. Azzam, M. Mourgoglou, X. Tolsa, and A. Volberg. On a two-phase problem for harmonic mea-
sure in general domains. Amer. J. Math., 141(5):1259–1279, 2019. 2, 4, 7, 12, 28, 39
[Bad11] M. Badger. Harmonic polynomials and tangent measures of harmonic measure. Rev. Mat. Iberoam.,
27(3):841–870, 2011. 3, 4, 10, 11, 39, 43
[Bad13] M. Badger. Flat points in zero sets of harmonic polynomials and harmonic measure from two sides.
J. Lond. Math. Soc., 87(1):111–137, 2013. 3
[BET17] M. Badger, M. Engelstein, and T. Toro. Structure of sets which are well approximated by zero sets
of harmonic polynomials. Anal. PDE, 10(6):1455–1495, 2017. 3
[BL15] M. Badger, S. Lewis. Local set approximation: Mattila-Vuorinen type sets, Reifenberg type sets,
and tangent sets. Forum Math. Sigma, 3, E24, 2015. doi:10.1017/fms.2015.26 4
[BW06] L. Barreira and C. Wolf. Pointwise dimension and ergodic decompositions. Ergod. Theory Dyn.
Syst. , 26:653–671, 2006. 30
[BCGJ89] C. J. Bishop, L. Carleson, J. B. Garnett, and P. W. Jones. Harmonic measures supported on curves.
Pac. J. Math., 138(2):233–236, 1989. 2
[BB11] A. Björn and J. Björn. Nonlinear Potential Theory on Metric Spaces. EMS Tracts in Mathematics,
Vol. 17, 2011. 7
[BB01] G. Bouchitté and G. Buttazzo. Characterization of optimal shapes and masses through Monge-
Kantorovich equation. J. Eur. Math. Soc., 3(2): 139–168, 2001. 46
[BBDP03] G. Bouchitté, G. Buttazzo, and L. De Pascale. A p-Laplacian approximation for some mass opti-
mization problems. J. Optim. Theory Appl., 118:1–25, 2003. 49
[BBS97] G. Bouchitté, G. Buttazzo, and P. Seppecher. Shape optimization solutions via Monge-Kantorovich
equation. C. R. Acad. Sci. Paris Sér. I Math., 324(10): 1185–1191, 1997. 46
[BCJ05] G. Bouchitté, T. Champion, and C. Jimenez. Completion of the space of measures in the Kan-
torovich norm. Riv. Mat. Univ. Parma, 7:27–139, 2005. 48, 49
[Bre11] H. Brezis. Functional Analysis, Sobolev Spaces and Partial Differential Equations. Universitext.
Springer, New York, 2011. 50
[Br90] M. Brzezina. On the base and the essential base in parabolic potential theory Czechoslov. Math. J.,
40 (1): 87–103, 1990. 20
[Cho06] S. Cho. Two-sided global estimates of the Green’s function of parabolic equations. Potential Anal.,
25:387–398, 2006. 16
[DeL08] C. De Lellis. Rectifiable Sets, Densities and Tangent Measures. Zurich Lectures in Advanced Math-
ematics. European Mathematical Society (EMS), Zürich, 2008. 33, 34
[Do84] J. L. Doob. Classical Potential Theory and its Probabilistic Counterpart. Grundlehren der mathe-
matischen Wissenschaften, Volume 262, Springer, New York, 1984. 16
[Ed11] D. A. Edwards. On the Kantorovich–Rubinstein theorem. Expo. Math., 29(4):387–398, 2011. 47
[Eng17A] M. Engelstein. A free boundary problem for the parabolic Poisson kernel Adv. Math., 314:835–
947, 2017. 3
[Eng17B] M. Engelstein. Parabolic NTA Domains in R2 . Commun. Partial Differ. Equ., 42:1524–1536,
2017. 3
[EG82] L. C. Evans and R. F. Gariepy. Wiener’s criterion for the heat equation. Arch. Ration. Mech. Anal.,
78:293–314, 1982. 19, 20
[EG15] L. C. Evans and R. F. Gariepy. Measure Theory and Fine Properties of Functions, Revised Edition.
Textbooks in Mathematics. New York: Chapman and Hall/CRC, New York, 2015. 37
[FG10] A. Figalli and N. Gigli. A new transportation distance between non-negative measures, with appli-
cations to gradients flows with Dirichlet boundary conditions. J. Math. Pures Appl., 94(2):107–130,
2010. 46, 47
[GZ82] R. F. Gariepy and W. P. Ziemer. Thermal capacity and boundary regularity. J. Differ. Equ., 45:374–
388, 1982. 21
[Gar07] J. B. Garnett. Bounded Analytic Functions, volume 236 of Graduate Texts in Mathematics.
Springer, New York, first edition, 2007. 67
[GH20] A. Genschaw and S. Hofmann. A weak reverse Hölder inequality for caloric measure J. Geom.
Anal., 30:1530–1564, 2020. 6, 13, 23, 26
[HL94] Q. Han and F. H. Lin. Nodal sets of solutions of parabolic equations: II. Comm. Pure Appl. Math.,
47(9):1219–1238, 1994. 11, 35, 37, 40, 41, 62
[He17] O. Hershkovits. Isoperimetric properties of the mean curvature flow. Trans. Amer. Math. Soc.,
369(6):4367–4383, 2017. 30, 31, 34, 36
[HLN04] S. Hofmann, J. L. Lewis, and K. Nyström. Caloric measure in parabolic flat domains. Duke Math.
J., 122:281–346, 2004. 3
[It18] T. Itoh. The Besicovitch covering theorem for parabolic balls in Euclidean space. Hiroshima Math.
J., 4:279–289, 2018. 5, 32, 33
[JK82] D. S. Jerison and C. E. Kenig. Boundary behavior of harmonic functions in nontangentially acces-
sible domains. Adv. Math., 46(1):80–147, 1982. 2
[KPT09] C. E. Kenig, D. Preiss, and T. Toro. Boundary structure and size in terms of interior and exterior
harmonic measures in higher dimensions. J. Amer. Math. Soc., 22(3):771–796, 2009. 2, 3, 4, 8, 32
[KT06] C. E. Kenig and T. Toro. Free boundary regularity below the continuous threshold: 2-phase prob-
lems. J. Reine Angew. Math., 596:1–44, 2006. 2, 3, 10, 11, 67
[Kor98] M. B. Korey. Ideal weights: asymptotically optimal versions of doubling, absolute continuity, and
bounded mean oscillation. J. Fourier Anal. Appl., 4(4–5):491–519, 1998. 29
[Lanc73] E. Lanconelli. Sul problema di Dirichlet per l’equazione del calore. Ann. Mat. Pura Appl., 97:83–
114, 1973. 19, 20
[Land69] E. M. Landis. Necessary and sufficient conditions for the regularity of a boundary point for the
Dirichlet problem for the heat equation. Dokl. Akad. Nauk SSSR, 185:517–520, 1969; Soviet Math.
Dokl., 10:380–384, 1969. 19
[Lieb96] G. Lieberman. Second Order Parabolic Differential Equations. World Scientific Publishing Co.,
Inc., River Edge, NJ, 1996. 24, 25, 59
[LZ19] F. Lin and Q. S. Zhang. On ancient solutions of the heat equation. Comm. Pure Appl. Math. 72
(3):2006–2028, 2019. 5
[Mag12] F. Maggi. Sets of Finite Perimeter and Geometric Variational Problems. An Introduction to Geo-
metric Measure Theory. Cambridge University Press, Cambridge, 2012. 37
[Mat95] P. Mattila. Geometry of Sets and Measures in Euclidean Spaces, volume 44 of Cambridge Studies
in Advanced Mathematics. Cambridge University Press, Cambridge, 1995. Fractals and rectifiabil-
ity. 18, 33, 34, 39, 43, 63, 66
[McC19] S. McCurdy. Estimates on the generalized critical strata of Green’s function Preprint, 2019.
https://arxiv.org/abs/1904.09361 3
[Nys06a] K. Nyström. Regularity below the continuous threshold in a two-phase free boundary problem.
Math. Scand., 99:257–286, 2006. 3, 6
[Nys06b] K. Nyström. On blow-ups and the classification of global solutions to a parabolic free boundary
problems. Indiana Univ. Math. J., 55:1233–1290, 2006. 3
[Nys06c] K. Nyström. Caloric measure and Reifenberg flatness. Ann. Acad. Sci. Fenn. Math., 31:405–436,
2006. 3, 10, 11
[Nys12] K. Nyström. On an inverse type problem for the heat equation in parabolic regular graph domains.
Math. Z., 31:197–222, 2012. 3
[Or19] T. Orponen. An integralgeometric approach to Dorronsoro estimates. Int. Math. Res. Not. (IMRN),
rnz317, https://doi.org/10.1093/imrn/rnz317. 31, 34
[Po96] C. Poon. Unique continuation for parabolic equations. Commun. Partial Differ. Equ., 21(3–4):521–
539, 1996. 35, 56, 62
[Pr87] D. Preiss. Geometry of measures in Rn : distribution, rectifiability, and densities. Ann. Math. (2),
125(3):537–643, 1987. 4, 31, 63, 65
[San15] F. Santambrogio. Optimal Transport for Applied Mathematicians. New York, NY, USA:
Birkhäuser, 2015. 46
[Sim87] J. Simon. Compact sets in the space Lp (0, T ; B). Ann. Mat. Pura Appl., 146:65–96, 1987. 50
[Sle16] W. Slegers. Metrics on Spaces of Measures. Bachelor Thesis, Mathematisch Instituut, Universiteit
Leiden, 2016. 47
[St70] E. M. Stein. Singular Integrals and Differentiability Properties of Functions. Princeton University
Press, 1970. 35
[TV18] X. Tolsa and A. Volberg. On Tsirelson’s theorem about triple points for harmonic measure. Int.
Math. Res. Not. (IMRN), 2018(12):3671–3683, 2018. 9, 64
[TW85] S. J. Taylor and N. A. Watson. A Hausdorff measure classification of polar sets for the heat equa-
tion. Math. Proc. Cambridge Phil. Soc., 97:325–344, 1985. 8, 17
[TZ17] T. Toro and Z. Zhao. Boundary rectifiability and elliptic operators with W 1,1 coefficients. Adv.
Calc. Var., 14(1):37–62, 2021. 3
[Ts97] B. Tsirelson. Triple points: from non-Brownian filtrations to harmonic measures. Geom. Funct.
Anal. (GAFA), 7:1096–1142, 1997. 9
[Vil03] C. Villani. Topics in Optimal Transportation Graduate Studies in Mathematics, Vol. 58, American
Math. Soc., Providence, RI, 2003. 46
[Vl02] V. S. Vladimirov. Methods of the Theory of Generalized Functions. Analytical Methods and Special
Functions, Vol. 6, London-New York City: Taylor & Francis, pp. XII+353, 2002. 15
[Wa12] N. A. Watson. Introduction to Heat Potential Theory. Mathematical Surveys and Monographs, Vol
182, American Mathematical Society, 2012. 8, 12, 13, 14, 15, 16, 17, 25, 27, 28, 29, 30, 37, 51,
54, 56, 61, 62
[Wa14] N. A. Watson. Regularity of boundary points in the Dirichlet problem for the heat equation. B.
Aust. Math. Soc., 90 (3):476–485, 2014. 20
[Zh02] Qi S. Zhang. The boundary behavior of heat kernels of Dirichlet Laplacians. J. Differ. Equ.,
182:416–430, 2002. 16
D EPARTAMENTO DE M ATEM ÁTICAS , U NIVERSIDAD DEL PA ÍS VASCO , BARRIO S ARRIENA S / N 48940
L EIOA , S PAIN AND , I KERBASQUE , BASQUE F OUNDATION FOR S CIENCE , B ILBAO , S PAIN .
Email address: michail.mourgoglou@ehu.eus
D EPARTAMENTO DE M ATEM ÁTICAS , U NIVERSIDAD DEL PA ÍS VASCO , BARRIO S ARRIENA S / N 48940
L EIOA , S PAIN .
Email address: carmelo.puliatti@ehu.eus

In Memory of Professor Michel Marias (1953-2020) : Bstract

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

In Memory of Professor Michel Marias (1953-2020) : Bstract

Uploaded by

Copyright:

Available Formats

BLOW-UPS OF CALORIC MEASURE IN TIME VARYING DOMAINS AND

APPLICATIONS TO TWO-PHASE PROBLEMS

MIHALIS MOURGOGLOU AND CARMELO PULIATTI

In memory of Professor Michel Marias (1953–2020)

Observe that hj is a homogeneous polynomial of degree j, which implies that Hhj is a

caloric polynomials of degree at most d. Thus, we define

Theorem I. Let Ω+ and Ω− be two disjoint n+1

We remark that if ∂a∗ Ω± ∩ S ∗ Ω± = ∅, then (P ∗ Ω± )0 = P ∗ Ω± . Any domain that

The condition (1.7) implies a vanishing mean oscillation assumption on dω − /dω + on E.

If Ω+ and Ω− also satisfy the TFCDC, then

That is, E ∈ LBW A(PΣ (d)).

vanishing Reifenberg flat.

2. P RELIMINARIES AND NOTATION

We denote points in Rn+1 by x̄ = (x, t) = (x′ , xn , t) := (x1 , . . . , xn , t) and define the

We say that the parabolic Hausdorff dimension of E is

3. H EAT P OTENTIAL T HEORY AND PDE ESTIMATES

Let Ω ⊂ Rn+1 be an open set. Define the parabolic operator

for all ϕ ∈ C 2,1 (Ω)

Let f be an extended real-valued function defined in ∂e Ω. The upper class Uf of f

Let GΩ (·, ·) be the non-negative real-valued function defined in Ω × Ω as

Given a Borel measure µ on Rn+1 and an open set Ω ⊂ Rn+1 ,

defines a non-negative supertemperature in Ω provided the integral is finite in a dense subset

concluding the proof of the claim.

2A different necessary and sufficent condition is given by [Land69].

In fact if ξ̄0 is regular, then (3.22) holds for all λ > 1.

Lemma 3.8. Let Ω ⊂ Rn+1 be an open set and ξ̄0 = (ξ0 , t0 ) ∈ ∂n Ω, If

Given a ∈ (0, 1), we define the translated time-backwards cylinder

Lemma 3.10. If a ∈ (0, 1) and ξ̄0 = (ξ0 , t0 ) ∈ Rn+1 , it holds that

Proof. For ξ¯0 ∈ ∂n Ω, we have that

and Γ(δρ x̄) = ρ−n Γ(x̄). We claim that

Set K = Cr− (Tξ̄,ρ

Moreover, since Gµ ≤ 1 in Rn+1 , it holds that

Therefore, the measure ρn µ

kx̄ − ȳk ≥ kx̄k − kȳk ≥ M0 r − r =: (M0 − 1)r,

and thus, by (3.2),

If SΩ ∩ Cβr (ξ̄0 ) ⊂ PΩ consists of regular points, then (3.34) holds with M0 = 2.

where T1 = Tmax (Cr (ξ̄0 )) = t0 + r 2 .

≤ c ω x̄ (CM r (ξ̄0 )).

for any ζ̄ ∈ ∂E(ξ̄0 , r/2) ∩ Ω.

Let λ > 0. Since vi is supertemperature on Ωi , it is lower semicontinuous, and remains

e i is open. Indeed, to show that Ω

ω(F ) 6= ω(E)/2 for every F ⊂ E. Given s ∈ R and v ∈ Sn , we define the half-space

4. H AUSDORFF AND TANGENT MEASURES

and Lip(1,1) -functions fi : Ei → R n+1 d ∞

We also define Fr (µ) := Fr (µ, 0̄). A standard argument shows that

Furthermore, if E ⊂ Rn+1 is µ-measurable, then the limit

5. N ODAL SET OF CALORIC FUNCTIONS

Given a caloric function h, we denote by Σh := {h = 0} the nodal set of h and by σh

Lemma 5.1 (Structure of the space-regular set). If h : Rn+1 → R is a non-zero caloric

6. C ALORIC MEASURE ASSOCIATED WITH A CALORIC FUNCTION

Given a caloric function h, by the discussion in Section 5, Σh is smooth away from a

where we used that h is caloric and h = 0 on Σh . As f (·) := | · | is smooth away from 0

Lemma 6.4. If h is a caloric function in Rn+1 and σ is the surface measure on Σh as

7. C ALORIC POLYNOMIAL MEASURES

and (7.2) follows.

and so, if for a fixed ε > 0 we further assume

the inequality (7.10) reads

which conclude (7.7) by choosing ε = C −1 2−n−m−4 .

where the last inequality holds because 0̄ ∈ supp ωhm .

The next result is an important application of the previous lemma.

Lemma 7.11. Let h ∈ P (d) and assume that

Then, there exists r0 > 0 and C0 ≥ 1 such that

|∇ψr | ≤ C/r, and |∂t ψr | + |∂x2i xj ψr | ≤ C/r 2 , for 1 ≤ i, j ≤ n.