Skip to main content

Full text of "Characterizing $W^{2,p}$~submanifolds by $p$-integrability of global curvatures"

See other formats


o 



o 
O 



> 



Characterizing W^^^ submanifolds by p-integrability of global curvatures 

Slawomir Kolasiriski* Pawel StrzeleckiJ Heiko von der Mosel^^ 

March 1, 2013 

Abstract 



We give sufficient and necessary geometric conditions, guaranteeing that an immersed com- 
pact closed manifold S™ C M" of class C^ and of arbitrary dimension and codimension (or, more 
(^^ generally, an Ahlfors-regular compact set E satisfying a mild general condition relating the size of 

CN holes in E to the flatness of E measured in terms of beta numbers) is in fact an embedded manifold 

^^ of class C^''^ n W'^^'P, where p > m and r = 1 — rn/p. The results are based on a careful analysis 

.^ of Morrey estimates for integral curvature-like energies, with integrands expressed geometrically, 

r \ in terms of functions that are designed to measure either (a) the shape of simplices with vertices 

^*^ on E or (b) the size of spheres tangent to E at one point and passing through another point of E. 

1-^, Appropriately defined maximal functions of such integrands turn out to be of class L'°(E) for 

d p > 771 if and only if the local graph representations of E have second order derivatives in L^ 

C and E is embedded. There are two ingredients behind this result. One of them is an equivalent 

I— I definition of Sobolev spaces, widely used nowadays in analysis on metric spaces. The second one 

is a careful analysis of local Reifenberg flatness (and of the decay of functions measuring that 
flatness) for sets with finite curvature energies. In addition, for the geometric curvature energy 
QQ involving tangent spheres we provide a nontrivial lower bound that is attained if and only if the 

00 admissible set E is a round sphere. 

^ MSC 2000: 28A75, 46E35, 53A07 

rn 
O 
(N 

• • 1 Introduction 

> 

^ In this paper we address the following question: under what circumstances is a compact, ?TT,-dimensional 

H set S in M", satisfying some mild additional assumptions, an ?Ti-dimensional embedded manifold of 

class VF^'P? For p > m = dim S we formulate two necessary and sufficient criteria for a positive 
answer. Each of them says that E is an embedded manifold of class W'^'P if and only if a certain ge- 
ometrically defined integrand is of class L^ with respect to the ?n,-dimensional Hausdorff measure on 
S. One of these integrands measures the flatness of all (?n, + l)-dimensional simplices with one vertex 
at a fixed point of S and other vertices elsewhere on S; see Definition |1.2| The other one measures 
the size of all spheres that touch an rTi-plane passing through a fixed point of S and contain another 
(arbitrary) point of S (Definition |1.3[ ). 

'Partially supported by the MNiSzW Research project N N201 611 140. 

^Partially supported by the DFG-MNiSzW research project Geometric curvature energies (Mo966/4-l). 



The extra assumptions we impose on the set S are: (1) Ahlfors regularity with respect to the m- 
dimensional Hausdorff measure Jif"^, and (2) roughly speaking, a certain relation between the flatness 
of S and the size of "holes" it might have: the flatter S is, the smaller these holes must be. To state the 
main result. Theorem [L4j formally, let us first specify these two conditions precisely and then define 
the geometric integrands mentioned above. Throughout the paper we denote with B"(a, s) an open 
n-dimensional ball of radius s centered at the point a G M", and we write a ^ bit a/C < b < Ca for 
some constant C > 1, and a < 6 (or a > 6), if only the left (or right) of these inequalities holds. 

1.1 Statement of results 

Definition 1.1 (the class of m-fine sets). Let S C M" be compact. We call S an m-fine set and write 
S G .^{m) if there exist constants Ay, > and Ms > 2 such that 

(i) (Ahlfors regularity) for all a; G S and r < diam S we have 

Jf"^(Sn]B"(x,r)) > Asr'"; (1.1) 

(ii) (control of "holes" in small scales) for each x G S and r < diam S we have 

Here, /3s and 9^, denote, respectively, the beta numbers and the bilateral beta numbers of S, 
defined by 

k{x,r) := -inil sup dist(z, x + iJ) : if G G(n, m) I , (1.2) 

"^ yz&lrm,{x,r) J 

6»s(a;,r) := - inf|d^^(S nl(x,r), (x + ii) n l(a;, r)) : ii G G{n,m)\ , (1.3) 

where G{n, m) stands for the Grassmannian of all m-dimensional linear subspaces of M", and where 

d^'{E,F) := sup{dist(2/,F) : y G E} + sup{dist{z,E) : z £ F} 

is the Hausdorff distance of sets in M". Intuitively, condition (ii) of Definition 1 1 . 1 1 ascertains that if 
E is flat at some scale r > 0, then the gaps and holes in S cannot be large. Their sizes are at most 
comparable to the degree of flatness of E. If an m-fine set S satisfies /3y{x, r) — )• uniformly w.r.t 
X G S as r — ;■ 0, then S is Reifenberg flat with vanishing constant, see e.g. G. David, C. Kenig and T. 
Toro 151 Definition 1.3] for a definition. However, note that neither the Reifenberg flatness of S, nor 
rectifiability of S itself is required in Definition |1.1[ Both these properties /o/tow from the finiteness 
of geometric curvature energies we consider here. 

It is relatively easy to see that ^{m) contains immersed C^ submanifolds of M" (cf. [ 16 , Example 
1.60] for a short proof), or embedded Lipschitz submanifolds without boundary. It also contains other 
sets such as the following stack of spheres S = (Ji^o ^« ^ i^i' where the 2-spheres Sj = S^(cj, r^) C 
M? with radii ri = 2~*~^ > are centered at the points Cj = {pi + pj+i)/2 for pi = (2~*, 0, 0) G M^, 





X 



¥ 



Figure 1: Left: a union of countably many splieres is in ^(2) n £/{5). Right: a set in ^{l) \ £/{5). 

i = 0,1,2, ... . Note that the spheres Sj and Sj+i touch each other at Pi+i, and the whole stack E is 
an admissible set in the class J^{2); see Figure [l] 

A slightly different class £/{S) of admissible sets was used by the second and third author in |"29| . 
Roughly speaking, the elements of £/{6) are Ahlfors regular unions of countably many continuous 
images of closed manifolds, and have to satisfy two more conditions: a certain degree of flatness and a 
related linking condition; all this holds up to a set of J^™-measure zero. The class £/{6) contains, for 
example, finite unions of C^ embedded manifolds that intersect each other along sets of J^"^ -measure 
zero (such as the stack of spheres in Figure[T]), and bi-Lipschitz images of such unions, but also certain 
sets with cusp singularities. For example, an arc with two tangent segments. 



A = {x €R^: xi,X2>0 and {xj + 



1 or max I Xi 

i=l,2 



1)} 



is in £/{6) for each 6 > 0. However, A is not in ^(1) as the /3yi(-, r) goes to zero as r — ;■ at the cusp 
points while 6a{x, r) remains constant there. On the other hand, the union of a segment and countably 
many circles that are contained in planes perpendicular to that segment, 

oo oo 

{{t,0,0):te[0,l]}u[J^jU[Jjj, 



where 

7j = {2~''(1, cos (/3, sin 9?) : c/? G [0, 27r]} 

and 7j is the image of 7j under the reflection (x, y, z) i— )• (1 — x, y, z), is not in s!^{5) as the linking 
condition is violated at all the points of the segment but it does belong to ^(1), as the circles prevent 
the f3{x, r) from going to zero at the endpoints of the segment. 

Both ^{m) and £/{5) contain sets of fractal dimension, e.g. sufficiently flat von Koch snowflakes. 
However, if one of our curvature energies of S is finite, it follows rather easily that the Hausdorff 
dimension of S must be m. 



Definition 1.2 (Global Menger curvature at a point). Let S G T{m) and x G S. Set 
K,G[mx) = K,g{x) := sup i^(x,xi,...,rcm+i) , 

where 

Jf'"+i(conv(x,a:i,...,Xm+i)) 

E:(x,xi,...,Xm+i) :=— ,,n,+2 ' (1-4) 

diam({x,xi,.. . ,a;m+i|j 

and conv(£') and diam(i?) denote the convex hull and the diameter of a set E, respectivel4M We say 
that ICg{x) is the global Menger curvature of S at x. 

When m = 1 and S is just a curve or a more general one-dimensional set then K{xo,xi,X2) 
is the ratio of the area of the triangle T = conv(xo, xi, X2) to the third power of the maximal edge 
length of T. Thus, K is controlled by R{T)~^, where R{T) is the circumradius of T; 

1 Area(r) 

K{xo,Xi,X2) < - 



4i?(T) \xo — xi\ \xi — X2I \x2 — xo\ 



For triangles with angles bounded away from and vr, both quantities are in fact comparable. There- 
fore, in this case our global curvature function /Cg does not exceed a constant multiple of the global 
curvature as defined by O. Gonzalez and J.H. Maddocks |,12], and widely used afterwards; see 
e.g. |[T3|, g, Q, ||2TJ, |[23|, |[27|, |[T0|, |[9|, and for global curvature on surfaces ||25|, ||26|. Also 



for m = 2, integrated powers of a function quite similar to K{xq, xi, 2:2) in ( L4 1 were used in |'28l to 



prove geometric variants of Morrey-Sobolev imbedding theorems for compact two-dimensional sets 



in M in an admissibility class slightly more general than the class £/{6) defined in |29|. 

To define the second integrand, we first introduce the tangent-point radius, which for the purposes 
of this paper is a function 

i?tp: S X S X G{n,m) -^ [0,+oo] 

given by 

Geometrically, this is the radius of the smallest sphere tangent to the affine m-plane x + H and 
passing through x and y. (If y happens to be contained in x + H, in particular if y = x, then we set 

l/Rtp{x,y;H)=0.) 

Definition 1.3 (Global tangent-point curvature). Assume that ff : E — > G{n,m) is an arbitrary 
map. Set 

/Ctp[S](x) = /Ctp(x) = ICtp{x,H{x)) := sup —— -— ^ . 

yes Rtp[x, y;-n[x)) 

Of course, the definition of /Ctp :$]—)■ [0, +00] depends on the choice of H. However, we shall 
often omit the particular map H from the notation, assuming tacitly that a choice of 'tangent' planes 

S 9 X I— )■ H{x) G G{n, m) has been fixed. 



'The function in \IA\ resembles tiie type of discrete curvatures considered by G. Lerman and JT. Whitehouse |l8| , jlTI 
but scales differently, see Remark 5.2 in 128J. 



Theorem 1.4. Let < m < n and S G J-{m). Assume p > m. The following conditions are 
equivalent: 

(1) S is an embedded W^'^-submanifold o/M" without boundary; 

(2) /Cg[S] GLP(S,^™); 

(3) There is a map ff : S — t- G(n, m) such that for this map 

/C,p[S]=/Ctp(-,^(-)) G LP(S,^"^). 

A quick comment on the equivalence of (1) and (3) should be made right away: it is a relatively 
simple exercise to see that for a C^ embedded manifold S the L^ norm of /Ctp(-, H{-)) can be finite 
for at most one continuous map ff : S — ;■ G(n, m) - the one sending every x G S to T^S G G(n, m). 

Let us also mention a toy case of the equivalence of conditions (1) and (2) in the above theorem. 
For rectifiable curves 7 in M" the equivalence of the arc-length parametrization F of 7 being injective 
and in W'^'P, and the global curvature of 7 being in L^ has been proved by the second and third author 
in pTJ. To be more precise, let Sl '■= M-fLX, L > 0, be the circle with perimeter L, and denote by 
r : S'i;, — ;■ M" the arclength parametrization of a closed rectifiable curve 7 : S^ — ;■ M" of length L. 
Then the global radius of curvature function pcil] : 5*/, — )• M, ; see, e.g., [13|, is defined as 

PGb]is):= inf /?(r(s),r(cT),r(T)), s€Sl (1.6) 

where, again, i?(- , • , •) denotes the circumradius of a triangle, and the global curvature kq [7] (s) of 7 
is given by 

KGb]{s):=^^^. (1.7) 

In |27| we prove for p > 1 that T G W'^'P{Sl, K") and T is injective (so that 7 is simple) if and only 
if Kg [7] £ L^- Examples show that this fails for p = 1 = dim 7: There are embedded curves of class 
W"^'^ whose global curvature kq is not in L^ . The first part of the proof (3) =^ (1) for tti = 1, namely 
the optimal C^'^ -regularity of curves with finite energy, is modelled on the argument that was used 
in |30| for a different geometric curvature energy, namely for J^ ^ \/Rl 

We conjecture that the implications (1) =^ (2), (3) of Theorem |1.4| fail fov p = m > 1. 



Remark. If (2) or (3) holds, then according to Theorem 1.4 S is embedded and locally, for some 

ii > 0, S n W{x, R) is congruent to a graph of a W'^'P function /: M"" ^ M"-™. Since p > m, 
we also know from a result of A. Calderon and A. Zygmund (see e.g. |[7| Theorem 1, p. 235]) that 
£»/ : M™ ^ L{W^, M"-™) is differentiable a.e. in the classic sense. 

Remark. One can complement Theorem 1.4 by the contribution of S. Blatt and the first author |[3| 
in the following way. Suppose that 2 < k < m + 2 and in Definition 1.2 one takes the supremum 
only with respect to (?n, + 2) — A: points of S, defining the respective curvature JCck as a function of 
A;-tuples (xq, xi, . . . , Xk-i) G T.^. Suppose that p > m{k — 1) and S is a C^ embedded manifold. 
Then, /C^fc is of class LP(E^ Jf ™'=) if and only if S is locally a graph of class IF1+^'P(M™, M""™), 
where s = 1 — m{k — l)p~^ G (0, 1). If A; = m + 2 and p > m{m + 2), then the assumption that 



S be a C^ manifold is not necessary; one can just assume S G J^{m). See L3] for details. We believe 
that the characterization of ||3| does hold for all 2 < /c < m + 2 without the assumption that S is of 
class C^. (To prove this, one would have to generalize the regularity theory presented in | |l6) to all 
curvatures /Cg fc). 

Blatt's preprint Q contains a similar characterization in terms of fractional Sobolev spaces of 
those C^ manifolds S for which the tangent-point energy JJ^^xS l/l-^tp)*^ is finite. 

Remark. W. AUard, in his classic paper |[T|, develops a regularity theory for m-dimensional varifolds 
whose first variation (i.e., the distributional counterpart of mean curvature) is in L^ for some p > m. 
His Theorem 8.1 ascertains that, under mild extra assumptions on the density function of such a 
varifold V, an open and dense subset of the support of ||y|| is locally a graph of class (7i.i-™/p_ 
For p > m Sobolev-Morrey imbedding yields W"^'^ C C*!'!-™/? and one might naively wonder 
if a stronger theorem does hold, implying Allard's (qualitative) conclusion just by Sobolev-Morrey. 
Indeed, J.P. Duggan [6] proved later an optimal result in this direction. For integral varifolds, W'^'P- 
regularity can be obtained directly via elliptic regularity theory, see U. Menne [19, Lemmata 3.6 
and 3.21]. 

In Allard's case the 'lack of holes' is built into his assumption on the first variation dV of V. 
Our setting is not so close to PDF theory: both 'curvatures' are defined in purely geometric terms 
and in a nonlocal way. Here, the 'lack of holes' follows, roughly speaking, from a delicate interplay 
between the inequality 9{x, r) < /3(x, r) built into the definition of .^{m) and the decay of /3(x, r) 
which follows from the finiteness of energy. A more detailed account on our strategy of proof here, is 
presented in the next subsection. 

At this stage we do not know for our curvature energies what the situation is like in the scale 
invariant case p = m. For two-dimensional integer multiplicity varifolds, however (or in the sim- 



pler situation of VF^'^-graphs over planar domains) Toro |31| was able to prove the existence of 
bi-Lipschitz parametrizations. For ?TT,-dimensional sets Toro f3T, eq. (1)] established a sufficient con- 
dition for the existence of bi-Lipschitz parametrizations in terms of 9. Her condition is satisfied, e.g., 
by S. Semmes' chord-arc surfaces with small constant, and by graphs of functions that are sufficiently 
well approximated by affine functions; see |32[ Section 5] for the details. 

Remark. Following the reasoning in | |27| Lemma 7] one can easily provide nontrivial lower bounds 
for the global tangent-point curvature for hypersurfaces (n = m + 1), and also for curves m = 1 < n; 
see Theorem 1.5 below. Indeed, setting E := ||/Ctp[S]||iP(s), where S C M" is a compact connected 
771-dimensional C^-submanifold without boundary, we can find at least one point x G S such that 
/Ctp[S](a;) < E/{J^'^{T,y/P), since otherwise we had a contradiction via 

^ = (^ (/Ctp[S](x))'djr™(x)) ' > -^A_^(s)Vp = E. 

Therefore R := inf^gs Rtp{x, y, T^S) > ^^'"(S) ^p/^. If there existed an open ball IB"(a, R) with 

(j; + T,.S)n9]B"(a,i?) = {x} 

such that S n ]B"(a, i?) / 0, then we could find a strictly smaller sphere tangent to S in j; and 
containing yet another point y G S contradicting the definition of R. Hence we have shown that the 



union of such open balls 

M := y {]B"(a, R) : a]B"(a, R) n {x + T^T.) = {x}} (1.8) 

contains no point of S. In other words, S is a compact embedded submanifold without boundary, 
contained in R" \ M, and one can ask for the area minimizing submanifold in M" \ M. In codimension 
one, i.e., for ttt, = n — 1, S = d^ for a bounded open set Q C M", and the union of balls defining M 
just consists of two such balls, one in Q and one in the unbounded exterior of S. So, due to the classic 
isoperimetric inequality (see, e.g. [8, Theorem 3.2.43]) one finds 

^"'^(s) > nwy"jr"((rj))'^ 

> nwy"^"(S(a,i?))'^ = jr"-i(aB"(a,i?)) = nw„i2"-^ 
which by definition of i? can be rewritten as 

ll^tpP]||iP(E) =E> (jr"-i(S))^^(na;„)^ (1.9) 

with equality if and only if S equals a round sphere. Hence, we obtain the following simple result. 

Theorem 1.5. Let p > 0. Among all compact embedded C^-hypersurfaces with given surface area, 
the round sphere uniquely (up to isometries) minimizes the energy ||/Ctp[S] \\lp{t,,j(°"-'^)- ^fP > n — 1, 
the same holds true for all [n — l)-fine sets T, G ,^{n — 1). 

Similarly, for tti = 1 one concludes that any of those great circles on any of the balls B"(a, R) 



generating M in (1.8 1 that are also geodesies on M uniquely minimize E among all closed simple 



C^ -curves S = 7 C M" \ M, which provides the lower bound 

ll^tp[7]||Lp(7) =E> 27r^^(7)p-\ (1.10) 



This is exactly what we found for curves in |27 Lemma 7 (3.1)], and is also consistent with ( |1.9| ) if 

n = 2 = ra+1. 

1.2 Essential ideas and an outline of the proof. 

This paper grew out of our interest in geometric curvature energies and earlier related research, cf. 



1 27 1, 1 24 1, 1 28 1, 1 29 1 and 1 16 1. While working on the integral Menger curvature energy of rectifiable 



curves 7 C M 

■^pi^)= \\\ 1^ -dM'\x)d,3^\y)d,^\z), p>3, 

we realized how slicing can be used to obtain optimal Holder continuity of arc-length parametriza- 
tionsQ(The scale invariant exponent p = 3 is critical here: polygons have infinite Mp-energy precisely 
for p > 3; see S. Scholtes pO) for a proof). 



^The second and the third author of this paper acknowledge with gratitude the stimulating conversations that they had in 
the spring of 2008 with Joan Verdera at CRM in Pisa. His insight that most of the work in (24| should and could be phrased 
in the language of beta numbers has helped us a lot in our subsequent research. 



One crucial difference between curves 7 and ?7i-dimensional sets S in M" for m, > 2 lies in the 
distribution of mass in balls on various scales: If 7 is a rectifiable curve and r < ^ diam7, then 
obviously =^^(7 n IB"(x, r)) > r for each x E 7. For m > 1 the measure ^™(S n ]B"(x, r)) might 
be much smaller than r*" due to complicated geometry of S at intermediate length scales. In |j28] we 
have devised a method, allowing us to obtain estimates of J^"^(T, n B"(x, r)) for ttt, = 2, n = 3 and 
all radii r < Rq, with i?o depending only on the energy level of S in terms of its integral Menger 



curvature. This method has been later reworked and extended in the subsequent papers |29 1, 1 16 1, to 
yield the so-called uniform Ahlfors regularity, i.e., estimates of the form 

^^'"(S n ]B"(x, r)) > ^ujmr"", for all r < Rq = Roienergy) , 

for other curvature energies and arbitrary < m < n (to cope with the case of higher codimension, we 
used a linking invariant to guarantee that S has large projections onto some m-dimensional planes). 
Combining such estimates for J^"^{T, n B"(x, r)) with an extension of ideas from [24] we obtained 



in 1 28 1, 1 29 1 and 1 16| a series of results, establishing C^'" regularity for surfaces, or more generally, 
for a priori non-smooth m-dimensional sets for which certain geometric curvature energies are finite. 
Finally, we also realized that the well-known pointwise characterization of H^^'^-spaces of P. Hajlasz 
| [T4] is the missing link, allowing us to combine the ideas from [16J and [29] in the present paper in 



order to provide with Theorem 1.4 a far-reaching, general extension of [27 Theorems 1 & 2] from 



curves to rn,-dimensional manifolds in ^ 



Let us now discuss the plan of proof of Theorem 1.4 and outline the structure of the whole paper. 

The easier part is to check that if S is an embedded compact VF^'^ manifold without boundary, 
then conditions (2) and (3) hold. We work in small balls B(x, R) centered on S, with R > chosen 
so that E n B(x, R) is a (very flat) graph of a W^^p function / : B'"(x, 2R) -^ M""™. Using Morrey's 
inequality twice, we first show that 

Pj:{a,r)<g{a)r, aGB(x,i?)nS, 0<r<R, 

for a function g ^ LP that is comparable to some maximal function of |-D^/|. Next, working with this 
estimate of beta numbers on all scales r = R/2^, k = 0, 1, 2, . . ., we show that in each coordinate 
patch each of the global curvatures /Cg and /Ctp can be controlled by two terms, 

/CG(a),resp./Ctp(a)<5(a) + C(i?) 

where C{R) is a harmless term depending only on the size of the patches. (It is clear from the defi- 
nitions that for embedded manifolds one can estimate both /Cg and /Ctp taking into account only the 
local bending of S and working in coordinate patches of fixed size; the effects of self-intersections 
are not an issue). This yields L^-integrability of ICg and /Ctp- We refer to Section 4 for the details. 

The reverse implications require more work. The proofs that (3) or (2) implies (1) have, roughly 
speaking, four separate stages. First, we use energy estimates to show that if ||/Cg||lp or ||/Ctp||LP ai^e 
less than E^^p for some finite constant E, then 

/ p \ K/{p-m) 

/3.(x,r)<f-] 



Here k denotes a number in (0, 1 — m/p), depending only on m,p with different explicit values for 
JCg or /Ctp, and A^, is the constant from Definition 1 1 . 1 1 measuring Ahlfors regularity of S. By the 
very definition of m-fine sets, such an estimate implies that the bilateral beta numbers of S tend to 
zero with a speed controlled by r'^. In particular, S is Reifenberg flat with vanishing constant, and an 
application of \^ Proposition 9.1] shows that S is an embedded manifold of class C^'". See Section 
3.1 for more details. 

Next, we prove the uniform Ahlfors regularity of S, i.e. we show that 

^^'"(Sn]B(x,r)) > ^jr'"(]B'^(x,r)) 

for all radii r G (0, Rq), where Rq depends only on the energy bound E and the parameters n, m,p, 
but not at all on S itself. Here, we rely on methods from our previous papers [lE] and | [28p9} . Roughly 
speaking, we combine topological arguments based on the linking invariant with energy estimates to 
show that for each r < Rq = Ro{E, n, m,p) the portion of S in B"(x, r) has large projection onto 
some plane H = H{r) G G{n, m). See Section 3.2. 

(There is a certain freedom in this phase of the proof; it would be possible to prove uniform 
Ahlfors regularity first, and estimate the decay of /^^(a^, t) afterwards. This approach has been used 



in I128J|29J.) 

After the second step we know that in coordinate patches of diameter comparable to Rq the man- 
ifold S coincides with a graph of a function / G C^'''(IB™, M"""*). The third stage is to bootstrap the 
Holder exponent k to the optimal r = 1 — m/p > k for both global curvatures /Cc and /Ctp. This 
is achieved by an iterative argument which uses slicing: If the integral of the global curvature to the 
power p over a ball is not too large, then this global curvature itself cannot be too large on a substantial 
set of good points in that ball. Geometric arguments based on the definition of the global curvature 
functions /Cc and /Ctp show that \Df{x) — Df{y)\ < \x — y\'^ on the set of good points. It turns out 
that there are plenty of good points at all scales, and in the Umit we obtain a similar Holder estimate 
on the whole domain of /. See Section 3.3. 

The fourth and last step is to combine the C^''^-estimates with a pointwise characterization of 
first order Sobolev spaces obtained by Hajlasz |14]. The idea is very simple. Namely, the bootstrap 
reasoning in the third stage of the proof (Section 3.3) yields the following, e.g. for the global Menger 
curvature JCg- On a scale Ri ?a Rq, the intersection S n B"(a, Ri) coincides with a flat graph of a 
function f-.Pc^W^^ M"-™ ~ P-^, with 



\Df{x)-Df{y)\<([ ICGiiiJm'di] 



for T = 1 — m/p. Such an inequality is true for every p > m so we can easily fix a number p' G (m, p) 
and show that 

\Df{x) - Df{y)\ < (M(x) + Miy))\x - y\ , (1.11) 

where M(-)^ is the Hardy-Li ttlewood maximal function of the global curvature. Since p/p' > 1, an 
application of the Hardy-Littlewood maximal theorem yields M^ G LP^p , or, equivalently, M G L^. 
Thus, by the well known result of Hajlasz (see Section 2.3), ( |1.11| ) implies that Df G W^'^. In fact, 
the LP norm of D^f is controlled by a constant times the L^-norm of the global Menger curvature JCg- 

9 



An analogous argument works for the global tangent-point curvature function /Ctp- This concludes the 
whole proof; see Section 3.4. 

For each of the global curvatures ICq , there are some technical variations in that scheme; here 
and there we need to adjust an argument to one of them. However, the overall plan is the same in both 
cases. 

The paper is organized as follows. In Section 2, we gather some preliminaries from linear algebra 
and some elementary facts about simplices, introduce some specific notation, and list some auxiliary 
results with references to existing literature. Section 3 forms the bulk of the paper. Here, following 
the sketch given above, we prove that L^ bounds for (either of) the global curvatures imply that S is 
an embedded manifold with local graph representations of class W'^'P. Finally, in Section 4 we prove 



the reverse implications, concluding the whole proof of Theorem 1.4 



Acknowledgement. The authors are grateful to the anonymous referee for her/his careful reading of 
this paper and the suggestions which have helped to improve the presentation of our work. 

2 Preliminaries 

2.1 The Grassmannian 

In this paragraph we gather a few elementary facts about the angular metric <(•,•) on the Grassman- 
nian G{n, m) of m-dimensional linear subspacesjof M". 

Here is a summary: for two m-dimensional linear subspaces 

[/ = spanjni, . . . ,Um} and y = span{t>i, . . . ,1;^} 
inM" such that the bases {ui, . . . ,Um), (^1, • • • ,Vm) are roughly orthonormal and such that |ni— fi| < 



e, we have the estimate <{U, V) < e. This will become especially useful in Section 3.3 

For U G G(n, m) we write ttu to denote the orthogonal projection of M" onto U and we set 

Qu = Id]R" — vTf/ = vTf/x, where Id]Rn : M" — ;• M" denotes the identity mapping. 

Definition 2.1. Let U,V ^ G{n, m). We set 

<([/, y) := ||7r(7 — TTvll = sup \'Ku{w) — T^v{w)\ . 

The function <(•, •) defines a metric on the Grassmannian G{n, m). The topology induced by this 
metric agrees with the standard quotient topology of G(n, m). We list several properties of < below. 



They will become useful for Holder estimates of the graph parameterizations of S in Section 3.3 The 
proofs are elementary and we omit them here. 



^^Formally, G{n, m) is defined as the homogeneous space 

Gin, m) ■- 0{n)/{0{m) x 0{n - m)) , 

where 0{n) is the orthogonal group; see e.g. A. Hatcher's book (15| Section 4.2, Examples 4.53, 4.54 and 4.55] for the 
reference. Thus G{n, m) could be treated as a topological space with the standard quotient topology. Instead, we work with 



the angular metric <(•, •), see Definition 2.1 



10 



Remark. Notice that 

<-{U,V) = \\ttu - Try II = pdRn - Qu - (IdRn -Qv)\\ = \\Qv - Qu\ 



Proposition 2.2 (Lemma 2.2 in |29 1). If the spaces U,V G G{n, m) have orthonormal bases {ei, . . . ,er 



and {fi,...,fm), respectively, and if\ei — fi\ < iDfor i = 1, . . . ,m, then < ([/, V) < 2m-d. 

Definition 2.3. Let V G G{n, m) and let (f i, . . . , Vm) be a basis of V. Fix some radius /> > and 
two constants e G (0, 1) and 6 € (0, 1). We say that (ui, . . . , Vm) is a (/), e, 6)-basis if 

(1 — e)/9 < |fi| < (1 + £)p for i = 1, . . . , m 
and I {vi,Vj)\ < Sp^ for i / j . 

Specifically, a [p, 0, 0)-basis will be called ortho- p-normal. 

Proposition 2.4. Let p > 0, e G (0, 1/2) a«(i 5 G (0, 1) Z^e some constants. Let (ui, . . . , t;^) be a 
{p, e, 5)-basis ofV£ G{n, m). Then there exist an ortho- p-normal-basis (■ui, . . . , -0^) of V and a 
constant C2 = C2 (m) such that 

\vi - Vi\ < (e + C26)p for i = 1, . . . , m . 

Proof. By scaling we may assume that p = 1. Define wi := Vi/\vi\ for i = 1, . . . ,m, fi := wi, 
vi := wi, and then recursively 

fc-i 
fk-=Wk-^{wk,Vi)vi, and Vk ■= fk/\fk\ for fc = 1, . . . ,m, 
i=l 

and observe that \wi—Vi\ = |1 — |fi|| < e and \{wi,Wj)\ < 5/(1 — e)^ < 45 for alii, j = 1, . . . ,m, and 
in addition, V = spanjwi, . . . , w-m} = span{z)i, ...,{)„} by construction. Notice that | |/fc| — 1| = 
ll/fcl - \wk\\ < \fk - Wk\, and therefore 

\fk -Vk\ = \\fk\ - 1| < \fk -Wkl 
so that by 

I^^A; - Vk\ < \vk - Wk\ + \wk - fk\ + \fk -Vk\<e + 2\fk - Wk\ 

the main task turns out to be to estimate at := Ifk — Wklior k = 1, . . . ,171, where we get immediately 
oi = by definition. If one estimates 

fc-i fc-i 



dk < ^\{wk,wi)\ + y^ iwi ■ 

<45{k-l) + Y,iai + \f. 

i=\ 
k-\ 

<4(5(fc-l) + 2^ai, 



i=\ 



11 



one can prove by induction that 



k-l-2 



ttk < 4:6 



i=0 



(A: - 1) + 2 ^ 3'{k -i-2) + 2 • 3'+^ ^ a^ for alH = 0, . . . , A; - 3. 



i=l 



Specifically for / = /c — 3 we obtain 

and therefore, for all A: = 1, . . . , tti, 

\vk -Vk\<e + 



fe-3 



{k - I) + 2'^?>\k - i - 2) 



i=Q 



m— 3 



(m-l) + 2^3*(m-i-2) =: e + C2{m)5 . 

i=0 



D 



Proposition 2.5. Let U,V €z G{n, m) and let (ei, . . . , em) be some orthonormal basis ofV. Assume 
that for each i = 1, . . . ,mwe have the estimate dist(ej, U) = \Qu{^i)\ ^ "d for some ^ G (0, l/v2)- 
Then there exists a constant C3 = C3 (m) such that 

<{U,V) <C3'&. 

Proof. Set Ui := 7ru{ei). For each i = 1, . . . , m we have |(5t/(ei)| < 1?, so 



\ui - ei\ = \Quiei)\ < ■& hence 
1 - t?^ < \/l - 192 < |ti.| < i< 1 + ^2 fori = l,..., 



m . 



(2.12) 



For any i y^ j the vectors ej and Cj are orthogonal, hence 



Therefore 



= {ei,ej) = {nuiei) + Qc/(ej),7rt/(ej) + Quiej)) 
= {■n'uiei),TTu{ej)) + {Qu{ei), Qu{ej)) ■ 

\{u„u,)\ = \{Qu{e^),Qu{eJ))\ < |Q[/(e,)||Qc7(e,)l < ^' 



(2.13) 



Estimates ( |2.12[ ) and ( |2. 13| ) show that (ui, . . . , Um) is a {p, e, (^)-basis of U with constants p = 1, 
£ = ^92 and 6 = {P'. Let (/i, . . . , /m) be the orthonormal basis of U arising from (ui, . . . , Um) by 
means of Proposition |2.4[ so that we obtain 

\fi-ei\ <\fi-u,\ + \u,-ei\ < {I + 02)^^ + -d . 

Using Proposition 2.2 and the fact that -d"^ < -d < 1 \n& finally get 

<([/, V) < 2m((l + C2)i9^ + ^) < 2m(l + C2 + 1)^ . 



Now we can set C3 = Ca^m) := 2m(l + C2(m) + 1) = 2m(2 + 02(171)). 



a 



12 



Proposition 2.6. Let {vi, . . . , Vm) be a (p, e, 5)-basis ofV£ G{n, m) with constants p > 0, e G 

(0, 1/2) and 6 E (0, 1). Let (ui, . . . , Um) be some basis ofU G G{n, m), such that \ui — Vi\ < i}pfor 
some ">} G {0, —^ — ^) and for each i = 1, . . . ,m. Furthermore, let us assume that 

C^{e + C25)<l/2. (2.14) 

Then there exists a constant C4 = C4^{m, e, 6) such that 

<{U,V) <C4t?. 

Proof. Set Cj := Vi/p and let (ei, . . . , e^) be the orthonormal basis of V arising from (ei, . . . , e^) 



by virtue of Proposition 2.4 Set /j := Ui/p 



\Qu{e'i)\ < \Qu{ei - ei)\ + \Qu{ei)\ < \ei - a] <([/, V) + \ei - fi\ 
<\ei-ei\<iU,V)+^. 



From Proposition 2.4 we have |ej — ej| < e + C2S, so 



\Qu{ei)\ < (e + C26) <{U, V) + 'd<2{e + C26) + ^^\+'9<^, 

4 y 2 

since C^{m) > 4 for all rrt G N; see the definition of C3{m) at the end of the proof of Proposition 



2.5| Hence Proposition 2.5 is applicable to the orthonormal basis (ei, . . . , e-m) of V, and we conclude 

<{U, V) < C3(e + C26) <([/, V) + Cs^ 
hence (1 - Csie + C2S)) <{U,V) < C^^ . 



Since we assumed ( 2.14 i we can divide both sides by 1 — C3(e + C2S) reaching the estimate 

C3 



Finally we set 



<{U,V)< 



Ca = C4:{m,e, 6) :- 



1-C3{e + C26) 



^. 



l-C3{m){e + C2{m)5)' 



D 



2.2 Angles and intersections of tubes 

The results of this subsection are taken from our earlier work [291. We are concerned with the inter- 
section of two tubes whose m-dimensional 'axes' form a small angle, i.e. with the set 



S{Hi,H2) := {y e M" : dist(y, Hi) < 1 for i = 1, 2}, 



(2.15) 



where Hi ^ H2 G G{n, m) are such that tt^^ restricted to H2 is bijective. Since the set {y G 
M" : dist(y, Hi) < 1} is convex, closed and centrally symmetricnfor each i = 1, 2, we immediately 
obtain the following: 

■^The term central symmetry is used here for central symmetry with respect to in K". 



13 



Lemma 2.7. S{Hi, H2) is a convex, closed and centrally symmetric set in M"; tihx {^{^11^2)) is a 
convex, closed and centrally symmetric set in Hi = M™. 

For the global tangent-point curvature /Ctp, the next lemma and its corollary provide a key tool in 



bootstrap estimates in Section 3.3 



Lemma 2.8. There exist constants 1 > ei = £i{m) > 0andc2{'m) < 00 with the following property. 
If Hi,H2 £ G{n,m) satisfy < ^{Hi,H2) = a < ei, then there exists an (m — 1) -dimensional 
subspace W C Hi such that 

TTH,{S{HuH2)) C{y£Hi: distiy^W) < 5c2/a} . 

For the proof, we refer to f29^. Lemma 2.6]. It is an instructive elementary exercise in classical 
geometry to see why this lemma is true f or m = 2 and n = 3. 

The next lemma is now practically obvious. 

Lemma 2.9. Suppose that H £ G{n, m) and a set S' C H is contained in {y £ H : dist(y, W) < d} 
for some d > 0, where W is an [m — l)-dimensional subspace of H. Then 

^™(5'n]B"(a,s)) <2™s"-i(i 

for each a £ H and each s > 0. 

Proof Writing each y £ S' D ]B"(a, s) as y = nwiy) + (y - T^wiu)), one sees that S' D ]B"(a, s) is 
contained in a rectangular box with {m — 1) edges parallel to W and of length 2s and the remaining 
edge perpendicular to W and of length 2d. D 

2.3 The voluminous simplices 

Several energy estimates for the global Menger curvature are based on considerations of simplices 
that are roughly regular, which means that they have all edges ^ d and volume Ri d'^+^. Here are the 
necessary definitions, making this vague description precise. 

Definition 2.10. Let T = conv(xo, • • • , Xm+i) be an {m + 1) -dimensional simplex in R". For each 
j = 0, . . . , m + 1 we define the faces fc, (T), the heights f)j(T) and the minimal height f)min(r) by 

fCj-(r) = conv(xo, . . .,Xj-i,Xj+i, . . . ,Xm+i) , 
f)j(r) = dist(xj,aff{xo,. . .,Xj_i,Xj+i,. . . ,x„+i}) 
and (i„jn(r) = min{f),(r) : i = 0, 1, . . . , m + 1} , 

where affjpo, ■ • • ,Pn} denotes the (at most A^-dimensional) affine plane spanned by A^ + 1 the points 

P0,...,P7VGM". 

Note that for any {m + 1) -dimensional simplex T the volume is given by 

J^'^^\T) = -^\),{T)M'"\^c,{T)) foranyiG{0,...,m + l}. (2.16) 

14 



The faces fCj(r) are lower-dimensional simplices themselves, so that a simple inductive argument 
yields the estimate 

J^'^^'iT) > . ^^,, Win(rr+^ (2.17) 

[m + Ij! 

Definition 2.11. Fix some r/ G [0, 1] and d > 0. Let T = conv(j;o> • • • ,Xm+i) be an [m + 1)- 
dimensional simplex in M". We say that T is {-q, d) -voluminous and write T G "^{rj, d) if the following 
conditionCjare satisfied 

diam(T) < d and f)min(T) > rid. 

Proposition 2.12. Let T = conv(3;o, . . . ,Xm+i) be an {rj,d) -voluminous simplex in M" and set 

a = gT/^. Let xq G M" be such that \xo — xo\ < ad and set T = conv(xo, xi, . . . , x^+i)- Then 

diam(f) < |d and f}™„(r) > lr]d = (|r/) (fd) . 
Thus, f G r(|r?, |d) . 
Proof. First we estimate the height i)o{T). Because |a;o — xo| < ad and r] G [0, 1] we have 

HT) > HT) -ad>{ri- a)d > ^rjd . (2.18) 

Fix two indices ii,i2G{l,2,...,m, + l} such that ii / i2- We shall estimate the height l)j^ (T). 
Without loss of generality we can assume that Xi^ is placed at the origin. Furthermore, permuting the 
vertices of T we can assume that ii = 1 and ^2 = 2. We need to estimate f)i(T'). Set 

P = spanjxo - X2, X3 - X2, ■ ■ ■ , Xm+i - X2} = spanjxo, X3, . . . , Xm+i} 
P = span{xo - X2, X3 - X2, . . . , x^+i - X2} = spanjxo, X3, . . . , x^+i} . 

Now we can write 

f)i(r)=dist(a;i,P) = |Qp(xi)| 

= \Qp{xi)-{Qp{x{)-Qp{xi))\ 

>\Qp{xi)\-\Qp[xi)-Qp{xi)\ (2.19) 

>rid- WQp - Qp|||xi| 
>(ry-<(P,P))d, 

so all we need to do is to estimate <(-P, P) from above unless ^(P, P) = 0, in which case we are 
done anyway. 

For that purpose let yo := ^pnp(^o) be the closest point to xq in the (m — 1) -dimensional 
subspace PnP. (Recall that X2 = 0.) Set 

v,:=^^^^^^e{PnP)\ 
\xo - yol 



'a similar class of 1-separated simplices has been considered by Lerman and Whitehouse in |17l Section 3.1] 



15 



and choose an orthonormal basis {v2, ■ ■ ■ , Vm) of PnP. Since yo € PnP C affjxi, X2, • • • , Xm+i} 
one has 

|a^o -yol > dist(xo,aff{xi,X2,...,Xm+i}) > i}rmn{T) > r]d, 



so that 



„ , . Qp{xo-yo) Qp{xo) dist(xo,P) ^ Ixq-xoI ^ a 
\xo-yo\ \xo-yo\ \xo-yo\ m rj 



(2.20) 



Choose any vector vi ^ P such that {vi,V2, ■ ■ . ,Vm) forms an orthonormal basis of P. Note that 
7Tp{vi) is orthogonal to Vj for each j = 2, . . . , m. Indeed, if j G {2, . . . , m}, then we have 

m 

{7rp{vi),Vj) = /^(ui,Uj)i;i,Wj\ + {{vi,vi)vi,Vj) = 0. 



i=2 



Hence, for 



w 



TTpivi) 



kp(wi)| 

we have P = span{w, V2, ■ ■ ■ , Vm} and (w, V2, ■ ■ ■ , Vm) is also an orthonormal basis of P. Moreover 

\w - vi\ <\w- 7rp(t!i)| + \Trp{vi) - vi\ = (1 - |vrp(i;i)|) + \Qp{vi)\ . 



Using ( 2.20 ) we obtain (1 — |vrp(fi)|) < a/rj, hence 



a 



\w — vi\ < 2— . 
V 



(2.21) 



Let /i G S" ^ be any unit vector in M". We calculate 

|7rp(/i) -7rp(/i)| 



'^{h,Vj)vj + {h,vi)vi - ^{h,Vj)v.j - {h, 
3=2 j=2 



'W)W 



< \{h, {vi — 'w))vi\ + \{h,w){vi — 'w)\ < 2\vi — w\ < 4 



a 



V 



This gives us the bound <(P, P) < 4^. Plugging this into ( |2.19| ) and recalling that a = |ry we get 



f),,(r) = f)i(t)>(7?-4^)d=-r/d. 



Since the index ii was chosen arbitrarily from the set {1, . . . , m + 1}, together with (2. 18 1 we obtain 



f)min(T)>^77d, 



which ends the proof. 



D 



16 



2.4 Other auxiliary results 

The following theorem due to Hajlasz gives a characterization of the Sobolev space W^'P and is now 
widely used in analysis on metric spaces. We shall rely on this result in Section 3.4. 

Theorem 2.13 (Hajlasz [ 14, Theorem 1]). Let Q, be a ball in M™ and 1 < p < oo. Then a function 
f G LP(0) belongs to W^'^{^) if and only if there exists a function g G L^(il) such that 

\f{x) - f{y)\ <\x-y\ {g{x) + g{y)) . {Ill) 

In fact, Hajlasz shows that if / G M^^'^, then ( |2.22| ) holds for g equal to a constant multiple of the 
Hardy-Li ttlewood maximal function M{\D^\) of|D/| defined as 

Mh{x) := sup -f h{y) dy. 

r>0 J 

B™(a:,r) 

Conversely, 

liyi.p ~ II/IIlp +inf ||c/||lp , 
g 



where the infimum is taken over all g for which ( |2.22| ) holds. This follows from the proof of Theorem 
1 in PI p. 405]. 



Recall that /? and 6 numbers were defined by ( |1.2[ ) and ( 1.3 1. 



Definition 2.14 (cf. |5], Definition 1.3). We say that a compact set S C M" is Reifenberg-flat (of 
dimension m) with vanishing constant if 

lim sup 6y;{x, r) = . 

The following proposition was proved by David, Kenig and Toro. We will rely on it in Section 3.1. 

Proposition 2.15 (cf. L5j, Proposition 9.1). Let k G (0, 1) be given. Suppose S is an m-dimensional 
compact Reifenberg-flat set with vanishing constant in M" and that there is a constant Cs such that 

Pt.{x, r) < Cj]r'^ for each x G S and r < 1. 

Then S is an m-dimensional C^''^-submanifold ofW^ without boundary^ 

3 Towards the W'^'P estimates for graphs 

In this section we prove the harder part of the main result, i.e. the implications (2) =^ ( 1 ) and (3) =^ ( 1 ). 
We follow the scheme sketched in the introduction. Each of the four steps is presented in a separate 
subsection. 



^Although boundaries of manifolds are not explicitly excluded in the statement of [5 , Proposition 9. 1] it becomes evident 
from the proof that no boundaries are present; see in particular |5 p. 433]. 



17 



3.1 The decay of /3 numbers and initial C^''^ estimates 

In this subsection we prove the following two results. 

Proposition 3.1. Let S C M" be an m-fine set, i.e. S G J^{m), such that 

II^gIIlp(e,^™) <e'p 

for some E < oo and some p > m. Then, the inequality 

Tp \ Ki/{p-m) 



/3s(x,r)<C 



Ki : = 



p — m 



A^y ' ^ p(m- + 1) + 2™, ' 

holds for all r E (0, diani S] and all x G S. The constant C depends on m, p only. 
Proposition 3.2. Let T, G ^{m) be an m-fine set such that 

ll^tp||LP(E,.if'") <E'P 
for some map H : T, ^ Gin, m), a constant E < oo and some p > m. Then, the inequality 

T? \ K2/(p-m) 



/3s(x,r)<C7 



K2 ■■-- 



p — m 



, As J ' ~ p + m 

holds for all r G (0, diani S] and all x G S. The constant C is an absolute constant. 

The argument is pretty similar in either case but it will be convenient to give two separate proofs. 
For the proof of Proposition 3.1 we mimic - up to some technical changes - the proof of |16] 
Corollary 2.4]. First we prove a lemma which is an analogue of [16^ Proposition 2.3]. 

Lemma 3.3. Let T, C M" be an m-fine set, and let xq,xi, ... , Xm+i be arbitrary points ofE. Assume 
that T = conv{xo, . . . ,Xm+i) is {r],d)-voluminous for some rj G (0,1) and some d G (0,oo). 
Furthermore, assume that ||/CG||LP(S,,ir'") ^ E'^'p for some E < oo and some p > m. Then there 
exists a constant C = C{m.,p) depending only on m, andp, such that 



Equivalently, 



where C = C'{m,,p) and 



r]<C' 



E 



Ki/(p-m) 



Kl 



p — m, 
p{m, + 1) + 2m 



Proof Set a = hr] . By Proposition 2.12 each (tti + l)-simplex 



T = conv(xo,a;i,...,Xm+i) 



18 



satisfying \xo — xq] < ad is (|?7, |d) -voluminous. Thus, for any such T we have according to ( |2.17| ) 



Kif) > /^;^ = C^ , (3.23) 



where C = C{m) = (g)'""^^ 9(^+1)! - Using ^^ we obtain 



9{m+l)!- 

> [ ICaixY dJf"'{x) 

This completes the proof of the lemma. D 



We are now ready to give the Proof of Proposition \3.1\ 

Fix some point x € S and a radius r G (0, diam(S)]. Let T = conv(xo, • • • , x^+i) be an 
{m + l)-simplex such that Xj G S n B(x, r) for i = 0, 1, . . . , m, + 1 and such that T has maximal 
^'"^^ -measure among all simplices with vertices in S n B(x, r), i.e. 

^'"+i(T) =max{jr""+i(conv(x'o,...,x;,+i)) : x'^ G Snl(x,r)}. 

The existence of T follows from the fact that the set S n ]B(x, r) is compact and from the fact that the 
function T 1— )• J^™+^(T) is continuous with respect to xq, ..., Xm+i', see, e.g., formula (|2.16[). 



Renumbering the vertices of T we can assume that f)min(r) = ^m+iiT). Thus, according to ( 2.16 l 
the largest m-face of T is conv(xo, . . . , Xm) ■ Let H = spanjxi — xq, • • • , Xm — xq}, so that xq + H 
contains the largest ?n,-face of T. Note that the distance of any point y G S n IB(x, r) from the affine 
plane xq + H has to be less then or equal to f)mm(r) = dist(xm+i, a^o + H), since if we could find a 
point y G S n B(x, r) with dist(y, xq + H) > i)mm{T), then the simplex conv(xo, • • • , Xm, y) would 
have larger ^'"+^-measure than T but this is impossible due to the choice of T. 

Since x G S n B(x,r), we know that dist(x,xo + H) < f]min(T). Thus, we obtain for all 
y G SnB(x,r) 

dist(y, x + H)< dist(y, xq + H) + dist(x, xo + H)< 2i)^^{T). (3.24) 

Hence 

/3s (x, r) < . (3.25) 

r 

Now we only need to estimate f)mm(7") = f)rn+i(r) from above. Of course T is (ry, 2r) -voluminous 
with r\ = f)min(T)/(2r). Lemma [33] implies that 



/3j:{x, r) < ^"""^ ' = 4?? < C7 ( — 1 



19 



which ends the proof of the proposition. D 



Now we come to the Proof of Proposition 3.2 



Fix X G E and r G (0, diam S]. We know by definition of the /3-numbers that (3 = /3s(x, r) < 1. 
We also know that for any z G S n 18(2;, (3r/2) that 

sup dist{-,x + Hz) > l3j:{x,r)r, 

x,r) 



where H^ G G{n, m) denotes the image of z under the mapping H : T, ^ G{n, m). Furthermore, for 
any e > we can find a point y^ G S n B(x, r) such that 

d\si{ye^x + Hz) > sup disi{-,x + Hz) — e> l3Y,{x-,r)r — e. 

EnB(a;,r) 

On the other hand, we have by ly^ — z\ < \y^ — x\ + \x — z\ <'^r 

1 9 

dist(ye, z + Hz) < -lCtp{z)\ye - z\^ < ICtp{z)-r'^ 



so that we obtain 



9 2, 



f'ICtpiz) > dist{y„z + Hz) 

> dist{y):,x + Hz) — \x — z\ 

> fiY.{x,r)r - e- fiY.{x,r)r/2, 

which upon letting e — ^ leads to 

4 
K.tp{z) > -/3E(x,r)/r. 

Estimating the energy as 

E> I ICtp{z)PdJ^"'iz) 

JsnB(x,/3r/2) 

> (^y (/3E(x,r))Pr-Pjr-(Sn]B(x,/3r/2)) > (J^' Q)" A^r— P(/3s(x,r))P+-, 

which gives the desired estimate f or C = 4 > ( |) ^'^ ^^"^""^ 2™/ (p+'^) . D 

Corollary 3.4 (C^''^ estimates, first version). Let Y, C W^ be an m-fine set and set IC^^^-) := 
/CGp](-)a«J/C(2)(.):=/Ctp[S](-).// 

)C^'\z)P dJ^-^iz) < E < oo 
s 

holds for i = lori = 2. Then S is an embedded closed manifold of class C^''^', where 

p — m p — m 

p{m + 1) + 2m ' p + m 

20 



Moreover we can find a radius R = R{n,m,p,As,Ms,E,diaiOLiT,) and a constant K 
K(n, m,p, A-£,My,, E, diam S) such that for each x £ T, there is a function 

of class C^''^\ such that fx{0) = and Dfx{0) = 0, and 

^r\V'{x,R) = x+ ( Graph /^n]B"(0,i?) 
where Graph fxCPx P = R" denotes the graph of fx, and 

||-L^/a;||c'0,Hi(]Bm(o,_R),]R{"-'")xn-) < ^■ 



Proof. The first non-quantitative part follows from our estimates on the /3-numbers in Proposition 3. 1 



and 3.2 in combination with [5 , Proposition 9. 1], cf. Proposition 2. 15 of the previous section. However, 
direct arguments (as in [16^ Corollary 3.18] for the global Menger curvature ICg, and in [29, Section 
5] for the global tangent-point curvature /Ctp), lead to the full statement of that corollary including the 
uniform estimates on the Holder-norm of Dfx and on the minimal size of the surfaces patches of S 
that can be represented as the graph of fx- Let us give the main ideas here for the convenience of the 
reader. 

Assume without loss of generality that x = and write k := kj for a ny i G {1,2} depending on 



the particular choice of integrand IC^^\ We know from Proposition 3.1 or 3.2 respectively, that there 
is a constant Ci = Ci(^e, E, m,p) such that 

/3(r) :=/3E(0,r) < Cir'^ for all r G (0, diam E]. (3.26) 

Since S G J^{m) we have 

6»(r) :=6's(0,r) < MsCir'" for all r € (0, diam S]. (3.27) 

The Grassmannian G{n,m) is compact, so we find for each r G (0, diamS] an ?n,-plane Hx{r) G 
G(n, m) such that 

sup d\st{z^Hx{r)) = (3{r)r. 

zeSnB(x,r) 

Taking an ortho-(r/3)-normal basis (fi(r), . . . ,Vm{r)) of Hx{r) for any such r G (0, diamS] we 
find by ( |3.27[ ) for each i = 1, . . . ,m, some point Zi{r) G S such that 



\ziir)-Viir)\<Mj:Cir^+'; (3.28) 

see Definition 1 1.1[ Now there is a radius Rq = Ro{As, E, m,p, Ms) > so small that we have the 
inclusion B(f j(r), M^Cir'^"'"^) C B(0, r/2) for each r G (0, i?o) and each i = 1, . . . , m, which then 
implies by p.26[ ) that 

dist{zi{r),Hx{r/2))<Cir''+^ for all r G (0, i?o)- (3.29) 

21 



The orthogonal projections u,;(r) := 7r//^(r/2)(^«(^)) iori = l,...,m, satisfy due to ( 3.28 l and ( 3.29 1 

\ui{r) - Vi{r)\ < \vi{r) - z,{r)\ + dist{zi{r), H^{r/2)) < {Mj^ + l)Cir^+\ 

Hence there is a smaller radius < Ri = Ri{Aj],E, m,p, Ms) < Rq such that for all r G (0, i?i) 
one has 

Cir^ < {M^ + l)Cir^ < ^ - ^, (3.30) 



so that Proposition 2.6 is applicable to the (r/3, 0, 0)-basis {vi{r), . . . , Vmir)) of V := Hx{r) and 
the basis (ui(r), . . . ,Um{r)) of U := H.j:{r/2) with i? := Cir''. (Notice that condition ( |2.14| ) in 
Proposition|2.6|is automatically satisfied since e = (^ = in the present situation.) Consequently, 



^{K,{r),H^{r/2)) < C^Cir^ for all r G (0, i?i). 



(3.31) 



Iterating this estimate, one can show that the sequence of ?n,-planes {Hx{r/2^)) is a Cauchy sequence 
in G{n, m), hence converges as A^ — ;■ oo to a limit ?n,-plane, which must coincide with the already 
present tangent plane TqH at x = 0, and the angle estimate ( |3.31[ ) carries over to 



^(T^S, Hx{r)) < Cr^ for all r e (0, Ri) . (3.32) 

Let y G S be such that |y — x| = r/2 and set Wi{r) = -k^ {^r){vi{r)). We have Zi{r) G ]B(y, r), so 

disi{zi{r),Hy{r)) < p^{y,r)r < Cir''+^ , 

hence \vi{r) - Wi{r)\ < \vi{r) - Zi{r)\ + dist{zi{r),Hy{r)) < (Ms + l)Cir'^+^ 
Applying once again Proposition |2.6| - which is possible due to p.30| ) - we obtain the inequality 

HH^ir), Hy{r)) < Ci{M^ + l)Cir'^ = C\x - y^ ■ 



This together with ( 3.32| ) (which by symmetry also holds in y replacing x) leads to the desired local 
estimate for the oscillation of tangent planes 



^{TxTs,TyT.)<C\x-y\'^ for all \x-y\<Ri/2, 



(3.33) 



where C = C{E, A^i rrijp, My,) and Ri = Ri{E, AY,m,p, Ms) do not depend on the choice of 
x,y G S. 

Next we shall find a radius i?2 = R2{E , AY,,m,p, My) such that for each x G S the affine 
projection 

vr^ : S n ]B(x, R2) ^ x + T^S 

is injective. This will prove that S n B(x, i?2) coincides with a graph of some function fx, which is 
C^'^^-smooth by ( [3331 ). 

Assume that there are two distinct points y, z G S n B(x, Ri) such that 7rx(y) = 'Kxiz). In other 
words (y—z) _L TxT,. Since y and z are close to each other the vector (y—z) should form a small angle 



with TzT, but then -^{TzT,, TxT,) would be large and due to p.33[ ) this can only happen if one of y or z 



22 



is far from x. To make this reasoning precise assume that |x — y| < |a; — zj and set H^ = Hx{\y — x\). 
Employing ( |3.26[ ) and ( |3.32| ) we get 



IQtMv - ^)\ < IQhAv - x)\ + IQtMv -x)- QhAv - x)\ 

< /3(x, \y - x\)\y - x\ + ^(T^S, H^)\y - x\ < C\y - x|i+'' < C\z - x|i+'^ , 

where C depends only on E, Ay., m and p. The same appUes to (z — x) so we also have 

\Qt.^{z-x)\<C\z-x\'+^. 

Next we estimate 

\z-y\ = |Qt.s(^ - y)\ < IQt^z -x)\ + IQr^y - x)\ < 2C\z - x\'+^ . (3.34) 

Setting Hz = Hz{\y — z\) and repeating the same calculations we obtain 

dist(y - z, TzT.) = |QT.E(y -z)\<C\y- z|^+« . 

This gives 

^(r,s, r,s) = IIQt.s - Qt^sII > |Qt.s(^ - y) - Qt^^{z - y)\\z - y\-^ 
> {\z -y\- \QtMz -y)\)\z-y\-'>l-C\y- z^ ■ 



On the other hand by ( |333] ) -^{T^T., TzT.) < C\x - z\'^ . Hence, applying ^334\ we obtain 



C\x - zl'' > 1 - C\y - z\^ > 1 - C\x - z\''+''^ ^^ \x-z\>(c + C\x-z\ 
This shows that if {y — z) ±TxTj then the point z has to be far from x. We set 

R2 = mm(l,{C + C)-^/'^^ 



^1/k 



and this way we make sure that vr^ : E n IB(x, R2) — )• x + T^jS is injective for each x G S, hence 
S n M{x, R2) is a graph of some function /^. : TJl n 18(0, R2) -^ {TJl)^. 



The oscillation estimate p.33| ) leads with standard arguments (as, e.g., presented in |[29l Section 
5]) to the desired uniform C^''^-estimates for fx on balls in T^S of radius R2 which depends on 
E, Ay , p, m, Ms, but not on the particular choice of the point x on S. D 

Remark 3.5. The statement of Corollary |3.4| can a posteriori be sharpened: One can show that one can 
make the constants R and K independent of My- This was carried out in detail in the first author's 
doctoral thesis; see fT6l Theorem 2.13], so we will restrict to a brief sketch of the argument here. 
Assume as before that x = and notice that /3(r) = /3(0, r) — ;• uniformly (independent of the 
point X and also independent of My according to ( |3.26| )). Since at this stage we know that S is a 
C^''^-submanifold of M" without boundary, it is clearly also admissible in the sense of |29 Definition 
2.9]. In particular S is locally flat around each point y G S - it is actually close to the tangent ?7i-plane 
TyY, near y - and S is nontrivially linked with sufficiently small {n — m — 1) -spheres contained in 

23 



the orthogonal complement of TyT,. Let Hx{r) for r G (0, diam S] be as in the proof of Corollary 3.4 
the optimal m-plane through x = Q such that 

dist(y, X + ifx(r)) </3(r)r for all y e S n]B(0, r). (3.35) 



One can use now the uniform estimate ( 3.26| ) (not depending on My) to prove that there is a radius 



i?3 = Rz{E, AY,,m,p) such that the angle ^(TqS, Hx{r)) is for each r G (0, R^) so small that, for 
any given p G Hx{r) n B(0, i?3), one can deform the linking sphere in the orthogonal complement 
of TqE with a homotopy to a small sphere in p + Hx{r)-^ without ever hitting S. Because of the 
homotopy invariance of linking one finds also this new sphere nontrivially linked with E. This implies 
in particular by standard degree arguments the existence of a point z E S contained in the {n — m)- 



dimensional disk in p + Hx{r) spanned by this new sphere; see, e.g. |29 Lemma 3.5]. On the other 
hand by ( |3.35[ ) S nB(0, r) is at most I3{r)r away from Hx{r) which implies now that this point z G S 
must satisfy |-z — p| < fi{r)r. This gives the uniform estimate 9{r) < Cj3{r) for all r < R^ and some 
absolute constant C. 



Now we know that the estimates in Corollary |3.4| do not depend on M^;. This constant may be 
replaced by an absolute one if we are only working in small scales. In the next section we show that 
this can be further sharpened: R and K depend in fact only on m, p and E, but not on the constant Aj^. 

3.2 Uniform Ahlfors regularity and its consequences 

In this section, we show that the L^-norms of the global curvatures JCq and /Ctp control the length 
scale in which bending (or 'hairs' , narrow tentacles, long thin tubes etc.) can occur on S. In particular, 
there is a number R depending only on n, m^p and E, where E is any constant dominating ||/Cg|Ilp 
or ||/Ctp||^p, such that for all x G S and all r < R the intersection S n B"(x,r) is congruent to 
Graph /i: n B"(x,r), where /^ : M™ — > M"""^ is a C^'"' function (with small C^ norm, if one 
wishes). Note that R does not at all depend on the shape or on other properties of S, just on its energy 
value, i.e. on the L^-norm of ICq or of /Ctp- 

By the results of the previous subsection, we already know that S is an embedded C^ compact 
manifold without boundary. This is assumed throughout this subsection. 

The crucial tool needed to achieve such control over the shape of S is the following. 

Theorem 3.6 (Uniform Ahlfors regularity). For each p > m there exists a constant C{n,m,p) 
with the following property. 7/'||/Cg'||lp or \\K,g\\lp is less than E^''^ for some E < oo, then for every 

^"^(SnB"(x,r)) > ^cj^r™ for all 0< r < Rq, (3.36) 

where Rq = C{n,m,p)E'^/(P-'^'> anduj^n = ^'"(B™(0, 1)). 



The proof of Theorem |3.6| is similar to the proof of Theorem 3.3 in |28 1 where Menger curvature 
of surfaces in R^ has been investigated. This idea has been later reworked and extended in various 
settings to the case of sets having codimension larger than 1 . 

Namely, one demonstrates that each S with finite energy cannot penetrate certain conical regions 
of R" whose size depends solely on the energy. The construction of those regions has algorithmic 



24 



nature. Proceeding iteratively, one constructs for each x G S an increasingly complicated set S which 
is centrally symmetric with respect to x and its intersection with each sphere d'E"'{x, r) is equal to the 
union of two or four spherical caps. The size of these caps is proportional to r but their position may 
change as r grows from to the desired large value, referred to as the stopping distance ds{x). The 
interior of S contains no points of S but it contains numerous {n — m — 1) -dimensional spheres which 
are nontrivially linked with S. Due to this, for each r below the stopping distance, S n B"(a;, r) has 
large projections onto some planes in G(n, m). However, there are points of S on dS, chosen so that 
the global curvature 1Cg{x), or /Ctp(x), respectively, must be > l/ds{x). 

To avoid entering into too many technical details of such a construction, we shall quote almost 
verbatim two purely geometric lemmata from our previous work that are independent of any choice 



of energy, and indicate how they are used in the proof of Theorem 3.6 
3.2.1 The case of global Menger curvature 



Recall the Definition 2.11 of the class ^(r?, d) of (r/, d) -voluminous simplices. The following propo- 



sition comes from the doctoral thesis of the first author, see |16| Proposition 2.5]. 

Proposition 3.7. Let 6 G (0, 1) and S be an embedded C^ compact manifold without boundary. 
There exists a real number r] = rj{6, m) G (0, 1) such that for every point xq G S there is a stopping 
distance d = ds(xo) > 0, and an {m + l)-tuple of points (xi, X2, • • • , Xm+i) G S"*^^ such that 

T = conv{xo, . . . , Xm+i} G ^{1], d) . 

Moreover, for all p G (0, d) there exists an m-dimensional subspace H = H[p) G G(n, m) with the 
property 

{xo + H)n ]B"(xo, \/l - S^p) C 7r,„+^(S n ]B"(xo, p)) . (3.37) 



Fixing 6 = 5{rn) G (0, v 1 — 4"^/™) small enough, we obtain 77 = rj{m) depending on m only. 
This yields the following. 

Corollary 3.8. For any xq G S and any p < ds{xo) we have 

jr-(S n B(xo, p)) > (1 - 5^r/'u;mP"' > ^comP"" • (3.38) 

Moreover, we can provide a lower bound for all stopping distances. For this, we need an elemen- 
tary consequence of the definition of voluminous simplices: 

Observation 3.9. If T = conv(xo, • • • , Xm+i) G r(r/, d) then by pTTj ) 



K{xo,...,Xjn+i)> 1 n^^T73W+2 = 7 r^r^- <^^-^^^ 

(m + l)!(a)™+^ [m + iy.d 

For rj = 'q{m) and d = ds{xo) this yields 

K.g{xq) > K{XQ, . . . , Xm+l) > -J-, V 

25 



for some constant a{m) depending only on m. By Proposition 2. 12 we know that for simplices T that 
arise from T by shifting xq by at most |??^d a similar estimate holds, possibly with a slightly smaller 
a{m) - still, depending only on m. Thus, 

ICg{z) > ^"^ , for all z G S n B"(xo, v^d/8) . (3.40) 

ds{xo) 

Using the assumption of Theorem|3.6|we now estimate 



p 



E> I Kaizfd.^'^iz) 

JSnB"(xn,r?2d/8) 



.2^/o^^ / a(jn^\ 



>=^™(Sn]B"(xo,r?'(i/8))( ^-^j by(|340i 



^ ^^a;„^""4(2;o)™-^a(m)P by Corollary |3;8] 



Note that r/ G (0, 1), so Corollary 3.8 is indeed applicable. Equivalently, 

dsixoY-"" > c/E 

for some c depending only on m and p. Upon taking the infimum w.r.t. xq £ T, (note that we use 
p > m here!), we obtain 

d(S):= inf dsixo)> (-) =:Ro 



An application of Corollary |3 . 8 1 implies now Theorem 3.6 in the case of /Cg- 



3.2.2 The case of global tangent-point curvature 

As we have already mentioned in the introduction, the L^ norm of the global tangent-point curvature 
/Ctp[S] can be finite for at most one choice of a continuous map H : T, 3 x i-^ H{x) G G{n,m). 
Thus, from now on we suppose 

H : T, 3 X I — > Tx'S G G{n, m) , 

since at this point we know already that S is a C^ submanifold of M" (without boundary). The general 
scheme of proof is similar to the case of global Menger curvature. Some of the technical details are 
different and we present them below. 

High energy couples of points and large projections 

The notion of a high energy couple expresses in a quantitative way the following rough idea: if there 
are two points x, y G S such that the distance from y to a substantial portion of the affine planes 
z + TjS (where z is very close to x) is comparable to |x — y|, then a certain fixed portion of the 
'energy', i.e. of the norm ||/Ctp||Lp, comes only from a fixed neighbourhood of x, of size comparable 
to \x — y\. 

Recall that Qt^t, stands for the orthogonal projection onto (T^S)-*-. 

26 



Definition 3.10 (High energy couples). We say that (x, y) G S x S is a (A, a, (i)-high energy couple 
if and only if the following two conditions are satisfied: 

(i) d/2 < |x - y| < 2d; 

(ii) The set 

S{x,y;a,d) := {z G ]B"(x, a^d) nS: IQtMv - z)\ > ad} 

satisfies 

J^"'{S{x,y;a,d)) > XJ^'^{3"'{0,a^d)) = Xujma^^'d"' . 

We shall be using this definition for fixed < a, A ^ 1 depending only on n and m. Intuitively, 
high energy couples force the L^-nomi of /Ctp to be large. 

Lemma 3.11. If {x,y) G S x S is a (A, a, d)-high energy couple with a < 2 ^^^ '^^ arbitrary 
A G (0, 1], then 



/Ctp(z) > — (3.41) 



a 

for all z G 5(a;, y\ a, d). 

Proof. For z G S(x^ y; a, d) and w G IB"(y, oP'd) we have 

dist(u;,z + T2S) = |Qt,s(m^-2)| = \QTj:.{y - z) ^ Qt,y.{w - y)\ 

> ad — Iw — y\ by Definition |3TT0 (ii) 



ad 1 

> -— as a < ;^ . 
2 "^ 

Moreover, |u) — z| < |x — y| + |a; — z| + |t(; — y| < 2d + 2a'^d < 3d. Thus, by the above computation, 

^ , ^ 2dist(u;,z + r,S) 
/Ctp(z) = sup ■ TK 

«,es \w-z\'^ 

2 dist(w , z + TzT,) ad a 

- ^^P \ 12 > T^TI^ = 7^ ■ 

«)eEnB"(y,a2r) |^ " z\ [6a) yu 

This completes the proof of the lemma. D 



The key to Theorem 3.6 in the case of /Ctp global curvature is to observe that high energy couples 



and large projections coexist on the same scale. 

Proposition 3.12 (Stopping distances and large projections). There exist constants r] = r]{m), 6 = 
S{'m), A = A(n, m) G (0, „ ) which depend only on n, m, and have the following property. 

Assume that S is an arbitrary embedded C^ compact manifold without boundary. For every x £ T, 
there exist a number d = ds{x) > and a point y G S such that 

(i) (x, y) is a (A, rj, d)—high energy couple; 



27 



(ii) for each r G (0, d] there exists a plane H{r) G G{n, m) such that 

TTH{r){^r\W{x,r)) D H{r) n B" {nH{r) (x) , r Vl - <52) 



and therefore 



^™(S n W{X, t)) > (1 - <52)W2^^^m > 1^^^. 



for all < r < ^^(x). 

For the proof of this lemma (for a much wider class of m-dimensional sets than just C^ embedded 
compact manifolds) we refer the reader to [29, Section 4]. 

Lemma 3.13. IfTid M" is an embedded C^ compact manifold without boundary, p > m and 

E> f ICtpixfdJ^"'{x), 



then the stopping distances ds{x) of Proposition 3.12 satisfy 






Ro 



(3.42) 



where c depends only on n, m and p. 



Proof. Let A and rj be the constants of Proposition 3.12 Use this proposition to select a ( A, ?7, (i)-high 
energy couple {x, y) £ T, x T,. Let 

S := S{x,y;r],ds{x)) 

be as in Definition 3.10 (ii). Applying Lemma [3. 11 



we estimate 



E > 



)Ctp{z)PdJ^'^{z) 



> jr"(5) 



9ds{x) 



by Lemma 3.11 



> XuJmr]'^"'^Pds{x)"'-P9-P by Definition [XTOl(ii) 



This implies 



dsix)P-"' > c/E 



for a constant c depending only on n, m, p. As in the case of ICg, upon taking the infimum of the left 
hand side w.r.t. x G S, we conclude the proof of the lemma. D 



Theorem 3.6 in the case of /Ctp follows now immediately. By the lower bound ( |3.42[ ) for stopping 



distances and Proposition 3.12 (ii), the inequality 



^^"^(S n M{x, r)) > (1 - d^r/^Umr"^ > Jw^r" 



holds for each x G S and each r < Rq, since Rq < d{T,) < ds{x). 



28 



3.2.3 An application: uniform size of C^'^-graph patches 

Now, returning to the proofs of Propositions [XT] and |3]2] we see that for all radii 



the estimate J^'^(E{x,r)) > Aj^ujmr"^ can be replaced by ( 3.36[ ), i.e. used with A-^ = 1/2. Thus, 
for such radii the decay estimates in Propositions 3.1 and 3.2 and the resulting C^ '^-estimates do 



not depend on A^ or diam S at all. An inspection of the argument leading to Corollary 3.4 gives the 



following sharpened version, with all estimates depending in a uniform way only on the energy. 

Corollary 3.14 (C^''^ estimates, second version). Assume that S C M" is an m-fine set and let 

/C(i)(-) := /CgP](-) a«^/C(2)(.) := /C.p[S](.). // 



/c«(zfdjr'"(z)<E 



< oo 



holds for i = lori = 2. Then E is an embedded closed manifold of class C^''^% where 

p — m p — m 



Ki 



p{m + 1) + 2m 



K2 



p + m 



Moreover we can find a radius Ri = a{n,m,p)E ^''^ ^' < Rq and a constant Ki = K{n,m,p) 
such that for each x £ T, there is a function 



Jx • J^x^ — • J~^ 



p 



-L ^^ iu>'i^—^n 



of class C^'^'S such that fx{0) = and Dfx{0) = 0, and 

Sn]B"(x,i?i) = x+ ('Graph/3;nB"(0,i?i)Y 
where Graph fxCPx P^ = M" denotes the graph of fx, and 

II 7-) f II ^ If TpKi/{p—m) 

l|-^^a;'IIC'''''«(B"(0,Ki),M("-™)x") — -f^l-^ 



As for Corollary 3.4 also here we do not enter into the details of construction of the graph 
parametrizations fx- These are described in [29^ Section 5.4] and in [16^ Section 3]. 

Remark 3.15. Note that shrinking a{n,m,p) if necessary, we can always assume that 

\Dfx{zi) - Dfx{z2)\ < Kii?-»/(P— ) . i?^« 

= Kia{n,m,p)''' E^^^^P-'^'^ E-^'^/^P-"'^ = Ki{m,p) ■ a{n,m,p)''' < eq 

for an arbitrary small eo = ^o l'^) > that has been a priori fixed. 



29 



3.3 Bootstrap: optimal Holder regularity for graphs 

In this subsection we assume that S is a flat m-dimensional graph of class C^'^\ satisfying 



lC^^{zYdM"^{z)< 



oo 



for z = 1 or i = 2, recall our notation from before: K.^^^ := K-g and /C^^^ := /Ctp. The goal is to show 
how to bootstrap the Holder exponent kj to r = 1 — m/p. 

Relying on Corollary 3.14 and Remark [3.15[ without loss of generality we can assume that 



S n B"(0, 20i?) = Graph / n B"(0, 20i?) 
for a fixed number i? > 0, where 

is of class C^'^'i and satisfies L»/(0) = 0, /(O) = 0, 

\Df\<eQ{m) onP (3.43) 

for some number eq to be specified later on. The ultimate goal is to show that oscBm(^ ^j Df < Cs'^ 
with a constant C depending only on the local energy of S; cf. p.50| ). The smallness condition ( |3.43| ) 



allows us to use all estimates of Section |2] for all tangent planes T^S with z G S n B"(0, 20i?). 

Let F: P -^ M*^ be the natural parametrization of S n B"(0, 2QR), given by F(^) = (^, /(^)) 
for ^ £ P; outside B"(0, 20R) the image of F does not have to coincide with S. The choice of eo 
guarantees 

^(rj.(gj)S,rp(52)S) < ei(m) for all XI, X2 G B"(0,5i?) nP, (3.44) 



where ei(m) is the constant from Lemma 2.8 



As in our papers p9| Section 6], |28 1 and p6| , developing the idea which has been used in p4| 
for curves, we introduce the maximal functions controlling the oscillation of Df at various places and 
scales. 



^*{q,A)= sup [osc Df] (3.45) 

BgCA \Be J 

where the supremum is taken over all possible closed m-dimensional balls Bg of radius g that are 
contained in a subset A C B"(0, 5R) n P, with g < 5R. Since / G C^''^ with k = ki or k = K2 we 
have a priori 

^*{g,A)<Cg^\ i = lor 1 = 2, (3.46) 

for some constant C which does not depend on g, A. 

To show that / G C^'"^ f or r = 1 — m/p, we check that locally, on each scale p, the oscillation 
of Df is controlled by a main term which involves the local integral of /C^*) (z)^ and has the desired 
form C/9^, up to a small error, which itself is controlled by the oscillation of Df on a much smaller 
scale p/N. The number N can be chosen so large that upon iteration this error term vanishes. 



30 



Lemma 3.16. Let f, F,Y,, R> and P be as above. Ifzi,Z2 G B"(0, 2i?)nP w/f/j \zi-Z2\ = t > 0, 
then for each sufficiently large N > Awe have 

\Df{zi)-Df{z2)\<A{mW{2t/N,B) + C{N,m,p)E]lH^ (3.47) 

where B := B'"(^iy^, f) is an m-dimensional disk in P, t := 1 — m/p, and 

Eb= I IC^'\z)P dJ^'^iz) (3.48) 

Jf{B) 

is the local curvature energy ofT, (with i = 1 or i = 2, respectively) over B. In the case of global 
tangent-point curvature /Ctp one can use ( |3.47 1 with A{m) = 2. 



Remark. Once this lemma is proved, one can fix an m-dimensional disk 

B™(6,s) cB"(0,i?)nP 



and use ( 3.47 1 to obtain for t < s 



$* (t, B™(6, s)) < A{m)^* {4t/N, B™(6, s + 2t)) 



m 



+ C{N,m,p)MUb,s + 2t)t\ t = \ , (3.49) 



where 

Mi,{b, r):= I [ lC^\zf dJ^'^(z) ) for i = 1, 2. 

We fix i and then a large N = N{i, m,p) > 4 such that A{m){4/N)'^' < 1/2. This yields A{my ■ 
{2/Ny'^^ — )• as j — )• cxD. Therefore, one can iterate ( 3.49[ ) and eventually show that 



osc Df < C'(m,p)Ml(b,5s)-s'' (3.50) 

B"'(b,s) ^ 

= C'im,p)( [ IC^'\zy dJ^"'{z)] -s", r = l--. 

yJF(B™(b,5s)) J P 

Thus, in particular, we have the following. 

Corollary 3.17 (Geometric Morrey-Sobolev embedding into C^""). Let p > m and T, c M" be an 

m-fine set 

' )C'^'\z)Pdjr'^{z)<E<oo 
s 

for i = 1 or i = 2. Then S is an embedded closed manifold of class C^'^, where r = 1 — m/p. 
Moreover we can find a radius R2 = a2(n, m,p)E~^'^P~'^' < Ri, where a2{n, m,p) is a constant 
depending only on n, m and p, and a constant K2 = K2 {n,m,p) such that for each x G S there is a 
function 

f : TJ] =: P ^ M™ -^ P-L ^ M""™ 

31 



of class C^'"^, such that f{0) = and Df{0) = 0, and 

S n B"(x, R2) = x+( Graph / n B"(0, it's)) , 
where Graph / C P x P = M" denotes the graph of f, and we have 



VJI/(Z1,22) 

for all zi,Z2€ B"(0, i?2) n P, where 

U{zi,Z2) = B™((zi + Z2)/2, 5|zi - Z2I) 



i/p 



Fi -2:21 



(3.51) 



The rest of this section is devoted to the proof of Lemma 3.16 for each of the global curvatures 
tew _ We follow the lines of | |16J and | |29J with some technical changes and necessary adjustments. 

3.3.1 Slicing: the setup. Bad and good points. 



We fix zi, Z2 and the disk B = B™(^i±^, t) as in the statement of Lemma 3.16 we have M"^{B) 



Umf^- Pick A^ > 4 and let Eb be the curvature energy of S over B, defined for i = 1 or i = 2 by 



( |3.48[ ). Assume that Df ^ const on B, for otherwise there is nothing to prove. 
Take 



K^:={EB-N^u^^f''>0 



and consider the set of bad points where the global curvature becomes large, 



(3.52) 



(3.53) 



We now estimate the curvature energy to obtain a bound for M'^{Yq). For this we restrict ourselves 
to a portion of S that is described as the graph of the function /. 



Eb 



K.^^{zY dJ^'^iz) 



F{B) 



> 



/C»( 



F{Yq) 



> 

(153) 



^0 



/C«(F(0)PJdet 
/C«(F(0)Prfe 



Wr™ 1 T r IdiRm 



di 



J^"'{Yo)KPt-"' = ^B^'"(yo)iV"(^™(5))'^ 



The last equality follows from the choice of Kq in (3.52i. Thus, we obtain 



1 t^ 



(3.54) 



32 



and since the radius of B equals t, we obtain 

B™(z„t/iV)\yo/0 fori = 1,2. 
Now, select two good points Uj G M"^{zj,t/N) \ Yq {j = 1, 2). By the triangle inequality, 

\Df[z^)-Df{z2)\ < \Df{zi)-Dfiu,)\ + \Dfiu2)-Df{z2)\ 

+ \Df{ui)-Df{u2)\ 
< 2^*{t/N,B) + \Df{ui)- Df{u2)\. 



Thus, we must only show that for good ui,U2 the last term in p.56| l satisfies 

|L»/(ni) - Df{u2)\ < A{m)<^*{2t/N,B) + C(iV,m,p)^^V . 



(3.55) 



(3.56) 



(3.57) 



This has to be done for each of the global curvatures )C^^\ (It will turn out that for /Ctp one can use 



just the second term on the right hand side of (3.57 1.) 



3.3.2 Angles between good planes: the 'tangent-point' case 



We first deal with the case of /Ctp which is less complicated. To verify ( |3.57| ), we assume that Df{ui ) / 
Df{u2) and work with the portion of the surface parametrized by the points in the good set 



By ( [3341 ), G satisfies 



G:=B\ Yo. 



M"^{G) > (1 - iV-"^)^™(S) =: Gi{p, m) f 



(3.58) 



(3.59) 



To conclude the whole proof, we shall derive - for each of the two global curvatures - an upper 
estimate for the measure of G, 



J^"'{G)<C2{p,m)Ko 



^m+T 



a 



(3.60) 



where a := ^{Hi,H2) / and Hi := Tpi^^.-^T, denotes the tangent plane to S at F{ui) G S for 
i = 1, 2. Combining ( |3.60[ ) and ( 3.59 ), we will then obtain 



a < [Ciy^C2KQe =: G^E]l'^f . 



(By an elementary reasoning analogous to the proof of Theorem 5.7 in 1 29 1, this also yields an estimate 
for the oscillation of Df.) 



Following 1 29 Section 6] closely, we are going to prove the upper estimate ( |3.60| ) for J^"^{G). 



By Corollary 3. 14 and Remark 3. 15 



SnB"(F(ui),20i?) = F(ni) + I Graph/i n B"(0, 20i2) I , 



33 



i.e, that portion of S near F{ui) G S is a graph of a C^'^^ function /i : Hi := Tp(„^)S — ;• H^ with 
|V/i| < eo{m) < 1. As G C -B = B™(^i±^,f) with Zi G B"(0,2i?) nP,t = \zi- Z2I < 4i?, and 
Ui £ B,"^{zi, t/N) (see ( |3.55| )), we have the inclusion 



G C ]B™(0, 6i?) C M'^im, 6R + 2R + t/N) C ]B'"(mi, WR), 

and, as F is 2-Lipschitz, F{G) C ]B"(F(ui), 20i?), i.e., F{G) C x+ (^ Graph /i n ]B"(0, 20ii)y 
Thus, since eoim) is small. 



'det 



^H^{F(G)) 



Id, 



T 



i)/i(0 J L DMO 



Id, 



de 



< 



^H^{F{G)) 



V2dC = V2J^'^{TTH,iFiG))). 



Therefore, 



J^™(G) < J^^'iFiG)) < V2J^'^{7r H,{F{G))), 



so that (3.60 1 would follow from 



jr'"(^Hi(i^(G))) <C4(m)Ko 



m+T 



a 



(3.61) 



To achieve this, we shall use the definition of /Ctp combined with the properties of intersections of 



tubes stated in Lemma [ZS] To shorten the notation, we write 

1 1 



Rtp{x,y;Tjy) Rtp{x,y) ' 

For an arbitrary ( € G and i = 1, 2 we have by ( |3.53| ) 

1 ^ 2\QHAF{C)-Fiui))\ 

Rtp{Fiu,),F{C)) \FiC)-F{uiW 

< )Ctp{F{ui)) < Kor^+^ . 

Let Pi = F{ui) + Hi be the affine tangent plane to T, at F{ui). Since F is Lipschitz with constant 

(1 + eo) < 2and|C-'Ui| < 2t, 



dist(F(C),Pi) = dist{F{0-F{ui),Hi) 

= \QHAnC)-F{u^))\ <8Kot'+^ 



(3.62) 



=: hn 



for ( £ G, i = 1,2. Select the points pi £ Pi, i = 1,2, so that \pi — p2\ = dist(Pi,P2)- The 
vector p2 — pi is then orthogonal to Hi and to H2, and since G is nonempty by ( |3.59[ ), we have 
\P1-P2\ < 2/io by ( [3321 ). 

Set p = {pi + P2)/2, pick a parameter ( £ G and consider y = F{() — p. We have 

y = (F(C) - Fim)) + (F(ni) - pi) + (pi - p), 



34 



so that -khAv) = ^m{F{C) - F{ui)) + (F(ni) - pi), and 

\y-nHAy)\ = \ipi-p)+F{C)-F{ui)-7rHAHC)-Fiui))\ 
= \{pi-p) + Qh^{F{C)-F{ui))\. 



Therefore, since |p — pi| < /iq and by ( |3.62| ), \y — 7r//^(y)| < /iq + /iq = 2/io. In the same way, we 
obtain \y — TTHiiv)] < 2/io- Thus, 



y 

2/io 



F{C)-P 
2/in 



eS{Hi,H2) 



where S{Hi, H2) = {x G M" : dist(x, Hj) < 1 for j = 1, 2} is the intersection of two tubes around 
the planes Hj considered in Section 2.2. Applying Lemma 2.8 which is possible due to the estimate 



( 3.44[ ) for ^{Hi,H2), we conclude that there exists an {m — 1) -dimensional subspace W (Z Hi such 
that 

TTHi {F{G) -p) C{xe Hi: dist(j;, W) < 2ho ■ 5c2/a} . 

On the other hand, since F is 2-Lipschitz, we certainly have 



(3.63) 



F(G) cB"(F(^^^^),2t 



and therefore 



7THAF{G)-p)cM''{a,2t), a:=7TH,{F{ 



Zl+ Z2. 



p). 



(3.64) 



Combining ( |3.63[ )- p.64[ ), we use Lemma 2.9 for the plane H := Hi £ G{n,m), the set S' := 
tthi {F{G) — p), and d := 2/io5c2/a, to obtain 



J^'^[^hAF{G))) < 4^-it™-! • 20/ioC2/a =: Ci{m)KQ 



m+T 



a 



(3.65) 



by definition of h^ in ( |3.62[ ), which is the desired ( |3.61| ), implying ( |3.60[ ) and thus completing the 
bootstrap estimates in the case of the global tangent-point curvature /Ctp- 

3.3.3 Angles between good planes: the 'Menger' case 



To obtain p.57[ ) for the global Menger curvature /Cg, one proceeds along the lines of 1 16 1, with a few 
necessary changes. 

The main difference between /Ctp and Kg is that the control of /Ctp directly translates to the control 
of the angles between the tangent planes. In the case of Kg an extra term is necessary. Namely, we 
choose xi , . . . , Xm G -P so that 



ui\ 



t 



l,2,...,m 



and the vectors Xj — ui form and ortho-p-normal basis of P with p = t/N; see Definition 2.3 



Analogously, we choose yi, . . . , y„i € P close to U2- Next, setting as before Hj = Tp/^.\Ti, we write 

\Df{ui) - Df{u2)\ < i{Hi,H2) (3.66) 

<^{Hi,X) + i{X,Y) + i{Y,H2), 



35 



with the constant in ( 3.66| ) depending on m only, where 



X = span (F(xi) - F(ni), F(x2) - F{ui), ..., F{xm) - F{ui)) 
Y = span (F(yi) - F{u2),F{y2) - F{u2), ..., F{y^) - F{u2)) 

are the secant m-dimensional planes, approximating the tangent ones. A technical but routine calcula- 
tion, relying on the fundamental theorem of calculus (see e.g. |[T6J Proof of Thm. 4.3] or (for m = 2) 
Step 4 of the proof of Theorem 6.1 in [28 1), shows that if the constant eo = eo("^) > controlling 
the oscillation of Df is chosen small enough then 



and consequently 



^{Hi,X) + ^{Y,H2) < Cim)<^*{2t/N,B) , 



\Dfiui) - Df{u2)\ < A(m)$*(2t/iV, B) + C(m) ^(X, Y) 



(3.67) 



where C{m) comes from ( 3.66 ). Thus, it remains to estimate the angle between the secant planes X, Y 
approximating the tangent ones Hi,H2. The estimate of ^(X, Y) is very similar to the computations 
carried out in Section 3.3.2 for the global-tangent point curvature. Here is the crux of the argument. 



We let G = -B \ lo be the good set defined in ( |3.58| ). Shrinking eq = eo{m) if necessary, we may 
assume that 

i{X,Y)<si{m) (3.68) 

where ei(7n) is sufficiently small. Then, 

jr™(G) < J^'^{F{G)) < 2H'^{7Tx{F{G))) , 
and the strategy is to show a counterpart of ( |3.61| ), namely 

j.m+T 



J^"'{7rx{F{G)))<C,Ko 



a 



a = i{X,Y). 



(3.69) 



Comparing this estimate with the lower bound ( |3.59| ) for the measure of G, one obtains 

^(X, y) < i^oi" = const • F^V 



which is enough to conclude the proof of Lemma 3.16 also in the case of the global Menger curvature 



Now, to verify ( [3391 ), we select a point C, ^B = B"'(^i±^, t) with 



\C-u,\^\F{C)-F{u,) 



1,2 



(one can arrange to have constants here close to 1 by the initial uniform smallness ofeQ{m) in ( 3.43 1). 
Then, the {m + l)-simplex T with vertices at F{ui), F{xi), ..., F{xm), F{Q is of diameter ss t. 
The face 

fc„^+l(T) = conv{F(ui), F(xi), . . . , F(x^)} 



36 



is spanned by m nearly orthogonal edges F{xi) — F{ui), of length roughly t/N each, and therefore 
^"■(fCm+i(r)) « t"". Thus, setting now Pi = F{ui) + X, and keeping in mind that m Yq (see 
( 3.53[ )), we obtain by means of ( |2.16| ) 

Kot~'+^ > ICg{F{u,)) 

t"'dist{F{C),Pi) 



>K{F{ui),F{xi),...,F{xm),F{C)) 



j-m+2 



Thus, 



dist(F(C), Pi) < C{m)Kot^+^ , (3.70) 

and the same estimate holds for dist(P(C), P2) where P2 = F{u2) + Y. Thus, we have a counterpart 



of (3.62i in the previous subsection. From that point we reason precisely like in Section 3.3.2, between 



( |3.62[ ) and ( |3.65| ), where at one point we need to use ( |3.68| ). This completes the proof of Lemma 3.16 
in the case of global Menger curvature Kg- 

3.4 ly^^ estimates for the graph patches 



We now show that Corollary 3. 17 [ combined with the result of Hajlasz, cf. Theorem 2.13[ easily yields 
the following. 

Theorem 3.18 (Sobolev estimates). Let S C M" be an m-fine set with 

lC^\zYd,^"'{z)<E<oo 



for i = lori = 2. Then S is an embedded closed manifold of class C^'^ Pi W"^'^, where r = 1 — m/p. 
Moreoverwe canfinda radius P3 = 03(71, n,p)E^^'^'P~''^' < R2,where a^{n, m,p) is a constant 
depending only on n, m, and p, and a constant K3 = Ks^n, m, p) such that for each x £ T, there is 
a function 



f : TxS =: P 



P 



J_ r^ TTDn — m 



of class C^'^ n VF^'P such that /(O) = and Df{0) = 0, and 

S n ]B"(x, P3) = X + f Graph / n ]B"(0, P3) ) , 



where Graph / C P x P = M" denotes the graph of f 

Proof. It remains to show that the graph parametrizations are in fact in W'^'^. To this end, we fix 



an exponent s £ {'m,p) and apply Corollary 3.17 with p replaced by s, to obtain from ( 3.51 1 the 
following estimate 



\Dfiz,)-Dfiz2) 



< 



< 



< 



™((^l+22)/2,5|^l-22|) 



/C«((z,/(^)))^dz 



1/. 



Zl - Z2 



l—m/s 



^({zi+Z2)/2,5\zi-Z2\) 



]c^\{zj{z))y dz^ V1-Z2 



{G{zi) + G{Z2))\ZI - Z2\ 



37 



where 

G{z) = {M]C^'-^\F{z)y)"'' for F{z) = izj{z)), 

and Mh denotes the standard Hardy-Littlewood maximal function of h. Since p > s, we have 
p/s > 1, so that (/C^*^ o Fy is in L^'^ and by the Hardy-Littlewood maximal theorem C 
M((/C(*) o FY) E LP/^ Thus, G G L^. An appUcation of Hajlasz' Theorem 



2.13 



concludes the 



proof of Theorem |3.18 D 



4 From W'^'^ estimates to finiteness of both energies 



In this section, we prove the implications (1) => (2), (3) of the main result. Theorem 1.4 Let us begin 
with a definition. 

Definition 4.1. Let S C M". We say that S is an m-dimensional, W^'^ -manifold (without boundary) 
if at each point x G E there exist an m-plane T^,!! G G{n,m), a radius R^ > 0, and a function 

/ E W^^P{Tj: n B'^(0, 2i?^), R"^™) such that 

Sn]B"(x,i?^) =x+ ('Graph/nB"(0,i?:,. 

We will use this definition only for p > m. In this range, by the Sobolev imbedding theorem, 
each W^'P-manifold is a manifold of class G^. 

Theorem 4.2. Let p > m and let T, be a compact, m-dimensional, W^'^ -manifold. Then the global 
curvature functions /Cg[S] and /Ctp[S] are of class ^^(S, J^™). 

Remark 4.3. As already explained in the introduction, here we assume that /Ctp is defined for the 
natural choice of m-planes Hx = T^S. As we mentioned before, if S is a C^ manifold and Hx i^T^X^ 
on a set of positive ^"^-measure, then the global curvature /Ctp defined for H^ instead of T^S has 
infinite L^-norm. 

4.1 Beta numbers for W^^ p graphs 

We start the proof with a general lemma that shall be applied later to obtain specific estimates for /Cg 

and/CtpinLP(S). 

Lemma 4.4. Let f € W'^'P(E"^{0,2R),M:'^~"'), where p > m and let S = Graph/. Then there 
exists a function g G U'{T. n ]B"((0, /(O), 2R),M"^) such that for each a e H n ]B"((0, /(O)), R) 
and any r < R 

I3j:{a,r) < g{a)r . 

Proof. Fix s G {m,p). Then, / G 1^^''*(]B™(0, 2i?)). Since s > m we have the embedding 

W^''(E'^iO,2R)) C C7i'°(B™(0,2i?)), 
where a = 1 — ™. Choose some point x G B'"(0, i?) and set as before 

F{z):={z,f{z)) and ^^(z) := F(z) - DF(x)(z - x) for z G B'"(0, 2ii). 

38 



Of course ^^^ is in W'^ 'PiM^'iO, 2 R),W') and therefore also in W^'%M''^''{0,2R),W). We now 
fix another point y in ]B™(x, R) and estimate the oscillation of ^x- Set 



By two consecutive applications of the Sobolev imbedding theorem in the supercritical case (cf. 1 11 
Theorem 7. 17]), keeping in mind that U is a ball of radius |x — y|, we obtain 

|*x(y) - ^x(x)| < C{n, m, s)\y - x|1-t (J \D^,{z)\' dz^ 

l/s 



C'\y-x\U \D^,{z)\'dz] 

C'\y - x\ (i \DF{z) - DF{x)\' dz\ 



l/s 



< C\y - x\ (i \z - x|"-"" I \D'^F{w)Y dw dz 

= C|y-x|2 / \D'f{w)\'dw\ 

<C\y-xfM{\D'f\'fl'{x). 

Here M denotes the Hardy-Littlewood maximal function and the constant C = C{n^ m, s) depends 
on n,m, and s. Since m < s < p we have f > 1 and \D'^f\^ £ L'p/^(E'^{0,2R)). Hence we also 
have M(|D2/I') e Lp/'{M'^{0,2R)). Therefore M{\D'^f\')^/' £ LP(]B'"(0, 2i?)). 
To estimate the /? number, note that 

l^.(y) - ^x(x)| = \F{y) - F{x) - DF{x){y - x)\ = \f{y) - f{x) - Df{x){y - x)\ . 

Choose two points a G S n W{F{0),R) and 6 e S n ]B"(F(0), 2R). Since S = Graph / there exist 
x,y e B™(0, 2R) such that F{x) = a and F{y) = b. 
Of course we have |y — x| < |6 — a|. Now we obtain 

dist(6, a + TaS) = dist{F{y),F{x) + r^(:,)S) 

<\F{y)-F{x)-DF{x){y-x)\ 
= \^^{y)-^^{x)\ 
<C\y-xfM{\D'fn'/^{x) 

Since ttir™ is bounded we find together with the previous considerations that the function g{a) := 
CM{\D'^f\''y/'{Trt>m{a)) is of class LP(S n ]B"(F(0), 2R),Jif"'). Choose a radius r € (0, R]. We 
have 

sup dist(6, a + TflS) < sup |b — a| g{a)<r g{a). 

6eSnB"(a,r) feeEnB"{a,T-) 

39 



Hence 



' ■■ ( ' ' 



/5e(o, ^) = - inf I sup dist{b,a + H) \ <- sup dist{b, a + TaT,) < g{a)r . 

r HeG{n,m) VbeSnB"(a,r) / ^ 6eSnB"(a,r) 

D 
We now need to estimate the global curvatures in terms of /3 numbers. Combining these estimates 



with the previous lemma, we will later be able to conclude the proof of Theorem 4.2 



4.2 Global Menger curvature for W'^'^ graphs 

Let us begin with an estimate for the global Menger curvature ICg- 

Lemma 4.5. Let S C M" be a closed m-dimensional set. Choose m + 2 points xq,- ■ ■ ,Xm+i ofT,; set 
T = conv(a;o, . . . , Xm+i) and d = diam(T). There exists a constant C = C{m, n) such that 

jr'"+i(r)<C7/3s(xo,d)d'"+^ 
and 

K{xo, ..., Xm+l) < C . 

Proof. If the affine space affjxo, • • • , Xm+i} is not {m + 1) -dimensional then Jif^~^^{T) = and 
there is nothing to prove. Hence, we can assume that T is an {m + 1) -dimensional simplex. The 
measure J^"^~^^{T) can be expressed by the formula (cf. ( |2.16| )) 

^'"+HT) = ^- dist(x„+i, aff{xo, . . . , x^})jr™(conv(xo, . . . , x^)) . 
m + 1 

In the same way, one can express the measure J^™(conv(xo, • • • , x^)) etc.; by induction, 

^'»+i(T) < - — ^d^+i . 

Hence, if f3j^{xQ, d) = 1, then there is nothing to prove, so we can assume that /32(xo, d) < 1. 
Fix an m-plane H G G{n, m) such that 

disi{y,XQ + H)<d(3'E{xQ,d) for all y G S n B"(xo,d) . (4.1) 

Set h := dl3j]{xQ, d) < d. Without loss of generality we can assume that xq lies at the origin. Let us 
choose an orthonormal basis (ui, . . . , w„) of M" as coordinate system, such that spanjui, . . . , Vm] = 



H. Because of ( |4.1| ) in our coordinate system we have 

Of course, T lies in some {m + 1) -dimensional section of the above product. Let 



V 

Q{a,b) 

Q 

and P 



aff{xo,. . .,Xm+i} = spanjxi,. .. ,Xm+i} , 
Q{d,h) 

vnQ. 



40 



Note that each of the sets V, Q and P contains T. Choose another orthonormal basis wi, . . . , Wn oi 
M" such that V = span{t(;i, . . . , Wm+i}- Set 

S := {x £ V-^ : \{x,Wi)\ < h for i = l,...,m} . 

Thus, 5 is just the cube [—h, /i]"^'"^^ placed in the orthogonal complement of V. Note that diam S = 
2h\/n — m — I. In this setting we have 

Px S =cQ{d + 2h^n -m-l,h + 2hVn - m - 1) . (4.2) 

Recall that h = (i/3s(xo, d) < d. We estimate 

j^"(r X 5) < jr"(p X s) 



^n—m 



< J^''{Q{d + 2hVn - m - 1, h + 2hVn - m - 1)) 
= (2d + 4/iVn - m - 1) " (2/i + 4/iVn - m - l) 

< (2d + 4dVn-m-ir(2/i + ihy/n -m- l)""™ 
= (2 + 4Vn-m-l)"(i"/3E(xo, d)"-"* . 

On the other hand we have 

jr"(T X 5) = jr™+i(r)^"-'"-i(5) 

= 2'^-"-i^"^+i(T)d"-™-i/3E(xo,d)"-™-i . 

Hence 

2"-"^-i^"^+i(r)(i"-™"i/3s(xo,d)"-'"~i < (2 + 4Vn - m - l)'^d"/3s(a;o, (i)""", 

or equivalently 

jr'"+i(T) < (2 + 4Vra - m - l)'"2~("-'"-^)d"^+^/3s(xo, d) . 

We may set C = C{n,m) = (2 + 4-v/n^^m^^^)"2^("^™^^). This completes the proof of the 
lemma. D 

Since S is a compact T^^'P-manifold (p > m) we may cover it by finitely many balls, in which 



S is described as a graph, such that Lemma 4.4 is satisfied in each of these graph patches with a 



respective function g defined only on that patch. More precisely, we find ai, . . . , a^ G S with 

N 

EC \jM"iak,R/2), 
fc=i 

such that for each A; = 1, . . . , A^, one has 

S n ]B"(afc, 2R) = ak + (Graph /^ n ]B"(afc, 2R)) , 

41 



where fk £ VF2'P(B'"(0, 2R),W'-"'), and there is a function gk G LP(S n B"(afc, 2i?), Jf"") with 
the property that for each a e T,nM^{ak,R) and any r < R one has the estimate 

/3s(a,r)<5fc(a)r. (4.3) 

Using a partition of unity subordinate to this finite covering, i.e., {'qk)k=i C C^(B"(afc, R/2)) with 
< ??fc < 1. ^k=i '? ~ ^' ^^ '^^^ extend the functions 7/^(7^ to all of S by the value zero outside of 
B"(afc, R/2) for each A: = 1, . . . , iV, and define finally g e LP(E, J^"^) as 

JV 

k=l 

Now, for any xq £ S there exists fc € {!,..., A^} such that xq G S n B'^(afc, i?/2), so that 
B"(xo, R/2) C B"(afc, fi), and we conclude with (|43) for any r < i? 



Af TV 

/3e{xo, r) = ^ VkPi:{xo, r) < ^ Vkgkixo)r = g{xo)r. 

k=l k=l 



Consequently, by Lemma [43 



^G(a^o) = sup K(xo,xi,...,Xm+i) 

xi,...,Xm+ies 

/3s(2;o,diam(xo,...,Xm+i)) 
< G sup 






_^gE diam(xo,...,Xm+i) 



< C sup g{xo) = Cg{xo) ■ 

xi,...,Xm+ies 

This leads to the following result. 

Corollary 4.6. Let Tibe a compact, m-dimensional, W^'^ -manifold for some p > m. Then /Cg[S] G 

4.3 Global tangent-point curvature for 1^^ ^ graphs 

The following simple lemma can be easily obtained from the definition of /Ctp. 

Lemma 4.7. Assume that S is a C^ embedded, compact m-dimensional manifold without boundary. 
Then, for some R = R{T,) > Owe have 

Mp(a;) < ^ + sup 



R r<R r 

Proof Choose i? > so that for each point x G S the intersection S n B"(x, 3R) is a graph of a C^ 
function /: T^S -> (T^S)-^ with oscillation of Df being small. Fix x G S. Set F(^) := {z, f{z)) 
for z G P = TiS. As before, we write 

x,y G S. 



/?tp(a:;,y;rj'5]) Rtp{x,y) ' 
42 



It is clear that for |x — y| > i? we have Rtp{x, y) > R/2 by definition. Thus 

2 1 

/Ctp(x) < ^ + sup 



R \x-y\<RRtp{x,y) 

It remains to estimate the last term. Now, ii x = F{^) and y = F{7]) E E n B"(x, R) with 

\y-x\ = \F{r,) - F{0\ ^ \r] - ^\ ^ Pj ^ ^ , j = 0, 1, , 2, . . . , 

then 

1 _ 2dist{y,x + T^^T,) ^ /3s(x,pj) 

Rtp{x,y) |y-x|2 ~ pj 

with an absolute constant. The lemma follows. D 



Combining the above lemma with Lemma 4.4 we conclude immediately that /Ctp G L^ for W'^'P- 



manifolds with p > m. The proof of the implications (1) =^ (2), (3) of Theorem 1.4 is now complete. 



References 



[1 
[2; 

[3 

[4; 

[5 

[6: 

[7 
[8 

[9: 
[lo: 
[11 

[12; 



William K. Allard, On the first variation of a varifold, Ann. of Math. (2) 95 (1972), 417^91. 

Simon Blatt, The energy spaces of the tangent-point energies, 2011. 

Simon Blatt and Slawomir Kolasinski, Sharp boundedness and regularizing effects of the integral Menger 
curvature for submanifolds, Adv. Math. 230 (2012), no. 3, 839-852. 

Jason Cantarella, Robert B. Kusner, and John M. Sullivan, On the minimum ropelength of knots and links, 
Invent. Math. 150 (2002), no. 2, 257-286. 

Guy David, Carlos Kenig, and Tatiana Toro, Asymptotically optimally doubling measures and Reifenberg 
flat sets with vanishing constant, Comm. Pure Appl. Math. 54 (2001), no. 4, 385^149. 

J. P. Duggan, Vl^^'P regularity for varifolds with mean curvature, Comm. Partial Differential Equations 11 
(1986),no. 9, 903-926. 

Lawrence C. Evans and Ronald F. Gariepy, Measure theory and fine properties of functions. Studies in 
Advanced Mathematics, CRC Press, Boca Raton, FL, 1992. 

Herbert Federer, Geometric measure theory. Die Grundlehren der mathematischen Wissenschaften, Band 
153, Springer- Verlag New York Inc., New York, 1969. 

Henryk Gerlach and Heiko von der Mosel, On sphe refilling ropes, Amer Math. Monthly 118 (2011), 
no. 10, 863-876. 

, What are the longest ropes on the unit sphere?. Arch. Ration. Mech. Anal. 201 (2011), no. 1, 



303-342. 



David Gilbarg and Neil S. Trudinger, Elliptic partial differential equations of second order. Classics in 
Mathematics, Springer- Verlag, Berlin, 2001, Reprint of the 1998 edition. 

Oscar Gonzalez and John H. Maddocks, Global curvature, thickness, and the ideal shapes of knots, Proc. 
Nad. Acad. Sci. USA 96 (1999), no. 9, 4769-4773 (electronic). 



43 



[13] Oscar Gonzalez, John H. Maddocks, Friedemann Schuricht, and Heiko von der Mosel, Global curvature 
and self-contact of nonlinearly elastic curves and rods, Calc. Var. Partial Differential Equations 14 (2002), 
no. 1, 29-68. 



[14; 

[15 

[16: 
[17: 

[18 

[19: 
[2o: 

[21 

[22: 

[23 

[24: 

[25 

[26: 

[27 

[28 
[29: 

[30 
[31 
[32 



Piotr Hajlasz, Sobolev spaces on an arbitrary metric space. Potential Anal. 5 (1996), no. 4, 403-415. 

Alan Hatcher, Algebraic topology, Cambridge University Press, 2002. 

Slawomir Kolasinski, Integral Menger curvature for sets of arbitrary dimension and codimension, Ph.D. 
thesis. Institute of Mathematics, University of Warsaw, 2011, arXiv:1011.2008v4. 

Gilad Lerman and J. Tyler Whitehouse, High-dimensional Menger-type curvatures. II. d-separation and 
a menagerie of curvatures, Constr. Approx. 30 (2009), no. 3, 325-360. 

, High-dimensional Menger-type curvatures. Parti: Geometric multipoles and multiscale inequal- 



ities. Rev. Mat. Iberoam. 27 (2011), no. 2, 493-555. 

UMch Menne, Second order rectifiability of integral varifolds of locally bounded first variation, 201 1, J. 
Geom. Anal., accepted, DOl: 10.1007/sl2220-01 1-9261-5. 

Sebastian Scholtes, For which positive p is the integral Menger curvature Aip finite for all simple poly- 
gons?, 2011, arXiv:1202.0504vl. 

Friedemann Schuricht and Heiko von der Mosel, Euler-Lagrange equations for nonlinearly elastic rods 
with self-contact. Arch. Ration. Mech. Anal. 168 (2003), no. 1, 35-82. 

, Global curvature for rectifiable loops. Math. Z. 243 (2003), no. 1, 31-11 . 

, Characterization of ideal knots, Calc. Var. Partial Differential Equations 19 (2004), no. 3, 281- 

305. 

Pawel Strzelecki, Marta Szumanska, and Heiko von der Mosel, Regularizing and self-avoidance effects of 
integral Menger curvature, Ann. Sc. Norm. Super. Pisa CI. Sci. (5) 9 (2010), no. 1, 145-187. 

Pawel Strzelecki and Heiko von der Mosel, On a mathematical model for thick surfaces. Physical and nu- 
merical models in knot theory, Ser. Knots Everything, vol. 36, World Sci. Publ., Singapore, 2005, pp. 547- 
564. 

, Global curvature for surfaces and area minimization under a thickness constraint, Calc. Var. 



Partial Differential Equations 25 (2006), no. 4, 431-467. 

, On rectifiable curves with L^ -bounds on global curvature: Self-avoidance, regularity, and mini- 
mizing knots.. Math. Z. 257 (2007), 107-130. 

, Integral Menger curvature for surfaces. Adv. Math. 226 (201 1), 2233-2304. 



, Tangent-point repulsive potentials for a class of non-smooth m-dimensional sets in M". 

Part I: Smoothing and self -avoidance effects, 2011, arXiv: 1102.3642; J. Geom. Anal., accepted, DOI: 
10.1007/S12220-011-9275-Z. 

, Tangent-point self-avoidance energies for curves, J. Knot Theory Ramifications 21 (2012), no. 5, 



28 pages. 



Tatiana Toro, Surfaces with generalized second fundamental form in L? are Lipschitz manifolds, J. Dif- 
ferential Geom. 39 (1994), no. 1, 65-101. 

, Geometric conditions and existence of bi-Lipschitz parameterizations, Duke Math. J. 77 (1995), 



no. 1, 193-227. 



44 



Slawomir Kolasinski 

Instytut Matematyki 

Uniwersytet Warszawski 

ul. Banacha 2 

PL-02-097 Warsaw 

POLAND 

E-mail: skola@mimuw.edu.pl 

Pawel Strzelecki 

Instytut Matematyki 

Uniwersytet Warszawski 

ul. Banacha 2 

PL-02-097 Warsaw 

POLAND 

E-mail; pawel st@mimuw. edu .pi 

Heiko von der Mosel 

Institut fiir Mathematik 

RWTH Aachen University 

Templergraben 55 

D-52062 Aachen 

GERMANY 

Email: heikoSinstmath . rwth-aachen . de 



45