Variational formulas for submanifolds of fixed degree

We consider in this paper an area functional defined on submanifolds of fixed degree immersed into a graded manifold equipped with a Riemannian metric. Since the expression of this area depends on the degree, not all variations are admissible. It turns out that the associated variational vector fields must satisfy a system of partial differential equations of first order on the submanifold. Moreover, given a vector field solution of this system, we provide a sufficient condition that guarantees the possibility of deforming the original submanifold by variations preserving its degree. As in the case of singular curves in sub-Riemannian geometry, there are examples of isolated surfaces that cannot be deformed in any direction. When the deformability condition holds we compute the Euler-Lagrange equations. The resulting mean curvature operator can be of third order.


Introduction
The aim of this paper is to study the critical points of an area functional for submanifolds of given degree immersed in an equiregular graded manifold. This can be defined as the structure (N, H 1 , . . . , H s ), where N is a smooth manifold and H 1 ⊂ H 2 ⊂ · · · ⊂ H s = T N is a flag of sub-bundles of the tangent bundle satisfying [H i , H j ] ⊂ H i+j when i, j 1 and i + j s, and [H i , H j ] ⊂ H s when i, j 1 and i + j > s. The considered area depends on the degree of the submanifold. The concept of pointwise degree for a submanifold M immersed in a graded manifold was first introduced by Gromov in [28] as the homogeneous dimension of the tangent flag given by The degree of a submanifold deg(M ) is the maximum of the pointwise degree among all points in M . An alternative way of defining the degree is the following: on an open neighborhood of a point p ∈ N we can always consider a local basis (X 1 , . . . , X n ) adapted to the filtration (H i ) i=1,...,s , so that each X j has a well defined degree. Following [36] the degree of a simple m-vector X j1 ∧ . . . ∧ X jm is the sum of the degree of the vector fields of the adapted basis appearing in the wedge product. Since we can write a m-vector tangent to M with respect to the simple m-vectors of the adapted basis, the pointwise degree is given by the maximum of the degree of these simple m-vectors.
We consider a Riemannian metric g = ·, · on N . For any p ∈ N , we get an orthogonal decomposition T p N = K 1 p ⊕ . . . ⊕ K s p . Then we apply to g a dilation induced by the grading, which means that, for any r > 0, we take the Riemannian metric g r making the subspaces K i p orthogonal and such that Whenever H 1 is a bracket generating distribution the structure (N, g r ) converges in Gromov-Hausdorff sense to the sub-Riemannian structure (N, H 1 , g |H 1 ) as r → 0. Therefore an immersed submanifold M ⊂ N of degree d has Riemannian area measure A(M, g r ) with respect to the metric g r . We define area measure A d of degree d by when this limit exists and it is finite. In (3.7) we stress that the area measure A d of degree d is given by integral of the norm the g-orthogonal projection onto the subspace of m-forms of degree equal to d of the orthonormal m-vector tangent to M . This area formula was provided in [36,35] for C 1 submanifolds immersed in Carnot groups and in [19] for intrinsic regular submanifolds in the Heisenberg groups. Given a submanifold M ⊂ N of degree d immersed into a graded manifold (N, (H i ) i ), we wish to compute the Euler-Lagrange equations for the area functional A d . The problem has been intensively studied for hypersurfaces, and results appeared in [22,15,8,9,16,2,30,31,33,48,46,37,12]. For submanifolds of codimension greater than one in a sub-Riemannian structure only in the case of curves has been studied. In particular it is well know that there exists minimizers of the length functional which are not solutions of the geodesic equation: these curves, discovered by Montgomery in [38,39] are called abnormal geodesics. In this paper we recognize that a similar phenomenon can arise while studying the first variational of area for surfaces immersed in a graded structure: there are isolated surfaces which does not admit degree preserving variations. Consequently we focus on smooth submanifolds of fixed degree, and admissible variations, which preserve it. The associated admissible vector fields, V = ∂Γt ∂t t=0 satisfies the system of partial differential equations of first order (5.3) on M . So we are led to the central question of characterizing the admissible vector fields which are associated to an admissible variation.
The analogous integrability problem for geodesics in sub-Riemannian manifolds and, more generally, for functionals whose domain of definition consists of integral curves of an exterior differential system, was posed by E. Cartan [7] and studied by P. Griffiths [26], R. Bryant [3] and L. Hsu [32]. These one-dimensional problems have been treated by considering a holonomy map [32] whose surjectivity defines a regularity condition implying that any vector field satisfying the system (5.3) is integrable. In higher dimensions, there does not seem to be an acceptable generalization of such an holonomy map. However, an analysis of Hsu's regularity condition led the authors to introduce a weaker condition named strong regularity in [11]. This condition can be generalized to higher dimensions and provides a sufficient condition to ensure the local integrability of any admissible vector field on M , see Theorem 7.2. Indeed, in this setting the admissibility system (5.3) in coordinates is given by where C j , B, A are matrices, F are the vertical components of the admissible vector field, G are the horizontal control components andp ∈ M . Since the strong regularity tells us that the matrix A(p) has full rank we can locally write explicitly a part of the controls in terms of the vertical components and the other part of the controls, then applying the Implicit Function Theorem we produce admissible variations.
In Remark 7.6 we recognize that our definition of strongly regular immersion generalizes the notion introduced by [28] of regular horizontal immersions, that are submanifolds immersed in the horizontal distribution such that the degree coincides with the topological dimension m. In [27], see also [43], the author shows a deformability theorem for regular horizontal immersions by means of Nash's Implicit Function Theorem [41]. Our result is in the same spirit but for immersions of general degree.
For strongly regular submanifolds it is possible to compute the Euler-Lagrange equations to obtain a sufficient condition for stationary points of the area A d of degree d. This naturally leads to a notion of mean curvature, which is not in general a second order differential operator, but can be of order three. This behavior doesn't show up in the one-dimensional case where the geodesic equations for regular curves have order less than or equal to two, see [11,Theorem 7.2] or [32,Theorem 10].
These tools can be applied to mathematical model of perception in the visual cortex: G. Citti and A. Sarti in [12] showed that 2 dimensional minimal surfaces in the three-dimensional sub-Riemannian manifold SE(2) play an important role in the completion process of images, taking orientation into account. Adding curvature to the model, a four dimensional Engel structure arises, see § 1.5.1.4 in [45], [17] and § 4.3 here. The previous 2D surfaces, lifted in this structure are codimension 2, degree four strongly regular surfaces in the sense of our definition. On the other hand we are able to show that there are isolated surfaces which do not admit degree preserving variations. Indeed, in Example 7.8 we exhibit an isolated plane, immersed in the Engel group, whose only admissible normal vector field is the trivial one. Moreover, in analogy with the one-dimensional result by [4], Proposition 7.9 shows that this isolated plane is rigid in the C 1 topology, thus this plane is a local minimum for the area functional. Therefore we recognized that a similar phenomenon to the one of existence of abnormal curves can arise in higher dimension. Finally we conjecture that a bounded open set Ω contained in this isolated plane is a global minimum among all possible immersed surfaces sharing the same boundary ∂Ω.
We have organized this paper into several sections. In the next one notation and basic concepts, such as graded manifolds, Carnot manifolds and degree of submanifolds, are introduced. In Section 3 we define the area of degree d for submanifolds of degree d immersed in a graded manifold (N, H i ) endowed with a Riemannian metric. This is done as a limit of Riemannian areas. In addition, an integral formula for this area in terms of a density is given in formula (3.6). Section 4 is devoted to provide examples of submanifolds of certain degrees and the associated area functionals. In Sections 5 and 6 we introduce the notions of admissible variations, admissible vector fields and integrable vector fields and we study the system of first order partial differential equations defining the admissibility of a vector field. In particular, we show the independence of the admissibility condition for vector fields of the Riemannian metric in § 6.2. In Section 7 we give the notion of a strongly regular submanifold of degree d, see Definition 7.1. Then we prove in Theorem 7.2 that the strong regularity condition implies that any admissible vector vector is integrable. In addition, we exhibit in Example 7.8 an isolated plane whose only admissible normal vector field is the trivial one. Finally in Section 8 we compute the Euler-Lagrange equations of a strongly regular submanifold and give some examples.

Preliminaries
Let N be an n-dimensional smooth manifold. Given two smooth vector fields X, Y on N , their commutator or Lie bracket is defined by [X, Y ] := XY − Y X. An increasing filtration (H i ) i∈N of the tangent bundle T N is a flag of sub-bundles Moreover, we say that an increasing filtration is locally finite when (iii) for each p ∈ N there exists an integer s = s(p), the step at p, satisfying H s p = T p N . Then we have the following flag of subspaces ) is a smooth manifold N endowed with a locally finite increasing filtration, namely a flag of sub-bundles (2.1) satisfying (i),(ii) and (iii). For the sake of brevity a locally finite increasing filtration will be simply called a filtration. Setting n i (p) := dim H i p , the integer list (n 1 (p), · · · , n s (p)) is called the growth vector of the filtration (2.1) at p. When the growth vector is constant in a neighborhood of a point p ∈ N we say that p is a regular point for the filtration.
We say that a filtration (H i ) on a manifold N is equiregular if the growth vector is constant in N . From now on we suppose that N is an equiregular graded manifold.
Given a vector v in T p N we say that the degree of v is equal to ℓ if v ∈ H ℓ p and v / ∈ H ℓ−1 p . In this case we write deg(v) = ℓ. The degree of a vector field is defined pointwise and can take different values at different points.
Let (N, (H 1 , . . . , H s )) be an equiregular graded manifold. Take p ∈ N and consider an open neighborhood U of p where a local frame {X 1 , · · · , X n1 } generating H 1 is defined. Clearly the degree of X j , for j = 1, . . . , n 1 , is equal to one since the vector fields X 1 , . . . , X n1 belong to H 1 . Moreover the vector fields X 1 , . . . , X n1 also lie in H 2 , we add some vector fields X n1+1 , · · · , X n2 ∈ H 2 \ H 1 so that (X 1 ) p , . . . , (X n2 ) p generate H 2 p . Reducing U if necessary we have that X 1 , . . . , X n2 generate H 2 in U . Iterating this procedure we obtain a basis of T M in a neighborhood of p such that the vector fields X ni−1+1 , . . . , X ni have degree equal to i, where n 0 := 0. The basis obtained in (2.3) is called an adapted basis to the filtration (H 1 , . . . , H s ).
Given an adapted basis (X i ) 1 i n , the degree of the simple m-vector field X j1 ∧ . . . ∧ X jm is defined by Any m-vector X can be expressed as a sum where J = (j 1 , . . . , j m ), 1 j 1 < · · · < j m n, is an ordered multi-index, and X J := X j1 ∧ . . . ∧ X jm . The degree of X at p with respect to the adapted basis (X i ) 1 i n is defined by It can be easily checked that the degree of X is independent of the choice of the adapted basis and it is denoted by deg(X).
If X = J λ J X J is an m-vector expressed as a linear combination of simple m-vectors X J , its projection onto the subset of m-vectors of degree d is given by and its projection over the subset of m-vectors of degree larger than d by In an equiregular graded manifold with a local adapted basis (X 1 , . . . , X n ), defined as in (2.3), the maximal degree that can be achieved by an m-vector, m n, is the integer d m max defined by (2.5) d m max := deg(X n−m+1 ) + · · · + deg(X n ).

2.1.
Degree of a submanifold. Let M be a submanifold of class C 1 immersed in an equiregular graded manifold (N, (H 1 , . . . , H s )) such that dim(M ) = m < n = dim(N ). Then, following [34,36], we define the degree of M at a point p ∈ M by where v 1 , . . . , v m is a basis of T p M . Obviously, the degree is independent of the choice of the basis of T p M . Indeed, if we consider another basis   Proof. As p ∈ N is regular, there exists a local adapted basis (X 1 , . . . , X n ) in an open neighborhood U 2 ⊂ U 1 of p. We express the smooth vector field V in U 2 as on U 2 with respect to an adapted basis (X 1 , · · · , X n ), where c ij ∈ C ∞ (U 2 ). Suppose that the degree deg(V p ) of V at p is equal to d ∈ N. Then, there exists an integer k ∈ {n d−1 + 1, · · · , n d } such that c dk (p) = 0 and c ij (p) = 0 for all i = d + 1, · · · , s and j = n i−1 +1, · · · , n i . By continuity, there exists an open neighborhood U ′ ⊂ U 2 such that c dk (q) = 0 for each q in U ′ . Therefore for each q in U ′ the degree of V q is greater than or equal to the degree of V (p), Taking limits we get lim inf  2.2. Carnot manifolds. Let N be an n-dimensional smooth manifold. An ldimensional distribution H on N assigns smoothly to every p ∈ N an l-dimensional vector subspace H p of T p N . We say that a distribution H complies Hörmander's condition if any local frame {X 1 , . . . , X l } spanning H satisfies dim(L(X 1 , . . . , X l ))(p) = n, for all p ∈ N, where L(X 1 , . . . , X l ) is the linear span of the vector fields X 1 , . . . , X l and their commutators of any order. A Carnot manifold (N, H) is a smooth manifold N endowed with an l-dimensional distribution H satisfying Hörmander's condition. We refer to H as the horizontal distribution. We say that a vector field on N is horizontal if it is tangent to the horizontal distribution at every point. A C 1 path is horizontal if the tangent vector is everywhere tangent to the horizontal distribution. A sub-Riemannian manifold (N, H, h) is a Carnot manifold (N, H) endowed with a positive-definite inner product h on H. Such an inner product can always be extended to a Riemannian metric on N . Alternatively, any Riemannian metric on N restricted to H provides a structure of sub-Riemannian manifold. Chow's Theorem assures that in a Carnot manifold (N, H) the set of points that can be connected to a given point p ∈ N by a horizontal path is the connected component of N containing p, see [40]. Given a Carnot manifold (N, H), we have a flag of subbundles (2.9) The smallest integer s satisfying H s p = T p N is called the step of the distribution H at the point p. Therefore, we have The integer list (n 1 (p), · · · , n s (p)) is called the growth vector of H at p. When the growth vector is constant in a neighborhood of a point p ∈ N we say that p is a regular point for the distribution. This flag of sub-bundles (2.9) associated to a Carnot manifold (N, H) gives rise to the graded structure (N, (H i )). Clearly an equiregular Carnot manifold (N, H) of step s is an equiregular graded manifold (N, H 1 , . . . , H s ). In particular a Carnot group turns out to be an equiregular graded manifold.
Given a connected sub-Riemannian manifold (N, H, h), and a C 1 horizontal path γ : [a, b] → N , we define the length of γ by By means of the equality

Area for submanifolds of given degree
In this section we shall consider a graded manifold (N, H 1 , . . . , H s ) endowed with a Riemannian metric g, and an immersed submanifold M of dimension m.
We recall the following construction from [28, 1.4.D]: given p ∈ N , we recursively define the subspaces K 1 Here ⊥ means perpendicular with respect to the Riemannian metric g. Therefore we have the decomposition of T p N into orthogonal subspaces Given r > 0, a unique Riemannian metric g r is defined under the conditions: (i) the subspaces K i are orthogonal, and (ii) When we consider Carnot manifolds, it is well-known that the Riemannian distances of (N, g r ) uniformly converge to the Carnot-Carathéodory distance of (N, H, h), [28, p. 144].
Working on a neighborhood U of p where a local frame (X 1 , . . . , X k ) generating the distribution H is defined, we construct an orthonormal adapted basis (X 1 , . . . , X n ) for the Riemannian metric g by choosing orthonormal bases in the orthogonal subspaces K i , 1 i s. Thus, the m-vector fields where J = (j 1 , j 2 , . . . , j m ) for 1 j 1 < · · · < j m n, are orthonormal with respect to the extension of the metric g r to the space of m-vectors. We recall that the metric g r is extended to the space of m-vectors simply defining Observe that the extension is denoted the same way.
3.1. Area for submanifolds of given degree. Assume now that M is an immersed submanifold of dimension m in a equiregular graded manifold (N, H 1 , . . . , H s ) equipped with the Riemannian metric g. We take a Riemannian metric µ on M . For any p ∈ M we pick a µ-orthonormal basis e 1 , . . . , e m in T p M . By the area formula we get where M ′ is a bounded measurable subset of M and A(M ′ , g r ) is the m-dimensional area of M ′ with respect to the Riemannian metric g r . Now we express By Lebesgue's dominated convergence theorem we obtain for any bounded measurable set M ′ ⊂ M .
Equation (3.6) provides an integral formula for the area A d . An immediate consequence of the definition is the following Remark 3.2. Setting d := deg(M ) we have by equation (3.6) and the notation introduced in (2.4) that the degree d area A d is given by for any bounded measurable set M ′ ⊂ M . When the ambient manifold is a Carnot group this area formula was obtained by [36]. Notice that the d area A d is given by the integral of the m-form In a more general setting, an m-dimensional submanifold in a Riemannian manifold is an m-current (i.e., an element of the dual of the space of m-forms), and the area is the mass of this current (for more details see [18]). Similarly, a natural generalization of an m-dimensional submanifold of degree d immersed in a graded manifold is an m-current of degree d whose mass should be given by A d . In [19] the authors studied the theory of H-currents in the Heisenberg group. Their mass coincides with our area (3.7) on intrinsic C 1 submanifolds. However in (3.8) we consider all possible m-forms and not only the intrinsic m-forms in the Rumin's complex [49,42,1].
Since M 0 is measurable, from (3.6) we obtain Remark 3.4. Another easy consequence of the definition is the following: if M is an immersed submanifold of degree d in graded manifold (N, This follows easily since in the expression we would have summands with negative exponent for r. In the following example, we exhibit a Carnot manifold with two different Riemannian metrics that coincide when restricted to the horizontal distribution, but yield different area functionals of a given degree Example 3.5. We consider the Carnot group H 1 ⊗ H 1 , which is the direct product of two Heisenberg groups. Namely, let R 3 ×R 3 be the 6-dimensional Euclidean space with coordinates (x, y, z, x ′ , y ′ , z ′ ). We consider the 4-dimensional distribution H generated by Let Ω be a bounded open set of R 2 and u a smooth function on Ω such that u t (s, t) ≡ 0. We consider the immersed surface Thus, the 2-vector tangent to M is given by When u s (s, t) is different from zero the degree is equal to 3, since both Z ∧ Y ′ and Z ′ ∧ Y ′ have degree equal to 3. Points of degree 2 corresponds to the zeroes of u s . We define a 2-parameter family g λ,ν of Riemannian metrics on H 1 ⊗ H 1 , for (λ, µ) ∈ R 2 , by the conditions (i) (X, Y, X ′ , Y ′ ) is an orthonormal basis of H, (ii) Z, Z ′ are orthogonal to H, and (iii) g(Z, Z) = λ, g(Z ′ , Z ′ ) = µ and g(Z ′ , Z) = 0. Therefore, the degree 3 area of Ω with respect to the metric g µ,ν is given by As we shall see later, these different functionals will not have the same critical points, that would depend on the election of Riemannian metric. Writing for non-negative integers z i and adding up on i from 1 to s we get sincem s = n − 1 and n s = n. We conclude that there exists i 0 ∈ {1, . . . , s} such that z i0 = 1 and z j = 0 for all j = i 0 . This implies If i 0 > 1 for all p ∈ M , then H ⊂ T M , a contradiction since H is a bracketgenerating distribution. We conclude that i 0 = 1 and so for X, Y ∈ H, the distribution H is non-integrable and satisfies Hörmander rank condition by Frobenius theorem. When we define a horizontal metric h on the distribution H then (M, H, h) is a sub-Riemannian structure. It is easy to prove that there exists an unique vector field T on M so that where L is the Lie derivative and X is any vector field on M . This vector field T is called the Reeb vector field. We can always extend the horizontal metric h to the Riemannian metric g making T a unit vector orthogonal to H. Let Σ be a C 1 hypersurface immersed in M . In this setting the singular set of Σ is given by and corresponds to the points in Σ of degree 2n. Observe that the non-integrability of H implies that the set Σ Σ 0 is not empty in any hypersurface Σ. Let N be the unit vector field normal to Σ at each point, then on the regular set Σ Σ 0 the g-orthogonal projection N h of N onto the distribution H is different from zero. Therefore out of the singular set Σ 0 we define the horizontal unit normal by and the vector field which is tangent to Σ and belongs to H 2 . Moreover, T p Σ∩(H 2 p H 1 p ) has dimension equal to one and T p Σ ∩ H 1 p equal to 2n − 1, thus the degree of the hypersurface Σ out of the singular set is equal to 2n + 1. Let e 1 , . . . , e 2n−1 be an orthonormal basis in T p Σ ∩ H 1 p . Then e 1 , . . . , e 2n−1 , S p is an orthonomal basis of T p Σ and we have Hence we obtain In [20] Galli obtained this formula as the perimeter of a set that has C 1 boundary Σ and in [50] Shcherbakova as the limit of the volume of a ε-cylinder around Σ over its height equal to ε. This formula was obtain for surfaces in a 3-dimensional pseudo-hermitian manifold in [9] and by S. Pauls in [44]. This is exactly the area formula independently established in recent years in the Heisenberg group H n , that is the prototype for contact manifolds (see for instance [15,9,10,47,30]).
Example 4.1 (The roto-translational group). Take coordinates (x, y, θ) in the 3dimensional manifold R 2 × S 1 . We consider the contact form the horizontal distribution H = ker(ω), is spanned by the vector fields and the horizontal metric h that makes X and Y orthonormal. Therefore R 2 × S 1 endowed with this one form ω is a contact manifold. Moreover (R 2 × S 1 , H, h) has a sub-Riemannian structure which is also a Lie group known as the roto-translational group. A mathematical model of simple cells of the visual cortex V1 using the sub-Riemannian geometry of the roto-translational Lie group was proposed by Citti and Sarti (see [13], [14]). Here the Reeb vector field is given by Let Ω be an open set of R 2 and u : Ω → R be a function of class C 1 . When we consider a graph Σ = Graph(u) given by the zero set level of the C 1 function f (x, y, θ) = u(x, y) − θ = 0, the projection of the unit normal N onto the horizontal distribution is given by Hence the 3-area functional is given by Therefore (E, H) is a Carnot manifold, indeed H satisfy the Hörmander rank condition since X 1 and X 2 generate all the tangent bundle. Here we follow a computation developed by Le Donne and Magnani in [34] in the Engel group. Let Ω be an open set of R 2 endowed with the Lebesgue measure. Since we are particularly interested in applications to the visual cortex (see [23], [45, 1.5.1.4] to understand the reasons) we consider the immersion Φ : Ω → E given by Φ = (x, y, θ(x, y), κ(x, y)) and we set Σ = Φ(Ω).
The tangent vectors to Σ are In order to know the dimension of T p Σ ∩ H p it is necessary to take in account the rank of the matrix Obviously rank(B) 3, indeed we have Moreover, it holds (4.6) , y)).
Since we are inspired by the foliation property of hypersurface in the Heisenberg group and roto-translational group, in the present work we consider only surface Σ = {(x, y, θ(x, y), κ(x, y))} verifying the foliation condition κ = X 1 (θ(x, y)). Thus, we have By the foliation condition (4.6) we have that the coefficient of X 3 ∧ X 4 is always equal to zero, then we deduce that deg(Σ) 4. Moreover, the coefficient of X 1 ∧X 4 never vanishes, therefore deg(Σ) = 4 and there are not singular points in Σ. When κ = X 1 (θ) a tangent basis of T p Σ adapted to 2.7 is given by When we fix the Riemannian metric g 1 that makes (X 1 , . . . , X 4 ) orthonormal we have that the A 4 -area of Σ is given by When we fix the Euclidean metric g 0 that makes (∂ 1 , ∂ 2 , ∂ θ , ∂ k ) we have that the A 4 -area of Σ is given by

Admissible variations for submanifolds
Let us consider an m-dimensional manifoldM and an immersion Φ :M → N into an equiregular graded manifold endowed with a Riemannian metric g = ·, · . We shall denote the image Φ(M ) by M and d := deg(M ). In this setting we have the following definition Let us see now that the variational vector field V associated to an admissible variation Γ satisfies a differential equation of first order. Let p = Φ(p) for somē p ∈M , and (X 1 , · · · , X n ) an adapted frame in a neighborhood U of p. Take a basis (ē 1 , . . . ,ē m ) of TpM and let e j = dΦp(ē j ) for 1 j m. As Γ t (M ) is a submanifold of the same degree as Φ(M ) for small t, there follows for all X J = X j1 ∧ . . . ∧ X jm , with 1 j 1 < · · · < j m n, such that deg(X J ) > deg(M ). Taking the derivative with respect to t in equality (5.2) and evaluating at t = 0 we obtain the condition for all X J such that deg(X J ) > deg(M ). In the above formula, ·, · indicates the scalar product in the space of m-vectors induced by the Riemannian metric g. The symbol ∇ denotes, in the left summand, the Levi-Civita connection associated to g and, in the right summand, the covariant derivative of vectors in X(M , N ) induced by g. Thus, if a variation preserves the degree then the associated variational vector field satisfies the above condition and we are led to the following definition. Thus we are led naturally to a problem of integrability: given V ∈ X 0 (M , N ) such that the first order condition (5.3) holds, we ask whether an admissible variation whose associated variational vector field is V exists.
Definition 5.4. We say that an admissible vector field V ∈ X 0 (M , N ) is integrable if there exists an admissible variation such that the associated variational vector field is V .
To get an m-vector in such a basis we pick any of the k 1 vectors in H 1 p ∩{v 1 , . . . , v n } and, for j = 2, . . . , s, we pick any of the k j vectors on (H j p H j−1 p ) ∩ {v 1 , . . . , v n }, so that • k 1 + · · · + k s = m, and • 1 · k 1 + · · · + s · k s d. So we conclude, taking n 0 = 0, that When we consider two simple m-vectors v i1 ∧ . . . ∧ v im and v j1 ∧ . . . ∧ v jm , their scalar product is 0 or ±1, the latter case when, after reordering if necessary, we have v i k = v j k for k = 1, . . . , m. This implies that the orthogonal subspace Λ d Hence we have Then we can choose an orthonormal basis (X J1 , . . . , X J ℓ ) in Λ d m (U ) ⊥ p at each point p ∈ U . 6.1. The admissibility system with respect to an adapted local basis.
In the same conditions as in the previous subsection, let ℓ = dim(Λ d m (U ) ⊥ p ) and (X J1 , . . . , X J ℓ ) an orthonormal basis of Λ d m (U ) ⊥ p . Any vector field V ∈ X(M , N ) can be expressed in the form  We write so that the local system (6.2) can be written as where c ijr is defined in (6.3) and, for 1 i ℓ, where β ij is defined in (6.4). We denote by B the ℓ × (n − ρ) matrix whose entries are b ir , by A the ℓ × ρ whose entries are a ih and for j = 1, . . . , m we denote by C j the ℓ × (n − ρ) matrix C j = (c ijh ) i=1,...,ℓ h=ρ+1,...,n . Setting Since the adapted change of basis preserves the degree of the m-vectors, the square matrix Λ = (λ JI ) of order n m acting on the m-vector is given by where Λ h and Λ v are square matrices of order n m − ℓ and ℓ respectively and Λ hv is a matrix of order n m − ℓ × ℓ. Moreover the matrix Λ is invertible since both {X J } and {Y I } are basis of the vector space of m-vectors. Remark 6.3. One can easily check that the inverse of Λ is given by the block matrix LetÃ be the associated matrix Setting and Ω = Ω h Ω v = (ω Jr ) r=1,...,n deg(J) d , a straightforward computation shows By Remark 6.3 we obtain Therefore, settingC LetC j be the associated matrix it is immediate to obtain the following equality LetB be the associated matrix A straightforward computation shows By Remark 6.3 we obtain (6.14) Finally, we have G = D hG + D hvF and F = D vF .
Proposition 6.4. Let g andg be two different metrics, then a vector fields V is admissible w.r.t. g if and only if V is admissible w.r.t.g.
Proof. We remind that an admissible vector field By (6.11), (6.14) and (6.13) we have In the previous equation we used that G = D hG + D hvF , F = D vF and Then the admissibility system (6.15) w.r.t. g is equal to zero if and only if the admissibility system (6.16) w.r.t. g.
Remark 6.5. When the metric g is fixed and (X i ) and (Y i ) are orthonormal adapted basis w.r.t g, the matrix D is a block diagonal matrix given by where D h and D v are square orthogonal matrices of orders ρ and (n − ρ), respectively. From equations (6.11), (6.14), (6.13) it is immediate to obtain the following equalitiesF where ψ 1 , . . . , ψ n ∈ C r (Φ −1 (U ), R). By Proposition 5.5 we deduce that V is admissible if and only if V ⊥ = n h=m+1 ψ h V h is admissible. Hence we obtain that the system (5.3) is equivalent to Definition 6.6. Let ι 0 (U ) be the integer defined in 6.1. Then we set k := n ι0 −m ι0 .

Integrability of admissible vector fields
In general, given an admissible vector field V , the existence of an admissible variation with associated variational vector field V is not guaranteed. The next definition is a sufficient condition to ensure the integrability of admissible vector fields.
Definition 7.1. Let Φ :M → N be an immersion of degree d of an m-dimensional manifold into a graded manifold endowed with a Riemannian metric g. Let ℓ = dim(Λ d m (U ) ⊥ q ) for all q ∈ N and ρ = n ι0 set in (6.1). When ρ ℓ we say that Φ is strongly regular atp ∈M if rank(A(p)) = ℓ, where A is the matrix appearing in the admissibility system (6.9).
The rank of A is independent of the local adapted basis chosen to compute the admissibility system (6.9) because of equations (6.17). Next we prove that strong regularity is a sufficient condition to ensure local integrability of admissible vector fields. We can rewrite the system (6.9) in the form where i 1 , . . . , i ρ−ℓ are the indexes of the columns of A that do not appear inÂ and A is the ℓ × (ρ − ℓ) matrix given by the columns i 1 , . . . , i ρ−ℓ of A. The vectors (E i ) i form an orthonormal basis of TM nearp.
On the neighborhood Wp we define the following spaces 1. X r 0 (Wp, N ), r 0 is the set of C r vector fields compactly supported on Wp taking values in T N . N ), and consider the map where Π v is the projection in the space of m-forms with compact support in Wp onto Λ r (Wp, N ), and Observe that F (Y ) = 0 if and only if the submanifold Γ(Y ) has degree less or equal to d. We consider on each space the corresponding || · || r or || · || r−1 norm, and a product norm. Then where we write in coordinates g hi X hi , and Y 3 = n r=ρ+1 f r X r .
Following the same argument we used in Section 5, taking the derivative at t = 0 of (5.2), we deduce that the differential DF (0)Y is given by Oberve that DF (0)Y = 0 if and only if Y is an admissible vector field, namely Y solves (7.1).
Our objective now is to prove that the map DG(0, 0, 0) is an isomorphism of Banach spaces.
where with an abuse of notation we identify Z 3 = ℓ i=1 z i X Ji and ℓ i=1 z i X hi . SinceÂ is invertible we have the following system Clearly Y 1 = Z 1 fixes g i1 , . . . , g i ρ−ℓ in (7.3), and Y 2 = Z 2 fixes the first and second term of the right hand side in (7.3). Since the right side terms are given we have determined Y 3 , i.e. g h1 , . . . , g h ℓ , such that Y 3 solves (7.3). Therefore DG(0, 0, 0) is surjective. Thus we have proved that DG(0, 0, 0) is a bijection. Let us prove now that DG(0, 0, 0) is a continuous and open map. Letting , we first notice DG(0, 0, 0) is a continuous map since identity maps are continuous and, by (7.3), there exists a constant K such that Moreover, DG(0, 0, 0) is an open map since we have This implies that DG(0, 0, 0) is an isomorphism pf Banach spaces.
Let now us consider an admissible vector field V with compact support on W p . We consider the map The mapG is continuous with respect to the product norms (on each factor we put the natural norm, the Euclidean one on the intervals and || · || r and || · || r−1 in the spaces of vectors on Φ(M )). Moreover G(0, 0, 0, 0) = (0, 0), since Φ has degree d. Denoting by D Y the differential with respect to the last three variables ofG we have that is a linear isomorphism. We can apply the Implicit Function Theorem to obtain unique maps such thatG(s, Y 1 (s), Y 2 (s), Y 3 (s)) = (0, 0). This implies that Y 1 (s) = 0, Y 2 (s) = 0, Y 3 (0) = 0 and that F (sV + Y 3 (s)) = 0.
Differentiating this formula at s = 0 we obtain Since V is admissible we deduce (Wp), equation (7.1) implies g hi ≡ 0 for each i = 1, . . . , ℓ. Therefore it follows ∂Y3 ∂s (0) = 0. Hence the variation Γ s (p) = Γ(sV + Y 3 (s))(p) coincides with Φ(q) for s = 0 and q ∈ Wp, it has degree d and its variational vector fields is given by Moreover, supp(Y 3 ) ⊆ supp(V ). Indeed, ifq / ∈ supp(V ), the unique vector field Y 3 (s), such F (Y 3 (s)) = 0, is equal to 0 atq. Remark 7.3. In Proposition 5.5 we stressed the fact that a vector field V = V ⊤ + V ⊥ is admissible if and only if V ⊥ is admissible. This follows from the additivity in V of the admissibility system (5.3) and the admissibility of V ⊤ . Instead of writing V with respect to the adapted basis (X i ) i we consider the basis E 1 , . . . , E m , V m+1 , . . . , V n described in Section 6.3.
Let A ⊥ , B ⊥ , C ⊥ be the matrices defined in (6.22), A ⊤ be the one described in Remark 6.7 and A be the matrix with respect to the basis (X i ) i defined in (6.7). When we change only the basis for the vector field V by (6.11) we obtaiñ A = AD h . Since A ⊤ is the null matrix andÃ = (A ⊤ | A ⊥ ) we conclude that rank(A(p)) = rank(A ⊥ (p)). Furthermore Φ is strongly regular atp if and only if rank(A ⊥ (p)) = ℓ k, where k is the integer defined in 6.6.

Some examples of regular submanifolds.
Example 7.4. Consider a hypersurface Σ immersed in an equiregular Carnot manifold N , then we have that Σ always has degree d equal to d n−1 max = Q − 1, see 4.1. Therefore the dimension ℓ, defined in Section 6, of Λ d m (U ) p is equal to zero. Thus any compactly supported vector field V is admissible and integrable. When the Carnot manifold N is a contact structure (M 2n+1 , H = ker(ω)), see 4.2, the hypersurface Σ has always degree equal to d 2n max = 2n + 1. Example 7.5. Let (E, H) be the Carnot manifold described in Section 4.3 where (x, y, θ, k) ∈ R 2 × S 1 × R = E and the distribution H is generated by Clearly (X 1 , . . . , X 4 ) is an adapted basis for H. Moreover the others no-trivial commutators are given by Let Ω ⊂ R 2 be an open set. We consider the surface Σ = Φ(Ω) where Φ(x, y) = (x, y, θ(x, y), κ(x, y)) and such that X 1 (θ(x, y)) = κ(x, y). Therefore the deg(Σ) = 4 and its tangent vectors are given byẽ Let g = ·, · be the metric that makes orthonormal the adapted basis (X 1 , . . . , X 4 ). Since (Λ 4 2 (N )) ⊥ = span{X 3 ∧ X 4 } the only no-trivial coefficient c 11r , for r = 3, 4 are given by On the other hand c 12h = ẽ 1 ∧ X k , X 3 ∧ X 4 = 0 for each h = 1, . . . , 4, since we can not reach the degree 5 if one of the two vector fields in the wedge has degree one. Therefore the only equation in (6.2) is given by Thus, we deduce Hence the equation (7.5) is equivalent to Since ι 0 (Ω) = 1, we have ρ = n 1 = 2, where ρ is the natural number defined in (6.1). In this setting the matrix C is given by Then the matrices A and B are given by Since rank(A(x, y)) = 1 and the matrixÂ(x, y), defined in the proof of Theorem 7.2, is equal to 1 for each (x, y) ∈ Ω we have that Φ is strongly regular at each point (x, y) in Ω and the open set W (x,y) = Ω. Hence by Theorem 7.2 each admissible vector field on Ω is integrable.
On the other hand we notice that k = n 1 −m 1 = 1. By the Gram-Schmidt process an orthonormal basis with respect to the metric g is given by where we set then a vector field V ⊥ = ψ 3 (x, y) v 3 + ψ 4 (x, y) v 4 normal to Σ is admissible if and only if ψ 3 , ψ 4 ∈ C r 0 (Ω) verify That is equivalent to In particular, since a ⊥ (x, y) > 0 we have that rank(a ⊥ (x, y)) = 1 for all (x, y) ∈ Ω. Along the integral curve γ ′ (t) =X 1 on Ω the equation (7.7) reads for each function f : Ω → R.
Remark 7.6. Let (N, H) be a Carnot manifold such that H = ker(θ) where θ is a R n−ℓ one form. Following [28,43] we say that an immersion Φ :M → N is horizontal when the pull-back Φ * θ = 0 and, given a point p ∈ Φ(M ), the subspace T p M ⊂ H p is regular if the map is onto for each horizontal vector V onM . Let X be an horizontal extension of V on N and Y be another horizontal vector field on N , then Assume that the local frame E 1 , . . . , E m generate T p M at p then the map (7.8) is given by θ([X, E j ](p)), for each j = 1, . . . , m. In [24, Section 3] the author notice that there exist special coordinates adjusted to the admissibility system such that the entries of the control matrix A are . . , V n are vector fields in the normal bundle. In this notation the surjectivity of this map coincides with the pointwise condition of maximal rank of the matrix (a ijh ). Since by equation (6.17) the rank of A is independent of the metric g we deduce that this regularity notion introduced by [28,27] is equivalent to strongly regularity at p (Definition 7.1) for the class of horizontal immersions. Here we provide an example of isolated surface immersed in the Engel group.
Since Υ v ∧ Υ w = X 1 ∧ X 3 the degree deg(Σ) = 3, where Σ = Υ(Ω) is a plane. An admissible vector field V = 4 k=1 f k X k verifies the system (6.2) that is given by Let K = supp(V ). First of all we have ∂f4 ∂x1 = 0. Since f 4 ∈ C ∞ (Ω) there follows Then let (x 1 , x 2 ) ∈ K we consider the curve along which f 4 and f 2 are constant. Since f 4 and f 2 are compactly supported at the end point, (x 1 +s 0 , x 3 ) ∈ ∂K we have f 4 (x 1 +s 0 , x 3 ) = f 2 (x 1 +s 0 , x 3 ) = 0. Therefore we gain f 4 = f 2 ≡ 0. Therefore the only admissible vector fields f 1 X 1 + f 3 X 3 are tangent to Σ. Assume that there exists an admissible variation Γ s for Υ, then its associated variational vector field is admissible. However we proved that the only admissible vector fields are tangent to Σ, therefore the admissible variation Γ s has to be tangent to Σ and the only normal one a trivial variation, hence we conclude that the plane Σ is isolated. Moreover, we have that k = 1 and the matrix A ⊥ defined in 7.1 is given by Since rank(A) = 1 < 3 we deduce that Υ is not strongly regular at any point in Ω.
In analogy with the rigidity result by [4], here we prove that Σ is isolated without using the admissibility system. This also implies that the plane Σ is rigid in the C 1 topology. Proposition 7.9. Let E 4 be the Engel group given by (R 4 , H), where the distribution H is generated by Let Ω ⊂ R 2 be a bounded open set. Then the immersion Υ : Ω → E 4 of degree 3 given by Υ(v, w) = (v, 0, w, 0) is isolated.
Proof. An admissible normal variation Γ s of Υ has to have the same degree of Υ and has to share the same boundary Υ(∂Ω) = ∂Σ, where clearly Σ = Υ(Ω). For a fix s, we can parametrize Γ s by where φ, ψ ∈ C 1 0 (Ω, R). Since deg(Φ(Ω)) = 3 we gain (7.10) Denoting by π 4 the projection over the 2-vectors of degree larger than 3, we have Therefore (7.10) is equivalent to The second equation implies that (7.11) is equivalent to Then we notice that the first and the third equations implies the second one as it follows Therefore the immersion Φ has degree three if and only if Only when the compatibility conditions ([29, Eq. (1.4), Chapter VI]) for linear system of first order are given we have a solution of this system. However the compatibility condition is given by Since φ ∈ C 1 0 (Ω) we obtain φ ≡ 0. Therefore also ψ v = 0, then ψ ≡ 0. Hence Φ = Υ.

First variation formula for submanifolds
In this section we shall compute a first variation formula for the area A d of a submanifold of degree d. We shall give some definitions first. Assume that Φ :M → N is an immersion of a smooth m-dimensional manifold into an ndimensional equiregular graded manifold endowed with a Riemannian metric g. Let µ = Φ * g. Fixp ∈M and let p = Φ(p). Take a µ-orthonormal basis (ē 1 , . . . ,ē m ) in TpM and define e i := dΦp(ē i ) for i = 1, . . . , m. Then the degree d area density Θ is defined by Assume now that V ∈ X(M , N ), then we set Finally, define the linear function f by Proof. By the definition of divergence we obtain (i) as follows To deduce (ii) we apply twice (i) as follows Theorem 8.5. Let Φ :M → N be an immersion of degree d of a smooth mdimensional manifold into an equiregular graded manifold equipped with a Riemannian metric g. Assume that there exists an admissible variation Γ :M × (−ε, ε) → N with associated variational field V with compact support. Then In this formula, (E i ) i is a local orthonormal basis of T M and (N j ) j a local orthonormal basis of T M ⊥ . The functions ξ ij are given by Proof. Since our computations are local and immersions are local embeddings, we shall identify locallyM and M to simplify the notation. We decompose V = V ⊤ + V ⊥ in its tangential V ⊤ and perpendicular V ⊥ parts. Since div dM and the functional f defined in (8.3) are additive, we use the first variation formula (8.4) and Proposition 8.3 to obtain To compute this integrand we consider a local orthonormal basis (E i ) i in T M around p and a local orthonormal basis (N j ) j of T M ⊥ with (N j ) j . We have We compute first (8.8) The group of summands in the second line of (8.8) is equal to V, H 2 , where To treat the group of summands in the first line of (8.8) we use (ii) in Lemma 8.4. recalling (8.7) we have so that applying the Divergence Theorem we have that the integral in M of the first group of summands in (8.8) is equal to We treat finally the summand This implies the result since H d = H 1 + H 2 + H 3 .
In the following result we obtain a slightly different expression for the mean curvature H d in terms of Lie brackets. This expression is sometimes more suitable for computations.
A straightforward computation shows that ξ i3 for i = 1, 2 defined in (8.9) are given by it follows that the third component of H d is equal to and the fourth component of H d is equal to Then first variation formula is given by for each ψ 3 , ψ 4 ∈ C ∞ 0 satisfying (7.7). Following Theorem 7.2 for each ψ 3 ∈ C ∞ 0 we deduce since a ⊥ > 0.