We show the fundamental theorems of curves and surfaces in the 3-dimensional Heisenberg group and find a complete set of invariants for curves and surfaces respectively. The proofs are based on Cartan’s method of moving frames and Lie group theory. As an application of the main theorems, a Crofton-type formula is proved in terms of p-area which naturally arises from the variation of volume. The application makes a connection between CR geometry and integral geometry.
A horizontally regular curve is a parametrized curve \gamma(t) such that \gamma^{\prime}_{\xi}(t)\neq 0 for all t\in I. We say that \gamma (t) is a horizontal curve if \gamma ^{\prime}(t)=\gamma _{\xi }^{\prime}(t) for all t\in I.
Given C^1-functions k(s), \tau(s), there exists a horizontally regular curve \gamma(s) with horizontal unit-speed having k(s) and \tau(s) as its p-curvature and contact normality, respectively. In addition, any regular curve \widetilde\gamma(s) with horizontal unit-speed satisfying the same p-curvature k(s) and contact normality \tau(s) differs from \gamma(s) by a pseudo-hermitian transformation g\in {\rm PSH}(1), namely,
for all s.
Given a C^1-function k(s), there exists a horizontal curve \gamma(s) with horizontal unit-speed having k(s) as its p-curvature. In addition, any horizontal curve \widetilde\gamma(s) with horizontal unit-speed satisfying the same p-curvature differs from \gamma(s) by a pseudo-hermitian transformation g\in {\rm PSH}(1), namely,
for all s.
Let \gamma(t)= (x(t),y(t),z(t) )\in H_1 be a horizontally regular curve, not necessarily with horizontal unit-speed. The p-curvature k(t) and the contact normality \tau(t) of \gamma(s) are
Suppose two horizontally regular curves in H_1 differ by a Heisenberg rigid motion, then their projections onto the xy-plane along the z-axis differ by a Euclidean rigid motion. In particular, two horizontal curves in H_1 differ by a Heisenberg rigid motion if and only if their projections are congruent in the Euclidean plane.
In H_1, the geodesics are the horizontally regular curves with constant p-curvature and zero contact normality.
Let F\colon U\rightarrow H_1 be a parametrized surface with coordinates (u,v) on U\subset \mathbb{R}^2. We say that F is a normal parametrization if
=0pt
1) F(U) is a surface without singular points,
2) F_{u}:=\frac{\partial F}{\partial u} defines the characteristic foliation on F(U),
3) |F_{u}|=1 for each point (u,v)\in U, where the norm is with respect to the Levi-metric.
We call (u,v) normal coordinates of the surface F(U).
Let U\subset \mathbb{R}^{2} be a simply connected open set. Suppose that a, b, c, l and m are functions defined on U satisfying the integrability conditions (?). Then there exists a normal parametrization F\colon U\rightarrow H_1 having a, b, c and l, m as the coefficients of first kind and second kind of F, respectively. In addition, any normal parametrization \widetilde{F}\colon U\rightarrow H_1 with the same coefficients of first kind and second kind differ from F by a Heisenberg rigid motion, namely, \widetilde{F}(u,v)=g\circ F(u,v) for all (u,v)\in U for some g\in {\rm PSH}(1).
Let U\subset \mathbb{R}^{2} be a simply connected open set. Suppose that a, b and c are smooth functions defined on U satisfying the integrability conditions (?). Then there exists a normal parametrization of p-minimal surface F\colon U\rightarrow H_1 having a, b and c as the coefficients of first kind of F, which also determines the coefficient m of the second kind as in (?). In addition, any normal parametrization of p-minimal surface \widetilde{F}\colon U\rightarrow H_1 with the same conditions differs from F by a Heisenberg rigid motion, namely, \widetilde{F}(u,v)=g\circ F(u,v) in U for some g\in {\rm PSH}(1).
Let (\Sigma ,g) be a 2-dimensional Riemannian manifold with Gaussian curvature K, and \alpha^\prime, l^\prime two real-valued functions defined on \Sigma. Assume that K, \alpha^{\prime} and l^{\prime} satisfy the integrability condition (?). Then for every non-singular point p\in \Sigma, there exists an open neighborhood U containing p and an embedding f\colon U\rightarrow H_1 such that
where \alpha, l are the induced p-variation and p-mean curvature on f(U) respectively. Moreover, f is uniquely determined up to a Heisenberg rigid motion.
Suppose \mathbb{X}\colon (u,v)\in\Omega \mapsto \Sigma\subset H_1 is a C^2-surface for some domain \Omega \subset \mathbb{R}^2. Let \mathcal{L} be the set of oriented horizontal lines in H_1 and n(\ell\cap\Sigma) be the number of intersections of the horizontal line \ell\in\mathcal{L} with the surface \Sigma. Then we have the Crofton formula
where {\rm d}L:= {\rm d}p\wedge {\rm d}\theta \wedge {\rm d}t is the kinematic density on \mathcal{L}.
Given two maps f, \widetilde{f}\colon M\rightarrow G, then \widetilde{f}^{*}\omega=f^{*}\omega if and only if \widetilde{f}=g\cdot f for some g\in G.
Suppose that \phi is a \mathfrak{g}-valued 1-form on a simply connected manifold M. Then there exists a map f\colon M\rightarrow G satisfying f^{*}\omega=\phi if and only if {\rm d}\phi=-\phi \wedge \phi. Moreover, the resulting map f is unique up to a group action.
Let \Phi\colon H_{1}\rightarrow H_{1} be a pseudo-hermitian transformation. Then \Phi=L_{p}\circ \Phi_{R} for some R\in {\rm SO}(2) and p\in H_{1}.
t suffices to consider the pseudo-hermitian transformation \Phi\colon H_{1}\rightarrow H_{1} such that \Phi(0)=0. Indeed, if \Phi(0)=p for some p\in H_1\setminus\{0\}, then the composition L_{p^{-1}}\circ \Phi is a transformation fixing the origin. Therefore, we reduce the proof of Theorem 3.1 to the following lemma:
Let \Phi be a pseudo-hermitian transformation on H_{1} such that \Phi (0)=0. Then, for any p\in H_{1}, the matrix representation of \Phi _{\ast }(p) with respect to the standard basis \big(\frac{\partial }{\partial x},\frac{\partial }{\partial y},\frac{\partial }{\partial z}\big) of \mathbb{R}^3 is
for some real constant \alpha _{0} which is independent of p, and hence \Phi _{\ast } is a constant matrix.
To prove Lemma I, first we calculate the matrix representation of \Phi _{\ast }(p) with respect to the basis (\mathring{e}_{1},\mathring{e}_{2},T). Since
the contact bundle \xi is invariant under \Phi _{\ast }. In addition, let h be the Levi-metric on \xi defined by h(X,Y)={\rm d}\Theta(X,JY), then
and hence h ( \Phi _{\ast }X,\Phi _{\ast }Y) =h( X,Y) for every X,Y\in \xi. Thus, \Phi _{\ast } is orthogonal on \xi. On the other hand, since
and
for all X\in \xi, we have \Phi _{\ast}T=T. From the above argument, we conclude that the matrix representation
for some real-valued function \alpha on H_{1}.
Next, we rewrite the matrix representation of \Phi _{\ast }(p) from the basis \big( \mathring{e} _{1},\mathring{e}_{2},\frac{\partial }{\partial z}\big) to the basis \big( \frac{ \partial }{\partial x},\frac{\partial }{\partial y},\frac{\partial }{ \partial z}\big). Let \Phi =(\Phi ^{1},\Phi ^{2},\Phi ^{3}), p= ( p_{1},p_{2},p_{3} ), \mathring{e} _{1}(p) =\frac{\partial }{\partial x}+p_{2}\frac{\partial }{ \partial z} and \mathring{e}_{2}(p) =\frac{\partial }{\partial y}-p_{1}\frac{\partial }{\partial z}, then
and
Thus,
where
and denote the subscripts as the partial derivatives for all \Phi^i’s. By (?) that \Phi _{z}^{1}=\Phi _{z}^{2}=0, it follows that the functions \Phi ^{1} and \Phi ^{2} both depend only on x and y, and so is \alpha. Moreover, use (?) again and the facts \Phi _{xy}^{1}=\Phi _{yx}^{1} and \Phi _{xy}^{2}=\Phi _{yx}^{2}, we have
which implies that \alpha _{x}=\alpha _{y}=0. Thus \alpha is a constant on H_{1}, say \alpha =\alpha _{0}. From (?) and notice that \Phi (0)=0, we finally get
which implies that \Phi _{x}^{3}=\Phi _{y}^{3}=0. Therefore
and the result follows.
Let \Phi be a pseudo-hermitian transformation on H_{1} such that \Phi (0)=0. Then, for any p\in H_{1}, the matrix representation of \Phi _{\ast }(p) with respect to the standard basis \big(\frac{\partial }{\partial x},\frac{\partial }{\partial y},\frac{\partial }{\partial z}\big) of \mathbb{R}^3 is
for some real constant \alpha _{0} which is independent of p, and hence \Phi _{\ast } is a constant matrix.
Any horizontally regular curve \gamma(t) can be reparametrized by its horizontal arc-length s such that |\gamma'_\xi(s)|=1.
efine s(t)=\int_{0}^{t}|\gamma _{\xi}^{\prime}(u)|{\rm d}u. Then any horizontal arc-length differs s up to a constant. By the fundamental theorem of calculus, we have \frac{{\rm d}s}{{\rm d}t}=|\gamma _{\xi}^{\prime}(t)|. Since
\gamma _{\xi}^{\prime}(s)=\frac{\gamma _{\xi}^{{\prime}}(t)}{|\gamma _{\xi}^{\prime}(t)|}, namely, |\gamma _{\xi }^{\prime}(s)|=1.
A lift of a mapping f\colon M\rightarrow G/H is defined to be a map F\colon M\rightarrow G such that the following diagram commutes:
where G is a Lie group, H is a closed Lie subgroup and G/H is the associated homogeneous space. In additional, another lift \tilde{F} of f has to satisfy
for some map g\colon M\rightarrow H.
In the next section, we shall set G={\rm PSH}(1), M=(a,b)\subset R, f=\gamma, F=\widetilde{\gamma}, G/H={\rm PSH}(1)/{\rm SO}(2), and identify {\rm PSH}(1)/{\rm SO}(2) with H_1.
et \gamma(t)=(x(t),y(t),z(t)) be a horizontally regular curve. The horizontal arc-length s is defined by
We first observe that there is the natural decomposition
where we abuse the notation by \frac{\partial}{\partial z}=T. Let \bar{\gamma}(s) be the reparametrization of \gamma(t) by the horizontal arc-length s. Since \gamma^{\prime}(t)=\bar{\gamma}^{\prime}(s)\frac{{\rm d}s}{{\rm d}t}, by comparing with the decomposition (L), one has
For the p-curvature, by (L), note that X(s)=\frac{{\rm d}t}{{\rm d}s}(x^{\prime}(t)\mathring{e}_{1}+y^{\prime}(t)\mathring{e}_{2}), and Y(s)=JX(s)=\frac{{\rm d}t}{{\rm d}s}(x^{\prime}(t)\mathring{e}_{2}-y^{\prime}(t)\mathring{e}_{1}). A straight-forward computation shows
so
Again by (L), the contact normality must be
and the result follows.
ecall [CCG] that the Hamiltonian system on H_{1} for the geodesics is
where
So the Hamiltonian system (R) can be expressed by
Since \dot{\xi}_{3}(t) =0, we have \xi _{3}(t) =c_{3} for some constant c_{3}. When c_{3}=0, one has x(t) =( c_{1}t+d_{1},c_{2}t+d_{2},(c_{1}d_{2}-c_{2}d_{1}) t+d_{3}), and this implies that k(t) =0 and \tau (t) =0; when c_{3}> 0, one has
where
Hence k(t) =-\frac{1}{[ ( a_{1}^{2}+a_{2}^{2}) ] ^{\frac{1}{2}}}<0 and \tau (t) =0; finally, when c_{3}< 0, one has
where
Hence k(t) =\frac{1}{[ ( a_{1}^{2}+a_{2}^{2}) ] ^{\frac{1}{2}}}>0 and \tau (t) =0.
The calculations above show that a horizontal curve is congruent to a geodesic if it has positive constant p-curvature. Conversely, it is easy to prove that any geodesic acted by a symmetry is still a geodesic. Therefore we complete the proof of Theorem 1.6.
The normal coordinates are determined up to a transformation of the form
for some smooth functions g(v), h(v) with \frac{\partial h}{\partial v}\neq 0.
uppose that (\widetilde{u},\widetilde{v}) is any normal coordinates around p, i.e.,
where \widetilde{X}=\pm X. We have the formula for the change of the coordinates
Expand F_{\widetilde{v}}=\widetilde{a}\widetilde{X}+\widetilde{b}\widetilde{ Y}+\widetilde{c}\widetilde{T} by the orthonormal basis \{\widetilde{X},\widetilde{Y},\widetilde{T} \}. The first identity of (S) implies
Since p is a non-singular point, we see that \widetilde{c}\neq 0 around p, and so
namely, \widetilde{v}=h(v) for some function h(v). In addition, comparing the coefficient of X in (S), (S), we have
and hence \widetilde{u}=\pm u+g(v) for some function g(v). Finally we compute
and the result follows.
Suppose \widetilde{\rm I}, \widetilde{\rm II}, \widetilde{\rm III}, \widetilde{\rm IV} are those defined as (?) with respect to the other normal coordinates (\widetilde{u},\widetilde{v}). Then we have
uppose \widetilde{a}, \widetilde{b}, \widetilde{c}, \widetilde{l}, \widetilde{m} are the coefficients of first and second kinds with respect to the normal coordinates (\widetilde{u}, \widetilde{v}). We point out that all such the coefficients have the same expression as in (?) w.r.t. the new coordinates except for \widetilde{X}=\pm X and \widetilde{Y}=J\widetilde{X}=\pm Y.
By Lemma 5.1, there exists two functions g(v) and h(v) such that
Now we compute the transformation laws of those coefficients:
Similarly, we have
and so F_{u}=\pm F_{\widetilde{u}} and F_{uu}=\pm\big(F_{\widetilde{u}\widetilde{u}}\frac{\partial \widetilde{u}}{\partial u}+F_{\widetilde{u}\widetilde{v}}\frac{\partial \widetilde{v}}{\partial u}\big)=F_{\widetilde{u}\widetilde{u}}. Thus
Similarly
From the transformation laws (?), (?), (?), (?), the result (5.2) follows.
In the proof (?), denote
then we have \alpha=\pm\widetilde{\alpha}. Actually, \alpha is a function defined on the non-singular part of \Sigma, independent of the choice of the normal coordinates up to a sign, such that \alpha e_{2}+T\in T\Sigma, and hence an invariant of \Sigma on the non-singular part. Similarly, from (?), so is for l, which actually is the p-mean curvature.
We point out that the signs appearing for \alpha and l are due to the different choices of the orientations. Indeed, if one chooses the normal coordinates with respect to a fixed orientation of the characteristic foliation, then we have \alpha =\widetilde{\alpha } and l=\widetilde{l}.
Let g_{\Theta} be the adapted metric on H_1. Then we have
defined on the non-singular part of \Sigma.
his lemma is a direct consequence of (?), (?), and g_{\Theta}|_{\Sigma}={\rm d}p\otimes {\rm d}p.
We have
y \omega ^{2}=\alpha \omega ^{3} and the second identity of (?), we have
where we have used the second formula of the structure equation (?) at the third equality above. On the other hand, from the Maurer–Cartan structure equation (?)
Combine two identities above and use the Cartan lemma, we see that there exists a function D such that
Similarly,
Again, by Cartan lemma, there exists a function A such that
Finally, use (?) again
where we have used the third formula of (?) and \hat{\omega} ^{2}\wedge \omega ^{3}=0. Therefore, there exists a function B such that
Similarly, by (B), (B), (B), we obtain
These complete the proof.
roposition 5.6 implies that
Therefore the integrability condition (?) is equivalent to {\rm d}\omega_{1}{}^{2}=0.
irst we show existence. Define a {\mathfrak{psh}}(1)-valued 1-form \phi on the non-singular part of \Sigma by
where
It is easy to check that \phi satisfies {\rm d}\phi+\phi \wedge \phi=0 if and only if the integrability condition (?) holds. Therefore, by Theorem 2.2, for any point p\in \Sigma, there exists an open set U containing p and an embedding f\colon U\rightarrow H_1 such that g=f^{*}(g_{\Theta}), \alpha^{\prime}=f^{*}\alpha and l^{\prime}=f^{*}l.
For uniqueness, by Proposition 5.6, the Darboux derivative is completely determined by the induced metric g_{\Theta}|_{\Sigma}, the p-variation \alpha and the p-mean curvature l. Therefore, by Theorem 2.1, the embedding into H_1 is unique up to a Heisenberg rigid motion.
An oriented horizontal line \ell in H_1 is an oriented line such that any point p\in \ell the tangent vector of the line at p lies on the contact plane \xi_p. For convenience we sometimes call a horizontal line or a line. Denote \mathcal{L} by the set of all oriented horizontal lines in H_1.
Any horizontal line \ell \in \mathcal{L} can be parametrized by a triple (p,\theta, t)\in \mathbb{R} \times S^1\times\mathbb{R}, and also be parametrized by a base point B=(p\cos\theta, p\sin\theta, t) with a horizontally unit-speed vector U=(\sin\theta, -\cos\theta, p), namely,
onsider the projection \pi(\ell) of the line l\in\mathcal{L} onto the xy-plane. Since \pi(\ell) can be uniquely determined by the pair (p,\theta), where p\in\mathbb{R} is the oriented distance from the origin to the line \pi(l) (see [Ch] or the remark below) and \theta\in [0,2\pi) is the angle from the positive x-axis to the normal (Fig. ?), the points (x,y)\in\pi(\ell) satisfy the equation
On the projection \pi(\ell), denote the foot point
and the unit tangent vector along the projection
where |u|_{\mathbb{R}^2} is the Euclidean length of u on the xy-plane; on the line \ell\in H_1, denote the lifting of the foot point b, called the base point by
Denote the tangent vector of \ell at point B by T(B). Since \ell is horizontal, which implies that T(B)\in\xi_B:=\operatorname{span}\{\mathring{e}_1(B), \mathring{e}_2(B)\}, and we have
for some \alpha, \beta\in \mathbb{R}. Notice that the projection \pi(T(B)) is exactly the unit tangent vector u along the projection \pi(\ell). Hence by comparing the first two components of (?) with (?) we have
Therefore by defining the horizontal vector
we have |U|_{\xi(B)}=1, the horizontally unit-speed, and conclude that the line \ell can be uniquely determined by the triple (p,\theta, t), i.e., the base point B, and be parametrized by B+sU for any s\in \mathbb{R} as shown in (8.2).
We point out that the lines we consider in H_1 are all oriented lines. Indeed, by convention in \mathbb{R}^2 there exists a bijection between the set of oriented lines and \mathbb{R}^2\times S^1, and the orientation of \ell\in H_1 follows that of \pi(\ell)\in \mathbb{R}^2. If we consider the non-oriented lines in \mathbb{R}^2 (and hence in H_1), then the coefficient on the right-hand side of (1.11) should be changed to 2.
Let \mathbb{X}(u,v)=(x(u,v),y(u,v),z(u,v))\in \Sigma be the parametrized surface in H_1. Then the configuration space D which describes the horizonal oriented lines intersecting \Sigma should be
where
uppose the line \ell(s) parametrized by (8.2) intersects the surface \Sigma at the point q.
At the point q, by Proposition 8.2, we have
for some s\in \mathbb{R}. By (?), (?), one has
which is compatible with \eqref{projformula} and we obtain the first constraint (8.4). Finally, use (?), (?) again to solve for the parameter s, and substitute s into \eqref{z(uv)}. It is easy to have the second constraint (8.4)
By a simple calculation and (8.4), we observe that
i.e., the horizontally unit-speed vector field U along the line have the same vector-value wherever being evaluated at the based point B or at the intersection q=\mathbb{X}(u,v).
Let \{\mathbb{X}(u,v);e_1(\theta),e_2(\theta),T \} be a frame defined by (?) and the corresponding coframe \{\mathbb{X}(u,v); \omega^1, \omega^2, \Theta\} with the connection 1-form {\omega_1}^2. We have
One concludes that
where \pi is the projection from D to \mathcal{L}, and \langle \,,\,\rangle is the Levi-metric.
n the surface since \mathbb{X}=(x,y,z)=x(1,0,y)+y(0,1,-x)+(0,0,z)=x\mathring{e}_1+y\mathring{e}_2+ z T, we have
Thus, by the moving frame formula (?) and the first constraint (8.4)
By the second constraint (8.4) and the parametrization of the line (8.2)
and the result follows.
Let E=\alpha\mathbb{X}_u + \beta\mathbb{X}_v be the tangent vector field defined on the surface \sum=\mathbb{X}(u,v). Then the vector E is on the contact bundle \xi (and hence in TH_1\cap\xi) if and only if pointwisely the coefficients \alpha and \beta satisfy
equivalently,
irst, we assume that E=\alpha\mathbb{X}_u + \beta\mathbb{X}_v=\alpha(x_u, y_u,z_u)+\beta(x_v, y_v,z_v)=c\mathring{e}_1+d\mathring{e}_2=(c,d,cy-dx) for some constants c and d. Compare each component of E to have
Substitute the last equation by the first two, we get the necessary condition \alpha z_u+\beta z_v= (\alpha x_u+\beta x_v)y - (\alpha y_u+\beta y_v)x.
The reverse part can be obtained by the direct computation
We have used the condition (8.7) in the last equality.
Suppose we choose the frames \{\mathbb{X}(u,v);e_1(\theta),e_2(\theta),T \} on D and the coframe with the connection 1-form defined by (?). We have the identity
where the singular foliation
defines the characteristic foliation of \Sigma, which is induced from the contact plane \xi.
y Proposition 8.6 and the moving frame formula (?)
To prove the vector E\in TM\cap\xi, it suffices to show that the coefficients \alpha:=(z_u+xy_u-yx_u) and \beta:= -(z_v+xy_v-yx_v) satisfy the condition (8.7), and we complete the proof by the previous Lemma 8.7.
In classical integral geometry [Ch, San], the quantity {\rm d}L:= {\rm d}p\wedge {\rm d}\theta\wedge {\rm d}t is called the (kinematic) density of the line (p,\theta,t)\in\mathbb{R}^3, which is always chosen to be positive depending the orientation. Hence, according to (8.6) and (8.8), in the following proof we have to consider the orientation of \{(u,v,\theta)\} to ensure the positivity of the quantity \langle E,e_2\rangle.
y Remark 8.9, we choose {\rm d}u\wedge {\rm d}v\wedge {\rm d}\theta as the orientation of D. Let D=D^+\cup D^-, where
By the structure equation (?),
We also have
Now we integrate the kinematic density {\rm d}L over the set \mathcal{L}. By using (8.6), (B), the Stock’s theorem, and (B), we have
where \Gamma^{\pm}:=\partial D^{\pm}\setminus\Gamma. We also point out that the number, 2, occurs in the first identity is due to the orientations for each horizontal line.
Next, we show that {\rm d}u\wedge {\rm d}v=0 on \Gamma^{\pm}. Indeed, by using the coordinates \{(u,v,\theta)\} for the configuration space D, any vector field defined on \Gamma^+ can be represented by A\wedge\frac{\partial}{\partial \theta}\in \partial \Sigma\times S^1 for some vector A defined on the tangent bundle T\partial\Sigma. The value {\rm d}u\wedge {\rm d}v evaluated on \Gamma^+ must be
Therefore, (B) becomes
we have used (B) and E is parallel to e_1 on \Gamma at the third equality.
The text on this page is from arXiv:1509.00950v3 by Hung-Lin Chiu, Yen-Chang Huang and Sin-Hua Lai, under the licence CC BY SA 4.0. The corpus made these changes: it read the items out of the LaTeX source, it expanded the macros of the paper, it put each proof after its statement, and it converted the text to HTML. The formulas are the TeX of the paper, and KaTeX renders them. A formula or a reference can look different from the PDF. The page is published under CC BY SA 4.0.
The record of each item is in analysis/objects/1509/1509.00950.jsonl.