On The Classification of Sub-Riemannian Structures

Download as pdf or txt
Download as pdf or txt
You are on page 1of 10

Communications in Mathematics 32 (2024), no.

1, 13–22 13
DOI: https://doi.org/10.46298/cm.10550
©2024 Rory Biggs and Odirile Ntshudisane
This is an open access article licensed under the CC BY-SA 4.0

On the classification of sub-Riemannian structures on a 5D


two-step nilpotent Lie group
Rory Biggs and Odirile Ntshudisane

Abstract. We classify the left-invariant sub-Riemannian structures on the unique


five-dimensional simply connected two-step nilpotent Lie group with two-dimensional
commutator subgroup; this 5D group is the first two-step nilpotent Lie group beyond
the three- and five-dimensional Heisenberg groups. Alongside, we also present a
classification, up to automorphism, of the subspaces of the associated Lie algebra
(together with a complete set of invariants).

1 Introduction
Invariant sub-Riemannian structures on Lie groups have proved to be a well-suited
differential geometric language for the study of several physical systems as well as being a
rich source of examples and counterexamples for a number of fundamental questions and
conjectures in sub-Riemannian geometry (see, e.g., [3], [12], [17], [20]). Much work has
been done in studying specific structures, their geodesics, and trying to classify various
families of structures, for instance studying the class of structures in three dimensions (see
e.g., [2], [8], [10], [11], [18], [19], [21]), in four dimensions (see e.g., [1], [4], [5], [6], [7]),
or for some sufficiently regular and thus amenable families of structures like those on the
(2n + 1)-dimensional Heisenberg groups (see, e.g., [9] and the references therein).
In this paper we consider the left-invariant sub-Riemannian structures on a five-dimen-
sional two-step nilpotent Lie group with two-dimensional commutator subgroup, which we
denote by T. This group is the first (lowest-dimensional) two-step nilpotent Lie group
MSC 2020: 17B40, 53C17, 22E25.
Keywords: Two-step nilpotent Lie group, sub-Riemannian structure, isometries.
Affiliation:
Rory Biggs – Department of Mathematics and Applied Mathematics, University of Pretoria,
0002 Pretoria, South Africa
E-mail: [email protected]
Odirile K.L. Ntshudisane – Department of Mathematics and Applied Mathematics, University
of Pretoria, 0002 Pretoria, South Africa
E-mail: [email protected]
14 Rory Biggs and Odirile Ntshudisane

beyond the three- and five-dimensional Heisenberg groups. We note that although the
four-dimensional Engel group (the simply connected Lie group with Lie algebra having
nonzero-commutator relations [E2 , E4 ] = E1 , [E3 , E4 ] = E2 ) has a smaller dimension than
T, the fact that it is a three-step nilpotent Lie group makes the sub-Riemannian structures
on the Engel group arguably more complicated (cf. [1], [5], [6], [7]).
In Section 2, we give a matrix representation for T, determine the group of automor-
phisms of its Lie algebra t, and classify the subspaces of t up to automorphism. In Section 3
we then proceed to classify the sub-Riemannian structures on T up to isometry (by making
use of the fact that all isometries are affine in this context [16]) and briefly describe the
isotropy subgroups of identity.

2 The Lie group T


There is only one five-dimensional two-step nilpotent simply connected (real) Lie group
with two-dimensional commutator subgroup (see, e.g., [22]). We denote this group T and
its Lie algebra t. The Lie group T has the following matrix representation (cf. [13])
  

 1 x 1 x 4 x 5 

0 1 x x
 
2 3

T=   : x1 , x2 , x3 , x4 , x5 ∈ R
 0 0 1 0  
 
0 0 0 1
 
  
 0 v1 v4 v5 
 5 
0 0 v v
  X 
2 3
t=    = vi Ei : v1 , . . . , v5 ∈ R .
 0 0 0 0 
 i=1 
0 0 0 0
 

The non-zero Lie brackets of t are given by

[E1 , E2 ] = E4 , [E1 , E3 ] = E5 .

The centre z = hE4 , E5 i of t coincides with its commutator subalgebra.

Lemma 2.1. The group of automorphisms of t is given by


  

 a 1 0 0 0 0 

a b c 0 0
 
 2 1 1
  



5×5
Aut(t) = a3 b2 c2 0
 0 ∈R  6 0, b1 c2 − c1 b2 6= 0
: a1 =


 a4 b3 c3 a1 b1 a1 c1  


 
a5 b 4 c 4 a1 b 2 a1 c 2
 

with respect to the ordered basis (E1 , E2 , E3 , E4 , E5 ).


Proof. Suppose ϕ ∈ Aut(t). That is, ϕ : t −→ t is a linear isomorphism that preserves
Lie brackets. Let [ϕij ] be the matrix representation of ϕ relative to the ordered basis
(E1 , E2 , E3 , E4 , E5 ).
On the classification of sub-Riemannian structures on a 5D two-step nilpotent Lie group 15

As ϕ·z = z, we have that ϕ14 = ϕ24 = ϕ34 = 0 and ϕ15 = ϕ25 = ϕ35 = 0. As ϕ preserves
the Lie bracket [E1 , E2 ] = E4 , we get ϕ44 = (ϕ11 ϕ22 − ϕ21 ϕ12 ) and ϕ54 = (ϕ11 ϕ32 − ϕ31 ϕ12 ).
Similarly, as ϕ preserves the Lie bracket [E1 , E3 ] = E5 , we have that ϕ45 = (ϕ11 ϕ23 −ϕ21 ϕ13 )
and ϕ55 = (ϕ11 ϕ33 − ϕ31 ϕ13 ). We thus have that
 
ϕ11 ϕ12 ϕ13 0 0
ϕ21 ϕ22 ϕ23 0 0 
 
[ϕij ] = ϕ31 ϕ32 ϕ33
 0 0 .

ϕ41 ϕ42 ϕ43 (ϕ11 ϕ22 − ϕ21 ϕ12 ) (ϕ11 ϕ23 − ϕ21 ϕ13 )
ϕ51 ϕ52 ϕ53 (ϕ11 ϕ32 − ϕ31 ϕ12 ) (ϕ11 ϕ33 − ϕ31 ϕ13 )

Preservation of the Lie bracket [E2 , E3 ] = 0 gives the conditions ϕ12 ϕ23 − ϕ22 ϕ13 = 0 and
ϕ12 ϕ33 − ϕ32 ϕ13 = 0. If ϕ12 6= 0, then ϕ23 = ϕ22
ϕ12
ϕ13
, ϕ33 = ϕ32
ϕ12
ϕ13
and so det ϕ = 0, which is
a contradiction. Thus ϕ12 = 0. Similarly, assuming that ϕ13 6= 0 leads to a contradiction
and thus ϕ13 = 0. Therefore,
 
ϕ11 0 0 0 0
ϕ21 ϕ22 ϕ23 0 0 
 
[ϕij ] = ϕ31 ϕ32 ϕ33
 0 0  
ϕ41 ϕ42 ϕ43 ϕ11 ϕ22 ϕ11 ϕ23 
ϕ51 ϕ52 ϕ53 ϕ11 ϕ32 ϕ11 ϕ33

with ϕ11 6= 0 and ϕ22 ϕ33 − ϕ23 ϕ32 6= 0. It is a simple matter to show that any such map
ϕ is an automorphism.

Subspace classification
Let s and w be two subspaces of a Lie algebra g. We say that s and w are equivalent
if there exists an automorphism ϕ ∈ Aut(g) such that ϕ · s = w. The subspace s is called
bracket generating if the smallest subalgebra of g containing s is g itself. If s is an ideal,
then it is said to be a fully characteristic ideal if ϕ · s = s for all ϕ ∈ Aut(g).
We identify some scalar invariants for subspaces of the Lie algebra t. A simple invariant
is the dimension of a subspace: if s is equivalent to w, then dim(s) = dim(w). Two more
invariants can be found by considering the dimension of the intersection of a given subspace
with any fully characteristic ideal. Accordingly, since the centre

z = hE4 , E5 i

and the subspace


c = hE2 , E3 , E4 , E5 i
are both fully characteristic ideals (by Lemma 2.1 in the case of c), we have that

dim(s ∩ z) = dim(w ∩ z) and dim(s ∩ c) = dim(w ∩ c).

whenever s and w are equivalent. The last scalar invariant is slightly more involved.
16 Rory Biggs and Odirile Ntshudisane

Lemma 2.2. If s is equivalent to w, then

dim(s ∩ z ∩ [E1 , s ∩ c]) = dim(w ∩ z ∩ [E1 , w ∩ c]).

Proof. Let ϕ ∈ Aut(t) such that ϕ · s = w. Then there exists ψ ∈ Aut(t) such that
ψ · E1 = a1 E1 , a1 6= 0, ψ|c = ϕ|c , and ψ|z = ϕ|z (see Lemma 2.1). Therefore

w ∩ z ∩ [E1 , w ∩ c] = (ϕ · s) ∩ z ∩ [E1 , (ϕ · s) ∩ c]
= ϕ · (s ∩ z) ∩ [E1 , ϕ · (s ∩ c)]
h i
1
= ψ · (s ∩ z) ∩ a1 ψ · E1 , ψ · (s ∩ c)
= ψ · (s ∩ z ∩ [E1 , s ∩ c])

and so dim(w ∩ z ∩ [E1 , w ∩ c]) = dim(s ∩ z ∩ [E1 , s ∩ c]).


With these invariants at hand, we now proceed to classify the subspaces of t. In Table 1
we list the equivalence class representatives identified alongside their associated values for
the scalar invariants.

Subspace s dim(s) dim(s ∩ c) dim(s ∩ z) dim(s ∩ z ∩ [E1 , s ∩ c])


hE1 i 0 0 0
hE2 i 1 1 0 0
hE4 i 1 1 0
hE1 , E2 i 1 0 0
hE1 , E4 i 1 1 0
hE2 , E3 i 2 0 0
2
hE2 , E5 i 2 1 0
hE2 , E4 i 2 1 1
hE4 , E5 i 2 2 0
hE1 , E2 , E3 i 2 0 0
hE1 , E2 , E5 i 2 1 0
hE1 , E2 , E4 i 2 1 1
3
hE1 , E4 , E5 i 2 2 0
hE2 , E3 , E4 i 3 1 1
hE2 , E4 , E5 i 3 2 1
hE1 , E2 , E3 , E4 i 3 1 1
hE1 , E2 , E4 , E5 i 4 3 2 1
hE2 , E3 , E4 , E5 i 4 2 2
hE1 , E2 , E3 , E4 , E5 i 5 4 2 2

Table 1: Subspace equivalence class representatives for t with values for scalar invariants
On the classification of sub-Riemannian structures on a 5D two-step nilpotent Lie group 17

Theorem 2.3. Any proper subspace of the Lie algebra t is equivalent to exactly one of the
following

SA: hE1 i, hE2 i, hE2 , E3 i, hE1 , E4 i, hE2 , E5 i, hE2 , E3 , E4 i, hE1 , E2 , E4 i


I: hE4 i, hE2 , E4 i, hE1 , E4 , E5 i, hE2 , E4 , E5 i, hE1 , E2 , E4 , E5 i
FCI: hE4 , E5 i, hE2 , E3 , E4 , E5 i
Gen: hE1 , E2 , E3 i, hE1 , E2 , E3 , E4 i
S: hE1 , E2 i, hE1 , E2 , E5 i.

Here, the subspaces are listed according to their class: subalgebras (SA), ideals (I), fully
characteristics ideals (FCI), bracket generating subspaces (Gen), or subspaces (S) with
none of these properties.
Proof. We treat a typical case for determining a class representative. Suppose s is a
subspace of the Lie algebra t with
dim(s) = 3. (1)
Further, suppose
dim(s ∩ z) = 1. (2)
Let X ∈ s ∩ z, X = x4 E4 + x5 E5 . Then
 
1 0 0 0 0
0 x4 −x5 0 0 
 
ϕ = 0 x5 x4 0
 0 
0 0 0 x4 −x5 
0 0 0 x5 x 4

is an automorphism such that ϕ · E4 = X. Thus s is equivalent to a subspace s̄ containing


E4 .
Now, by a simple dimensionality argument, 2 ≤ dim(s̄ ∩ c) ≤ 3. Let us suppose

dim(s̄ ∩ c) = 2. (3)

Since E4 ∈ s̄ ∩ c, there exists V̄ , W̄ ∈ s̄ such that hW̄P , E4 i = s̄ ∩ c, hV̄ ,P


W̄ , E4 i = s̄ and
V̄ ∈
/ c. This implies that v̄1 6= 0 and w̄1 = 0; here V̄ = v̄i Ei and W̄ = w̄i Ei .
Finally, suppose that
dim(s̄ ∩ z ∩ [E1 , s̄ ∩ c]) = 1. (4)
Then

1 = dim(hV̄ , W̄ , E4 i ∩ z ∩ [E1 , hV̄ , W̄ , E4 i ∩ hE2 , E3 , E4 , E5 i])


= dim(hE4 i ∩ [E1 , hW̄ , E4 i])
= dim(hE4 i ∩ [E1 , hW̄ i])
= dim(hE4 i ∩ hw̄2 E4 + w̄3 E5 i)
18 Rory Biggs and Odirile Ntshudisane

and so it follows that w̄2 6= 0 and w̄3 = 0. Therefore


 
v̄1 0 0 0 0
v̄2 w̄2 0 0 0
ϕ0 = 
 
 v̄3 0 1 0 0 

v̄4 w̄4 0 v̄1 w̄2 0 
v̄5 w̄5 0 0 v̄1

is an automorphism such that ϕ0 · hE1 , E2 , E4 i = hV̄ , W̄ , E4 i = s̄. Thus s̄ (and therefore s)


is equivalent to hE1 , E2 , E4 i.
By considering all other possible values of the invariants in (1), (2), (3), and (4) one
obtains all possible class representatives. As all representatives obtained are differentiated
by the set of scalar invariants (see Table 1), they are mutually non-equivalent. Standard
computations determine whether each class representative is a subalgebra, ideal, fully
characteristic or generating subspace.
Since the four scalar invariants identified evaluate distinctly for each equivalence class
(see Table 1), these invariants form a complete set.

Corollary 2.4. Two subspaces s and w of t are equivalent if and only if

dim(s) = dim(w),
dim(s ∩ z) = dim(w ∩ z),
dim(s ∩ c) = dim(w ∩ c), and
dim(s ∩ z ∩ [E1 , s ∩ c]) = dim(w ∩ z ∩ [E1 , w ∩ c]).

Here z = hE4 , E5 i and c = hE2 , E3 , E4 , E5 i.

3 Sub-Riemannian structures on T
A left-invariant sub-Riemannian structure is a triple (G, D, g) where G is a real, finite-
dimensional, connected Lie group, D is a smooth bracket generating left-invariant distri-
bution on G, and g is a left-invariant Riemannian metric on D. Equivalently: D(1) is a
bracket generating linear subspace of the Lie algebra g of G with D(x) = d1 Lx · D(1) for
every x ∈ G, where Lx : G → G, y 7→ xy; g1 is a positive definite, symmetric bilinear form
on D(1) with gx (d1 Lx · A, d1 Lx · B) = g1 (A, B) for every A, B ∈ D(1).
Let (G, D, g) and (G0 , D0 , g0 ) be two left-invariant sub-Riemannian structures. An isom-
etry between (G, D, g) and (G0 , D0 , g0 ) is a diffeomorphism φ : G → G0 such that φ∗ D = D0
and g = φ∗ g0 ; that is,

dx φ · D(x) = D0 (φ(x)) and gx (X, Y ) = gφ(x)


0
(dx φ · X, dx φ · Y ),

for all x ∈ G and X, Y ∈ D(x). By definition, left translations Lx are isometries. Isometries
preserve the Carnot–Carathéodry distance associated to the sub-Riemannian structure.
On the classification of sub-Riemannian structures on a 5D two-step nilpotent Lie group 19

It turns out that for left-invariant sub-Riemannian structures on simply connected


nilpotent Lie groups, every isometry is the composition of a left-translation and a Lie
group isomorphism [16]. (Indeed in [16] this is proved more generally for nilpotent metric
Lie groups.) Therefore, since all left translations are isometries, if two such structures
are isometric then there exists a Lie group isomorphism between them that realizes the
isometry. We note that there is a one-to-one correspondence between the Lie group au-
tomorphisms on a simply connected Lie group and the Lie algebra automorphisms on its
Lie algebra (see, e.g., [14]). Consequently, we have the following simple algebraic charac-
terization for two sub-Riemannian structures on a simply connected nilpotent Lie group G
with Lie algebra g to be isometric.

Proposition 3.1. (cf. [7], [9]) Two left-invariant sub-Riemannian structures (G, D, g) and
(G, D0 , g0 ) on a simply connected nilpotent Lie group G are isometric if and only if there
exists an automorphism ψ ∈ Aut(g) such that

ψ · D(1) = D0 (1) and g1 = ψ ∗ g10 .

Here (ψ ∗ g10 )(A, B) = g10 (ψ · A, ψ · B) for A, B ∈ D(1).


Accordingly, the distribution D of any left-invariant sub-Riemannian structure (T, D, g)
on T, is isometric to a structure with distribution at identity being one of the bracket
generating subspaces listed in Theorem 2.3. All that remains to be done is to normalize
the metrics g by Lie algebra automorphisms using Proposition 3.1. Doing this we arrive
at the following classification of left-invariant sub-Riemannian structures on T.

Theorem 3.2. Any left-invariant sub-Riemannian structure (T, D, g) is isometric to ex-


actly one of the following:
(
H3 (1) = hE1 , E2 , E3 i
(T, H3 , h3 ) :
h31 = diag(1, 1, 1)
(
H4 (1) = hE1 , E2 , E3 , E4 i
(T, H4 , h4,α ) :
h4,α
1 = α · diag(1, 1, 1, 1), α > 0
(
H5 (1) = hE1 , E2 , E3 , E4 , E5 i
(T, H5 , h5,(α,β) ) : 5,(α,β)
h1 = diag(1, 1, 1, α, β), α ≥ β > 0.

Here the metrics are written with respect to the bases given for their respective distributions.

Remark 3.3. (T, H5 , h5,(α,β) ) corresponds to the result in [15, Proposition 6] for the clas-
sification of invariant Riemannian structures on T.
Proof. We treat the rank 4 structures (i.e., those with dim D(g) = 4, g ∈ T) as a typical
case. Let (T, D, g) be a rank 4 left-invariant sub-Riemannian structure. By Theorem 2.3
there exists ψ0 ∈ Aut(t) such that ψ0 · D(1) = H4 (1). By Proposition 3.1, (T, D, g) is
isometric to (T, H4 , g1 ) for some metric g1 on H4 .
20 Rory Biggs and Odirile Ntshudisane

We can write g11 as a positive definite symmetric matrix with respect to the basis
(E1 , E2 , E3 , E4 ) for H4 (1):  
h1 a1 a2 a3
a1 h2 a4 a5 
g11 = 
a2 a4 h3 a6  .

a3 a5 a6 h4
Now  
1 0 0 0 0
 0 1 0 0 0
 
ψ1 =  0
 a 0 1 0 0
− 3 − a5 − a6

h4 h4 h4
1 0
0 0 0 0 1
is an automorphism of t such that ψ1−1 · H4 (1) = H4 (1) and g12 = (ψ1 )∗ g11 has matrix
 0
h1 a01 a02 0

a01 h02 a04 0 
g12 = 
a02 a04 h03 0 

0 0 0 h04

with respect to (E1 , E2 , E3 , E4 ) for some constants a01 , a02 , a04 , h01 , . . . h04 ∈ R. Note here
that g11 = (ψ1−1 )∗ g12 , or equivalently g12 (A, B) = (ψ1 )∗ g11 (A, B) = g11 (ψ1 · A, ψ1 · B) for
A, B ∈ H4 (1). That is to say, (T, H4 , g1 ) is isometric to (T, H4 , g2 ) by Proposition 3.1.
Continuing on in this way, we have
 
1 0 0 0 0
 a01 h203 −a02 a04 1 − a040 0 0 
 
0
 a4 −h2 h3 0 0 h2  b 1 0 0 0
 a0 h0 −a0 a0  3 ∗ 2
 0 b2 0 0 
ψ2 =  a0 2 −h0 h0 0 1 0 0 
 2 2 1 4
∈ Aut(t), g 1 = ψ g
2 1 = 
 0 0 b3 0 

4 2 3

 a0 
 0 0 0 1 − h40  0 0 0 b4
2
0 0 0 0 1

for some b1 , . . . , b4 ∈ R. Note that a04 2 − h02 h03 6= 0 and h02 6= 0 since g12 is positive definite.
Finally, s s !
r r r
b2 b1 b1 b2 b 1 b2 b1 b22
ψ4 = diag , , , ,
b4 b4 b3 b4 b24 b3 b24

is an automorphism such that g14 = ψ3∗ g13 = b1b4b2 I4 = h4,α 1 with α = b1b4b2 . It therefore
follows by transitivity that (T, D, g) is isometric to (T, H4 , h4,α ) for some α > 0.
Now suppose (T, H4 , h4,α ) and (T, H4 , h4,β ) are isometric for some α, β > 0. By Propo-
sition 3.1 there exists ψ ∈ Aut(t) such that ψ · H4 (1) = H4 (1) and h4,α = ψ ∗ h4,β . Utilizing
Lemma 2.1 and computing these conditions in coordinates, it is fairly straightforward to
show that this implies that α = β. Hence, each different α > 0 yields a non-isometric
structure.
On the classification of sub-Riemannian structures on a 5D two-step nilpotent Lie group 21

Since isometries preserving the identity element are automorphisms of the group, it is
not difficult to find the (linearized) isotropy subgroup of identity (i.e., the subgroup of the
isometry group fixing the identity).

Corollary 3.4. The isotropy subgroups of identity associated to the respective left-invariant
sub-Riemannian structures on T are given by
(i) Iso1 (T, H3 , h3 ) ∼
= Z2 × O(2),
(ii) Iso1 (T, H4 , h4,α ) ∼
= Z2 × Z2 × Z2 ,
(iii) Iso1 (T, H5 , h5,(α,β) ) ∼
= Z2 × Z2 × Z2 , for α > β > 0,
5,(α,α) ∼
Iso1 (T, H5 , h ) = Z2 × O(2) where α > 0.

Remark 3.5. The isotropy groups of (T, H5 , h5,(α,β) ), α ≥ β > 0 correspond to the result
in [15, Proposition 7] for invariant Riemannian structures on T.

Acknowledgements
The research leading to these results has received funding from the National Research
Foundation (NRF) of South Africa.

References
[1] Adams M.R., Tie J.: On sub-Riemannian geodesics on the Engel groups: Hamilton’s equations.
Math. Nachr. 286 (14-15) (2013) 1381–1406.
[2] Agrachev A., Barilari D.: Sub-Riemannian structures on 3D Lie groups. J. Dyn. Control Syst. 18
(1) (2012) 21–44.
[3] Agrachev A., Barilari D., Boscain U.: A comprehensive introduction to sub-Riemannian
geometry. Cambridge University Press (2020).
[4] Almeida D.M.: Sub-Riemannian homogeneous spaces of Engel type. J. Dyn. Control Syst. 20 (2)
(2014) 149–166.
[5] Ardentov A.A., Sachkov Y.L.: Extremal trajectories in the nilpotent sub-Riemannian problem on
the Engel group. Mat. Sb. 202 (11) (2011) 31–54.
[6] Ardentov A.A., Sachkov Y.L.: Conjugate points in nilpotent sub-Riemannian problem on the
Engel group. J. Math. Sci. (N.Y.) 195 (3) (2013) 369–390.
[7] Bartlett C.E., Biggs R., Remsing C.C.: Control systems on nilpotent Lie groups of dimension ≤ 4:
equivalence and classification. Differential Geom. Appl 54 (2017) 282–297.
[8] Biggs R.: Isometries of Riemannian and sub-Riemannian structures on three-dimensional Lie
groups. Commun. Math. 25 (2) (2017) 99–135.
[9] Biggs R., Nagy P.T.: On sub-Riemannian and Riemannian structures on the Heisenberg groups.
J. Dyn. Control Syst. 22 (3) (2016) 563–594.
[10] Boscain U., Rossi F.: Invariant Carnot-Caratheodory metrics on S 3 , SO(3), SL(2), and lens
spaces. SIAM J. Control Optim. 47 (4) (2008) 1851–1878.
22 Rory Biggs and Odirile Ntshudisane

[11] Butt Y.A., Sachkov Y.L., Bhatti A.I.: Extremal trajectories and Maxwell strata in
sub-Riemannian problem on group of motions of pseudo-Euclidean plane. J. Dyn. Control Syst. 20
(3) (2014) 341–364.
[12] Calin O., Chang D.C.: Sub-Riemannian geometry. Cambridge University Press (2009).
[13] Ghanam R., Thompson G.: Minimal matrix representations of five-dimensional Lie algebras.
Extracta Math. 30 (1) (2019) 95–133.
[14] Hilgert J., Neeb K.H.: Structure and geometry of Lie groups. Springer (2012).
[15] Homolya S., Kowalski O.: Simply connected two-step homogeneous nilmanifolds of dimension 5.
Note Mat. 26 (1) (2006) 69–77.
[16] Kivioja V., Le Donne E.: Isometries of nilpotent metric groups. J. Éc. polytech. Math. 4 (2017)
473–482.
[17] Liu W., Sussman H.J.: Shortest paths for sub-Riemannian metrics on rank-two distributions.
Mem. Amer. Math. Soc. 118 (564) (1995) x+104.
[18] Mazhitova A.D.: Sub-Riemannian geodesics on the three-dimensional solvable non-nilpotent Lie
group SOLV− . J. Dyn. Control Syst. 18 (3) (2012) 309–322.
[19] Moiseev I., Sachkov Y.L.: Maxwell strata in sub-Riemannian problem on the group of motions of
a plane. ESAIM Control Optim. Calc. Var. 16 (2) (2010) 380–399.
[20] Montgomery R.: A tour of subriemannian geometries, their geodesics and applications. American
Mathematical Society (2002).
[21] Sachkov Y.L.: Conjugate and cut time in the sub-Riemannian problem on the group of motions of
a plane. ESAIM Control Optim. Calc. Var. 16 (4) (2010) 1018–1039.
[22] Šnobl L., Winternitz P.: Classification and identification of Lie algebras. American Mathematical
Society (2014).

Received: December 10, 2020


Accepted for publication: July 03, 2021
Communicated by: Ilka Agricola

You might also like