Acta Polytechnica doi:10.14311/AP.2017.57.0373 Acta Polytechnica 57(6):373–378, 2017 © Czech Technical University in Prague, 2017 available online at http://ojs.cvut.cz/ojs/index.php/ap FUNCTIONAL REALIZATIONS OF LIE ALGEBRAS AS NOETHER POINT SYMMETRIES OF SYSTEMS Rutwig Campoamor-Stursberg Instituto de Matemática Interdisciplinar-UCM, Plaza de Ciencias 3, E-28040 Madrid, Spain correspondence: rutwig@ucm.es Abstract. Functional realizations of Lie algebras are applied to the problem of determining Lie and Noether point symmetries of Lagrangian systems in N dimensions, particularly in the plane. This encompasses both the case of symmetry-preserving perturbations of a given system, as well as the generic analysis on the structure of (regular) Lagrangians in order to admit a symmetry algebra belonging to a specific isomorphy class. Keywords: Lagrangian; Lie point symmetry; Noether symmetry; perturbation. 1. Introduction The study of the precise relation between constants of the motion of a dynamical system and the symme- try properties of the corresponding equations of the motion (alternatively, an associated Lagrangian) goes back to the seminal paper of E. Noether [1], a key result that, combined with the Lie group theoreti- cal approach to differential equations, has become an essential tool in many branches of applied math- ematics and Mechanics (see, e.g., [2, 3] and references therein). Usually, the Noether symmetry analysis is carried out mainly for conservative systems, due to require- ments of the Hamiltonian formalism and the corre- sponding quantization of the systems [4]. However, there is no restriction, at least from the Lie alge- braic point of view, to treat both conservative and dissipative systems simultaneously. This can be done introducing appropriate functional realizations of Lie algebras, the symmetry generators of which depend on the coordinates of the extended configuration space, and considering the constraints imposed by this time- dependence. This extended approach allows to cover physically relevant systems, such as time-modulated oscillators, and has further applications in the con- text of perturbation theory, such as the study of some geometric properties of the orbits of a given system [5]. Within the context of inverse of problems in dynam- ics [5, 6], in this work we consider an inverse approach to dynamics based on functional realizations of Lie al- gebras that are required to satisfy the Noether symme- try condition for a (regular) Lagrangian system. This allows two approaches, either starting from a given system and analyzing perturbations that preserve cer- tain of the symmetries, or determining families of systems invariant with respect to these realizations. For the purpose of illustration, the plane case is con- sidered, although the Ansatz can also be formulated in arbitrary dimensions. 2. Lie and Noether point symmetries of systems Let L (t, q, q̇) be a regular Lagrangian in n-dimensions and let d dt (∂L ∂q̇i ) − ∂L ∂qi = 0, 1 ≤ i ≤ n (1) be the corresponding equations of the motion. By regularity, we can always rewrite the equations of motion in normal form, i.e., q̈i = ωi(t, q̇j,qj) = gij ( ∂L ∂qj − ∂2L ∂t∂q̇j − ∂2L ∂q̇j∂qk q̇k ) , (2) for 1 ≤ i ≤ n, with gij denoting the inverse Hessian matrix of L. As is well known, the system (2) can be reformulated in equivalent form as the first order partial differential equation Af = ( ∂ ∂t + q̇i ∂ ∂qi + ωi ∂ ∂q̇i ) f = 0. (3) We call a vector field X = ξ(t, q) ∂ ∂t + ηj(t, q) ∂∂qj ∈ X(RN+1) a Lie point symmetry of the equations (2) if its first prolongation Ẋ = X + η̇j (t, q, q̇) ∂∂q̇j with η̇j = −dξdt q̇j + dηj dt satisfies the commutator [ Ẋ, A ] = − dξ dt A. (4) Lie point symmetries span a finite-dimensional Lie algebra LPS, called the Lie (point) symmetry algebra of the system, that be seen to be always a subalgebra of the simple algebra sl(n + 2,R), corresponding to the free system q̈ = 0 [7, 8]. Whenever the system arises from a Lagrangian, a constant of the motion of (2) is defined as a function F (t, q, q̇) satisfying the condition dF dt = ∂F ∂t + q̇i ∂F ∂qi + q̈i ∂F ∂q̇i = A(F) = 0. (5) 373 http://dx.doi.org/10.14311/AP.2017.57.0373 http://ojs.cvut.cz/ojs/index.php/ap Rutwig Campoamor-Stursberg Acta Polytechnica In the following, we will mainly consider Lie point and Noether point symmetries and the conservation laws associated to them. In this context, a vector field X = ξ (t, q) ∂ ∂t + ηj (t, q) ∂ ∂qj (6) is called a Noether point symmetry of (2) if it satisfies the constraint Ẋ (L) + A (ξ) L− A (V ) = 0, (7) for some function V (t, q) that does not depend on the velocities, and that we usually refer to as the gauge term of the symmetry generator. Expanding the symmetry condition (7) provides the following partial differential equation ξ (t, q) ∂L ∂t + ηj (t, q) ∂L ∂qj + η̇j (t, q, q̇) ∂L ∂q̇j + dξ dt L (t, q, q̇) − ∂V ∂t − q̇j ∂V ∂qj = 0. The first Noether theorem (see e.g. [1]) states that to any symmetry (6) there corresponds a constant of the motion determined by the rule J = ξ ( q̇k ∂L ∂q̇k −L ) −ηk ∂L ∂q̇k + V (t, q) . (8) It follows in particular that a constant of the motion J is an invariant of Ẋ, i.e., the condition Ẋ(J) = 0 is satisfied. 2.1. Perturbations that preserve symmetry subalgebras Given a system described by a Lagrangian L (t, q, q̇) and having an r-dimensional Lie algebra LNS of Noether point symmetries, fixing a certain subalgebra L0 < LNS of dimension r0 < r we can ask whether a perturbed Lagrangian L̂ = L + εS (t, q, q̇) exists such that the symmetry generators Xj (1 ≤ j ≤ r0) of L0 are still Noether point symmetries of L̂. Clearly, the equations of the motion (1) of L̂ are given by d dt (∂L ∂q̇i ) − ∂L ∂qi + d dt (∂S ∂q̇i ) − ∂S ∂qi = 0, (9) so that the symmetry condition for the subalgebra L0 leads, for each generator Xk (1 ≤ k ≤ r0), to the differential equation Ẋk ( L̂ ) + A (ξk) L̂− A (Vk) = Ẋk (L) + A (ξk) L− A (Vk) + Ẋk (S) + dξk dt S = 0. (10) Now the vector fields Xk are Noether symmetries of the Lagrangian L, and fixing the gauge terms Vk(t, q), the equation (10) simplifies to a PDE involving merely the perturbation term S (t, q, q̇): Ẋk (S) + dξk dt S (t,q,q̇) =ξk (t,q) ∂S ∂t +ηjk (t,q) ∂S ∂qj +η̇jk (t,q,q̇) ∂S ∂q̇j + dξk dt S (t,q,q̇) = 0, (11) where 1 ≤ k ≤ r0. In the case of velocity-independent perturbation terms, a straightforward verification shows that a Noether symmetry is preserved only if ∂ξk ∂qj = 0 holds for all 1 ≤ k ≤ r0 and 1 ≤ j ≤ n. This allows to restrict the perturbation analysis to sub- algebras L0, the generators of which have components of the type ξk = ϕk(t). This agrees with the usually observed pattern of Lie point symmetries of nonlinear second-order systems of differential equations of the type q̈ = ω (t, q) [7, 8]. 3. Functional realizations of sl(2, R) as Noether symmetry algebra As has been already pointed out in many contexts, the non-compact Lie algebra sl(2,R) plays a relevant role within the group-theoretic analysis of differen- tial equations, in particular, concerning the (super)- integrability of plane systems and the linearization analysis [9–11]. This fact suggests to consider this Lie algebra more closely in the context of inverse problems, as done in [12]. For these reasons, in the following we restrict our analysis to sl(2,R). Prior to analyzing a functional realization of sl(2,R), we make some observations on the generic structure of Noether point symmetries. Consider to this extent a Lagrangian adopting the kinetic form T = 1 2 Aij(q)q̇iq̇j, (12) such that A11(q)A22(q) −A12(q)2 6= 0 holds. Lemma 1. Let X = ξ(t, q) ∂ ∂t + ηj(t, q) ∂∂qj be a Noether point symmetry of a Lagrangian (12). Then the components have the generic structure ξ(t, q) = C1t2 + C2t + C3, ηj(t, q) = ηj1(q)t + ηj2(q). (13) Moreover, the gauge term does not explicitly depend of time, i.e., V = V (q). Proof. Developing the symmetry condition (7) and keeping only the independent term, as well as the terms with highest power in q, we obtain that ∂V ∂t = 0 and the conditions ∂ξ ∂q1 A11(q) = 0, ∂ξ ∂q2 A22(q) = 0, ∂ξ ∂q2 A12(q) + 1 2 ∂ξ ∂q1 A22(q) = 0, ∂ξ ∂q1 A12(q) + 1 2 ∂ξ ∂q2 A22.(q) = 0. 374 vol. 57 no. 6/2017 Noether Point Symmetries of Systems As A11(q)A22(q)−A12(q)2 6= 0, a short computation shows that the condition ∂ξ ∂q1 = ∂ξ ∂q2 = 0 must be nec- essarily satisfied. Introducing this into the symmetry condition, the following relations are obtained for the terms linear in q: ∂η2 ∂t A12(q) + ∂η1 ∂t A11(q) − ∂V ∂q1 = 0, ∂η1 ∂t A12(q) + ∂η2 ∂t A22(q) − ∂V ∂q2 = 0. (14) Multiplying the first equation by A12, the second by −A11 and adding them leads to the expression ∂η2 ∂t ( A212 −A11A22 ) + A12 ∂V ∂q1 + A11 ∂V ∂q2 = 0, (15) from which we conclude that η2 is at most linear in t, hence it admits a decomposition η2(t, q) = η21(q) t + η22(q). For η1(t, q) the assertion is obtained similarly. Finally, for the terms quadratic in q in the symmetry condition (7), we have expression of the type − 1 2 Aij(q) dξ dt + tΨ1(q) + Ψ2(q) = 0, (16) showing that ξ(t) is at most quadratic in t, from which the assertion follows. We observe that the generalization of the latter result to n-dimensions is straightforward. Now let f(q) and g(q) be arbitrary non-vanishing functions and consider the Lie algebra generated by the following vector fields: X1 = t2 ∂ ∂t + tf(q) ∂ ∂q1 + tg(q) ∂ ∂q2 , X2 = 1 2 d dt X1 = t ∂ ∂t + 1 2 f(q) ∂ ∂q1 + 1 2 g(q) ∂ ∂q2 , (17) X3 = 1 2 d2 dt2 X1 = ∂ ∂t . It follows at once that [X1,X2] = −X1, [X1,X3] = −2X2 and [X2,X3] = −X3, showing that the Lie algebra is isomorphic to sl(2,R) for any choices of f and g. We observe that the structure of the sl(2,R) Lie algebra generalizes naturally that studied in [8, 12]. In these conditions, it can be asked which is the most general (kinetic) Lagrangian such that it admits this Lie algebra as an algebra of Noether point symmetries. It suffices to impose the invariance with respect to X1 and X3 in order to ensure that the system is sl(2,R)- invariant. The symmetry condition (7) applied to X3 is trivially satisfied for the gauge term V (t, q) = 0. Analyzing now the invariance with respect to X1, and inspecting first the terms linear in q̇, leads to the constraints f(q)A11(q) + g(q)A12(q) − ∂V ∂q1 = 0, g(q)A22(q) + f(q)A12(q) − ∂V ∂q2 = 0, for the gauge term V (q). As f(q)g(q) 6= 0, this allows us to set f2(q)A11(q)−g2(q)A22(q)−f(q) ∂V ∂q1 −g(q) ∂V ∂q2 = 0. In order to completely satisfy the symmetry condition, the functions Aij(q) must be solutions to the following system of PDEs: f(q) ∂A11 ∂q1 + g(q) ∂A11 ∂q2 + 2A11(q) ( ∂f ∂q1 − 1 ) + 2A12(q) ∂g ∂q1 = 0, f(q) ∂A22 ∂q1 + g(q) ∂A22 ∂q2 + 2A22(q) ( ∂g ∂q2 − 1 ) + 2A12(q) ∂f ∂q2 = 0, f(q) ∂A12 ∂q1 + g(q) ∂A12 ∂q2 + A12(q) ( ∂f ∂q1 + ∂g ∂q2 − 2 ) + A11(q) ∂f ∂q2 + A22(q) ∂g ∂q1 = 0. (∗) There are in principle two different ways to analyze these systems related to a Lagrangian of type (12): either we fix the latter and search for functions f(q) and g(q) such that (17) is a Lie algebra of Noether symmetries, or we fix the components of the symmetry generators and try to determine the most general kinetic term (12) invariant under the algebra. The same problem can be formulated allowing a potential term. We shall exhibit examples of the two approaches leading to nontrivial solutions of the equations. 3.1. Separable kinetic Lagrangians Let us illustrate the preceding situation first for the case of the separable Lagrangians T = 1 2 ( qk1 q̇ 2 1 + q l 2q̇ 2 2 ) , where k,l 6= −2 are constants. This case contains in particular that of the free Euclidean Lagrangian, that is well known to allow a sl(2,R)-subalgebra of Noether symmetries [4]. The symmetry condition (∗) thus requires the solving for the unknown functions f(q) and g(q), as well as the gauge term V (q) for the symmetry generator X1. The resulting equations are lg(q) + 2q2 ( ∂g ∂q2 − 1 ) = 0, ql2g(q) − ∂V ∂q2 = 0, kf(q) + 2q1 ( ∂f ∂q1 − 1 ) = 0, ql1f(q) − ∂V ∂q1 = 0. As they are first-order PDEs, they can be solved with standard methods (see e.g. [13]), and the solution can be expressed as f(q) = 2q1 k + 2 + a1q −k/2 1 , g(q) = 2q2 l + 2 + a2q −l/2 2 , V (q) 2 = qk+21 (k + 2)2 + ql+22 (l + 2)2 + a1q (k+2)/2 1 k + 2 + a2q (l+2)/2 2 l + 2 375 Rutwig Campoamor-Stursberg Acta Polytechnica for the component functions and gauge term, respec- tively. Clearly, as the system related to T is lineariz- able, it admits five additional Noether symmetries, all of which possessing a zero term in ∂ ∂t [11]. Looking now for a potential U(t, q) that pre- serves the symmetry, it follows at once from the X3-invariance that ∂U∂t = 0, and thus the perturbed system is conservative. The invariance by X1 implies that U must satisfy the first-order differential equation (see (11)) f(q) ∂U ∂q1 + g(q) ∂U ∂q2 + 2U(q) = 0, (18) with the functions as obtained above. A routine but tedious computation leads to the general solution U(q) = Ψ(u) qk1 ( 2q1 + a2(k + 2)q −k/2 1 )2 , (19) where u = 22l 2/(2l+4)2q (1+l)/2 2 + a1(l + 2) 4q(k+1)/21 + 2a2(k + 2) . (20) We skip the proof that the perturbed system T + εU possesses exactly a symmetry algebra of Noether point symmetries isomorphic to sl(2,R). We observe that, considering k and l as parameters, this approach also allows us to relate different dynam- ical systems with (k,l) 6= (k′, l′) that have an isomor- phic symmetry algebra, the corresponding symmetry generators being related by the parameterization of the functions f(q) and g(q). 3.2. Systems with fixed symmetry Let us now consider the second possibility, namely, fixing the functions f(q) and g(q) in (17). To this extent, consider f(q) = qn1 , g(q) = qn2 for simplicity, where n 6= 0. In this case, the symmetry condition for X1 leads to the system( nqn−1i + nq n−1 j −2 ) Aij(q) + qn1 ∂Aij ∂q1 + qn2 ∂Aij ∂q2 = 0, qn1 A11 + q n 2 A12 − ∂V ∂q1 = 0, qn1 A12 + q n 2 A22 − ∂V ∂q2 = 0 for (ij) = (11), (12), (22). It is not too difficult to see that if A11(q) = A22(q) 6= 0 holds, then the preceding system possesses a nontrivial solution only for the values n = 0, 1. In order to find manage- able solutions for arbitrary n, we thus assume that A11(q) = A22(q) = 0. Then the system can be re- duced, and for n 6= 1 admits the solution A12(q) = (q1q2)−n exp (−(q1−n1 + q1−n2 ) n− 1 ) , V (q) = exp (−(q1−n1 + q1−n2 ) n− 1 ) . For n = 1, we merely get A12(q) = 1 (the free pseudo- Euclidean Lagrangian) and the gauge term V (q) = q1q2. Further, considering a potential U(q) requires to solve the additional PDE qn1 ∂U ∂q1 + qn1 ∂U ∂q2 + 2U(q) = 0, (21) that can be easily seen to provide the solution U(q) = exp (2q1−n1 n− 1 ) Ψ (qn−11 + qn−12 (q1q2)n−1 ) . (22) If we admit generalized potentials depending on q̇, the integrability condition (11) can be separated with respect to the variable t, because of the X3-invariance. Skipping the details, it follows from a routine computa- tion that the most general U preserving the subalgebra sl(2,R) in the preceding realization is given by U(q, q̇) = exp (2q1−n1 n− 1 ) Ψ (qn−11 + qn−12 (q1q2)n−1 ,u1,u2 ) , where the auxiliary variables are defined as ui = q̇iq1−ni exp ( − 2q1−n1 n− 1 ) , i = 1, 2. (23) For generic choices of the function Ψ, the system determined by T − εU always possesses a Noether point symmetry algebra isomorphic to sl(2,R). 4. Time-dependent Lagrangians As follows from Lemma 1, for conservative systems a Noether point symmetry depends at most quadrati- cally on time. If we skip the conservative character of the system, a more wide class of possibilities is given, and (explicitly time-dependent) constants of the mo- tion can still be guaranteed whenever an appropriate subalgebra of Noether symmetries is chosen [8]. Let θ(t) and ρ(t) be two arbitrary functions, and let {u(t),v(t)} be an independent set of solutions of the second-order ODE z̈(t) + ρ(t)ż(t) + θ(t)z(t) = 0. (24) In these conditions, the function ξ (t) = C1u(t)2 + C2u(t)v(t) + C3v(t)2 determines the general solution of the third-order ODE ... ξ + 3ρ̇ξ̈ + ( ρ̇ + 2ρ2 + 4θ ) ξ̇ + ( 4ρθ + 2θ̇ ) ξ = 0. (25) Let us consider vector fields of the generic shape X = ξ (t) ∂ ∂t + 1 2 ξ̇ (t) qi ∂ ∂qi , (26) and define Xi (1 ≤ i ≤ 3) as the vector field associated to the constant Cj = δ j i . Computing the brackets, taking into account the constraint (25), we get the relations [X1, X2] = W(u,v)X1, [X1, X3] = 2W(u,v)X2, [X2, X3] = W(u,v)X3, (27) 376 vol. 57 no. 6/2017 Noether Point Symmetries of Systems where W(u,v) = uv̇ − u̇v denotes the Wronskian of {u(t),v(t)}. Now, if W (u,v) reduces to a constant λ,1 it is straightforward to verify that the vector fields X1, X2 and X3 span a Lie algebra isomorphic to sl(2,R). For convenience, we further define the function V (q) = 1 4 ξ̈ (t) ( q21 + q 2 2 ) , which shall serve as generic gauge term for the sym- metry condition (11). Analyzing the symmetry condi- tion (7), the vector fields Xi are the symmetry gener- ators of a Noether point symmetry of a Lagrangian L (t, q, q̇) whenever the first-order PDE 1 2 ∂L ∂q̇1 ( ξ̈q1 − ξ̇q̇1 ) + 1 2 ∂L ∂q̇2 ( ξ̈q2 − ξ̇q̇2 ) + ξ ∂L ∂t + ξ̇ 2 ( q1 ∂L ∂q1 + q2 ∂L ∂q2 ) − ... ξ 4 ( q21 + q 2 2 ) − 1 2 ξ̈q̇1q1 − 1 2 ξ̈q̇2q2 + ξ̇L = 0 (28) is satisfied. The general solution to the latter is given by L (t, q, q̇) = ( ξξ̈ − ξ̇2 ) 4ξ2 ( q21 + q 2 2 ) + ξ̇ (q1q̇1 + q̇2q2) 2ξ + 1 ξ Ψ ( q1√ ξ , q2√ ξ , ξ̇q1 − 2q̇1ξ√ ξ , ξ̇q2 − 2q̇2ξ√ ξ ) , with Ψ an arbitrary function of its arguments. We observe that the previous class of Lagrangians con- tains, in particular, a family giving rise to oscillatory systems with a time-dependent frequency: L = 1 2 ( q̇21 + q̇ 2 2 ) + 2 ¨ξ(t)ξ(t) − 1 ξ(t)2 ( q21 + q 2 2 ) . (29) It should be remarked that the realization of type (26) exhibits the most general form that the term in ∂ ∂t can have, as follows at once from the following property: Lemma 2. Let X = ξ(t, q) ∂ ∂t + ηj(t, q) ∂∂qj be a Noether point symmetry of a regular Lagrangian L = Aij(t, q)q̇iq̇j−U(t, q). Then the condition ∂ξ∂q = 0 always holds. The proof is completely analogous to that of Lemma 1, and follows immediately from the inspec- tion of the terms in the symmetry condition (7) having the highest power in the velocities q̇, as well as the regularity of the Lagrangian. 5. Conclusions In this work we have illustrated, using functional re- alizations of Lie algebras based on the simple Lie algebra sl(2,R), different possibilities to formulate a kind of inverse problem in dynamics, imposing that 1This condition is ensured whenever we set ρ(t) = 0 in equation (24). See e.g. [14], p. 512. the generators appear as Noether point symmetries of Lagrangian dynamical systems. This allows either to consider symmetry-preserving perturbations of a given system, as developed in [12], or to derive the most general Lagrangian invariant by the functional real- ization of the Lie algebra. The cases of conservative and dissipative systems can be treated simultaneously, considering realizations explicitly depending on time- dependent functions. Albeit the examples have been restricted to the plane by simplicity, there is no ob- struction to formulate the problem in arbitrary dimen- sion. However, in order to ensure that the system is integrable [4], it is convenient to consider a realization of a symmetry algebra sl(2,R) ⊂ g, so that formula (8) provides a sufficient number of independent constants of the motion. In this situation, it should be taken into account that this approach by means of Noether point symmetries in the N-dimensional (conservative) case is somewhat restricted, as the corresponding preserved symmetry algebras are subalgebras of the Noether point symmetry algebra of the free system defined by the kinetic term T of the Lagrangian. For the Euclidean case, this corresponds to subalgebras of the Schrödinger algebra S(N), and hence any semisimple Lie algebra considered in this frame must be taken as a subalgebra of sl(2,R) ⊕ so(N) (see e.g. [12, 15]). However, for non-Euclidean geometries, the situation may differ. As an illustrative example consider the free Lagrangian L0 = 1q2n ( q̇21 + · · · + q̇2n ) in the upper- half space U = {q ∈ Rn| qn > 0} endowed with the Poincaré metric ds2 = 1 (qn)2 ( dq1 ⊗dx1 + · · · + dqn ⊗dqn ) (30) It is well known that it admits the conformal group SO (1,n) as isometry group [16], i.e., the correspond- ing symmetry generators are Killing vectors. It is easy to show that for any n ≥ 2, the algebra LPS of Lie point symmetries of the dynamical system associated to L0 is isomorphic to the direct sum so (1,n) ⊕ r2, with r2 the 2-dimensional affine Lie algebra. Only the Lie point symmetry Y = t ∂ ∂t fails to satisfy the condition (7), as it leads to the PDE − 1 2q2n n∑ k=1 ( q̇2k + q̇k ∂V ∂qk ) − ∂V ∂t , (31) which has no solution for a gauge term V (t, q), thus the Noether point symmetries are given by the Lie algebra LNS ' so (1,n)⊕R. As a consequence, for any n ≥ 2, the algebra LNS of Noether point symmetries of a second-order system q̈k = 2q̇kq̇n qn − q2n 2 ∂U ∂qk , 1 ≤ k ≤ n− 1, q̈n = q̇2n − q̇21 −···− q̇2n−1 qn − q2n 2 ∂U ∂qn (32) with Lagrangian L = L0 −U (t, q) corresponds to a subalgebra of so (1,n) ⊕R. These two cases indicate 377 Rutwig Campoamor-Stursberg Acta Polytechnica that a detailed study of the symmetry algebras of free Lagrangians corresponding to nonequivalent metrics in N ≥ 2 dimensions would give rise to a hierarchy of Lie algebras that allows to systematize the symmetry analysis of perturbed systems. For some important types of differential equations, this approach has al- ready provided interesting results (see [15, 17] and references therein). Further work along these lines is currently in progress. Finally, as follows from the symmetry condition (7), a Noether point symmetry of a regular Lagrangian L necessarily possesses the generic form X = ξ(t) ∂ ∂t + ηj(t, q) ∂∂qj . This suggests to study specifically re- alizations of Lie algebras of this type, in order to characterize those isomorphism classes of Lie algebras that appear as Noether symmetries of a system, but do not correspond to an isometry generator of the associated kinetic Lagrangian. Some developments in this direction have been proposed in [18], in con- nection with various geometric properties. A similar approach can be found in [15] and some previous work, where symmetries of certain types of non-autonomous systems defined in Riemann spaces have been studied, providing new insights to the geometrical interpreta- tion of dynamical quantities. An interesting problem worthy to be analyzed in the context of the symmetry analysis of dynamical systems is the possibility of com- bining symmetry groups with certain of the geometric properties of the orbits determined by the solutions of a system [5], a question that has still not exhausted the possibilities of the group-theoretical approach. Acknowledgements The author expresses his gratitude to Prof. M. Znojil for the invitation to the AAMP XIV conference. During the preparation of this work, the author was financially supported by the research project MTM2016- 79422-P of the AEI/FEDER (EU). References [1] E. Noether. Invariante Variationsprobleme. Nachr Ges Wiss Göttingen, Math-Phys Kl 1918:235–257, 1918. [2] L. V. Ovsyannikov. Group Analysis of Differential Equations. Academic, New-York, 1982. [3] L. Dresner. Application of Lie’s Theory of Ordinary and Partial Differential Equations. IOP Publishing, Bristol, 1999. [4] A. M. Perelomov. Integrable Systems of Classical Mechanics and Lie Algebras. Birkhäuser Verlag, Basel, 1990. [5] A. S. Galiullin. Inverse Problems of Dynamics. Mir Publishers, Moscow, 1984. [6] S. E. Jones, B. G. Vujanovic. On the inverse Lagrangian problem. Acta Mech 73:245–251, 1988. doi:10.1007/BF01177044. [7] V. M. Gorringe, P. G. L. Leach. Lie point symmetries for systems of second-order linear differential equations. Quaest Math 11:95–117, 1988. doi:10.1080/16073606.1988.9631946. [8] R. Campoamor-Stursberg. On certain types of point symmetries of systems of second-order ordinary differential equations. Comm Nonlinear Sci Num Simulat 19:2602–2614, 2014. doi:10.1016/j.cnsns.2014.01.006. [9] P. G. L. Leach. Equivalence classes of second-order ordinary differential equations with only a three dimensional Lie algebra of point symmetries and linearisation. J Math Anal Appl 284:31–48, 2003. doi:10.1016/S0022-247X(03)00147-1. [10] A. Ballesteros, et al. N-dimensional sl(2)-coalgebra spaces with non-constant curvature. Phys Letters B 652:376–383, 2007. doi:10.1016/j.physletb.2007.07.012. [11] G. Gubbioti, M. C. Nucci. Are all classical superintegrable systems in two-dimensional space linearizable? J Math Phys 58:012902, 2017. doi:10.1063/1.4974264. [12] R. Campoamor-Stursberg. Perturbations of Lagrangian systems based on the preservation of subalgebras of Noether symmetries. Acta Mech 227:1941–1956, 2016. doi:10.1007/s0070. [13] E. Kamke. Differentialgleichungen. Lösungsmethoden und Lösungen. Band II. Akademische Verlagsgesellschaft, Leipzig, 1962. [14] E. Kamke. Differentialgleichungen. Lösungsmethoden und Lösungen. Band I. Akademische Verlagsgesellschaft, Leipzig, 1962. [15] L. Karpathopoulos, A. Paliathanasis, M. Tsamparlis. Lie and Noether point symmetries for a class of nonautonomous dynamical systems. J Math Phys 58:082901, 2017. doi:10.1063/1.4998715. [16] Y. Fuji, K. Yamagishi. Killing spinors on spheres and hyperbolic manifolds. J Math Phys 27:979–981, 1986. doi:10.1063/1.527118. [17] M. Tsamparlis, A. Paliathanasis, A. Qadir. Noether symmetries and isometries of the minimal surface Lagrangian under constant volume in a Riemannian space. Int J Geom Methods Mod Phys 12:1550033, 2015. [18] R. Campoamor-Stursberg. An alternative approach to systems of second-order ordinary differential equations with maximal symmetry. Realizations of sl(n + 2, R) by special functions. Comm Nonlinear Sci Num Simulat 37:200–211, 2016. doi:10.1016/j.cnsns.2016.01.015. 378 http://dx.doi.org/10.1007/BF01177044 http://dx.doi.org/10.1080/16073606.1988.9631946 http://dx.doi.org/10.1016/j.cnsns.2014.01.006 http://dx.doi.org/10.1016/S0022-247X(03)00147-1 http://dx.doi.org/10.1016/j.physletb.2007.07.012 http://dx.doi.org/10.1063/1.4974264 http://dx.doi.org/10.1007/s0070 http://dx.doi.org/10.1063/1.4998715 http://dx.doi.org/10.1063/1.527118 http://dx.doi.org/10.1016/j.cnsns.2016.01.015 Acta Polytechnica 57(6):373–378, 2017 1 Introduction 2 Lie and Noether point symmetries of systems 2.1 Perturbations that preserve symmetry subalgebras 3 Functional realizations of sl(2,R) as Noether symmetry algebra 3.1 Separable kinetic Lagrangians 3.2 Systems with fixed symmetry 4 Time-dependent Lagrangians 5 Conclusions Acknowledgements References