signature=f4cc4caf9dfae5cdabe9df2bfa43b008,Inverse problem for Lagrangian systems on Lie algebroids ...

Appendix: Relation to other approaches

In Sect. 4.2 we recover the Helmholtz conditions given in [21] as the vanishing of \(d^{\mathcal {L}^{\tau }E}\Theta _{\Gamma ,F}\) on a certain basis of sections of \(\mathcal {L}^{\tau }E\longrightarrow E\).

In the previous section we worked in the basis \(\{\tilde{T}_{\alpha },\tilde{V}_{\alpha }\}\) of local sections of \(\mathcal {L}^{\tau }E\), constructed from a basis \(\{e_{\alpha }\}\) of local sections of E. Another common basis of sections of \(\mathcal {L}^{\tau }E\) is \(\{e_{\alpha }^{C},e_{\alpha }^{V}\}\), the set of complete and vertical lifts of \(\{e_{\alpha }\}\). The relationship between both is

$$\begin{aligned} \tilde{T}_{\alpha }=e_{\alpha }^{C}+C^{\gamma }_{\alpha \beta }y^{\beta }\tilde{V}_{\gamma } \quad \mathrm{{\,and\,}} \quad \tilde{V}_{\gamma }=e_{\gamma }^{V}. \end{aligned}$$

As in the tangent bundle case, a SODE on a Lie algebroid defines a connection (see [21]). Then the horizontal lift of a section \(X\in \Gamma (E)\) can be defined from its complete and vertical lift and the SODE as

$$\begin{aligned} X^{H}=\frac{1}{2}\left( X^{C}-[\Gamma ,X^{V}]\right) , \end{aligned}$$

and we get another basis \(\{H_{\alpha }:=e_{\alpha }^{H},e_{\alpha }^{V}\}\) of sections of \(\mathcal {L}^{\tau }E\). The relationship with the above is given by

$$\begin{aligned} H_{\alpha }=e_{\alpha }^{H}=\tilde{T}_{\alpha }+\frac{1}{2}\left( \frac{\partial \Gamma ^{\gamma }}{\partial y^{\alpha }}-C^{\gamma }_{\alpha \beta }y^{\beta } \right) \tilde{V}_{\gamma }=\tilde{T}_{\alpha }+\Lambda ^{\gamma }_{\alpha }\tilde{V}_{\gamma }. \end{aligned}$$

Note that if \(\Gamma \) is variational we have \(\mathcal {L}_{\Gamma }F^{*}\lambda _{E}=\Theta _{\Gamma ,F}\) for some local diffeomorphism F and hence \(\mathcal {L}_{\Gamma }d^{\mathcal {L}^{\tau }E}F^{*}\lambda _{E}=d^{\mathcal {L}^{\tau }E}\Theta _{\Gamma ,F}\). Then the equations

$$\begin{aligned}&\mathcal {L}_{\Gamma }d^{\mathcal {L}^{\tau }E}F^{*}\lambda _{E}(H_{\eta },H_{\beta })=0, \quad \mathcal {L}_{\Gamma }d^{\mathcal {L}^{\tau }E}F^{*}\lambda _{E}(H_{\eta },\tilde{V}_{\beta })=0 \quad \mathrm{{\,and\,}}\\&\mathcal {L}_{\Gamma }d^{\mathcal {L}^{\tau }E}F^{*}\lambda _{E}(\tilde{V}_{\eta },\tilde{V}_{\beta })=0, \end{aligned}$$

together with

$$\begin{aligned} \mathcal {L}_{\Gamma }d^{\mathcal {L}^{\tau }E}F^{*}\lambda _{E}\left( H_{\eta },\tilde{V}_{\beta }\right) -\mathcal {L}_{\Gamma }d^{\mathcal {L}^{\tau }E}F^{*}\lambda _{E}\left( H_{\beta },\tilde{V}_{\eta }\right) =0 \end{aligned}$$

yield the Helmholtz conditions given in [21].

In order to check this we first compute \([\Gamma ,\tilde{V}_{\eta }]\) and \([\Gamma ,H_{\eta }]\) in terms of the basis \(\{H_{\alpha },\tilde{V}_{\alpha }\}\):

$$\begin{aligned} \,[\Gamma ,\tilde{V}_{\eta }]= & {} -\tilde{T}_{\eta }-\frac{\partial \Gamma ^{\alpha }}{\partial y^{\eta }}\tilde{V}_{\alpha }\nonumber \\= & {} -\left( H_{\eta }-\Lambda _{\eta }^{\beta }\tilde{V}_{\beta }\right) -\frac{\partial \Gamma ^{\alpha }}{\partial y^{\eta }}\tilde{V}_{\alpha }=-H_{\eta }+\frac{1}{2}\left( C^{\gamma }_{\beta \eta }y^{\beta }-\frac{\partial \Gamma ^{\gamma }}{\partial y^{\eta }}\right) \tilde{V}_{\gamma },\\ \,[\Gamma ,H_{\eta }]= & {} [\Gamma ,\tilde{T}_{\eta }]+[\Gamma ,\Lambda ^{\gamma }_{\eta }\tilde{V}_{\gamma }]=[\Gamma ,\tilde{T}_{\eta }]+\rho (\Gamma )(\Lambda ^{\gamma }_{\eta })\tilde{V}_{\gamma }+\Lambda ^{\gamma }_{\eta }[\Gamma ,\tilde{V}_{\gamma }]\\= & {} -\left( \rho ^{i}_{\eta }\frac{\partial \Gamma ^{\alpha }}{\partial x^{i}}\tilde{V}_{\alpha }+y^{\alpha }C^{\gamma }_{\eta \alpha }(H_{\gamma }-\Lambda ^{\nu }_{\gamma }\tilde{V}_{\nu }) \right) + \left( y^{\alpha }\rho ^{i}_{\alpha }\frac{\partial \Lambda ^{\gamma }_{\eta }}{\partial x^{i}}+\Gamma ^{\alpha }\frac{\partial \Lambda ^{\gamma }_{\eta }}{\partial y^{\alpha }} \right) \tilde{V}_{\gamma }\\&+\, \Lambda ^{\gamma }_{\eta }\left( -H_{\gamma }+\frac{1}{2}\left( C^{\nu }_{\beta \gamma }y^{\beta }-\frac{\partial \Gamma ^{\nu }}{\partial y^{\gamma }} \right) \tilde{V}_{\nu }\right) = \frac{1}{2}\left( y^{\beta }C_{\beta \alpha }^{\gamma }-\frac{\partial \Gamma ^{\gamma }}{\partial y^{\alpha }} \right) H_{\gamma } \\&+\left( y^{\alpha }\rho ^{i}_{\alpha }\frac{\partial \Lambda ^{\gamma }_{\eta }}{\partial x^{i}}+\Gamma ^{\alpha }\frac{\partial \Lambda ^{\gamma }_{\eta }}{\partial y^{\alpha }}+\Lambda ^{\nu }_{\eta }\Lambda ^{\gamma }_{\nu }-\Lambda ^{\nu }_{\eta }\frac{\partial \Gamma ^{\gamma }}{\partial y^{\nu }}-\rho ^{i}_{\eta }\frac{\partial \Gamma ^{\gamma }}{\partial x^{i}}+y^{\alpha }C^{\nu }_{\eta \alpha }\Lambda ^{\gamma }_{\nu } \right) \tilde{V}_{\gamma }. \end{aligned}$$

We introduce the notation

$$\begin{aligned} D_{\eta }^{\gamma }:= & {} \frac{1}{2}\left( y^{\beta }C_{\beta \alpha }^{\gamma }-\frac{\partial \Gamma ^{\gamma }}{\partial y^{\alpha }} \right) \quad \mathrm{{\,and\,}}\nonumber \\ \Phi ^{\gamma }_{\eta }:= & {} \left( y^{\alpha }\rho ^{i}_{\alpha }\frac{\partial \Lambda ^{\gamma }_{\eta }}{\partial x^{i}}+\Gamma ^{\alpha }\frac{\partial \Lambda ^{\gamma }_{\eta }}{\partial y^{\alpha }}+\Lambda ^{\nu }_{\eta }\Lambda ^{\gamma }_{\nu }-\Lambda ^{\nu }_{\eta }\frac{\partial \Gamma ^{\gamma }}{\partial y^{\nu }}-\rho ^{i}_{\eta }\frac{\partial \Gamma ^{\gamma }}{\partial x^{i}}+y^{\alpha }C^{\nu }_{\eta \alpha }\Lambda ^{\gamma }_{\nu } \right) \end{aligned}$$

so that \([\Gamma ,\tilde{V}_{\eta }]=-H_{\eta }+D^{\gamma }_{\eta }\tilde{V}_{\gamma }\) and \([\Gamma ,H_{\eta }]=D_{\eta }^{\gamma }H_{\gamma }+\Phi ^{\gamma }_{\eta }\tilde{V}_{\gamma }\).

We will also need the expression of \(d^{\mathcal {L}^{\tau }E}F^{*}\lambda _{E}\) in terms of \(\{\theta ^{\alpha }:=\tilde{V}^{\alpha }-\Lambda _{\beta }^{\alpha }\tilde{T}^{\beta },\tilde{T}^{\alpha }\}\), the dual basis of \(\{H_{\alpha },\tilde{V}_{\alpha }\}\):

$$\begin{aligned} d^{\mathcal {L}^{\tau }E}F^{*}\lambda _{E}= & {} \left( \rho (H_{\gamma })(F_{\alpha })-\frac{1}{2}F_{\nu }C^{\nu }_{\gamma \alpha } \right) \tilde{T}^{\gamma }\wedge \tilde{T}^{\alpha }+\frac{\partial F_{\alpha }}{\partial y^{\gamma }}\theta ^{\gamma }\wedge \tilde{T}^{\alpha }\\= & {} A_{\gamma \alpha }\tilde{T}^{\gamma }\wedge \tilde{T}^{\alpha }+\frac{\partial F_{\alpha }}{\partial y^{\gamma }}\theta ^{\gamma }\wedge \tilde{T}^{\alpha }\, \end{aligned}$$

where \(A_{\gamma \alpha }=\rho (H_{\gamma })(F_{\alpha })-\frac{1}{2}F_{\nu }C^{\nu }_{\gamma \alpha }\).

Now we introduce the notation \(T_{F}:=d^{\mathcal {L}^{\tau }E}F^{*}\lambda _{E}\) and write the Helmholtz conditions in local coordinates as follows:

$$\begin{aligned} \mathcal {L}_{\Gamma }T_{F}(H_{\eta },H_{\beta })= & {} \Gamma (T_{F}(H_{\eta },H_{\beta }))-T_{F}([\Gamma ,H_{\eta }],H_{\beta })-T_{F}(H_{\eta },[\Gamma ,H_{\beta }]) \nonumber \\= & {} \Gamma (A_{\eta \beta })-\Gamma (A_{\beta \eta })-\left[ A_{\gamma \beta }D^{\gamma }_{\eta }-A_{\beta \gamma }D^{\gamma }_{\eta }+\frac{\partial F_{\beta }}{\partial y^{\gamma }}\Phi ^{\gamma }_{\eta } \right] \nonumber \\&-\left[ A_{\eta \gamma }D^{\gamma }_{\beta }-A_{\gamma \eta }D^{\gamma }_{\beta }-\frac{\partial F_{\eta }}{\partial y^{\gamma }}\Phi ^{\gamma }_{\beta } \right] = 0\end{aligned}$$

(21)

$$\begin{aligned} \mathcal {L}_{\Gamma }T_{F}(H_{\eta },\tilde{V}_{\beta })= & {} \Gamma (T_{F}(H_{\eta },\tilde{V}_{\beta }))-T_{F}([\Gamma ,H_{\eta }],\tilde{V}_{\beta })-T_{F}(H_{\eta },[\Gamma ,\tilde{V}_{\beta }]) \nonumber \\= & {} \Gamma \left( -\frac{\partial F_{\eta }}{\partial y^{\beta }}\right) +\frac{\partial F_{\gamma }}{\partial y^{\beta }}D^{\gamma }_{\eta }-\left[ -A_{\eta \beta }+A_{\beta \eta }-\frac{\partial F_{\eta }}{\partial y^{\gamma }}D^{\gamma }_{\beta } \right] =0 \end{aligned}$$

(22)

$$\begin{aligned} \mathcal {L}_{\Gamma }T_{F}(\tilde{V}_{\eta },\tilde{V}_{\beta })= & {} -T_{F}([\Gamma ,\tilde{V}_{\eta }],\tilde{V}_{\beta })-T_{F}(\tilde{V}_{\eta },[\Gamma ,\tilde{V}_{\beta }]) \nonumber \\= & {} T_{F}(H_{\eta },\tilde{V}_{\beta })+T_{F}(\tilde{V}_{\eta },H_{\beta })=-\frac{\partial F_{\eta }}{\partial y^{\beta }}+\frac{\partial F_{\beta }}{\partial y^{\eta }}=0 \end{aligned}$$

(23)

Now we compute

$$\begin{aligned}&\mathcal {L}_{\Gamma }T_{F}(H_{\eta },\tilde{V}_{\beta })-\mathcal {L}_{\Gamma }T_{F}(H_{\beta },\tilde{V}_{\eta }) \\&\quad = \Gamma \left( -\frac{\partial F_{\eta }}{\partial y^{\beta }} \right) +\frac{\partial F_{\gamma }}{\partial y^{\beta }}D^{\gamma }_{\eta }-\left[ -A_{\eta \beta }+A_{\beta \eta }-\frac{\partial F_{\eta }}{\partial y^{\gamma }}D^{\gamma }_{\beta } \right] \\&\quad \quad -\Gamma \left( -\frac{\partial F_{\beta }}{\partial y^{\eta }} \right) -\frac{\partial F_{\gamma }}{\partial y^{\eta }}D^{\gamma }_{\beta } +\left[ -A_{\beta \eta }+A_{\eta \beta }-\frac{\partial F_{\beta }}{\partial y^{\gamma }}D^{\gamma }_{\eta } \right] \end{aligned}$$

and use (23) to obtain

$$\begin{aligned} A_{\eta \beta }=A_{\beta \eta }. \end{aligned}$$

(24)

Substituting (24) into (21) and (22) these equations become

$$\begin{aligned} \frac{\partial F_{\beta }}{\partial y^{\gamma }}\Phi ^{\gamma }_{\eta }=\frac{\partial F_{\eta }}{\partial y^{\gamma }}\Phi ^{\gamma }_{\beta } \quad \mathrm{{\,and\,}} \quad \Gamma \left( \frac{\partial F_{\eta }}{\partial y^{\beta }}\right) -\frac{\partial F_{\gamma }}{\partial y^{\beta }}D^{\gamma }_{\eta }-\frac{\partial F_{\eta }}{\partial y^{\gamma }}D^{\gamma }_{\beta }=0, \end{aligned}$$

which are the equations given in [21]. This can be checked directly by making the substitution

$$\begin{aligned} \Lambda ^{\nu }_{\eta }\frac{\partial \Gamma ^{\gamma }}{\partial y^{\nu }}=2\Lambda ^{\nu }_{\eta }\Lambda ^{\gamma }_{\nu }+\Lambda ^{\nu }_{\eta }C^{\gamma }_{\nu \tau }y^{\tau }. \end{aligned}$$

Beware that the notation in [21] is \(N^{\gamma }_{\eta }=-\Lambda ^{\gamma }_{\eta }\).

Note that the Helmholtz conditions for invariant Lagrangians on the tangent bundle of a Lie group G given in [7] are also recovered. Indeed, by dropping the terms where derivatives with respect to \(x^{i}\) appear and substituting \(\Lambda ^{\gamma }_{\eta }=\frac{\partial \Gamma ^{\gamma }}{\partial y^{\eta }}-C^{\gamma }_{\eta \beta }y^{\beta }\) we get

$$\begin{aligned} \Phi ^{\gamma }_{\eta }&=\frac{1}{2}\Gamma ^{\alpha }\frac{\partial ^{2}\Gamma ^{\gamma }}{\partial y^{\alpha }\partial y^{\eta }}-\frac{1}{2}\Gamma ^{\alpha }C^{\gamma }_{\eta \alpha }-\frac{1}{4}\frac{\partial \Gamma ^{\nu }}{\partial y^{\eta }}\frac{\partial \Gamma ^{\gamma }}{\partial y^{\nu }}-\frac{1}{4}C^{\nu }_{\eta \beta }y^{\beta }C^{\gamma }_{\nu \tau }y^{\tau }\\&\quad \;-\frac{1}{4}\frac{\partial \Gamma ^{\nu }}{\partial y^{\eta }}C^{\gamma }_{\nu \tau }y^{\tau }+\frac{3}{4}C^{\nu }_{\eta \beta }y^{\beta }\frac{\partial \Gamma ^{\gamma }}{\partial y^{\nu }}. \end{aligned}$$