Abstract
This work represents the continuation of our paper (Arlotti and Lods, in An \(L^{p}\)-approach to the well-posedness of transport equations associated to a regular field—part I. Mediterr J Math. 2018. https://doi.org/10.1007/s00009-019-1425-8). In \(L^{p}\)-spaces \(1< p <\infty \) we investigate well-posedness of transport equations with general external Lipschitz fields and general measures associated to a large variety of boundary conditions modelled by abstract boundary operators H. In particular, a new explicit formula for the corresponding transport semigroup is given. Some applications are also presented.
Similar content being viewed by others
References
Arlotti, L., Lods, B.: Substochastic semigroups for transport equations with conservative boundary conditions. J. Evol. Equ. 5, 485–508 (2005)
Arlotti, L., Banasiak, J., Lods, B.: A new approach to transport equations associated to a regular field: trace results and well-posedness. Mediterr. J. Math. 6, 367–402 (2009)
Arlotti, L., Banasiak, J., Lods, B.: On general transport equations with abstract boundary conditions. The case of divergence free force field. Mediterr. J. Math 8, 1–35 (2011)
Arlotti, L.: Explicit transport semigroup associated to abstract boundary conditions, Discrete Contin. Dyn. Syst. A, Dynamical systems, differential equations and applications. In: 8th AIMS Conference. Suppl. I, pp. 102–111 (2011)
Arlotti, L., Lods, B.: An \(L^{p}\)-approach to the well-posedness of transport equations associated to a regular field–part I. Mediterr. J. Math. (2018). https://doi.org/10.1007/s00009-019-1425-8
Banasiak, J., Arlotti, L.: Perturbations of Positive Semigroups with Applications. Springer, Berlin (2006)
Banasiak, J., Falkiewicz, A., Namayanja, P.: Semigroup approach to diffusion and transport problems on networks. Semigroup Forum 93, 427–443 (2016)
Batkai, A., Kramar Fijavž, M., Rhandi, A.: Positive Operator Semigroups. From Finite to Infinite Dimensions. Operator Theory: Advances and Applications, vol. 257. Birkhauser/Springer, Cham (2017)
Boulanouar, M.: On a mathematical model of age-cycle length structured cell population with non-compact boundary conditions. Math. Methods Appl. Sci. 38, 2081–2104 (2015)
Boulanouar, M.: On a mathematical model of age-cycle length structured cell population with non-compact boundary conditions (II). Math. Methods Appl. Sci. 39, 1855–1876 (2016)
Dautray, R., Lions, J.L.: Mathematical Analysis and Numerical Methods for Science and Technology, vol. 6. Evolution Problems II. Springer, Berlin (2000)
Engel, K.-J., Kramar Fijavž, M.: Exact and positive controllability of boundary control systems. Netw. Heterog. Media 12, 319–337 (2017)
Greiner, G.: Perturbing the boundary conditions of a generator. Houston J. Math. 13, 213–229 (1987)
Lods, B., Mokhtar-Kharroubi, M.: On the theory of a growing cell population with zero minimum cycle length. J. Math. Anal. Appl. 266, 70–99 (2002)
Pazy, A.: Semigroups of Linear Operators and Applications to Partial Differential Equations. Springer, Berlin (1983)
Voigt, J.: Functional Analytic Treatment of the Initial Boundary Value Problem for Collisionless Gases. Habilitationsschrift, München (1981)
Author information
Authors and Affiliations
Corresponding author
Additional information
Publisher's Note
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Appendix A: On the Family \((U_{k}(t))_{t \geqslant 0}\)
Appendix A: On the Family \((U_{k}(t))_{t \geqslant 0}\)
We prove that the family of operators \((U_{k}(t))_{t \geqslant 0}\) introduced in Definition 3.6 is well defined and satisfies Theorem 3.8. We first need to establish general properties of \({\mathcal {T}}_{\mathrm {max},\,p}\).
1.1 A.1 Additional Properties of \({\mathcal {T}}_{\mathrm {max},\,p}\)
We establish here several results, reminiscent of [4] about how \({\mathcal {T}}_{\mathrm {max},\,p}\) and some strongly continuous family of operators can interplay. We start with the following where we recall that \({\mathscr {D}}_{0}\) has been defined in the beginning of Section 3.
Proposition A.1
Let \((U(t))_{t\geqslant 0}\) be a strongly continuous family of \({\mathscr {B}}(X)\). For any \(f \in X\), set
Assume that
- (i)
For any \(f \in {\mathscr {D}}_{0},\) the mapping \(t \in [0,\infty ) \mapsto U(t)f \in X\) is differentiable with
$$\begin{aligned} \dfrac{\mathrm {d}}{\mathrm {d}t}U(t)f=U(t){\mathcal {T}}_{\mathrm {max},\,p}f, \quad t \geqslant 0. \end{aligned}$$ - (ii)
For any \(f \in {\mathscr {D}}_{0}\) and any \(t \geqslant 0\), it holds that \(U(t)f \in {\mathscr {D}}({\mathcal {T}}_{\mathrm {max},\,p})\) with \({\mathcal {T}}_{\mathrm {max},\,p}U(t)f=U(t){\mathcal {T}}_{\mathrm {max},\,p}f.\)
Then, the following holds
- (1)
for any \(f \in X\) and \(t >0\), \(I_{t}[f] \in {\mathscr {D}}({\mathcal {T}}_{\mathrm {max},\,p})\) with
$$\begin{aligned} {\mathcal {T}}_{\mathrm {max},\,p}I_{t}[f]=U(t)f-U(0)f. \end{aligned}$$ - (2)
for any \(f \in {\mathscr {D}}_{0}\) the mapping \(t \in [0,\infty ) \mapsto {\mathsf {B}}^{\pm }U(t)f \in Y_{p}^{\pm }\) is continuous and,
$$\begin{aligned} {\mathsf {B}}^{\pm }I_{t}[f]=\int _{0}^{t}{\mathsf {B}}^{\pm }U(s)f \mathrm {d}s \quad \forall t >0. \end{aligned}$$
Let now \(f \in X\) be such that \({\mathsf {B}}^{-}I_{t}[f] \in L^p_-\) for all \(t >0\) with \(t \in [0,\infty ) \mapsto {\mathsf {B}}^{-}I_{t}[f] \in L^p_-\text { continuous,}\) then,
Proof
Under assumptions \(i)-ii)\), for any \(f \in {\mathscr {D}}_{0}\), since both the mappings \(t \mapsto U(t)f\) and \(t \mapsto {\mathcal {T}}_{\mathrm {max},\,p}U(t)f\) are continuous and \({\mathcal {T}}_{\mathrm {max},\,p}\) is closed one has \(I_{t}[f] \in {\mathscr {D}}({\mathcal {T}}_{\mathrm {max},\,p})\) with
This proves that (1) holds for \(f \in {\mathscr {D}}_{0}\) and, since \({\mathscr {D}}_{0}\) is dense in X, the result holds for any \(f \in X\).
Let us prove (2). Pick \(f \in {\mathscr {D}}_{0}\). Since the mapping \(t \geqslant 0\mapsto U(t)f \in {\mathscr {D}}({\mathcal {T}}_{\mathrm {max},\,p})\) is continuous for the graph norm on \({\mathscr {D}}({\mathcal {T}}_{\mathrm {max},\,p})\) while \({\mathsf {B}}^{\pm }\,:\,{\mathscr {D}}({\mathcal {T}}_{\mathrm {max},\,p}) \mapsto Y_{p}^{\pm }\) is continuous (see [5, Remark 3.4]), the mapping \(t \geqslant 0 \mapsto {\mathsf {B}}^{\pm }U(t)f \in Y_{p}^{\pm }\) is continuous. Moreover, \({\mathsf {B}}^{\pm }I_{t}[f]=\int _{0}^{t}{\mathsf {B}}^{\pm }U(s)f \mathrm {d}s\) still thanks to the continuity on \({\mathscr {D}}({\mathcal {T}}_{\mathrm {max},\,p})\) with the graph norm and point (1).
Let now \(f \in X\) be given such that \({\mathsf {B}}^{-}I_{t}[f] \in L^p_-\) for all \(t >0\) with \(t \geqslant 0 \mapsto {\mathsf {B}}^{-}I_{t}[f] \in L^p_-\) continuous. For all \(t \geqslant 0,\) \(h \geqslant -t\), denote now
Since \(I_{t}[f] \in {\mathscr {D}}({\mathcal {T}}_{\mathrm {max},\,p})\) and \({\mathsf {B}}^{-}I_{t}[f] \in L^p_-\), one has clearly \(I_{t,t+h}[f] \in {\mathscr {D}}({\mathcal {T}}_{\mathrm {max},\,p})\) and \({\mathsf {B}}^{-}I_{t,t+h}[f] \in L^p_-\) and Green’s formula (see [5, Equation (2.12)]) yields
Since \({\mathcal {T}}_{\mathrm {max},\,p}I_{t,t+h}[f]=U(t+h)f-U(t)f\), one gets
The continuity of \(s \geqslant 0\mapsto U(s)f \in X\) together with the one of \(s \geqslant 0 \mapsto {\mathsf {B}}^{-}I_{s}[f] \in L^p_-\) gives then that
i.e. \(t \geqslant 0 \mapsto {\mathsf {B}}^{+}I_{t}[f] \in L^p_+\) is continuous. \(\square \)
We can complement the above with the following whose proof is exactly as that of [4, Proposition 3] and is omitted here:
Proposition A.2
Let \((U(t))_{t\geqslant 0}\) be a strongly continuous family of \({\mathscr {B}}(X)\) satisfying the following, for any \(f \in {\mathscr {D}}_{0}\):
- (i)
For any \(t \geqslant 0\),
$$\begin{aligned}{}[U(t)f]({\mathbf {x}})=0 \quad \forall {\mathbf {x}}\in \mathbf {\Omega }\text { such that } \tau _{-}({\mathbf {x}}) \geqslant t. \end{aligned}$$ - (ii)
For any \({\mathbf {y}}\in \Gamma _{-}\), \(t >0\), \(0< r< s < \tau _{+}({\mathbf {y}})\), it holds
$$\begin{aligned} \left[ U(t)f\right] (\Phi ({\mathbf {y}},s))=\left[ U(t-s+r)f\right] (\Phi ({\mathbf {y}},r)). \end{aligned}$$ - (iii)
the mapping \(t \geqslant 0 \mapsto U(t)f \in X\) is differentiable with \(\frac{\mathrm {d}}{\mathrm {d}t}U(t)f=U(t){\mathcal {T}}_{\mathrm {max},\,p}f\) for any \(t \geqslant 0.\)
Then, the following properties hold
- (1)
For any \(f \in X\) and any \(t \geqslant 0\) and \(\mu _{-}\)-a.e. \({\mathbf {y}}\in \Gamma _{-}\), given \(0< s_{1}< s_{2} < \tau _{+}({\mathbf {y}})\), there exists \(0< r < s_{1}\) such that
$$\begin{aligned} \int _{s_{1}}^{s_{2}}\left[ U(t)f\right] (\Phi ({\mathbf {y}},s))\mathrm {d}s=\int _{t-s_{2}+r}^{t-s_{1}+r}\left[ U(\tau )f\right] (\Phi ({\mathbf {y}},r))\mathrm {d}\tau . \end{aligned}$$ - (2)
For any \(f \in {\mathscr {D}}_{0}\) and \(t \geqslant 0\), one has \(U(t)f \in {\mathscr {D}}({\mathcal {T}}_{\mathrm {max},\,p})\) with \({\mathcal {T}}_{\mathrm {max},\,p}U(t)f=U(t){\mathcal {T}}_{\mathrm {max},\,p}f.\)
1.2 A.2 Proof of Theorem 3.8
We now come to the proof of Theorem 3.8 which will consist of showing that the family of operators \((U_{k}(t))_{t \geqslant 0}\) introduced in Definition 3.6 is well defined and satisfies the properties listed in Theorem 3.8. The proof is made by induction and we start with a series of Lemmas (one for each of the above properties in Theorem 3.8) showing that \(U_{1}(t)\) enjoys all the listed properties.
As already mentioned, the fact that the mapping
is a measure isomorphism, for any \(f \in {\mathscr {D}}_{0}\) and \(t >0\), the function \(U_{1}(t)f\) is well defined and measurable on \(\mathbf {\Omega }\). Moreover, using [5, Proposition 2.2] and Fubini’s Theorem:
Therefore,
thanks to Proposition 3.5. Therefore \(\Vert U_{1}(t)f\Vert _{p} \leqslant {\left| \left| \left| H \right| \right| \right| }\,\Vert f\Vert _{p}\) for all \(f \in {\mathscr {D}}_{0}\) with moreover
Since \({\mathscr {D}}_{0}\) is dense in X, this allows to define a unique extension operator, still denoted by \(U_{1}(t) \in {\mathscr {B}}(X)\) with
Now, one has the following
Lemma A.3
The family \((U_{1}(t))_{t\geqslant 0}\) is strongly continuous on X.
Proof
Let \(t >0\) be fixed. Set \(\mathbf {\Omega }_{t}=\{{\mathbf {x}}\in \mathbf {\Omega }_{-}\,;\;\tau _{-}({\mathbf {x}}) \leqslant t \}.\) One has \([U_{1}(t)f]({\mathbf {x}})=0\) for any \({\mathbf {x}}\in \mathbf {\Omega }{{\setminus }} \mathbf {\Omega }_{t}\) and any \(f \in X.\) Let us fix \(f \in {\mathscr {D}}_{0}\) and \(h >0.\) One has
Now,
and, repeating the reasoning before Lemma A.3 one gets
Since \(U_{0}(s+h)f-U_{0}(s)f=U_{0}(s)\left( U_{0}(h)f-f\right) \) one gets from Proposition 3.5 that
This proves that
Let us investigate the second integral in (A.3). One first notices that, given \({\mathbf {x}}=\Phi ({\mathbf {y}},s)\) with \({\mathbf {y}}\in \Gamma _{-}\), \(0< s < \min (t,\tau _{+}({\mathbf {y}}))\), it holds
Therefore
and, since \((U_{0}(t))_{t\geqslant 0}\) is a contraction semigroup, we get
Using (A.2), we get \(\lim _{h\rightarrow 0^{+}}\int _{\mathbf {\Omega }_{t+h}{\setminus } \mathbf {\Omega }_{t}}|U_{1}(t+h)f|^{p}\mathrm {d}\mu =0\) and we obtain finally that \(\lim _{h \rightarrow 0^{+}}\Vert U_{1}(t+h)f-U_{1}(t)f\Vert _{p}^{p}=0.\) One argues in a similar way for negative h and gets
Since \({\mathscr {D}}_{0}\) is dense in X and \(\Vert U_{1}(t)\Vert _{{\mathscr {B}}(X)} \leqslant {\left| \left| \left| H \right| \right| \right| }\) we deduce that above limit vanishes for all \(f \in X\). This proves the result. \(\square \)
One has also the following
Lemma A.4
For all \(t \geqslant 0,\) \(h \geqslant 0\) and \(f \in X\) it holds
Proof
It is clearly enough to consider \(t >0\), \(h >0\) since \(U_{1}(0)f=0\) while \(U_{0}(0)\) is the identity operator. Notice that, for any \(f \in {\mathscr {D}}_{0}\) and any \(0 \leqslant t_{1} \leqslant t_{2}\), for \({\mathbf {x}}=\Phi ({\mathbf {y}},s) \in \mathbf {\Omega }_{t_{1}}\) we have
Now, given \(f \in X\) and \(0 \leqslant t_{1} \leqslant t_{2}\) the above formula is true for almost every \({\mathbf {x}}=\Phi ({\mathbf {y}},s) \in \mathbf {\Omega }_{t_{1}}\) by a density argument. Therefore, for almost every \({\mathbf {x}}=\Phi ({\mathbf {y}},s) \in \mathbf {\Omega }_{t}\) and any \(\delta >0\) it holds
so that, using the definition of \(U_{1}(r)\) again
from which we deduce that \(U_1(t+h)f({\mathbf {x}})=U_1(t)U_0(h)f({\mathbf {x}})\) for almost any \({\mathbf {x}}\in \mathbf {\Omega }_{t}.\)With the notations of the previous proof, one has from (A.4) that
This proves the result, since \(U_{1}(t)f\) vanishes on \(\mathbf {\Omega }_{t+h}{{\setminus }}\mathbf {\Omega }_{t}\) while \(U_{0}(t)f\) vanishes on \(\mathbf {\Omega }_{t}.\) \(\square \)
One has now the following
Lemma A.5
For any \(f \in {\mathscr {D}}_{0}\), the mapping \(t \geqslant 0 \mapsto U_{1}(t)f \in X\) is differentiable with \(\frac{\mathrm {d}}{\mathrm {d}t}U_{1}(t)f=U_{1}(t){\mathcal {T}}_{\mathrm {max},\,p}f\) for any \(t \geqslant 0.\)
Proof
In virtue of the previous Lemma, it is enough to prove that \(t \geqslant 0 \mapsto U_{1}(t)f \in X\) is differentiable at \(t=0\) with
Consider \(t >0\). One has
Now, since \(f \in {\mathscr {D}}({\mathcal {T}}_{0,\,p})\), one has from (3.9)
so that
Using [5, Proposition 2.2], one has
so that (A.5) yields
This proves the result.\(\square \)
Lemma A.6
For any \(f \in {\mathscr {D}}_{0}\) and any \(t >0\), one has \(U_{1}(t)f \in {\mathscr {D}}({\mathcal {T}}_{\mathrm {max},\,p})\) with \({\mathcal {T}}_{\mathrm {max},\,p}U_{1}(t)f=U_{1}(t){\mathcal {T}}_{\mathrm {max},\,p}f\).
Proof
The proof follows from a simple application of Proposition A.2 where the assumptions (i)–(iii) are met thanks to the previous Lemmas. \(\square \)
Let us now establish the following
Lemma A.7
For any \(f \in X\) and any \(t >0\), one has \({\mathcal {I}}^{1}_{t}[f]:=\int _{0}^{t}U_{1}(s)f \mathrm {d}s \in {\mathscr {D}}({\mathcal {T}}_{\mathrm {max},\,p})\) with
and \({\mathsf {B}}^{\pm } {\mathcal {I}}_{t}^{1}[f] \mathrm {d}s \in L^{p}_{\pm }\),
Moreover the mappings \(t \geqslant 0\mapsto {\mathsf {B}}^{\pm } {\mathcal {I}}_{t}^{1}[f] \mathrm {d}s \in L^{p}_{\pm }\) are continuous. Finally, for any \(f \in {\mathscr {D}}_{0}\) and any \(t \geqslant 0\), the traces \({\mathsf {B}}^{\pm }U_{1}(t)f \in L^{p}_{\pm }\) and the mappings \(t \geqslant 0\mapsto {\mathsf {B}}^{\pm }U_{1}(t)f \in L^{p}_{\pm }\) are continuous.
Proof
Thanks to the previous Lemmas, the family \((U_{1}(t))_{t\geqslant 0}\) satisfies assumptions (i)–(ii) of Proposition A.1. One deduces then from the same Proposition (point (1)) that, for any \(f \in X\) and any \(t >0\), \({\mathcal {I}}^{1}_{t}[f] \in {\mathscr {D}}({\mathcal {T}}_{\mathrm {max},\,p})\) with \({\mathcal {T}}_{\mathrm {max},\,p}{\mathcal {I}}_{t}^{1}[f]=U_{1}(t)f-U_{1}(0)f=U_{1}(t)f.\)
To show that \({\mathsf {B}}^{-}{\mathcal {I}}_{t}^{1}[f] \) can be expressed through formula (A.6) we first suppose \(f \in {\mathscr {D}}_{0}\). For such an f both \(U_{0}(t)f \) and \(U_{1}(t)f\) belong to \({\mathscr {D}}({\mathcal {T}}_{\mathrm {max},\,p})\) for any \(t\geqslant 0\) with \({\mathsf {B}}^{-}U_{1}(t)f=H{\mathsf {B}}^{+}U_{0}(t)f \in L^p_-\). Using this equality, the continuity of H and Proposition A.1 (point 2) applied both to \((U_{1}(t))_{t\geqslant 0}\) and \((U_{0}(t))_{t\geqslant 0}\) one gets
i.e., (A.6) for \(f \in {\mathscr {D}}_{0}.\)
Consider now \(f \in X\) and let \((f_{n})_{n} \in {\mathscr {D}}_{0}\) be such that \(\lim _{n}\Vert f_{n}-f\Vert _{p}=0.\) According to Eq. (3.8), the sequence \(({\mathsf {B}}^{+}{\mathcal {I}}_{t}^{0}[f_{n}])_{n}\) converges in \(L^p_+\) towards \({\mathsf {B}}^{+}{\mathcal {I}}_{t}^{0}[f]\). Since (A.6) holds true for \(f_{n}\), and H is continuous, then the sequence \(({\mathsf {B}}^{-}{\mathcal {I}}_{t}^{1}[f_{n}])_{n}\) converges in \(L^p_-\) to \(H{\mathsf {B}}^{+}{\mathcal {I}}_{t}^{0}[f].\) One deduces from this that \({\mathsf {B}}^{-}{\mathcal {I}}_{t}^{1}[f] \in L^p_-\) with (A.6).
Moreover the mapping \(t\geqslant 0 \mapsto {\mathsf {B}}^{-}I_{t}^{1}[f] \in L^p_-\) is continuous since both H and the mapping \(t \geqslant 0\mapsto {\mathsf {B}}^{+}{\mathcal {I}}_{t}^{0}[f] \in L^p_+\) are continuous (see Proposition 3.3). This property and Proposition A.1 imply that \( {\mathsf {B}}^{+}{\mathcal {I}}_{t}^{1}[f] \in L^p_+\) and that the mapping \(t \mapsto {\mathsf {B}}^{+}{\mathcal {I}}_{t}^{1}[f] \in L^p_+\) is continuous too.
Finally observe that, if \(f \in {\mathscr {D}}_{0}\), then \(f \in {\mathscr {D}}({\mathcal {T}}_{\mathrm {max},\,p})\) and for any \(t \geqslant 0\) one has
Thus one can state that for any \(f \in {\mathscr {D}}_{0}\) and any \(t \geqslant 0\), the traces \({\mathsf {B}}^{\pm }U_{1}(t)f \in L^{p}_{\pm }\) and the mappings \(t \geqslant 0\mapsto {\mathsf {B}}^{\pm }U_{1}(t)f \in L^{p}_{\pm }\) are continuous. \(\square \)
Let us now investigate Property (7):
Lemma A.8
One has
Proof
Given \(f \in {\mathscr {D}}_{0}\), for any \(s >0\) and \(\mu _{+}\)-a.e. \({\mathbf {z}}\in \Gamma _{+}\):
Thus,
Now, using Fubini’s Theorem and, for a given \({\mathbf {z}}\in \Gamma _{+}\), the change of variable \(s \mapsto s-\tau _{-}({\mathbf {z}})\), we get
Using Fubini’s Theorem again
where we used [5, Eq. (2.5) in Prop. 2.2]. Therefore, it is easy to check that
which is the desired result. \(\square \)
We finally have the following
Lemma A.9
Given \(\lambda >0\) and \(f \in X\), set \(F_{1}=\int _{0}^{\infty }\exp (-\lambda t)U_{1}(t)f\mathrm {d}t.\) Then \(F_{1} \in {\mathscr {D}}({\mathcal {T}}_{\mathrm {max},\,p})\) with \({\mathcal {T}}_{\mathrm {max},\,p}F_{1}=\lambda \,F_{1}\) and \({\mathsf {B}}^{\pm }F_{1} \in L^{p}_{\pm }\) with
Proof
Let us first assume \(f \in {\mathscr {D}}_{0}\). Then, for any \({\mathbf {y}}\in \Gamma _{-}\), \(s \in (0,\tau _{+}({\mathbf {y}}))\):
i.e. \(F_{1}(\Phi ({\mathbf {y}},s))=\exp (-\lambda s)\left[ H{\mathsf {B}}^{+}C_{\lambda }f\right] ({\mathbf {y}}).\) This exactly means that \(F_{1}=\Xi _{\lambda }HG_{\lambda }f\). By a density argument, this still holds for \(f \in X\) and we get the desired result easily using the properties of \(\Xi _{\lambda }\) and \(G_{\lambda }.\) \(\square \)
The above lemmas prove that the conclusion of Theorem 3.8 is true for \(k=1.\) One proves then by induction that the conclusion is true for any \(k \geqslant 1\) exactly as above. Details are left to the reader.
Rights and permissions
About this article
Cite this article
Arlotti, L., Lods, B. An \(L^{p}\)-Approach to the Well-Posedness of Transport Equations Associated to a Regular Field: Part II. Mediterr. J. Math. 16, 145 (2019). https://doi.org/10.1007/s00009-019-1426-7
Received:
Revised:
Accepted:
Published:
DOI: https://doi.org/10.1007/s00009-019-1426-7