No label defined (Q4699480)

Implicit ordinary differential equations (*) \( 0=F(t,p,x,\dot x)\) are investigated, with \(F: U\to\mathbb{R}^n\), \(U\subseteq\mathbb{R}\times\mathbb{R}^k\times\mathbb{R}^n\times\mathbb{R}^n\) open, \(k\in\mathbb{N}\cup\{0\}\), \(n\in\mathbb{N}\), and \(p\) is a parameter. Solutions are assumed to be \(C^1\) and defined on open intervals. The work contains an Introduction (11 pages), a list of symbols (1 page), a list of references (8 pages), and three chapters: 1. Generalized vector fields and implicit differential equations (29 pages), 2. Generic properties of linear active networks (36 pages), 3. Singular and impasse points (48 pages). In the Introduction, it is shown by means of examples that the behavior of solutions to implicit equations is very different from that of explicit ones. The failure of the circuit simulation program PSPICE is demonstrated in two cases. In Chapter 1, generalized vector fields are defined to be pairs \((M,\Theta)\), \(M\) being a Banach manifold and \(\Theta\subseteq TM\), a \(C^1\)-mapping \(x\) defined on some open interval is a solution of \((M,\Theta)\) if \(Tx(t)\in\Theta\) for all \(t\). Following ideas of \textit{W. C. Rheinboldt} [Math. Comput. 43, 473-482 (1984; Zbl 0581.65058)], \textit{A. Szatkowski} [Int. J. Syst. Sci. 23, No. 2, 167-186 (1992; Zbl 0747.34003)] and \textit{S. Reich} [Circuits Syst. Signal Process. 9, No. 4, 367-382 (1990; Zbl 0719.34114)], a vector field equivalent to \((M,\Theta)\) is constructed by means of an iterative procedure, the number of iterations of which is called the geometric index, a term carrying over to (*) by considering \((U,(d(\text{id}_U))^{-1}F^{-1}(0))\). (\(d\varphi\) is the bundle chart associated to the chart \(\varphi\).) For semi-implicit differential-algebraic equations, i.e., under the assumptions that \(F\) is \(C^1\), \(F(x,y)=(F_1(x,y),F_2(x))\), and \(D_2F_1\) is surjective on \(U\), conditions for the geometric index of (*) to be \(0\) and \(1\) and an index reduction method are given, and it is shown that the results remain valid if the mapping \(F\) is assumed to act between real Banach spaces. It is further shown that the geometric index of linear DAEs with constant coefficients equals the index of the pencil of their coefficient matrices. Chapter 2 concerns linear electrical networks that may contain ideal transformers (T), nullors (N), independent (i.S.) and controlled sources, resistors (R), inductors (L), capacitors (C), and gyrators (G). These networks depend on some parameter \(p\in\mathbb{R}^k\) representing resistances, capacitances etc.; the work is devoted to properties of the network equations \[ A(p)\dot x+B(p)x=f(t)\tag{ce} \] that are generic w.r.t. \(p\), where \(A(p)\) and \(B(p)\) are square matrices and \(x\) consists of all branch voltages and branch currents. The equivalence of solvability criteria based on matroids [\textit{A. Recski}, Matroid theory and its applications in electric network theory and in statics. Algorithms and Combinatorics, 6. Berlin etc.: Springer-Verlag (1989; Zbl 0723.05023)] and those based on pairs of conjugate trees [\textit{M. Fosséprez}, Nonlinear circuits: Qualitative analysis of nonlinear, nonreciprocal circuits. John Wiley \& Sons (1992)] is demonstrated, and these criteria are generalized. A method to obtain complete sets of state coordinates is established, which includes the following extension of \textit{P. R. Bryant} [Proc. IEE (GB), Part C, 106, 174-188 (1959)]: The voltages across the forest capacitors together with the currents through the co-forest inductors in any normal pair of conjugate trees (a term introduced in this work) may be given initial values independently from each other. For networks containing i.S.s, Ns, Rs, Cs, and Ls, a method of augmentation is proposed that yields network equations of index \(1\) and requires a minimum number of additional network elements. All criteria may be checked by algorithms polynomial in the number of network branches. For networks containing i.S.s, Rs, Cs, Ls, Ts and Gs, it is shown that the index of (ce) is either \(1\) or \(2\), depending only on the network graph. Chapter 3 investigates the quasi-linear equation \[ 0=A(x)\dot x-g(x),\tag{ql} \] with \(A: U\to{\mathcal L}(\mathbb{R}^n,\mathbb{R}^n)\), \(g: U\to\mathbb{R}^n\), \(U\subseteq\mathbb{R}^n\) open, \(x_0\in U\), and \(n>0\). Sufficient conditions are given for (ql) (i) to be \(C^s\)-conjugate in some neighborhood of \(x_0\) to \[ 0=x_1^r \dot x_1\pm 1,\quad 0=\dot x_2 = \dots =\dot x_n,\tag{nl} \] in some neighborhood of \(0\) (\(s,r >0\)), and (ii) to have (exactly) two solutions that converge to, but are not continuable beyond \(x_0\), generalizing results of \textit{P. J. Rabier} [J. Math. Anal. Appl. 144, No. 2, 425-449 (1989; Zbl 0683.34017)], \textit{M. Medved} [Riv. Mat. Pura Appl. 10, 95-107 (1992; Zbl 0766.58050)], \textit{P. J. Rabier} and \textit{W. C. Rheinboldt} [J. Math. Anal. Appl. 181, No. 2, 429-454 (1994; Zbl 0803.34004)] and \textit{G. Reißig} [IEEE Trans. Circuits and Systems Part I, 43(2), 122-133 (1996)]. (In the semi-explicit case \(A(x)=\left(\begin{smallmatrix}\text{id}&0\\ 0&0\end{smallmatrix}\right)\), slightly generalized solutions are considered: the first components are assumed to be \(C^1\), and the remaining continuous.) Counterexamples to the main theorem of \textit{L. O. Chua} and \textit{A.-C. Deng} [Int. J. Circuit Theory Appl. 17, No. 2, 213-235 (1989; Zbl 0676.94022)] are given.