A problem of integral geometry for a family of curves with incomplete data (Q906084)

Given a bounded open set \(G\) of \(\mathbb R^n\), for each fixed \(x\in G\) and \(\omega\in \Omega\), where \(\Omega\) is the unit sphere of \(\mathbb R^n\), the author considers the set of curves \(L(x,\omega)=\{y: y=\phi(x,\omega,t)\}\), of the form \(y=\phi(x,\omega,t)=x+t\omega +t^2\alpha(x,\omega,t)\) for \((x,\omega, t)\in T\), meaning \(x\in G\) (initial value), \(\omega\in \Omega\) (initial velocity), \(t\in (0, l(x,\omega))\), \(\alpha\) is assumed to be of class \(C^2\) on \(T\). For each \(x\) the Jacobian of the map \(\phi(x, \cdot,\cdot)\) maps \(\Omega\times (0, l(x,\omega))\) into \(G_x=G\backslash \{x\}\), vanishes only at \(t=0\) and is bounded away from zero outside any neighborhood of \(x\). It is assumed that for some system of disjoint open sets \(G_i\subset G\), \(i=1, \dots, p\), with piecewise \(C^2\) boundary \(\partial G_i\) the union \(G_0=\cup_i G_i\), satisfies \(\bar{G}_0=\bar{G}\) and a convexity type property w.r.t. the curves \(L(x, \omega)\), namely that for any \((x,\omega)\in G_0\times \Omega\), the curve \(L(x,\omega)\) intersects \(\partial G_0\) in at most countably many points. It is also considered the class \(\mathbb K\) of bounded functions \(g:G\times G\times \Omega\to \mathbb R\), such that \(g\in C^1(G\times G_i\times \Omega)\), \( \forall i=1,\dots, p\), and \(g(x,y,\omega)\) is Lipschitz on \( y\in G_i\), and so we may define its extension at \( z\in \partial G_i\) denoted by \([g(x,z,\omega)]_i\). Given a point \(z\in \partial G_0\) that is a boundary point for only two sets \(G_j\) and \(G_l\) (i.e., a contact point of \(\partial G_0\)), the difference \([g(x,z,\omega)]_{jl}:=[g(x,z,\omega)]_j-[g(x,z,\omega)]_l\) is called the size of jump or discontinuity of \(g\) at \((x,z,w)\). Given a set of orthogonal transformations \(A_1, \dots ,A_N\) of \(\mathbb{R}^n\), consider the continuous function \(H\) defined on \(G_0\times \Omega\) by line integrals of first kind \[ H(x,\omega) =\sum_{i=1}^{N} \int_{L(x,A_i\omega)}g_i(x,y,A_i\omega)d_y\sigma \] where \(g_i(x,y,\omega)\in \mathbb{K}\), and extended to \(G\times \Omega\) as a bounded function, and extended to boundary points \(\partial G_0, \times \Omega\) by its upper limits. The author considers the integral geometry problem of finding a hypersurface \(\partial G_0\), assuming that a given function \(H\) defined \(G\times \Omega\) satisfies an equation as above for some \(g_i\) and \(A_i\). In particular the set of all contact points \(y\) of the family \(g_i(x,y,\omega)\) must be found, but the unknowns which are not to be determined are the integrands \(g_i\), the transformations \(A_i\) and the curves \(L(x,\omega)\). He explains how useful this kind of problem can be in applied science, and the fact that the integration along curves \(L(x, A_i\omega)\) with \(x\) an interior point includes more traditional settings where \(x\) is an initial point, after reformulation of the integral equations. Multiplying both sides of the equation by an arbitrary test function \(\beta(x,\omega) \in C^2(G\times \Omega)\) and integrating over \(\Omega\), the author obtains in Theorem 1 a gradient estimate for \(\Phi(x)=\int_{\Omega}\beta(x,\omega)H(x,\omega)d\omega\) in terms of a function \(M(z)=\int_{\Omega_z}\sum_i |q_i(z,z,\omega)|_{j,l}d\omega\), for a contact point \(z\in \partial G_j\cap \partial G_l\), \(\Omega_z\) the set of points of \(\Omega\) orthogonal to the inward unit normal \(n_j(z)\) to \(\partial G_j\) at \(z\), and \(q_i(x,y,\omega)=\beta(x,A^{-1}_i\omega)g_i(x,y,\omega)\), concluding that if \(M(z)\neq 0\) then \(|\nabla \Phi(x)|\) is unbounded only near \(\partial G_0\). This leads to uniqueness for the solution of the boundary problem, namely if \(G^k_j\), \(A^k_i\), \(L^k(x,\omega)\), \(g_i^k(x,y,\omega)\) are two systems \(k=1,2\), \(j=1,\dots, p_k\), \(i=1, \dots, N_k\), that solve the equation for the same function \(H\) on \(G\times \Omega\), then \(\partial G_0^1=\partial G_0^2\). The author claims that some of the assumptions can be weakened and still lead to the same conclusion.