1933 ESCHRODINGER infinitesimal, Fermat's principle in these circumstances demands that the light ray should curve earthward (see Fig. 2), so that it remains a little longer in the higher " faster"layers and reaches its destination more quickly than by the shorter straight path(broken line in the figure; disregard the square, Fig. 2. wWWW for the time being). I think, hardly any of you will have failed to observe that the sun when it is deep on the horizon appears to be not circular but flattened: its vertical diameter looks to be shortened this is a result of the curvature of the rays. According to the wave theory of light, the light rays, strictly speaking, ave only fictitious significance. They are not the physical paths of some particles of light, but are a mathematical device, the so-called orthogonal trajectories of wave surfaces, imaginary guide lines as it were, which point in the direction normal to the wave surface in which the latter advances(cf s.g. 3 which shows the simplest case of concentric spherical wave surfaces and accordingly rectilinear rays, whereas Fig 4 illustrates the case of curved Fig. 4306 1933 E.SCHRÖDINGER infinitesimal, Fermat’s principle in these circumstances demands that the light ray should curve earthward (see Fig. 2), so that it remains a little longer in the higher "faster" layers and reaches its destination more quickly than by the shorter straight path (broken line in the figure; disregard the square, Fig. 2. WWW1 W1 for the time being). I think, hardly any of you will have failed to observe that the sun when it is deep on the horizon appears to be not circular but flattened: its vertical diameter looks to be shortened. This is a result of the curvature of the rays. According to the wave theory of light, the light rays, strictly speaking, have only fictitious significance. They are not the physical paths of some particles of light, but are a mathematical device, the so-called orthogonal trajectories of wave surfaces, imaginary guide lines as it were, which point in the direction normal to the wave surface in which the latter advances (cf. Fig. 3 which shows the simplest case of concentric spherical wave surfaces and accordingly rectilinear rays, whereas Fig. 4 illustrates the case of curved Fig. 3. Fig. 4