1953 年 19 巻 87 号 p. 120-123
As several types of difference-method hitherto employed in solving partial defferential equations are based on numerical differentiation, the inherent error in the solution is more or less large even if we adopt higher differences, and consequently, in order to get the higher accuracy, it is necessary to have many lattice points in the given domain, and we are obliged to have a considerable amount of labour for caluculation. In the present paper, in order to get rid of such inconvenience, the author has proposed a new method based on numerical integration, and applying it to the heading problem, he made it clear that far more precise solution can be obtained in spite of a coarse network.