Kioxia Corporation

Toshiba Memory Corporation

Tetsuaki Matsunawa

Taiki Kimura

Yuki Watanabe

According to one embodiment, a mask design method includes the following configuration. The method includes setting evaluation points on a circuit pattern created based on design data, setting a parameter that defines a shape of the mask pattern on the mask pattern corresponding to the circuit pattern and calculating an optical image intensity on the evaluation points set on the circuit pattern based on the mask pattern and the parameter. The method includes calculating an evaluation value relating to the optical image intensity by an objective function based on the optical image intensity and optimizing the parameter using a value of a partial differential for the parameter in the objective function to minimize the evaluation value calculated by the objective function.

Mask design method and storage medium thereof

According to the present embodiment, the pattern generation device includes a misalignment value calculation unit configured to acquire a layout information, calculate a layout function from the layout information, and calculate a misalignment value by a convolution of the layout function and an integral kernel having a predetermined parameter, and a pattern correction unit configured to correct a pattern to generate a modified layout information using a calculated result by the misalignment value calculation unit, and output the modified layout information.

Pattern generation device, pattern generation method, and method of manufacturing semiconductor device

A non-transitory computer readable storage medium according to an embodiment stores a mask evaluation program evaluating a mask used to manufacture an integrated circuit device. The program causes a computer to realize a convolutional neural network. The convolutional neural network output a calculated value of second data when first data is input. The first data corresponds to a circuit pattern of the mask. The second data corresponds to a pattern formed by the mask. The convolutional neural network has a filter and a weighting coefficient learned to reduce an error of the calculated value and an actual measured value of the second data by using the first data and the actual measured value of the second data.

Growing community of inventors

Fujisawa, Japan

Tetsuaki Matsunawa