Proceedings Article | 26 May 1995
KEYWORDS: Photomasks, Critical dimension metrology, Semiconducting wafers, Photoresist processing, Image processing, Lithography, Process control, Deep ultraviolet, Printing, Scanning electron microscopy
In this paper, wafer level experimental results on the characteristics of a low-k1 submicron lithography processes using binary mask and deep ultraviolet (DUV) wavelength (248 nm) printing tool will be discussed. In the study, we investigated the nonlinear response of wafer resist critical dimension (CD) at different pitches to the mask CD. The CD nonlinearity is due to the combination of relatively poor optical aerial image quality at low-k1 imaging and resist processes. Operating a lithographic process at this nonlinear region presents an issue since only limited process windows are available. In addition, it requires both proximity correction for larger features on the mask and tighter mask CD control. The later requirement, in practice, is rather difficult to achieve with current mask technology. In order to avoid operating low-k1 lithography processes at CD nonlinear region, improvement of aerial image quality is obviously needed. Although aerial image quality improvement can be achieved with different advanced lithographic enhancement techniques such as phase shift masks and off- axis illumination, there are different process issues related to each of the techniques as well. With conventional illumination source and binary mask, one of the option left to enhance aerial image quality is mask biasing. Our experimental results show that by positively biasing the mask, i.e., mask chrome CD line dimension is larger than targeted resist CD, CD nonlinearity for low-k1 processes can be improved. For low- k1 processes, the isofocal bias of different feature sizes (different k1 factor) was also found to be different. A strong dependence of isofocal bias on pitch was observed. Again, mask biasing can help to shift the isofocal CD of critical dimensions to the target CD. However, in our study, we found that for certain resist, opposite mask biases are required to achieve CD linear response and zero isofocal bias. In this case, tradeoffs between mask CD control, proximity effect, and process window need to be considered.
