|
A mask making process is the result of data preparation (e.g., bias), patterning and etch. The effect of these steps can be simulated starting from a circuit design. The purpose of this paper is to determine the transfer function for a production flow of a mask, describing the transition from the design to the finished reticle. The simulation results will be compared with actual measurements. 1D measurements of a mask, as shown for example in Ref.[1], provide a lot of useful information. The fingerprint that a writing tool leaves on the experimental result actually makes it possible to determine if the reticle is produced with an e-beam writer or an optical laser tool. From the deviations from the nominal value the "effective" spot size of the laser tool can be determined. In case an e-beam writer is used, the performance of the e-beam proximity correction in use becomes apparent. The mean-to-target offset can be resolved if no data biasing is performed. From SEM pitures of the reticle the 2D corner rounding effects can be determined from the combined influence of the spot size, scattering (in case of an e-beam writer) and data biasing. Global sizing can be resolved from 2D measurements. If the pattern bias is performed for each individual feature it is included in the rounding. Sufficient data about sizing, etch bias and spot size can be obtained if SEM pictures are taken from several critical sites on the reticle. For example, dense lines/spaces or contacts/dots constitute the minimum objects to be printed. Recent reticles ordered at IMEC include a monitor structure. Among other features, this performance monitor contains dots and contact holes with sizes between 90 nm and 250 nm with additional serifs at the corners. The serifs have sizes of 10 - 40 nm in 10 nm steps. All feature sizes are specified at 1x unless otherwise indicated. They are extending multiples of 10 nm from the corner position. In our work we assessed the integrity of varying serifs on dots and contact holes of 120 nm to estimate the "effective" spot size and etch bias of a reticle. The contact holes are simulated with a spot size of 80 nm and an etch bias of 70 nm. SEM pictures are in agreement with the etch bias as seen from 1D results. One might consider that another blurring function and bias condition will give the same result. A case study shows that by interchanging these parameters the resulting mask pattern yields different results. The simulations show that the SEM pictures of the 2D structures within the monitor pattern contain enough information to extract the transfer function. In some situations it might be impossible to measure 2D features (e.g., additional measurements at mask shop become too expensive; a pellicle already mounted impedes mask assessment at the wafer fab; too limited time in contrast with the lengthy procedure to extract feature contours from images; etc.). In such case 1D measurements obtained from the mask shop can already allow to estimate the etch bias. Also, the spot size can be determined from the printed minimum feature size.
|
