Method for setting and adjusting process parameters to maintain acceptable critical dimensions across each die of mass-produced semiconductor wafers

A method for partially bypassing process time of a slow-to-evaluate simulator comprising the steps of:

• (a) providing drive means for driving the simulator through a regular gridwork of process parameter permutations;
• (b) storing the results of the driven simulations in memory;
• (c) generating a surface model of the stored results;
• (d) using the surface model in place of the slow-to-evaluate simulator for generating simulation results for afterwards-supplied process parameter inputs that are Substantially near or within the region of the regular gridwork defined by the drive means;
• and
• (e) using an interpolator to correct errors between the data stored in the driven results region of said memory and the generated output of said use of the surface model.

A machine-implemented method for simulating a given mass production process comprising the steps of:

• (a) coupling the output of an uncalibrated process simulator to a post-simulation calibrator;
• (b) supplying actual process parameter data to the process simulator and storing the simulated results as nontransformed simulation results;
• (c) comparing the stored simulated results against actual measured results and responsively modifying the post-simulation calibrator such that, for the stored simulation results, the post-simulation calibrator generates transformed simulation results more in agreement with the actual results than are the nontransformed simulation results;
• and
• (d) after calibration, coupling the uncalibrated process simulator to a simulated process parameter generator that supplies simulated process parameters and using the output of the modified post-simulation calibrator as representative of the performance of the actual production line given the simulated process parameters supplied by the generator.

A machine-implemented method for simulating a given mass production process comprising the steps of:

• (a) coupling the output of an uncalibrated process simulator to a post-simulation calibrator;
• (b) supplying actual process parameter data to the process simulator and storing the simulated results;
• (c) comparing the stored simulated results against actual measured results and responsively modifying the post-simulation calibrator such that, for the stored simulated results, the calibrator generates transformed simulation results more in agreement with the actual results;
• and
• (d) after calibration, coupling the process simulator to a simulated process parameter generator that generates simulated process parameters and using the output of the calibrated transform module as representative of the performance of the actual production line given the simulated process parameters supplied by the generator;
• wherein the process parameter generator is a noisy process parameter generator.

A method for determining effects of different production variances and means on statistics of resulting dimensions of critically-dimensioned features that repeat cross-reticle in mass-produced integrated circuits fabricated on a given actual production line,

• said method using in-scribe measurement data obtained from in-scribe features of wafers of the mass-produced integrated circuits, and said method further comprising the steps of:
• (a) creating a machine-implemented, noise-perturbed, process modeler that includes a plurality of noise-perturbed model parameters each having a respective, simulated mean and simulated variance, where the created process modeler produces simulated in-scribe results which statistically closely match the obtained in-scribe measurement data of the actual production line and where the created process modeler is modifiable to produce simulated cross-reticle results;
• (b) increasing the simulated variance of one or more noise-perturbed model parameters that correspond to cross-reticle physical attributes;
• (c) running statistical simulations on the otherwise-matched process model using the increased one or more variance values;
• and
• (d) determining from the simulated runs which one or more of the increased process parameter variances contributes most to increasing the simulated cross-reticle variance of dimensions of one or more of the critically-dimensioned features.

A machine-implemented, bootstrapping simulation method that starts with a slow-to-evaluate simulator, said method comprising the steps of:

• (a) running the slow-to-evaluate simulator through a first set of input parameters;
• (b) recording the slowly-simulated results for the first set of input parameters;
• (c) using the recorded slowly-simulated results to define a fast-to-evaluate simulator that generates results substantially equal to said recorded slowly-simulated results in less time than does the slow-to-evaluate simulator when given said first set of input parameters;
• and
• (d) using the defined fast-to-evaluate simulator in place of the slow-to-evaluate simulator when presented with a second set of input parameters that are within or substantially close to an input space defined by said first set of input parameters.

A machine-implemented method for use with a wafer processing line having a wafer coating station, a lithographic radiation exposing station receiving wafers from the coating station, a wafer developing station receiving wafers from the exposing station, and a wafer etching station receiving wafers from the developing station, wherein at least one of said coating, exposing, developing and etching stations has a plurality of controllable process parameters;

• said method comprising the steps of:
• (a) sampling measurements of in-scribe features of wafers passing through said processing line;
• (b) developing a first statistical profile of the sampled in-scribe measurements;
• (c) defining a second statistical profile for acceptable dimensions of in-die features of the wafers passing through said processing line;
• (d) extrapolating from the developed first statistical profile of the sampled in-scribe measurements, a third statistical profile of predicted dimensions of the in-die features of the wafers passing through said processing line;
• (e) testing the predicted third statistical profile for compliance with the acceptable second statistical profile;
• and
• (f) in cases where said third statistical profile is not in compliance with said second statistical profile, automatically selecting at least one controllable parameter of said stations of the wafer processing line for adjustment, said adjustment being one that simulation-wise brings the third statistical profile of later-processed wafers into compliance with the second statistical profile.

A method for determining effects of different production variances and means on statistics of resulting dimensions of critically-dimensioned features that repeat across mass-produced integrated circuit devices fabricated on a given actual production line,

• said method using sample measurement data obtained from critically-dimensioned features appearing in first areas of the mass-produced devices, and said method further comprising the steps of:
• (a) creating a machine-implemented, noise-perturbed, process modeler that includes a plurality of noise-perturbed model parameters each having a respective simulated mean and a respective simulated variance, where the created process modeler produces simulated sample data results which statistically closely match the obtained sample measurement data of the actual production line and where the created process modeler is modifiable to produce simulated measurement results for critically-dimensioned features of the devices in areas other than in the sampled first areas;
• (b) changing the simulated variance of one or more noise-perturbed model parameters that correspond to physical attributes in second areas of the devices, the second areas being different than the sampled first areas
• (c) running statistical simulations on the otherwise-matched process model using the changed one or more variance values;
• and
• (d) determining from the simulated runs which one or more of the changed process parameter variances contributes most significantly, if at all, to changing the simulated variance of dimensions in the second areas of one or more of the critically-dimensioned features.