A Genetic Algorithm-Enhanced Sensor Marks Selection Algorithm for Wavefront Aberration Modeling in Extreme-UV (EUV) Photolithography

Round 1

Reviewer 1 Report (Previous Reviewer 1)

The article has largely been re-written. The non-clear parts of the experimental deaign have been thoroughly described. I therefore suggest the publication of this paper.

The language of the manuscript is fine, only several typos can be found.

Author Response

Thank you very much for your approval and for the previous comments.

Reviewer 2 Report (Previous Reviewer 2)

Genetic Algorithms Enhanced Sensor Marks Selection Algorithm For Advanced Photolithography Process

Aris Magklaras, Panayiotis Alefragis, Christos Gogos, Christos Valouxis and Alexios Birbas

 

Revised version

The article's main contribution has to be the GA solving the problem, as mentioned in the article's title. Section 3.2 tries to give explanations concerning the GA used to solve the treated problem. However, there still are some concerns.

 

1. The authors do not give solutions' encoding. What does a solution of the GA represent? What is the chromosome structure? The authors speak about solutions, but an explicit description is not given.

 

2. In line #490, it is written

"In our case, for ensuring that we have 221 points in the final solution, the fitness 490

function first checks if the sum of the elements in the solution is equal to 221...."

 

So, a solution might have more or less 221 "elements" ??!!. The reader's suppositions concerning the solution structure are not admissible. Hence, THE AUTHORS MUST GIVE THE SOLUTION STRUCTURE.

 

3. Equation (1) giving the fitness function has an unusual (and inadmissible) form. This function must have arguments leading us to a solution, which the GA considers. This form of equation (1) is also related to remark (1)

 

4. How a chromosome's mutation is achieved? The resulting solution needs to be repaired, and how? (In a few words).

 

5. The objective of Section 3.2 may not be to introduce the reader to the GAs. That is why this section has an excessive extension, and lines #420 – 429 (including the classic Fig. 9) must be erased.

 

6. I maintain my remark concerning the experimental results presentation. For example, the authors give statistics concerning the A, D, and G- optimality, but statistics of the aggregate value of the GA's fitness function are not presented.

 

Please keep in mind that even if your problem is a real-life industry problem, the results of the GA must be presented, being your manuscript's main topic.

All my remarks have intended to help you to achieve a publishable manuscript.

 

 

Marginal remarks

7. In line #203, the last column PHIk(x,y) could have errors concerning the indexes of y. The index could not vary from 1 to N.

 

8. Within Figure 6, the basis functions are denoted with small PHI (phi) different from equation #203.

 

Conclusion

Some of the drawbacks revealed in my first review have been eliminated. However, the algorithm and the experimental results are not yet properly presented. So, it can not be published in this version, but I encourage the authors to revise the abovementioned concerns.

 

 

Author Response

Dear reviewer, 

 

Thank you very much for your comments. It is very much appreciated that you put so much effort in reviewing our document and eventually improving our paper. You can find the response to your comments in the attached file below. 

Wkr,

Aris

Author Response File: Author Response.pdf

Round 2

Reviewer 2 Report (Previous Reviewer 2)

The article's main contribution has to be the GA solving the problem, as mentioned in the article's title. In my previous review, I mentioned that the GA is not presented appropriately, and the definitory elements of the proposed GA are missing.

Because

1) the authors gave a minimum of details concerning the solutions' encoding, objective function and genetic operators,

2) the authors added a minimum of information concerning the fitness function in the results section,

 I consider that the MANUSCRIPT CAN BE PUBLISHED in this third version.

This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.

Round 1

Reviewer 1 Report

Before the article is published several key points need to by clarified to make the article more understandable to the reader.

If I undestand correctly the Authors first select a pool of points using their PDUMS algorithms, which are candidates for being selected for measurement. Then a subset of these points is selected by consecutively picking a point that (l. 177) "contributes most to the compound criterion". This is not clear - on what basis is this point selected, what exactly is calculated to determine that a given point taken from a pool contributes most? In l. 147 the Authors state that  "The only information that is available (...) in the matrix C". How can this matrix be available without comparing the model results to measurements? Or are they compared before determining C? How is the matrix C determined? This is a crucial point which must be clarified, as this matrix is the basis for calculating the G, D, A optimality criteria.  Moreover, it is not clear how G, D and A optimality criteria determine how much information a given selection of points will provide about the quality of the printed pattern.

Concerning the genetic algorithm, the text does not make it clear what exactly it optimises. The positions (x,y) of individual points or the selection of points from the pool prepared by PDUMS? Also, the details of the GA are completely omitted, such as representation of individuals in the genome, the genome type (binary, real number?), population size, selection scheme (roulette wheel, tournament, other?), cross-over type, fitness function (here the reader only learns that all 3 criteria are combined in the fitness function, but the formula is not provided), cross-over and mutation probabilities.

Fig. 7: the points are arranged on a square lattice, which is strange considering that they were generated using the Poisson disc scheme.

Furthermore, please explain why in Fig. 5 the excluded volumes are smaller for points at the wafer rim as compared to those closer to its center.

Several language mistakes and typos should also by corrected, such as:

 - l. 60: is selected then we

 - l. 64: in this paper

 - l. 81 & 85: Section X deals with

 - l. 93: is how to select

 - l. 188: "r distances apart" sounds incorrect

 - l. 230 & 247: the a population in a GA consists of individuals and not "people"

 - l. 249: worldwide --> global

Author Response

Please see the attachement.

Author Response File: Author Response.pdf

Reviewer 2 Report

Comments for author File: Comments.pdf

Author Response

Please see attachement.

Author Response File: Author Response.pdf