Simulation Study of Vertical p–n Junction Photodiodes’ Optical Performance According to CMOS Technology

Round 1

Reviewer 1 Report

Thank you for providing me the opportunity to contribute as a reviewer. Following are my thoughts regarding the manuscript titled "Simulation study of vertical p-n junction photodiodes’ optical performance according to CMOS technology" by G. Ferreira et al., as submitted to (ID 1608340).

In this work the authors investigate the effect of gate width and junction structure on the spectrophotometric characteristics of CMOS photodiodes, using a simulation software. A comparison of the proposed model’s results to the literature is also included, as well as a realization of a testing design for future fabrication. Therefore, and given the importance of CMOS technology and the completeness of the work, I consider that it is well within the scope of Applied Sciences and of interest to most readers of this journal.

The introductory section shows the appropriate background material, sufficient for someone not an expert in this area to understand the context and significance of the work, however a couple of details regarding CMOS comparison to CCD should be addressed in order for the article to reflect the state of the art in this field.

First, CCD superior fill factor (L48-49) is not true when compared to back-illuminated CMOS. Second, CMOS more susceptible to noise is not currently accurate, since active pixel sensor (APS) CMOS overcome this problem. Please consult for example [doi: 10.1109/TNS.2013.2276123] and [doi: 10.1109/ASSCC.2013.6690968]. Refs 7, 8 are rather old in this respect.

Table 2 - 0.7 μm line: Authors state they assumed the values of the four concentrations, for lack of reported ones; however, shouldn’t they follow a monotonic pattern versus gate width? Please explain.

Moreover, the following comments should be addressed as well:

Eq5: Please consider checking whether minus should be omitted

L139 “denoted by n₁ and n₂”: Please consider correcting to “n₁ and p₁”

Eq11: Please consider correcting G(X) to G(x), as well as defining x

L253: “e.g.”: Should be “i.e.”

Furthermore there exist disperse grammar/syntax errors of minor importance, e.g. L 124, 277, 287, 390-1, 398.

Finally, the References do not contain DOIs, even though they exist for most of them.

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 2 Report

The paper presents a detailed simulation analysis of p-n junction photodiodes for three CMOS technologies and three different junction structures.

The analysis is well described and discussed, with proper connections to existing experimental results reported in the literature.

I think the paper deserves publication, but I have one question that should be addressed. In the choice of acceptor and donor concentrations (page 6, table 2) for the 0.7 micron technological node the authors use the default values from COMSOL, which are slightly different from the above numbers chosen for the other two technological nodes. I think this is not justified. It would have been better to use the numbers of the 0.35 micron node, the closest one. I wonder if the lower performances obtained for the 0.7 micron devices could be due to the lower concentration values of acceptors and donors arbitrarily assumed for this technological node.

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 3 Report

Line 39 - seems like the Planck units should be J*s instead of J.s (the dot looks like a period instead of multiplication sign). I see this in all your equations as well. Make sure you are using the correct symbol for multiplication (the dot, not the period).

Figure 2 is very well done, nice work!

Lines 48/49 - would be nice to have a reference for the 'fill factor' statement

Lines 156-158 - Can you elaborate why you couldn't include the oxide layers in the sim? Ideally with reference(s). Or, is there some simple correction factor for the oxide layer that can be applied to your results?

Equations 12/13 - some parentheses formatting issues in 12 and an extra set of parentheses in 13; you don't have to fix them but they look a little ugly.

Table 2 - The default values for 0.7um junction depth from COMSOL look wrong based on the values from the other 2 CMOS junction depths. Concentrations increase from 0.18 to 0.35, but then the defaults have lower concentrations than 0.18. I would use simple linear regression to extrapolate the values instead of using the COMSOL defaults. I think this explains some of your strange results for the thickest junction depths. I would use linear regression (or another appropriate function, e.g. polynomial regression or something fancier like ML) to approximate the junction depths for the 0.7um thickness and re-run those calculations. This is also affecting your conclusions in line ~400 that the oxide layer explains the difference between experimental and your calculated efficiency. It's probably the doping concentration that's explaining it - I would check the concentrations from the paper(s) you are referencing and put those in Table 6.

Table 3 - double-check you aren't forgetting anything else here.

Line 249 - why is "metal1" in italics and not simply "metal"?

Table 6 - would be good to have doping concentrations here, or in another table. Perhaps breaking this into multiple tables (e.g. grouped by junction thickness or photodiode structure) would help make it easier to read.

Line 434 - "The results show that, globally, the greater the technology" - "the greater the technology" makes no sense in English, but I think you mean to simply say the greater the junction depth as you did in parenthesis. Simply say that.

Overall good work, but I think you need to re-run the simulations for the 0.7um depth or remove these results entirely since the doping concentrations seem to be way off (far too small). This is the only major revision I think you should take, other feedback is minor in comparison.

On a more nuanced note, isn't the concentration of dopants depth-dependent and not uniform? And COMSOL is making an assumption that the concentrations are uniform? I'm not sure you mentioned this assumption in the paper but I think this is probably an important point which explains some of the differences between the simulation and reality. Probably there is a way to do this in the simulation, which you might want to look into.

Author Response

Please see the attachment

Author Response File: Author Response.docx

Reviewer 4 Report

Thank you for submitting your manuscript. The following questions should be addressed before the publication.

1. The nonexistence of the oxide layer between the simulation and fabricated device makes the direct comparison of those two results difficult. Could the author provide any simulation with the oxide layer being introduced to make a fair comparison?
2. Could the author provide any theoretical analysis of why n+/p-substrate diode provide better performance in terms of responsivity and quantum efficiency?
3. The author mentioned that the quantum efficiency difference between experimental characterization and the proposed simulation could be affected by multiple aspects. Could the author provide a detailed analysis of how do those aspects could potentially affect the results?
4. The author mentioned that the detector is designed for malaria diagnosis. For this application, what is the performance of the existing detectors? Also, will it make a big difference if the quantum efficiency is 70% vs 90%?
5. The font size of Equation 12, 13, and 14 is significantly larger than other equations. Please verify if it is the pdf conversion issue.

Author Response

Please see the attachment

Author Response File: Author Response.docx

Round 2

Reviewer 3 Report

Great work making improvements! Although I'm surprised the results didn't change with the changing of the concentrations in the simulation.

Reviewer 4 Report

Thank you for resubmitting your manuscript. The questions brought by the reviewer previously have been addressed properly. This article is recommended for direct publication.