The Effect of the Dielectric Layer with Surface Plasma Treatment on Characteristics of Amorphous Silicon Antifuse
Round 1
Reviewer 1 Report
Authors presents a Nitrogen plasma-treated hydrogenated Amorphous Si as the antifue used in the FPGA. It may be interesting to the readers. However, I can’t recommend this manuscript to publish for the reasons.
1. The antifuse developed in this manuscript is dedicated for the FPGA. Therefore, author better to explain in detail that how antifuse was used in FPGA. Consequently, the basic characteristics and requirement of the antifuse in FPGA.
2. Accordingly, author is better to review the current antifuse performance in the state-of-the-art FPGA.
3. The nitrogen plasma treatment is a straightforward technique for a-Si:H. the author need to point point out the unique advantages of the plasma treatment.
4. How is the thickness, N content, microstructure, etc. of the nitrogen-treated layer? And, how it affects the leakage current?
5. In addition to the leakage current, will this layer affect the operation speed and other performance characteristics of the FPGA?
6. Please provide the statistics properties of the current-voltage characteristics like Fig. 4.
7. The resistance for the treated a-Si:H are not higher than the untreated sample in Fig. 6. What’s happen?
8. Please provide the leakage current in current density. And, please bench mark the current leakage current from the literature in Table one or separated.
Author Response
1.Authors presents a Nitrogen plasma-treated hydrogenated Amorphous Si as the antifuse used in the FPGA. It may be interesting to the readers. However, I can’t recommend this manuscript to publish for the reasons.
- The antifuse developed in this manuscript is dedicated for the FPGA. Therefore, author better to explain in detail that how antifuse was used in FPGA. Consequently, the basic characteristics and requirement of the antifuse in FPGA.
Reply: In the revised manuscript, we have appended some content related to FPGA, which is scattered in section1:
FPGA typically consists of three components: programmable logic blocks, programmable routing, and I/O blocks. Among the three, the programmable routing interconnect comprises of almost 90% of the total chip area, and thus contributes to most of the delay, area and power consumption [6]. Typically, an FPGA uses 20 to over 100 interconnects per logic gate to link logic blocks [6]. Antifuses are among the commonly used interconnect devices for FPGAs.
……
The uncertainty of the programmed resistance value due to the breakdown-position uncertainty leads to inconsistent signal delay (RC delay) time, which results in poor performance for high-speed FPGA. Therefore, with gate oxide antifuse used as programmable interconnect device, high-performance FPGA can not be realized.
……
With the advantages of lower capacitance, faster operation and smaller size, α-Si:H antifuse is very promising for high-speed and high-density FPGA [14].
……
The mature and commercialized state-of-art antifuse FPGAs are all based on metal-to-metal antifuses. Microsemi’s antifuse consists of a two layer dielectric composed of silicon nitride and amorphous silicon sandwiched between a raised tungsten plug and the top metal layer. The QuickLogic antifuse structure consists of two metal layers separated by a single layer amorphous silicon dielectric. In the literature[7], it is found that in the reliability test, the two types of FPGAs all have some individual phenomena of programming failure. The standby current of the antifuse FPGA product can reach 8mA due to the antifuse leakage [18].
- Accordingly, author is better to review the current antifuse performance in the state-of-the-art FPGA.
Reply: In the revised manuscript, we have supplemented this review of the state-of-the-art antifuse FPGA in section 1 Introduction.
The mature and commercialized State-of-art antifuse FPGAs are all based on metal to metal antifuses. MicroSemi’s antifuse consists of a two layer dielectric composed of silicon nitride and amorphous silicon sandwiched between a raised tungsten plug and the top metal layer. The QuickLogic antifuse structure consists of two metal layers separated by a single layer amorphous silicon dielectric. In the literature[7], it is found that in the reliability test, the two types of FPGAs all have some individual phenomena of programming failure. The standby current of the antifuse FPGA product can reach 8mA due to the antifuse leakage [18].
- The nitrogen plasma treatment is a straightforward technique for a-Si:H. the author need to point point out the unique advantages of the plasma treatment.
Reply: The nitrogen plasma treatment is indeed a straightforward technique for a-Si:H. However，the Al/α-Si:H,N/α-Si:H/Al(N-5min) antifuse formed by nitrogen plasma treatment has the unique advantage of eliminating the antifuse switch-off effect, which is a persistent problem of the antifuse reliability, inherent in SiNx, SiOx, a-Si:H, etc. At the same time, the nitrogen plasma treatment makes α-Si develop its strengths and avoid its weaknesses. In order to obtain the ideal programming voltage, α-Si does not need to be too thin, can be several times larger than SiNx, SiOx, easy to obtain the film thickness with smaller relative error, accurate and stable parameters, good process controllability. The nitrogen plasma treatment did not increase the additional thickness of film, maintaining the existing advantages of α-Si. The ON-state resistance of α-Si was small and distributed centrally. The nitrogen plasma treatment facilitated the α-Si antifuse to obtained a more concentrated small resistance, which slightly improved the ON-state resistance characteristics of a-Si:H antifuse. The programming voltage of α-Si is concentrated, the nitrogen plasma treatment granted the α-Si antifuse a more concentrated programming voltage, Slightly improved programming performance of a-Si:H antifuse. The α-Si antifuse has the disadvantage of large leakage current. The nitrogen plasma treatment significantly reduced the antifuse OFF-state leakage current, which was close to SiNx and SiOx. In summary, The nitrogen plasma treatment technology has a simple process, good compatibility, easy control, good uniformity, consistency and stability, resulting in comprehensive and superior performance at minimal cost, such as stable programming voltage, small resistance and concentrated distribution, small leakage current, and elimination of switch-off, ensures the performance, yield and reliability of antifuse FPGA, and high cost performance.
- How is the thickness, N content, microstructure, etc. of the nitrogen-treated layer? And, how it affects the leakage current?
Reply: The nitrogen plasma treatment is applied to thin film surface to improve surface characteristics. It is not aimed at generating the independent layer with specific thickness. The surface Morphologies of α-Si:H,N film is shown in Fig. 3. The film is extremely thin and unmeasurable. The nitrogen plasma treatment does not increase the film thickness extra, therefore does not affect the breakdown voltage, but through surface improvement to obtain low leakage and other performance improvements. The treated α-Si:H has the fine surface, because the saturation of the treated a-Si:H surface dangling bonds was maximized and there were the minimum number of surface defects. So the surface characteristics of the treated α-Si:H are better than the untreated sample with the rough surface, resulting in the effect of reducing the leakage current.
- In addition to the leakage current, will this layer affect the operation speed and other performance characteristics of the FPGA?
Reply：A large number of distributed antifuses exist in FPGA. Reducing the leakage current of the amorphous silicon antifuses can effectively improve the reliability of FPGA and obtain a smaller static current.
Second, Due to the surface treatment, the antifuse on-state resistance after programming is small and has a tight distribution. On the one hand, the delay of programmable interconnect line of FPGA is reduced and the operation speed is correspondingly faster. On the other hand, the antifuse programming in FPGA has better uniformity and consistency, and the FPGA programming yield is increased.
Third, Due to the surface treatment, the switch-off effect of the amorphous silicon antifuse is eliminated, which avoids the potential logic error and improves the reliability of FPGA.
- Please provide the statistics properties of the current-voltage characteristics like Fig. 4.
Reply：In the revised manuscript, we have supplemented the current-voltage characteristics of the newly developed Al/α-Si:H,N/α-Si:H/Al(N-5min) antifuse shown in Fig.4(b), which illustrates the statistics properties of the newly developed antifuse.
- The resistance for the treated a-Si:H are not higher than the untreated sample in Fig. 6. What’s happen?
Reply：Fig. 3 shows the AFM surface morphologies of the treated α-Si:H and the untreated sample. The rough surface of the untreated sample caused the uneven electric field, Leading to the antifuse programming is not uniform, and the programmed resistance are distributed over a wide range. The treated α-Si:H has the fine surface, because the saturation of the treated a-Si:H surface dangling bonds was maximized and there were the minimum number of surface defects. So the surface characteristics of the treated α-Si:H are better than the untreated sample, and the treated α-Si:H programming is more stable, resulting in the tight distribution for the programmed resistances. In conclusion, The resistance of the untreated sample are partially higher than that of the treated
- Please provide the leakage current in current density. And, please bench mark the current leakage current from the literature in Table one or separated.
Reply：(1) In the revised manuscript, the leakage currents from the literature have been marked separately. In the newly added Table II, the leakage current is given in terms of current density. In order to be consistent with the experimental data in FIG. 4, the leakage current in Table I is not currently given in terms of current density, but there is no ambiguity because section 2 gives the exact dimensions of each structure.
(2) The all dielectric/α-Si structures in Table I are experimental samples prepared by the authors themselves, which partly based on the literature, but it is not exactly the same as the literature. The contents in the table I are all experimental data, not literature data.
Reviewer 2 Report
I have the following concern:
1. Rewrite the abstract from zero, focus of the problem, idea, algorithm to solve it and result not exceed half page.
2. Write complete algorithm represent the main algorithm of the proposal method or as point base on suitable format
3. Add new section under title the hypothesis and limitations of the develop method present by author(s).
4. Add section under title discuss explain on it the advantages/disadvantage of this Construction based on the researcher(s)'opinion to become as Guide of the other future researchers work at the same field, at the same time to increase the weight of your manuscript.
7. Preparing table show the compare among the previous works and put that table at the end of section " Related work" in addition, must add (two - three lines) after each previous work show in that lines you as authors similar and different with that previous works at any point.
8. Rewrite the Conclusion through extended it also must present complete flowchart or block diagram of the Methodology.
Author Response
- I have the following concern:
- Rewrite the abstract from zero, focus of the problem, idea, algorithm to solve it and result not exceed half page.
Reply：Rewrite the abstract:
With the advantages of lower capacitance, faster operation and smaller size, the amorphous silicon antifuse is very promising for high-speed and high-density FPGA. However, the leakage current of the conventional amorphous silicon antifuse in the off-state is high, which decreases its performance and reliability. Although the leakage current of the SiNx or SiOx antifuse is small, its proper thickness of the dielectric layer is not easily controlled by PECVD processes. And the highly undesirable switch-off behavior is common to almost all metal-to-metal antifuse structures. Focusing on the study of amorphous silicon multilayer dielectric structures, A novel antifuse with the structure of Al/α-Si:H,N/α-Si:H/Al was proposed, which was manufactured by nitrogen plasma treatment for α-Si:H film surface. Through surface plasma treatment, the hydrogen content of the dielectric layer is stable, the film surface is smoother, the leakage current is reduced, the switch-off is eliminated, the programming voltage is more concentrated and the on-state resistance distribution is more compact. The results demonstrated that surface plasma treatment with proper time for the dielectric layer could significantly improve the performance and reliability of the Al/α-Si:H,N/α-Si:H/Al antifuse. Furthermore, the fabrication process of the α-Si:H,N/α-Si:H structure has excellent compatibility, controllability and simplicity.
- Write complete algorithm represent the main algorithm of the proposal method or as point base on suitable format
Reply：The current work focuses on the newly developed Al/α-Si:H,N/α-Si:H/Al antifuse. The proposal method for fabricating the novel Al/α-Si:H,N/α-Si:H/Al antifuse is the nitrogen plasma treatment.
In the revised manuscript, we have supplemented a complete flowchart shown in Fig.1(b), which illustrates the fabrication process of the novel Al/α-Si:H,N/α-Si:H/Al antifuse.
The algorithm in the current work only involves the calculation of Hydrogen content and is quoted from literature [19,20]. Therefore, only a brief introduction is made in the revised manuscript, with notes at the citations.
- Add new section under title the hypothesis and limitations of the develop method present by author(s).
Reply：In the revised manuscript, a new section, 3.8 Hypothesis and Limitations of the Fabrication Process of the α-Si:H,N/α-Si:H Structure, has been added.
In the fabrication process of the newly developed Al/α-Si:H,N/α-Si:H/Al antifuse, the recommended time of surface treatment with nitrogen plasma for α-Si:H thin film is 5 minutes. The surface treatment was conducted by PECVD under the hypothesis process conditions: the radio frequency (RF) power, total pressure and substrate temperature were 60 Watt, 0.6 Torr and 300℃ respectively. If the process method and process conditions change, the recommended time of 5 minutes is no longer applicable.
The nitrogen plasma treatment is a surface modification treatment for α-Si:H. The surface treatment degree needs an optimal value, which should not be too small or too large.
With the time of surface plasma treatment for dielectric films being increased, nitrogen plasma could saturate more silicon dangling bonds of the α-Si:H film surface and the defects of the α-Si:H film surface decreased gradually. When the time of surface plasma treatment accumulated to the suitable amount, the saturation of the surface dangling bonds was maximized and there were the minimum number of surface defects. After proper time, nitrogen plasma could not further saturate silicon dangling bonds of the surface, on the contrary, it would cause the etching effect. According to the surface treatment mechanism and the current α-Si:H,N/α-Si:H film preparation practice, it is easy to locate the time optimal value of surface treatment through process comparison experiments. Therefore, In order to obtain the high performance amorphous silicon antifuse, the nitrogen plasma treatment with proper time for the dielectric film surface is a highly recommended method.
- Add section under title discuss explain on it the advantages/disadvantage of this Construction based on the researcher(s)'opinion to become as Guide of the other future researchers work at the same field, at the same time to increase the weight of your manuscript.
Reply：In order to better improve the overall effect of the paper, we rewrited the Abstract and the Conclusion, divided section3 into section3.1 to 3.8.
In the revised manuscript, the advantages/disadvantage of this Construction is scattered in section3.7, section3.8 and section4.
- Preparing table show the compare among the previous works and put that table at the end of section " Related work" in addition, must add (two - three lines) after each previous work show in that lines you as authors similar and different with that previous works at any point.
Reply：An added table, Table II, shows the comparison between the previous works and the current work. At the same time, Literature annotation is also carried out in table II.
Table II clearly shows the positive significance of the author's current work, the α-Si:H,N/α-Si:H, relative to the previous works：
(1) In the table II, the α-Si:H,N/α-Si:H antifuse has the highest reliability due to its elimination of the switch-off effect.
(2) To achieve ideal programming voltage（8-10 V）, the film thickness of SiNx, SiOx, or NO, too thin, is difficult to achieve stable control by PECVD processes, but the thicker thickness (60nm/80nm) of the α-Si film is easy to be deposited stably. So, the α-Si film is an ideal choice for fabricating the antifuse.
(3) The OFF-state leakage of the α-Si:H,N/α-Si:H antifuse is very low, close to that of SiNx and SiOx, and significantly better than the α-Si:H.
(4) The ON-state resistance of the α-Si:H,N/α-Si:H antifuse is small and tight, close to that of SiNx and SiOx, and better than the α-Si:H.
(5) The Programming voltage distribution of the α-Si:H,N/α-Si:H antifuse is compact , better than the α-Si:H, and significantly better than the SiNx and SiOx.
- Rewrite the Conclusion through extended it also must present complete flowchart or block diagram of the Methodology.
Reply：(1). Rewrite the Conclusion:
At present, mainstream FPGA manufacturers adopt metal-to-metal antifuses based on amorphous silicon, although there are many advantages, there are still shortcomings，such as large standby current, accidental programming failure, etc. Investigation and experiments have found that the metal-to-metal antifuses is prevalent during undesirable switch-off behavior, the leakage current of the conventional amorphous silicon antifuse in the off-state is high, however, Although the leakage current of the SiNx or SiOx antifuse is small, it is difficult to control the thin dielectric layer thickness.
Focusing on the study of amorphous silicon multilayer dielectric structures, A novel antifuse with the structure of Al/α-Si:H,N/α-Si:H/Al was proposed, which was manufactured by nitrogen plasma treatment for α-Si:H film surface. The experimental results revealed that the nitrogen plasma treatment could significantly improve the performance of amorphous silicon dielectric film, for nitrogen plasma could further saturate silicon dangling bonds to decrease the defects of the dielectric film surface.
The nitrogen plasma treatment is indeed a straightforward technique for a-Si:H. However，the Al/α-Si:H,N/α-Si:H/Al antifuse formed by nitrogen plasma treatment has the unique advantage of eliminating the antifuse switch-off effect, which is a persistent problem of the antifuse reliability, inherent in SiNx, SiOx, a-Si:H, etc. At the same time, the nitrogen plasma treatment makes α-Si develop its strengths and avoid its weaknesses. In order to obtain the ideal programming voltage, α-Si does not need to be too thin, can be several times larger than SiNx, SiOx, easy to obtain the film thickness with smaller relative error, accurate and stable parameters, good process controllability. The nitrogen plasma treatment did not increase the additional thickness of film, maintaining the existing advantages of α-Si. The on-state resistance of α-Si was small and distributed centrally. The nitrogen plasma treatment facilitated the α-Si antifuse to obtained a more concentrated small resistance, which slightly improved the on-state resistance characteristics of a-Si:H antifuse. The programming voltage of α-Si is concentrated, the nitrogen plasma treatment granted the α-Si antifuse a more concentrated programming voltage, Slightly improved programming performance of a-Si:H antifuse. The α-Si antifuse has the disadvantage of large leakage current. The nitrogen plasma treatment significantly reduced the antifuse off-state leakage current, which was close to SiNx and SiOx. In summary, The nitrogen plasma treatment technology has a simple process, good compatibility, easy control, good uniformity, consistency and stability, resulting in comprehensive and superior performance at minimal cost, such as stable programming voltage, small resistance and concentrated distribution, small leakage current, and elimination of switch-off, ensures the performance, yield and reliability of antifuse FPGA, and high cost performance.
(2). In the revised manuscript, we have supplemented a complete flowchart of the Methodology shown in Fig.1(b), which illustrates the fabrication process of the newly developed Al/α-Si:H,N/α-Si:H/Al antifuse.
Author Response File: 
Author Response.pdf
Round 2
Reviewer 1 Report
The authors addressed all my comments/suggestions. I recommend this manuscript for publication.
Reviewer 2 Report
It can be accepted now
