1. Introduction
In the development of modern advanced silicon process technology, one of the primary focuses is on increasing front-end-of-line (FEOL) device density by reducing device dimensions, including a channel length reduction for device speed advantages. However, the FEOL shrinkage enforces the back-end-of-line (BEOL) to be scaled down to match with the FEOL scaling for interconnects. BEOL scaling implies not only a reduction in the metal width, but also the metal thickness to prevent aggravating the RC constant time imposed by the metal interconnects.
Though the thinner metal layers allow for an increased number of metal layers to be packed in an adequate die thickness for handling, packaging, and so on, every metal layers becomes closer to the silicon substrate. Hence, due to the more complicated high frequency interactions among the metal layers and with the silicon substrate present, the magnetic and electric interactions in RF chips yield lower performances for passive RF components, such as inductors, transmission lines, and metal capacitors.
The additional metal stack can increase the distance of the UTM to the substrate [
1,
2,
3,
4]. Due to this, the mask process increased. And if the UTM-to-substrate distance is increased, the characteristics of the RF passive device are improved, but other active circuits may be deteriorated due to the increase in the routing resistance.
2. Problem Statement
As the position of ultra-thick-metal (UTM) layer becomes closer to the silicon substrate, negative effects on passive RF components occur.
Figure 1a shows the quasi-TEM modes occurring in the substrate due to microwave and millimeter wave frequencies, as well as the deep penetration of the E-field and H-field into the substrate.
Figure 1b shows the E-field within the substrate according to frequency, using an HFSS EM simulator. We obtained the material and thickness information for EM simulations from foundry technology files, which were then calibrated through the silicon validation of a general inductor pattern. Based on these considerations, the next part presents a solution to circumvent these losses. Firstly, the proximity of the UTM layer to the substrate increases an electrical coupling to the substrate, as shown in
Figure 1c. The increased electrical coupling lowers the self-resonance frequency (SRF), with an increased coupling capacitance to the substrate. Note that the dropped SRF handicaps the maximum operation frequency. The increased electrical coupling also gives rise to an aggravated displacement current loss because of the lossy substrate grounding, as illustrated in
Figure 1d. Secondly, the penetration of electrical fields into the silicon substrates becomes deeper, as the operating frequency is increased. The increasing depth of the penetration worsens the dielectric loss due to the lossy silicon substrate material. Lastly, more magnetic flux interacts with the silicon substrate as the UTM layer becomes closer to the silicon substrate, and hence one can notice the increasing Eddy current loss through the increasing magnetic coupling, as depicted in
Figure 1e. As shown in
Figure 1f, a total effective magnetic flux surrounding the UTM layer generated by AC current flows is diminished as the space between the UTM layer and the silicon substrate narrows. The consequence of this is a reduced inductance density for an equivalent inductor dimension.
Therefore, the research objective is to find a fair position for the UTM layer without compromising full circuit integration for three purposes. One purpose is to minimize the magnetic loss and the electrical loss on RF passive components, and the second is not to impose excessive ohmic losses on metal interconnects through multi-via chain from FEOL devices to the top of the metal layers, which is the UTM layer. The third purpose, therefore, is to accomplish an RF circuit performance enhancement that can exchange with area optimizations and power minimizations. In this paper, the proposed optimization method proves the advancement by evaluating the quality of several passive components, such as the inductor and transmission lines, in the RF (<10 GHz) and mmWave (>10 GHz) frequency ranges.
3. Experiments
Loss mechanisms associating with the UTM altitude described earlier are used to simply drive the desires of the most favorable UTM position for the process-aided RF circuit performance enhancement. The observation of the electric field and magnetic field profiles in an integrated silicon material is the first step to understand the various loss mechanisms at various frequency ranges. There are three scenarios in the observation of the depths of the E-field in a UTM layer inductor, as follows: (1) substrate surface with n-type epitaxial layer (N
epi), as suggested in [
5]; (2) without N
epi; and (3) with patterned-ground-shield (PGS) [
6,
7,
8].
As presented in
Figure 2a, the penetration depth of the E-field determines the total amount of dielectric loss of the lossy substrate material, and
Figure 2b summarizes the dielectric loss amount for each inductor configuration. Four key performance metrics of inductors, such as (1) the peak quality factor (Q
max), (2) the self-resonance-frequency (SRF), (3) the coupling capacitance to the substrate (C
ox), and (4) the effective line inductance (L
eff), are evaluated for various distances of the UTM layer from the silicon substrate, as shown in
Figure 2c–f. The results suggest the improvement of peak Q, and the SRF is saturated beyond a distance of 7 um, while L
eff is slightly increased as the UTM becomes further away from the substrate.
4. Process Optimization
The increased height of the via underneath the UTM layer escalates the UTM position. A similar previous work showed the effectiveness of the escalated UTM position by simply adding more metal layers at the cost of extra masks and processing steps [
2]. Instead of adding complexity in the BEOL process, fabricating the taller via underneath the UTM does the trick, as far as concerns the process of choice balances via resistance for the benefit of the escalated UTM position at various frequency ranges.
Three process aspects determining via contact resistance (Rc) density are (1) via aspect ratio, (2) pitch of the via process implying via diameter (Dvia) and the spacing between vias (Svia), and (3) a composition of materials for via fabrication. The optimization of the via height incorporates all three of these aspects.
Figure 3 describes the via structures under consideration, and
Table 1 and
Table 2 tabulate the compositions of the materials for via fabrication. As the via height increases, a portion of the main via material, Cu, dominates the overall R
c, while the thickness of the barrier materials is unchanged. The aspect ratio of the via needs to be maintained in order to not affect a profile of vertical via shape, which causes potential voids during via fabrication. Meanwhile, reliability requirements, such as time-dependent dielectric breakdown (TDDB), determine a minimum via spacing, which depends slightly on via height. One can maintain the minimum via spacing, as long as the reliability requirement is satisfied.
As per the results,
Figure 4a summarizes the trend of via contact resistance (R
c) according to via heights.
Figure 4b also summarizes via contact resistance density according to via height complying to reliability requirements with a minimum allowable via spacing.
4.1. DC Result of Process Optimization
The investigation of the altitude of the UTM layer takes into account the UTM position on E-field/M-field couplings, as well as the change in via resistance as a result of process modifications. The conclusion suggests escalating the via height to position the UTM layer at an optimal cost-effective position away from the silicon substrate, as illustrated in
Figure 5. One can find the photos of the vertical via profile before and after process modification in
Figure 6a.
Figure 6b illustrates the total resistance from the full via stacks from the lowest via metal to the highest metal layer (also known as the UTM layer), which drops by as much as one-third.
4.2. Measurement Results of Micro-Strip
Circuit designs for RF/mmWave applications require a 50 ohm transmission line. However, one layer of a technologically advanced metal has a small width/thickness, so insertion loss is serious when implementing microstrip line and coplanar waveguide (CPW); so, UTM-1 and UTM-2 metals are used as ground metals. Since UTM-2 has a short distance, the UTM width is reduced to 50 ohms, so that the insertion loss is inevitably large. In this paper, it is possible to fabricate a 50 ohm transmission line using UTM-1 and UTM-2 by increasing the height of the via below the UTM layer.
We provide the total effect of the escalated UTM position, and the re-engineered via process on a microstrip line design in
Figure 7.
Figure 7a displays the normalized characteristic impedances of the microstrip line fabricated in the UTM layer with various widths and altitudes.
Figure 7b depicts the insertion loss according to the UTM positions. One can understand how to find an optimum UTM position for one’s target applications by observing
Figure 7.
4.3. Measurement Results of Passive Components
Figure 8 demonstrates the performance improvement of an inductor fabricated in a UTM layer at the targeted position, and one can notice the 14.8% Q-factor enhancement and the 13.1% SRF increase as a result of nothing but the elevated UTM position, as shown in
Figure 8a and b, respectively. The experiment with a microstrip line also achieves a higher characteristic impedance with a lower insertion loss (IL), as shown in
Figure 8c and d, respectively.
Table 3 and
Table 4 summarize the measurements of the fabricated inductor and microstrip line with only elevated UTM processes.
4.4. Measurement Results of RF Circuit
For the demonstration of the effectiveness of the escalated UTM position, we fabricated a low-noise amplifier (LNA) [
9,
10] tuned at 29 GHz in a 28 nm FD-SOI process, as shown in
Figure 9a,b. A cascoded LNA topology is chosen for its simplicity and the footprints of all device components are untouched both before and after the process modification, except for the via layer elevating the UTM. Therefore, one can assume the majority of performance improvements are in relation to the process change only. Note that the processing cost change is negligible, since no additional layers are used.
We measured more than 30 chips at the same wafer and location. Among them, this is a demonstration of the data of a chip with a medium value. The reduction in the IL in the input and output matching networks contributes to the signal gain, S
21, by as much as 31.8%, as shown in
Figure 9c and
Table 5. Also, the process change drops the min-imum noise figure by 9.1% thanks to the reduced input IL, according to
Figure 9d and
Table 5.
4.5. Measurement Results of Reliability
In this paper, the minimum via spacing according to via height was determined to satisfy reliability requirements such as time-dependent dielectric breakdown. And the via resistance of the optimized process is lowered by 1/3, which can be expected to improve the performance of the logic circuit.
The reliability of the proposed elevated UTM process was tested in the 28FDS process, and the back-end-of-line reliability focuses on electro migration (EM), stress migration (SM), and time-dependent dielectric breakdown (TDDB).
EM measures the change in resistance caused by the movement of matter due to the continuous movement of ions within a conductor, which, in turn, is caused by changes in momentum between conducting electrons and diffusing metal atoms.
Figure 10 shows that the EM structure within the process window passed the accelerated reliability test. This guarantees that devices manufactured using the proposed process can be used for more than 10 years.
SM is a measurement that checks for voids, hillocks, and openings in metal wiring due to stress.
Table 6 shows that all SM structures passed the 500 h accelerated reliability test. It was confirmed that the interconnection of the proposed elevated UTM process was reliable. And it shows improved reliability results due to the increase in via contact area.
TDDB is an item that measures the oxide breakdown phenomenon and oxide life-time over time by applying a high temperature and a high voltage to the capacitor. As shown in
Table 7, all TDDB structures passed the reliability test. This shows that there is no leakage current flowing into the hard breakdown of the capacitor during electrical stress.
5. Conclusions
In this paper, the distance of the UTM layer to the substrate is increased by elevating the via height. This distance increase is carried out without additional mask processes by increasing the via height in order to maintain the same aspect ratio and is conducted with the same equipment. And a methodology for selecting the optimal width/height is performed, while maintaining the stability of the etch process by keeping the aspect ratio. Through this, the Rc per unit area was kept lower than before.
It is possible to fabricate a 50 ohm transmission line using UTM-1 and UTM-2 by increasing the height of the via below the UTM layer and the performance of the inductor could be dramatically improved. The effectiveness of the escalated UTM is shown using inductor and transmission line measurements using 28FDS technology, and has also been studied in LNA design with/without the escalated UTM and the elevated UTM via. The measurement of the LNA design reports an overall 30% improvement in loss counting integration loss by UTM via, the electric loss in the inductor, and the magnetic loss appearing in the transmission line for PAD connection.
The simple modification in UTM position gains a significant amount of performance improvements on interconnects and passive components without paying for extra process steps. The well-designed BEOL configuration can improve the performance with the cost of a negligible increase in turnaround time.
Author Contributions
Validation, S.K.; formal analysis, S.K.; investigation, S.K.; writing—original draft preparation, S.K.; writing—review and editing, H.L.; Y.J.; supervision, H.L.; Y.J. All authors have read and agreed to the published version of the manuscript.
Funding
This study received no external funding.
Data Availability Statement
The original contributions presented in the study are included in the article, further inquiries can be directed to the corresponding author.
Conflicts of Interest
Authors Sutae Kim and Hyungjin Lee were employed by the company Samsung. The remaining authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
References
- Lee, H.J.; Callender, S.; Rami, S.; Shin, W.; Yu, Q.; Marulanda, J.M. Intel 22 nm Low-Power FinFET (22FFL) Process Technology for 5G and Beyond. In Proceedings of the IEEE Custom Integrated Circuits Conference, Boston, MA, USA, 22–25 March 2020; pp. 1–7. [Google Scholar] [CrossRef]
- Ong, S.N.; Lehmann, S.; Chow, W.H.; Zhang, C.; Schippel, C.; Chan, L.H.; Andee, Y.; Hauschildt, M.; Tan, K.K.; Watts, J.; et al. A 22nm FDSOI Technology Optimized for RF/mmWave Applications. In Proceedings of the IEEE Radio Frequency Integrated Circuits Symposium (RFIC), Philadelphia, PA, USA, 10–12 June 2018; pp. 72–75. [Google Scholar] [CrossRef]
- Yu, Q.; Garrett, J.; Waldemer, J.; Ma, Y.; Ravikumar, S.; Liu, G.; Rami, S. A mmWave Switch Using Novel Back-End-Of-Line (BEOL) in 22 nm FinFET Technology. In Proceedings of the IEEE MTT-S International Microwave Symposium Digest, Atlanta, GA, USA, 6–11 June 2021; pp. 34–37. [Google Scholar] [CrossRef]
- Yu, Q.; Kim, G.S.; Garrett, J.; Thomson, D.; Dogiamis, G.; Monroe, N.; Han, R.; Ma, Y.; Waldemer, J.; Nam, Y.S.; et al. Low-Loss On-Chip Passive Circuits Using C4 Layer for RF, mmWave and sub-THz Applications. In Proceedings of the IEEE MTT-S International Microwave Symposium Digest, Denver, CO, USA, 19–24 June 2022; pp. 325–328. [Google Scholar] [CrossRef]
- Dubuc, D.; Tournier, E.; Telliez, I.; Parra, T.; Boulanger, C.; Graffeuil, J. High quality factor and high self-resonant frequency monolithic inductor for millimeter-wave Si-based IC’s. In Proceedings of the IEEE International Microwave Symposium Digest, Seattle, WA, USA, 2–7 June 2002; pp. 193–196. [Google Scholar] [CrossRef]
- Yue, C.P.; Wong, S.S. On-chip spiral inductors with patterned ground shields for Si-based RF ICs. IEEE J. Solid-State Circuits 1998, 33, 743–752. [Google Scholar] [CrossRef]
- Liang, H.B.; Lin, Y.S.; Chen, C.C.; Lee, J.H. Optimization of PGS pattern of transformers/inductors in standard RF BiCMOS technology for RFIC applications. In Proceedings of the IEEE Radio Frequency Integrated Circuits Symposium (RFIC), San Francisco, CA, USA, 10–13 June 2006. [Google Scholar] [CrossRef]
- Wu, J.N.; Chang, Y.M.; Shen, Z.Y.; Chen, H.C. Investigation on CMOS on-Chip Inductors Using Various Patterned Ground/Floating Shield Techniques. In Proceedings of the 2021 International Conference on Electronic Communications, Internet of Things and Big Data (ICEIB), Yilan County, Taiwan, 10–12 December 2021; pp. 113–116. [Google Scholar] [CrossRef]
- Zhang, C.; Zhang, F.; Syed, S.; Otto, M.; Bellaouar, A. A Low Noise Figure 28 GHz LNA in 22 nm FDSOI Technology. In Proceedings of the IEEE Radio Frequency Integrated Circuits Symposium (RFIC), Boston, MA, USA, 2–4 June 2019; pp. 207–210. [Google Scholar] [CrossRef]
- Shin, W.; Callender, S.; Pellerano, S.; Hull, C. A Compact 75 GHz LNA with 20 dB Gain and 4 dB Noise Figure in 22 nm FinFET CMOS Technology. In Proceedings of the IEEE Radio Frequency Integrated Circuits Symposium (RFIC), Philadelphia, PA, USA, 10–12 June 2018; pp. 284–287. [Google Scholar] [CrossRef]
| Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to people or property resulting from any ideas, methods, instructions or products referred to in the content. |
© 2024 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access article distributed under the terms and conditions of the Creative Commons Attribution (CC BY) license (https://creativecommons.org/licenses/by/4.0/).