- Nano Express
- Open Access
Charge Splitting In Situ Recorder (CSIR) for Real-Time Examination of Plasma Charging Effect in FinFET BEOL Processes
Nanoscale Research Letters volume 12, Article number: 534 (2017)
- The Correction to this article has been published in Nanoscale Research Letters 2017 12:584
A novel device for monitoring plasma-induced damage in the back-end-of-line (BEOL) process with charge splitting capability is first-time proposed and demonstrated. This novel charge splitting in situ recorder (CSIR) can independently trace the amount and polarity of plasma charging effects during the manufacturing process of advanced fin field-effect transistor (FinFET) circuits. Not only does it reveal the real-time and in situ plasma charging levels on the antennas, but it also separates positive and negative charging effect and provides two independent readings. As CMOS technologies push for finer metal lines in the future, the new charge separation scheme provides a powerful tool for BEOL process optimization and further device reliability improvements.
Plasma-enhanced processes are widely used in the formation of fin field-effect transistor (FinFET) circuits, which composed of many high-aspect ratio structures and fine metal lines . During manufacturing, etching and deposition step for realizing these 3D compositions can lead to significant plasma-induced stress to the FinFET devices [2,3,4]. As CMOS FinFET technology advances, the metal linewidth and pitch reduce more aggressively than its height, driving the need for forming high aspect ratio trenches defined by extremely fine lines. Unavoidably, this promotes the severity for plasma-induced damage (PID) to the transistors, and its corresponding effect on circuit reliability becomes one of the key concerns in developing FinFET technologies [5,6,7]. In forming small contacts, vias, and fine metal lines, strong power and high-selectivity plasma are generally applied . Moreover, in etching the bulk fin, sputtering of reactive ions on the fin surface can lead to defects in the bulk fin, critical to the characteristics of the transistors . In order to enable fin metal gate and dense interconnect structures, complex metal stacks are more often used in advanced FinFET technologies [10, 11]. In addition, high-k gate dielectric used in advanced technology usually leads to enhanced stress-induced trapping after plasma process [12,13,14]. During plasma charging, discharging path through narrow fins and to the substrate can lead to more non-uniform stress levels across a whole wafer . The plasma-induced stress on the transistor gate oxide is known to result in the further degradation of the gate dielectric integrity [16, 17].
The plasma-induced damage on gate dielectric film can lead to performance degradation in highly non-uniform charging scenarios, even yield loss [18,19,20], as a result of reliability failures [21, 22]. Thus, in advanced FinFET technologies, test devices with enlarged antenna structures are generally used for monitoring PID effects, which provide feedbacks for further process optimization.
The most common and widely used measure of PID is the time-to-breakdown characteristic of the test samples with large antenna structures. The latent damage on these PID patterns are typically reflected by measuring the time-dependent degradation of the gate dielectric layers, hence, failing to give the real-time feedback of the plasma processes . In addition, conventional test devices cannot tell the sources and the polarity of plasma-induced charging rate and or maximum potential build-up on the antenna. A PID recorder with a floating gate coupled by antenna structure has been proposed with in situ detection capability in our previous study [24, 25]. In this work, we proposed a revised PID recorder with a charge splitting feature. Through a forward diode and a reverse diode connected to a common antenna structure, the new design provides separate paths for positive and negative charges. Therefore, charging levels of both polarities can be independently recorded. This new charge splitting in situ recorder (CSIR) requiring only small antennas enables future study of plasma charging effect in middle-end of the line (MEOL) processes.
Plasma Charging Polarity
Previous studies reported that, during etching process in forming poly- or metal layers, the plasma inhomogeneity as well as the variations of the antenna potential can lead to a drastic difference in the charging rate or even the polarities may change at different locations [24, 25]. The macro-environment in the plasma chamber and micro-patterns can both affect the distribution of charging rates on a wafer . Namely, plasma charging rate in back-end-of-line (BEOL) etching varies spatially and timely. During radio frequency (RF) plasma processes, the surface of the wafer collects charging current, J p, which is composed of an ion current J i and an electron current J e . The ion current is almost constant with time and is determined by the ion density J i and the Bohm velocity . Since the plasma potential V p(t) is higher than the gate potential V G for most of the time, the electron current flows only during the short periods when the plasma potential is near its minimum. During Q FG process, the gate voltage may increase or decrease over many RF cycles, depending on which component of the currents is larger, until a steady-state gate voltage is reached when the tunneling current balances J p on the antenna. As shown in Fig. 1, the distribution of plasma charging rate, J P (x,y,t), across the wafer during etching process at different stages changes in both magnitude as well as in polarities, where it can be expressed as in Eq. (1) where J e represents electron current density, and J i represents ion current density.
The different plasma charging polarities result in either positive or negative antenna charge, Q P, accumulated at a different time and location. To clarify, at time t 1, a negative J p leads to negative antenna charge Q−. At t 2, a positive J p induces a positive antenna charge Q+ on the identical location on the wafer, as illustrated in Fig. 1. Thus, positive or negative charge may accumulate on a same antenna at different time during the etching process. From previous reports , the peak levels of J e and J i are around − 0.15 and 0.35 mA/cm2, respectively. It has been found [28, 29] that DC and AC/bi-directional gate stress on n-channel and p-channel FinFET results in different latent damage to the gate dielectric film. High voltage stresses with positive or negative DC bias and AC voltage with a switching frequency of 0.1 Hz are applied to conventional FinFET test samples, respectively. As shown in Fig. 2, the time-to-breakdown (T BD) of a transistor stressed by positive, negative, and gate stress in both directions are compared. The results indicate that DC gate stress will cause worse damage on the samples, while AC gate stress results in less severe damage to these transistors, as suggested by the longer T BD for samples subjected to bi-directional stress. Figure 2 also shows that the oxide degradation depends not only on the charging polarity, but also on the type of wells under the n-channel and p-channel transistors, which is expected to be caused by the difference in the discharging paths of these test devices during process. Hence, conventional PID detector, which uses T BD as the indicator for damage severity cannot reflect the plasma charging level during the process. On the other hand, the plasma charging recorder proposed in our previous work records the stress level by injecting or ejecting electron to/from a floating gate (FG) coupled by a charge-collecting antenna. The recorded data, floating gate charge (Q FG), is read subsequently after fabrication [24, 25]. The recording is then measured by threshold voltage shift on the read transistor, of which the channel is controlled and directed by the same floating gate. The raised potential on the antenna with Q P from plasma charging can induce both positive and negative antenna voltage during the formation of a single metal layer. Further, for different metal layers, different manufacturing parameters are used. For example, etching time, chemical used, and chamber temperatures may vary. These parameters can affect the antenna charge distribution across a wafer during etching. In other cases, a transistor with connections to multiple metal layers subject to even more complex plasma charging sequences, as illustrated Fig. 3a.
At different stages of the BEOL process, the plasma charging current at a particular antenna can switch between ion and electron current, i.e., the net Q P can also shift from positive to negative. The recordings on samples with antenna consisting of metal 2, metal 3, metal 4, and multiple metal layers are summarized in Fig. 3b. Data suggests net charging of a single metal layer  on a particular change polarity from layer to layers. In addition, the averaging effect found on the Q FG of the samples with antenna structures of multiple metal layers is further supported by the measured data in Fig. 3b. With both positive and negative V G on the antenna, the final Q FG will then be averaged out by electron injection and ejection into/from the FG which may occur sequentially. This compensation effect will limit the recorder to reveal the real stress conditions a device experienced during plasma processes. The revised CSIR is designed to address the problem on how to individually record positive and negative charging effects without interference and to supply more detailed data on the charging situation in the plasma chamber.
Test Pattern for Charge Separation
In this research, the positive ion charging and negative electron charging on the antenna can be separated with a new charge splitting in situ recorder (CSIR) proposed, as illustrated in Fig. 4a. A CSIR consists of two floating gates, FG1 and FG2 which record the different types of charging effect separately. The antenna structure connects to the two coupling gates through a forward diode (D1) and a reverse diode (D2), respectively. In the left half of the structure, the positive charges will flow into the coupling gate 1 (CG1) through D1. When CG1 is positively charged, the voltage is coupled to the floating gate through the contact slots on both sides. The floating gate will be negatively charged as electrons inject from the substrate. The right half of the structure on the other hand is the negative charging path, allowing current to flow from the antenna into the coupling gate 2 (CG2) through D2, resulting in positive Q FG. Figure 4b further shows the cross-sectional view of the CSIR with on-chip pn diodes, directing the positive and negative charging paths to the separate coupling gates, CG1 and CG2, which couple the potential on the antenna to the FG1 and the FG2, respectively.
When the left half of the recorder are enabled in the CSIR under a positively charged antenna, the right half is inactive as charge is blocked by the reverse diode, and vice versa. Both on-chip diodes are composed of n+/p-well. For D2, to sustain negative voltage in its p-well, the p-region needs to be surrounded by a deep n-well, blocking the charging path directly to the substrate. The simulated potential distribution on the cross-section in a CSIR under positive and negative charging periods of the antenna is shown in Fig. 5a and b, respectively. Assuming that the potential on an antenna reaches 5 V, through the diode on the left, positive charge flows to the control gate on the left, which results in a high positive voltage (V CG1). At the same time, positive charge is blocked by the diode on the right, resulting in a close-to-zero V CG2. The difference in potential on the two control gates are verified by the simulated potential contours in Fig. 5a. The effect of negative charging on the antenna is shown in Fig. 5b. Simulated potential profiles verify that the on-chip pn diodes can effectively direct and block the potential to CG1 and CG2, complimentarily, as designed. This way, positive and negative charging effects corresponding to different sources in the plasma treatments can be independently obtained, preventing charge compensation and interference issues on the detector.
Results and discussion
The measured threshold voltage shift (ΔV T) on device controlled by FG1 with forward diode and that by FG2 with reverse diode and samples without diode are compared in Fig. 6. Data along the center line of a wafer reveal that a recorder with a single floating gate does subject to charge neutralization even within the processing of a single metal layer. The averaging effect of a recorder without diode proves that the peak charging rates will be not reflected truthfully. On the other hand, readings from the new CSIR can provide positive and negative charging levels, independently. To further investigate the plasma charging effect in metal 2 (M2) formation, the collected charge on FG1 and FG2 of the CSIR of each dies can be independently calculated by Eq. (2),
where Q FG is the charge in the floating gate. C T is the total capacitance of the floating gate, as illustrated in Fig. 7. ΔV T is the threshold voltage shift detected at the read gate of the recorder, while α RGis the coupling ratio from the read gate.
When floating gate charge is initially zero and Q FG reaches the saturated level when the electric field across the gate dielectric layer is reduced to zero, the final antenna gate potential at the end of a plasma process can be expressed as followed,
in which, V ANT is antenna gate potential by plasma charging and α ANT represents the coupling ratio to the floating gate from the antenna gate. V FB is the flatband voltage from the metal gate to the fin-substrate. Under a given process time, the average plasma charging current density, J p can be then projected by Eq. (4).
where Δt is the duration of a plasma process [28, 29] and C ANT is the total capacitance of the metal antenna, while A ANT is the charging area of an antenna. All the parameters used in the above calculations are summarized in Table 1.
The distribution of positive and negative charging rates across a wafer during processing of the top (metal 9) and bottom metal (metal 2) layers are further compared in Fig. 8. It implies that charging on the antenna structure is more prominent at higher metal levels (metal 9), because on metal 9, its higher plasma energy causes J p to be larger than J p of metal 2 in terms of magnitude. Also, data suggests that both electron and ion charging rates peak around the center for both cases. As expected, dies closed to the center of the wafer experience high charging level, which can be attributed to the longer discharging path during plasma treatment. This location effect is found to be identical for both more electron and ion charging dominant conditions. The projected plasma charging rate, J P (x,y), averaged over the formation of a single metal layer, metal 2 (M2) and metal 9 (M9), are further compared in Fig. 9. These wafer maps reveal that electron charging rate seems to be at a plateau except at the edge, while ion charging rates showed a higher variation in the middle section of the wafer. In the future, these wafer maps under different processing conditions are expected to provide insights to the plasma chamber, or further optimization guidelines by better compensating the charging effects.
Antenna Ratio Effect
Traditional PID monitoring devices are typically designed to amplify the PID effect by connecting the transistor’s gates directly to a large antenna, evaluating the stress levels by increase of the total Q P expected to be discharged through a small channel region [30, 31]. Antenna ratio (AR) is proportional to the stress current density through the gate dielectric during plasma processes . Large Q P on the antenna is known to induce latent damage and/or traps in the dielectric layer, which ultimately lead to reliability degradation . As expected, higher AR on conventional FinFETs does significantly elevate the stress levels, causing a more severe TBD degradation, namely, device failure within a shorter period of operation, see Fig. 10. On the other hand, in a CSIR, the plasma charging level recorded as the floating gate charge, Q FG, shows very little antenna effect. Namely, it does not respond to increasing antenna area, as revealed by the data summarized in Fig. 11.
In the new floating-gate-based CSIR, antenna ratio (AR) will affect the peak potential on the coupling gates during plasma charging. In scaled technologies, the parasitic capacitances on the connection and coupling structure are expected to reduce, leading to less AR sensitivity on the recording results. The reasons for leading to such an obvious difference to AR effect between CSIR and conventional detectors are as followed. In these floating gate recorders, the charge accumulated on the antenna, Q P will not be discharged through the channel area. Increased Q P raises V CG, leading to electron injection or ejection into/from the floating gates. As shown in the simulated capacitance in the Fig. 11a, the capacitance of the antenna, C ant, increases proportionally to antenna area, A ant. With the total charge current directly proportional to antenna area, increased AR in a CSIR will not affect the potential on the antenna. Measurement data reveal that the Q FG level remains about the same for SCIR with AR exceeding 100×.
This feature not only saves test pattern area, but also enables finding J P (x,y) with higher spatial resolution for studying patterning effect on PID. Besides, a detector with small antenna can facilitate the design of test patterns for understanding PID in middle-end of the line (MEOL) and contact processes.
Finally, performance summary of the new CSIR for monitoring PID in advanced BEOL FinFET process is summarized in Table 2. The sense range of the traditional detector is AR, while the sense range of the new in situ recorder is based on floating gate length. Further, the area of new in situ recorder can be very small. Most importantly, the new CSIR can provide the real-time feedback of plasma process and separate levels of ion charging and electron charging rate, independently.
A novel charge splitting in situ recorder (CSIR) for monitoring plasma-induced damage is first-time proposed and demonstrated. The CSIR provides a powerful tool for understanding electron charging and ion charging rates in a plasma chamber simultaneously. Wafer maps can facilitate further study between the correlation to device reliability and these individual charging effects.
Wang Y, Han Q, Zhang H (2017) Study of poly etch for performance improvement with alternative spin-on materials in FinFET technology node. China Semiconductor Technology International Conference (CSTIC), Shanghai, pp 1–3
Wang Z, Tanner P, Salm C, Mouthaan T, Kuper F, Andriesse M, van der Drift E (1999) Plasma-induced charging damage of gate oxides. STW, Mierlo, p 593
Han Q-H, Meng X-Y, Zhang H-Y (2015) Challenges and solutions to FinFET gate etch process. China Semiconductor Technology International Conference, Shanghai, pp 1–3
Eriguchi K, Ono K (2008) Quantitative and comparative characterizations of plasma process-induced damage in advanced metal–oxide–semiconductor devices. J Phys D 41:024002
Ma S, MVittie JP (1996) Prediction of plasma charging induced gate oxide tunneling current and antenna dependence by plasma charging probe, Processing of 1st P2ID, p 20
McVittie JP (1997) "Process charging in ULSI: mechanisms, impact and solutions," International Electron Devices Meeting. IEDM Technical Digest. Washington, DC, USA, pp. 433-436.
Eriguchi K, Nakakubo Y, Matsuda A, Takao Y, Ono K (2009) Plasma-induced defect-site generation in Si substrate and its impact on performance degradation in scaled MOSFETs. IEEE Electron Device Lett 30:1275–1277
Li X, Liu C (2016) The alignment performance study for the gate layer in FinFet processes. China Semiconductor Technology International Conference (CSTIC), Shanghai, pp 1–3
Wang Z, Ackaert J, Scarpa A, Salm C, Kuper FG, Vugts M (2005) “Strategies to cope with plasma charging damage in design and layout phases,” International Conference on Integrated Circuit Design and Technology. ICICDT., International, pp. 91–98
Dostalik WW, Krishnan S, Kinoshita T, Rangan S (1993) “Electron shading effects in high density plasma processing for very high aspect ratio structures,” 3rd International Symposium on Plasma Process-Induced Damage, International, pp. 160–163
Cho C, Kim D, Choo H, Park I (2007) “Effect of the substrate, metal-line and surface material on the performance of RFID tag antenna,” IEEE Antennas and Propagation Society International Symposium. Honolulu, HI. pp. 1761–1764
Oates AS (2003) Reliability issues for high-k gate dielectrics. IEEE International Electron Devices Meeting 2003, Washington, pp 38.2.1–38.2.4
Okada K, Ota H, Nabatame T, Toriumi A (2007) Dielectric breakdown in high-K gate dielectrics—mechanism and lifetime assessment. IEEE International Reliability Physics Symposium Proceedings. 45th Annual, Phoenix, pp 36–43
Ramprabu G, Nagarajan S (2013) Performance analysis of communication processor for WMSN. Int J Adv Res Comput Sci 4(2):209–212
Hatta SWM, Soin N, Rahman SHA, Wahab YA, Hussin H (2014) “Effects of the fin width variation on the performance of 16 nm FinFETs with round fin corners and tapered fin shape”, IEEE International Conference on Semiconductor Electronics (ICSE2014). Kuala Lumpur, pp 53-536
Eriguchi K, Takao Y, Ono K (2014) “A new aspect of plasma-induced physical damage in three-dimensional scaled structures—sidewall damage by stochastic straggling and sputtering” IEEE International Conference on IC Design & Technology, Austin, TX. pp. 1–5
Matsuda A, Nakakubo Y, Takao Y, Eriguchi K, Ono K (2013) “Atomistic simulations of plasma process-induced Si substrate damage - Effects of substrate bias-power frequency”, in Proc. 2013 International Conference on IC Design & Technology, (ICICDT), Pavia. pp. 191–194
Liao PJ, Liang SH, Lin HY, Lee JH, Lee Y-H, Shih JR, Gao SH, Liu SE, Wu K (2013) “Physical origins of plasma damage and its process/gate area effects on high-k metal gate technology”, in 2013 IEEE International Reliability Physics Symposium (IRPS), Anaheim, CA, pp. 4C.3.1-4C.3.5.
Eriguchi K, Kamei M, Takao Y, Ono K (2012) “High-k MOSFET performance degradation by plasma process-induced charging damage—impacts on device parameter variation,” Integrated Reliability Workshop Final Report (IRW), South Lake Tahoe, CA, pp. 80-84
Kyung SM, Chang YK, Yoo OS, Byoung JP, Sung WK, Chadwin DY, Heh D, Bersuker G, Lee BH, Yeom GY (2008) “Plasma induced damage of aggressively scaled gate dielectric (EOT≪1.0nm) in metal gate/high-k dielectric CMOSFETs”, in IEEE International Reliability Physics Symposium. Phoenix, AZ, pp. 723-724.
Dick James (2012) “High-k metal gates in leading edge silicon devices,” IEEE Advanced Semiconductor Manufacturing Conference (ASMC). Saratoga Springs, NY, pp. 346-353.
Simon P, Luchies JM, Maly W (1999) “Identification of plasma-induced damage conditions in VLSI designs,” ICMTS 1999. Proceedings of 1999 International Conference on Microelectronic Test Structures (Cat. No.99CH36307). Goteborg, pp. 1-6.
Choi S, Park YJ (2015) Numerical simulation of percolation model for time dependent dielectric breakdown (TDDB) under non-uniform trap distribution. 2015 International Conference on Simulation of Semiconductor Processes and Devices (SISPAD), Washington, pp 405–408
Wu CH, Tsai YP, Lin CJ, King YC (2015) “Mapping of wafer-level plasma induced charge contour by novel on-chip in-situ recorders in advance FinFET technologies,” IEEE International Electron Devices Meeting (IEDM). Washington, DC. pp. 7.1.1–7.1.4
Tsai YP, Wu CH, Lin CJ, King YC (2016) Wafer-level mapping of plasma-induced charging effect by on-chip in situ recorders in FinFET technologies. IEEE Trans Electron Devices 63:2497–2502
Wafer Charging Bulletin on Wafer Charging Monitors, Inc. (2004) Woodside, CA, U.S.A. Vol. 7, No. 1 (http://www.charm-2.com).
Z. Wang 2004 “Detection of and protection against plasma charging damage in modern IC technologies,” PhD Thesis. ISBN 90–365–2079-7
Goebel DM, Katz I (2008) Fundamentals of Electric Propulsion: Ion and Hall Thrusters (Wiley, New York).
Yeh MS et al (2014) Comprehensive study of n-channel and p-channel twin poly-Si FinFET nonvolatile memory. IEEE Trans Nanotechnol 13(4):814–819
Chen FF (1996) Plasma-induced oxide damage: a status report. Department of Electrical Engineering, UCLA
Loffler R, Haffner M, Visanescu G, Weigand H, Wang X, Zhang D, Fleischer M, Meixner AJ, Fortagh J, Kern DP (2011) Optimization of plasma-enhanced chemical vapor deposition parameters for the growth of individual vertical carbon nanotubes as field emitters. Carbon 49:4197–4203
Eriguchi K, Takao Y, Ono K (2011) A new prediction model for effects of plasma-induced damage on parameter variations in advanced LSIs. IEEE International Conference on IC Design & Technology, Kaohsiung, pp 1–4
Martin A, Bukethal C, Ryden KH (June 2009) Fast wafer level reliability monitoring: quantification of plasma-induced damage detected on productive hardware. IEEE Trans Device Mater Reliab 9(2):135–144
The authors would like to thank the support from the Ministry of Science and Technology (MOST), Taiwan, and Taiwan Semiconductor Manufacturing Company (TSMC).
This study is supported by the Ministry of Science and Technology (MOST), the Taiwan Semiconductor Manufacturing Company (TSMC), and the internal funding of the department.
The authors declare that they have no competing interests.
Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
The original version of this article was revised: Figure 3 was corrected.
A correction to this article is available online at https://doi.org/10.1186/s11671-017-2336-x.
About this article
Cite this article
Tsai, YP., Hsieh, TH., Lin, C.J. et al. Charge Splitting In Situ Recorder (CSIR) for Real-Time Examination of Plasma Charging Effect in FinFET BEOL Processes. Nanoscale Res Lett 12, 534 (2017). https://doi.org/10.1186/s11671-017-2309-0
- Plasma-induced damage
- Advanced FinFET technology
- Charge splitting