Advanced quantitative analysis of epitaxial SiGe composition on production wafer for logic devices

Transcription

1 Received: 25 July 2017 Accepted: 3 October 2017 DOI: /sia.6340 RESEARCH ARTICLE Advanced quantitative analysis of epitaxial SiGe composition on production wafer for logic devices Christopher Penley 1 Michael Walker 1 Michael Wilson 1 Daniel Billingsley 2 Arvind Raviswaran 2 Shane Phillips 1 1 Analysis Engineering, Samsung Austin Semiconductor, Samsung Blvd, Austin, TX 78754, USA 2 Diffusion Engineering, Samsung Austin Semiconductor, Samsung Blvd, Austin, TX 78754, USA Correspondence Christopher Penley, Analysis Engineering, Samsung Austin Semiconductor, Samsung Blvd, Austin, TX 78754, USA. c.penley@samsung.com Quantitative analyses of in situ boron doped SiGe composition on production wafers for 14 and 20 nm logic devices were successfully characterized using advanced, small area Time of Flight Secondary Ion Mass Spectrometry analysis methods. The quantification of dopant levels in SiGe offered an improved method for tool to tool matching, process monitoring, and improvement, and performing this function accurately was necessary to enable advanced SiGe technology development that ensured world class manufacturing requirements were met. The boron concentration was measured with Time of Flight Secondary Ion Mass Spectrometry, based on a single Si 1 x Ge x /Si on silicon substrate standard, which exhibited excellent matching with corresponding inline XRF intensity data. Time of Flight Secondary Ion Mass Spectrometry accurately characterized SiGe composition with adequate sensitivity to detect small boron concentration variations to fine tune process parameters. This provided insight to the relationship between overlap capacitance and Rodlin measurements with SiGe dopant levels and ultimately led to device performance improvement. KEYWORDS boron, FinFET, SiGe, TOF SIMS, XRF 1 INTRODUCTION Selective epitaxial growth (SEG) of boron doped silicon germanium (SiGe) acts as one of the key strain engineering methods of enhancing transistor performance with the continuous reduction of transistor size because of its improved electrical properties. In particular, SiGe is increasingly important in pfet performance and V th tuning. Benefits offered by SiGe for device performance have been recognized by providing compressive p channel stress and increasing hole mobility due to the lattice mismatch between SiGe and the Si substrate. 1,2 The capability to characterize the SiGe composition directly on production wafers is a highly desirable need in semiconductor manufacturing. Literature work has revealed many successful inline metrology methods to determine the Ge concentration via X ray diffraction, X ray photoelectron spectroscopy, and X ray fluorescence (XRF) on production wafers. 3,4 Additionally, there has been some exceptional research and development work on SiGe characterization with a focus on proper quantification methods for Ge concentration via SIMS in large part due to the complicated nature of significant variations in both sputter and ionization yield observed with respect to Ge content. 5-7 However, limited data exists on quantifying SiGe B concentration and the effects on device performance. 7-9 The amount of work on boron quantification in SiGe has been limited, especially on production wafers, due to difficulties of existing inline measurement techniques (ie, μrs and XRF) lacking capabilities to provide long term, high precision B concentration. Time of Flight Secondary Ion Mass Spectrometry (TOF SIMS) was determined to be the most appropriate surface analytical technique for this analysis. Time of Flight Secondary Ion Mass Spectrometry provided quantitative elemental analysis, distribution of each element of interest in a thin film and achieved superior detection sensitivity compared to other analytical techniques previously mentioned. Also, it has the capability to perform small area analysis. The disadvantage of the technique would be the destructive nature when collecting depth profiles, limiting its usefulness in wafer fabrication environments. In 14 and 20 nm logic device technologies, the SiGe process is composed of 3 layers: L1, L2, and L3. The different layers are comprised of varying Ge fraction and boron dopant concentration. L2 layer maintains a high Ge fraction and high boron doped film which 90 Copyright 2017 John Wiley & Sons, Ltd. wileyonlinelibrary.com/journal/sia Surf Interface Anal. 2018;50:90 95.

2 PENLEY ET AL. 91 acts as a compressive stressor layer. The dopant concentration is well known to be a critical contributor to device performance but not fully understood. In this study, the highly critical L2 SiGe films were successfully analyzed on production wafers for 20 nm manufacturing requirements as well as for 14 nm FinFET process development via TOF SIMS with excellent agreement with inline XRF. Through TOF SIMS quantification of boron dopant levels in SiGe on production wafers, device characteristics were effectively controlled and improved using the compositional characterization utilized in this work. determined from the Si 1 x Ge x /Si standard, and B concentration was averaged across the L2 SiGe layer. X ray fluorescence analyses were collected inline using a high power X ray tube, 4 kw Rh anode, providing spectra for boron quantification. The appropriate reference standards were used for XRF analysis to accurately quantify the elemental composition. 3 RESULTS AND DISCUSSION 3.1 TOF SIMS monitoring of boron in SiGe 2 EXPERIMENTAL PROCEDURES Time of Flight Secondary Ion Mass Spectrometry measurements were collected off line using a 30 kev Bi + analysis beam interlaced with a 500 ev O + 2 sputter beam, detecting positive secondary ions of Si +, Ge +, and B +. A low energy electron beam of 20 ev was used for charge compensation while profiling SiGe films. Small area depth profiles were acquired on nonpatterned optical scatterometery (OS) sites with dimensions of approximately 60 μm 80μm. The OS sites of interest contain all the 3 SiGe layers (ie, L1, L2, and L3) deposited directly on silicon substrate. The analysis method used the distinct features of TOF SIMS, allowing simultaneous collection of mass spectra, depth profile, and image information with the capability to reconstruct each from only the area of interest post data collection. In this study, the initial data collection extended beyond the dimensions of the OS site of interest and only data for the SiGe layers were reconstructed using a region of interest centered within the OS site. With the larger data collection area, some favorable by products included: maintaining an adequate dose density ratio between Bi + and O + 2 while also maintaining sufficient data point density within the SiGe film of interest, captured SiGe film uniformity across OS sites, maximized area of OS site characterized using region of interest, and improved the consistency of quantifying the B concentration from the center of the OS site. It was found to be extremely precise and reproducible for boron quantification in the SiGe layer. Depth calibration for SIMS profiles were based on the thickness of Si 1 x Ge x standard films as measured with transmission electron microscopy. SIMS analysis confirmed the sputter rate increased as the Ge content increased, this sputter rate response was consistent with literature. 5,6 Since the relative sensitivity factor for boron was expected to vary along with the matrix composition of SiGe due to ion yield dependence on Ge concentration, 7 multiple boron doped Si 1 x Ge x standards, with varying Ge concentration ranging from 20 to 60 at. %, were used to quantify possible B concentration error across a specific Ge concentration range, of less than 5 at. %, required for this work. The error in B concentration induced by ionization yield effects from varying Ge concentration was determined to be negligible for this study. Thus, in this work, the quantitative analysis of B and Ge concentration with TOF SIMS was accomplished through the single standard approach (ie, for all measured SiGe films on production wafers a Si 1 x Ge x /Si standard was used). The calculated relative sensitivity factor was Establishing a quantitative measurement technique for monitoring the SiGe process was essential for manufacturing 20 nm and the development of 14 nm process nodes. The development of TOF SIMS analysis on a production wafer provided an additional SiGe process control metric measuring B concentration in the film and offered support for tool to tool matching (TTTM). Time of Flight Secondary Ion Mass Spectrometry B concentration measurements on production wafers required adequate precision to be used as a long term SiGe process monitoring candidate. However, determining the precision on production wafers was difficult due to the destructive nature of SIMS and wafer uniformity concerns. Thus, the precision was determined by collecting and comparing depth profiles from respective adjacent OS sites, with a similar SiGe film stack, located within the same die shot. The full dataset included the analysis of 4 wafers using at least 2 wafer die shots per wafer across multiple days. An RSD mean of less than 1% for the B concentration measurement was calculated between respective OS sites, sufficient for monitoring SiGe dopant levels. Time of Flight Secondary Ion Mass Spectrometry depth profiles confidently differentiated the L3 and L2 SiGe films (L1 was masked between the L2 and Si substrate). However, TOF SIMS was only able to provide reliable B and Ge concentration for L2 SiGe layer (L3 lied within the surface transient region). Figure 1 shows a typical set of TOF SIMS depth profiles collected during weekly monitoring for TTTM. The depth profiles presented in Figure 1 were a collection of SiGe profiles from 20 SiGe process chambers, including wafer center and mid (approximately 90 mm off center) locations, a total of 40 TOF SIMS depth profiles were overlaid. Through process development and months of TOF SIMS B concentration monitoring in SiGe, a statistical process control chart was established to monitor and control the SiGe process. Using the established control limits, process variation could be differentiated between common cause (normal process variation) and special cause (assignable cause). Normal process variations were observed from tool to tool, chamber to chamber, and with respect to time. An example of B concentration TOF SIMS depth profile shifts observed during process monitoring is shown in Figure 2 for. A total of 16 OS sites measured from including 4 chambers, 2 wafer locations, and from subsequent weeks were overlaid. When reviewing the data initially for this specific tool, it was difficult to determine if the difference was due to normal process variation or not. After reviewing additional TOF SIMS depth profiles and the statistical process control chart across multiple SiGe toolsets within the same time span, a true shift in B concentration was noted across the entire SiGe toolset fleet

3 92 PENLEY ET AL. a root cause for the concentration shift. After a thorough investigation, the root cause was identified to be a change in the incoming B 2 H 6 gas, as shown in Figure 4. A fluctuation of less than 2% in the B 2 H 6 gas flowed into the SiGe toolsets during the specific time period identified in Figures 2 and 3 causing the elevated B concentration. Time of Flight Secondary Ion Mass Spectrometry SiGe analysis on production wafers provided the precision to identify a B concentration shift and lead to determining the root cause. Thus, a corrective action could be implemented for the out of control signals to reduce the variation of product quality and prevent reoccurrence. 3.2 Inline XRF calibration for boron FIGURE 1 Time of Flight Secondary Ion Mass Spectrometry depth profiles collected for weekly tool to tool matching monitoring, multiple tools, on 20 nm production wafers within optical scatterometery sites A group of 20 nm production wafers processed with varying B 2 H 6 and GeH 4 gas flow rates were used to measure a range of boron dopant levels in SiGe films. Time of Flight Secondary Ion Mass Spectrometry B concentration was calculated from depth profiles collected within OS sites and compared with the boron intensity from XRF. As shown in Figure 5, positive correlation was observed between TOF SIMS and XRF across a larger B concentration range. The calibration work assisted in implementing an empirical model to gauge B concentration through inline XRF measurements on production wafers thus opening up the possibility of monitoring B concentration inline. Although, a couple of possible outliers were noted within the group of wafers analyzed. Further work, reported in subsequent sections, used for fine tuning the SiGe process through varying process flow rates and temperatures revealed additional concerns of using inline XRF to measure small differences in B concentration. 3.3 TOF SIMS within wafer uniformity of SiGe With the proven precision of TOF SIMS B concentration measurements on 20 nm production wafers for TTTM and for root cause analysis, within wafer uniformity of SiGe dopants were examined. Boron concentration uniformity information in SiGe on production wafers would be favorable to help reach consistent yield and performance across a wafer. Although small B concentration differences were expected across a 300 mm wafer, it was valuable information to quantify the real variation to implement process adjustments to mitigate the effect on device performance. Two 14 nm production wafers were analyzed by acquiring depth profiles from multiple OS sites spread out radially across the wafer. As shown in Figure 6, similar within wafer uniformity variation was observed for both production wafers. Each wafer noted a linear decrease when moving radially outward from the wafer center to wafer edge. FIGURE 2 Time of Flight Secondary Ion Mass Spectrometry depth profiles collected for weekly tool to tool matching monitoring, same tool, on 20 nm production wafers within optical scatterometery sites as well. Figure 3 illustrates a control chart from a selected time period of 2 months from 3 SiGe toolsets displaying normal process variation and a true shift in B concentration, indicated by the red arrows. With a shift in B concentration across all toolsets and chambers, on average 13% higher, causing each tool to exceed not only control limits but also, in some cases, the upper spec limit required identifying 3.4 TOF SIMS SiGe process development characterization The TOF SIMS analysis method was successfully implemented to monitor the SiGe process for manufacturing 20 nm production wafers. The same analysis methodology was then transferred to assist in 14 nm SiGe process development. In general, the 20 and 14 nm SiGe processes are similar with each device node comprised of 3 SiGe layers of varying Ge fraction and boron dopant concentration. However, as devices shrink, the ability to control process variations and to fine tune

4 PENLEY ET AL. 93 FIGURE 3 An example of using Time of Flight Secondary Ion Mass Spectrometry for SiGe boron concentration process monitoring on 20 nm production wafers. (USL, upper spec limit; UCL, upper control limit; LCL, lower control limit; LSL, lower spec limit) FIGURE 4 B 2 H 6 gas flow monitoring charts for multiple process chambers observed a similar change point aligned with the measured increase in Time of Flight Secondary Ion Mass Spectrometry boron concentration 1.20 Normalized SIMS Boron (at/cm 3 ) R² = Normalized XRF Boron (kcps) Boron Conc. (at/cm 3 ) R² = R² = Distance from Wafer Center (mm) FIGURE 5 Comparison of boron content in SiGe films on 20 nm production wafers measured with off line Time of Flight Secondary Ion Mass Spectrometry and inline X ray fluorescence (XRF) each individual process becomes even more critical for large scale manufacturing of smaller nodes. To optimize the SiGe process for manufacturing 14 nm devices, it was essential to fully understand the effect of adjustments in gas flow rates and process temperatures. Both SiGe process gas flow rate and temperature relationships with SiGe composition were examined via TOF SIMS. The relationship between B 2 H 6 gas flow rate and B concentration was investigated by TOF SIMS on 20 and 14 nm devices. As expected, Figure 7 showed, FIGURE 6 Within wafer uniformity boron concentration distribution for 14 nm production wafers the incorporation of boron in SiGe increased linearly with flow rate. For both 20 and 14 nm devices, the boron sensitivity was approximately a 3% change in B concentration for a 1% adjustment in B 2 H 6 flow, within the specific process window. In Figure 8, TOF SIMS B concentration showed a strong correlation with L2 SiGe process temperature. The 2 variables noted an inversely proportional relationship within the operating process window where the B concentration decreased with increasing L2 SiGe

5 94 PENLEY ET AL. Normalized Boron Conc. (at/cm 3 ) nm WFR CTR nm WFR MID 20 nm WFR CTR nm WFR MID Normalized B 2 H 6 Flow Rate (sccm) FIGURE 7 Boron concentration dependence on B 2 H 6 flow rate for 14 and 20 nm production wafers Normalized Boron Conc. (at/cm 3 ) R² = R² = WFR CTR WFR MID Normalized L2 SiGe Temp. ( C) Spectrometry analysis on production wafers provided insight to the SiGe film composition at varying process temperatures. Understanding the effects of varying the process temperature on boron dopant levels allowed for SiGe deposition temperature to be increased; thus, reducing deposition time and increasing wafer throughput. 3.5 Device characterization Accurate modeling of device performance with respect to process parameters is essential. For 14 nm FinFET, it was determined dopant levels in SiGe were increasingly more significant for device performance compared with 20 nm. Thus, it was necessary to optimize the SiGe process under tighter control limits across the entire fleet of tools to reduce PMOS device variation. As shown previously in Figures 7 and 8, TOF SIMS analysis results illustrated L2 SiGe B concentration had a linear response to adjustments to B 2 H 6 flow and an inversely proportional relationship to SiGe process temperature, respectively. With these correlations known, B concentration splits were set up, processed, and characterized to mitigate device variation through improving TTTM. In Figures 9 and 10, the impact of B concentration on device performance was characterized during 14 nm device development through varying B 2 H 6 flow. The decrease in S/D doping through B 2 H 6 flow rate reduction during L2 SiGe growth reduced FIGURE 8 Boron concentration dependence on L2 SiGe process temperature for 14 nm production wafers temperature; the inverse was true for Ge concentration. The dopant trend suggested a couple of temperature dependent processes: (1) desorption of B 2 H 6 at higher temperatures, limiting boron incorporation into SiGe film and (2) influence on GeH 4 incorporation, which influenced the interaction of B 2 H 6 on the SiGe film. Careful examination of multiple boron dopant profiles suggested less uniform distribution through L2 layers at lower process temperatures. This was supported by the fact that boron profiles from wafer center locations were consistently less uniform and it was well documented that the wafer center was the coldest area in the SiGe process chamber; also, this phenomenon was confirmed by the fact higher B concentration was noted at wafer center locations. The ability to optimize the temperature for the SiGe deposition process while maintaining required film properties for device performance plays a vital role in a manufacturing environment. Time of Flight Secondary Ion Mass Tool A (B Adj) FIGURE 10 Comparison of C ov vs Rodlin, C ov Rodlin relationship between Tool A and (B Adj) indicates comparable device performance after SiGe process adjustment C ov (ff/µm) Tool A (B Adj) FIGURE 9 Comparison of C ov dependence for SiGe boron concentration by varying B 2 H 6 flow rate, (B Adj), to improve tool totool matching device performance

6 PENLEY ET AL. 95 overlap capacitance (C ov ) between the gate and the heavily doped S/D and increased transistor resistance (Rodlin). ORCID Christopher Penley 4 SUMMARY The manufacturing of advanced logic nodes relies on the delivery of a consistent SEG process. SiGe boron concentration was proven to serve as a strong indicator of device performance; therefore, the monitoring of B concentration in SiGe was an effective and quantitative measure for process control via off line TOF SIMS and inline XRF. For 14 nm FinFETs, optimizing the SEG of SiGe becomes more challenging and requires Ge and B concentration characterization to improve device performance. Mitigating SiGe variability is necessary to reduce device variation. The previous approach used for 20 nm SiGe process control relied on inline optical measurements to monitor critical device dimensions. During 14 nm development and the early stages of manufacturing, it was determined the previous inline monitoring methods were not going to suffice for controlling the SiGe process and for TTTM. In this body of work, TOF SIMS analysis was able to fill the void by successfully quantifying B concentration on production wafers with adequate sensitivity required to directly relate dopant levels to device performance, ie, C ov and Rodlin characteristics. For production wafers, TOF SIMS was recommended for measuring boron dopant levels to fine tune the SiGe process for device performance and inline XRF could possibly be examined for inline monitoring for gross differences in the future. Further examination will be required to determine the long term stability of inline XRF for monitoring B concentration and how well it could monitor boron concentration for the SiGe process. REFERENCES 1. Lander R, Ponomarev Y, van Berkum J, de Boer W. High hole mobilities in fully strained Si 1 x Ge x layers. IEEE Trans Electron Devices. 2001;48: Shimamune Y, Fukuda M, Koiizuka M, et al. Technology breakthrough of low temperature, low defect, and low cost SiGe selective epitaxial growth. Symp on VLSI Tech. 2007; L'herron B, Loubet N, Liu Q, et al. Silicon germanium (SiGe) composition and thickness determination via simultaneous small spot XPS and XRF measurements. ASMC, th Annual Semi. 2014; Hung PY, Kasper N, Nadeau J, Ok I, Hobbs C, Vigliante A. Application of inline high resolution x ray diffraction in monitoring Si/SiGe and conventional Si in SOI fin shaped field effect transistor processes. J Vac Sci Technol B. 2012;30: Jiang ZX, Kim K, Lerma J, et al. Quantitative SIMS analysis of SiGe composition with low energy O + 2 beams. Appl Surf Sci. 2006;252: Zhu Z, Ronsheim P, Turansky A, et al. SIMS quantification of SiGe composition with low energy ion beams. Surf Interface Anal. 2011;43: Py M, Barnes JP, Lafond D, Hartmann JM. Quantitative profiling of SiGe/ Si superlattices by time of flight secondary ion mass spectrometry: the advantages of the extended full spectrum protocol. Rapid Commun Mass Spectrom. 2011;25: Hashimoto T, Tokunaga K, Fukumoto K, et al. SiGe HBT technology based on a 0.13 μm process featuring an f max of 325 GHz. J Electron Devices Soc. 2014;2: Eguchi S, Miyashita I, Kagotoshi Y, Toyoda H, Kanai A, Machida N. A study of boron concentration uniformity in selective epitaxial growth for SiGe HBT. SiGe Technology and Device Meeting, ISTDM Third International. 2006; 1 2. ACKNOWLEDGEMENTS The authors would like to thank various experts for their support and background discussions during the different phases of this work: Bang Won Kim and Patrick Aniekwu at Samsung Austin Semiconductor for Metrology analysis support. A special thank you to the Samsung Austin Semiconductor management and technical staff for their support of this project as well. How to cite this article: Penley C, Walker M, Wilson M, Billingsley D, Raviswaran A, Phillips S. Advanced quantitative analysis of epitaxial SiGe composition on production wafer for logic devices. Surf Interface Anal. 2018;50: doi.org/ /sia.6340