- Email: [email protected]
Practical quantitative scanning microwave impedance microscopy
MR-12595; No of Pages 4 Microelectronics Reliability xxx (2017) xxx–xxx
Contents lists available at ScienceDirect
Microelectronics Reliability journal homepage: www.elsevier.com/locate/microrel
Practical quantitative scanning microwave impedance microscopy Oskar Amster a,⁎, Fred Stanke a, Stuart Friedman a, Yongliang Yang a, St.J. Dixon-Warren b, B. Drevniok b a b
PrimeNano, Inc., 8701 Patrick Henry Dr., Bidg 8, Santa Clara, CA 95054, USA TechInsights, 1891 Robertson Road, Ottawa, Ontario K2H 5B7, Canada
a r t i c l e
i n f o
Article history: Received 27 May 2017 Received in revised form 17 July 2017 Accepted 18 July 2017 Available online xxxx Keywords: Scanning microwave impedance microscopy Scanning microwave microscopy Nanoscale C-V curves Electrical characterization doping concentration
a b s t r a c t Scanning microwave impedance microscopy (sMIM) is an emerging technique that has the potential to displace conventional scanning capacitance microscopy (SCM), and other electrical scanning probe microscopy (SPM) techniques, for the profiling of dopants in semiconductor samples with sub-micron spatial resolution. In this work, we consider the practical application of sMIM for quantitative measurement of the dopant concentration profile in production semiconductor devices. We calibrate the sMIM using a doped calibration sample prior to performing the measurements on an “unknown” production device. We utilize nanoscale C-V curves to establish a calibration curve for both n- and p-type carriers in a single reference and apply the calibration curve to an “unknown” device presenting the measurements in units of doping concentration. The calibrated results are compared to SRP measurements on the same area of the device. © 2017 Elsevier Ltd. All rights reserved.
1. Introduction Scanning probe microscopy (SPM) based electrical measurement techniques [1], such as scanning capacitance microscopy (SCM) [2], scanning spreading resistance (SSRM) [3] and scanning microwave microscopy (SMM) [4] have shown value for the profiling of dopants in semiconductor samples, but suffer from significant limitations. Conventional SCM is performed using a resonant circuit originally invented by RCA in the 1980s [3]. SSRM is restricted to samples that provide a DC path from the probe tip to ground. Scanning microwave impedance microscopy (sMIM) is an emerging technique that has the potential to displace conventional SCM, and other electrical SPM techniques, for the profiling of dopants in semiconductor samples with deep sub-micron spatial resolution [5]. Application of microwave frequencies to image doped semiconductors has a long history, but only more recently been successfully integrated with AFM for imaging. 1.1. Experimental arrangement Fig. 1 shows a schematic diagram of sMIM hardware. The sMIM measurements discussed here were performed using commercially available atomic force microscopes (AFM) configured to capture the output signals from the sMIM hardware [6]. The sMIM electronics couples a 3 GHz microwave signal to the AFM probe tip via a 50-Ω impedance interface module and a micro-fabricated coaxial transmission line in a ⁎ Corresponding author. E-mail address: [email protected] (O. Amster).
custom-designed, shielded, metal SPM probe. The reflected microwave signal is amplified and demodulated by the sMIM electronics providing two analog signals representing the Real and Imaginary components of the complex reflection coefficient into the SPM controller where it is synchronized with the topography during scanning [7]. With the addition of a lock-in amplifier, the sMIM can provide the dC/dV amplitude and phase signals correlating with the doping concentration and carrier type, respectively, of a doped semiconductor analogous to SCM. In an sMIM experiment, microwaves are coupled through a custom AFM cantilever to the probe [8] where they interact as evanescent waves with the portion of the sample immediately under the tip. A fraction of the microwaves is reflected and the amplitude and phase (or equivalently, the real and imaginary parts) of the reflection are determined by the local electrical properties of the sample. For a linear sample the permittivity and conductivity determine the reflection, while for non-linear sample like a doped semiconductor, the tip-bias-dependentdepletion-layer structure contributes significantly. SMM is a related technique that also uses microwave interactions to probe electrical properties. Recent work has shown promising preliminary results despite using unshielded probes [9]. The sMIM-C data improves with a dual pass scan when used for quantification applications. The sample is scanned using contact-mode where the first pass follows the surface topography and a second pass is made 100's of nanometers above the sample surface following the same topography forming a constant offset height image. The difference of these two images is used for subsequent analysis for quantification. The difference image reduces effects from stray capacitance and drift. As a result, sMIM measurements can provide valuable nano-scale
http://dx.doi.org/10.1016/j.microrel.2017.07.082 0026-2714/© 2017 Elsevier Ltd. All rights reserved.
Please cite this article as: O. Amster, et al., Practical quantitative scanning microwave impedance microscopy, Microelectronics Reliability (2017), http://dx.doi.org/10.1016/j.microrel.2017.07.082
2
O. Amster et al. / Microelectronics Reliability xxx (2017) xxx–xxx
Fig. 1. Schematic of the sMIM electronics, matching circuit with shielded coaxial line to the probe-sample interface.
information about semiconductor devices, processes and defects. Results of dopant-level characterization are presented here. Fig. 3. The numerical calibration derived from linear fits to the doped regions for both ntype and p-type. The specific points are taken from the sMIM-C values at 0 V bias as shown in Fig. 2b, having been extracted from values in Fig. 2a along the line between the asterisks.
2. Discussion of results 2.1. sMIM-C quantification validation Quantitative profiling of the dopants in a semiconductor sample with sMIM requires calibration of the sMIM-C signal. The most practical method uses a “known” dopant calibration sample, such as those available from IMEC [10] and Infineon [11]. Fig. 2 shows sMIM-C, C-V response for an Infineon dopant calibration sample where the fast-axis is x-position and the slow-axis is DC bias. The vertical black dashed lines mark the center of the doped regions in x. A black set of asterisks mark the scan at which a 0 V bias occurs. Fig. 2b. shows the profile of sMIM-C acquired at that bias. This profile has peaks in sMIM-C at the x-locations of the doped regions. These peaks of sMIM-C give the “log-linear” calibration data for sMIM-C in terms of log doping, as shown in Fig. 3. The numerical calibration derived from linear fits shown in Fig. 3 are: sMIMP ¼ 0:176 log10 ðNA Þ−2:48; CC ¼ 0:993
ð1Þ
sMIMN ¼ 0:183 log10 ðND Þ−2:55; CC ¼ 0:999
ð2Þ
other. The points for the two lowest doping levels are likely corrupted by noise. In retrospect, longer integration times would have been appropriate. Fig. 3 clearly shows a monotonic behavior for both p-type and n-type dopants on same axis. The higher doping concentrations are to the right of the chart, giving a higher sMIM-C signal, and the less doped material gives a lower signal. Fig. 4 shows sMIM-C, C-V curves extracted along the vertical lines in Fig. 2; a) for the p-type regions and b) for the n-type regions. These data are proportional to C-V, with accumulation on the left and inversion on the right for the p-type and accumulation on the right and inversion on the left for the n-type regions. The values in accumulation were forced to be the same a posteriori, for ease of comparison to the reader.
The fits neglect the points for the highest doping concentrations, of both N- and P-types, because they are next to each other at in the center of the collection of doped stripes, and corrupt the measurements of each
Fig. 2. Data collected as an image, a) has the fast-scan axis as x and the “slow-scan axis” as DC tip bias. Black asterisks mark the scan line that is at a bias of 0 V in panel a, And panel b shows the profile of sMIM-C acquired at that bias. These peaks of sMIM-C give the “loglinear” calibration data for sMIM-C in terms of log doping, as shown in the figure below.
Fig. 4. sMIM-C extracted along vertical lines in Fig. 2 for a) p-type and b) n-type doped regions as a function of bias. These data are proportional to C(V), with accumulation on the left and inversion on the right for the p-type and vice-versa for the n-type. The values in accumulation were forced to be the same a posteriori.
Please cite this article as: O. Amster, et al., Practical quantitative scanning microwave impedance microscopy, Microelectronics Reliability (2017), http://dx.doi.org/10.1016/j.microrel.2017.07.082
O. Amster et al. / Microelectronics Reliability xxx (2017) xxx–xxx
3
2.2. Practical application of sMIM-C quantification Once the instrument is calibrated, an “unknown” sample could be profiled with sMIM-C, and the dopant concentrations extracted in an empirical manner using the calibration chart. In addition, the results of the sMIM-C measurements could be cross referenced with other techniques, such as SRP or SIMS. The SRP technique provides accurate profiling of semiconductor dopants; but the spatial resolution is very limited [12]. The following example uses the same methodology described previously using a single planar doped reference sample with two doped calibration samples in cross-section, one p-type and the other n-type. Fig. 5 shows the sMIM-C profiles obtained on samples of a) the IMEC T8 (P-type) and b) the IMEC T9 (N-type) dopant staircase samples. A monotonic behavior is clearly observed, with the high-doped material, at the right of each chart, giving a higher sMIM-C signal, and the lowdoped material giving a lower signal. These profiles were obtained under identical scanning conditions. Calibration is required each time the SPM probe is replaced. Representative calibration curves shown in Fig. 6 are generated from the sMIM data in Fig. 5 correlated with the dopant concentration values for each step in the T8 and T9 staircase samples, as was measured independently by secondary ion mass spectrometry (SIMS). The sMIM-C signal is proportional to the log of the dopant concentration in accordance with previous work done. Fig. 7 shows the cross-sectional views of the NMOS power transistors found on the Linear Technologies LTC3612 die, which is used as the “unknown” sample. Below, Fig. 7(a) shows a cross sectional scanning electron microscopy (SEM) image of the NMOS transistor structure, while (b) and (c) show the corresponding sMIM-C and SCM dC/dV phase images of the same area. The dC/dV phase image provides direct information on the dopant type, with the N-type source/drain diffusions yielding a negative (yellow) signal and the P-type well and substrate yielding a positive (blue-purple) signal. The sMIM-C image shows additional information about the device structure. The polysilicon gate is well defined as well as the gradient from N++ to N+ diffusion region as well
Fig. 6. Calibration curves calculated from graphing the known doping concentrations of the two reference samples, a) IMEC T8 (P-type) on the top and b) T9 (N-type) on the bottom, doped staircase reference samples with the measured sMIM-C average values of each plateau from the extracted profiles shown in Fig. 5.
as the P well and the substrate region material features which are not present in the lock-in image. Fig. 8a presents the results of the sMIM image in units of log doping using the calibration data from the IMEC reference sample. Fig. 8b is a line profile extracted from the scaled image where the data is presented in log of dopant concentration with a comparison to a SRP measurement made on the same area. Agreement between sMIM and SRP is better than an order of magnitude. Since the dopant concentration in a semiconductor device can vary by up to six orders of magnitude, this is useful information at the dramatically better spatial resolution for sMIM as compared to SRP. The dopant concentration in the N + source/drains was also estimated from the sMIM data, and a value of 1018 cm−3 was found. sMIM-C measurements are higher resolution and can see local variations of a sample due to the real doping or artifacts of the sample prep that can affect the measurement. Systematic uncertainties for measuring the cross-sectioned reference samples and “unknown” devices result from variations in sMIM-C measurements at the edge (tope surface) where there are variations in sample quality and with electromagnetic edge effects. Further work is being done to reduce these and other uncertainties. 3. Conclusion
Fig. 5. sMIM-C Profiles of the a) IMEC T8 and b) T9 Standard Dopant Staircase Samples.
The results presented demonstrate that sMIM-C can be used as a method for quantitative measurements of dopant concentrations on semiconductor devices with nanoscale spatial resolution. The methodology is applied using both a planar doped reference sample with P-type and N-type doped regions as well as cross-section staircase samples. Calibration sMIM-C is based on this linear relationship to the log of doping concentration over a useful range. These results match previous modeling results with finite element analysis. The combination of the sMIM-C image with the dC/dV phase and Amplitude images provide a richer visualization of the device structure's
Please cite this article as: O. Amster, et al., Practical quantitative scanning microwave impedance microscopy, Microelectronics Reliability (2017), http://dx.doi.org/10.1016/j.microrel.2017.07.082
4
O. Amster et al. / Microelectronics Reliability xxx (2017) xxx–xxx
Fig. 7. Cross-sectional analysis of the NMOS power transistors in the Linear Technology LTC3612. A SEM image in a) with details of the area in sMIM-C b) and d(sMIM-C)/dV phase image in c).
materials variations such as dielectric materials, highly doped features, variations in doping concentration, as well as the metallic regions. The sMIM technique does not yet achieve the accuracy of conventional SRP (±20%). sMIM does provide better than 0.5 order-of-magnitude measurements of the dopant concentration, with the resolution to differentiate qualitatively on the order of 0.1 order-of-magnitude doping concentration. The technique provides a much higher spatial resolution over SIMS and SRP.
Acknowledgements Supported in part by Dept. of Energy SBIR DE-SC0009856. References [1] G. Binnig, C.F. Quate, C. Gerber, Atomic-force microscope, Phys. Rev. Lett. 56 (1986) 930–933. [2] J.R. Matey, J. Blanc, Scanning capacitance microscopy, J. Appl. Phys. 57 (5) (1985) 1437–1444. [3] Enrico Brinciotti, Georg Gramse, Soeren Hommel, Thomas Schweinboeck, Andreas Altes, Matthias A. Fenner, Juergen Smoliner, et al., Probing resistivity and doping concentration of semiconductors at the nanoscale using scanning microwave microscopy, Nanoscale 7 (35) (2015) 14715–14722, http://dx.doi.org/10.1039/C5NR04264J. [4] Yongliang Yang, Keji Lai, Qiaochu Tang, Worasom Kundhikanjana, Michael a Kelly, Kun Zhang, Zhi-xun Shen, Xinxin Li, Batch-fabricated cantilever probes with electrical shielding for nanoscale dielectric and conductivity imaging, J. Micromech. Microeng. (2012) http://dx.doi.org/10.1088/0960-1317/22/11/115040. [5] J.R. Matey, J. Blanc, Scanning capacitance microscopy, J. Appl. Phys. 57 (5) (1985) 1437. [6] B. Drevniok, St.J. Dixon-Warren, O. Amster, S. Friedman, Y. Yang, Extending electrical scanning probe microscopy measurements of semiconductor devices using microwave impedance microscopyISTFA 2015: Conference Proceedings from the 41st International Symposium for Testing and Failure Analysis, November 1–5, 2015, Portland, Oregon, USA, 2015. [7] http://www.primenanoinc.com/smim_wp/wp-content/uploads/2015/02/ PrimeNano-About-sMIM-141107.pdf (retrieved April 6, 2016). [8] Stuart Friedman, Oskar Amster, Yongliang Yang, Recent advances in scanning Microwave Impedance Microscopy (sMIM) for nano-scale measurements and industrial applications, Proc. SPIE 9173, Instrumentation, Metrology, and Standards for Nanomanufacturing, Optics, and Semiconductors VIII, 917308, August 27, 2014. [9] Yongliang Yang, Keji Lai, Qiaochu Tang, Worasom Kundhikanjana, Michael a Kelly, Kun Zhang, Zhi-xun Shen, Xinxin Li, Batch-fabricated cantilever probes with electrical shielding for nanoscale dielectric and conductivity imaging, J. Micromech. Microeng. (2012) http://dx.doi.org/10.1088/0960-1317/22/11/115040. [10] Natasja Duhayon, Experimental study and optimization of scanning capacitance microscopy for two-dimensional carrier profiling of submicron semiconductor devices(PhD Thesis) IMEC vzw Interuniversitair Micro-Elektronica Centrum vzw, Leuven (Heverlee), 2006. [11] Schweinböck, Infineon Dopant Calibration Sample, Product Description Presentation, 2014. [12] SOLECON Labs Technical Note, “How Big a Pattern Do We Need for Spreading Resistance Analysis?”, http://www.solecon.com/pdf/how_big_a_pattern_do_we_need_ for_sra.pdf (Retrieved April 6, 2016).
Fig. 8. a) log (dopant concentration) of an NMOS power device from calibrated sMIM-C measurements b) along with the SRP data obtained on the same device.
Please cite this article as: O. Amster, et al., Practical quantitative scanning microwave impedance microscopy, Microelectronics Reliability (2017), http://dx.doi.org/10.1016/j.microrel.2017.07.082
Contents lists available at ScienceDirect
Microelectronics Reliability journal homepage: www.elsevier.com/locate/microrel
Practical quantitative scanning microwave impedance microscopy Oskar Amster a,⁎, Fred Stanke a, Stuart Friedman a, Yongliang Yang a, St.J. Dixon-Warren b, B. Drevniok b a b
PrimeNano, Inc., 8701 Patrick Henry Dr., Bidg 8, Santa Clara, CA 95054, USA TechInsights, 1891 Robertson Road, Ottawa, Ontario K2H 5B7, Canada
a r t i c l e
i n f o
Article history: Received 27 May 2017 Received in revised form 17 July 2017 Accepted 18 July 2017 Available online xxxx Keywords: Scanning microwave impedance microscopy Scanning microwave microscopy Nanoscale C-V curves Electrical characterization doping concentration
a b s t r a c t Scanning microwave impedance microscopy (sMIM) is an emerging technique that has the potential to displace conventional scanning capacitance microscopy (SCM), and other electrical scanning probe microscopy (SPM) techniques, for the profiling of dopants in semiconductor samples with sub-micron spatial resolution. In this work, we consider the practical application of sMIM for quantitative measurement of the dopant concentration profile in production semiconductor devices. We calibrate the sMIM using a doped calibration sample prior to performing the measurements on an “unknown” production device. We utilize nanoscale C-V curves to establish a calibration curve for both n- and p-type carriers in a single reference and apply the calibration curve to an “unknown” device presenting the measurements in units of doping concentration. The calibrated results are compared to SRP measurements on the same area of the device. © 2017 Elsevier Ltd. All rights reserved.
1. Introduction Scanning probe microscopy (SPM) based electrical measurement techniques [1], such as scanning capacitance microscopy (SCM) [2], scanning spreading resistance (SSRM) [3] and scanning microwave microscopy (SMM) [4] have shown value for the profiling of dopants in semiconductor samples, but suffer from significant limitations. Conventional SCM is performed using a resonant circuit originally invented by RCA in the 1980s [3]. SSRM is restricted to samples that provide a DC path from the probe tip to ground. Scanning microwave impedance microscopy (sMIM) is an emerging technique that has the potential to displace conventional SCM, and other electrical SPM techniques, for the profiling of dopants in semiconductor samples with deep sub-micron spatial resolution [5]. Application of microwave frequencies to image doped semiconductors has a long history, but only more recently been successfully integrated with AFM for imaging. 1.1. Experimental arrangement Fig. 1 shows a schematic diagram of sMIM hardware. The sMIM measurements discussed here were performed using commercially available atomic force microscopes (AFM) configured to capture the output signals from the sMIM hardware [6]. The sMIM electronics couples a 3 GHz microwave signal to the AFM probe tip via a 50-Ω impedance interface module and a micro-fabricated coaxial transmission line in a ⁎ Corresponding author. E-mail address: [email protected] (O. Amster).
custom-designed, shielded, metal SPM probe. The reflected microwave signal is amplified and demodulated by the sMIM electronics providing two analog signals representing the Real and Imaginary components of the complex reflection coefficient into the SPM controller where it is synchronized with the topography during scanning [7]. With the addition of a lock-in amplifier, the sMIM can provide the dC/dV amplitude and phase signals correlating with the doping concentration and carrier type, respectively, of a doped semiconductor analogous to SCM. In an sMIM experiment, microwaves are coupled through a custom AFM cantilever to the probe [8] where they interact as evanescent waves with the portion of the sample immediately under the tip. A fraction of the microwaves is reflected and the amplitude and phase (or equivalently, the real and imaginary parts) of the reflection are determined by the local electrical properties of the sample. For a linear sample the permittivity and conductivity determine the reflection, while for non-linear sample like a doped semiconductor, the tip-bias-dependentdepletion-layer structure contributes significantly. SMM is a related technique that also uses microwave interactions to probe electrical properties. Recent work has shown promising preliminary results despite using unshielded probes [9]. The sMIM-C data improves with a dual pass scan when used for quantification applications. The sample is scanned using contact-mode where the first pass follows the surface topography and a second pass is made 100's of nanometers above the sample surface following the same topography forming a constant offset height image. The difference of these two images is used for subsequent analysis for quantification. The difference image reduces effects from stray capacitance and drift. As a result, sMIM measurements can provide valuable nano-scale
http://dx.doi.org/10.1016/j.microrel.2017.07.082 0026-2714/© 2017 Elsevier Ltd. All rights reserved.
Please cite this article as: O. Amster, et al., Practical quantitative scanning microwave impedance microscopy, Microelectronics Reliability (2017), http://dx.doi.org/10.1016/j.microrel.2017.07.082
2
O. Amster et al. / Microelectronics Reliability xxx (2017) xxx–xxx
Fig. 1. Schematic of the sMIM electronics, matching circuit with shielded coaxial line to the probe-sample interface.
information about semiconductor devices, processes and defects. Results of dopant-level characterization are presented here. Fig. 3. The numerical calibration derived from linear fits to the doped regions for both ntype and p-type. The specific points are taken from the sMIM-C values at 0 V bias as shown in Fig. 2b, having been extracted from values in Fig. 2a along the line between the asterisks.
2. Discussion of results 2.1. sMIM-C quantification validation Quantitative profiling of the dopants in a semiconductor sample with sMIM requires calibration of the sMIM-C signal. The most practical method uses a “known” dopant calibration sample, such as those available from IMEC [10] and Infineon [11]. Fig. 2 shows sMIM-C, C-V response for an Infineon dopant calibration sample where the fast-axis is x-position and the slow-axis is DC bias. The vertical black dashed lines mark the center of the doped regions in x. A black set of asterisks mark the scan at which a 0 V bias occurs. Fig. 2b. shows the profile of sMIM-C acquired at that bias. This profile has peaks in sMIM-C at the x-locations of the doped regions. These peaks of sMIM-C give the “log-linear” calibration data for sMIM-C in terms of log doping, as shown in Fig. 3. The numerical calibration derived from linear fits shown in Fig. 3 are: sMIMP ¼ 0:176 log10 ðNA Þ−2:48; CC ¼ 0:993
ð1Þ
sMIMN ¼ 0:183 log10 ðND Þ−2:55; CC ¼ 0:999
ð2Þ
other. The points for the two lowest doping levels are likely corrupted by noise. In retrospect, longer integration times would have been appropriate. Fig. 3 clearly shows a monotonic behavior for both p-type and n-type dopants on same axis. The higher doping concentrations are to the right of the chart, giving a higher sMIM-C signal, and the less doped material gives a lower signal. Fig. 4 shows sMIM-C, C-V curves extracted along the vertical lines in Fig. 2; a) for the p-type regions and b) for the n-type regions. These data are proportional to C-V, with accumulation on the left and inversion on the right for the p-type and accumulation on the right and inversion on the left for the n-type regions. The values in accumulation were forced to be the same a posteriori, for ease of comparison to the reader.
The fits neglect the points for the highest doping concentrations, of both N- and P-types, because they are next to each other at in the center of the collection of doped stripes, and corrupt the measurements of each
Fig. 2. Data collected as an image, a) has the fast-scan axis as x and the “slow-scan axis” as DC tip bias. Black asterisks mark the scan line that is at a bias of 0 V in panel a, And panel b shows the profile of sMIM-C acquired at that bias. These peaks of sMIM-C give the “loglinear” calibration data for sMIM-C in terms of log doping, as shown in the figure below.
Fig. 4. sMIM-C extracted along vertical lines in Fig. 2 for a) p-type and b) n-type doped regions as a function of bias. These data are proportional to C(V), with accumulation on the left and inversion on the right for the p-type and vice-versa for the n-type. The values in accumulation were forced to be the same a posteriori.
Please cite this article as: O. Amster, et al., Practical quantitative scanning microwave impedance microscopy, Microelectronics Reliability (2017), http://dx.doi.org/10.1016/j.microrel.2017.07.082
O. Amster et al. / Microelectronics Reliability xxx (2017) xxx–xxx
3
2.2. Practical application of sMIM-C quantification Once the instrument is calibrated, an “unknown” sample could be profiled with sMIM-C, and the dopant concentrations extracted in an empirical manner using the calibration chart. In addition, the results of the sMIM-C measurements could be cross referenced with other techniques, such as SRP or SIMS. The SRP technique provides accurate profiling of semiconductor dopants; but the spatial resolution is very limited [12]. The following example uses the same methodology described previously using a single planar doped reference sample with two doped calibration samples in cross-section, one p-type and the other n-type. Fig. 5 shows the sMIM-C profiles obtained on samples of a) the IMEC T8 (P-type) and b) the IMEC T9 (N-type) dopant staircase samples. A monotonic behavior is clearly observed, with the high-doped material, at the right of each chart, giving a higher sMIM-C signal, and the lowdoped material giving a lower signal. These profiles were obtained under identical scanning conditions. Calibration is required each time the SPM probe is replaced. Representative calibration curves shown in Fig. 6 are generated from the sMIM data in Fig. 5 correlated with the dopant concentration values for each step in the T8 and T9 staircase samples, as was measured independently by secondary ion mass spectrometry (SIMS). The sMIM-C signal is proportional to the log of the dopant concentration in accordance with previous work done. Fig. 7 shows the cross-sectional views of the NMOS power transistors found on the Linear Technologies LTC3612 die, which is used as the “unknown” sample. Below, Fig. 7(a) shows a cross sectional scanning electron microscopy (SEM) image of the NMOS transistor structure, while (b) and (c) show the corresponding sMIM-C and SCM dC/dV phase images of the same area. The dC/dV phase image provides direct information on the dopant type, with the N-type source/drain diffusions yielding a negative (yellow) signal and the P-type well and substrate yielding a positive (blue-purple) signal. The sMIM-C image shows additional information about the device structure. The polysilicon gate is well defined as well as the gradient from N++ to N+ diffusion region as well
Fig. 6. Calibration curves calculated from graphing the known doping concentrations of the two reference samples, a) IMEC T8 (P-type) on the top and b) T9 (N-type) on the bottom, doped staircase reference samples with the measured sMIM-C average values of each plateau from the extracted profiles shown in Fig. 5.
as the P well and the substrate region material features which are not present in the lock-in image. Fig. 8a presents the results of the sMIM image in units of log doping using the calibration data from the IMEC reference sample. Fig. 8b is a line profile extracted from the scaled image where the data is presented in log of dopant concentration with a comparison to a SRP measurement made on the same area. Agreement between sMIM and SRP is better than an order of magnitude. Since the dopant concentration in a semiconductor device can vary by up to six orders of magnitude, this is useful information at the dramatically better spatial resolution for sMIM as compared to SRP. The dopant concentration in the N + source/drains was also estimated from the sMIM data, and a value of 1018 cm−3 was found. sMIM-C measurements are higher resolution and can see local variations of a sample due to the real doping or artifacts of the sample prep that can affect the measurement. Systematic uncertainties for measuring the cross-sectioned reference samples and “unknown” devices result from variations in sMIM-C measurements at the edge (tope surface) where there are variations in sample quality and with electromagnetic edge effects. Further work is being done to reduce these and other uncertainties. 3. Conclusion
Fig. 5. sMIM-C Profiles of the a) IMEC T8 and b) T9 Standard Dopant Staircase Samples.
The results presented demonstrate that sMIM-C can be used as a method for quantitative measurements of dopant concentrations on semiconductor devices with nanoscale spatial resolution. The methodology is applied using both a planar doped reference sample with P-type and N-type doped regions as well as cross-section staircase samples. Calibration sMIM-C is based on this linear relationship to the log of doping concentration over a useful range. These results match previous modeling results with finite element analysis. The combination of the sMIM-C image with the dC/dV phase and Amplitude images provide a richer visualization of the device structure's
Please cite this article as: O. Amster, et al., Practical quantitative scanning microwave impedance microscopy, Microelectronics Reliability (2017), http://dx.doi.org/10.1016/j.microrel.2017.07.082
4
O. Amster et al. / Microelectronics Reliability xxx (2017) xxx–xxx
Fig. 7. Cross-sectional analysis of the NMOS power transistors in the Linear Technology LTC3612. A SEM image in a) with details of the area in sMIM-C b) and d(sMIM-C)/dV phase image in c).
materials variations such as dielectric materials, highly doped features, variations in doping concentration, as well as the metallic regions. The sMIM technique does not yet achieve the accuracy of conventional SRP (±20%). sMIM does provide better than 0.5 order-of-magnitude measurements of the dopant concentration, with the resolution to differentiate qualitatively on the order of 0.1 order-of-magnitude doping concentration. The technique provides a much higher spatial resolution over SIMS and SRP.
Acknowledgements Supported in part by Dept. of Energy SBIR DE-SC0009856. References [1] G. Binnig, C.F. Quate, C. Gerber, Atomic-force microscope, Phys. Rev. Lett. 56 (1986) 930–933. [2] J.R. Matey, J. Blanc, Scanning capacitance microscopy, J. Appl. Phys. 57 (5) (1985) 1437–1444. [3] Enrico Brinciotti, Georg Gramse, Soeren Hommel, Thomas Schweinboeck, Andreas Altes, Matthias A. Fenner, Juergen Smoliner, et al., Probing resistivity and doping concentration of semiconductors at the nanoscale using scanning microwave microscopy, Nanoscale 7 (35) (2015) 14715–14722, http://dx.doi.org/10.1039/C5NR04264J. [4] Yongliang Yang, Keji Lai, Qiaochu Tang, Worasom Kundhikanjana, Michael a Kelly, Kun Zhang, Zhi-xun Shen, Xinxin Li, Batch-fabricated cantilever probes with electrical shielding for nanoscale dielectric and conductivity imaging, J. Micromech. Microeng. (2012) http://dx.doi.org/10.1088/0960-1317/22/11/115040. [5] J.R. Matey, J. Blanc, Scanning capacitance microscopy, J. Appl. Phys. 57 (5) (1985) 1437. [6] B. Drevniok, St.J. Dixon-Warren, O. Amster, S. Friedman, Y. Yang, Extending electrical scanning probe microscopy measurements of semiconductor devices using microwave impedance microscopyISTFA 2015: Conference Proceedings from the 41st International Symposium for Testing and Failure Analysis, November 1–5, 2015, Portland, Oregon, USA, 2015. [7] http://www.primenanoinc.com/smim_wp/wp-content/uploads/2015/02/ PrimeNano-About-sMIM-141107.pdf (retrieved April 6, 2016). [8] Stuart Friedman, Oskar Amster, Yongliang Yang, Recent advances in scanning Microwave Impedance Microscopy (sMIM) for nano-scale measurements and industrial applications, Proc. SPIE 9173, Instrumentation, Metrology, and Standards for Nanomanufacturing, Optics, and Semiconductors VIII, 917308, August 27, 2014. [9] Yongliang Yang, Keji Lai, Qiaochu Tang, Worasom Kundhikanjana, Michael a Kelly, Kun Zhang, Zhi-xun Shen, Xinxin Li, Batch-fabricated cantilever probes with electrical shielding for nanoscale dielectric and conductivity imaging, J. Micromech. Microeng. (2012) http://dx.doi.org/10.1088/0960-1317/22/11/115040. [10] Natasja Duhayon, Experimental study and optimization of scanning capacitance microscopy for two-dimensional carrier profiling of submicron semiconductor devices(PhD Thesis) IMEC vzw Interuniversitair Micro-Elektronica Centrum vzw, Leuven (Heverlee), 2006. [11] Schweinböck, Infineon Dopant Calibration Sample, Product Description Presentation, 2014. [12] SOLECON Labs Technical Note, “How Big a Pattern Do We Need for Spreading Resistance Analysis?”, http://www.solecon.com/pdf/how_big_a_pattern_do_we_need_ for_sra.pdf (Retrieved April 6, 2016).
Fig. 8. a) log (dopant concentration) of an NMOS power device from calibrated sMIM-C measurements b) along with the SRP data obtained on the same device.
Please cite this article as: O. Amster, et al., Practical quantitative scanning microwave impedance microscopy, Microelectronics Reliability (2017), http://dx.doi.org/10.1016/j.microrel.2017.07.082