This is the accepted version of the article: Ruiz, Ana; Seoane, Natalia; Claramunt, S.; [et al.]. «Combined nanoscale KPFM characterization and device simulation for the evaluation of the MOSFET variability related to metal gate workfunction fluctuations». Microelectronic Engineering, Vol. 216 (Aug. 2019), art. 111048. DOI 10.1016/j.mee.2019.111048 This version is available at https://ddd.uab.cat/record/249156 under the terms of the license Combined nanoscale KPFM characterization and device simulation for the evaluation of the MOSFET variability related to metal gate workfunction fluctuations A. Ruiz1, N. Seoane2, S. Claramunt1, A. García-Loureiro2, M. Porti1, M. Nafria1 1 Electronic Engineering Department, Universitat Autònoma de Barcelona, Barcelona 08193, Spain 2 CITIUS, Universidade de Santiago de Compostela, Santiago de Compostela 15782, Spain Electronic address: ana.ruiz@uab.cat 1. Summary 3. Experimental set-up. In this work, a more realistic approximation based A 100 nm thick TiN layer grown over a HfO2/Si on 2D nanoscale experimental data obtained on a substrate was used to obtain the experimental data. metal layer is presented to investigate the impact of The layer was fabricated by continuous e-gun the metal gate polycrystallinity on the MOSFET evaporation of metallic TiN. The TiN formation was variability. The nanoscale data (obtained with a ensured by passing the Ti atoms through a reactive Kelvin Probe Force Microscope, KPFM) were nitrogen-enriched atmosphere (nitrogen partial introduced in a device simulator to analyze the effect pressure of 8×10-4 mbars). X-Ray Diffraction (XRD) of a TiN metal gate work functions (WF) fluctuations was used to determine the structure of the TiN layer. on the MOSFET electrical characteristics. The results The XRD spectrum shows a polycrystalline structure demonstrate that the device characteristics are with two peaks [14], that correspond to the [111] and affected not only by the WF fluctuations, but also [200] orientations [9]. being the [111] orientation their spatial distribution, which is specially relevant in dominant over the [200] one. very small devices. The effect on these characteristics of the spatial distribution on the gate area of such The nanoscale morphological and electrical fluctuations is also evaluated. properties of the TiN layer were measured with a Nano-Observer AFM from Concept Scientific Instruments. This technique allows bimodal single 2. Introduction pass AM-KPFM measurements, so that it is possible Extensive studies have been devoted to investigate to obtain during the same scan topographical and variability sources affecting ultra-scaled MOSFET sample-tip contact potential difference (CPD) 2D technologies, as Random Dopant Distributions, Line maps with nanometer resolution [15,16]. Compared to Edge Roughness, high-k dielectric polycrystallization a lift mode based AM-KPFM (normally with a worse [1–5], interface traps [6–8] and metal resolution [17–19], our bimodal single pass AM- polycrystallization [9–11]. Charges captured in traps KPFM can provide a similar resolution to that located at the semiconductor/insulator interface can obtained with a FM-KPFM [17,20] and minimizes lead to fluctuations in the threshold voltage (VT) and cross-talk between topography and the CPD data the on-current [12,13] introducing, therefore, device- [18,21]. Since the CPD image corresponds to the WF to-device variability. Such VT fluctuations also difference between the tip and the sample, assuming a depend on the position of the traps along the channel constant value of the tip WF during the test, the [13]. Regarding metal gates, the metal measured CPD fluctuations can be related to the WF polycrystallization results in grains with different variations of the TiN layer. Therefore, CPD maps give sizes and orientations, which have associated different information of the local value of the WF of the WFs. The random distribution of grains (and the sample. KPFM images were obtained with highly corresponding WF) results in VT variability (TVV) doped Si tips to get a better resolution. [9–11]. Some works have already analyzed the impact of the WF fluctuations on the variability of MOSFETs 4. Experimental results. [9,10]. However, the statistical models that were used Fig. 1a corresponds to a 520 nm × 240 nm make assumptions that might not be representative of topographical image of the TiN layer. It shows a real devices. granular structure, where the Grains (Gs) are In this work, the impact of metal surrounded by grain boundaries (GBs), which polycrystallization on the electrical parameters of correspond to the depressions in the image. Fig. 1b MOSFETs is studied, based on KPFM experimental shows the measured WF map of the same surface WF maps. These results are compared to those region. Note that GBs tend to show lower WFs (WF < obtained using the approximated WF distributions 4.1 eV) than nanocrystals. normally used in customary approaches for this kind We used the open-source software Gwyddion to of variability source. An analysis of the influence of identify the Gs and obtain the TiN granular pattern the location of the WF fluctuations along the channel (Fig. 1c), where the GBs have been neglected. When is also studied. only nanocrystals are considered, Gs with different maximum WF values are measured, as shown in Fig. 2, which corresponds to a 2D-histogram that relates the maximum WF and the maximum height (with 5. Simulations respect to the mean height value of the topographical In order to evaluate the device variability image shown in Fig. 1a) of each grain in Fig. 1c. Note associated to the WF fluctuations of the TiN layer, that although a continuous distribution of maximum different WF distributions representing the metal gate WFs is obtained, WFs are mostly concentrated at ~4.3 of a MOSFET were considered. WF maps as those eV and, with less frequency, around ~4.5 eV, shown in Fig. 1 have been divided to generate 100 suggesting two predominant WFs and indicating a non-overlapping 50 nm × 50 nm gate WF profiles. much higher number of nanocrystals with low WF Next, these profiles were introduced in a 3D in-house- than with high WF. This result, confirmed by XRD built drift-diffusion device simulator [22] as metal [14], is compatible with the presence of two grain gate, to evaluate the device electrical characteristics of orientations in the TiN layer ([111] and [200]) whose MOSFETs with a gate area of 50 × 50 nm2. In order WFs are separated by 200 mV [9] . to compare the results obtained with our methodology, based on nanoscale experimental data, with those shown in the literature, we proposed the study of two cases: (i) the actual WF distribution (as in Fig. 1b, with GBs and a continuous distribution of Gs WF), and (ii) an approximated WF distribution, with two discrete values of the WF, as in Fig. 1d. Note that the last is the customary approach for this kind of TVV source [9]. To generate the approximated WF maps (Fig. 1d), the grain pattern in Fig. 1c is the starting point. We generated an approximated WF maps where GBs have been neglected and, for the Gs in Fig. 1c, two discrete and Fig. 1. Topography (a) and WF (b) maps obtained with constant values, corresponding to the [111] and [200] KPFM on a TiN layer. Binary mask (c) of the grains orientations (confirmed from XRD analysis), were identified in the topographical map and an example only considered. We have assigned 4.3 eV to the generated WF map (d) from the binary mask. Orange color corresponds to 4.5 eV and purple color to 4.3 eV. nanocrystals with WF lower than 4.4 eV and 4.5 eV to those with higher values, leading to WF maps as that shown in Fig. 1d. The MOSFET behavior for devices with metal gate WF distributions obtained from the WF maps as in Fig. 1b and 1d have been compared. Fig. 2. 2D-histogram showing the maximum WF vs maximum height of the grains, obtained from Fig. 1c. In the X axis, the height of each grain is determined with respect Fig. 3. Distributions of IOFF (top), VT (middle) and ION to the mean value of the topographical image (Fig. 1a). (bottom) related to WF fluctuations in 50 nm × 50 nm Si MOSFETs (at a drain bias of 50 mV) when using actual WF The data obtained from KPFM images add new data (left column) an approximated WF (right column) from information on the properties of the polycrystalline maps as those shown in Fig. 1b and 1d respectively. The metal layer, not taken into account in previous works. mean value (< >) and standard deviation (σ) of the statistical ensembles are shown. Besides the dispersion in the WF of the nanocrystals (not only two discrete values are measured), GBs with Fig. 3 shows the off-current (IOFF), threshold lower WF than Gs are also observed, which could also voltage (VT) and on-current (ION) distributions affect the variability of devices. Therefore, all these obtained for these two cases. Note that, the features should be taken into account when studying approximate case (i. e., disregarding the GBs and the the MOSFET variability and when considering the WF continuous distribution), in the right column, WF fluctuations as variability source. leads to a ~10% and ~18% reduction in σVT and leading to variations of VT. A similar effect was σlog(IOFF), respectively, and to a considerable shift observed for other sources of variability, as is the case (~0.12 V in the case of the VT) in the mean value of of interface traps in ultra-scaled MOSFETs [6]. the distributions. To better understand these results, the effect of the spatial location of the Gs in the MOSFET metal gate has also been evaluated. As an example, Fig. 4 shows the gate of two devices in which only one nanocrystal has been included. The average WF of the grains is 4.3 eV for grain (a) and 4.18 eV for grain (b). Since, in this section, we want to evaluate only the impact on VT of the position of the grains in the channel, the WF of the portion of the gate not covered by the nanocrystal has been set to same average value, that is, 4.3 eV for Fig. 4a and 4.18 eV for Fig. 4b. The G has been swept along the device gate, from the source end (X=0 nm) to the drain end (X=50 nm), and for each position of the grain, the device was simulated and the corresponding VT extracted. Fig. 5 shows the VT variation with respect to the homogeneous case when the grain position is swept, at both low and high drain biases. Note that, although the distribution of WF is the same in all cases, the specific location of the G may have a different impact on the VT values (depending on its position along L), leading to both positive and negative threshold voltage shifts with Fig. 5. VT spatial sensitivity to a TiN grain WF fluctuations respect to the grain average WF value. In addition, the when its position is swept along the channel of the Si device is clearly more sensitive to WF variations at MOSFET at both low and high drain biases like those the source end than at the drain end (as seen in Fig. 5 shown in Fig 4a and 4b have seen considered for the for both grains). At the drain end, the presence of the simulation. ΔVT is calculated as the difference between the VT of a device with a gate WF equal to the of the G grain does not have any influence on VT. The drain and the VT of the device that includes the nanocrystal. bias also has an influence on the TVV, as seen on Fig. 5. However, when the grains were swept along the y- direction (i.e. along the width of the gate area), no 6. Conclusion significant changes are observed in VT. When analyzing the impact of the WF fluctuations of polycrystalline metal gates on the MOSFET variability, the presence of GBs and the WF continuous distribution of the Gs need to be considered to obtain more realistic data. Moreover, the device characteristics and variability are also affected by the spatial distribution of such fluctuations: in small devices, where few nanocrystals may be present in the gate area, their location should also be considered in order to correctly estimate the variability at device level. Acknowledgments Fig. 4. Two examples of isolated nanocrystals (enclosed in This work has been partially supported by the pink rectangles), with average WF values of 4.3 eV, grain (a), and 4.18, grain (b). The larger green squares Spanish AEI and ERDF (TEC2016-75151-C3-1-R, show the real dimensions of the 50 × 50 nm device gate. TEC2014-53909-REDT and RYC-2017-23312). References Therefore, the results demonstrate that, not only the WF values of the grains in the metal gate can impact [1] K. Shubhakar, K.L. Pey, S.S. Kushvaha, S.J. O’Shea, N. Raghavan, M. Bosman, M. Kouda, K. Kakushima, H. on the MOSFET variability, but also VT depends on Iwai, Grain boundary assisted degradation and breakdown how those fluctuations are distributed along the study in cerium oxide gate dielectric using scanning channel direction. This dependence of VT on the G tunneling microscopy, Appl. Phys. Lett. 98 (2011) position can be explained taking into account the 072902. doi:10.1063/1.3553190. impact of the WF fluctuations on the electrostatics of [2] V. Iglesias, M. Porti, M. Nafría, X. Aymerich, P. Dudek, the device. Depending on the G location, the potential T. Schroeder, G. Bersuker, Correlation between the distribution in the channel is affected differently, nanoscale electrical and morphological properties of crystallized hafnium oxide-based metal oxide semiconductor structures, Appl. Phys. Lett. 97 (2010) R. Rodriguez, M. Nafria, X. Aymerich, C. Marquez, F. 262906. doi:10.1063/1.3533257. Gamiz, Threshold voltage and on-current Variability related to interface traps spatial distribution, Eur. Solid- [3] C. Couso, M. Porti, J. Martin-Martinez, A.J. Garcia- State Device Res. Conf. 2015–Novem (2015) 230–233. Loureiro, N. Seoane, M. Nafria, Local Defect Density in doi:10.1109/ESSDERC.2015.7324756. Polycrystalline High-k Dielectrics: CAFM-Based Evaluation Methodology and Impact on MOSFET [13] V. Velayudhan, F. Gamiz, J. Martin-Martinez, R. Variability, IEEE Electron Device Lett. 38 (2017) 637– Rodriguez, M. Nafria, X. Aymerich, Influence of the 640. doi:10.1109/LED.2017.2680545. interface trap location on the performance and variability of ultra-scaled MOSFETs, Microelectron. Reliab. 53 [4] O. Pirrotta, L. Larcher, M. Lanza, A. Padovani, M. Porti, (2013) 1243–1246. doi:10.1016/j.microrel.2013.07.052. M. Nafría, G. Bersuker, Leakage current through the poly-crystalline HfO2: Trap densities at grains and grain [14] A. Ruiz, N. Seoane, S. Claramunt, A. Garcia-Loureiro, boundaries, J. Appl. Phys. 114 (2013) 134503. M. Porti, C. Couso, J. Martin-Martinez, M. Nafria, doi:10.1063/1.4823854. Workfunction fluctuations in polycrystalline TiN observed with KPFM and their impact on MOSFETs [5] K. Murakami, M. Rommel, B. Hudec, A. Rosová, K. variability, Appl. Phys. Lett. 114 (2019) 093502. Hušeková, E. Dobročka, R. Rammula, A. Kasikov, J.H. doi:10.1063/1.5090855. Han, W. Lee, S.J. Song, A. Paskaleva, A.J. Bauer, L. Frey, K. Fröhlich, J. Aarik, C.S. Hwang, Nanoscale [15] V. Iglesias, M. Lanza, K. Zhang, A. Bayerl, M. Porti, M. Characterization of TiO2 Films Grown by Atomic Layer Nafra, X. Aymerich, G. Benstteter, Z.Y. Shen, G. Deposition on RuO2 Electrodes, ACS Appl. Mater. Bersuker, Degradation of polycrystalline HfO2-based gate Interfaces. 6 (2014) 2486–2492. doi:10.1021/am4049139. dielectrics under nanoscale electrical stress, Appl. Phys. Lett. 99 (2011) 103510. doi:10.1063/1.3637633. [6] V. Velayudhan, J. Martin-Martinez, R. Rodriguez, M. Porti, M. Nafria, X. Aymerich, C. Medina, F. Gamiz, [16] T. Berthold, G. Benstetter, W. Frammelsberger, R. TCAD simulation of interface traps related variability in Rodríguez, M. Nafría, Nanoscale characterization of bulk decananometer mosfets, 2014 5th Eur. Work. C. copper oxide films by Kelvin Probe Force Microscopy, Var. VARI 2014. (2014). Thin Solid Films. 584 (2015) 310–315. doi:10.1109/VARI.2014.6957078. doi:10.1016/j.tsf.2015.01.071. [7] S.M. Amoroso, L. Gerrer, S. Markov, F. Adamu-Lema, [17] J.L. Garrett, D. Somers, J.N. Munday, The effect of patch A. Asenov, Comprehensive statistical comparison of potentials in Casimir force measurements determined by RTN and BTI in deeply scaled MOSFETs by means of heterodyne Kelvin probe force microscopy, J. Phys. 3D atomistic simulation, Eur. Solid-State Device Res. Condens. Matter. 27 (2015) 214012. doi:10.1088/0953- Conf. (2012) 109–112. 8984/27/21/214012. doi:10.1109/ESSDERC.2012.6343345. [18] S.A. Burke, J.M. Ledue, Y. Miyahara, J.M. Topple, S. [8] B. Kaczer, T. Grasser, P.J. Roussel, J. Franco, R. Fostner, P. Grütter, Determination of the local contact Degraeve, L.A. Ragnarsson, E. Simoen, G. Groeseneken, potential difference of PTCDA on NaCl: Acomparison of H. Reisinger, Origin of NBTI variability in deeply scaled techniques, Nanotechnology. 20 (2009) 264012. pFETs, IEEE Int. Reliab. Phys. Symp. Proc. (2010) 26– doi:10.1088/0957-4484/20/26/264012. 32. doi:10.1109/IRPS.2010.5488856. [19] T. Glatzel, S. Sadewasser, M.C. Lux-Steiner, Amplitude [9] H.F. Dadgour, K. Endo, V.K. De, K. Banerjee, Grain- or frequency modulation-detection in Kelvin probe force orientation induced work function variation in nanoscale microscopy, Appl. Surf. Sci. 210 (2003) 84–89. metal-gate transistors - Part II: Implications for process, doi:10.1016/S0169-4332(02)01484-8. device, and circuit design, IEEE Trans. Electron Devices. 57 (2010) 2515–2525. doi:10.1109/TED.2010.2063270. [20] T. Ouisse, F. Martins, M. Stark, S. Huant, J. Chevrier, Signal amplitude and sensitivity of the Kelvin probe force [10] N.M. Idris, A. Brown, J. Watling, A. Asenov, Simulation microscopy, Appl. Phys. Lett. 88 (2006) 043102. Study of Workfunction Variability in MOSFETs with doi:10.1063/1.2168251. Polycrystalline Metal Gates, Ultim. Integr. Silicon. 57 (2010) 165–168. http://eprints.gla.ac.uk/45013/. [21] G. Li, B. Mao, F. Lan, L. Liu, Practical aspects of single- pass scan Kelvin probe force microscopy, Rev. Sci. [11] K. Ohmori, T. Matsuki, D. Ishikawa, T. Morooka, T. Instrum. 83 (2012) 113701. doi:10.1063/1.4761922. Aminaka, Y. Sugita, T. Chikyow, K. Shiraishi, Y. Nara, K. Yamada, Impact of additional factors in threshold [22] A.J. Garcia-Loureiro, N. Seoane, M. Aldegunde, R. voltage variability of metal/high-k gate stacks and its Valin, A. Asenov, A. Martinez, K. Kalna, Implementation reduction by controlling crystalline structure and grain of the density gradient quantum corrections for 3-D size in the metal gates, Tech. Dig. - Int. Electron Devices simulations of multigate nanoscaled transistors, IEEE Meet. IEDM. (2008). doi:10.1109/IEDM.2008.4796707. Trans. Comput. Des. Integr. Circuits Syst. 30 (2011) 841– 851. doi:10.1109/TCAD.2011.2107990. [12] V. Velayudhan, J. Martin-Martinez, M. Porti, C. Couso,