The growing importance of applications based on machine learning is driving the need to develop dedicated, energy-efficient electronic hardware. Compared with von Neumann architectures, which have separate processing and storage units, brain-inspired in-memory computing uses the same basic device structure for logic operations and data storage1,2,3, thus promising to reduce the energy cost of data-centred computing substantially4. Although there is ample research focused on exploring new device architectures, the engineering of material platforms suitable for such device designs remains a challenge. Two-dimensional materials5,6 such as semiconducting molybdenum disulphide, MoS2, could be promising candidates for such platforms thanks to their exceptional electrical and mechanical properties7,8,9. Here we report our exploration of large-area MoS2 as an active channel material for developing logic-in-memory devices and circuits based on floating-gate field-effect transistors (FGFETs). The conductance of our FGFETs can be precisely and continuously tuned, allowing us to use them as building blocks for reconfigurable logic circuits in which logic operations can be directly performed using the memory elements. After demonstrating a programmable NOR gate, we show that this design can be simply extended to implement more complex programmable logic and a functionally complete set of operations. Our findings highlight the potential of atomically thin semiconductors for the development of next-generation low-power electronics.
Subscribe to Journal
Get full journal access for 1 year
only $3.90 per issue
All prices are NET prices.
VAT will be added later in the checkout.
Tax calculation will be finalised during checkout.
Rent or Buy article
Get time limited or full article access on ReadCube.
All prices are NET prices.
The data that support the findings of this study are available at http://doi.org/10.5281/zenodo.4073060.
Kautz, W. H. Cellular logic-in-memory arrays. IEEE Trans. Comput. C-18, 719–727 (1969).
Stone, H. S. A logic-in-memory computer. IEEE Trans. Comput. C-19, 73–78 (1970).
Le Gallo, M. et al. Mixed-precision in-memory computing. Nat. Electron. 1, 246–253 (2018).
Horowitz, M. Computing’s energy problem (and what we can do about it). In 2014 IEEE Intl Solid-State Circuits Conf. Digest of Technical Papers (ISSCC) 10–14 (IEEE, 2014).
Wang, Q. H., Kalantar-Zadeh, K., Kis, A., Coleman, J. N. & Strano, M. S. Electronics and optoelectronics of two-dimensional transition metal dichalcogenides. Nat. Nanotechnol. 7, 699–712 (2012).
Manzeli, S., Ovchinnikov, D., Pasquier, D., Yazyev, O. V. & Kis, A. 2D transition metal dichalcogenides. Nat. Rev. Mater. 2, 17033 (2017).
Radisavljevic, B., Radenovic, A., Brivio, J., Giacometti, V. & Kis, A. Single-layer MoS2 transistors. Nat. Nanotechnol. 6, 147–150 (2011).
Bertolazzi, S., Brivio, J. & Kis, A. Stretching and breaking of ultrathin MoS2. ACS Nano 5, 9703–9709 (2011).
Iannaccone, G., Bonaccorso, F., Colombo, L. & Fiori, G. Quantum engineering of transistors based on 2D materials heterostructures. Nat. Nanotechnol. 13, 183–191 (2018); correction 15, 520 (2018).
Wu, B., Wan, A., Iandola, F., Jin, P. H. & Keutzer, K. SqueezeDet: unified, small, low power fully convolutional neural networks for real-time object detection for autonomous driving. In 2017 IEEE Conf. Computer Vision and Pattern Recognition Workshops (CVPRW) 446–454 (IEEE, 2017).
Graves, A., Mohamed, A. & Hinton, G. Speech recognition with deep recurrent neural networks. In 2013 IEEE Intl Conf. Acoustics, Speech and Signal Processing 6645–6649 (IEEE, 2013).
Kononenko, I. Machine learning for medical diagnosis: history, state of the art and perspective. Artif. Intell. Med. 23, 89–109 (2001).
Naffziger, S. High-performance processors in a power-limited world. In 2006 Symp. VLSI Circuits, 2006. Digest of Technical Papers 93–97 (IEEE, 2006).
McKee, S. A. Reflections on the memory wall. In Proc. First Conf. Computing Frontiers CF’04 162 (ACM Press, 2004).
Xu, X. et al. Scaling for edge inference of deep neural networks. Nat. Electron. 1, 216–222 (2018).
Arute, F. et al. Quantum supremacy using a programmable superconducting processor. Nature 574, 505–510 (2019).
Kwon, J. et al. Three-dimensional monolithic integration in flexible printed organic transistors. Nat. Commun. 10, 54 (2019).
Shulaker, M. M. et al. Three-dimensional integration of nanotechnologies for computing and data storage on a single chip. Nature 547, 74–78 (2017).
Yu, S. Neuro-inspired computing with emerging nonvolatile memories. Proc. IEEE 106, 260–285 (2018).
Chen, W.-H. et al. CMOS-integrated memristive non-volatile computing-in-memory for AI edge processors. Nat. Electron. 2, 420–428 (2019).
Li, C. et al. Analogue signal and image processing with large memristor crossbars. Nat. Electron. 1, 52–59 (2018).
Fuller, E. J. et al. Parallel programming of an ionic floating-gate memory array for scalable neuromorphic computing. Science 364, 570–574 (2019).
Fiori, G. et al. Electronics based on two-dimensional materials. Nat. Nanotechnol. 9, 768–779 (2014); erratum 9, 1063 (2014).
Splendiani, A. et al. Emerging photoluminescence in monolayer MoS2. Nano Lett. 10, 1271–1275 (2010).
Desai, S. B. et al. MoS2 transistors with 1-nanometer gate lengths. Science 354, 99–102 (2016).
Radisavljevic, B., Whitwick, M. B. & Kis, A. Integrated circuits and logic operations based on single-layer MoS2. ACS Nano 5, 9934–9938 (2011).
Wachter, S., Polyushkin, D. K., Bethge, O. & Mueller, T. A microprocessor based on a two-dimensional semiconductor. Nat. Commun. 8, 14948 (2017).
Bertolazzi, S., Krasnozhon, D. & Kis, A. Nonvolatile memory cells based on MoS2/graphene heterostructures. ACS Nano 7, 3246–3252 (2013).
Li, D. et al. Nonvolatile floating-gate memories based on stacked black phosphorus-boron nitride-MoS2 heterostructures. Adv. Funct. Mater. 25, 7360–7365 (2015).
Tan, C., Liu, Z., Huang, W. & Zhang, H. Non-volatile resistive memory devices based on solution-processed ultrathin two-dimensional nanomaterials. Chem. Soc. Rev. 44, 2615–2628 (2015).
Wang, J. et al. Floating gate memory-based monolayer MoS2 transistor with metal nanocrystals embedded in the gate dielectrics. Small 11, 208–213 (2015).
Cao, W., Kang, J., Bertolazzi, S., Kis, A. & Banerjee, K. Can 2D-nanocrystals extend the lifetime of floating-gate transistor based nonvolatile memory? IEEE Trans. Electron Dev. 61, 3456–3464 (2014).
Sebastian, A., Pannone, A., Subbulakshmi Radhakrishnan, S. & Das, S. Gaussian synapses for probabilistic neural networks. Nat. Commun. 10, 4199 (2019).
Sivan, M. et al. All WSe2 1T1R resistive RAM cell for future monolithic 3D embedded memory integration. Nat. Commun. 10, 5201 (2019).
Sun, L. et al. Self-selective van der Waals heterostructures for large scale memory array. Nat. Commun. 10, 3161 (2019).
Kim, H., Ovchinnikov, D., Deiana, D., Unuchek, D. & Kis, A. Suppressing nucleation in metal–organic chemical vapor deposition of MoS2 monolayers by alkali metal halides. Nano Lett. 17, 5056–5063 (2017).
Cun, H. et al. Wafer-scale MOCVD growth of monolayer MoS2 on sapphire and SiO2. Nano Res. 12, 2646–2652 (2019).
Raja, A. et al. Dielectric disorder in two-dimensional materials. Nat. Nanotechnol. 14, 832–837 (2019).
Papandroulidakis, G., Vourkas, I., Vasileiadis, N. & Sirakoulis, G. Ch. Boolean logic operations and computing circuits based on memristors. IEEE Trans. Circuits Syst. II 61, 972–976 (2014).
Resta, G. V. et al. Polarity control in WSe2 double-gate transistors. Sci. Rep. 6, 29448 (2016).
Resta, G. V. et al. Doping-free complementary logic gates enabled by two-dimensional polarity-controllable transistors. ACS Nano 12, 7039–7047 (2018).
Dumcenco, D. et al. Large-area epitaxial monolayer MoS2. ACS Nano 9, 4611–4620 (2015).
We thank Z. Benes (CMI) for help with electron-beam lithography, L. Navrátilová (CIME) for the preparation of the device cross-section and R. Zamani (CIME) for help with TEM imaging. We acknowledge support from the European Union’s Horizon 2020 research and innovation programme under grant agreements 829035 (QUEFORMAL), 785219 and 881603 (Graphene Flagship Core 2 and Core 3), from the Marie Curie-Sklodowska COFUND (665667), from the H2020 European Research Council (ERC, grant 682332) as well as from the CCMX Materials Challenge grant ‘Large area growth of 2D materials for device integration’. Device preparation was carried out in part in the EPFL Centre of MicroNanotechnology (CMI). TEM imaging was carried out in the EPFL Interdisciplinary Centre for Electron Microscopy (CIME).
The authors declare no competing interests.
Peer review information Nature thanks Devin Verreck, Takhee Lee and the other, anonymous, reviewer(s) for their contribution to the peer review of this work.
Publisher’s note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
Extended data figures and tables
The drain–source conductance GDS is shown versus time. Blue curve, VPROG = –12.5 V; red curve, VPROG = +12.5 V. To predict the trend of the decay, we fit both curves using the following expression, f(x) = Axk (dashed black lines). We expect that the device has a 10-year retention.
a, Device variability. IDS versus VG curves for six different devices on the same die. b, Fresh IDS versus VG curves, corresponding to the first VG sweep carried out on these devices. Maximal gate voltage ±VG,MAX (corresponding to VPROG) is insufficient for inducing charge transfer into the floating gate memory. This shows the behaviour of the FGFET in the initial state. c, IDS versus VG for different values of VDS (red curve, 50 mV; blue curve, 100 mV; green curve, 250 mV; orange curve, 500 mV). The progressive increase of the current without a decrease in the memory window demonstrates that the memory effect is not due to capacitive charges in the contacts. d, IDS versus VG for different sweep rates. The decrease of the memory window is a function of the sweep rate. The decrease is most probably a result of charge dynamics limiting the charging and discharging of the floating gate.
a, Energy band diagrams of different materials comprising the FGFET before being brought into contact. EC and EV are the positions of the bottom of the conduction band and the top of the valence band, respectively. b, Programming of the floating-gate memory by electron injection into the floating gate with the application of a positive gate voltage (upper panel). Lower panel, accompanying positive shift in the threshold voltage. c, Erase operation with electron extraction from the floating gate under the application of a negative gate voltage (upper panel). Lower panel, accompanying negative shift in the threshold voltage.
IDS is shown as a function of the number of program/erase (P/E) cycles. a, Each P/E cycle consists of a 100-ms +7.5 V pulse for the erase operation, and a 100-ms −7.5 V pulse for the program operation. b, As a but with a +10.0 V pulse for erasing and a −10.0 V pulse for programming. Both measurements are taken using a constant VDS = 50 mV and on the same device.
Extended Data Fig. 5 Example of the graphical estimation of the noise margin for the inverter programmed with VPROG = 8.5 V.
Output voltage VOUT as a function of input voltage VIN (blue dots), and its mirror reflection (red dots). VOH and VOL are defined as the points where the slope of the transfer curve (VOUT as a function of VIN, blue dots) is equal to −1, whereas VIL and VIH are the corresponding values of VIN.
a, Circuit schematic for a two-input NOR. b, Logic over time for different programming states Q1, Q2. Red: time traces of input voltages VIN1 and VIN2. Orange: output curves for Q1,2 = 33, constant LOW and Q1,2 = 11, constant HIGH. Blue: output curves for Q1,2 = 21, inverter A (IN1); Q1,2 = 12, inverter B (IN2). Green: output curve for Q1,2 = 22, NOR A,B. (Here and in Extended Data Figs. 7, 8, we denote (for example) programming state ‘Q1 = 3, Q2 = 3’ by ‘Q1,2 = 33’.).
a, Circuit schematic for a three-input NOR. b, Logic over time for different programming states Q1, Q2, Q3. Q1–3 = 111, constant HIGH; Q1–3 = 211, inverter A (IN1); Q1–3 = 112, inverter C (IN2); Q1–3 = 122, NOR B,C; Q1–3 = 212, NOR A,C; Q1–3 = 221, NOR A,B; Q1–3 = 222, NOR A,B,C.
a, Two-input schematic of the logic-in-memory concept. b, Interface model for input polarity control. c, NAND gate, Q1–4 = 2211; d, NOR gate, Q1–4 = 2332; e, XOR gate, Q1–4 = 2222. We derive the XOR canonical form by applying De Morgan’s laws.
a, b, Hardware implementation of the four-memory programmer (a) and of the nine-memory programmer (b). c, Software working diagram of the programming (top) and test (bottom) blocks. d, Example of programming (left) and test (right) blocks working, using a nine-memory programmed into the following state Q1–9 = 222111111 to perform a three-input NAND operation.
Raman spectrum of transferred MoS2 from a single crystal (which also provided the material used in this paper) using 532-nm laser excitation and a 3,000 lines mm−1 grating. The observed wavenumber difference between the A1g and E2g Raman modes of MoS2 is consistent with a monolayer. Black line is a fit to the data points (red circles).
a, Atomically resolved STEM image showing a large region of monolayer MoS2. Inset, fast Fourier transform (FFT) amplitude spectrum further shows the crystalline monolayer MoS2 structure. b, A magnified filtered STEM image taken from a shows the 2H crystal structure of monolayer MoS2. c, STEM simulation image of monolayer MoS2. The intensity line profiles at bottom right of b and c are taken along the dashed lines in those images, and show the peak positions of Mo atoms and S atoms.
a, Wide-field view of the device fabricated using the logic-in-memory process. b, Magnified view of the contact area boxed in a. c, Cross-section image of the gate stack consisting of (from bottom to top) Pt bottom gate, HfO2 blocking oxide, Pt floating gate, HfO2 tunnel oxide. The MoS2 2D channel is on top of the gate stack.
Six supplementary notes and six supplementary tables, describing the operation of the floating gate memories, memory endurance test, memory configurations for realising different logic circuits, comparison with other logic-in-memory technologies, basic material characterisation and the approach used for programming the memory circuits.
About this article
Cite this article
Migliato Marega, G., Zhao, Y., Avsar, A. et al. Logic-in-memory based on an atomically thin semiconductor. Nature 587, 72–77 (2020). https://doi.org/10.1038/s41586-020-2861-0
Nature Electronics (2021)
Nano-Micro Letters (2021)
Investigation of BiFeO3/HfO2 gate stack for ferroelectric field effect transistors in IOT applications
Journal of Materials Science: Materials in Electronics (2021)