In a neuromorphic computing system, the complex CMOS neuron circuits have been the bottleneck for efficient implementation of weighted sum operation. The phenomenon of metal-insulator-transition (MIT) in strongly correlated oxides, such as NbO2, has shown the oscillation behavior in recent experiments. In this work, we propose using a MIT device to function as a compact oscillation neuron, achieving the same functionality as the CMOS neuron but occupying a much smaller area. Pt/NbOx/Pt devices are fabricated, exhibiting the threshold switching I-V hysteresis. When the NbOx device is connected with an external resistor (i.e., the synapse), the neuron membrane voltage starts a self-oscillation. We experimentally demonstrate that the oscillation frequency is proportional to the conductance of the synapse, showing its feasibility for integrating the weighted sum current. The switching speed measurement indicates that the oscillation frequency could achieve>33 MHz if parasitic capacitance can be eliminated.

Implementation of brain-inspired neural networks with conventional CPU/GPU platforms based on sequential von Neumann architecture is computationally expensive and power hungry.1 Alternative approaches, such as neuro-inspired architectures, have attracted great attention due to the tolerance to fault and error and massively parallel computation for speed-up.2,3 Resistive memory (RRAM) has been demonstrated to emulate the synapses in the neural network.4–6 Integration of RRAM into a crossbar array architecture can implement the weighted sum (i.e., vector-matrix multiplication) in a parallel fashion,7–9 accelerating the most time/energy-consuming operation in the neuro-inspired learning algorithms, as shown in Fig. 1(a). When an input vector is fed into the crossbar array, the weighted sum current (modulated by the weight or conductance of each RRAM synapse) will sink to the neuron node at the end of the column. The neuron node thus integrates this analog current and converts into spikes or digital outputs. Typically, a CMOS integrate-and-fire neuron circuit is needed to convert the analog current into spikes (essentially as an analog-to-digital converter). The counter further counts the number of spikes and converts them into digital bits. However, today's CMOS integrate-and-fire neuron typically requires tens of transistors, as shown in Fig. 1(b).10 Figure 2(a) shows the simulated waveform of the membrane voltage (Vin) and output spike (Vspike) for different weighted sum current values (6 μA vs. 1 μA). The number of output spikes is designed to be proportional to the amplitude of the input weighted sum current. Apparently, such a CMOS neuron is complex and occupies a much larger size than the column pitch of the crossbar array, thereby reducing the parallelism as the time-multiplexing is needed to sequentially read out all the weighted sum from the array.

FIG. 1.

(a) The resistive crossbar array implements the synaptic network, and the neuron node at the end of the column integrates the weighted sum current from the array. (b) One example of the CMOS integrate-and-fire neuron design.10 The membrane voltage integrates and discharges after triggering the output spike.

FIG. 1.

(a) The resistive crossbar array implements the synaptic network, and the neuron node at the end of the column integrates the weighted sum current from the array. (b) One example of the CMOS integrate-and-fire neuron design.10 The membrane voltage integrates and discharges after triggering the output spike.

Close modal
FIG. 2.

(a) The waveform of the CMOS integrate-and-fire neuron for different weighted sum current values (6 μA vs. 1 μA).10 (b) Proposed design of an oscillation neuron with a MIT device at the end of the column that emulates the Vin node.

FIG. 2.

(a) The waveform of the CMOS integrate-and-fire neuron for different weighted sum current values (6 μA vs. 1 μA).10 (b) Proposed design of an oscillation neuron with a MIT device at the end of the column that emulates the Vin node.

Close modal

The metal-Insulator-Transition (MIT) phenomenon has been observed in strongly correlated oxide devices such as VO211 and NbO2.12 Recently, the MIT induced oscillation has been investigated for various applications.13–17 Design of the MIT device as a compact oscillation neuron in the RRAM array architectures and its advantages over the CMOS neuron, such as the area, latency, energy and leakage power, has been thoroughly investigated by circuit-level simulations.18 In this work, we experimentally demonstrate the compact oscillation neuron at the device-level using NbOx in order to replace the CMOS neuron, as shown in Fig. 2(b). Due to the threshold switching I-V with hysteresis, the node voltage on the MIT device will oscillate and emulate the Vin node in the CMOS neuron, and the oscillation frequency is expected to be proportional to the weighted sum current.18 The switching speed measurement indicates that the oscillation frequency could achieve >33 MHz if parasitic capacitance can be eliminated.

Pt/NbOx/Pt devices were fabricated in ASU NanoFab. The active device area (overlap region between top and bottom electrodes) is about 10 × 10 μm2. First, an e-beam evaporated Pt/Ti (25 nm/3 nm) bottom electrode was patterned by optical lithography techniques on a SiO2/Si substrate (200 nm/500 μm, respectively). A blanket NbOx thin film (15 nm) was deposited by using reactive sputtering at 100 °C using an Nb target and an O2/Ar gas mixture in the ratio of 1/10. The Pt (25 nm) electrode was formed on top of the NbOx blanket layer by using e-beam evaporation with a subsequent optical lithography and lift-off process to define the perpendicular cross-point structure. The bottom electrode pads were exposed by optical lithography and wet etching of the NbOx layer. A schematic of the Pt/NbOx/Pt device is shown in the inset of Fig. 3.

FIG. 3.

Measured I-V threshold switching characteristics of the Pt/NbOx/Pt device. The inset shows the schematic of the fabricated Pt/NbOx/Pt device.

FIG. 3.

Measured I-V threshold switching characteristics of the Pt/NbOx/Pt device. The inset shows the schematic of the fabricated Pt/NbOx/Pt device.

Close modal

A forming voltage of ∼3 V is needed to trigger the subsequent threshold switching. The ON/OFF ratio (ROFF/RON) is about 50–100. Figure 3 shows the typical threshold switching I-V characteristics of the Pt/NbOx/Pt devices. As the voltage was swept from 0 to 3 V with a current compliance of 1 mA, an abrupt increase in current was observed at about 1.9 V, called threshold voltage (Vth). While sweeping the voltage from 3 V to 0, the current switched to the OFF state at about 1.6 V, called hold voltage (Vhold). Similar behavior is also observed for the negative voltage sweeps (not shown), consistent with the nonpolar nature of insulation-metal-transition. The conduction mechanism in the NbOx device is believed to be more filamentary in the ON state and more bulk-dominated in the OFF state.19,20 Therefore, it is observed that Vth and RON are insensitive to the device area, while ROFF increases with the downscaling of the device size.13 However, these characteristics are subject to temperature changes. It was reported that RON, ROFF, Vhold, and Vth decrease with higher temperature,21,22 where smaller RON and ROFF can be attributed to a typical semiconducting behavior of the oxide. These resistance and voltage shifts will induce unwanted frequency deviations in the oscillation neuron, and thus, a compensation method for oscillation frequency may be required at the circuit level to remedy this issue. Nevertheless, the MIT phenomenon in NbO2 is much more thermally stable than that in VO2, where the MIT phenomenon disappears above 67 °C.11 

X-ray Photoelectron Spectroscopy (XPS) is employed to characterize the stoichiometry and the binding energy of the NbOx thin film. To avoid the effects of surface contamination, all the samples were surface etched for 4 nm using Ar+ sputtering in the XPS chamber before collecting the XPS information. Figure 4(a) shows the XPS spectrum of Nb3d in the NbOx thin film. The fitting was performed using CasaXPS data processing software. The two peaks of the green fitting curve located at 205.5 eV and 208.3 eV correspond to NbO2 3d5/2 and 3d3/2, respectively, while the two peaks of the blue fitting curve located at 207.3 eV and 210.1 eV correspond to Nb2O5 3d5/2 and 3d3/2, respectively, which indicates that both NbO2 and Nb2O5 phases co-exist in the NbOx thin film. It is known that NbO2 exhibits the MIT behavior, while Nb2O5 may show the resistive switching.23 The O1s peak shows that the Nb-O bonds are located at ∼531 eV and other non-bridging oxygen bonds are located at ∼532.4 eV, as shown in Fig. 4(b).

FIG. 4.

The XPS spectra of (a) Nb3d and (b) O1s in the deposited NbOx thin film, showing the co-existence of NbO2 and Nb2O5 phases. NbO2 is known to exhibit the MIT behavior.

FIG. 4.

The XPS spectra of (a) Nb3d and (b) O1s in the deposited NbOx thin film, showing the co-existence of NbO2 and Nb2O5 phases. NbO2 is known to exhibit the MIT behavior.

Close modal

To make the neuron node oscillate, we connect the NbOx device with a load resistor (RL) as synapse to demonstrate the oscillation neuron function, as shown in Fig. 5(a). The resistance of the load resistor is chosen in between NbOx device's ON state (RON) and OFF state (ROFF), and there is a parasitic capacitance (C) at the neuron node. When the voltage VDD is applied, the membrane voltage on the capacitor will be charged because most of the voltage drop is on the NbOx device (ROFF > RL). Once the voltage exceeds Vth, the NbOx device switches to RON, and the capacitor starts discharging since the voltage drop on the NbOx device becomes small (RON < RL). Once the membrane voltage decreases below Vhold, the NbOx device switches to ROFF. This charging and discharging process repeats, and thus, the voltage of the neuron node oscillates between Vhold and Vth. From Ref. 18, the analytical solution of the charging time trise from Vhold to Vth can be expressed as

trise=RrC×log(VthVDDRrRLVholdVDDRrRL),
(1)

where Rr = RLǁROFF. Similarly, the discharging time tfall from Vth to Vhold can be calculated as

tfall=RfC×log(VholdVDDRfRLVthVDDRfRL),
(2)

where Rf = RLǁRON. If we assume ROFF ≫ RL ≫ RON, then Rr ≈ RL and Rf ≈ RON, which makes trise proportional to RL and tfall to be a constant much smaller than trise. The ideal oscillation frequency f can be obtained by using Eq. (1)

f=WLC×log(VholdVDDVthVDD),
(3)

where WL = 1/RL is the weight of the RRAM synapse. f is then proportional to WL. Therefore, the oscillation frequency represents a weighted sum if the NbOx device connected to all the RRAM synaptic weights in one column of the array. Figures 5(b)–5(d) show the measured oscillation frequency with different RL values, i.e., different synaptic weights. A voltage pulse is applied to Channel 1 (CH1), and the node voltage is monitored on Channel 2 (CH2) using an oscilloscope. As the charging is through the load resistor and the discharging is through the NbOx device at RON, the RC delay of the charging is larger than that of the discharging, which makes the voltage oscillation a triangular waveform. The oscillation frequencies are 2 MHz, 0.7 MHz, and 0.4 MHz with the different load resistance 3.6 kΩ, 11.5 kΩ, and 16.1 kΩ, respectively. This suggests that the oscillation frequency is proportional to the equivalent resistance of the column (i.e., weighted sum results) if the NbOx neuron can be connected to the crossbar array. In our experiments, the oscillation frequency is limited by the parasitic capacitance (C1=573 pF estimated) in the testing setup as the resistor is externally connected to the pad of the NbOx device.

FIG. 5.

(a) Circuit configuration of an oscillation neuron node with the Pt/NbOx/Pt MIT device and a load resistor (RL) as synapse. Oscillation characteristics with various RL values: (b) RL=3.6 kΩ and frequency= 2 MHz. (c) RL=11.5 kΩ and frequency= 0.7 MHz. (b) RL=16.1 kΩ and frequency= 0.4 MHz. The oscillation frequency is proportional to the synaptic conductance. C1 is estimated to be 573 pF.

FIG. 5.

(a) Circuit configuration of an oscillation neuron node with the Pt/NbOx/Pt MIT device and a load resistor (RL) as synapse. Oscillation characteristics with various RL values: (b) RL=3.6 kΩ and frequency= 2 MHz. (c) RL=11.5 kΩ and frequency= 0.7 MHz. (b) RL=16.1 kΩ and frequency= 0.4 MHz. The oscillation frequency is proportional to the synaptic conductance. C1 is estimated to be 573 pF.

Close modal

In order to explore the intrinsic switching speed of the NbOx device, a voltage pulse with a high and low waveform is employed. The high pulse is to test the off-to-on speed, and the low pulse is to test the on-to-off speed. The measurement setup is schematically shown Fig. 6(a). There is a sudden increase in current, when time reaches the off-to-on switching time. To reduce the parasitic capacitance effect, CH2 of the oscilloscope is directly connected in the device and the input impedance of CH2 is set to 50 Ω. Figure 6(b) shows that the device will not turn on at 1.8 V. The device switches to the ON state at 2.0 V, and the off-to-on switching speed is around 33 ns, as shown in Fig. 6(c). Figures 6(d) and 6(e) show that the switching speed does not obviously change when increasing the pulse amplitude from 2.0 V to 2.4 V, indicating that the intrinsic switching speed is ∼33 ns, and thus, the oscillation frequency could achieve >33 MHz if parasitic capacitance can be eliminated.

FIG. 6.

(a) Circuit configuration of the Pt/NbOx/Pt MIT device speed measurement. The high-low voltage waveform is used: the high pulse is to test the off-to-on speed and the low pulse is to test the on-to-off speed. Different high pulses are used: (b) 1.8 V, (c) 2.0 V, (d) 2.2 V, and (e) 2.4 V. The results show that the off-to-on speed is around 33 ns and on-to-off is almost instant. The switching speed is independent of the pulse amplitude, indicating that the intrinsic switching speed is ∼33 ns, and thus, the oscillation frequency could achieve >33 MHz.

FIG. 6.

(a) Circuit configuration of the Pt/NbOx/Pt MIT device speed measurement. The high-low voltage waveform is used: the high pulse is to test the off-to-on speed and the low pulse is to test the on-to-off speed. Different high pulses are used: (b) 1.8 V, (c) 2.0 V, (d) 2.2 V, and (e) 2.4 V. The results show that the off-to-on speed is around 33 ns and on-to-off is almost instant. The switching speed is independent of the pulse amplitude, indicating that the intrinsic switching speed is ∼33 ns, and thus, the oscillation frequency could achieve >33 MHz.

Close modal

Using the above parameters, we performed SPICE simulations of the oscillation neuron based on our developed Verilog-A behavior model of the NbOx device to find out the feasible range of synaptic weights (1/RL) for oscillation. As shown in Fig. 7, the SPICE simulation result shows good consistency with the measured oscillation frequencies in Figs. 5(b)–5(d), and its deviation from the ideal frequency estimation [Eq. (3)] is also observed, which is due to noticeable tfall. Overall, the results suggest that there is a limited weight range for oscillation and larger frequency deviation occurs when the weight approaches either ON or OFF states of NbOx. To support a wider weight range, we need a larger ON/OFF ratio of NbOx, possibly by downscaling of the device size to reduce the off-state leakage or a more advanced device engineering by stacking a tunneling layer.24 It is estimated that at least an ON/OFF ratio of 1000 is required to support a 128 × 128 array.18 Compared to the CMOS neuron, the compact oscillation neuron can potentially achieve a >12.5X area and a >5X energy reduction as suggested in Ref. 18 if the NbOx device characteristics can be improved with a larger ON/OFF ratio of 1000 and smaller operating voltages (Vhold and Vth <1 V).

FIG. 7.

Oscillation frequency as a function of synaptic weight (1/RL). The SPICE result shows good consistency with the measured frequencies in Figs. 5(b)–5(d). Oscillation failure occurs when the weight is too large or too small. The frequency deviation from the ideal estimation [Eq. (3)] is due to noticeable tfall.

FIG. 7.

Oscillation frequency as a function of synaptic weight (1/RL). The SPICE result shows good consistency with the measured frequencies in Figs. 5(b)–5(d). Oscillation failure occurs when the weight is too large or too small. The frequency deviation from the ideal estimation [Eq. (3)] is due to noticeable tfall.

Close modal

In summary, an oscillation neuron using the NbOx device is experimentally demonstrated with a synapse (load resistor). By varying the load resistance, the oscillation frequency is proportional to the synaptic conductance, showing the feasibility for integrating the weighted sum. The intrinsic switching speed of the NbOx device is ∼33 ns, indicating that the oscillation frequency could achieve >33 MHz if the parasitic capacitance can be eliminated. This work reports a potential building block for the neuromorphic system.

This work was supported by NSF-CCF-1552687 and NSF-ECCS-1701565.

1.
S. B.
Eryilmaz
,
D.
Kuzum
,
S.
Yu
, and
H.-S. P.
Wong
, “
Device and system level design consideration for analog-non-volatile-memory based neuromorphic architectures
,” in
IEEE International Electron Devices Meeting
(
2015
).
2.
G. W.
Burr
,
P.
Narayanan
,
R. M.
Shelby
,
S.
Sidler
,
I.
Boybat
,
C. D.
Nolfo
, and
Y.
Leblebici
, “
Large-scale neural networks implemented with non-volatile memory as the synaptic weight element: comparative performance analysis (accuracy, speed, and power)
,” in
IEEE International Electron Devices Meeting
(
2015
).
3.
D.
Kuzum
,
S.
Yu
, and
H.-S. P.
Wong
, “
Synaptic electronics: Materials, devices and applications
,”
Nanotechnology
24
,
382001
(
2013
).
4.
S. H.
Jo
,
T.
Chang
,
I.
Ebong
,
B. B.
Bhadviya
,
P.
Mazumder
, and
W.
Lu
, “
Nanoscale memristor device as synapse in neuromorphic systems
,”
Nano Lett.
10
(
4
),
1297
1301
(
2010
).
5.
D.
Garbin
,
E.
Vianello
,
O.
Bichler
,
Q.
Rafhay
,
C.
Gamrat
,
G.
Ghibaudo
,
B.
DeSalvo
, and
L.
Perniola
, “
HfO2-based OxRAM devices as synapses for convolutional neural networks
,”
IEEE Trans. Electron Devices
62
(
8
),
2494
2501
(
2015
).
6.
S.
Ambrogio
,
S.
Balatti
,
V.
Milo
,
R.
Carboni
,
Z. Q.
Wang
,
A.
Calderoni
,
N.
Ramaswamy
, and
D.
Ielmini
, “
Neuromorphic learning and recognition with one-transistor-one-resistor synapses and bistable metal oxide RRAM
,”
IEEE Trans. Electron Devices
63
(
4
),
1508
1515
(
2016
).
7.
L.
Gao
,
I.-T.
Wang
,
P.-Y.
Chen
,
S.
Vrudhula
,
J.-S.
Seo
,
Y.
Cao
,
T.-H.
Hou
, and
S.
Yu
, “
Fully parallel write/read in resistive synaptic array for accelerating on-chip learning
,”
Nanotechnology
26
(
45
),
455204
(
2015
).
8.
M.
Prezioso
,
F.
Merrikh-Bayat
,
B. D.
Hoskins
,
G. C.
Adam
,
K. K.
Likharev
, and
D. B.
Strukov
, “
Training and operation of an integrated neuromorphic network based on metal-oxide memristors
,”
Nature
521
,
61
64
(
2015
).
9.
P.-Y.
Chen
,
D.
Kadetotad
,
Z.
Xu
,
A.
Mohanty
,
B.
Lin
,
J.
Ye
,
S.
Vrudhula
,
J.-S.
Seo
,
Y.
Cao
, and
S.
Yu
, “
Technology-design co-optimization of resistive cross-point array for accelerating learning algorithms on chip
,” in
IEEE Design, Automation and Test in Europe (DATE)
(
2015
).
10.
D.
Kadetotad
,
Z.
Xu
,
A.
Mohanty
,
P.-Y.
Chen
,
B.
Lin
,
J.
Ye
,
S.
Vrudhula
,
S.
Yu
,
Y.
Cao
, and
J.-S.
Seo
, “
Parallel architecture with resistive crosspoint array for dictionary learning acceleration
,”
IEEE J. Emerging Sel. Top. Circuits Syst.
5
(
2
),
194
204
(
2015
).
11.
Y.
Zhou
,
X.
Chen
,
C.
Ko
,
Z.
Yang
,
C.
Mouli
, and
S.
Ramanathan
, “
Voltage-triggered ultrafast phase transition in vanadium dioxide switches
,”
IEEE Electron Device Lett.
34
(
2
),
220
222
(
2013
).
12.
D. M.
Pickett
and
R. S.
Williams
, “
Sub-100 fJ and sub-nanosecond thermally driven threshold switching in niobium oxide crosspoint nanodevices
,”
Nanotechnology
23
,
215202
(
2012
).
13.
K.
Moon
,
E.
Cha
,
J.
Park
,
S.
Gi
,
M.
Chu
,
K.
Baek
,
B.
Lee
,
S.
Oh
, and
H.
Hwang
, “
High density neuromorphic system with Mo/Pr0.7Ca0.3MnO3 synapse and NbO2 IMT oscillator neuron
,” in
Proceedings of International Electron Devices Meeting (IEDM)
(
2015
).
14.
S.
Li
,
X.
Liu
,
S. K.
Nandi
,
D. K.
Venkatachalam
, and
R. G.
Elliman
, “
High-endurance megahertz electrical self-oscillation in Ti/NbOx bilayer structures
,”
Appl. Phys. Lett.
106
,
212902
(
2015
).
15.
M. D.
Pickett
,
G.
Medeiros-Ribeiro
, and
R. S.
Williams
, “
A scalable neuristor built with Mott memristors
,”
Nat. Mater.
12
,
114
117
(
2013
).
16.
J.
Sakai
, “
High-efficiency voltage oscillation in VO2 planer-type junctions with infinite negative differential resistance
,”
J. Appl. Phys.
103
,
103708
(
2008
).
17.
N.
Shukla
,
W. Y.
Tsai
,
M.
Jerry
,
M.
Barth
,
V.
Narayanan
, and
S.
Datta
, “
Ultra low power coupled oscillator arrays for computer vision applications
,” in
IEEE Symposium on VLSI Technology
(
2016
).
18.
P.-Y.
Chen
,
J.-S.
Seo
,
Y.
Cao
, and
S.
Yu
, “
Compact oscillation neuron exploiting metal-insulator-transition for neuromorphic computing
,” in
IEEE/ACM International Conference on Computer-Aided Design (ICCAD)
(
2016
).
19.
X.
Liu
,
S. M.
Sadaf
,
M.
Son
,
J.
Shin
,
J.
Park
,
J.
Lee
,
S.
Park
, and
H.
Hwang
, “
Diode-less bilayer oxide (WOx–NbOx) device for cross-point resistive memory applications
,”
Nanotechnology
22
(
47
),
475702
(
2011
).
20.
S. K.
Nandi
,
X.
Liu
,
D. K.
Venkatachalam
, and
R. G.
Elliman
, “
Threshold current reduction for the metal-insulator transition in NbO2-x-selector devices: The effect of ReRAM integration
,”
J. Phys. D: Appl. Phys.
48
(
19
),
195105
(
2015
).
21.
S.
Kim
,
J.
Park
,
J.
Woo
,
C.
Cho
,
W.
Lee
,
J.
Shin
,
G.
Choi
,
S.
Park
,
D.
Lee
,
B. H.
Lee
, and
H.
Hwang
, “
Threshold-switching characteristics of a nanothin-NbO2-layer-based Pt/NbO2/Pt stack for use in cross-point-type resistive memories
,”
Microelectron. Eng.
107
,
33
36
(
2013
).
22.
C.
Funck
,
S.
Menzel
,
N.
Aslam
,
H.
Zhang
,
A.
Hardtdegen
,
R.
Waser
, and
S.
Hoffmann-Eifert
, “
Multidimensional simulation of threshold switching in NbO2 based on an electric field triggered thermal runaway model
,”
Adv. Electron. Mater.
2
(
7
),
1600169
(
2016
).
23.
S.
Kim
,
X.
Liu
,
J.
Park
,
S.
Jung
,
W.
Lee
,
J.
Woo
,
J.
Shin
,
G.
Choi
,
C.
Cho
,
S.
Park
,
D.
Lee
,
E.-J.
Cha
,
B.-H.
Lee
,
H. D.
Lee
,
S. G.
Kim
,
S.
Chung
, and
H.
Hwang
, “
Ultrathin (<10 nm) Nb2O5/NbO2 hybrid memory with both memory and selector characteristics for high density 3D vertically stackable rram applications
,” in
Symposium on VLSI Technology Digest of Technical Papers
(
2012
).
24.
X.
Liu
,
S. K.
Nandi
,
D. K.
Venkatachalam
,
K.
Belay
,
S.
Song
, and
R. G.
Elliman
, “
Reduced threshold current in NbO2 selector by engineering device structure
,”
IEEE Electron Device Lett.
35
(
10
),
1055
1057
(
2014
).