Pankaj Kumbhare

Pankaj Kumbhare

Auto-generated profile, based on publicly available information

Pankaj Kumbhare

Department of Electrical Engineering, IIT Bombay, Mumbai, India

Research Focus & Metrics

Academic Career Length

10Years (2017-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
13 February 2018 in IEEE Electron Device Letters
Verified
Journal Article
Closed access
SLS. Lashkare
S. Lashkare
SCS. Chouhan
S. Chouhan
TCT. Chavan
T. Chavan
ABA. Bhat
A. Bhat
PKP. Kumbhare
P. Kumbhare
UGU. Ganguly
U. Ganguly
Resistance random access memories (RRAM) or memristors with an analog change of conductance are widely explored as an artificial synapse, e.g., Pr0.7Ca0.3MnO3 (PCMO) RRAM-based synapses. In addition to synapses, scaled neurons are essential to enable a neuromorphic hardware. In this letter, we propose a PCMO RRAM for integrate and fire (IF) neuron. The analog conductance increase during SET process enables integration function. Upon exceeding a conductance threshold (i.e., fire) during a READ operation, a hard RESET is performed to reduce the conductance. The SET, READ, and RESET are performed in different phases of a clock to enable a PCMO for IF neuron. The availability of a non-volatile PCMO-based synapse makes PCMO for IF neuron attractive. Finally, PCMO-based neuron in spiking neural network yields software-equivalent classification accuracy as demonstrated on standard Fischer's Iris flower data set.

4 July 2017 in IEEE Electron Device Letters
Verified
Journal Article
Closed access
SLS. Lashkare
S. Lashkare
NPN. Panwar
N. Panwar
PKP. Kumbhare
P. Kumbhare
BDB. Das
B. Das
UGU. Ganguly
U. Ganguly
Resistance random access memories (RRAMs) are widely explored to show spike time dependent plasticity (STDP) as a learning rule to show biological synaptic behavior, as these devices possess analog conductance change. To implement STDP, preand post-neuronal waveforms are superposed. Only the peak voltage changes the conductance of memory. But due to the remaining part of the waveform (which don't affect the conductance change), there is a significant amount of inadvertent current flow leading to unnecessary energy consumption. In this letter, we experimentally demonstrate that, the PCMO-based RRAM, a memristor (1M) in series with the NPN selector can be used as a synapse to reduce the undesirable energy consumption. First, we propose the preand post-neuronal waveform engineering required for PCMO-based memristor + Selector (1S1M) to reduce energy consumption. Second, we demonstrate experimentally that 1S1M synapse gives >4700× reduction in energy consumption compared with 1M synapse for a single neuronal waveform. Third, we implement and compare anti-STDP for 1M and 1S1M synapse to show no loss of generality. Thus, we experimentally demonstrate 1S1M-based energy efficient synapse for brain inspired computing.

12 July 2022 in IEEE Transactions on Electron Devices
Verified
Journal Article
Closed access
AWAmey M. Walke
Amey M. Walke
MPMihaela I. Popovici
Mihaela I. Popovici
KBKaustuv Banerjee
Kaustuv Banerjee
SCSergiu Clima
Sergiu Clima
PKPankaj Kumbhare
Pankaj Kumbhare
JHJan Van Houdt
Jan Van Houdt
High endurance of 10 11 cycles is demonstrated in ~9–10-nm stoichiometric Hafnium Zirconate (HZO) metal–ferroelectric–metal (MFM) capacitors deposited using Cl precursors with La and Y dopants. La doping is shown to offer higher remnant polarization than Y. Investigation of doped layers with asymmetric polarization versus electric field ( ${P}$ – ${E}$ ) measurements and unipolar fatigue cycles suggests that in the pristine state, the HZO is comprised of ferroelectric domains with internal built-in electric field-induced pinned coercive field ( ${E}_{c}$ ). Doping is shown to increase the pinning effect and two distinct groups of ferroelectric domains emerge, which is antialigned at zero applied electric field ( ${E}$ ). The antiferroelectric-like (pinched ${P}$ – ${E}$ loop) behavior is therefore attributed to internal built-in ${E}$ -induced pinning of the domains during growth and/or annealing steps. The initial wake-up is attributed to gradual depinning of the domains with bipolar electric pulses. Suppression of monoclinic phase was observed in doped layers that survive 10 11 cycles. The wake-up is shown to be dependent on total duration and magnitude of bipolar electric pulses. Precycling scheme is demonstrated for stable operation at a lower field.

18 July 2017 in IEEE Transactions on Electron Devices
Verified
Journal Article
Closed access
PKP. Kumbhare
P. Kumbhare
ICI. Chakraborty
I. Chakraborty
AKA. Khanna
A. Khanna
UGU. Ganguly
U. Ganguly
Enhancement of nonlinearity (NL) in low-resistance state (LRS) currents of resistance random access memory (RRAM) devices is a key challenge for the selectorless RRAM array. The conventional approach is based on adding multiple dielectrics, e.g., tunnel layers to enable NL at the expense of stack simplicity. In this brief, we present a Pr 0.7 Ca 0.3 MnO 3 (PCMO)-based selectorless RRAM device that exhibits high NL. The presented single oxide layer device (W/PCMO/Pt) enables selectorless RRAM without an extra tunnel barrier layer. The demonstrated device is forming free. It shows a low SET current density (10 4 A/cm 2 ) and a high NL in LRS current for READ operation (105 ±5) and for SET operation (44±4) along with a large memory window (160±2) in dc cycling. Further, the device shows very low device-to-device variability (σ/μ <; 0.23) along with excellent retention (ten years at 200 °C), good pulsed endurance (no degradation for >10 4 cycles), and a possible multilevel cell capability. A self-heating-based sharp current increase produces the NL to enable selectorless operation. Thus, a selectorless PCMO-based RRAM based on a novel mechanism is presentedand benchmarked against the literature.

1 September 2019 in APL Materials
Verified
Journal Article
Open access
PDF file
DKDevesh Khilwani
Devesh Khilwani
VMVineet Moghe
Vineet Moghe
SLSandip Lashkare
Sandip Lashkare
VSVivek Saraswat
Vivek Saraswat
PKPankaj Kumbhare
Pankaj Kumbhare
UGUdayan Ganguly
Udayan Ganguly
The neural network enables efficient solutions for Nondeterministic Polynomial-time (NP) hard problems, which are challenging for conventional von Neumann computing. The hardware implementation, i.e., neuromorphic computing, aspires to enhance this efficiency by custom hardware. Particularly, NP hard graphical constraint optimization problems are solved by a network of stochastic binary neurons to form a Boltzmann Machine (BM). The implementation of stochastic neurons in hardware is a major challenge. In this work, we demonstrate that the high to low resistance switching (set) process of a PrxCa1−xMnO3 (PCMO) based RRAM (Resistive Random Access Memory) is probabilistic. Additionally, the voltage-dependent probability distribution approximates a sigmoid function with 1.35%–3.5% error. Such a sigmoid function is required for a BM. Thus, the Analog Approximate Sigmoid (AAS) stochastic neuron is proposed to solve the maximum cut—an NP hard problem. It is compared with Digital Precision-controlled Sigmoid (DPS) implementation using (a) pure CMOS design and (b) hybrid (RRAM integrated with CMOS). The AAS design solves the problem with 98% accuracy, which is comparable with the DPS design but with 10× area and 4× energy advantage. Thus, ASIC neuro-processors based on novel analog neuromorphic devices based BM are promising for efficiently solving large scale NP hard optimization problems.

26 April 2018 in IEEE Transactions on Electron Devices
Verified
Journal Article
Closed access
PKP. Kumbhare
P. Kumbhare
UGU. Ganguly
U. Ganguly
Selector-less resistive random-access memory (RRAM) devices require high nonlinearity (NL) in low-resistance state current (ILRS). In this paper, we investigate the effect of composition (here, Ca%)-dependent material properties, viz., ion migration barrier, and thermal conductivity, on memory performance of recently demonstrated Pr1-xCaxMnO3-based selector-less RRAM with high NL. First, the NL increases as ``x'' decreases. This is attributed to higher self-heating in PCMO(x) films as composition-dependent thermal conductivity (κ ) decreases as ``x'' is decreased. This enables selector-less operation. Second, larger memory window is observed as ``x'' increases due to decrease in the voltage required for onset of RESET process (i.e., VMIN, RESET) as ``x'' is increased. Lower VMIN, RESET at higher ``x'' is due to higher oxygen-ion conductivity attributed to lower migration barrier (Em). Third, retention is measured at different temperatures to extract composition-dependent Em. The poor memory retention performance for x = 1 is consistent with lower Em estimated. Finally, enhanced endurance is observed for lower Em due to low-energy requirement for switching. Based on these observations, an interrelation between Em, retention, and endurance is established. Therefore, an interplay of composition-dependent thermal conductivity and oxygen-ion migration barrier enables tunable memory characteristics in Pr1-xCaxMnO3-based RRAM.