Memory cell and method of retrieving information therefrom
The memory cell with a ferroelectric capacitor and readout circuit allows non-destructive readout by measuring the derivative of capacitance, addressing the challenge of data loss in FeRAM read operations.
Patent Information
- Application Number
- PCT/IL2026/050382
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2025-05-06
- Filing Date
- 2026-05-06
- Publication Date
- 2026-11-12
AI Technical Summary
Existing ferroelectric random-access memory (FeRAM) technologies face challenges in non-destructive readout, as conventional read processes erase the polarization state, necessitating subsequent re-write operations.
A memory cell with a ferroelectric capacitor and a memory readout circuit that applies a bias voltage pulse below the coercive threshold, measuring the derivative of capacitance to determine the memory state without altering the polarization state.
Enables non-destructive readout of ferroelectric memory states, preserving data integrity and reducing the need for subsequent re-writing.
Smart Images

Figure IL2026050382_12112026_PF_FP_ABST
Abstract
Description
[0001] MEMORY CELL AND METHOD OF RETRIEVING INFORMATION THEREFROM
[0002] RELATED APPLICATION
[0003] This application claims the benefit of priority of U.S. Provisional Patent Application No. 63 / 800,549 filed on May 6, 2025, the contents of which are incorporated herein by reference in their entirety.
[0004] FIELD AND BACKGROUND OF THE INVENTION
[0005] The present invention, in some embodiments thereof, relates to solid state electronics, and more particularly, but not exclusively, to a memory cell and method of retrieving information therefrom.
[0006] Ferroelectric Random-Access Memory (FeRAM or FRAM) is a type of non-volatile memory where data is stored based on the electrical polarization of a ferroelectric material. FeRAM was commercialized in the mid-1990s based on ferroelectric materials such as PZT and BTO. In 2011 it was reported that a thin (sub 10 nm) HfO2-based film can exhibit ferroelectricity, which makes the technology both scalable and CMOS-compatible. This finding brought ferroelectric-based memories to the front of research in non-volatile memory technology.
[0007] Known are FeRAMs having ferroelectric capacitors (FeCAP). During a read operation, a switching pulse is applied, typically via an access transistor, to the FeCAP, and the resulting current and, consequently, charge (probed via, e.g., voltage across capacitance) is measured. A high current or voltage indicates a state change (polarization switching), while a low current or voltage signifies no change (non- switching, only capacitor charging or discharging displacement current). This method allows the determination of the stored value but is destructive, as the read process erases the polarization state of the FeCAP, necessitating a subsequent re-write to restore the data.
[0008] Recent advancements propose the use of asymmetric FeCAP structures and a special read procedure. These asymmetric FeCAPs allow non-destructive readout due to distinct capacitance values around a bias of 0 V for different polarization states. The readout can be performed either by directly measuring the capacitance (using a periodic AC input with multiple repetitions to accumulate sufficient signal) or by applying a pulse and integrating the resulting current; the difference in total charge, caused by the change in capacitance near a bias of 0 V, provides the distinction.Some background art includes U.S. Patent Nos. 5,666,305, 3,132,326, 5,262,982 and Japanese patent application No. JPH06275062.
[0009] SUMMARY OF THE INVENTION
[0010] According to an aspect of some embodiments of the present invention there is provided a memory cell. The memory cell comprises a ferroelectric capacitor configured to be polarizable into at least two distinct polarization states, each corresponding to a different non-volatile memory state, and a memory readout circuit. The memory readout circuit is configured to apply a pulse of bias voltage to the capacitor, to measure an electric quantity indicative of a sign of a derivative of a capacitance of the capacitor, and to determine a memory state of the capacitor based on the electric quantity.
[0011] According to some embodiments of the invention the memory readout circuit is configured to ensure that the bias voltage is less than a coercive voltage of the capacitor throughout a duration of the pulse.
[0012] According to some embodiments of the invention the derivative of a capacitance of the capacitor is with respect to the bias voltage.
[0013] According to some embodiments of the invention the derivative of a capacitance of the capacitor is with respect to time.
[0014] According to some embodiments of the invention the ferroelectric capacitor is symmetric. According to some embodiments of the invention the ferroelectric capacitor is asymmetric. According to some embodiments of the invention the electric quantity comprises a difference between a peak current resulting from the bias voltage and a predetermined current threshold.
[0015] According to some embodiments of the invention the predetermined current threshold is between a first peak current characteristic of a first polarization state of the polarization states and a second peak current characteristic of a second polarization state polarization states.
[0016] According to some embodiments of the invention the predetermined current threshold is within a predetermined tolerance from a multiplication of a value of the capacitance during the pulse by a time-derivative of the bias voltage during the pulse.
[0017] According to some embodiments of the invention the electric quantity comprises a difference between an electrical charge resulting from the bias voltage and a predetermined electrical charge threshold.
[0018] According to some embodiments of the invention the electric quantity comprises a difference between an electrical charge or a peak current resulting from the bias voltage and anelectrical charge or a peak current characteristic of a reference memory cell being at a predetermined memory state.
[0019] According to some embodiments of the invention the pulse is a bipolar pulse.
[0020] According to some embodiments of the invention the electric quantity comprises a combination between an electrical charge or charge density (e.g., areal charge density) resulting from a positive-polarity pulse of the bipolar pulse, and an electrical charge or charge density resulting from a negative-polarity pulse of the bipolar pulse.
[0021] According to some embodiments of the invention the combination is a linear combination. According to some embodiments of the invention the combination is a difference.
[0022] According to some embodiments of the invention the capacitor has a ferroelectric layer made of a material selected from the group consisting of hafnium zirconate, lead zirconate titanate, barium titanate, strontium bismuth tantalate, bismuth ferrite, sodium bismuth titanate, lithium niobate, polyvinylidene fluoride, calcium titanate, zinc oxide, potassium niobate.
[0023] According to some embodiments of the invention the capacitor has a ferroelectric layer made of a ferroelectric heterostructure.
[0024] According to an aspect of some embodiments of the present invention there is provided a memory array. The memory array comprises a plurality of memory cells at addressable locations, and a respective plurality of selector devices arranged to individually access memory cells responsively to address signals. According to some embodiments of the invention at least one of the memory cells is the memory cell as delineated above and optionally and preferably as further detailed below.
[0025] According to an aspect of some embodiments of the present invention there is provided a memory array. The memory array comprises: a plurality of ferroelectric capacitors, configured to be polarizable into at least two distinct polarization states, each corresponding to a different nonvolatile memory state; a respective plurality of access transistors associated with the ferroelectric capacitors, wherein a terminal of each ferroelectric capacitor is connected to a first channel terminal of a respective access transistor; a word line connected to a gate terminal of each access transistor; a bit line connected to a second channel terminal of each access transistor. The memory array also comprises a readout circuit configured to apply, via the word line, a gate voltage to one of the access transistors so as to establish conduction between respective channel terminals of the access transistor, to apply, via the bit line, a pulse of bias voltage to a respective ferroelectric capacitor, to measure an electric quantity indicative of a sign of a derivative of a capacitance ofthe ferroelectric capacitor, and to determine a memory state of the ferroelectric capacitor based on the electric quantity.
[0026] According to some embodiments of the invention the array is a two-dimensional array. According to some embodiments of the invention the plurality of ferroelectric capacitors are arranged in rows and columns. According to some embodiments of the invention the word line is connected to the gate terminal of each access transistor in a row, and the bit line is connected to a second channel terminal of each access transistor in a column.
[0027] According to some embodiments of the invention the array is a three-dimensional array. According to some embodiments of the invention the plurality of ferroelectric capacitors are arranged in layers, wherein each layer comprises ferroelectric capacitors arranged in rows and columns. According to some embodiments of the invention the word line is connected to the gate terminal of each access transistor in a row, and the bit line is connected to a second channel terminal of each access transistor in a column, and wherein the array comprises vertical lines connecting channel terminals of different layers.
[0028] According to an aspect of some embodiments of the present invention there is provided a memory readout system, comprises a memory readout circuit configured to apply a pulse of bias voltage to a ferroelectric capacitor, to measure an electric quantity indicative of a sign of a derivative of a capacitance of the capacitor, and to determine a memory state of the capacitor based on the electric quantity.
[0029] According to some embodiments of the invention the memory cell or memory array or memory readout system serves as a component in a system selected from the group consisting of neuromorphic circuitry, a display system, an imager system, a digital signal processing system, a programmable logic system, a data storage system, a sensor system, a mobile computing system, a robotic system, a smart home system, a medical system, a network switch system, a barcode scanner system, and a battery management system.
[0030] According to an aspect of some embodiments of the present invention there is provided a method of reading information from a ferroelectric capacitor configured to be polarizable into at least two distinct polarization states, each corresponding to a different non-volatile memory state, the method comprises: applying a pulse of bias voltage to the ferroelectric capacitor; measuring an electric quantity indicative of a sign of a derivative of a capacitance of the capacitor; and determining a memory state of the capacitor based on the electric quantity.
[0031] According to an aspect of some embodiments of the present invention there is provided a method of neuromorphic computing. The method comprises executing the method for readinginformation from a ferroelectric capacitor as delineated above and optionally and preferably as further detailed below, and performing at least one neuromorphic computation operation based on the information.
[0032] According to an aspect of some embodiments of the present invention there is provided a method of operating a display system. The method comprises executing the method for reading information from a ferroelectric capacitor as delineated above and optionally and preferably as further detailed below, and controlling at least one display-driving operation based on the information.
[0033] According to an aspect of some embodiments of the present invention there is provided a method of operating an imager system. The method comprises executing the method for reading information from a ferroelectric capacitor as delineated above and optionally and preferably as further detailed below, and performing at least one image sensing and / or image-processing operation based on the information.
[0034] According to an aspect of some embodiments of the present invention there is provided a method of digital signal processing. The method comprises executing the method for reading information from a ferroelectric capacitor as delineated above and optionally and preferably as further detailed below, and performing at least one digital signal processing operation based on the information.
[0035] According to an aspect of some embodiments of the present invention there is provided a method of operating a programmable logic system. The method comprises executing the method for reading information from a ferroelectric capacitor as delineated above and optionally and preferably as further detailed below, and configuring and / or reconfiguring at least one programmable logic resource based on the information.
[0036] According to an aspect of some embodiments of the present invention there is provided a method of operating a data storage system. The method comprises executing the method for reading information from a ferroelectric capacitor as delineated above and optionally and preferably as further detailed below, and performing at least one data storage operation based on the information.
[0037] According to an aspect of some embodiments of the present invention there is provided a method of operating a sensor system. The method comprises executing the method for reading information from a ferroelectric capacitor as delineated above and optionally and preferably as further detailed below, and performing at least one sensor-related operation based on the information.According to an aspect of some embodiments of the present invention there is provided a method of operating a mobile computing system. The method comprises executing the method for reading information from a ferroelectric capacitor as delineated above and optionally and preferably as further detailed below, and performing at least one mobile computing operation based on the information.
[0038] According to an aspect of some embodiments of the present invention there is provided a method of operating a robotic system. The method comprises executing the method for reading information from a ferroelectric capacitor as delineated above and optionally and preferably as further detailed below, and performing at least one robotic control operation based on the information.
[0039] According to an aspect of some embodiments of the present invention there is provided a method of operating a smart home system. The method comprises executing the method for reading information from a ferroelectric capacitor as delineated above and optionally and preferably as further detailed below, and performing at least one smart home control operation based on the information.
[0040] According to an aspect of some embodiments of the present invention there is provided a method of operating a medical system. The method comprises executing the method for reading information from a ferroelectric capacitor as delineated above and optionally and preferably as further detailed below, and performing at least one medical-device operation based on the information.
[0041] According to an aspect of some embodiments of the present invention there is provided a method of operating a network switch system. The method comprises executing the method for reading information from a ferroelectric capacitor as delineated above and optionally and preferably as further detailed below, and performing at least one packet-processing and / or routing operation based on the information.
[0042] According to an aspect of some embodiments of the present invention there is provided a method of operating a barcode scanner system. The method comprises executing the method for reading information from a ferroelectric capacitor as delineated above and optionally and preferably as further detailed below, and performing at least one barcode scanning operation based on the information.
[0043] According to an aspect of some embodiments of the present invention there is provided a method of operating a battery management system. The method comprises executing the method for reading information from a ferroelectric capacitor as delineated above and optionally andpreferably as further detailed below, and performing at least one battery management operation based on the information.
[0044] Unless otherwise defined, all technical and / or scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which the invention pertains. Although methods and materials similar or equivalent to those described herein can be used in the practice or testing of embodiments of the invention, exemplary methods and / or materials are described below. In case of conflict, the patent specification, including definitions, will control. In addition, the materials, methods, and examples are illustrative only and are not intended to be necessarily limiting.
[0045] Implementation of the method and / or system of embodiments of the invention can involve performing or completing selected tasks manually, automatically, or a combination thereof. Moreover, according to actual instrumentation and equipment of embodiments of the method and / or system of the invention, several selected tasks could be implemented by hardware, by software or by firmware or by a combination thereof using an operating system.
[0046] For example, hardware for performing selected tasks according to embodiments of the invention could be implemented as a chip or a circuit. As software, selected tasks according to embodiments of the invention could be implemented as a plurality of software instructions being executed by a computer using any suitable operating system. In an exemplary embodiment of the invention, one or more tasks according to exemplary embodiments of method and / or system as described herein are performed by a data processor, such as a computing platform for executing a plurality of instructions. Optionally, the data processor includes a volatile memory for storing instructions and / or data and / or a non-volatile storage, for example, a magnetic hard-disk and / or removable media, for storing instructions and / or data. Optionally, a network connection is provided as well. A display and / or a user input device such as a keyboard or mouse are optionally provided as well.
[0047] BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWINGS
[0048] The patent or application file contains at least one drawing executed in color. Copies of this patent or patent application publication with color drawing(s) will be provided by the Office upon request and payment of the necessary fee.
[0049] Some embodiments of the invention are herein described, by way of example only, with reference to the accompanying drawings. With specific reference now to the drawings in detail, it is stressed that the particulars shown are by way of example and for purposes of illustrativediscussion of embodiments of the invention. In this regard, the description taken with the drawings makes apparent to those skilled in the art how embodiments of the invention may be practiced.
[0050] In the drawings:
[0051] FIG. 1 is a schematic illustration of a quasi-static capacitance-voltage (C-V) curve which depends on a polarization state, and which can be used for deriving a transient current in response to a low-bias voltage ramp in a ferroelectric capacitor (FeCap) according to some embodiments of the present invention.
[0052] FIGs. 2A-C show a FeCap structure and measurement setup, according to some embodiments of the present invention. FIG. 2A is a schematic illustration of the FeCap structure, FIG. 2B is an optical image supplemented with electronic circuit symbols, and FIG. 2C is enlarged optical image of part of FIG. 2B.
[0053] FIGs. 3 A and 3B show C-V measurements (FIG. 3 A) of two representative ferroelectric capacitors, after PRG and ERS operations, and write endurance measurement (FIG. 3B) showing evolution of remanent polarization (2Pr) and breakdown after 4- 108switching cycles, as obtained in experiments performed according to some embodiments of the present invention. The increase in leakage current indicates device degradation preceding breakdown.
[0054] FIG. 4 shows polarization-voltage (P-V) measurements from positive-up-negative-down (PUND) tests at varying sweep frequencies, as obtained in experiments performed according to some embodiments of the present invention. An increase in coercive voltage (Vc) with higher frequency (shorter ramp time) is demonstrated.
[0055] FIG. 5 shows 5RC delay estimation, as obtained by modeling performed according to some embodiments of the present invention. Shown is a 5RC delay estimation of square FeCAP structure versus length scale L, for different thickness of HZO.
[0056] FIGs. 6 A and 6B describe a non-destructive readout (NDRO) according to some embodiments of the present invention. FIG. 6A illustrates NDRO measurement scheme in unipolar and bipolar configurations, and FIG. 6B shows current waveform in bipolar configuration, as measured in experiments performed according to some embodiments of the present invention, demonstrating positive and negative read pulses after PRG and ERS operations. Read pulse ramp-up / down time was 800 ps. In FIG. 6B, the MW is expressed in terms of Al, which is proportional to 2V-dC / dV.
[0057] FIGs. 7A-C show unipolar NDRO, as obtained in experiments performed according to some embodiments of the present invention. FIG. 7A shows measured current I during unipolar NDRO at room temperature after ERS and PRG vs. number of read cycles. The first Ik (dots) andthe subsequent 1013cycles show Min, Max, Avg, and Std values of 1600 reads / decade. FIG. 7B is a histogram of results from FIG. 7A. FIG. 7C shows the difference between the transient current peaks vs. read cycles. The difference between the transient current peaks can be used to represent MW. At 0.7 V after 5- 1012cycles, ERS state is lost due to positive accumulative switching. Lower read at 0.6 V shows narrower initial MW, preserved at 1013read cycles.
[0058] FIGs. 8A-C show bipolar NDRO (positive read), as obtained in experiments performed according to some embodiments of the present invention. FIG. 8A shows measured current peaks during positive read of bipolar NDRO at room temperature after ERS and PRG vs. number of read cycles. The first Ik read cycles (dots) and subsequent 1013cycles show Min, Max, Avg, and Std values of 1600 reads per decade. FIG. 8B shows histogram of results from FIG. 8 A. FIG. 8C shows the difference between the transient current peaks (which difference can be used to represent MW) vs. read cycles. A drift in polarization states is observed, attributed to retention effects.
[0059] FIGs. 9A-C show bipolar NDRO (negative read), as obtained in experiments performed according to some embodiments of the present invention. FIG. 9A shows measured current peaks during negative read of bipolar NDRO at room temperature after ERS and PRG vs. number of read cycles. The first Ik read cycles (dots) and subsequent 1013cycles show Min, Max, Avg, and Std values of 1600 reads per decade. FIG. 9B shows histogram of results from FIG. 9A. FIG. 9C shows the difference between the transient current peaks (which difference can be used to represent MW) vs. read cycles.
[0060] FIGs. 10A-C show Bipolar NDRO at temperature of 125°C, as obtained in experiments performed according to some embodiments of the present invention. FIG. 10A shows measured current during bipolar NDRO at 125°C after ERS and PRG vs. read cycles. The first Ik read cycles (dots) and subsequent 1013cycles show Min, Max, Avg, and Std values of 1600 reads per decade. FIG. 10B is a histogram of results from FIG. 10A. FIG. 10C the difference between the transient current peaks (which difference can be used to represent MW) vs. number of read cycles during reads of bipolar NDRO at 125°C. The degradation of the MW is attributed to retention due to more than 20 hours accumulated test time. Preservation of MW after 1013read cycles is demonstrated at 125°C.
[0061] FIGs. 11A and 11B show retention measurements, performed according to some embodiments of the present invention. FIG. 11A shows the difference between the transient current peaks (which difference can be used to represent MW) using the NDRO method of the present embodiments at room temperature and 85°C. The read conditions are 0.7 V and 700 ps for the room temperature measurements, and 0.7 V and 800 ps for the 85°C measurements. Dashedlines represent simple log fitting extrapolation to 10 years. FIG. 1 IB shows readout current using the NDRO method of the present embodiments as a function of the retention time after PRG and ERS operations at 125°C, and readout currents from the measurement shown in FIGs. 10A-C, plotted against the effective accumulated measurement time (dashed).
[0062] FIGs. 12A-C are schematic illustrations of relationships between capacitance and voltages in FeCAPs.
[0063] FIGs. 13 A and 13B are schematic illustrations of an NDRO technique based on peak current measurements during an applied pulse, according to some embodiments of the present invention.
[0064] FIG. 14 is a schematic illustration showing a bipolar NDRO technique in which the read operation alternates between positive and negative pulses, according to some embodiments of the present invention.
[0065] FIGs. 15A and 15B show integration scheme for two memory states, according to some embodiments of the present invention. The integrated quantity in these embodiments is a transient current response.
[0066] FIG. 16A and 16B are histograms showing example results Qp, QN of the integration described in FIGs. 15A and 15B for a positive-polarity pulse (FIG. 16A) and a negative-polarity pulse (FIG. 16B) of a bipolar read pulse, according to some embodiments of the present invention.
[0067] FIG. 17A and 17B are graphs showing example results Qp, QN of the integration described in FIGs. 15A and 15B for a positive-polarity pulse (FIG. 17A) and a negative-polarity pulse (FIG.
[0068] 17B) of a bipolar read pulse, according to some embodiments of the present invention.
[0069] FIGs. 18A and 18B show direct state decoding based on a difference AQ=QP-QN, according to some embodiments of the present invention.
[0070] FIGs. 19A and 19B show measured value of the difference AQ following a bipolar NDRO after ERS and PRG, where FIG. 19A shows AQ as a function of the number of read cycles, and FIG. 19B shows histogram of results from FIG. 19A.
[0071] FIG. 20 is a schematic illustration of a memory cell, according to some embodiments of the present invention.
[0072] FIGs. 21A and 21B are schematic illustrations of memory array, according to some embodiments of the present invention.
[0073] FIG. 22 is a schematic illustration of memory array, in embodiments of the present invention in which the array comprises access transistors, word lines and bit lines.DESCRIPTION OF SPECIFIC EMBODIMENTS OF THE INVENTION
[0074] The present invention, in some embodiments thereof, relates to solid state electronics, and more particularly, but not exclusively, to a memory cell and method of retrieving information therefrom.
[0075] Before explaining at least one embodiment of the invention in detail, it is to be understood that the invention is not necessarily limited in its application to the details of construction and the arrangement of the components and / or methods set forth in the following description and / or illustrated in the drawings and / or the Examples. The invention is capable of other embodiments or of being practiced or carried out in various ways.
[0076] Referring now to the drawings, FIG. 20 illustrates a memory cell 10, according to some embodiments of the present invention. Memory cell 10 comprises a ferroelectric capacitor 12 configured to be polarizable into two or more distinct polarization states, each corresponding to a different non-volatile memory state. Ferroelectric capacitor 12 can comprise two conductive electrodes 14, 16, and a dielectric layer 18 disposed between the electrodes 14 and 16. Dielectric layer 18 comprises a material exhibiting ferroelectricity, and is referred to below as "ferroelectric layer." Ferroelectric layer 18 is configured to assume a remanent polarization state in response to an applied electric field and to retain at least a portion of this polarization state after removal of the applied electric field. Ferroelectric layer 18 is also configured to reversibly switch the polarization between two or more different states in response to different applied electric fields (e.g., of opposite polarity), wherein the switching occurs upon application of an electric field exceeding a coercive threshold.
[0077] Conductive electrodes 14, 16, can be made of any conductive material such as, but not limited to, gold, silver, copper, aluminum, tungsten, titanium, tantalum, molybdenum, nickel, cobalt, chromium, ruthenium, rhodium, iridium, platinum, palladium, osmium, rhenium, niobium, vanadium, hafnium, zirconium, titanium nitride, tantalum nitride, tungsten nitride, hafnium nitride, zirconium nitride, molybdenum nitride, ruthenium oxide, iridium oxide, strontium ruthenate, lanthanum nickelate, indium tin oxide, fluorine-doped tin oxide, aluminum-doped zinc oxide, polysilicon, doped polysilicon, metal silicides, conductive carbides, conductive oxides, conductive nitrides, conductive alloys, multilayer electrode stacks, and combinations thereof. In some embodiments of the present invention one or both of electrodes 14 and 16 comprise a metal layer in direct contact with ferroelectric layer 18, optionally and preferably together with one or more additional conductive layers serving, for example, as adhesion layers, diffusion barriers, seedlayers, stress-engineering layers, routing layers, probe-pad layers, capping layers, oxidationresistant layers, and the like.
[0078] Representative examples of ferroelectric material suitable for use as layer 18 include, without limitation, hafnium zirconate, lead zirconate titanate, barium titanate, strontium bismuth tantalate, bismuth ferrite, sodium bismuth titanate, lithium niobate, polyvinylidene fluoride, calcium titanate, zinc oxide, potassium niobate. Also contemplated are embodiments in which ferroelectric layer 18 is made of a ferroelectric heterostructure. Representative examples of ferroelectric heterostructures include, without limitation, multilayer hafnium-oxide-based stacks, such as HfO2 / ZrO2, HfO2 / HfZrO2, HfZrO2 / A12O3, HI rO2 / SiO2, HfZrO2 / TiO2, HfZrO2 / HfO2 / Al2O3, HfZrO2 / ZrO2 / HfZrO2, doped hafnium-oxide superlattices, nanolaminates comprising alternating ferroelectric and dielectric sublayers, perovskite-based heterostructures such as PZT / BaTiOs, BaTiOs / SrTiOs, PbTiOs / SrTiOs, BiFcOs / SrTiOs. Additional heterostructures include, without limitation, Aurivillius-phase multilayers, van der Waals ferroelectric hetero structures, polymer / inorganic ferroelectric composites, ferroelectric / antiferroelectric stacks, and ferroelectric / semiconductor stacks. The hetero structure can be based on one or more desired characteristics such as, but not limited to, coercive voltage, remanent polarization, leakage current, stress state, crystallographic phase stabilization, and imprint.
[0079] In some embodiments of the present invention ferroelectric capacitor 12 is symmetric, and in some embodiments of the present invention ferroelectric capacitor is 12 asymmetric.
[0080] As used herein, "symmetric ferroelectric capacitor" refers to a ferroelectric capacitor for which the characteristic dependence of the capacitance on a voltage across the capacitor (or equivalently an electric field between the electrodes) corresponding to different polarization states are generally mirror-symmetric (within a tolerance of 10%) with respect to voltage inversion in a vicinity of zero bias, such that the capacitance at or around zero bias does not substantially distinguish between the polarization states.
[0081] As used herein, "asymmetric ferroelectric capacitor" refers to a ferroelectric capacitor for which the characteristic dependences of the capacitance on a voltage across the capacitor (or equivalently an electric field between the electrodes) corresponding to different polarization states are not mirror-symmetric with respect to voltage inversion in a vicinity of zero bias, such that the capacitance at or around zero bias differs between polarization states.It is expected that during the life of a patent maturing from this application many relevant ferroelectric capacitors will be developed and the scope of the term ferroelectric capacitor is intended to include all such new technologies a priori.
[0082] Memory cell 10 also comprises a memory readout circuit 20 configured to apply a pulse of bias voltage to capacitor 12 (between electrodes 14 and 16) so as to generate an electrical field within layer 18. The applied pulse can be unipolar or bipolar. Preferably, the level of the applied bias voltage is less than the coercive threshold throughout a duration of the applied pulse. The advantage of this embodiment is that it allows non-destructive readout of the non-volatile memory state of capacitor 12.
[0083] As used herein, "non-destructive readout" refers to a read operation in which the stored polarization state is sensed without switching the ferroelectric polarization from one memory state to another, such that, after completion of the read operation, the capacitor remains in the same memory state that existed before the read operation, except for a variation that is smaller than 10%, or smaller than 5%, or smaller than 2%, or smaller than 1%, of a memory window between the memory states. The memory window can be defined as the separation between readout charges or areal charge densities corresponding to different memory states. It is appreciated that since the readout charges can be derived from other measured observables (e.g., current, capacitance), the memory window correlates with such measured observables. Therefore, the memory window can be represented by each such measured observable. Thus, the memory window can be represented by observables such as current or capacitance. In experiments performed by the Inventors, the memory window was conveniently represented in the current domain as the difference between transient current peaks corresponding to different memory states.
[0084] Optionally and preferably one or more parameters defining the pulse are selected to increase the value of the memory window to be above a predetermined threshold (e.g., above IfC / pm2), or to maximize it. Representative examples of pulse parameters that can be selected to increase the memory window include, without limitation, the amplitude of the pulse, the width of the pulse, the rise and / or fall times of the pulse, the slew rate of the pulse, the shape of the pulse (e.g., rectangular shape, trapezoidal shape, triangular shape, multi-level shape, pulse with shaped edges), the duty cycle, the repetition rate, the number of pulses per read, and so on. When the pulse is a bipolar pulse, each of these parameters can be selected independently for the positivepolarity pulse and the negative-polarity pulse of the bipolar pulse. Additional parameters which can be selected when the pulse is bipolar include, without limitation, the ratio between the amplitudes of the positive-polarity pulse and the negative-polarity pulse, the delay between thepositive-polarity and the negative-polarity pulse (including, but not limited to, the case of zero delay), the bipolar symmetry or lack of bipolar symmetry, and the extent of bipolar asymmetry.
[0085] Readout circuit 20 is also configured to measure an electric quantity exhibited by capacitor 12 in response to the applied pulse of bias voltage. Preferably, the electric quantity is indicative of a sign of a derivative of a capacitance of capacitor 12. The derivative can be with respect to the bias voltage or with respect to time. Readout circuit 20 is also configured to determine the memory state of capacitor 12 based on the measured electric quantity. Typically, circuit 20 measures a transient current, and / or a transient voltage, and then computes or otherwise obtains a value of the electric quantity from the measured current and / or voltage. The memory state of capacitor 12 is then determined by applying a decision operation to the obtained electric quantity. For example, circuit 20 can determine the memory state based on a sign of the electric quantity, e.g., a negative value indicates a PRG (T) state and a positive value indicates an ERS ('0') state. In another embodiment, a thresholding procedure is applied to electric quantity, wherein values of electric quantity on one side of a predetermined threshold are associated with the PRG (T) state and values of electric quantity on an opposite side of the predetermined threshold are associated with the ERS ('0') state. Use of more than one threshold is also contemplated and is useful when capacitor 12 stores more than two memory states. In such embodiments, different ranges of values of the electric quantity are mapped to respective memory states. For example, when capacitor 12 is configured to store three memory states, circuit 20 can compare the value of the electric quantity to a first and a second thresholds, such that a value below the first threshold indicates a first state, a value between the first and second thresholds indicates a second state, and a value above the second threshold indicates a third state. Any number of thresholds can be employed to distinguish among a corresponding plurality of memory states.
[0086] In some embodiments of the present invention circuit 20 applies a single pulse and determines the memory state from a response to that single pulse, and in some embodiments circuit 20 applies two or more pulses, of the same polarity or of opposite polarities, and combines information extracted from the responses to the multiple pulses in order to determine the memory state.
[0087] In some embodiments of the present invention readout circuit 20 can be formed on-chip with capacitor 12, optionally and preferably on a same die or substrate. Alternatively, circuit 20 can be partially or wholly off-chip, e.g., in a separate integrated circuit.
[0088] Circuit 20 can be embodied as analog circuitry, digital circuitry, or can comprise both analog circuitry for executing some operations described herein and digital circuitry for executingother operations described herein. Thus, circuit 20 can determine the electric quantity directly in the analog domain, for example, by integration, peak detection, subtraction, ratio generation, filtering, rectification, correlation, or threshold comparison, and / or in the digital domain, for example, by digitizing a waveform or one or more sampled values and executing one or more arithmetic, logical, statistical, or decision operations thereon. Also contemplated are embodiments in which a first part of the processing is analog and a second part is digital, for example, where a transient current is first converted to a voltage, integrated, filtered, amplified, or sampled, and thereafter digitized and processed to obtain the electric quantity and / or the memory decision. Circuit 20 can be dedicated to a single capacitor 12 or shared among a plurality of capacitors, as further detailed below.
[0089] Following are examples for electric quantities that are suitable for the present embodiments. Circuit 20 can measure or compute any of these electric quantities, including combinations of two or more electric quantities.
[0090] In some embodiments of the present invention the electric quantity comprises a difference between a peak current resulting from bias voltage and a predetermined current threshold. The predetermined current threshold can be selected in more than one way. For example, the predetermined current threshold can be between a first peak current characteristic of a first polarization state and a second peak current characteristic of a second polarization state, e.g., between characteristic values associated with PRG (T) and ERS ('0'). In some embodiments of the present invention the predetermined current threshold is selected to lie midway between two characteristic peak currents, and in some embodiments of the present invention it is offset from the midpoint, for example, to compensate for noise, drift, asymmetry, parasitic current, retention drift, temperature dependence, process spread, desired decision margin, or differing error costs for the two states.
[0091] In some embodiments of the present invention the predetermined current threshold can be within a predetermined tolerance e.g., ±10%) from C-dV / dt, where C is a value of the capacitance during the pulse and dV / dt is a time-derivative of the bias voltage during the pulse. In these embodiments the predetermined current threshold corresponds to an expected contribution of C-dV / dt to the current, and the electric quantity can indicate whether the measured peak current is greater or smaller than that expected contribution. The capacitance value C can be an instantaneous value, an average value, a calibrated value, a nominal value, a zero-bias value, a value associated with a portion of the pulse, a value inferred from prior measurements, or any other representative value.In some embodiments of the present invention the electric quantity is based on an electrical charge resulting from the bias voltage. For example, the electric quantity can comprise a difference between the electrical charge resulting from the bias voltage and a predetermined electrical charge threshold. Suitable techniques for determining the electrical charge include, without limitation, integrating current over time by digital or analog processing, accumulating charge on an integration capacitor, charge sharing, and switched-capacitor techniques. The predetermined electrical charge threshold is optionally and preferably between characteristic charge values of the two polarization states, and can be selected by calibration or modeling.
[0092] In some embodiments of the present invention a reference memory cell 22 is used by circuit 20 to measure the electric quantity. Reference memory cell 22 can be similar to memory cell 10, namely comprising a ferroelectric capacitor 12r and optionally and preferably also a circuit 20r, except that it is prepared in advance at a predetermined memory state, namely pre-programmed to a known state for use as a reference. Circuit 20 of cell 10 can communicate with circuit 20r of cell 22, or directly with capacitor 12r, to measure a characteristic electrical charge or peak current of cell 22. In these embodiments, the electric quantity can comprise a difference between the electrical charge or peak current resulting from the bias voltage applied to cell 10 and the characteristic electrical charge or peak current of reference memory cell 22. The reference memory cell 22 may provide a static reference or a dynamically refreshed reference, and one or more reference memory cells may be used. For example, one reference memory cell may be maintained in a PRG (T) state, one reference memory cell may be maintained in an ERS ('0') state, and a decision threshold may be generated from one or both reference responses. Also contemplated are embodiments in which the reference memory cell is selected to track local variations in one or more of temperature, parasitics, noise conditions, and the like.
[0093] When the applied pulse is bipolar, the electric quantity can comprise a difference between the electrical charge or charge density (e.g., areal charge density) resulting from a positive-polarity pulse of the bipolar pulse, and the electrical charge or charge density resulting from a negativepolarity pulse of the bipolar pulse. For example, circuit 20 can extract a first quantity QP from the transient response to the positive-polarity pulse and a second quantity QN from the transient response to the negative-polarity pulse, compute a combination of QP and QN, and then determine the memory state based on the computed combination. Examples of combinations between QP and QN suitable for the present embodiments include, without limitation, a linear combination (with uniform or nonuniform weights), a ratio (signed ratio, absolute ratio, normalized ratio), aproduct, a quotient, a polynomial function, and a nonlinear transform. In experiments performed by the inventors the combination was a linear combination in the form AQ=QP-QN.
[0094] Preferably, the positive-polarity pulse and the negative-polarity pulse of the bipolar read pulse are applied consecutively within a single read event, whereby QP and QN are obtained from the positive-polarity and negative-polarity pulses of the bipolar pulse of the same read event. The advantage of these embodiments is that common-mode effects, slow drift, baseline offset, gain drift, parasitic response, and other variations that are present in both pulse responses are at least partially reduced in the computed combination. The positive-polarity pulse and the negativepolarity pulse can have equal or unequal magnitudes, equal or unequal widths, equal or unequal slew rates, equal or unequal shapes, and can be contiguous or separated by a delay. The quantities QP and QN can be expressed as total charge, signed charge, unsigned charge, magnitude of charge, integrated current, weighted integrated current, charge density (e.g., areal charge density), normalized charge density (e.g., normalized areal charge density) or any other quantity derived from the respective responses.
[0095] In some embodiments of the present invention the electric quantity is obtained, at least in part, by windowed and / or weighted integration. In these embodiments circuit 20 evaluates an expression of the form Qw=Jw(t)|I(t)|dt, where w(t) is a weighting function that selectively amplifies time intervals that are more informative for distinguishing between memory states and suppress other time intervals. The weighting function can define one or more windows around a rising edge of the transient response, a falling edge of the transient response, both edges of the transient response, a central portion of the transient response, a tail portion of the response, and the like. The weighting can be binary, piecewise constant, continuously varying, or otherwise selected. The advantage of windowed integration is that it reduces the influence of less informative parts of the waveform and / or parasitic components that occur outside the selected window.
[0096] In some embodiments of the present invention the electric quantity is obtained following baseline subtraction or common-mode cancellation. In these embodiments circuit 20 can subtract a local baseline, a reference segment, a pre -pulse segment, a post-pulse segment, a quiet-time segment, a reference waveform, or another estimate of common-mode behavior from the measured current or voltage, for example, according to T(t)=I(t)-Ibaseiine(t), where Ibaseiine(t) is a baseline response current. The baseline can be obtained from an earlier portion of the same read event, from a separate calibration event, from a reference memory cell, from neighboring cells of a memory array, from repeated reads, or from a stored model. The advantage of such subtraction isthat it reduces offsets, drift, leakage, feedthrough, capacitive coupling, instrumentation artifacts, routing artifacts, and other components that are not strongly indicative of the stored state.
[0097] In some embodiments of the present invention the electric quantity is obtained, at least in part, using matched filtering or correlation. In these embodiments circuit 20 correlates the measured waveform with one or more stored templates, for example, according to M=fl(t)T(t)dt, where T(t) is a template waveform characteristic of a given state, operating condition, or expected response feature. One template can correspond to PRG (T), another template can correspond to ERS ('0'), and the state decision can be made by comparing two or more correlation values. The templates can be generated from previously measured responses, calibration data, simulation, adaptive learning, averaged historical responses, or analytical models. The advantage of using matched filtering is that it improves detection under noisy conditions. The template waveform can be a current waveform, a voltage waveform, or otherwise transformed responses.
[0098] In any of the above embodiments, circuit 20 can apply one or more nonlinear transforms to the measured response before computing the electric quantity. Representative examples of such nonlinear transforms include, without limitation, rectification, squaring, clipping, compression, logarithmic transformation, exponential transformation, and the like. The advantage of such nonlinear processing is that it can increase signal-to-noise ratio, improve robustness to outliers, and emphasize a selected region of the waveform.
[0099] In any of the above embodiments the electric quantity can be obtained from multiple samples. In these embodiments, the determination of the memory state of capacitor 12 is based on a multi-sample decision. For example, circuit 20 can compute a decision quantity for multiple bipolar pulse pairs, multiple single pulses, multiple repeated reads, or multiple time windows, and then apply majority voting, weighted voting, confidence scoring, averaging, median selection, outlier rejection, sequential testing, or another statistical decision procedure. The advantage of this approach is that it provides a tradeoff between read latency and read robustness. The number of samples may be fixed or adaptive. For example, circuit 20 can terminate the decision early once a confidence criterion is met, or can apply additional pulses if the confidence is below a threshold.
[0100] In any of the above embodiments, the electric quantity is optionally and preferably normalized so as to reduce sensitivity to drift, parasitics, gain variation, and / or operating-point changes. For example, when QP-QN is calculated, it can be normalized by QP+QN so that the electrical quantity is (QP-QN) / (QP+QN).
[0101] Reference is now made to FIGs. 21A-B, which are schematic illustrations of a memory array 30 according to some embodiments of the present invention. Memory array 30 comprises aplurality of memory cells at addressable locations, and a respective plurality of selector devices 32 arranged to individually access the memory cells responsively to address signals. Each of the memory cells comprises a ferroelectric capacitor as further detailed hereinabove. For example, one or more of the memory cells of array 30 can optionally and preferably be memory cell 10 as further detailed hereinabove. In some embodiments of the present invention the selector devices 32 are part of the cells of the array, so that each array cell comprises also its respective selector device 32.
[0102] The memory cells of array 30 are preferably arranged over a grid, so that the addressable location of each memory cell can be defined using the grid coordinates as known in the art. The grid can be rectangular, hexagonal, triangular, radial, polar, or any other grid or addressable arrangement known in the art. In some embodiments of the present invention array 30 is a two-dimensional array, as schematically illustrated in FIG. 21 A, and in some embodiments of the present invention array 30 is a three-dimensional array, as schematically illustrated in FIG. 2 IB. When the array is three-dimensional, the memory cells can be arranged in layers 40, as illustrated, optionally and preferably with each layer itself forming a two-dimensional arrangement such as rows and columns. Also contemplated are embodiments in which the three-dimensional arrangement is not strictly layered, for example, interleaved arrangements, monolithic stacked arrangements, wafer-stacked arrangements, heterogeneous stacked arrangements, folded arrangements, terraced arrangements, pillar-based arrangements, and other volumetric organizations in which memory cells are distributed over more than one elevation level.
[0103] Selector devices 32 serve for selectively coupling individual memory cells to read and / or write lines (not shown) of array 30, so that a selected memory cell can be accessed while other memory cells are at least partially isolated. Selector devices 32 can provide at least one of: addressing functionality, isolation functionality, current-limiting functionality, and sneak path suppression. Suitable devices which can use as selector devices 32 include, without limitation, field-effect transistors, access transistors, MOS transistors, CMOS transistors, thin-film transistors, junction transistors, bipolar transistors, pass transistors, transmission gates, diodes, pn diodes, Schottky diodes, threshold switches, ovonic threshold switches, two-terminal selectors, memristive selectors, tunneling devices, ferroelectric transistors, and any combination thereof. In operation, an address signal can place a selected selector device 32 in a relatively conductive state, thereby electrically coupling its associated memory cell to one or more lines used for read and / or write, while selector devices associated with unselected cells remain in a relatively non-conductive state so as to reduce disturbance and undesired current paths. In some embodiments of the presentinvention the selector device 32 is controlled by a gate signal, and in some embodiments the selector device exhibits threshold behavior and is enabled when a voltage across it exceeds a threshold condition.
[0104] Each of the memory cells 10 of array 30 may have its own individual readout circuit 20 (see FIG. 20). Alternatively, two or more of the memory cells of array 30 (e.g., all the memory cells of array 30) may share a readout circuit, in which case it is not necessary for each individual memory cell to have a separate readout circuit. When two or more of the memory cells of array 30 share a readout circuit, the shared readout circuit 20 is configured to transmit address signals so as to operate one or more selector devices 32 and thereby select a particular memory cell or subset of memory cells for access, and thereafter to apply the pulse of bias voltage to the selected ferroelectric capacitor, optionally and preferably so as to allow non-destructive readout, to measure an electric quantity exhibited by the capacitor of the selected cell, preferably an electric quantity indicative of a sign of a derivative of its capacitance, and to determine the memory state of the selected capacitor based on the measured electric quantity, all as further detailed above. In some embodiments of the present invention the shared circuit 20 sequentially reads different cells of array 30, and in some embodiments the shared circuit 20 reads two or more cells in parallel using replicated front ends, multiplexing, time-division operation, frequency-division operation, code-division operation, or any other shared-access technique.
[0105] A more detailed example of memory array 30 for the case in which all the memory cells share the same readout circuit 20 is illustrated in FIG. 22. While the configuration illustrated in FIG. 22 shows a two-dimensional array of memory cells, this need not necessarily be the case, since configurations in which the array 30 is three-dimensional are also contemplated. In the configuration illustrated in FIG. 22, which is not to be considered as limiting, each array cell comprises ferroelectric capacitor 12 and a selector device 32, where the selector devices 32 are embodied as access transistors, wherein the storage terminal 34 of each capacitor 12 is connected to a first channel terminal 36 of a respective access transistor 32. Access transistor 32 can be of any known type, including, without limitation, a planar MOSFET, a fin field-effect transistor (FinFET), a gate-all-around field-effect transistor (GAAFET), a nanosheet field-effect transistor, a silicon-on-insulator (SOI) MOSFET, a high-voltage MOSFET (including EDMOS), a thin-film transistor (TFT) (including oxide TFTs and polycrystalline- silicon TFTs), a pass transistor, or a transmission-gate device, and can be an n-type transistor, a p-type transistor, or part of a complementary configuration. In some embodiments of the present invention the accesstransistors 32 are part of the cells of the array, so that each array cell comprises a capacitor 12 and an access transistor device 32 connected thereto.
[0106] The storage terminal 34 can be connected to one of the electrodes of the ferroelectric capacitor (for example, electrode 14 of capacitor 12, see FIG. 20). The other electrode of the ferroelectric capacitor 12 (for example, electrode 16) can be connected to a reference, common, plate, ground, or other bias node of array 30. For clarity of illustration, the other electrode is shown as connected to ground, but this depiction is not to be considered as limiting, and the electrode can instead be connected to any other line or node.
[0107] Array 30 also comprises one or more word lines 38 connected to the gate terminal 42 of each access transistor 32, and one or more bit lines 44 connected to a second channel terminal 46 of each access transistor 32. Word line(s) 38 and bit line(s) 44 are arranged according to the geometrical organization of the array. For example, when the ferroelectric capacitors are arranged in rows and columns, each word line 38 can be connected to the gate terminal 42 of each access transistor 32 in a row, and each bit line 44 can be connected to a second channel terminal 46 of each access transistor 32 in a column. In these embodiments, selection of a particular word line and a particular bit line identifies a particular array location. Also contemplated are embodiments in which the word lines and / or bit lines are segmented, folded, shared only by subsets of cells, or connected through multiplexers, decoders, repeaters, switches, or routing networks.
[0108] In embodiments in which array 30 is a three-dimensional array, the ferroelectric capacitors 12 are optionally and preferably arranged in layers, wherein each layer can comprise ferroelectric capacitors arranged in rows and columns. In such embodiments the word line(s) 38 can be connected to the gate terminal 42 of each access transistor 32 in a row, the bit line(s) 44 can be connected to a second channel terminal 46 of each access transistor 32 in a column, and array 30 can also comprise vertical lines 48 (perpendicular to the plane of FIG. 22, in this example, hence shown as zero-dimensional objects), connecting channel terminals and / or gate terminals of different layers. The vertical lines 48 can be embodied as through-silicon vias, monolithic intertier vias, vertical buses, pillars, conductive plugs, stacked contacts, or any other suitable vertical interconnect structures. Also contemplated are embodiments in which one set of lines is local to each layer and another set of lines is shared across layers, embodiments in which different layers use different line orientations, and embodiments in which layer selection is performed by one or more additional control lines, decoders, or selector devices.
[0109] Readout circuit 20, which is optionally and preferably shared among all the memory cells of array 30, is configured to transmit address signals so as to select a particular memory cell ofarray 30, and thereafter to perform the aforementioned readout operations in order to read, optionally and preferably non-destructively, the polarization state of the capacitor of the selected cell, hence also the memory state of this cell. For example, readout circuit 20 can apply, via one of the word lines 38, a gate voltage to one of access transistors 32 so as to establish conduction between respective channel terminals 36, 46 of the access transistor 32. Once the conduction is established, readout circuit 20 applies, via one of the bit lines 44, a pulse of bias voltage to the ferroelectric capacitor 12 connected to that access transistor. Circuit 20 can then measure the electric quantity and determine the memory state of the ferroelectric capacitor as further detailed hereinabove. For clarity of presentation, communication lines between readout circuit 20 and the word line(s) 38, bit lines(s) 44, and optionally vertical lines 48, are not illustrated in FIG. 22, but one of ordinary skills in the art, provided with the details described herein would know how to connect readout circuit 20 to these lines.
[0110] Circuit 20, as further detailed hereinabove, can be provided as a component in a memory readout system that is connectable to any memory cell or memory array operating on ferroelectric principles. Thus, the readout functionality described herein can be embodied as a standalone or partially standalone system, module, integrated circuit, instrument, controller, or subsystem configured to interface with one or more ferroelectric capacitors, ferroelectric memory cells, ferroelectric memory arrays, FeRAM devices, or other memory structures in which information is represented by ferroelectric polarization states. In such embodiments, the memory readout system can provide pulse generation, addressing, sensing, analog and / or digital processing, calibration, compensation, and state determination functionality as further detailed hereinabove.
[0111] Following is a description of potential applications which can benefit from memory cell 10 and / or circuit 20 and / or memory array 30 according to some embodiments of the present invention.
[0112] In some embodiments of the present invention memory cell 10 and / or circuit 20 and / or memory array 30 are incorporated in neuromorphic circuitry, for example, as non-volatile synaptic storage elements, weight- storage elements, state variables, local adaptation memory, event-history memory, inference-memory blocks, or memory associated with spiking neural networks, analog neural networks, mixed-signal neural networks, and in-memory or near-memory computing architectures. In some embodiments of the present invention memory cell 10 and / or circuit 20 and / or memory array 30 are incorporated in a display system, for example, to store image data, calibration data, gamma-correction data, display-driving parameters, local compensation values, frame data, persistence settings, user preferences, or control code in flat-panel displays,microdisplays, wearable displays, automotive displays, and augmented-reality or virtual-reality display modules. In some embodiments of the present invention memory cell 10 and / or circuit 20 and / or memory array 30 are incorporated in an imager system, for example, as embedded memory associated with image sensors, event cameras, machine-vision sensors, depth cameras, thermal imagers, biomedical imagers, or camera modules, optionally to store pixel data, calibration data, defect maps, timing parameters, local processing states, Al inference parameters, or imageprocessing instructions.
[0113] In some embodiments of the present invention memory cell 10 and / or circuit 20 and / or memory array 30 are incorporated in a digital signal processing system, for example, to store coefficients, filter parameters, transform parameters, temporary states, lookup tables, firmware, context data, training data, control variables, or captured data in systems performing audio processing, radar processing, wireless-baseband processing, video processing, compression, decompression, encryption, decryption, beamforming, speech processing, sensor fusion, or edge inference. In some embodiments of the present invention memory cell 10 and / or circuit 20 and / or memory array 30 are incorporated in a programmable logic system, such as a field-programmable gate array, a complex programmable logic device, a configurable accelerator, or a reconfigurable data path, optionally to store configuration bits, routing data, lookup-table contents, state-machine data, calibration data, or user-programmable parameters. In some embodiments of the present invention memory cell 10 and / or circuit 20 and / or memory array 30 are incorporated in a data storage system, including, without limitation, solid-state storage devices, storage-class memory devices, archival systems, high-speed buffers, write caches, metadata stores, transactional logs, and hybrid memory hierarchies that cooperate with DRAM, SRAM, flash memory, or magnetic storage.
[0114] In some embodiments of the present invention memory cell 10 and / or circuit 20 and / or memory array 30 are incorporated in a sensor system, for example, in systems associated with pressure sensors, temperature sensors, magnetic sensors, gas sensors, biosensors, inertial sensors, optical sensors, acoustic sensors, industrial sensors, environmental sensors, or multi-sensor platforms, optionally for storing sensed data, calibration constants, event histories, alarm thresholds, model parameters, or edge-processing states. In some embodiments of the present invention memory cell 10 and / or circuit 20 and / or memory array 30 are incorporated in a mobile computing system, including, without limitation, smartphones, tablets, wearable devices, laptops, portable Al systems, handheld terminals, asset trackers, and internet-of-things nodes, optionally where reduced standby power and non-volatility are beneficial. In some embodiments of thepresent invention memory cell 10 and / or circuit 20 and / or memory array 30 are incorporated in a robotic system, for example, in autonomous robots, collaborative robots, drones, industrial manipulators, service robots, medical robots, warehouse robots, and agricultural robots, optionally for local state retention, controller parameter storage, perception memory, motion-planning data, learned policies, or mission logs.
[0115] In some embodiments of the present invention memory cell 10 and / or circuit 20 and / or memory array 30 are incorporated in a smart home system, including, without limitation, smart locks, thermostats, lighting controllers, appliance controllers, security sensors, control hubs, voiceassistant devices, occupancy sensors, and energy-management nodes. In some embodiments of the present invention memory cell 10 and / or circuit 20 and / or memory array 30 are incorporated in a medical system, for example, in patient monitors, implantable or wearable medical devices, diagnostic devices, imaging equipment, infusion systems, laboratory instruments, neural interfaces, prosthetic controllers, or therapeutic devices, optionally where non-volatility, low-power operation, and high read endurance are advantageous. In some embodiments of the present invention memory cell 10 and / or circuit 20 and / or memory array 30 are incorporated in a network switch system, router, communication interface, modem, optical transceiver, packet processor, or data-center accelerator, optionally to store forwarding tables, control-plane data, packetprocessing state, telemetry data, encryption keys, queue states, classifier parameters, or fail-safe recovery data.
[0116] In some embodiments of the present invention memory cell 10 and / or circuit 20 and / or memory array 30 are incorporated in a barcode scanner system, including handheld scanners, fixed scanners, industrial readers, point-of-sale readers, vision-based code readers, and logistics terminals, optionally to store decoded data, local operating parameters, calibration constants, firmware states, or user profiles. In some embodiments of the present invention memory cell 10 and / or circuit 20 and / or memory array 30 are employed in a battery management system, for example, in electric vehicles, energy- storage systems, backup-power systems, consumerelectronics batteries, industrial battery packs, and charging infrastructure, optionally to store cellbalancing data, aging models, calibration values, safety thresholds, state-of-charge history, state-of-health data, fault logs, or control firmware parameters.
[0117] According to an aspect of some embodiments of the present invention there is provided a method of neuromorphic computing. The method comprises reading information from a ferroelectric capacitor as described herein, and performing at least one neuromorphic computation operation based on the information. Representative examples of neuromorphic computationoperations which can be performed by the method according to some embodiments of the invention include, without limitation, synaptic weight update, vector-matrix multiplication, accumulation of neuron input currents, thresholding / activation evaluation, and in-memory inference.
[0118] According to an aspect of some embodiments of the present invention there is provided a method of operating a display system. The method comprises reading information from a ferroelectric capacitor as described herein, and driving a display based on the information. Representative examples of operations for driving the display which can be performed by the method according to some embodiments of the invention include, without limitation, applying gamma correction, applying per-pixel or per-region compensation, adjusting brightness or backlight level, updating a frame buffer, and selecting a refresh mode.
[0119] According to an aspect of some embodiments of the present invention there is provided a method of operating an imager system. The method comprises reading information from a ferroelectric capacitor as described herein, and performing at least one image sensing and / or image processing operation based on the information. Representative examples of image sensing and / or image-processing operations which can be performed by the method according to some embodiments of the invention include, without limitation, pixel readout control, exposure control, defect-pixel correction, dark-frame or offset correction, and image denoising.
[0120] According to an aspect of some embodiments of the present invention there is provided a method of digital signal processing. The method comprises reading information from a ferroelectric capacitor as described herein, and performing at least one digital signal processing operation based on the information. Representative examples of digital signal processing operations which can be performed by the method according to some embodiments of the invention include, without limitation, finite-impulse-response filtering, fast Fourier transform processing, beamforming, compression, decompression, encryption, and decryption.
[0121] According to an aspect of some embodiments of the present invention there is provided a method of operating a programmable logic system. The method comprises reading information from a ferroelectric capacitor as described herein, and configuring and / or reconfiguring at least one programmable logic resource based on the information. Representative examples of configuration and / or reconfiguration operations which can be performed by the method according to some embodiments of the invention include, without limitation, loading configuration bits, configuring routing interconnects, programming lookup-table contents, initializing state-machine registers, and applying calibration parameters.According to an aspect of some embodiments of the present invention there is provided a method of operating a data storage system. The method comprises reading information from a ferroelectric capacitor as described herein, and performing at least one data storage operation based on the information. Representative examples of data storage operations which can be performed by the method according to some embodiments of the invention include, without limitation, reading a cache line, updating metadata, performing a journal or transactional-log update, maintaining wear or health statistics, and restoring state after a power event.
[0122] According to an aspect of some embodiments of the present invention there is provided a method of operating a sensor system. The method comprises reading information from a ferroelectric capacitor as described herein, and performing at least one operation that relates to sensing based on the information. Representative examples of operations which can be performed by this method according to some embodiments of the invention include, without limitation, applying calibration constants, threshold-based event detection, storing sampled sensor data, sensor fusion parameter update, and alarm or trigger generation.
[0123] According to an aspect of some embodiments of the present invention there is provided a method of operating a mobile computing system. The method comprises reading information from a ferroelectric capacitor as described herein, and performing at least one mobile computing operation based on the information. Representative examples of mobile computing operations which can be performed by the method according to some embodiments of the invention include, without limitation, storing and restoring device state, maintaining low-power retention of user settings, caching frequently used data, secure key retrieval, and power-mode control.
[0124] According to an aspect of some embodiments of the present invention there is provided a method of operating a robotic system. The method comprises reading information from a ferroelectric capacitor as described herein, and performing at least one robotic control operation based on the information. Representative examples of robotic control operations which can be performed by the method according to some embodiments of the invention include, without limitation, closed-loop actuator control, motion-planning state update, localization or mapping state retention, safety-state handling, and mission-log update.
[0125] According to an aspect of some embodiments of the present invention there is provided a method of operating a smart home system. The method comprises reading information from a ferroelectric capacitor as described herein, and performing at least one smart home control operation based on the information. Representative examples of smart home control operations which can be performed by the method according to some embodiments of the invention include,without limitation, applying access-control rules, updating schedules, storing user preferences, executing automation scenes, and event-trigger handling.
[0126] According to an aspect of some embodiments of the present invention there is provided a method of operating a medical system. The method comprises reading information from a ferroelectric capacitor as described herein, and performing at least one medical operation based on the information. Representative examples of medical operations which can be performed by the method according to some embodiments of the invention include, without limitation, storing parameters of a patient or medical device, maintaining calibration of a medical device, alarm-state retention, updating therapy parameters, updating stimulation parameters, and logging diagnostic measurements.
[0127] According to an aspect of some embodiments of the present invention there is provided a method of operating a network switch system. The method comprises reading information from a ferroelectric capacitor as described herein, and performing at least one packet processing and / or routing operation based on the information. Representative examples of packet processing and / or routing operations which can be performed by the method according to some embodiments of the invention include, without limitation, lookup of forwarding tables, updating routing entries, applying quality-of-service rules, maintaining flow state, and storing telemetry or counters.
[0128] According to an aspect of some embodiments of the present invention there is provided a method of operating a barcode scanner system. The method comprises reading information from a ferroelectric capacitor as described herein, and performing at least one barcode scanning operation based on the information. Representative examples of barcode scanning operations which can be performed by the method according to some embodiments of the invention include, without limitation, storing decoded barcode data, applying decoding parameters, maintaining calibration constants, storing user profiles, and managing scan-trigger state.
[0129] According to an aspect of some embodiments of the present invention there is provided a method of operating a battery management system. The method comprises reading information from a ferroelectric capacitor as described herein, and performing at least one battery management operation based on the information. Representative examples of battery management operations which can be performed by the method according to some embodiments of the invention include, without limitation, state-of-charge estimation, state-of-health estimation, cell-balancing control, protection-threshold enforcement, and fault or event logging.
[0130] As used herein the term “about” refers to ± 10 %The terms "comprises", "comprising", "includes", "including", “having” and their conjugates mean "including but not limited to".
[0131] The term “consisting of’ means “including and limited to”.
[0132] The term "consisting essentially of" means that the composition, method or structure may include additional ingredients, steps and / or parts, but only if the additional ingredients, steps and / or parts do not materially alter the basic and novel characteristics of the claimed composition, method or structure.
[0133] As used herein, the singular form "a", "an" and "the" include plural references unless the context clearly dictates otherwise. For example, the term "a compound" or "at least one compound" may include a plurality of compounds, including mixtures thereof.
[0134] Throughout this application, various embodiments of this invention may be presented in a range format. It should be understood that the description in range format is merely for convenience and brevity and should not be construed as an inflexible limitation on the scope of the invention. Accordingly, the description of a range should be considered to have specifically disclosed all the possible subranges as well as individual numerical values within that range. For example, description of a range such as from 1 to 6 should be considered to have specifically disclosed subranges such as from 1 to 3, from 1 to 4, from 1 to 5, from 2 to 4, from 2 to 6, from 3 to 6 etc., as well as individual numbers within that range, for example, 1, 2, 3, 4, 5, and 6. This applies regardless of the breadth of the range.
[0135] Whenever a numerical range is indicated herein, it is meant to include any cited numeral (fractional or integral) within the indicated range. The phrases “ranging / ranges between” a first indicate number and a second indicate number and “ranging / ranges from” a first indicate number “to” a second indicate number are used herein interchangeably and are meant to include the first and second indicated numbers and all the fractional and integral numerals therebetween.
[0136] It is appreciated that certain features of the invention, which are, for clarity, described in the context of separate embodiments, may also be provided in combination in a single embodiment. Conversely, various features of the invention, which are, for brevity, described in the context of a single embodiment, may also be provided separately or in any suitable subcombination or as suitable in any other described embodiment of the invention. Certain features described in the context of various embodiments are not to be considered essential features of those embodiments, unless the embodiment is inoperative without those elements.
[0137] Various embodiments and aspects of the present invention as delineated hereinabove and as claimed in the claims section below find experimental support in the following examples.EXAMPLES
[0138] Reference is now made to the following examples, which together with the above descriptions illustrate some embodiments of the invention in a non-limiting fashion.
[0139] Example 1
[0140] Ultrafast N on-Destructive Readout of EeRAM by Lo -Voltage Transient Current Conventional FeRAM readout methods are destructive, requiring polarization switching of the FE capacitor (FeCAP) and write-back, which reduces endurance, increases latency, and energy consumption. Known techniques for non-destructive readout (NDRO) relied on a capacitance memory window (MW), but such capacitance sensing typically requires relatively long sensing times (e.g., microsecond-scale or longer). These techniques also generally require an asymmetric FeCAP structure to produce a measurable state-dependent capacitance near zero bias, which can compromise retention (e.g., due to built-in field or imprint effects).
[0141] This Example presents an NDRO method utilizing ultrafast transient response which applies to both symmetric and asymmetric structures. This Example experimentally demonstrates sub-ns read operations without altering the polarization state, achieving >1013read cycles (limited by test time). This structure- agnostic method improves retention (tested at 125°C), endurance, and simplifies implementation, thus paving the way for fast, energy-efficient FeRAM -based solutions.
[0142] Introduction
[0143] Ferroelectric random-access memory (FeRAM) is a promising CMOS -compatible nonvolatile memory, offering high-speed, low power, high endurance, and good retention [1,2,3]. However, conventional read methods, which rely on current sensing during polarization switching, are destructive, requiring write-back operations that add complexity and reduce endurance, increasing energy consumption and latency [4,5], which currently limits the speed of the technology compared to dynamic random-access memory (DRAM).
[0144] Previous studies on NDRO for FeRAM have primarily relied on capacitance measurements, either through slow small-signal techniques or pulse-based current integration, the latter introducing additional computational complexity [4,5]. Both methods necessitate asymmetric FeCAP structures, which compromise retention due to built-in voltage effects, ultimately reducing device reliability. An alternative approach involves applying lower- voltage pulses to induce partial rather than full polarization switching during read operations. However,this method remains inherently destructive, as it is constrained by the finite number of ferroelectric domains available (which scale with FeCAP area). Consequently, for practical applications, it is limited to fewer than ~10k read cycles.
[0145] This Example describes a NDRO method and system based on fast transient measurements. This technique probes the current response at low-voltage using a high-speed pulse, and exhibits an MW due to the capacitance derivative with voltage. The technique offers a simple, non-destructive, ultrafast, low-power and low-energy readout, enables high endurance, and is compatible with both symmetric and asymmetric FeCAP structures, addressing retention limitations. The technique of the present embodiments was validated through experimental results, including: 1) sub-ns readout, 2) more than 1013cycles (limited only by test time), and 3) state retention at 125 °C. Furthermore, the Inventors demonstrated both unipolar and bipolar read schemes, with the latter preventing accumulative polarization changes in a single direction, potentially enhancing device reliability.
[0146] Non-Destructive Readout (NDRO)
[0147] According to some embodiments of the present invention an ultrashort, low-voltage pulse below the coercive voltage (V<Vc, which depends on time [6]) is applied to sense the polarization state optionally and preferably without altering it. The transient current in response to a low-bias voltage ramp in the FeCap can be phenomenologically understood from its quasi-static C-V curve (schematically shown in FIG. 1). While the capacitance itself shows no MW at V = 0 V in an ideally symmetric FeCap, the sign of its voltage derivative depends on the polarization state: positive after an ERS operation and negative after PRG. The expressions of simplified a chargecontrol approach according to some embodiments of the present invention can be written as:
[0148] dQ dV dC
[0149] Q = C - V - - = i = c - — + v ■ —
[0150] dt dt dt
[0151] dV dC dV dV ( dC\
[0152] — + V - - — C ± 7 - .
[0153]
[0154] dt dV dt dt \ dV /
[0155] The term V-dC / dV is proportional to half the MW, and so the transient current is proportional to the voltage derivative of the capacitance C. The NDRO technique of the present embodiments employs an ultrafast sub-Vc pulse, where the resulting current response reveals a MW due to the dependence of capacitance derivative on the FeCAP polarization state. A larger MW can be achieved according to some embodiments of the present invention by increasing the dC / dV term and applying a higher read voltage V, while ensuring that V<Vc to maintain non-destructive operation.RF-Compatible FeCAP and Measurement Setup
[0156] The Inventors experimentally demonstrated the NDRO technique of the present embodiments using a FeCap structure and measurement configuration shown in FIGs. 2A-C. The FeCAP stack (FIG. 2A) included 8 nm FE HZO with TiN top and bottom electrodes. TE and BE are related to the first metal layer in electrical contact with the FE layer. The metal layer affects the ferroelectricity. Different metals can be used according to some embodiments of the present invention. Additional metal(s) sandwich may ensure good contacts leading to probe pads and sufficiently thick metal. The FeCAP stack was designed and fabricated in RF-compatible structure with ground-signal-ground (GSG) transmission lines to allow high-speed sub-ns pulses. FeCAP dimensions were IV = L = 4 pm (total area 16 pm2), which resulted in RC delay lower than 100 ps. The C-V characteristics in FIG. 3A show C of about 600 fF ensuring compatibility with sub-ns pulse operation. Typical write-endurance measurements are displayed in FIG. 3B, demonstrating breakdown after about 4- 108write (switching) cycles, as evident by an increased leakage current, which limits the FeCAP performance.
[0157] Polarization curves measured by positive-up-negative-down (PUND) scheme at varying sweep rate are shown in FIG. 4, demonstrating an increase in coercive voltage (Vc) with higher frequency (shorter ramp time). This behavior is consistent with known ferroelectric domain switching kinetics and previous studies on the time dependence of Vc, such as Ref. [6]. Due to the sub-ns rise time, the effective Vc is relatively large, making Vread< Vc in this case (at tp< 1 ns). As a result, higher voltages can be applied with shorter pulses without exceeding Vc, thus providing a wider MW. Note that Vc as well as the RC delay also depend on the capacitor area, as discussed in [6] . A thinner HZO layer can allow lower voltage operation but increases the RC time constants, requiring a smaller FeCAP area for sub-ns switching, as shown in FIG. 5. The 5RC delay estimation indicates that fast switching is achievable with sufficiently small FeCAP dimensions, which reduces the displacement current. Although this, a priori, makes the readout more challenging, the peak current remains sufficiently high for accurate detection even at sub-ns timescales.
[0158] FIGs. 6A and 6B illustrate the measurement scheme and readout current following PRG and ERS operations, demonstrating that the peak current during an ±0.6 V, 800 ps pulse accurately indicates the FeCAP polarization state for both positive and negative read pulses. This behavior is observed during both voltage ramp-up (positive current) and ramp-down (negative current), confirming that the current difference is not caused by partial polarization switching, emphasizing its non-destructive nature.NDRO Experimental Demonstration
[0159] Unipolar NDRQ scheme
[0160] Unipolar read performance at room temperature is illustrated in FIGs. 7A-C. The first Ik read samples are displayed, followed by 1013cycles on a log scale, showing results of 1600 reads per decade. Peak current extraction reveals distinct levels corresponding to polarization states. Histograms of all read samples demonstrate clear state separation, while Min, Max, Avg, and Std values of the reads are plotted versus read cycle number. After about 5- 1012positive unipolar read cycles the ERS state switches due to accumulative stress at Vread=0.7 V. To prevent unintended switching, a lower read voltage can be used, typically with a narrower MW, since MW is proportional to 2V-dC / dV. Experimental demonstration is illustrated in FIG. 7C. Specifically, at Vread=0.6 V the MW remains narrower than at Vread=0.7 V, but no unintended switching occurs, even after 1013read cycles.
[0161] Bipolar NDRO scheme
[0162] To mitigate destructive disturbances due to the accumulative effect of unipolar pulses, a bipolar read scheme alternates positive and negative pulses, balancing stress in both directions and reducing the probability of undesired switching. FIGs. 8A-C and 9A-C show measured bipolar readouts for 1013cycles (positive and negative read bias, respectively). The MW (which is represented in FIGs. 8C and 9C by the difference between the transient current peaks) in the positive part (FIGs. 8A-C) of the bipolar read is narrower compared to the negative part (FIGs.
[0163] 9A-C) due to the asymmetry of the specific FeCap. This asymmetry arises from the capacitive MW [4,5], which is subtracted from the dC / dV term in the positive part and added in the negative part (opposite to the asymmetry shown in FIG. 3A). Note that the MW is preserved after 1013read cycles (limited only by measurement time), demonstrating that no partial switching is required for the read operation. A drift in polarization states is observed, attributed to retention effects, as the total measurement time for 1013cycles exceeds 20 hours. The NDRO of the present embodiments can be optimized by positive and / or negative readout (FIGs. 8A-C and 9A-C), and / or by adjusting the read amplitude (Vread) and pulse width (tp) of the positive and / or negative pulse separately, as demonstrated schematically with different parameters in FIGs. 6A-B.
[0164] High-Temperature (125°C ) Cycling and Retention
[0165] FIGs. 10A-C demonstrate 1013bipolar read cycles at 125°C. The degradation of the MW between about 1011and about 1013cycles follows the trend of retention, with the data overlaid against effective accumulated test time in FIGs. 11A-B. The accumulated test time of about 1013cycles exceeds 20 hours, with current shifts resembling state retention characteristics (FIGs. 11 A-B), confirming that the readout remains non-destructive even at about 1013read cycles. FIGs. 11A- B present retention measurements using the NDRO of the present embodiments at room temperature, revealing state drift, which degrades further at 125°C, consistent with the expected Tagantsev trend [7,8]. Note that the non-destructive nature of the technique enables retention studies with a single write event (per PRG or ERS operation) for the entire duration, providing a large number of sampling points over time and / or longer retention timescales without increasing the total measurement time. This is unlike standard retention tests, which require a write operation for each data point due to destructive readout. The MW is represented in FIGs. IOC and 11A by the difference between the transient current peaks.
[0166] Table 1, below compares key readout properties in this example with conventional NDRO [4,5] methods and conventional FeRAM read approaches [1,2].
[0167] Table 1
[0168] Comparison of Readout Schemes
[0169] Refs. [1,2] Ref. [4] Ref. [5]
[0170] This Example (Destructive) (Cap) (Cap-Pulsed)
[0171] Destructive Yes No No No
[0172] Read time >10 ns -100 ms >100 ns -1 ns FeCap structure Any Asymmetric Asymm. Any
[0173] Eimit retention No Yes Yes No Measured parameter Pol. Charge Cap. Cap. Trans, current Endurance Demo 1012Demo 2- 103Demo 1011Demo 1013Area (pm2) 0.028 4800 4800 16
[0174]
[0175] Summary
[0176] This Example described an NDRO technique for FeRAM by fast transient currents, achieved through sub-ns sub-Vc pulses. This technique enjoys the properties of ultrafast, low- energy, non-destructive readout by detecting distinct polarization states through transient current peaks, compatible with both symmetric and asymmetric FeCAPs. The technique described herein supports both unipolar and bipolar read schemes, enhancing the reliability and reducing write disturbances. Experimental results validate the approach, demonstrating robust operation at elevated temperatures (about 125°C), and high read endurance (more than 1013cycles) confirming state separation and long-term reliability. Energy consumption (about CV2) is reduced due to the low read voltage. The system of the present embodiments can be used in scalable FeRAM applications in non-volatile memory technology and neuromorphic hardware.Example 2
[0177] Method for Non-destructive Read of Polarization
[0178] This Example introduces a method for non-destructive readout of the polarization state stored in FeCAPs by leveraging the varying derivative of capacitance. This approach is versatile and can be applied to any FeCAP structure, including symmetric designs. The readout can be fast, can have high endurance and / or long retention, and can simplify state detection by identifying the peak current without requiring additional computational processing (such as integration).
[0179] In FeCAPs, the relationship between the charge (Q), the voltage (V), and the capacitance (C) can be described in its simplified form by Q=C V. In ferroelectric materials, the capacitance (C) is not constant with voltage (see FIGs. 12A-C), and in general the current I can be expressed as the time derivative of charge:
[0180] I=dQ / dt=C- dV / dt+V- dC / dt=(dV / dt)- (C+V- dC / dV).
[0181] The derivative dC / dV depends on the polarization state of the FeCAP, which can result in either a positive or negative contribution to the current. With reference to FIGs. 12A-C, the derivative of capacitance with voltage is positive after negative pulse, and negative after a positive pulse. Consequently, the total current becomes either higher or lower with respect to C- dV / dt based on the stored polarization state. This behavior is similar in both symmetric and asymmetric structures. Therefore, unlike reading the capacitance value which does not exhibit different capacitance values near 0 V bias in symmetric devices, the derivative can be used to read the memory window in any structure. Thus, the technique is effective even for symmetric FeCAPs where the capacitance C around bias of 0V is identical for both polarization states.
[0182] To determine the stored state, the peak current during the pulse is optionally and preferably measured. Since the peak current is proportional to the total current, the polarization state can be identified by comparing the peak current to a predetermined threshold that distinguishes between the two states (FIGs. 13A-B). This method enables efficient, reliable, and non-destructive readout of the stored data. The Inventors have successfully demonstrated experimentally (see Example 1) such a read operation with pulses having pulse width on a nanosecond scale, limited by measurement capability. This speed can be increased, for example, using on-chip read circuitry.
[0183] The applied voltage V is optionally and preferably kept lower than the coercive voltage of the ferroelectric layer Vc. This allows keeping the read pulse non-destructive. Vc itself depends on time: shorter pulses result in higher Vc compared to its steady state value, therefore short read pulses are preferred to preserve the ferroelectric polarization state (non-destructive operation), as short pulses reduce the chances of unintended polarization switch. Furthermore, short pulsesincrease the dV / dt term, thus enhancing the signal that needs to be read to determine the polarization state dV / dt- V-dC / dV.
[0184] The read operation of the present embodiments can be performed using either positive and / or negative pulses to measure their respective peak currents (see FIGs. 13A-B). Alternating between positive and negative pulses during the read operation (see example in FIG. 14) allows determination of the stored state and is advantageous as it reduces the likelihood of accumulation of unipolar pulses, which could otherwise lead to unintended switching. This alternating approach improves reliability and long-term stability of the device.
[0185] Example 3
[0186] Self-Referenced Readout
[0187] The read signal for PRG and ERS can be made farther apart, and the reference sensitivity can be further reduced by converting bipolar transient sensing into a self-referenced, chargedomain subtraction metric. In this Example, the bipolar readout conditions are the same as in Example 1, above.
[0188] The magnitude of the transient current generated by the positive-polarity pulse of the bipolar read waveform can be integrated to provide a value for QP, and the magnitude of the transient current generated by the negative-polarity pulse of the bipolar read waveform can be integrated to provide a value for QN. The integration scheme for the two memory states PRG (T) and ERS ('0') is shown in FIGs. 15A and 15B, respectively, and example results of such integrations are shown in FIGs. 16A and 17A for QP, and FIGs. 16B and 17B for QN. Note that since there is no polarization switching, the quantities QP and QN do not represent polarization charges. The Inventors found that QP is smaller than QN when the memory state is PRG (T) and larger than QN when the memory state is ERS ('0'). Thus, according to some embodiments of the present invention the difference AQ = QP - QN can be defined. As shown in FIGs. 18A and 18B, AQ is negative for PRG (FIG. 18 A) and positive for ERS (FIG. 18B), so that the sign of AQ provides a direct state decoding. Note that since QP and QN are extracted from two consecutive pulses within a single read operation, slow drift and other common-mode variations in transient response are suppressed in AQ, eliminating dependence on an external reference.
[0189] FIGs. 19A and 19B show AQ as the NDRO MW under cycling, where one NDRO cycle is defined as a readout pulse pair comprising a positive pulse and a negative pulse. The two AQ distributions remain fully separated through 5x 1012NDRO cycles. Min / max / avg / std are computed from 800 reads per decade (FIG. 19A). The histogram in FIG. 19B shows a mean MW of about 25 fC / pm2, with full separation between PRG and ERS readouts. This corresponds to about 1013individual read pulses, since Example 1 counted each positive or negative pulse as one read, whereas in this Example one NDRO cycle is a two-pulse readout (positive and negative). Thus, 5xl012cycles in this example correspond to about 1013pulse reads in Example 1.
[0190] Table 2, below, compares key readout properties in this example with methods disclosed in Refs. 1 and 5 and with the technique described in Example 1, above.
[0191] Table 2
[0192] Comparison of Readout Schemes
[0193] Ref. [1] Ref. [5] Example 1 This Example Destructive Yes No No No
[0194] Read time >10 ns >100 ns -1 ns -2 ns
[0195] Limit retention No Yes No No
[0196] Endurance Demo 1012Demo 10111013(uni); 5xl012
[0197] 5xl012(bi)
[0198] Area (pm2) 0.028 4800 16 16
[0199] MW (fC / pm2) -600 - 0.4 < 10 -25
[0200]
[0201] As demonstrated, the readout scheme which includes calculating AQ improves the MW relative to Example 1, while preserving nanosecond-class read time using a 16 pm2FeCAP. The RF-compatible design reduces parasitic RC delay (FIGs. 2A-C). Further area scaling can reduce RC and allow faster readout. The method of the present embodiments does not require stack asymmetry, avoiding associated retention / imprint constraints, and it also avoids the refresh limits of partial-switching NDRO by operating in an ultrafast sub-Vc, non-switching regime.
[0202] The value of MW can optionally and preferably be increased by accumulating the value of AQ. In these embodiments N consecutive bipolar read pulses are applied, where N is an integer larger than 1. For each bipolar read pulse, QP, QN, and AQ are calculated as further detailed hereinabove, and an effective difference AQeff is calculated by summing over the individual differences AQ calculated for each bipolar read pulse. The effective difference AQeff is thus proportional to N. Formally AQeii~aN, where a is a proportionality constant. This procedure provides a tradeoff between the speed and the accuracy. Larger values of N lengthen the readout time but provide higher value of AQeff and therefore larger MW and higher accuracy, while smaller values of N provide shorter read time at an expense of reduced accuracy. Accumulation of AQ is particularly useful when operating near minimal readout voltage or under higher noise or parasitic currents.The value of MW per read bipolar pulse can be increased by a judicious selection of the pulse parameters. In these embodiments, the waveform of the read bipolar pulse is selected to increase, preferably maximize, the value of |AQ|, and optionally and preferably also the signal-to-noise ratio. The pulse is preferably selected such that, throughout the pulse duration, the voltage Vread is below the coercive voltage Vc, so as to preserve the non-destructive nature of the read operation. Representative examples of pulse parameters that can be selected to increase |AQ|, include, without limitation, amplitude VP of the positive-polarity pulse, amplitude VN of the negative-polarity pulse, ratio between |VP| and |VN| (including, but not limited to, the case of |VP / VN|=1), the width of the positive-polarity pulse, the width of the negative-polarity pulse, the rise and / or fall times of the positive-polarity pulse, the rise and / or fall times of the negative-polarity pulse, the slew rate of the positive-polarity pulse, the slew rate of the negative-polarity pulse, the delay between the positive-polarity and the negative-polarity pulse (including, but not limited to, the case of zero delay), the shapes of the positive-polarity pulse and the negative-polarity pulse (e.g., rectangular shape, trapezoidal shape, triangular shape, multi-level shape, pulse with shaped edges), where the shape of each of the positive-polarity pulse the negative-polarity pulse can be selected independently, the duty cycle, the repetition rate, the number of bipolar pulses per read, bipolar symmetry or lack of bipolar symmetry, extent of bipolar asymmetry, use of precharge or bias point selection before the read pulse(s), and so on.
[0203] Although the invention has been described in conjunction with specific embodiments thereof, it is evident that many alternatives, modifications and variations will be apparent to those skilled in the art. Accordingly, it is intended to embrace all such alternatives, modifications and variations that fall within the spirit and broad scope of the appended claims.
[0204] It is the intent of the applicant(s) that all publications, patents and patent applications referred to in this specification are to be incorporated in their entirety by reference into the specification, as if each individual publication, patent or patent application was specifically and individually noted when referenced that it is to be incorporated herein by reference. In addition, citation or identification of any reference in this application shall not be construed as an admission that such reference is available as prior art to the present invention. To the extent that section headings are used, they should not be construed as necessarily limiting. In addition, any priority document(s) of this application is / are hereby incorporated herein by reference in its / their entirety.REFERENCES
[0205] [1] N. Ramaswamy et al., "NVDRAM: A 32Gb Dual Layer 3D Stacked Non-volatile Ferroelectric Memory with Near-DRAM Performance for Demanding Al Workloads," 2023 International Electron Devices Meeting (IEDM), San Francisco, CA, USA, 2023, pp. 1-4, doi: 10.1109 / IEDM45741.2023.10413848.
[0206] [2] J. Okuno et al., "A highly reliable 1.8 V 1 Mb Hfo.5Zro.s02-based 1T1C FeRAM Array with 3-D Capacitors," 2023 International Electron Devices Meeting (IEDM), San Francisco, CA, USA, 2023, pp. 1-4, doi: 10.1109 / IEDM45741.2023.10413661.
[0207] [3] M. M. Dahan, H. Mulaosmanovic, O. Levit, S. Dunkel, S. Beyer, and E. Yalon, “Sub-nanosecond switching of Si:HfO2 ferroelectric field-effect transistor,” Nano Lett., vol. 23, no. 4, pp. 1395— 1400, Feb. 2023, doi: 10.1021 / acs.nanolett.2c04706.
[0208] [4] S. Mukherjee et al., "Capacitive Memory Window With Non-Destructive Read in Ferroelectric Capacitors," in IEEE Electron Device Letters, vol. 44, no. 7, pp. 1092-1095, July 2023, doi: 10.1109 / LED.2023.3278599.
[0209] [5] S. Mukherjee et al., "Pulse-Based Capacitive Memory Window with High Non-Destructive Read Endurance in Fully BEOL Compatible Ferroelectric Capacitors," 2023 International Electron Devices Meeting (IEDM), San Francisco, CA, USA, 2023, pp. 1-4, doi: 10.1109 / IEDM45741.2023.10413879.
[0210] [6] W. D. Zhang and A. Q. Jiang, "Size-Scaling Effect on Domain Switching Time and Coercive Field of TiNZHfo.5Zro.5O2 / TiN Thin-Film Capacitors," in IEEE Transactions on Electron Devices, vol.
[0211] 70, no. 12, pp. 6324-6328, Dec. 2023, doi: 10.1109 / TED.2023.3325417.
[0212] [7] A. K. Tagantsev, I. Stolichnov, N. Setter, and J. S. Cross, “Nature of nonlinear imprint in ferroelectric films and long-term prediction of polarization loss in ferroelectric memories,” J. Appl. Phys., vol. 96, no. 11, pp. 6616-6623, Dec. 2004, doi: 10.1063 / 1.1805190.
[0213] [8] P. Vishnumurthy, R. Alcala, T. Mikolajick, U. Schroeder, L. A. Antunes and A. Kersch, "Ferroelectric HfCE-based Capacitors for FeRAM: Reliability from Field Cycling Endurance to Retention (invited)," 2024 IEEE International Reliability Physics Symposium (IRPS), Grapevine, TX, USA, 2024, pp. 1-10, doi: 10.1109 / IRPS48228.2024.10529299.
Claims
WHAT IS CLAIMED IS:
1. A memory cell, comprising:a ferroelectric capacitor configured to be polarizable into at least two distinct polarization states, each corresponding to a different non-volatile memory state; anda memory readout circuit configured to apply a pulse of bias voltage to said capacitor, to measure an electric quantity indicative of a sign of a derivative of a capacitance of said capacitor, and to determine a memory state of said capacitor based on said electric quantity.
2. The memory cell according to claim 1, wherein said derivative of a capacitance of said capacitor is with respect to said bias voltage.
3. The memory cell according to claim 1, wherein said derivative of a capacitance of said capacitor is with respect to time.
4. The memory cell according to claim 1, wherein said memory readout circuit is configured to ensure that said bias voltage is less than a coercive voltage of said capacitor throughout a duration of said pulse.
5. The memory cell according to claim 1, wherein said ferroelectric capacitor is symmetric.
6. The memory cell according to any of claims 2-4, wherein said ferroelectric capacitor is symmetric.
7. The memory cell according to claim 1, wherein said ferroelectric capacitor is asymmetric.
8. The memory cell according to any of claims 2-4, wherein said ferroelectric capacitor is asymmetric.
9. The memory cell according to claim 1, wherein said electric quantity comprises a difference between a peak current resulting from said bias voltage and a predetermined current threshold.
10. The memory cell according to any of claims 2-8, wherein said electric quantity comprises a difference between a peak current resulting from said bias voltage and a predetermined current threshold.
11. The memory cell according to claim 9, wherein said predetermined current threshold is between a first peak current characteristic of a first polarization state of said polarization states and a second peak current characteristic of a second polarization state polarization states.
12. The memory cell according to claim 10, wherein said predetermined current threshold is between a first peak current characteristic of a first polarization state of said polarization states and a second peak current characteristic of a second polarization state polarization states.
13. The memory cell according to claim 9, wherein said predetermined current threshold is within a predetermined tolerance from a multiplication of a value of said capacitance during said pulse by a time-derivative of said bias voltage during said pulse.
14. The memory cell according to claim 10, wherein said predetermined current threshold is within a predetermined tolerance from a multiplication of a value of said capacitance during said pulse by a time-derivative of said bias voltage during said pulse.
15. The memory cell according to claim 1, wherein said electric quantity comprises a difference between an electrical charge resulting from the bias voltage and a predetermined electrical charge threshold.
16. The memory cell according to any of claims 2-8, wherein said electric quantity comprises a difference between an electrical charge resulting from the bias voltage and a predetermined electrical charge threshold.
17. The memory cell according to claim 1, wherein said electric quantity comprises a difference between an electrical charge or a peak current resulting from said bias voltage and an electrical charge or a peak current characteristic of a reference memory cell being at a predetermined memory state.
18. The memory cell according to any of claims 2-8, wherein said electric quantity comprises a difference between an electrical charge or a peak current resulting from said bias voltage and an electrical charge or a peak current characteristic of a reference memory cell being at a predetermined memory state.
19. The memory cell according to claim 1, wherein said pulse is a bipolar pulse.
20. The memory cell according to any of claims 2-18, wherein said pulse is a bipolar pulse.
21. The memory cell according to claim 19, wherein said electric quantity comprises a combination between an electrical charge or charge density resulting from a positive-polarity pulse of said bipolar pulse, and an electrical charge or charge density resulting from a negative-polarity pulse of said bipolar pulse.
22. The memory cell according to claim 20, wherein said electric quantity comprises a combination between an electrical charge or charge density resulting from a positive-polarity pulse of said bipolar pulse, and an electrical charge or charge density resulting from a negative-polarity pulse of said bipolar pulse.
23. The memory cell according to claim 21, wherein said combination is a linear combination.
24. The memory cell according to claim 22, wherein said combination is a linear combination.
25. The memory cell according to claim 21, wherein said combination is a difference.
26. The memory cell according to claim 22, wherein said combination is a difference.
27. The memory cell according to any of claims 1-26, wherein said capacitor has a ferroelectric layer made of a material selected from the group consisting of hafnium zirconate, lead zirconate titanate, barium titanate, strontium bismuth tantalate, bismuth ferrite, sodium bismuth titanate, lithium niobate, polyvinylidene fluoride, calcium titanate, zinc oxide, potassium niobate.
28. The memory cell according to any of claims 1-27, wherein said capacitor has a ferroelectric layer made of a ferroelectric heterostructure.
29. A memory array, comprising a plurality of memory cells at addressable locations, and a respective plurality of selector devices arranged to individually access memory cells responsively to address signals, wherein at least one of said memory cells is the memory cell according to any of claims 1-28.
30. A memory array, comprising:a plurality of ferroelectric capacitors, configured to be polarizable into at least two distinct polarization states, each corresponding to a different non-volatile memory state;a respective plurality of access transistors associated with said ferroelectric capacitors, wherein a terminal of each ferroelectric capacitor is connected to a first channel terminal of a respective access transistor;a word line connected to a gate terminal of each access transistor;a bit line connected to a second channel terminal of each access transistor; anda readout circuit configured to apply, via said word line, a gate voltage to one of said access transistors so as to establish conduction between respective channel terminals of said access transistor, to apply, via said bit line, a pulse of bias voltage to a respective ferroelectric capacitor, to measure an electric quantity indicative of a sign of a derivative of a capacitance of said ferroelectric capacitor, and to determine a memory state of said ferroelectric capacitor based on said electric quantity.
31. The array of claim 30, being a two-dimensional array.
32. The array of claim 31 , wherein said plurality of ferroelectric capacitors are arranged in rows and columns.
33. The array of claim 32, wherein said word line is connected to the gate terminal of each access transistor in a row, and said bit line is connected to a second channel terminal of each access transistor in a column.
34. The array of claim 30, being a three-dimensional array.
35. The array of claim 34, wherein said plurality of ferroelectric capacitors are arranged in layers, wherein each layer comprises ferroelectric capacitors arranged in rows and columns.
36. The array of claim 35, wherein said word line is connected to the gate terminal of each access transistor in a row, and said bit line is connected to a second channel terminal of each access transistor in a column, and wherein the array comprises vertical lines connecting channel terminals of different layers.
37. The memory cell or memory array according to any of claims 1-36, serving as a component in a system selected from the group consisting of neuromorphic circuitry, a display system, an imager system, a digital signal processing system, a programmable logic system, a data storage system, a sensor system, a mobile computing system, a robotic system, a smart home system, a medical system, a network switch system, a barcode scanner system, and a battery management system.
38. A memory readout system, comprising a memory readout circuit configured to apply a pulse of bias voltage to a ferroelectric capacitor, to measure an electric quantity indicative of a sign of a derivative of a capacitance of said capacitor, and to determine a memory state of said capacitor based on said electric quantity.
39. The system according to claim 38, wherein said memory readout circuit is configured to ensure that said bias voltage is less than a coercive voltage of said capacitor throughout a duration of said pulse.
40. A method of reading information from a ferroelectric capacitor configured to be polarizable into at least two distinct polarization states, each corresponding to a different nonvolatile memory state, the method comprising:applying a pulse of bias voltage to the ferroelectric capacitor;measuring an electric quantity indicative of a sign of a derivative of a capacitance of the capacitor; anddetermining a memory state of the capacitor based on said electric quantity.
41. A method of neuromorphic computing, comprising executing the method of claim 40 for reading information from a ferroelectric capacitor, and performing at least one neuromorphic computation operation based on said information.
42. A method of operating a display system, comprising executing the method of claim 40 for reading information from a ferroelectric capacitor, and controlling at least one displaydriving operation based on said information.
43. A method of operating an imager system, comprising executing the method of claim 40 for reading information from a ferroelectric capacitor, and performing at least one image sensing and / or image-processing operation based on said information.
44. A method of digital signal processing, comprising executing the method of claim 40 for reading information from a ferroelectric capacitor, and performing at least one digital signal processing operation based on said information.
45. A method of operating a programmable logic system, comprising executing the method of claim 40 for reading information from a ferroelectric capacitor, and configuring and / or reconfiguring at least one programmable logic resource based on said information.
46. A method of operating a data storage system, comprising executing the method of claim 40 for reading information from a ferroelectric capacitor, and performing at least one data storage operation based on said information.
47. A method of operating a sensor system, comprising executing the method of claim 40 for reading information from a ferroelectric capacitor, and performing at least one sensor-related operation based on said information.
48. A method of operating a mobile computing system, comprising executing the method of claim 40 for reading information from a ferroelectric capacitor, and performing at least one mobile computing operation based on said information.
49. A method of operating a robotic system, comprising executing the method of claim 40 for reading information from a ferroelectric capacitor, and performing at least one robotic control operation based on said information.
50. A method of operating a smart home system, comprising executing the method of claim 40 for reading information from a ferroelectric capacitor, and performing at least one smart home control operation based on said information.
51. A method of operating a medical system, comprising executing the method of claim 40 for reading information from a ferroelectric capacitor, and performing at least one medicaldevice operation based on said information.
52. A method of operating a network switch system, comprising executing the method of claim 40 for reading information from a ferroelectric capacitor, and performing at least one packet-processing and / or routing operation based on said information.
53. A method of operating a barcode scanner system, comprising executing the method of claim 40 for reading information from a ferroelectric capacitor, and performing at least one barcode scanning operation based on said information.
54. A method of operating a battery management system, comprising executing the method of claim 40 for reading information from a ferroelectric capacitor, and performing at least one battery management operation based on said information.
