Non-volatile polarization induced strain coupled 2D FET memory

Description

CROSS-REFERENCE TO RELATED APPLICATIONS

None.

STATEMENT REGARDING GOVERNMENT FUNDING

None.

TECHNICAL FIELD

The present disclosure generally relates to a non-volatile memory, and in particular, to a non-volatile memory based on polarization induced strain in a ferroelectric material.

BACKGROUND

This section introduces aspects that may help facilitate a better understanding of the disclosure. Accordingly, these statements are to be read in this light and are not to be understood as admissions about what is or is not prior art.

A non-volatile memory (NVM) is a type of memory that maintains its value when power has been removed. In the post-Moore law era, many investigators have been working on alternative approaches to NVM. Chief amongst these alternatives are Heat Assisted Magnetic Recording (HAMR), spin-transfer torque RAM (STT-RAM), Phase-change memory (PCM), Resistive RAM (RRAM), ferroelectric RAM (FERAM or FRAM), and ferroelectric transistor (FEFET) memory. Each of these have approaches have advantages and disadvantages. For example, HAMR suffers from low performance. FERAM and FEFET based memories show distinct advantages due to electric field (E) driven low-power writes. However, there are issues in ferroelectric (FE) based NVM such as destructive read in FERAMs. FEFET based NVMs without inter-layer metal show concerns such as traps and depolarization fields and FEFETs with floating inter-layer metal exhibit gate leakage.

Several investigators have addressed the issue of destructive read in FRAMs. For example, U.S. Pub. 2019-0363247 for Zemen et al. describes an NVM with a strain inducing layer which induces a strain in a ferromagnetic layer which can maintain a magnetization direction and read as a value held by the cell.

However, additional improvements are needed to improve readability of the cell without flipping the cell value during the read operations.

Therefore, there is an unmet need for a novel approach in ferroelectric-based memory that addresses destructive reads as well as provide a high level of fidelity when reading the cell value.

SUMMARY

A polarization induced strain coupled two dimensional field effect transistor (PoSt FET) memory cell is disclosed. The PoSt FET memory cell includes a transistor. The transistor includes a source contact, a drain contact, a gate contact, a back contact, a channel disposed atop the gate contact. The channel and the gate are separated by an electrically insulating material. The PoSt FET memory cell also includes a piezoelectric (PE)/ferroelectric(FE) (PE/FE) layer disposed between the gate contact and the back contact and configured to store bit information in form of ferroelectric polarization (P). A ratio of cross-sectional area of the channel to cross-sectional area of the PE/FE layer is between about 0.03 to about 0.07.

A method of maintain a digital bit value in a polarization induced strain coupled two dimensional field effect transistor (PoSt FET) memory cell is also disclosed. The method includes providing a transistor configured to maintain a bit value. The transistor includes a source contact, a drain contact, a gate contact, and a back contact. The transistor also includes a channel disposed atop the gate contact, wherein the channel and the gate are separated by an electrically insulating material. Furthermore, the transistor includes a piezoelectric (PE)/ferroelectric(FE) (PE/FE) layer disposed between the gate contact and the back contact and configured to store bit information in form of ferroelectric polarization (P), wherein a ratio of cross-sectional area of the channel to cross-sectional area of the PE/FE layer is between about 0.03 to about 0.07, and wherein the bit value is maintained based on polarization of the PE/FE layer, whereby the polarization induced strain in the PE/FE layer is transferred to the channel.

A memory array is also disclosed. The memory array includes a plurality of polarization induced strain coupled two dimensional field effect transistor (PoSt FET) memory cells, disposed in one or more rows and one or more columns. Each PoSt FET memory cell includes a transistor. The transistor includes a source contact, a drain contact, a gate contact, a back contact, a channel disposed atop the gate contact. The channel and the gate are separated by an electrically insulating material. The PoSt FET memory cell also includes a piezoelectric (PE)/ferroelectric(FE) (PE/FE) layer disposed between the gate contact and the back contact and configured to store bit information in form of ferroelectric polarization (P). A ratio of cross-sectional area of the channel to cross-sectional area of the PE/FE layer is between about 0.03 to about 0.07. The gate contact of each of the PoSt FET memory cells in each row of the one or more rows is coupled to an associated word line (WL) for said row. The back contact of each of the PoSt FET memory cells in each column of the one or more columns is coupled to an associated write bit line (WBL) for said column. The drain contact of each of the PoSt FET memory cells in each column of the one or more columns is coupled to an associated read bit line (RBL) for said column. A PoSt FET memory cell in a row and a column is written to by activating an associated WBL and WL and a PoSt FET memory cell in a row and a column is read from by activating an associated RBL and WL.

BRIEF DESCRIPTION OF DRAWINGS

FIG. 1a is a cross-section of a polarization induced Strain coupled 2D FET (PoSt FET) memory cell, according to the present disclosure as well as a symbolic representation of PoSt FET memory cell.

FIG. 1b is a schematic representation of a write operation.

FIG. 1c is a schematic representation of a read operation.

FIG. 1d is a schematic representation of the PoSt FET memory cell of FIG. 1a coupled to a sense amplifier.

FIG. 2a is a graph of polarization in C/m² vs. electric field in kV/cm which shows a hysteresis curve for switching polarization from P+ to P−.

FIG. 2b is a graph of strain (scaled as 10⁻³) vs. Electric field in kV/cm.

FIG. 3a is a simulation platform used to analyze the PoSt FET memory cell of FIG. 1.

FIG. 3b is a graph of I_DS vs. V_GS for a κ of 0.04.

FIG. 4 is a PoSt FET Memory-based array.

FIG. 5 is a graph of σ_TMD in GPa and ΔE_G in meV vs. W_PE in nm.

FIG. 6 is a graph of Area (in F² where F is minimum feature size of a technology) due to increased strain as κ decreases.

FIG. 7 is a graph of I_LRS/I_HRS vs. κ.

DETAILED DESCRIPTION

For the purposes of promoting an understanding of the principles of the present disclosure, reference will now be made to the embodiments illustrated in the drawings, and specific language will be used to describe the same. It will nevertheless be understood that no limitation of the scope of this disclosure is thereby intended.

In the present disclosure, the term “about” can allow for a degree of variability in a value or range, for example, within 10%, within 5%, or within 1% of a stated value or of a stated limit of a range.

In the present disclosure, the term “substantially” can allow for a degree of variability in a value or range, for example, within 90%, within 95%, or within 99% of a stated value or of a stated limit of a range.

A novel approach in ferroelectric-based memory is presented in the present disclosure that addresses the aforementioned issues of the prior art cells such as destructive reads as well as provide a high level of fidelity when reading the cell value. Towards this end, a novel low power non-volatile memory utilizing FE polarization-based bit storage/switching and piezoelectricity induced dynamic bandgap modulation for bit sensing is presented. While retaining the low power electric-field driven write attributes of FE based non-volatile memory, the presented memory provides low power non-destructive read solution. A polarization induced Strain coupled 2D FET (PoSt FET) memory cell 100 is provided in the present disclosure, as presented in FIG. 1a which is a cross-sectional view of the cell. Referring to FIG. 1a, the PoSt FET memory cell 100 provides the following features: (a) FE polarization-based bit-storage, (b) E driven write, and (c) coupling of piezoelectricity of FE with dynamic bandgap (EG) tuning of Transition Metal Dichalcogenides (TMDs) for read. The PoSt FET cell 100 is a 4-terminal device including metal contacts for source 102 (S), drain 104 (D), gate 110 (G), and back 114 (B). The PoSt FET memory cell 100 further includes a channel 106 disposed atop the gate 110. The channel 106 of the PoSt FET memory cell 100 is based on a two-dimensional Transition Metal Dichalcogenides (2D TMD) (selected from the group consisting of MoS₂, MoSe₂, WS₂, WSe₂) controlled by gate 110 which is disposed on a dielectric 108. A piezoelectric (PE)/ferroelectric (FE) material 112 (such as PZT, silicon doped HfO₂ and combinations thereof HfO₂) disposed between the gate 110 the back 114 stores the bit information in the form of ferroelectric polarization (P), with +(−) P representing logic ‘1’ (‘0’). Combinations thereof TMD and PE from the above groups may be used in the device. To write (i.e., to switch P), a voltage across PE (V_GB) is applied such that the electric field (|E|) is greater than the coercive field (E_C), where V_C (i.e., the voltage across the PE/FE material 112) is E_C times the thickness of the PE/FE (or simply PE or simply FE) material 112. The polarity of V_GB determines the logic (P+ or P−) state written in the device. Referring to 1b, a schematic of write operation is shown by way of a change in polarization in the PE/FE material 112. When the voltage across the gate and back (i.e., the voltage across the PE/FE material which is V_GB) is greater than V_C, the polarization of the PE/FE material is positive; however, when the voltage across the gate and back (i.e., V_GB) is less than −V_C, the polarization of the PE/FE material is negative.

Alternatively, referring to FIG. 1c, a schematic of read operation is shown by way of sensing drain to source current, I_DS. For the read operation two conditions must occur: 1) a gate to back voltage (i.e., V_GB) must be applied which is less than V_C, thereby avoiding flipping the stored value; and 2) a V_GS voltage must be applied that is higher than the transistor's threshold voltage in order to turn on the transistor to thereby establish I_DS which can be measured and compared to a threshold to determine the stored bit value as described below. To read the stored bit value, strain induced in the PE/FE material 112 is ascertained and related to the bit value stored (i.e., written) in the PE/FE material 112. The induced strain in the PE/FE material 112 is transferred to the 2D TMD material of the channel 106. To sense the stored bit, polarization-induced strain (S) in PE/FE material 112 is used which transduces as out-of-plane stress in the 2D TMD material of the channel 106, leading to dynamic modulation in the bandgap (ΔE_G). This, in turn, tunes the drain current (I_DS) leading to low/high resistance states (LRS/HRS) of PoSt FET memory cell 100. For P+, S>0 and ΔE_G<0 which yields high I_DS=I_LRS. Similarly, P− yields S<0, ΔE_G>0 and low I_DS=I_HRS. To efficiently transduce strain from PE to TMD, we utilize the hammer and nail effect, wherein the area of TMD above the gate (A_TMD) acts as the nail while PE serves as the hammer. For that, A_TMD is designed to be smaller than the cross-sectional area of PE (A_PE).

Referring to FIG. 2a, a graph of polarization in C/m² vs. electric field in kV/cm is provided which shows a hysteresis curve for switching polarization from P+ to P−. A width of the hysteresis curve across polarization at about 0, quantified by the coercive field/voltage (E_C/V_C), is desirable so that there can be sufficient margins for non-destructive reads. Piezoelectric material PZT-5H with E_C of about 9.5 kV/cm satisfies this condition. Moreover, the magnitude of polarization determines the strain in PE required for distinguishing bit value during a read. FIG. 2a of PZT-5H is based on remanent polarization (P_R) of about 0.30 C/m², and E_C of about 9.5 kV/cm. Higher polarization provides larger strain.

Referring to FIG. 2b, a graph of strain (scaled as 10⁻³) vs. Electric field in kV/cm is also shown. The slope of this plot at about strain=0 (quantified by the parameter piezoelectric coefficient, d₃₃) determines the distinguishability of P+ and P− states during read. Higher the slope, better distinguishability of P+ and P− is achieved. FIG. 2b of PZT-5H shows a d₃₃ of about 650 picometers(pm)/V. Piezoelectric materials with large piezoelectric effect (d₃₃) are desirable. Simultaneously, it should also show good ferroelectric properties, viz., high remnant polarization (P_R) and high coercive field (E_C). Both of these requirements are fulfilled by PZT-5H. Although PMN-PT has impressive piezoelectric effect (d₃₃=850 pm/V), it is not a suitable material for PoSt FET due to its low coercive field (E_C=4 KV/cm).

The mathematical relationship associated with the dynamic modulation in the bandgap (ΔE_G) is provided below which is also used in modeling the induced PE polarization. To analyze the described PoSt FET, we developed a simulation framework shown in FIG. 3a. We model the P-E response of PZT-5H using Landau-Khalatnikov (LK) equation and calibrate it with experiment as shown in FIG. 2a. We utilize the S-E response from the same experiments to extract piezoelectric coefficients (d₃₃ and d₃₁) of PZT-5H. These parameters are used to model pressure transduced to TMD (σ_TMD) using COMSOL MULTIPHYSICS SUITE. We simulate the full 3D structure of the presented PoSt FET (including the hammer and nail effect) in COMSOL and employ the strain-charge form of the constitutive equations for PE with proper hard boundary conditions to obtain σ_PE and σ_TMD. σ_TMD is converted to ΔE_G and self-consistently coupled with the TMD charge/potential model.

E = α ⁢ ⁢ P + β ⁢ P 2 + γ ⁢ P 5 + ρ ⁢ dP dt ( 1 )
where α, β, γ are static coefficients of ferroelectric;
ρ is the kinetic coefficient.
Equation 1 is the Landau Khalatnikov (LK) equation used for modeling polarization.
ΔE_G=α_TMDσ_TMD  (2)
where α_TMD is pressure coefficient, and
δ_TMD is stress in 2D TMD channel modeled in COMSOL Multiphysics Suite
I_DS=f(E₀+ΔE_G,V_GS,V_DS)  (3)
Where E₀ is the intrinsic bandgap which is 1.5 eV,
ΔE_G is bandgap modulation,
V_GS is gate to source voltage, and
V_DS is drain to source voltage.

A comparison between a typical 2D FET and the PoSt FET of the present disclosure shows a higher current for I_LRS and a lower current for I_HRS as shown in FIG. 3b which includes a graph of I_DS vs. V_GS and a zoomed-in version of the graph for a κ of 0.04. It should be noted that κ depends on device geometry. In particular, κ is the ratio of the cross-sectional area of the ferroelectric-piezoelectric material to the 2D TMD material (i.e., L_2DW_2D/L_PEW_PE, where 2D refers to the 2D TMD material). Results show a range of acceptable κ. The minimum acceptable κ is between about 0.03 to about 0.07 and that is based on distinguishability and sense margin (>1 uA) shown in Table 1. At the minimum allowable area of PE which is 9*3λ² (κ=0.22) a distinguishability of 1.6 and sense margin <1 uA, are achieved which are undesirable.

In order to model PoSt FET, P is coupled with S equations of PE/FE with strain induced dynamic bandgap modulation (ΔE_G) of TMD as provided in equations 1-3. The P-E and S-E plots are then calibrated which are obtained from the model as provided in FIGS. 2a and 2b with experiments performed for PZT. S induced ΔE_G as provided in equations 1-4 depends on parameters of the PE and 2D materials viz. [a] 2D out-of-plane compliance parameter (C₃₃); [b] coefficient of bandgap change in 2D (a); and [c] piezo compliance (S₃₃). ΔE_G also depends on stress reduction factor (κ) which is extracted based on aspect ratio of 2D and PE. Examples of these parameters are provided in Table 2 for demonstration purposes only, thus no limitation is intended hereby.

The results from COMSOL simulations (shown in FIG. 3a) show that the hammer and nail effect cause σ_TMD to be boosted compared to σ_PE as shown in FIG. 3a, when the area of nail/TMD (A_TMD) is lower than that of PE (A_PE), the hammer. The device parameter κ=A_TMD/A_PE is a measure of this effect, where smaller κ is expected to provide larger σ_TMD. We observe about 12× increase in σ_TMD compared to σ_PE at κ=0.04 (see Table 1). Optimization of the width of PE (W_PE) can enable tuning of κ (see FIG. 3a), and hence σ_TMD and ΔE_G. By increasing W_PE from 90 nm to 180 nm, κ decreases from 0.07 to 0.03, leading to 1.78× increase in σ_TMD and ΔE_G.

To understand the unique read mechanism, the transfer (I_DS−V_GS) and output (I_DS31 V_DS) characteristics PoSt FET are presented. It should be noted that a gate voltage (V_G) lower than coercive voltage (V_C=0.6V) of PZT is applied to ensure that the stored P state is undisturbed in this analysis. We apply gate voltage (V_G) which is lower than coercive voltage (V_C=0.6V) of PZT-5H to ensure that the stored P state is undisturbed in this analysis. For P+, PS FET shows 2.3× higher I_DS (I_LRS) than standard 2D TMD FET while for P−, I_DS is 3.4× lower (I_HRS) due to strain driven ∓E_G. Based on the I_DS−V_GS characteristics, we identify that 0.3V<V_GS<0.4V provides optimal I_LRS/I_HRS, sufficient current necessary for read operation and ample read disturb margin (V_C−V_GS˜200 mV). We choose V_GS=V_READ=0.35V that gives I_LRS/I_HRS of about 8×. Note that I_LRS/I_HRS can be improved by material optimizations and device optimization (e.g. by reducing κ).

Utilizing the PoSt FET memory cell presented above, a PoSt FET Memory-based array is presented in FIG. 4. Based on the unique read mechanism of PoSt FET, a decoupled read-write access-transistor-less memory array is also presented herein. Referring to FIG. 4, gates of PoSt FET memory cells are coupled to a word line (WL), while the back terminal and drain are coupled to the write bit-line (WBL) and read bit-line (RBL), respectively. The write ports are connected in a cross-point fashion (WL shared along the row and WBL shared along the column). A V/2 biasing for write is utilized as provided in Table 3 provided below.

From Table 3, it is seen that WBL and WL of the accessed cell is such that |V_GB|=V_DD>V_C appears across PE, resulting in P switching. For example, to write P+, we apply V_DD/2 on WL and −V_DD/2 on WBL, resulting in V_GB=V_DD. Similarly, to write P−, we apply −V_DD/2 on WL and V_DD/2 on WBL for which V_GB=−V_DD. By applying V_DD/2 biasing, half-accessed cells get |V_DD/2|<V_C across them due to either WL (in half-accessed column) or WBL (in half-accessed row). |V_DD/2|<V_C in these cells prevents polarization switching. Moreover, in this biasing scheme WBL/WL of un-accessed cells can be kept at 0V. This minimizes write energy expended by these cells that is otherwise observed with assertion of WLs/BLs with conventional V_DD swing. Note that we achieve non-volatile memory operation without any access transistors (unlike other memory cells used in arrays of the prior art). For read, a V_READ is applied on WL and V_DD on RBL of the accessed cell. The read current is sensed on RBL (shared along the column). No current flows through the half-accessed/un-accessed cells as they have either their gate voltage (WL) or drain voltage (RBL) or both set to 0.

PoSt FET shows a compact cell area in the range of 27F²−47.25F² (where F is the feature size) for 0.07>κ>0.03 as shown in FIG. 5 FIG. 6 is a graph of σ_TMD in GPa and ΔE_G in meV vs. W_PE in nm. Distinguishability I_LRS/I_HRS increases to about 12× at κ=0.03 compared to 3× at κ=0.07. This is because bandgap change |ΔE_G| (for both P+ and P−) increases as shown in FIG. 6 due to increased strain as κ decreases, according to equation 1-3, above. Decreasing κ further lowers resistance of PeFET in P+(higher I_LRS) but increases resistance for P− (low I_HRS) which ultimately results in improved distinguishability. A larger κ (=0.07) offers higher integration density, lower write energy and lower read power, albeit at the cost of lower I_LRS/I_HRS, as shown in FIG. 7 which is a graph of I_LRS/I_HRS vs. κ.

Those having ordinary skill in the art will recognize that numerous modifications can be made to the specific implementations described above. The implementations should not be limited to the particular limitations described. Other implementations may be possible.