betaFeedback

WorksPhysics and Astronomy

Electrostatic and Quantum Transport Engineering of Multilayer Ferroelectric Tunnel Junctions for Improved TER and RA Trade off

Abstract

Ferroelectric tunnel junctions (FTJs) are promising for non-volatile memory and neuromorphic computing, but achieving a high tunnel electroresistance ratio (TER) with a low resistance-area (RA) product remains challenging. In this work, TiN/Hf0.5_{0.5}Zr0.5_{0.5}O2_2 (HZO)-based FTJs containing zero to three TiO2_2 and IGZO-based quantum wells (QWs) are analyzed using the Miller ferroelectric (FE) polarization model integrated with the non-equilibrium Green's function (NEGF) formalism to examine the effects of QW material and thickness on polarization and transport. TiO2_2-based structures exhibit higher polarization and TER than IGZO-based structures. At a read bias of 0.1 V, the uniform TiO2_2 double-QW structure, with 1-nm-thick QW and insulating layer (IL) layers, gives a TER of 2.71×1010%2.71\times10^{10}\% and an RA product of 5×104 Ω⋅cm25\times10^4~Ω\cdot\text{cm}^2, compared with 5.95×107%5.95\times10^7\% and 2×105 Ω⋅cm22\times10^5~Ω\cdot\text{cm}^2, respectively, for IGZO. Both uniform and non-uniform TiO2_2 double-QW structures exhibit resonant tunneling (RT) and negative differential resistance (NDR) associated with QW-confined states. Increasing the first QW thickness to 1.38 nm, while keeping the second QW and both IL layers at 1 nm, shifts the confined-state energies and modifies resonant transport. The selected non-uniform structure achieves a TER of 2.95×1011%2.95\times10^{11}\% and an RA product of 1.25×103 Ω⋅cm21.25\times10^3~Ω\cdot\text{cm}^2 at a read bias of 0.1 V. Elastic scattering changes the amplitude and width of QW resonant-transmission peaks, with a stronger effect on the low resistance state (LRS) than on the high resistance state (HRS). These results show that QW material and thickness optimization can improve the TER–RA trade-off, while electrode screening and elastic scattering further influence polarization-dependent transport in multilayer FTJs.

Introduction

Ferroelectric tunnel junctions (FTJs) offer potential for non-volatile memory (NVM) and neuromorphic computing because their electrical resistance can be controlled through the polarization of a ferroelectric (FE) barrier. Reversing the FE polarization changes the effective tunneling barrier, switching the device between a LRS and a HRS [1, 2]. The difference in current density between these two states is expressed by the TER, defined as

TER⁡(%)=[JLRSJHRS−1]×100,(1)\operatorname{TER}(\%) = \left[\frac{J_{LRS}}{J_{HRS}} - 1\right] \times100, \tag*{(1)}

where JLRSJ_{LRS} and JHRSJ_{HRS} are the current densities in the LRS and HRS, respectively [3]. HfO2_2-based FTJs have received considerable attention because of their CMOS compatibility, scalability, and potential for low-energy operation [4, 5, 6]. However, achieving a high TER together with a low LRS RA product remains challenging, particularly for thicker FE barriers, where increased tunneling resistance limits the read current [7]. Introducing QWs into an asymmetric tunneling barriers provides resonant transport paths when their confined energy levels align with the transport energy window [8, 9]. The resulting transmission in the two polarization states depends on the QW material properties, particularly dielectric permittivity, electron affinity, and effective mass, as well as QW thickness, thereby influencing the TER and RA characteristics.

TiO2_2 and In-Ga-Zn-O (IGZO) have both previously been integrated into hafnia-based FE structures. TiO2_2 has been used as a dielectric interlayer in HZO-based FTJs, where its insertion improves FE properties and TER [10]. It has also been used as an interlayer to enhance ferroelectricity in HZO films [11]. IGZO has also been integrated with HZO in ferroelectric devices [12]. In our previous work, we investigated an asymmetric HZO-based FTJ incorporating a single TiO2_2 QW using a Preisach-based FE polarization model coupled with NEGF method [13]. A single QW FTJ incorporating a Ta2_2O5_5 well was also investigated using the transfer-matrix method (TMM) [14]. Multiple QW hafnia-based FTJs were later analyzed using the Miller polarization model and TMM, assuming ideal electrode screening [15]. Apart from QW material selection, layer thickness also affects the confined-state energies and resonant transport. The IL thickness determines the wave-function penetration and coupling between neighboring QWs, whereas the QW thickness controls the confinement energies and positions of the confined states. Previous studies of superlattice structures have reported changes in transmission with non-uniform layer thicknesses [16, 17]. Such changes can modify resonant transport and NDR [18]. However, a systematic investigation of the influence of QW material properties and thickness configurations on resonant transport and the TER–RA trade-off remains limited in multilayer FTJs. The combined effects of finite electrode screening and elastic scattering on these characteristics also require further investigation. In this work, multilayer HfO2_2-based FTJs containing zero to three QWs are analyzed using the Miller phenomenological FE polarization model integrated with NEGF formalism. TiO2_2 and IGZO-based QWs are first compared to examine how their dielectric and electronic properties affect polarization and the TER-RA characteristics. The TiO2_2-based double-QW structure is then examined in detail. Its uniform configuration is used to investigate RT, polarization-dependent transport, and NDR. The effects of finite electrode screening on the TER-RA characteristics and elastic scattering on resonant transmission are also evaluated. Finally, the QW thicknesses are varied while keeping both IL thicknesses equal to examine how non-uniform QW thicknesses affect confined-state alignment, electron transport, and NDR. These results provide physical insight into how QW material and thickness, electrode screening, and elastic scattering affect polarization-dependent transport, and offer a theoretical basis for optimizing the TER-RA trade-off in multilayer FTJs.

Device Modeling

The device configurations and their corresponding energy-band profiles are shown in Figure 1. All structures consist of TiN top and bottom electrodes (TE and BE) and an HZO FE layer. An IL is retained in the 0-QW structure, whereas the other configurations contain 1, 2, or 3 QW–IL pairs. The general layer sequence is written as M/FE/(QW/IL)n_n/M, where n=1,2,3n=1,2,3 represents the number of QW–IL pairs. The thicknesses and relative permittivities of the FE layer, ii-th QW, and corresponding IL are denoted by (tfe,εfe)(t_{\mathrm{fe}},\varepsilon_{\mathrm{fe}}), (tqw,i,εqw,i)(t_{\mathrm{qw},i},\varepsilon_{\mathrm{qw},i}), and (til,i,εil,i)(t_{\mathrm{il},i},\varepsilon_{\mathrm{il},i}), respectively. The electrostatic potential within the metal electrodes is described by the Thomas–Fermi screening model as follows: [2], [19], [20]:

Device schematics and corresponding energy-band profiles of TiN/HZO-based FTJs with different QW configurations

Figure 1. Device schematics and corresponding energy-band profiles of TiN/HZO-based FTJs with different QW configurations: (a) 0-QW (FE/IL bilayer), (b) 1-QW, (c) 2-QW, and (d) 3-QW.

Vm1(m2)(x)={−σsλm1ε0εm1ex/λm1,x≤0,σsλm2ε0εm2e−(x−L)/λm2,x≥L.(2)V_{\mathrm{m1(m2)}}(x)= \begin{cases} -\dfrac{\sigma_s\lambda_{\mathrm{m1}}}{\varepsilon_0\varepsilon_{\mathrm{m1}}}e^{x/\lambda_{\mathrm{m1}}}, & x\leq0,\\ \dfrac{\sigma_s\lambda_{\mathrm{m2}}}{\varepsilon_0\varepsilon_{\mathrm{m2}}}e^{-(x-L)/\lambda_{\mathrm{m2}}}, & x\geq L. \end{cases} \tag*{(2)}

Here, σs\sigma_s is the screening charge density, and λm1(m2)\lambda_{\mathrm{m1(m2)}} and εm1(m2)\varepsilon_{\mathrm{m1(m2)}} denote the screening lengths and effective relative permittivities of the TE and BE, respectively. The applied potential and built-in potential satisfy the following relation:

Va+Vbi=Vm1(0)−Vm2(L)+Vfe+∑i=1n(Vqw,i+Vil,i).(3)V_a+V_{\mathrm{bi}}=V_{\mathrm{m1}}(0)-V_{\mathrm{m2}}(L)+V_{\mathrm{fe}}+\sum_{i=1}^{n}(V_{\mathrm{qw},i}+V_{\mathrm{il},i}). \tag*{(3)}

The built-in potential is given as:

Vbi=ϕm2−ϕm1−ϕ3+ϕ4.(4)V_{\mathrm{bi}}=\phi_{\mathrm{m2}}-\phi_{\mathrm{m1}}-\phi_3+\phi_4. \tag*{(4)}

where ϕm1(m2)\phi_{\mathrm{m1(m2)}} denote the electrode-interface barrier contributions, while ϕ3\phi_3 and ϕ4\phi_4 represent the internal-interface barrier contributions. Applying Gauss’s law at the metal/FE interface and assuming continuity of the electric displacement field across the internal FE/QW and QW/IL interfaces give [21]

σs=Pfe+ε0εfeEfe=ε0εqw,iEqw,i=ε0εil,iEil,i.(5)\sigma_s=P_{\mathrm{fe}}+\varepsilon_0\varepsilon_{\mathrm{fe}}E_{\mathrm{fe}}=\varepsilon_0\varepsilon_{\mathrm{qw},i}E_{\mathrm{qw},i}=\varepsilon_0\varepsilon_{\mathrm{il},i}E_{\mathrm{il},i}. \tag*{(5)}

The corresponding electric fields in the QW and IL layers are therefore

Eqw,i=σsε0εqw,i,Eil,i=σsε0εil,i.(6)E_{\mathrm{qw},i}=\frac{\sigma_s}{\varepsilon_0\varepsilon_{\mathrm{qw},i}},\qquad E_{\mathrm{il},i}=\frac{\sigma_s}{\varepsilon_0\varepsilon_{\mathrm{il},i}}. \tag*{(6)}

The potential drop across the non-FE layers and electrode screening regions is

VS=σsε0[∑i=1n(tqw,iεqw,i+til,iεil,i)+λm1εm1+λm2εm2].(7)V_S=\frac{\sigma_s}{\varepsilon_0}\left[\sum_{i=1}^{n}\left(\frac{t_{\mathrm{qw},i}}{\varepsilon_{\mathrm{qw},i}}+\frac{t_{\mathrm{il},i}}{\varepsilon_{\mathrm{il},i}}\right)+\frac{\lambda_{\mathrm{m1}}}{\varepsilon_{\mathrm{m1}}}+\frac{\lambda_{\mathrm{m2}}}{\varepsilon_{\mathrm{m2}}}\right]. \tag*{(7)}

For convenience, the combined electrical-thickness contribution of these layers and the electrode screening regions is denoted by SS:

S=∑i=1n(tqw,iεqw,i+til,iεil,i)+λm1εm1+λm2εm2.(8)S=\sum_{i=1}^{n}\left(\frac{t_{\mathrm{qw},i}}{\varepsilon_{\mathrm{qw},i}}+\frac{t_{\mathrm{il},i}}{\varepsilon_{\mathrm{il},i}}\right)+\frac{\lambda_{\mathrm{m1}}}{\varepsilon_{\mathrm{m1}}}+\frac{\lambda_{\mathrm{m2}}}{\varepsilon_{\mathrm{m2}}}. \tag*{(8)}

Thus, VS=σsS/ε0V_S=\sigma_sS/\varepsilon_0. Substituting the displacement-field relation and Vfe=EfetfeV_{\mathrm{fe}}=E_{\mathrm{fe}}t_{\mathrm{fe}} into the voltage balance gives

Va+Vbi=Efetfe+(Pfeε0+εfeEfe)S.(9)V_a+V_{\mathrm{bi}}=E_{\mathrm{fe}}t_{\mathrm{fe}}+\left(\frac{P_{\mathrm{fe}}}{\varepsilon_0}+\varepsilon_{\mathrm{fe}}E_{\mathrm{fe}}\right)S. \tag*{(9)}

Solving for the voltage drop across the FE layer yields

Vfe=tfe(Va+Vbi−PfeSε0)tfe+εfeS.(10)V_{\mathrm{fe}}=\frac{t_{\mathrm{fe}}\left(V_a+V_{\mathrm{bi}}-\dfrac{P_{\mathrm{fe}}S}{\varepsilon_0}\right)}{t_{\mathrm{fe}}+\varepsilon_{\mathrm{fe}}S}. \tag*{(10)}

The FE polarization is described using the phenomenological Miller model. Its voltage-dependent evolution is given by [22]

∂Pfe∂Vfe=Γm∂Psat∂Vfe,(11)\frac{\partial P_{\mathrm{fe}}}{\partial V_{\mathrm{fe}}}=\Gamma_m\frac{\partial P_{\mathrm{sat}}}{\partial V_{\mathrm{fe}}}, \tag*{(11)}

where PfeP_{\mathrm{fe}} is the polarization along the applied-voltage trajectory, PsatP_{\mathrm{sat}} is the saturated-branch polarization, and Γm\Gamma_m is the switching factor, defined as

Γm=1−tanh⁡[(Pfe−Psat±Ps−Pfe)1/2].(12)\Gamma_m=1-\tanh\left[\left(\frac{P_{\mathrm{fe}}-P_{\mathrm{sat}}}{\pm P_s-P_{\mathrm{fe}}}\right)^{1/2}\right]. \tag*{(12)}
HZO FE [15]TiN electrode
ParameterValueParameterValue
tFEt_{\mathrm{FE}}5 nmΦm\Phi_{\mathrm{m}}4.3 eV [26]
EcE_{\mathrm{c}}1.2 MV/cmme∗/mom_{\mathrm{e}}^{*}/m_{\mathrm{o}}15 [27]
PrP_{\mathrm{r}}12 μ\muC/cm2^{2}εr\varepsilon_{\mathrm{r}}3 [28]
PsP_{\mathrm{s}}13 μ\muC/cm2^{2}λm\lambda_{\mathrm{m}}0.5 Å [14]
MaterialRoleme∗/mom_{\mathrm{e}}^{*}/m_{\mathrm{o}}χ\chi (eV)εr\varepsilon_{\mathrm{r}}
HZOFE0.142.8025
HfO2_{2}IL0.142.8025
TiO2_{2}QW0.50 [29]4.00 [30]48 [31]
IGZOQW0.34 [32]4.16 [33]10 [34]

Table 1. FE AND MATERIAL PARAMETERS USED IN THE FTJ SIMULATIONS.

tfet_{\mathrm{fe}}: FE thickness; EcE_{\mathrm{c}}: coercive field; me∗m_{\mathrm{e}}^{*}: electron effective mass; mom_{\mathrm{o}}: free-electron mass; χ\chi: electron affinity; εr\varepsilon_{\mathrm{r}}: relative permittivity; Φm\Phi_{\mathrm{m}}: metal work function; λm\lambda_{\mathrm{m}}: metal screening length.

where PsP_{\mathrm{s}} is the saturation polarization. The saturated polarization is given by

Psat=Pstanh⁡[Vfe∓Vc2Vcln⁡(Ps+PrPs−Pr)],(13)P_{\mathrm{sat}}=P_{\mathrm{s}}\tanh\left[\frac{V_{\mathrm{fe}}\mp V_{\mathrm{c}}}{2V_{\mathrm{c}}}\ln\left(\frac{P_{\mathrm{s}}+P_{\mathrm{r}}}{P_{\mathrm{s}}-P_{\mathrm{r}}}\right)\right], \tag*{(13)}

where PrP_{\mathrm{r}} is the remanent polarization and VcV_{\mathrm{c}} is the coercive voltage. The upper and lower signs correspond to increasing and decreasing voltage sweeps, respectively. The polarization is obtained by numerically integrating Eq. (11) along the applied-voltage waveform. The current density is evaluated using the Landauer formalism [23, 24, 25]:

J=−2eAh∑ky,kz∫−∞+∞Tm(E)[fm1(E)−fm2(E)] dE,(14)J=-\frac{2e}{Ah}\sum_{k_{y},k_{z}}\int_{-\infty}^{+\infty}T_{\mathrm{m}}(E)\left[f_{\mathrm{m1}}(E)-f_{\mathrm{m2}}(E)\right]\,dE, \tag*{(14)}

where ee is the elementary charge, AA is the cross-sectional area, and hh is Planck’s constant. Here, Tm(E)T_{\mathrm{m}}(E) denotes the transmission probability, while fm1(E)f_{\mathrm{m1}}(E) and fm2(E)f_{\mathrm{m2}}(E) are the Fermi–Dirac distributions of the TE and BE, respectively. Within the NEGF formalism, the total retarded and in-scattering self-energies are written as

Σr=Σm1r+Σm2r+Σor,(15)\Sigma^{\mathrm{r}}=\Sigma_{\mathrm{m1}}^{\mathrm{r}}+\Sigma_{\mathrm{m2}}^{\mathrm{r}}+\Sigma_{\mathrm{o}}^{\mathrm{r}}, \tag*{(15)}
Σin=Γm1fm1+Γm2fm2+Σoin,(16)\Sigma^{\mathrm{in}}=\Gamma_{\mathrm{m1}}f_{\mathrm{m1}}+\Gamma_{\mathrm{m2}}f_{\mathrm{m2}}+\Sigma_{\mathrm{o}}^{\mathrm{in}}, \tag*{(16)}

where Σm1(m2)r\Sigma_{\mathrm{m1(m2)}}^{\mathrm{r}} are the retarded contact self-energies and Γm1(m2)\Gamma_{\mathrm{m1(m2)}} are the corresponding contact broadening matrices. Elastic scattering is incorporated through the retarded and in-scattering self-energies

Σor=D∘Gr,Σoin=D∘Gn,(17)\Sigma_{\mathrm{o}}^{\mathrm{r}}=D\circ G^{\mathrm{r}},\qquad\Sigma_{\mathrm{o}}^{\mathrm{in}}=D\circ G^{\mathrm{n}}, \tag*{(17)}

where ∘\circ represents element-wise multiplication, DD is the scattering-correlation matrix, GrG^{\mathrm{r}} and GnG^{\mathrm{n}} denote the retarded and electron correlation function, respectively. For local scattering, the matrix elements are given by

Dpq=δpqD0,(18)D_{pq}=\delta_{pq}D_{0}, \tag*{(18)}

where D0D_{0} controls the dephasing strength with pp and qq denoting local basis indices. In the computational framework, the Miller model is integrated with the NEGF formalism. The FE polarization is evolved under the applied voltage using the fourth-order

Runge–Kutta (RK4) method. At each RK4 stage, the polarization and applied voltage are used to calculate the screening charge density σs\sigma_{\mathrm{s}} and the voltage drop across the FE layer through the electrostatic relations in Eqs. (2)–(10). The resulting FE voltage is then used to evaluate the polarization derivative using the Miller model [Eqs. (11)–(13)]. The updated polarization and screening charge density determine the electric-field and electrostatic-potential profiles across the device layers. These profiles, along with the material-dependent band offsets, are used in the NEGF formalism to calculate the transmission and current density. The current density is subsequently obtained using the Landauer formalism [Eq. (14)]. Forward and reverse voltage sweeps provide the PfeP_{\mathrm{fe}}–VaV_{\mathrm{a}} hysteresis and JJ–VaV_{\mathrm{a}} characteristics for different QW–IL configurations. The effect of elastic scattering is also evaluated for the TiO2_{2}-based 2-QW FTJ using the scattering formulation in Eqs. (15)–(18).

The simulation parameters are listed in Table I. The framework is validated by comparing the calculated JJ–VaV_{\mathrm{a}} and PfeP_{\mathrm{fe}}–VaV_{\mathrm{a}} characteristics with the reference results reported in Ref. [15] for a double-QW FTJ with a TiN/HZO/(Ta2_{2}O5_{5}/HfO2_{2})2_{2}/TiN stack, as shown in Figure 2. The simulated polarization loop reproduces the overall hysteretic behavior of the reference data, while the corresponding current characteristics show good agreement, supporting the consistency of the polarization, electrostatic, and quantum-transport calculations.

Comparison of simulated current-voltage and polarization-voltage characteristics

Figure 2. Comparison of the simulated JJ-VaV_{\mathrm{a}} and PfeP_{\mathrm{fe}}-VaV_{\mathrm{a}} characteristics with the reference data reported in Ref. [15] for a double-QW FTJ with a TiN/HZO/(Ta2_{2}O5_{5}/HfO2_{2})2_{2}/TiN stack.

Results and Discussion

Effect of QW Number and Material

To investigate the influence of QW number and material on electron transport, multilayer FTJs incorporating TiO2_{2} and IGZO-based QWs are compared. Finite screening effects at both the TE and BE are considered, with a screening length of λm=0.5\lambda_{\mathrm{m}}=0.5 Å. The corresponding JJ-VaV_{\mathrm{a}} characteristics under a triangular voltage waveform are shown in Figure 3(a) and Figure 3(b), respectively. Both materials exhibit changes in current density with QW number, reflecting modifications in the RT pathways introduced by the QW–IL layers. To examine the changes in transport with QW number, the corresponding zero-bias transmission spectra are shown in Figure 3(c) and Figure 3(d). These spectra exhibit resonant peaks associated with QW-confined states. The energy separation, ΔE\Delta E, quantifies the difference between the energy positions of the first resonant peaks in the LRS and HRS. For one, two, and three QWs, ΔE\Delta E decreases from 0.59 to 0.50 and 0.42 eV for TiO₂, and from 0.48 to 0.33 and 0.22 eV for IGZO. TiO₂-based FTJs exhibit a larger peak separation than IGZO-based structures for each QW configuration, indicating a greater spectral distinction between the two resistance states. The RA products are shown in Figure 3(e) and Figure 3(f) for TiO₂ and IGZO-based FTJs, respectively. At a read bias of 0.1 V, the TiO₂-based 2-QW FTJ exhibits an RA product of 5×1045 \times10^{4} Ω⋅cm2\Omega\cdot\mathrm{cm}^{2}, compared with 2×1052 \times10^{5} Ω⋅cm2\Omega\cdot\mathrm{cm}^{2} for the IGZO-based structure. The corresponding TER characteristics are presented in Figure 3(g) and Figure 3(h), respectively. At the same read bias, the TER is 2.71×1010%2.71 \times10^{10}\% for TiO₂ and 5.95×107%5.95 \times10^{7}\% for IGZO. Thus, the TiO₂-based double-QW structure provides both a lower RA product and a higher TER than its IGZO counterpart. This difference is related to the dielectric and electronic properties

Transport characteristics of multilayer FTJs incorporating TiO₂-based and IGZO-based QWs

Figure 3. Effect of QW number on the transport characteristics of multilayer FTJs incorporating TiO₂-based (left panels) and IGZO-based (right panels) QWs. (a), (b) JJ-VaV_{\mathrm{a}} characteristics; (c), (d) transmission spectra; (e), (f) RA products; and (g), (h) TER.

of the QW materials. Although TiO₂ has a slightly lower electron affinity (4.00 eV) than IGZO (4.16 eV), its dielectric permittivity is considerably higher. The higher permittivity results in a lower electric field within the QW and modifies the voltage distribution and electrostatic potential profile across the multilayer stack. Meanwhile, differences in electron affinity and effective mass influence the band alignment and quantum-confinement energies, respectively. These changes affect the confined-state energies and their alignment in the LRS and HRS, thereby modifying resonant transmission and current modulation.

The polarization hysteresis loops in Figure 4(a) and Figure 4(b) compare TiO₂ and IGZO-based FTJs, respectively, with 0–3 QWs. TiO₂-based structures retain a larger remanent polarization as compared to IGZO-based structures, while increasing the QW number reduces the polarization in both materials. This trend originates from the increased effective electrical thickness of the non-FE layers, SS, which modifies the voltage distribution [Eqs. (8) and (9)] and reduces the voltage available across the FE layer [Eq. (10)]. The corresponding polarization versus FE-layer voltage is presented in Figure 4(c) and Figure 4(d). The non-zero polarization at Vfe=0V_{\mathrm{fe}}=0 reflects the persistence of polarization under the internal depolarizing field. At zero voltage, the magnitudes of the depolarizing field and voltage are expressed as

Polarization characteristics of multilayer FTJs incorporating TiO₂-based and IGZO-based QWs

Figure 4. Polarization characteristics of multilayer FTJs incorporating TiO₂-based (left panels) and IGZO-based (right panels) QWs with different QW numbers. (a), (b) PfeP_{\mathrm{fe}}-VaV_{\mathrm{a}} hysteresis loops; (c), (d) PfeP_{\mathrm{fe}}-VfeV_{\mathrm{fe}} characteristics.

∣Edep,0∣=∣Pfe∣Sε0(tfe+εfeS)(19)\left|E_{\mathrm{dep},0}\right| = \frac{\left|P_{\mathrm{fe}}\right|S}{\varepsilon_{0}(t_{\mathrm{fe}}+\varepsilon_{\mathrm{fe}}S)} \tag*{(19)}
∣Vdep,0∣=∣Pfe∣tfeε0εfeSS+tfe/εfe.(20)\left|V_{\mathrm{dep},0}\right| = \frac{\left|P_{\mathrm{fe}}\right|t_{\mathrm{fe}}}{\varepsilon_{0}\varepsilon_{\mathrm{fe}}}\frac{S}{S+t_{\mathrm{fe}}/\varepsilon_{\mathrm{fe}}}. \tag*{(20)}

The depolarizing field depends on both the electrical thickness SS and the FE polarization PfeP_{\mathrm{fe}}. Increasing the QW number increases the electrical thickness, which tends to enhance the depolarizing field. At the same time, the reduction in polarization tends to lower its magnitude. Thus, the net depolarizing field is determined by the combined variation of SS and PFeP_{\mathrm{Fe}}. The material dependence arises from the dielectric contribution of the QWs. For equal physical thickness, tQW/εTiO2<tQW/εIGZOt_{\mathrm{QW}}/\varepsilon_{\mathrm{TiO_2}} < t_{\mathrm{QW}}/\varepsilon_{\mathrm{IGZO}}, since εTiO2=48\varepsilon_{\mathrm{TiO_2}}=48 and εIGZO=10\varepsilon_{\mathrm{IGZO}}=10. For the same polarization and remaining device parameters, the smaller electrical-thickness contribution of TiO2_2 results in a lower depolarizing field and voltage, thereby favoring stronger polarization retention.

Figure 5 shows the QW-number and thickness dependence of normalized polarization (PFe/PrP_{\mathrm{Fe}}/P_{\mathrm{r}}) and depolarizing field (Edep/EcE_{\mathrm{dep}}/E_{\mathrm{c}}) at zero applied voltage for TiO2_2 and IGZO-based FTJs. For TiO2_2-based FTJs, Figure 5(a) shows that the normalized polarization decreases with increasing QW number, while a slight reduction is observed with increasing uniform QW thickness (tQW1=tQW2=tQWt_{\mathrm{QW1}}=t_{\mathrm{QW2}}=t_{\mathrm{QW}}). In comparison, IGZO-based FTJs exhibit a slightly greater reduction in normalized polarization with increasing QW number and thickness, as shown in Figure 5(b). The corresponding normalized depolarizing-field magnitude in Figure 5(c) increases with QW number but remains almost unchanged with increasing QW thickness. For IGZO-based FTJs, the normalized depolarizing-field magnitude also increases with QW number and exhibits slightly higher values than those of TiO2_2-based structures, as shown in Figure 5(d). This material-dependent behavior is associated with the dielectric permittivities of the QW materials. Since εIGZO=10\varepsilon_{\mathrm{IGZO}}=10 is lower than εTiO2=48\varepsilon_{\mathrm{TiO_2}}=48, IGZO contributes more electrical thickness for the same physical QW thickness, resulting in a higher depolarizing field. In contrast, the smaller electrical-thickness contribution of TiO2_2 favors better polarization retention.

Four plots showing QW thickness dependence of normalized polarization and depolarizing field for TiO$_2$- and IGZO-based QWs

Figure 5. QW-thickness dependence of normalized polarization and depolarizing field at zero applied voltage for FTJs incorporating TiO2_2-based (left panels) and IGZO-based (right panels) QWs. (a), (b) Normalized polarization, PFe/PrP_{\mathrm{Fe}}/P_{\mathrm{r}}; (c), (d) normalized depolarizing field, Edep/EcE_{\mathrm{dep}}/E_{\mathrm{c}}. Results are shown for 0–3 QWs, with LRS and HRS distinguished by different line styles.

The applied-voltage dependent polarization hysteresis for different QW configurations is compared in Figure 6(a) and Figure 6(b) for TiO2_2 and IGZO-based FTJs, respectively.

Applied-voltage-dependent polarization hysteresis for TiO$_2$- and IGZO-based FTJs with 0–3 quantum wells

Figure 6. Applied-voltage-dependent polarization hysteresis of multilayer FTJs incorporating (a) TiO2_2-based (top row) and (b) IGZO-based (bottom row) QWs. The columns from left to right correspond to 0–3 QWs.

For both materials, increasing the QW number extends the polarization-switching region over a wider voltage range, indicating that higher applied voltages are required to reverse the polarization. For a given QW configuration, increasing the maximum applied-voltage amplitude drives the polarization closer to saturation. However, incomplete polarization switching is more evident in IGZO-based structures with higher QW numbers, where saturation is not reached even at the maximum applied-voltage amplitude. In comparison, TiO2_2-based structures approach polarization saturation at comparatively lower voltage amplitudes. These differences are associated with the larger effective electrical-thickness contribution of IGZO, which reduces the voltage available across the FE layer and consequently limits polarization switching. Considering the combined polarization and transport characteristics discussed above, the TiO2_2-based 2-QW FTJ is selected for further investigation of its transport mechanisms and structural optimization.

TiO2_2-based 2-QW FTJ

Uniform QW and IL Layers

The transport characteristics of the TiO2_2-based 2-QW FTJ with 1-nm-thick QWs and ILs are shown in Figure 7. The JJ-VaV_{\mathrm{a}} characteristics in Figure 7(a) show distinct current densities in the LRS and HRS. The LRS current also exhibits NDR, which is associated with RT through QW-confined states. As shown by the bias-dependent transmission spectra in Figure 7(b), resonant transmission occurs at Va=0.6V_{\mathrm{a}}=0.6 V when the confined states align with the electrode transport-energy window. This alignment enhances electron transmission and increases the current. With further increase in bias, the confined states shift away from the transport-energy window, reducing the resonant contribution to the current and resulting in NDR. The zero-bias transmission spectra in Figure 7(c) further reveal different transmission probabilities for the LRS and HRS, resulting from polarization-dependent changes in the tunneling barrier. The corresponding RA and TER characteristics are shown in Figure 7(d). At a read bias of

Transport characteristics of the TiO$_2$-based 2-QW FTJ

Figure 7. Transport characteristics of the TiO2_2-based 2-QW FTJ with 1 nm thick QWs and ILs: (a) JJ-VaV_a characteristics; (b) bias-dependent transmission spectra showing resonant transmission at Va=0.6V_a=0.6 V; (c) zero-bias transmission spectra; and (d) bias-dependent TER and RA product.

Vr=0.1V_r=0.1 V, the uniform double-QW structure exhibits a TER of 2.71×1010%2.71 \times10^{10}\% and an RA product of 5×104 Ω⋅cm25 \times10^{4}\ \Omega\cdot\mathrm{cm}^{2}.

The influence of electrode screening on the transport and electrostatic properties of the TiO2_2-based 2-QW FTJ is examined in Figure 8. As shown in Figure 8(a), increasing the screening length λm\lambda_m at Vr=0.1V_r=0.1 V reduces the LRS-RA product while increasing the HRS-RA, resulting in a higher TER. The corresponding energy-band profiles in Figure 8(b) illustrate the changes in the potential distribution and QW energy levels with screening length. The effect of the normalized screening parameter, R=λm/ϵmR=\lambda_m/\epsilon_m, is shown in Figure 8(c). With increasing RR, the LRS-RA increases, whereas the HRS-RA initially decreases and then increases slightly at larger RR. Accordingly, the TER decreases over most of the investigated range. The corresponding band profiles in Figure 8(d) show the associated changes in the potential profile and QW confinement. These results demonstrate that electrode screening modifies the electrostatic potential profile and, consequently, the tunneling characteristics and current modulation between the resistance states.

Effect of electrode screening on the transport characteristics and energy-band profiles

Figure 8. Effect of electrode screening on the transport characteristics and energy-band profiles of the TiO2_2-based 2-QW FTJ. (a), (b) RA products and TER as functions of the screening length λm\lambda_m and the corresponding energy-band profiles, respectively; (c), (d) RA products and TER as functions of the screening parameter R=λm/ϵmR=\lambda_m/\epsilon_m and the corresponding energy-band profiles, respectively. All remaining simulation parameters are kept unchanged.

The effect of elastic scattering on resonant transport in the TiO2_2-based 2-QW FTJ is examined in Figure 9. The transmission spectra at Va=0V_a=0 V and 0.6 V in Figure 9(a) and Figure 9(b), respectively, show that increasing the dephasing strength DoD_o weakens and broadens the resonant peaks associated with QW states due to reduced phase coherence. The resulting transmission within the bias window determines the tunneling current and is reflected in the LRS and HRS RA products shown in Figure 9(c) and Figure 9(d), respectively. The LRS RA exhibits a stronger dependence on DoD_o, particularly at higher bias, whereas the HRS RA curves remain relatively close. Consequently, the TER varies with both bias and dephasing strength, as shown in Figure 9(e). The TER decreases with increasing bias over the investigated range, indicating a reduction in the resistance-state difference. Overall, elastic scattering affects LRS and HRS transport differently, thereby influencing the TER–RA characteristics.

Effect of elastic dephasing strength on the transport characteristics

Figure 9. Effect of elastic dephasing strength DoD_o on the transport characteristics of the TiO2_2-based 2-QW FTJ. Transmission spectra of the LRS and HRS at VaV_a (a) 0 V and (b) 0.6 V. Bias-dependent RA products for (c) the LRS and (d) the HRS, and (e) the corresponding TER for different DoD_o values.

Non-Uniform QW Layers

The effect of non-uniform QW thickness on the transport characteristics of the TiO2_2-based 2-QW FTJ is investigated by varying tqw1t_{qw1} for different values of tqw2t_{qw2}. The LRS and HRS RA products at Vr=0.1V_r=0.1 V for the investigated thickness combinations are presented in Figure 10(a), while the corresponding TER is shown in Figure 10(b). The TER reaches its maximum at tqw1=1.38 nmt_{\mathrm{qw1}}=1.38\,\mathrm{nm}. This dependence arises from changes in quantum confinement and the relative alignment of QW-confined states, which affect the LRS and HRS transmission probabilities differently. Considering the TER enhancement together with the corresponding RA values, the configuration with tqw1=1.38 nmt_{\mathrm{qw1}}=1.38\,\mathrm{nm}, tqw2=1 nmt_{\mathrm{qw2}}=1\,\mathrm{nm}, and ti1=ti2=1 nmt_{\mathrm{i1}}=t_{\mathrm{i2}}=1\,\mathrm{nm} is selected as the optimum TER–RA configuration. The corresponding JJ-VaV_{\mathrm{a}} characteristics in Figure 10(c) show distinct current densities for the LRS and HRS, along with an NDR feature at Va=0.36 VV_{\mathrm{a}}=0.36\,\mathrm{V}. Similar to the uniform structure, this behavior is associated with RT through QW-confined states. The bias-dependent transmission spectra in Figure 10(d) show changes in resonant transmission as the confined-state energies shift relative to the electrode transport-energy window. The zero-bias transmission spectra in Figure 10(e) also reveal different transmission probabilities for the LRS and HRS due to polarization-dependent barrier modulation. These differences are reflected in the voltage-dependent TER and RA characteristics in Figure 10(f). The selected non-uniform double-QW structure achieves a TER of 2.95×1011%2.95 \times10^{11}\% and an RA product of 1.25×103 Ω⋅cm21.25 \times10^{3}\,\Omega\cdot\mathrm{cm}^{2} at Vr=0.1 VV_{\mathrm{r}}=0.1\,\mathrm{V}. Thus, it provides a more favorable TER–RA trade-off, with approximately 10.9 times higher TER and 40 times lower RA than the uniform double-QW structure. This improvement results from the unequal QW thicknesses, which modify the confined-state alignment and polarization-dependent transmission in the two resistance states.

Effect of non-uniform QW thickness on the transport characteristics of the TiO2-based 2-QW FTJ

Figure 10. Effect of non-uniform QW thickness on the transport characteristics of the TiO2_2-based 2-QW FTJ. (a) LRS and HRS RA products and (b) corresponding TER as functions of tqw1t_{\mathrm{qw1}} for different values of tqw2t_{\mathrm{qw2}}, evaluated at Vr=0.1 VV_{\mathrm{r}}=0.1\,\mathrm{V}. (c) JJ–VaV_{\mathrm{a}} characteristics of the selected geometry (tqw1=1.38 nmt_{\mathrm{qw1}}=1.38\,\mathrm{nm}, tqw2=1 nmt_{\mathrm{qw2}}=1\,\mathrm{nm}, and ti1=ti2=1 nmt_{\mathrm{i1}}=t_{\mathrm{i2}}=1\,\mathrm{nm}); (d) bias-dependent transmission spectra showing resonant transmission near Va=0.36 VV_{\mathrm{a}}=0.36\,\mathrm{V}; (e) zero-bias transmission spectra for the LRS and HRS; and (f) bias-dependent TER and RA products.

Conclusion

In this work, the Miller phenomenological polarization model is coupled with the NEGF formalism to investigate the electrostatics and quantum transport characteristics of multilayer HfO2_2-based FTJs. A comparison of TiO2_2 and IGZO-based QWs shows that their dielectric and electronic properties strongly influence polarization retention and TER–RA characteristics. TiO2_2-based structures exhibit higher polarization retention and TER than IGZO-based structures, with the double-QW configuration offering the most favorable TER–RA trade-off. The transport analysis further shows that RT and NDR originate from transport through QW-confined states. In addition, finite electrode screening modifies the electrostatic potential profile, while elastic dephasing affects resonant transmission. The investigation of non-uniform QW thicknesses further demonstrates that variations in individual QW thicknesses modify the confined-state energies and their relative alignment, thereby influencing the resistance-state characteristics and TER–RA trade-off. These findings show how QW material selection and thickness configuration influence polarization-dependent transport and the resulting TER–RA characteristics of multilayer FTJs.

References

  1. [1]H. Kohlstedt, N. Pertsev, J. Rodríguez Contreras, and R. Waser, “Theoretical current-voltage characteristics of ferroelectric tunnel junctions,” Phys. Rev. B, vol. 72, no. 10, p. 125341, Sep. 2005, doi: 10.1103/PhysRevB.72.125341.
  2. [2]M. Y. Zhuravlev, R. F. Sabirianov, S. Jaswal, and E. Y. Tsymbal, “Giant electroresistance in ferroelectric tunnel junctions,” Phys. Rev. Lett., vol. 94, no. 24, p. 246802, Jun. 2005, doi: 10.1103/PhysRevLett.94.246802.DOI
  3. [3]M. Kobayashi, Y. Tagawa, F. Mo, T. Saraya, and T. Hiramoto, “Si-doped hfo2-based ferroelectric tunnel junctions with a composite energy barrier for non-volatile memory applications,” Materials, vol. 15, no. 6, pp. 134–139, 2019, doi: 10.1109/JEDS.2018.2885932.
  4. [4]Y. Lee, S. Song, W. Ham, and S.-E. Ahn, “Si-doped hfo2-based ferroelectric tunnel junctions with a composite energy barrier for non-volatile memory applications,” Materials, vol. 15, no. 6, p. 2251, 2022, doi: 10.3390/ma15062251.DOI
  5. [5]V. Deshpande, K. S. Nair, M. Holzer, S. Banerjee, and C. Dubourdieu, “CMOS back-end-of-line compatible ferroelectric tunnel junction devices,” Solid-State Electronics, vol. 186, p. 108054, 2021, doi: 10.1016/j.sse.2021.108054.DOI
  6. [6]R. Athle and M. Borg, “Ferroelectric tunnel junction memristors for in-memory computing accelerators,” Adv. Intell. Syst., vol. 6, no. 3, p. 2300554, Mar. 2024, doi: 10.1002/aisy.202300554.
  7. [7]N. Jao, Y. Xiao, A. K. Saha, S. K. Gupta, and V. Narayanan, “Design space exploration of ferroelectric tunnel junction toward crossbar memories,” IEEE J. Explor. Solid-State Comput. Devices and Circuits, vol. 7, no. 2, pp. 115–122, Dec. 2021, doi: 10.1109/XJCDC.2021.3117566.DOI
  8. [8]Z. Ma, T. Zhang, K. Liang, Y. Qi, D. Wang, J. Wang, and J. Jiang, “Ferroelectric tunnel junctions with multi-quantum well structures,” Applied Physics Letters, vol. 104, no. 22, 2014, doi: 10.1063/1.4880896.DOI
  9. [9]J. Su, X. Zheng, Z. Wen, T. Li, S. Xie, K. M. Rabe, X. Liu, E. Y. Tsymbal, “Resonant band engineering of ferroelectric tunnel junctions,” Phys. Rev. B, vol. 104, no. 6, p. L060101, 2021, 10.1103/PhysRevB.104.L060101.DOI
  10. [10]J. Hwang, Y. Goh, and S. Jeon, “Effect of insertion of dielectric layer on the performance of hafnia ferroelectric devices,” IEEE Trans. on Electron Devices, vol. 68, no. 2, pp. 841–845, Feb. 2021, doi: 10.1109/ted.2020.3043728.DOI
  11. [11]H. Joh, T. Jung, and S. Jeon, “Stress engineering as a strategy to achieve high ferroelectricity in thick hafnia using interlayer,” IEEE Trans. on Electron Devices, vol. 68, no. 5, pp. 2538–2542, May 2021, 10.1109/TED.2021.3068246.
  12. [12]F. Mo, Y. Tagawa, C. Jin, M. Ahn, T. Saraya, T. Hiramoto, and M. Kobayashi, “Low-voltage operating ferroelectric fet with ultrathin igzo channel for high-density memory application,” IEEE Journal of the Electron Devices Society, vol. 8, pp. 717–723, 2020, doi: 10.1109/JEDS.2020.3008789.
  13. [13]B. Khattar, A. Tripathi, M. Brahma, and A. Sharma, “Asymmetric resonant ferroelectric tunnel junctions for simultaneous high tunnel electroresistance and low resistance-area product,” Applied Physics Devices, vol. 1, no. 4, p. 046106, 2025, doi: 10.1063/5.0286849.DOI
  14. [14]P. Chang and Y. Xie, “Evaluation of HfO₂-based ferroelectric resonant tunnel junction by band engineering,” IEEE Electron Device Lett., vol. 44, no. 1, pp. 168–171, Jan. 2023, 10.1109/LED.2022.3225298.
  15. [15]P. Chang, J. Li, Y. Guo, Y. Xie, and L. Zeng, “Physical modeling of resonant tunneling in ferroelectric tunnel junctions with multiple quantum well superlattice heterostructures,” IEEE Transactions on Electron Devices, vol. 71, no. 3, pp. 2210–2216, 2024, 10.1109/TED.2024.3356474.DOI
  16. [16]M. Alexander, M. Nido, W. Rühle, and K. Köhler, “Resonant-tunneling transfer times between asymmetric GaAs/Al₀.₃Ga₀.₆₅As double quantum wells,” Physical Review B, vol. 41, no. 17, p. 12295, 1990, doi: 10.1103/PhysRevB.41.12295.
  17. [17]E. Diez, I. Gómez, F. Dominguez-Adame, R. Hey, V. Bellani, and G. Parravicini, “Gaussian semiconductor superlattices,” Physica E: Low-dimensional Systems and Nanostructures, vol. 7, no. 3-4, pp. 832–835, 2000, doi: 10.1016/S1386-9477(00)00071-0.DOI
  18. [18]I. Gómez, F. Dominguez-Adame, E. Diez, and V. Bellani, “Electron transport across a gaussian superlattice,” Journal of Applied Physics, vol. 85, no. 7, pp. 3916–3918, 1999, doi: 10.1063/1.369764.DOI
  19. [19]C. Kittel and P. McEuen, Introduction to solid state physics. John Wiley and Sons, 2018.
  20. [20]D. Pantel and M. Alexe, “Electroresistance effects in ferroelectric tunnel barriers,” Phys. Rev. B, vol. 82, no. 13, p. 134105, Oct. 2010, doi: 10.1103/physrevb.82.134105.DOI
  21. [21]S.-C. Chang, A. Naeemi, D. E. Nikonov, and A. Gruverman, “Theoretical approach to electroresistance in ferroelectric tunnel junctions,” Phys. Rev. Appl., vol. 7, no. 2, p. 024005, Feb. 2017, doi: 10.1103/PhysRevApplied.7.024005.DOI
  22. [22]S. Miller, J. Schwank, R. Nasby, and M. Rodgers, “Modeling ferroelectric capacitor switching with asymmetric nonperiodic input signals and arbitrary initial conditions,” Journal of applied physics, vol. 70, no. 5, pp. 2849–2860, 1991, doi: 10.1063/1.349348.DOI
  23. [23]Z. Zhou, L. Jiao, J. Zhou, Q. Kong, S. Luo, C. Sun, Z. Zheng, X. Wang, D. Zhang, G. Liu et al., “Time-dependent landau-ginzburg equation-based ferroelectric tunnel junction modeling with dynamic response and multi-domain characteristics,” IEEE Electron Device Lett., vol. 43, no. 1, pp. 158–161, Jan. 2022, doi: 10.1109/LED.2021.3128998.DOI
  24. [24]S. Datta, Quantum transport: atom to transistor. Cambridge university press, 2005.DOI
  25. [25]A. Sharma, A. Tulpurkar, and B. Muralidharan, “Ultrasensitive nanoscale magnetic-field sensors based on resonant spin filtering,” IEEE Trans. on Electron Devices, vol. 63, no. 11, pp. 4527–4534, Nov. 2016, doi: 10.1109/TED.2016.2606354.DOI
  26. [26]F. Fillot, T. Morel, S. Minoret, I. Matko, S. Maîtrejean, B. Guillaumot, B. Chenavier, and T. Billon, “Investigations of titanium nitride as metal gate material, elaborated by metal organic atomic layer deposition using tdmat and NH₃,” Microelectroron. Eng., vol. 82, no. 3-4, pp. 248–253, Dec. 2005, doi: 10.1016/j.mee.2005.07.083.DOI
  27. [27]D. Shah, H. Reddy, N. Kinsey, V. M. Shalaev, and A. Boltasseva, “Optical properties of plasmonic ultrathin tin films,” Advanced Optical Materials, vol. 5, no. 13, p. 1700065, 2017, doi: 10.1002/adom.201700065.DOI
  28. [28]S. Tripura Sundari, R. Rameshshan, F. Jose, S. Dash, and A. Tyagi, “Investigation of temperature dependent dielectric constant of a sputtered tin thin film by spectroscopic ellipsometry,” J. Appl. Phys., vol. 115, no. 3, Jan. 2014, doi: 10.1063/1.4862485.DOI
  29. [29]J. Robertson, “Band offsets, schottky barrier heights, and their effects on electronic devices,” J. Vac. Sci. Technol. A, vol. 31, no. 5, p. 050821, Aug. 2013, doi: 10.1116/1.4818426.DOI
  30. [30]S. Mondal, A. Kumar et al., “Tunable dielectric properties of TiO₂ thin film based mos systems for application in microelectronics,” Superlattices and Microstruct., vol. 100, pp. 876–885, Dec. 2016, doi: 10.1016/j.spmi.2016.10.054.DOI
  31. [31]S. Monaghan, P. Hurley, K. Cherkaoui, M. Negara, and A. Schenk, “Determination of electron effective mass and electron affinity in HfO₂ using mos and mosfet structures,” Solid-State Electronics, vol. 53, no. 4, pp. 438–444, Apr. 2009, doi: 10.1016/j.sse.2008.09.018.DOI
  32. [32]J. Park, D.-K. Kim, J.-I. Park, I. M. Kang, J. Jang, H. Kim, P. Lang, and J.-H. Bae, “Numerical analysis on effective mass and traps density dependence of electrical characteristics of a-igzo thin-film transistors,” Electronics, vol. 9, no. 1, p. 119, 2020, doi: 10.3390/electronics9010119.
  33. [33]S. K. Dargar and V. M. Srivastava, “Design and analysis of igzo thin film transistor for amoled pixel circuit using double-gate tri active layer channel,” Heliyon, vol. 5, no. 4, 2019, doi: 10.1016/j.heliyon.2019.e01452.DOI
  34. [34]Y. Li, Y. Pei, R. Hu, Z. Chen, Y. Zhao, Z. Shen, B. Fan, J. Liang, and G. Wang, “Effect of channel thickness on electrical performance of amorphous igzo thin-film transistor with atomic layer deposited alumina oxide dielectric,” Current Applied Physics, vol. 14, no. 7, pp. 941–945, 2014, doi: 10.1016/j.cap.2014.04.011.

Paper details

vlnc:W_wqsf5rc2
Version
v1 · latest
Published
Fresh — added to the Hub within the last 30 days
Added
License
CC BY 4.0
Usage

views · 2

downloads · 0

A view counts once per browser per day, a download each time the PDF is requested.

  • Physics and Astronomy

Not recorded for this paper

View statement
Looking for related papers…
Contents