Tuning Negative Capacitance in ${\mathrm{Pb}\mathrm{Zr}}_{0.2}{\mathrm{Ti}}_{0.8}{\mathrm{O}}_3/{\mathrm{Sr}\mathrm{Ti}\mathrm{O}}_3$ Heterostructures via Layer Thickness Ratio

In this work, we exploit the ferroelectric-dielectric layer thickness ratio r as an effective tuning parameter to control the ferroelectric polarization and transient negative capacitance (NC) state in epitaxial ${\mathrm{Pb}\mathrm{Zr}}_{0.2}{\mathrm{Ti}}_{0.8}{\mathrm{O}}_3/{\mathrm{Sr}\mathrm{Ti}\mathrm{O}}_3$ bilayer heterostructures. The remnant polarization decreases monotonically with decreasing r, with the system exhibiting an abrupt transition from ferroelectric to dielectric dominated behavior at a critical ratio $r_c$ of 8 to 7, which is consistent with the evolution of the free-energy profile modeled via Landau theory. For samples with large r, the polarization switching dynamics during the transient NC regime can be well described by the nucleation and growth model, with the narrow distribution of the characteristic switching time ($t_0$) pointing to a domain-wall-motion-limited behavior. Right below $r_c$, we observe a significantly broadened distribution of $t_0$, which can be attributed to the emergence of a multidomain state. The transient NC mode is quenched in samples with r < 6, confirming that the ferroelectric order is suppressed. Our study provides critical information for optimizing the materials design for complex oxide-based NC transistors, paving the path for their application in low-power nanoelectronics.

Figure 1

(a) Sample schematic. (b) AFM topography image and (c) XRD θ−2θ scan taken on a 60 nm PZT/40 nm STO/10 nm LNO (r = 1.5) sample. STO (002)_f and (002)_s label the Bragg peaks of the STO film and substrate, respectively. The (001) peak for the STO film is close to the substrate peak and not resolved. (d) HRTEM image of a 100 nm PZT/5 nm STO/10 nm LNO (r = 20) sample. The dotted lines mark the interfaces.

Figure 2

(a) P(E) hysteresis loops taken on PZT/STO stacks with different r. (b) Effective dielectric constant versus  V_bias taken on 100 nm PZT, STO, and PZT/STO stack capacitors with different r. (c) 2P_r as a function of 1/r for PZT (open circle), PZT/STO (open circles), and STO/PZT (solid triangles). The red dashed line is the simulation result using the single-domain model.

Figure 3

(a) P(E) hysteresis loops taken on a 100 nm PZT capacitor at 1, 2.5, 5, 10, and 25 kHz (along the arrow direction), superimposed with the simulated P(E) (dashed line). (b) Simulated g(P) for PZT/STO stacks with different r at zero bias. (c)–(d) Simulated polarization of PZT/STO stacks as a function of total voltage (V) and voltage across the FE layer (V_f) for (c) r = 7.9 and (d) 7.7. t_STO= 10 nm. The arrows in (c) illustrate the polarization switching as the free energy becomes unstable.

Figure 4

(a) Voltage across the r = 8 stack capacitor (V_cap) during charging and discharging upon applied square wave source voltage (V_s). Inset: circuit schematic. Load resistor R = 9.86 k$\mathrm{\Omega }$. (b) Transient NC in samples with r = 8, 7, 6. (c)–(e) Q_s versus V_cap extracted for samples shown in (b).

Figure 5

(a),(b) Three sets of V_c(t) data in the transient NC regime (open symbols) for the (a) r = 8 and (b) r = 7 samples with fits to the KAI/NLS model (solid line) and LK model (dashed line). (c),(d) Distribution functions of log (t₀) extracted from the NLS fitting results for the r = 8 (red) and r = 7 (black) samples plotted on (c) linear and (d) logarithmic scales.