Ferroelectric HZO—the hafnium zirconium oxide compound Hf0.5Zr0.5O2—is a candidate material for compact nonvolatile memory and low-power transistor technologies. Its appeal lies in the ability to produce ferroelectric behavior in thin films, while its composition and deposition approach are compatible with CMOS-oriented manufacturing. But realizing that behavior depends on more than the oxide itself: electrode materials, film thickness, and thermal processing all influence the result.
Experiments with platinum (Pt), nickel (Ni), and palladium (Pd) capping layers show that elemental metals can help induce ferroelectricity in HZO. This broadens the electrode choices beyond the titanium nitride and tantalum nitride stacks often used in earlier HZO research, an important consideration when integrating the material into different device structures and process flows.
Why ferroelectric HZO matters for device integration
Ferroelectric materials can retain polarization after an applied electric field is removed, a property that makes them relevant to nonvolatile memory. HZO is of interest because it can exhibit ferroelectricity at nanometer-scale thicknesses, unlike some traditional ferroelectric materials that require thicker films to limit leakage. It can also be deposited using atomic layer deposition (ALD), a thin-film process suited to controlled deposition.
Potential device applications include ferroelectric field-effect transistors (FeFETs) and negative capacitance field-effect transistors (NC-FETs). In an NC-FET, a ferroelectric layer is integrated into the transistor gate stack to create an effective negative capacitance under suitable conditions. The study discussed here evaluates metal-ferroelectric-metal (MFM) capacitors, however; it does not demonstrate a complete FeFET or NC-FET. Its findings concern the materials and processing choices that may inform later device integration.
How capping layers influence ferroelectricity
HZO’s ferroelectric behavior is associated with its orthorhombic crystal phase. The electrodes on either side of the film do more than provide electrical contact: they mechanically constrain the film during processing. That confinement can influence which crystal structures form as the material is annealed.
Using different electrode combinations also provides integration flexibility. Electrode properties, including work function, can affect transistor behavior, while the choice of material may matter when working with sensitive channel materials or particular fabrication constraints. Elemental metal electrodes can also offer patterning options, such as lift-off, rather than requiring an additional etch step for a deposited electrode film. These are integration considerations, not guarantees that a given metal will suit every process.
The capacitors in the study used ALD-grown HZO and Pt, Ni, or Pd electrodes. Without annealing, the measured polarization–electric field response was linear rather than ferroelectric. After rapid thermal annealing, hysteresis appeared, with the temperature at which it emerged varying by electrode stack. For example, Pd/HZO/Ni showed ferroelectric behavior at 500°C, while the tested Pt/HZO/Ni stack required 550°C.
Thickness and annealing must be considered together
The strongest reported result came from a Pt/HZO/Ni capacitor with approximately 20 nm of HZO, annealed at 650°C for 30 seconds. It reached a remnant polarization of 19 μC/cm² and a coercive field of 1.07 MV/cm. Remnant polarization describes the polarization remaining after the applied field is removed; coercive field is the field associated with switching the polarization.
These values should be understood in the context of the tested capacitor structure and process conditions. They are not a universal recipe for HZO, nor evidence that the same performance will transfer unchanged to a transistor stack.
Thickness was especially important in the Pd/HZO/Ni tests. Films between 10 and 15 nm did not show ferroelectric behavior under the reported conditions, while films 20 nm or thicker did. The authors noted that the thinner samples could become resistive, with leakage paths along grain boundaries as a possible concern. Although thinner HZO films are often explored for higher polarization, this result illustrates why thickness cannot be optimized independently of film quality, electrodes, and annealing.
Annealing also had limits. In the tested stacks, increasing temperature generally promoted ferroelectric behavior, but devices became resistive above approximately 800°C. For Pt/HZO/Ni, diffraction measurements showed stronger evidence of the orthorhombic phase after annealing, consistent with the electrical measurements. The results underscore the need to balance phase formation against the effects of excessive thermal processing.
What the capacitor tests reveal about reliability
Retention and fatigue measurements provide an initial view of memory-related behavior, but they should not be treated as full qualification data for an application. In the reported tests, the measured samples showed similar retention trends over more than 700 cycles. During fatigue testing, switching polarization fell substantially in the first 10,000 cycles and then leveled off after roughly 20,000 cycles.
Those observations indicate both promise and a development need. The electrode stacks produced ferroelectric behavior, but endurance remains a consideration for memory use. A materials result in an MFM capacitor is one step in evaluating a device; practical implementation also depends on the target architecture and its processing requirements.
Practical implications for process and device teams
For engineers assessing HZO integration, the central takeaway is that electrode selection is a process variable, not simply a matter of choosing a conductor. The Pt/HZO/Ni stack delivered the highest reported remnant polarization in this comparison, while other combinations showed different annealing thresholds and phase behavior. Film thickness likewise changed whether ferroelectric response was observed.
Elemental Pt, Ni, and Pd therefore expand the available options for investigating ferroelectric HZO, particularly where conventional nitride electrodes are unsuitable for a device structure or fabrication flow. The results do not establish a single best stack for all applications. Instead, they point to a practical evaluation strategy: consider electrode pairing, HZO thickness, annealing window, leakage, and endurance together. That process-level view is essential when moving from a promising capacitor result toward memory or low-voltage transistor integration.