Materials Matter

A blog from Intermolecular

Home / Company / News / Why Does Nano-Stacking Make Ferroelectric Materials Better?

Blog Post

Why Does Nano-Stacking Make Ferroelectric Materials Better?

Ferroelectric materials could speed up computing by enabling new types of logic and memory devices. Hafnium oxide, blended with zirconium oxide, is a promising ferroelectric for these devices. Scientists at Intermolecular, part of the Performance Materials Business of Merck KGaA Darmstadt Germany, recently demonstrated a new technique that will accelerate the engineering of devices using these critical materials.

To improve ferroelectric properties of hafnium-zirconium oxide blends, scientists stacked nanometer-thick films of HfO2 and ZrO2.[1] These nanolaminates are made using Intermolecular’s atomic layer deposition (ALD) technology, with which the thickness of each layer is controlled to the precision of a single atom. To determine how stacking affects performance, bilayers were prepared with just 3 nanometers(nm) of either ZrO2 or HfO2 deposited on a titanium nitride electrode, followed by 3nm of the other material (Figure 1). Even though the overall composition was the same, the difference in ferroelectric performance was dramatic: the sample with HfO2 deposited first had over twice the remanent polarization, a key metric of ferroelectric performance, than the sample with ZrO­2 deposited first.[2]

Figure 1. (Hf,Zr)O2 bilayer stack configurations and corresponding ferroelectric polarization curves.

To find out why this happens, we need to pinpoint whether HfO2, ZrO2, or both materials are responsible for the change in performance. The answer lies in the crystal structures of each material. The atoms that make up HfO2 and ZrO2 can arrange in several different configurations. Subtle changes in material chemistry, interface structure or processing conditions can favor the ferroelectric orthorhombic crystal structure over the monoclinic or tetragonal phases. Nanometers-thick films of these materials typically contain a mixture of crystal phases, so a larger fraction of orthorhombic crystallites in the film yields better ferroelectric performance.

To find out how much of each type of crystal structure makes up a film, the best tool is usually X-ray diffraction (XRD). While XRD can easily distinguish the monoclinic phase from other structures, the tetragonal and orthorhombic phases have remarkably similar diffraction patterns (Figure 2). In fact, for few-nanometer crystallites, the subtle differences between their diffraction patterns might be totally invisible.

Figure 2. Simulated X-ray diffraction patterns and crystal structures for the monoclinic (m), ferroelectric orthorhombic (oFE), and tetragonal (t) phases of HfO2.

XRD probes the long-range crystalline order of a material, which is similar for the tetragonal and orthorhombic phases. However, the local atomic arrangements of these two phases are quite different (Figure 3). In the tetragonal structure, the cation (Hf4+ or Zr4+) is bonded to 8 oxygen ions. In the orthorhombic and monoclinic structures, the cation is only bonded to 7 oxygen ions, and the bond lengths are different. One type of X-ray absorption spectroscopy, called EXAFS, measures the atomic neighborhood surrounding one element in a sample. Such a measurement could tell apart the tetragonal and orthorhombic phases (see Figure 3).

Figure 3. Simulated EXAFS spectra and Hf—O local structure for different crystal structures of HfO2.

It can be difficult to determine the signals from each crystal phase that make up an EXAFS spectrum, especially for films that have mixtures of multiple phases of HfO2 and ZrO2. A new analysis developed at Intermolecular overcomes this problem. First, a spectrum is simulated for each crystal phase that could be present in the film. Then, a mixture of spectra is fitted to the measured spectrum to estimate the real crystal phase composition of the film. By combining the insights from XRD and EXAFS, the distribution of all crystal phases in both HfO2 and ZrO2 can be fully determined (Figure 4).

Figure 4. Approximate crystal phase fractions in HfO2—ZrO2 bilayers. Clockwise from the top, pie chart slices represent monoclinic (red), orthorhombic (green), and tetragonal (blue) phase fractions.

We now see why the sample with HfO2 deposited first outperforms the opposite stacking sequence (Figure 1). While there is a somewhat higher orthorhombic phase fraction of HfO2 for this sample, the orthorhombic ZrO2 phase fraction is also much higher. This is a surprising result since the HfO2 layer is often expected to be the main contributor to ferroelectric performance.

Thanks to the clarity of this new method, the interfacial mechanisms underlying the superior ferroelectric performance of HfO2—ZrO2 nanolaminates may soon come to light, aiding the design and production of next-generation devices.



[1]       S. L. Weeks, A. Pal, V. K. Narasimhan, K. A. Littau, T. Chiang, ACS Appl. Mater. Interfaces 2017, 9, 13440.

[2]       M. E. McBriarty, V. K. Narasimhan, S. L. Weeks, A. Pal, H. Fang, T. A. Petach, A. Mehta, R. C. Davis, S. V Barabash, K. A. Littau, Phys. status solidi 2019, 1900285.