Thickness-Dependent Crystallization of Ultrathin Antimony Thin Films for Monatomic Multilevel Reflectance and Phase Change Memory Designs

Phase change materials, with more than one reflectance and resistance states, have been a subject of interest in the fields of phase change memories and nanophotonics. Although most current research focuses on rather complex phase change alloys, e.g., Ge2Sb2Te5, recently, monatomic antimony thin films have aroused a lot of interest. One prominent attractive feature is its simplicity, giving fewer reliability issues like segregation and phase separation. However, phase transformation and crystallization properties of ultrathin Sb thin films must be understood to fully incorporate them into future memory and nanophotonics devices. Here, we studied the thickness-dependent crystallization behavior of pulsed laser-deposited ultrathin Sb thin films by employing dynamic ellipsometry. We show that the crystallization temperature and phase transformation speed of as-deposited amorphous Sb thin films are thickness-dependent and can be precisely tuned by controlling the film thickness. Thus, crystallization temperature tuning by thickness can be applied to future memory and nanophotonic devices. As a proof of principle, we designed a heterostructure device with three Sb layers of varying thicknesses with distinct crystallization temperatures. Measurements and simulation results show that it is possible to address these layers individually and produce distinct and multiple reflectance profiles in a single device. In addition, we show that the immiscible nature of Sb and GaSb could open up possible heterostructure device designs with high stability after melt-quench and increased crystallization temperature. Our results demonstrate that the thickness-dependent phase transformation and crystallization dynamics of ultrathin Sb thin films have attractive features for future memory and nanophotonic devices.

2 SI 1 -Spectroscopic ellipsometry data fitting and AFM thickness extraction. thickness relationships were derived by combining initial AFM thickness results with spectroscopic fitting for thicker and thinner films.
In spectroscopic ellipsometry, linearly polarized light is reflected by a sample. The change in phase and intensity, before and after reflection, is used to extract information about the sample.
The interaction of the initial linearly polarized light with the sample will produce transmission 3 and reflection light intensities upon reaching multiple interfaces inside the sample. The reflection and transmission of partial intensities on the interfaces can be represented by a collection of equations called the Fresnel equations. The total reflected light intensities for both the perpendicular (s-polarized) and parallel (p-polarized) lights (Rs and Rp) for the specific incidence angle are directly related to the ellipsometry measured parameters (ψ and Δ). The complex reflection coefficient ( ), which is the ratio of the Rp and Rs, relate the reflectance intensities with the measured parameters as: Where and Δ represent the change in intensity and phase of the detected light from the original linearly polarized light. 1,2 The measurement parameters of ψ and Δ by themselves do not represent any material properties.
Therefore, they do not provide enough information about the measured sample (except for bulk samples and dynamic ellipsometry measurement discussed in section SI-3). In turn, they have to be converted into optical material properties like an index of refraction (n) and extinction coefficient (k) and physical parameters like film thickness and roughness. Therefore, a model has to be constructed based on individual layers present in the sample to extract the physical and optical properties of the measured sample. Each layer in the model is represented by dispersion relations containing fitting parameters. For our samples, a commercially available software, WVASE, produced by the J.A. Woollam Company, was used to perform the data fitting.
Measurement data of ψ and Δ for Sb thin films of variable thickness were collected in the 300 -1700 nm spectrum range. The measurements were done for both as-deposited and crystalline films. Figure S1  One fitting parameter was roughness, described by an effective medium approximation (EMA), where the value was assumed by 50% of the topmost layer and 50% void . 2 A roughness layer thickness value of zero produced the best results in our fitting, indicating a smooth surface of our deposited thin films. Next, AFM scans in tapping mode were performed to confirm our deposited thin films' full coverage and smoothness. An example of such a scan is shown in Figure S1 (b) for an Sb thin film of 400 pulses (4.7 nm in thickness). The Gwyddion software package was used to analyze the AFM images. 4 The mean roughness value (Ra) of < 0.2 nm was found in this case, indicating a very smooth surface. The full coverage and smoothness of the 500 pulses Sb film is also visible on the right side of the scratch in Figure S1  respectively. We extracted an average of 6.3 nm thickness from the line profiles, which was used as initial input for the ellipsometry data fitting. After this fitting, a thickness of 6.0 nm was extracted. Based on these results for the 500 pulses Sb film, the fitting was also performed for the Sb film produced with fewer pulses and then the input thickness for the fitting was adapted directly proportional to the number of pulses used. films. An additional TEM image, over a large area, for the heterostructure device based on multiple Sb thin film layers with varying thicknesses is given in Figure S2  Multiple measurement parameters can be used to study the phase transformatio n. This study mainly focused on the parameter and the pseudo index of refraction (<n>) values extracted by assuming the heterostructure as a bulk. Figure S3 ( happened at room temperature before the measurement, and annealing to higher temperatures then does not produce any property changes. Figure S3 (b) shows spectroscopic scan measurements of the 600P Sb thin film before and after annealing to 220 O C. For all incidence angles used, no change in the measured values is observable since no phase transformation occurred.