Novel ultra-low-loss CMOS copper nanoplasmonics

Metal-based components have been the essential elements of photonic systems since the early days of their development. In the past, the functionality of such components was limited only to their ability to efficiently reflect light. In modern systems, however, novel metal-based components are required to push the photonic technologies to new frontiers. In addition, photonic and optoelectronic devices are diffraction limited, i.e., their dimensions cannot be reduced below the wavelength of light (because of the diffraction of light). Traditionally, this has meant that truly nanoscale photonic devices and circuits could not be created.

It may be possible, however, to overcome this fundamental diffraction limitation with the use of gold and silver. These two metals show a highly negative permittivity at optical frequencies, which means that light cannot easily propagate through them (i.e., only to a depth of about 25nm). Accordingly, photons can be efficiently localized in wavelength-sized (λ-sized) photonic cavities.¹ In addition, light can be converted with these metals into surface plasmon polaritons (SPPs), i.e., collective electromagnetic excitations at the interface between a metal and an insulator.² It is therefore possible to switch from conventional 3D photonics to 2D surface plasmon photonics and to control light at the scale of 10–100nm (far beyond the diffraction limit of light). The SPP electromagnetic field is strongly confined to the metal–dielectric interface. Plasmonic components are thus being used in the development of many practical applications, including next-generation photonic circuitry,³ highly sensitive biosensors,⁴ as well as fast and dense magnetic recorders.⁵ It is not possible, however, to translate these devices from the laboratory to industry because they are based on silver and gold, i.e., noble metals that are chemically inert and thus incompatible with industry-standard manufacturing processes. Although the nanophotonics community is trying to find suitable alternatives to noble metals for these purposes,^{6, 7} the results so far do not surpass silver and gold in terms of the losses in deep-subwavelength plasmonic devices. This is therefore currently one of the most critical bottlenecks to the commercialization of plasmonics and metal-based nanophotonics.

Despite the common belief (e.g., shown in many experimental studies^7,8) that copper is substantially more ‘lossy’ than gold, we decided to focus on this metal in our work because it is a key material in modern electronics. For example, in previous work,⁹ we studied the optical properties of polycrystalline gold films. Our aim was to understand the differences in the measurements that had been obtained by different research groups. We found that the optical properties were determined by the internal polycrystalline structure of the metal film, which was unique for each sample. Furthermore, in our first theoretical calculations we found that the dielectric function of polycrystalline copper can be as good as—if not better than—that of polycrystalline gold.

To experimentally verify our theoretical calculations, we have also developed a deposition process that is based on the electron beam evaporation method. The copper films we thus deposit exhibit remarkable optical characteristics, as confirmed by spectroscopic ellipsometry measurements (see Figure 1). The ellipsometric results, however, are very sensitive to the quality of the film surface and to the presence of a sub-nanometer-thick oxide layer. These features therefore strongly affect the accuracy of the measurements.

Figure 1. Ellipsometric measurements of the imaginary part of the dielectric function(ε”) of the as-deposited copper (Cu) film immediately after deposition (red line). The data from previous studies^{8, 11} is also shown (yellow and blue lines).

To overcome this problem, we used a CMOS-compatible process¹⁰ to fabricate an array of deep-sub-wavelength copper plasmonic waveguides (see Figure 2). We conducted scanning near-field optical microscopy measurements of these waveguides and obtained an SPP propagation length of 30μm at a light wavelength of 1550nm. By measuring the SPP dispersion and fitting our experimental data with the numerical simulations, we found that the imaginary part of the copper's dielectric function was 6.0 (at a wavelength of 1550nm). For comparison, the imaginary part of the dielectric function of gold (reported by Johnson and Christy⁸) is nearly double (i.e., 11 at a wavelength of 1550nm). Our results also indicate that our fabrication process for the nanoscale plasmonic waveguides provides a substantial improvement to the optical properties of the as-deposited copper films. This is because the sample is kept at temperature of 390°C for 20 minutes during our plasma-enhanced chemical vapor deposition process. The fabricated copper sample therefore exhibits optical properties that are two times better than those of the gold samples produced by Johnson and Christy.

Figure 2. (a) Scanning near-field optical microscopy image of a silicon (Si) chip with an array of nanoscale Cu plasmonic waveguides. (b) Scanning electron microscopy image of the plasmonic waveguide's cross-section. SiN: Silicon nitride.

In summary, we have demonstrated for the first time that it is possible to fabricate truly nanoscale ultra-low-loss copper plasmonic components with a relatively simple CMOS-compatible process. Furthermore, the experimentally measured optical characteristics of our components are superior to those of previously produced similar gold components.¹⁰ Our copper components are also advantageous because copper—unlike gold—can be used in industry-standard manufacturing processes. These results open up the prospect of practical implementation of copper components for plasmonic devices as well as their commercialization. We believe, however, that the technological limits of our approach have not yet been achieved. In our future work we therefore plan to further improve the optical characteristics of polycrystalline copper films and nanostructures.

This work was supported by the Russian Science Foundation (grant 14-19-01788).