Material requirements for plasmonics and metamaterials

The unifying factor for almost all plasmonic structures and optical metamaterial devices lies in their constituent materials and fabrication methods. We normally manufacture and arrange nanoscale unit cells, which contain a metal and sometimes dielectrics, in a predesigned fashion to achieve the properties predicted by the design. But which constituent materials should be used? Metals have traditionally been the material of choice for plasmonics and optical metamaterials because they can support the needed SP oscillations responsible for the unusual optical properties of the resultant device. To date, researchers have mostly used silver and gold for all applications. These metals have the lowest optical losses compared to other metals, and hence they are the best materials for optical applications. However, even silver and gold still exhibit excessive losses, arising in part from interband electronic transitions, at operational frequencies in the visible or the near-infrared (NIR) ranges.^{Reference Naik, Shalaev and Boltasseva12,Reference West, Ishii, Naik, Emani, Shalaev and Boltasseva13}

Another problem, in addition to the excessive losses, which are detrimental for any optical device, is that the optical properties of noble metals cannot be tuned. The dielectric permittivity cannot be adjusted or changed easily, which is an undesirable feature for many advanced designs of metamaterials and in the related field of transformation optics.^{Reference Pendry, Schurig and Smith3} In particular, the fact that the real part of the dielectric permittivity of metals is very large in magnitude in the optical range is limiting the realization of metamaterial devices. This is the case for metamaterials with an effective dielectric permittivity close to zero, which leads to a range of very interesting properties, such as subwavelength resolution.^{Reference Engheta10,Reference Podolskiy and Narimanov11}

For such a metamaterial consisting of alternating layers of metal and dielectric, in order to “cancel out” metallic and dielectric responses, we need to make the metal layers very thin and combine them with thicker dielectric layers, while maintaining all the structures on a subwavelength scale. Due to fabrication limitations, creating uniform, smooth metal layers with thicknesses below 10 nm becomes increasingly difficult.^{Reference Naik, Shalaev and Boltasseva12} Structuring metal on the nanoscale also introduces additional losses that are again detrimental to the device performance.^{Reference Naik, Shalaev and Boltasseva12} In addition to the high losses and the inability to adjust or tune the optical properties, noble metals are quite expensive. Further, the most important factor limiting practical applications of plasmonics and metamaterial devices is that these metals are not compatible with existing standards in semiconductor technologies. To develop real-life devices, researchers need to find materials that are compatible with complementary metal oxide semiconductor (CMOS) technologies.

Material requirements can be summarized as follows. Low-loss metallic components need to be combined with the ability to adjust or tune the optical properties of the metallic material according to the design. Switchable devices, where drastic changes in the optical properties of the constituent materials can be achieved by applying external excitation, and low-cost, semiconductor compatible materials need to be developed. The current status of plasmonics and metamaterials research can be compared to the early stages of progress in silicon chip technology. In the 1980s, only a few elements from the periodic table were present in a silicon chip. With the development of silicon chip technology, the introduction of more elements and more advanced functions into the chip led to a large expansion in the realm of materials used to create more sophisticated devices.

Looking for alternatives

What materials can be used instead of Ag and Au? For plasmonic applications and optical metamaterials, materials that support SPs are necessary; thus, the research community needs to find materials with a negative real part of their dielectric permittivity in the optical range. A recent paper^{Reference Boltasseva and Atwater4} outlined different classes of materials that could replace metals in various plasmonic and metamaterial applications, depending on the wavelength range (see Figure 1).

Figure 1.

Material space for plasmonics and metamaterials applications. The important material parameters such as carrier concentration (maximum doping concentration for semiconductors), carrier mobility, and interband losses form the optimization space for various applications. While the spherical bubbles represent materials with low interband losses, elliptical bubbles represent those with larger interband losses in the corresponding part of the electromagnetic spectrum. Adapted with permission from Reference 4.

Here, the visible and NIR spectral ranges are discussed, but there are a number of different materials that hold promise for applications in the ultraviolet (UV) and mid-infrared (mid-IR) regions.^{Reference Naik, Shalaev and Boltasseva12,Reference West, Ishii, Naik, Emani, Shalaev and Boltasseva13} Figure 1 is not a complete collection of all of the materials that can be used in plasmonics and metamaterials applications, but it shows some of the main options available. In the figure, various classes of materials are grouped using the two parameters that determine the optical properties of conducting materials: the carrier concentration and carrier mobility. For a good plasmonic material, the carrier concentration has to be high enough to provide a negative real permittivity, but it should also be tunable (such that other permittivity values are accessible with changes in carrier concentration). Lower carrier mobilities translate to higher Drude damping losses (associated with the scattering of conduction electrons by defects, surface states, and lattice vibrations) and thus higher material losses.^{Reference Naik, Shalaev and Boltasseva12–Reference Khurgin and Boltasseva15} Additional losses due to interband transitions are highly undesirable and severely limit potential applications. The ideal material for plasmonics would lie at the left end of the plot on the horizontal plane, where interband losses are zero. Such a lossless metal is, however, still elusive.^{Reference Khurgin and Sun14}

Before looking at other material classes, it should be noted that researchers have examined metals other than Ag, Au, and their alloys (Reference 13 and references therein). It was found that some of the metals can have performances comparable to that of Ag and Au (such as Cu or alkali metals) and/or can be used in the wavelength ranges where Au and Ag do not provide the required performance (such as Al that can be used as a good plasmonic material in the UV range). However, the application realm of these materials is still very limited due to different factors related to their optical and chemical properties and fabrication issues (References 12 and 13 and references therein).

A holistic approach to the search for alternative plasmonic materials can be the following. One of the main problems with using metals is that they have very high carrier concentrations, resulting in a high plasma frequency, which in turn leads to high optical losses.^{Reference Khurgin and Boltasseva15} Thus, searching for less metallic plasmonic materials—or intermediate carrier density materials—holds promise. Two general approaches are employed to find such new materials: First, we can take a semiconductor and dope it sufficiently such that the properties become plasmonic in the optical range. The resulting carrier concentration must increase to such a degree that the real part of dielectric permittivity becomes negative in the visible or NIR range. Alternatively, we can approach the problem from another direction by taking a metal and “compounding” it with a nonmetal to dilute it slightly^{Reference Naik, Shalaev and Boltasseva12} (intermetallics such as transition metal nitrides, metal borides, silicides, and oxides).

Turning semiconductors into plasmonic materials: Transparent conducting oxides

Semiconductors are wonderful materials that allow a great deal of freedom in adjusting and choosing the properties of these materials (e.g., doping). The greatest challenge in using semiconductors for plasmonic and metamaterials applications is to make them metallic in the optical range. This requires that a very high carrier concentration needs to be achieved in doped semiconductors—on the order 10²¹ cm^–3, and any additional losses associated with free carrier scattering and defects must be avoided. Thus, wide bandgap semiconductors are required in order to have negligible interband transitions and a high carrier mobility to ensure low damping losses. The plasmonic operation of most semiconductors that are currently used for metamaterial demonstrations is limited to the mid-IR region.^{Reference Hoffman, Alekseyev, Howard, Franz, Wasserman, Podolskiy, Narimanov, Sivco and Gmachl16,Reference Law, Podolskiy and Wasserman17} This is because Si, Ge, and GaAs cannot be doped to sufficiently high levels for them to become metallic in the optical range.

The class of materials that has recently attracted a great deal of interest in the plasmonics community is transparent conducting oxides (TCOs), such as the well-known indium tin oxide (ITO). It was found by several research groups that TCOs can achieve carrier concentrations sufficient for them to become plasmonic in the NIR range (see Reference 12 and references therein).

My group has spent about three years researching ZnO, which is a wide bandgap semiconductor that is widely used for display panels. The challenge here is once again to dope it sufficiently high, such that it becomes plasmonic in the optical range.^{Reference Naik, Shalaev and Boltasseva12,Reference Naik, Liu, Kildishev, Shalaev and Boltasseva18–Reference Kim, Naik, Emani, Guler and Boltasseva20} Figure 2a shows a comparison among three different TCOs after optimization: ITO, ZnO doped with Ga, and ZnO doped with Al. The real part of the dielectric permittivity crosses the zero axis at ≤1.5 μm, which is an important telecommunication wavelength used in fiber-optic communication systems, enabling technology for our modern Information Age. At this particular point, the material turns from dielectric to metallic. When it becomes metallic with a small but negative real part of the dielectric permittivity, it can be used for making an epsilon-near-zero or a negatively refracting metamaterial that operates in the NIR.^{Reference Naik, Liu, Kildishev, Shalaev and Boltasseva18}Figure 2b shows the imaginary parts of the dielectric permittivities—or material loss—in these materials compared to those in Au and Ag. It is clear that the imaginary part of the dielectric permittivity for the transparent conducting oxides in the NIR region (where they become metallic) is several times smaller than that for Ag, which has the smallest material loss among metals. An important step toward utilizing transparent conducting oxides in plasmonic and metamaterial applications is to bring the technology for growing and patterning these materials to the same level as for metals. Recent demonstrations have shown that we can control the properties of TCOs, and grow, deposit (also on 3D structures), and nano-pattern them (see inset in Figure 2a) according to a specific design.^{Reference Naik, Kim and Boltasseva19–Reference Frölich and Wegener21}

Figure 2.

(a) Real (ε′) and (b) imaginary (ε″) parts of the dielectric function of transparent conducting oxide (TCO) films: Al-doped ZnO, AZO (Al concentration 2 wt%), Ga-doped ZnO, GZO (Ga concentration 6 wt%), and indium tin oxide, ITO (Sn concentration 10 wt%), deposited using pulsed laser deposition. The deposition conditions were optimized to produce the lowest losses and the highest plasma frequency. Inset: 54° tilted scanning electron microscope image of an array of GZO nanodisks with a mean diameter, D = 500 nm, and height, h = 270 nm.

One of the important properties expected from a new plasmonic material is the ability to tune its optical properties to provide modulation/switching capability. For TCO structures, the spectral position of the localized SP resonance can be changed not only by changing the size and shape of the nanoparticle (as normally done in plasmonics), but also by changing the doping concentration^{Reference Frölich and Wegener21} by applying temperature, through utilizing electrochemical methods,^{Reference Garcia, Buonsanti, Runnerstrom, Mendelsberg, Llordes, Anders, Richardson and Milliron22} or through using other post-processing means.^{Reference Li, Guo, Zhang, Zhou, Odom, Seideman, Ketterson and Chang23} Thus, in contrast to conventional plasmonic materials where the optical properties are not configurable, in the case of TCOs, the position of the SP resonance can be chosen both by adjusting the material properties and by means of annealing or other post-processing techniques.

“Dilute” metals: Transition metal nitrides

Another way of finding materials with smaller free carrier concentrations than noble metals is to consider various classes of ceramic compounds where metal is mixed with a nonmetal. These include silicides, borides, nitrides, and other compounds. One extremely interesting, and so far most promising, example of such ceramics is titanium nitride (TiN). TiN looks like gold—the beautiful domes of Russian churches were originally coated with gold, but recently, most of them are shining because of TiN coatings. Although TiN has a slightly different tint than gold, the golden appearance suggests that the optical properties of TiN should resemble those of gold (Figure 3). TiN has a number of other useful properties: It is inexpensive, extremely hard (drill bits are made of TiN), and biocompatible—an important fact for applications in sensing and therapeutics.^{Reference Lal, Link and Halas5} TiN is also a CMOS compatible material, and it is currently used in CMOS production lines as a gate metal and as a barrier layer. In addition, unlike noble metals, TiN has a very high melting temperature, which makes it highly suitable for applications such as thermal radiation engineering and thermophotovoltaics.

Figure 3.

Optical image of thin films of metal nitrides deposited on centimeter-sized glass substrates. The films were deposited by DC reactive sputtering. The thickness of films is in the range of 100 to 120 nm. The metallic luster indicates the possibility of plasmonic behavior in the visible spectrum. The color of the films correlates with the plasma frequency or carrier concentration in these films. Reprinted with permission from Reference 12. © 2013 Wiley.

Figure 4 compares the optical properties of alternative plasmonic materials with those of conventional metals. The real part of the dielectric permittivity of TiN is small compared to that of Au and Ag. TiN is metallic, but it behaves as a diluted metal in comparison to the noble metals, while its material losses are comparable to that of gold.^{Reference Naik, Kim and Boltasseva19} In addition to TiN, we have also examined other transition metal nitrides, such as ZrN and HfN (Figures 3 and 4), and found that they also exhibit plasmonic properties.^{Reference Naik, Shalaev and Boltasseva12} Transition metal nitrides can be made nonstoichiometrically, so their properties can be adjusted/tuned depending on the deposition conditions, thus allowing some tailoring of the dielectric permittivity for a specific application.

Figure 4.

Comparison of optical properties of alternative plasmonic materials with those of conventional metals.^{Reference Naik, Kim and Boltasseva19} Optical constants of low loss thin films of TiN, ZrN, and transparent conducting oxides (TCOs) (AZO, Al-doped ZnO; GZO, Ga-doped ZnO; and ITO, indium tin oxide) are plotted along with those of gold and silver. The arrows show the wavelength ranges in which nitrides and TCOs are respectively metallic. (a) TCOs and nitrides have smaller negative permittivity values than those of metals, while (b) compares material losses and shows that losses in TCOs are many times smaller than those in either gold or silver. The losses in nitrides are slightly higher than metals due to interband transitions at the crossover. Note: ε′, real part of the dielectric function; ε″, imaginary part of the dielectric function.

Performance of plasmonic and metamaterials structures: What is the best material?

In a practical device, we need to optimize the functionality expected from a particular design of a plasmonic or metamaterial structure. Even though the imaginary part of the dielectric permittivity is an important indicator of the resultant structure characteristics, it does not necessarily define the device performance. In each particular case, we must define a so-called figure of merit (FOM) to characterize the performance of the device.^{Reference Naik, Shalaev and Boltasseva12} For example, in the case of plasmonic waveguides, this might be the mode confinement or the propagation length, whereas in the case of plasmonic nanoparticles, it might be the field enhancement. FOMs defined for different structures include both the real and imaginary part of the dielectric permittivity. In each case, based on the metric and the operational wavelengths region, one can begin to define the material candidates to enable the best device performance.

For example, if we look for a SP resonance that occurs in a metallic nanosphere, a metric that provides a good estimate of the performance of this nanoparticle would be the field enhancement that can be obtained. A simple approximation comparing the ratio of the enhanced field intensity on the surface of a spherical nanoparticle to the intensity of the incident field is shown in Figure 5a for nanoparticles made of different materials at two different wavelengths, 500 and 1700 nm (selected for illustrative purposes) (based on data reported in Reference 12). An important thing to note here is that localized SP resonances will occur at different wavelengths for different materials.^{Reference Naik, Shalaev and Boltasseva12} Thus, to achieve a localized SP resonance using noble metals in the NIR, more complex nanoparticle geometries need to be developed since the resonance for simple spherical particles will be located in the visible range.^{Reference Lal, Link and Halas5,Reference Naik, Shalaev and Boltasseva12} Changing to a new material such as a TCO would solve the problem easily. The same order of magnitude is obtained for the enhanced field, and it is much easier to realize: TCOs could enable localized SP resonance applications in the NIR without the need for complex geometries such as core–shell structures. We also examined the optical performance of the nanospheres made of transitional metal nitrides (Figure 5a). Titanium nitride and zirconium nitride have performances similar to that of Au and exhibit localized SP resonance modes in the visible part of the spectrum. In particular, titanium nitride could be of significant interest for biological applications, because TiN is biocompatible, and its SP resonance mode lies in the biological transparency window (700–1000 nm) (also known as the therapeutic window; it defines the range of wavelengths where light has its maximum depth of penetration into tissue).^{Reference Naik, Shalaev and Boltasseva12}

Figure 5.

(a) Comparison of the figures of merit (FOMs) for localized surface plasmon resonance (LSPR): the maximum field intensity enhancement on the surface of a spherical nanoparticle (inset) of the plasmonic material computed from quasistatic approximations^{Reference Naik, Shalaev and Boltasseva12} at two wavelengths, 500 and 1700 nm. The nanoparticles are assumed to be in a host of refractive index 1.33. The same order of magnitude (or even higher) can be obtained for the enhanced field in new plasmonic materials compared to noble metals. (b) The FOM of hyperbolic metamaterials (HMMs) (inset), defined as the ratio of the real (Re(k _z)) and imaginary (Im(k _z)) parts of the wave vector perpendicular to the layers, along the z direction^{Reference Hoffman, Alekseyev, Howard, Franz, Wasserman, Podolskiy, Narimanov, Sivco and Gmachl16} for two wavelengths, 500 and 1700 nm. The calculations are for planar alternating layers of metal/dielectric layers formed by different pairs of materials. In the optical range, alternative plasmonic materials clearly outperform noble metals by orders of magnitude. Note: I _max, maximum field intensity on the surface of a spherical nanoparticle; I ₀, intensity of the incident field; E ₀, incident field; ITO, indium tin oxide; GZO, Ga-doped ZnO; AZO, Al-doped ZnO. Columns are color graded for illustrative purposes.

The second example is a special class of optical metamaterials called “hyperbolic metamaterials.”^{Reference Engheta10,Reference Podolskiy and Narimanov11} Simply stated, these materials are highly anisotropic—they are dielectric in one direction and metallic in the other. Such extreme anisotropy means that the dispersion of light propagating inside this material becomes hyperbolic, which leads to a range of very interesting effects such as negative refraction and^{Reference Hoffman, Alekseyev, Howard, Franz, Wasserman, Podolskiy, Narimanov, Sivco and Gmachl16,Reference Naik, Liu, Kildishev, Shalaev and Boltasseva18} subwavelength resolution^{Reference Engheta10,Reference Podolskiy and Narimanov11} and could enable novel applications in nanophotonics and thermal radiation engineering. In order to realize such materials, we normally either create a material by making alternating layers of metal and dielectric or by making metallic nanowires in a dielectric host. Both of these approaches can produce hyperbolic media. However, to target practical applications, planar devices would be required. Thus, we consider hyperbolic metamaterials made as a layered medium.

The recent discovery of the enhancement of the photonic density of states (PDOS) within a broad bandwidth in hyperbolic metamaterials could revolutionize PDOS engineering,^{Reference Jacob, Kim, Naik, Narimanov, Boltasseva and Shalaev24} potentially enabling light sources with dramatically increased photon extraction efficiency and possibly leading to non-resonant single photon sources.^{Reference Lounis and Orrit25} These metamaterials can be combined with wide-spectrum, room-temperature quantum emitters, such as quantum dots, to provide greatly enhanced spontaneous emission rates. This application of metamaterials^{Reference Jacob, Kim, Naik, Narimanov, Boltasseva and Shalaev24} can also transform an isotropic emission profile into a directional one, which could lead to new types of light sources. However, metals with their large negative permittivity and high optical losses limit the metamaterial performance. Moreover, in order to achieve high PDOS, the individual metamaterial layers need to be as thin as possible.^{Reference Jacob, Smolyaninov and Narimanov26} Noble metal films cannot be patterned into ultra-thin layers without compromising their quality. Thus, realization of multilayered metamaterials with deep subwavelength layers and good performance based on metals is extremely challenging.

To conduct a comparative study of different materials for hyperbolic metamaterials applications, a performance metric must, once again, be defined. For applications in which light is transmitted through the structure (e.g., potential concepts of new types of light sources), a FOM, defined as the ratio of the real and imaginary parts of the wave vector perpendicular to the layers,^{Reference Hoffman, Alekseyev, Howard, Franz, Wasserman, Podolskiy, Narimanov, Sivco and Gmachl16} is a good indicator of the performance (Figure 5b). In Figure 5b, the FOM is plotted for various superlattices with metal filling fractions of 50% for two wavelengths, 500 and 1700 nm (for FOM curves in the whole wavelength range where hyperbolic dispersion exists, see Reference 12). The dielectrics used with the plasmonic materials are based on ease of fabrication, and, for simplicity, their index is assumed to be constant. Thus, TCOs are paired with ZnO with an index of 2.0; titanium and zirconium nitrides are paired with aluminum nitride (index 1.70); and silver and gold are paired with a commonly used dielectric material, aluminum oxide (index 1.78). In the visible and NIR ranges, alternative plasmonic materials clearly outperform noble metals by orders of magnitude. However, this conclusion holds only for hyperbolic metamaterials in a planar geometry, and alternative plasmonic materials may not outperform noble metals for other geometries.

We have previously demonstrated the first metal-free hyperbolic metamaterial that exhibits negative refraction in the NIR region by replacing conventional metal layers with highly doped ZnO and combining with dielectric ZnO layers.^{Reference Naik, Liu, Kildishev, Shalaev and Boltasseva18} Recently, we have pursued the transition to a completely new concept of creating layered metamaterials. This can be compared to heterostructure growth used to make semiconductor devices. Instead of creating metamaterials by assembling layers by usual means—depositing polycrystalline layers of metal and dielectric—now, for the first time, a layered metamaterial as a superlattice (Figure 6a) can be grown. This means that we now have the long-awaited opportunity to create truly subwavelength layers that are perfectly smooth and can be made as thin as a few nanometers. By selecting TiN as a metallic component in this metamaterial and choosing a dielectric with exactly the same lattice constant as TiN (such as (Al, Sc)N), growing the entire metamaterial as a heterogeneous nitride superlattice is possible.

Figure 6.

(a) Schematic of a layered metamaterial with subwavelength layers grown as a superlattice (gray and black spheres schematically represent atoms of Ti and N, respectively, while light and dark blue spheres represent atoms of Al or Sc and N, respectively); (b) a high-resolution transmission electron microscope image of the interface (blue arrow) between TiN and Al_0.6Sc_0.4N layers in the superlattice together with the fast Fourier transforms corresponding to the diffractograms on either sides of the interface (TiN and Al_0.6Sc_0.4N layers).^{Reference Naik, Saha, Liu, Saber, Stach, Irudayaraj, Sands, Shalaev and Boltasseva27} Sharp interfaces and perfect lattice matching are evident.

In our experiments, we developed an approach to achieve lattice match by alloying AlN with 40 at.% ScN, which allows stabilization of this dielectric material in the cubic rock salt phase required for epitaxial superlattice growth^{Reference Naik, Saha, Liu, Saber, Stach, Irudayaraj, Sands, Shalaev and Boltasseva27} (Figure 6b). A high-resolution transmission electron microscopy (TEM) image of the interface of TiN and Al_0.6Sc_0.4N layers is shown in Figure 6b. The figure also shows electron diffraction patterns recorded from the TiN and Al_0.6Sc_0.4N layers, demonstrating perfect lattice matching. The superlattice approach gives the possibility to grow metamaterial layers with unit cells as small as 2 nm and with extremely sharp interfaces. Titanium nitride as the plasmonic material enables the long-awaited leap from previously demonstrated hyperbolic metamaterials based on thick, polycrystalline layers of metals and dielectrics to high-performance, truly binary superlattice metamaterials consisting of ultra-thin, smooth, epitaxial layers that could unlock the full range of unusual properties of hyperbolic metamaterials.

We are currently using superlattice hyperbolic metamaterials to study enhancement in the PDOS by investigating spontaneous emission (radiative decay rate or lifetime) changes of emitters placed close to the metamaterial. TiN-based superlattice metamaterials are expected to provide a higher PDOS enhancement compared to their metal-based counterparts, opening up remarkable possibilities for the realization of practical hyperbolic metamaterial devices that are CMOS- and bio-compatible, thermally stable, and controllable properties and have low loss.

A more important consequence of our studies is that TiN could boost the performance in many more classes of plasmonic and metamaterial devices. The developed approach can be applied to other areas of plasmonics and optical metamaterials and enable the first CMOS-compatible plasmonic interconnects and circuits as well as improved imaging, sensing, light harvesting, and medical devices. Such applicability is based on biocompatibility as well as chemical and thermal stability of titanium nitride. High-performance, low-loss TiN-based metamaterials could also lead to a new generation of devices for quantum optical technologies.