Towards nano-scale photonics with micro-scale photons: the opportunities and challenges of mid-infrared plasmonics

1.1 Localized surface plasmons

For simplicity, consider a spherical metallic nanoparticle (of diameter a) embedded in a dielectric medium of ε_d and excited by a homogeneous electromagnetic wave of amplitude E_o, as shown schematically in Figure 2A. Since the size of the particle is much smaller than the wavelength (a<<λ), the problem of electromagnetic field distribution inside and in the vicinity of the nanoparticle can be solved in the quasistatic approximation, where the time-varying electromagnetic field exciting the particle is considered to be constant in space, on a length scale larger than the nanoparticle. In this limit, it can be shown that the field inside the particle is homogeneous and is related to the incident field via [16].

Figure 2

Schematics of (A) localized surface plasmon (LSP) and (B) surface plasmon polariton (SPP) modes at metal dielectric interfaces, showing both the charge density oscillations in the metal and the electric fields in the surrounding dielectric.

As can be clearly seen from Eq. (4), the induced field inside the metallic sphere (and therefore the field in the immediate vicinity of this particle) is strongly enhanced when ε_m→-2ε_d. This field enhancement is primarily limited by absorption inside the metallic particle and also by the frequency detuning from the resonant condition. Enhancement of local fields can be used in a number of sensing applications as well as solar cell technology [17–19]. At the same time, the strong inhomogeneity of the electric field promotes a stronger enhancement of higher moments of the field and can be utilized for nonlinear optics applications [20, 21]. For the majority of noble metals, the resonant condition for SP excitation in spherical particles happens in the near-UV frequency range.

It is critical to note that, being a quasi-static excitation, the LSP resonance does not depend on the size of the particle, a. It can, however, be controlled by adjusting the shape of the particle and by varying the dielectric environment around the particle. These two mechanisms can be utilized to tune the SP resonance to some degree, with the former shown in Figure 3 [22]. In practice, in isolated nanostructures, such tuning is effective across the visible and near-IR frequency ranges [23–27]. A wider spectral tuning can be obtained in complex composite systems, such as fractals or percolation films [21].

Figure 3

Spectral location of the LSP resonance for subwavelength metallic nano-rod particles as a function of increasing aspect ratio; dots correspond to experimental data, line corresponds to quasi-static calculations. Deviation between quasi-static theory and experiments results from antenna-like resonances in longer nanoellipsoids. Copyright (2005), from ref. [22], with permission of the author.

At mid-IR frequencies, where |ε_m|→∞, the field inside plasmonic nanoparticles is vanishingly small, reflecting the fact that at these frequencies the properties of noble metals resemble those of perfect conductors. This effectively precludes subwavelength metallic particles from supporting such localized surface plasmon (LSP) excitations in the mid-IR. However, the visible-photonics of noble metals can be successfully mimicked in mid-IR “designer metals” fabricated, for example, by highly doped semiconductors, an effect discussed at greater length later in this article.

1.2 Surface plasmon polaritons

Consider now the interface between a metal with permittivity ε_m (occupying the half-space z<0) and a dielectric with permittivity ε_d (occupying the half-space z>0), shown in Figure 2B. Under some conditions, such an interface can support a guided mode propagating along the interface and exponentially decaying into both the metal and dielectric half-spaces. The properties of such a wave can be derived by considering the boundary conditions of Maxwell’s equations. Assuming that the mode propagates along the x direction, the field of such a wave can be derived based on the magnetic component [16]:

with

Continuity of the x component of the electric field

Equation (6) represents dispersion of an interface-guided mode, most often referred to as a surface plasmon polariton (SPP), propagating along the metal-dielectric boundary. It can be seen that the solution to this dispersion equation corresponding to a propagating mode

Due to their exponential decay inside the dielectric, SPPs are often heralded as highly confined modes, potentially enabling sub-wavelength manipulation of light. Quantitatively, the light confinement can be characterized by the characteristic decay length of light into the dielectric,

where λ_o is the free-space wavelength and the ratio n_SPP=k_xc/ω represents the effective refractive index of the SPP mode. It is clearly seen in Figure 4A that the strong subwavelength confinement can be achieved only in the vicinity of the SPP resonance, where |ε_m|→ε_d so that n_SPP>>1. For noble metals, these conditions are only satisfied at visible (or near-IR) frequencies.

Figure 4

(A) Calculated penetration of SPP mode into dielectric (normalized to free-space wavelength λ_o) for an SPP at a Ag/air interface. (B) Calculated propagation length (normalized to free-space wavelength λ_o) of an SPP at Ag/air interface. Note the large penetration depth, and propagation length, for SPPs at mid-IR wavelengths. Permittivity of Ag is taken from Figure 1.

At mid-IR wavelengths, conventional SPPs decay over macroscopic distances, moving away from the metal-dielectric interface, and effectively become surface-current guided modes propagating along the metal-dielectric interface with the speed of light in the dielectric. While such modes can propagate for very long distances (>>λ_o, Figure 4B), due to their weak penetration into the lossy metal, the large decay lengths into the dielectric (Figure 4A) lessens their utility for applications benefiting from strong confinement of the electric field.

1.3 Spoof plasmons

This drastic difference between SPPs in the visible and mid-IR frequency domains has fueled research into alternative means to confine propagating radiation to subwavelength length scales at mid-IR frequencies. In contrast to localized surface plasmon (LSP) resonances, where such efforts are effectively limited to choosing an alternative material platform (also a potential solution for enhanced mode confinement of SPPs), the behavior of SPPs can additionally be mimicked in the mid-IR frequency range using structured perfectly-conducting systems via spoof plasmons [28–30].

Originally introduced in [28], the spoof SPP is a collective excitation of the waveguide modes propagating in perforated films and free-space radiation. An example of such a structure is shown in Figure 5. When the surface of the structure is a perfect metal, each corrugation of the system can be considered to behave as a (subwavelength) waveguide with the core characterized by permittivity ε_h and permeability μ_h. Therefore, the field inside such a corrugation can be represented as a linear combination of the waveguide modes. In the limit when a <<λ₀ the contribution from the fundamental mode becomes dominant.

Figure 5

Example of a THz spoof SPP structure. (A) Schematic of metal structure with subwavelength grooves of width a, depth h, and periodicity d, which can be approximated as an effective medium with ε_y, ε_z, and μ_x. (C) Real (red) and imaginary (blue) dispersion relation (k vs f) for a system with a/d=0.3, a=3 μm, and h=30 μm. Black dashed line shows dispersion for a perfect conductor. Copyright (2010), Springer, from ref. [31], with permission of the author.

In the same limit, the collective response of the multiple corrugations can be approximated by effective medium theory, which yields the plasma-like response with effective plasma frequency [28]

The corrections to the effective medium theory due to the finite conductivity of noble metals, important for the mid-IR frequency range have been developed in Ref. [31]. As expected, these corrections result in absorption losses, somewhat similar to the propagation losses in visible-frequency SPPs. As noted in [31], however, the quality factors of spoof plasmons are typically smaller than those of their visible-frequency counterparts. While the periodicity of the spoof plasmon structures (and the associated propagating modes) can be significantly subwavelength, accurately mimicking short wavelength SPPs, it should be noted that the grooves themselves approximate quarter wave antennae, and as such, require that the spoof plasmon geometry be limited in size by the wavelength of the incident light, at least in one dimension. Nonetheless, spoof SPP structures offer significant design flexibility, as they can be designed across a wide range of wavelengths, in particular those in which metals have large, negative permittivity, such that the spoof mode mimics the highly confined and lossy modes of classical SPPs at short wavelengths.

1.4 Coupling to and from SPPs

Modulation, or control, of the plasmon confinement scale is useful in applications ranging from on-chip photonics to sensing to directional control of light [32–36]. However, coupling these modes to free-space radiation is essential for integration of plasmonic devices and structures with macro-scale optical components and devices.

Regardless of the nature of the SPP or spoof SPP, its effective modal index is always above that of the adjoining dielectric. This index mismatch presents an effective mismatch in the momentum spectrum of light, and thus prevents direct coupling between the guided mode and propagating plane waves in the dielectric. The coupling can be introduced [32, 33], and controlled [30, 32, 37] by either introducing a high-index layer, weakly coupled to the SPP (such as Kretchmann or Otto configurations) or by corrugating the surface of the metal.

Naturally, SPPs can propagate not only on the flat metal surfaces, but also on curved structures, such as cylinders or spheroids, etc. With some exceptions (propagating of a guided mode along the axis of an infinitely long cylinder), introduction of curvature or finite-size corrugations into the system results in the coupling between the SPPs and free-space modes, thus leading to radiation losses in SPP waves, but at the same time introducing the mechanism to couple diffraction-limited waves to confined SPP modes. Such coupling is typically strongly enhanced when the size of the SPP-supporting structure is close to an integer number of half-wavelengths of the SPP [38].

1.5 Surface plasmons and antennae

One of the most often used families of SPP-supporting structures is known as optical antennae. However, the vast majority of antenna work, until recently, has focused on the very long wavelengths used for radio communications or radar applications (λ~mm→m range). At such long wavelengths, metals can effectively be treated as perfect conductors, and plasmonic effects are negligible. Thus the plasmonic component of traditional antenna modes is rarely considered. In the visible frequencies, the permittivity of the metal is still negative, but quite small in magnitude (ε_m~ε_d) and penetration of an antenna mode into the metal can be significant, compared to the wavelength of the incident light and the dimensions of the antenna itself. At these frequencies the deviation of n_SPP from 1 can be significant, and thus the dimensions of these antennae approach those typically associated with structures supporting localized surface plasmon resonances. In this sense, antenna systems effectively bridge the divide between the quasi-static SP resonators and wavelength-scale SPP platforms [38, 39].

As shown in Figure 3, the resonant LSP frequency of sub-wavelength structures can be redshifted by increasing the aspect ratio of these structures. However, commensurate with the increasing resonant wavelength is a rapidly increasing magnitude of the negative metal permittivity. A number of works have focused on plasmonic antenna resonances in the mid-IR. However, the resonant wavelength of these structures is dependent on the structures’ dimensions and scales with increasing antenna size. In this sense, such structures do not exhibit the size-independent resonance associated with the sub-wavelength metallic nanoparticles described by Eq. (4). Though the localized modes of such structures cannot be perfectly modeled using a perfectly conducting metal, the higher aspect ratio of antennae and higher metal permittivity shifts the physics behind noble-metal based SPP-antennae [39, 40] toward their well-known RF counterparts.

2.1 Why the mid-infrared?

The accepted spectral limits of the mid-IR wavelength range vary between disciplines and applications. In this work, we will consider the mid-IR to consist of the wavelength range between 3–30 µm. Within this range, several sub-ranges are notable for a variety of technical applications. These are the mid-wave infrared (MWIR, 3–5 µm), the long-wave infrared (LWIR, 8–12 µm) and the very-long-wave infrared (VLWIR, 12–30 µm) [41]. Taken together, the entirety of the mid-IR is of vital technical importance for a range of applications, chief among them optical sensing and thermal signature detection and control. Additionally, mid-IR frequencies have been suggested for free-space optical communications [42].

The interest in mid-IR optical sensing is due to the large number of molecules with fundamental vibrational absorption resonances in this wavelength range. These molecules are of significance for numerous medical, environmental, and industrial sensing applications. Figure 6 shows the infrared absorbance for five representative molecules with strong and distinct absorption spectra in the mid-IR [43]. Each spectrum constitutes a unique spectral fingerprint for the molecule. For this reason, the mid-IR is often referred to as the ‘molecular fingerprint’ region of the optical spectrum.

Figure 6

Mid-IR absorbance for five molecules of interest for medical, environmental and industrial sensing applications. Data from the EPA spectral database [43].

In addition, the mid-IR is also home to the thermal radiation (emitted by any finite temperature object) of hot bodies across a wide range of temperatures. Planck’s Law describes the thermal radiation, as a function of wavelength and temperature, from a perfect blackbody, an object that absorbs all incident radiation. Because real world objects are not perfect blackbodies, this thermal emission is typically modified by an emissivity ε, where 0<ε(λ, T)<1, which modulates the object’s thermal emission as a function of wavelength and temperature, resulting in an adjusted blackbody spectral emittance (in W-m^-3), given as:

where T is the object’s temperature, λ is the wavelength of the emitted light, c is the speed of light in a vacuum, h is Planck’s constant, and k is the Boltzmann constant. Figure 7B shows the calculated emission spectra from a perfect blackbody for a range of temperatures. From this Figure, and quantitatively from Wien’s displacement law, it can be seen that the peak of the blackbody emission varies inversely with temperature:

Figure 7

(A) Atmospheric transmission spectra [44] and (B) blackbody spectral emittance for a range of temperatures. Dashed line traces wavelength of peak blackbody emission as given by Eq. (10). Shaded regions correspond to MWIR (blue) and LWIR (red) wavelength ranges. (C) MWIR thermal image of Blackhawk helicopter (Thermal image compliments of FLIR Systems, Inc) and (D) LWIR thermal image of the author’s (DW) research team (taken with FLIR BX320 thermal imaging camera).

where b=2.897×10^-3 K-m, is Wien’s displacement constant. The peak emission wavelength (λ_p) is in the mid-IR for temperatures ranging from approximately 200–1400 K. Moreover, the technological importance of the MWIR and LWIR bands can be seen to originate from the overlap of these bands with the 3–5 µm and 8–13 µm atmospheric transmission windows (Figure 7A) [44]. These atmospheric transmission bands act as natural bandpass filters for thermal detection or imaging applications. The MWIR is then the band associated with thermal emission from ‘hot’ objects (e.g., turbines and engines), while the LWIR is more usually associated with thermal emission from ‘room temperature’ biological objects (e.g., humans), as is evident in Figure 7C, D. Thus, for target acquisition and tracking, and infrared countermeasures, as well as night vision imaging and thermal auditing, mid-IR materials and optoelectronic devices in the MWIR and LWIR are of vital importance.

In addition to the technological importance of the mid-IR for sensing and security applications, the 3–30 µm range is of interest for a variety of fundamental scientific investigations. The longer wavelengths of the mid-IR, when compared to the visible (400–700 nm) and near-IR (700 nm–1.4 µm), allow for the investigation of materials with subwavelength features fabricated using standard lithographic techniques, or ultra-subwavelength structures fabricated using more complex and expensive patterning technologies. The additional available range in feature size, normalized to operation wavelength, afforded by the longer wavelengths of the mid-IR makes the mid-IR a valuable test-bed for the investigation of optical phenomena across the UV-IR frequency ranges.

Furthermore, operation in the mid-IR allows for integration of plasmonic and metamaterial structures with traditional semiconductor materials, transparent at these longer wavelengths. This allows for the incorporation of well-understood electronic material systems with plasmonic architectures using technologically mature fabrication techniques. Such plasmonic/semiconductor integration would open the door to hybrid optoelectronic-plasmonic devices, giving a unique level of control over both the dielectric and metal constituents of plasmonics structures and devices.

2.2 Mid-IR optoelectronics

Interest in the mid-IR has been historically driven by the aforementioned applications. However, mid-IR systems development has been hampered by the lack of an optical infrastructure commensurate with the technological importance of the mid-IR. Mid-IR counterparts for basic optical and optoelectronic components in the near-IR/VIS are either an order of magnitude more expensive or do not exist. Significant effort over the past decades has been made to ameliorate this discrepancy, with research efforts to develop mid-IR detectors, sources, and optoelectronics.

While the standard mid-IR detector is the HgCdTe (MCT) photodiode, interest in more efficient, higher operating temperature detectors has fueled research into quantum well and quantum dot infrared photodetectors, as well as a new class of Sb-based detectors, such as type-II superlattice and nBn detectors [45–53].

Perhaps no single recent development has driven interest in the mid-IR more than the initial demonstration, and subsequent rapid development, of the quantum cascade laser (QCL) [54–57], a semiconductor laser source which generates light from transitions between quantized states in the conduction band of complex multi-period quantum well (QW) heterostructures (Figure 8). QCLs provide a compact, high power, wavelength flexible, and now, commercially available [58], coherent source for numerous mid-IR applications, giving systems developers and researchers alike a powerful tool for investigating novel phenomena, materials, and structures, as well as developing new gas sensing and countermeasure systems in the mid-IR.

Figure 8

(A) Schematic of electron transport and optical transitions in a QCL, with inset showing the front facet of a fabricated QCL. (B) Light output and applied bias as a function of current (I) and temperature (T) for a QCL emitting at λ~4.6 µm, shown in (A), inset. Inset shows 80 K (red) and 300 K (purple) emission spectra for the QCL shown in (A), inset. Figures reproduced with permission from Peter Liu, Princeton University.

The long-time characterization work-horse of the mid-IR is the Fourier transform infrared (FTIR) spectrometer, an interferometry-based spectroscopy tool invaluable for characterization of the optical properties of bulk materials, thin films, and quantum heterostructures in the mid-IR. Spatially resolved (though diffraction-limited) spectral data can be obtained by coupling a mid-IR microscope (using all-reflective optics) to an FTIR [59].

However, the emergence of mid-IR optical components based upon subwavelength-geometry metamaterial and plasmonic structures requires a new class of imaging and spectroscopic tools. Key among these are near-field scanning systems which can provide spatially-resolved spectroscopic data on length scales much smaller than the wavelength of the probe light [12, 60–64].

2.3 Why mid-IR plasmonics?

Though the rapid maturation of mid-IR source technology has been impressive, the development of a commensurate optical infrastructure, necessary to support the envisioned QCL-based sensing and countermeasure systems, has been slower to develop. Mid-IR plasmonics may well prove to be an enabling technology for the development of this infrastructure. Plasmonics has played a role in the development of the QCL, from early QCL waveguides to the more recent QCL beam steering and polarization control by facet-patterning. Further development of optical and optoelectronic surface plasmon-based technologies may also aid in the future growth of the mid-IR optical infrastructure required to support QCL-based systems.

A prime motivation for plasmonics devices and structures at shorter wavelengths lies in the SPs confinement to subwavelength mode volumes, enhancing light-matter interactions for emitters and sensors, and opening the door to light-guiding via plasmonic optical interconnects. This enhanced interaction has applications in the mid-IR as well, for chemical and biological sensing as well as detector and emitter enhancement.

For sensing applications, the strong and distinct mid-IR vibrational absorption resonances of numerous molecular species make the mid-IR an important wavelength range. The ability to confine mid-IR light to nano-scale volumes with plasmonic structures has the potential to enhance the interaction between incident light and these molecules. In a similar fashion, mid-IR plasmonics structures are of interest for enhancing coupling to mid-IR detectors, improving the efficiency and responsivity of these devices. Alternatively, plasmonic structures can be integrated with QCLs, acting to shape or steer the emitted light, or to couple the emitted light to on-chip waveguiding structures.

Finally, mid-IR plasmonic structures have significant potential applications in controlling and modulating thermal emission from heated objects. Here there exists no direct analogue to shorter wavelength plasmonic structures, as the mid-IR is home to the peak blackbody emission for most all terrestrial finite temperature objects. Plasmonic materials allow for designer-engineered optical properties, with controllable absorption resonances. By designing selectively absorbing plasmonic materials, the emissivity of a thermal emitter can be controlled, allowing for the development of selective thermal emitters for the cloaking or mimicking of heated objects.

2.4 Summary

Applications for mid-IR technologies span disparate fields, from security and defense, to communication, to environmental, industrial, medical, and chemical engineering, with additional potential for fundamental scientific investigations. The rapid development of the QCL has spurred increasing interest and subsequently, research investment in the mid-IR. From these efforts, new tools, phenomena, and materials have the opportunity to form an enabling optical infrastructure. As we shall show in the following sections, mid-IR plasmonic and hybrid plasmonic/optoelectronic structures and devices have the potential to provide a core component of this optical infrastructure, enabling the development of new technologies with benefits to a range of scientific disciplines, and more importantly, to society as a whole.

3.1 Extraordinary optical transmission

The field of plasmonics, though active for arguably a century beforehand, was recently spurred by the demonstration of extraordinary optical transmission (EOT) by Ebbesen et al. in 1998 [65]. In EOT, strong transmission peaks were observed through metal films perforated with periodic arrays of subwavelength apertures (over 100% transmission, when normalized to the open area of the grating). Because the EOT was observed at frequencies corresponding to the expected coupling frequencies to SPPs on the patterned films, it was initially suggested that the EOT phenomenon was wholly a result of the excitation and subsequent re-radiation of SPPs. Though future work would show this phenomenon to be more complex [66, 67], it is widely agreed that SPPs play a major role in the EOT effect in periodic systems.

Early demonstrations of the mid-IR EOT phenomenon utilized traditional metals (Ag, Au, Ni, Cu), and resulted in strikingly similar transmission spectra to their shorter wavelength counterparts [68–72]. Figure 9 shows the EOT through a metal grating on a GaAs substrate. Here, the fundamental peak (1, 0), shows a peak transmittance of 33%, well above the grating open area (~20%), and not corrected for losses external to the plasmonic structure.

Figure 9

Normal incidence transmission spectrum for a Au mid-IR EOT grating with periodicity Λ=2.8 µm and aperture diameter D=1.4 µm, fabricated on a GaAs substrate. Peak transmission through the structure is ~33%, larger than the percentage of exposed GaAs (~20%) at the surface.

Though the far field transmission and reflection of the mid-IR EOT grating closely resembles that of a shorter wavelength structure, the near-field properties will differ significantly. As described earlier, the penetration depth of a mid-IR SPP can be well over a wavelength, and it is this weak confinement which results in the low propagation losses and long propagation lengths of the excitation. These attributes allow the SPP mode to be mapped without the need for subwavelength resolution [73, 74]. Figure 10 shows results from work where propagation of SPPs on mid-IR EOT gratings was characterized using a knife-edge experimental technique, a measurement technique possible with the longer propagation lengths of these long-wavelength SPPs.

Figure 10

(A) Experimental set-up for knife-edge measurement of propagating SPPs on a Ti/Au EOT grating fabricated on a GaAs substrate. (B) Contour plot showing intensity of transmitted/scattered light as a function of the position of the slit in (A) as a function of laser frequency (x-axis). Overlaid black line shows broadband transmission of the EOT grating. Adapted from [74].

3.2 Mid-IR sensing

Before the interest in EOT gratings and associated plasmonic geometries, there already existed a significant body of work on IR absorption enhancement from metallic surfaces. Infrared reflection absorption spectroscopy (IRAS) [75] is a long-used tool for measuring molecular absorption resonances, not reliant on plasmon-enhancement of the absorption signal (Figure 11A). Replacing the smooth metal films of IRAS with islanded metal films allows for coupling to localized excitations on the film surface, and thus normally incident absorption measurements, a technique commonly referred to as surface enhanced IR absorption (SEIRA) spectroscopy [76, 77].

Figure 11

Experimental schematics for (A) IRAS (B) reflection-mode SEIRA, (C) Kretschmann configuration SEIRA, with inset showing expanded view of metal, and (D) SPEIRA measurements.

The physical mechanism for SEIRA is similar to that of surface enhanced Raman spectroscopy (SERS), with SEIRA measurements of molecular absorption showing an absorption enhancement of 10–1000, when compared to standard, direct transmission, IR absorption measurements [78]. This additional enhancement results from the vibrational absorption of molecules in the strengthened near-fields associated with localized resonances on the islanded metallic surfaces. Figure 11B shows a schematic of a reflection SEIRA experiment. SEIRA relies on the inherent multiscale roughness of thin evaporated or electroplated metallic films in order to provide a surface with a broad spectral range of resonances. However, little effort was made to tailor the metal surface geometry to enhance absorption at specific wavelengths.

The localized resonances giving the absorption enhancement in SEIRA are often referred to, in the literature, as ‘plasmonic’, though they are perhaps closer to classical antenna resonances. Earlier work investigating SEIRA on thin metal films in the prism (Kretschmann) configuration (Figure 11C) noted that absorption enhancement in this configuration could be attributed to both localized field enhancement on the textured metal surface as well as coupling to propagating surface plasmons [79], an effect known to give enhancement of Raman signals [80]. Other groups have investigated prism-coupling purely to SPP modes for mid-IR biological sensing applications with promising results [81].

Subsequent efforts looked to use EOT gratings for detection of thin layers of molecular coatings, using so-called surface plasmon enhanced infrared absorption (SPEIRA) spectroscopy measurements [82, 83]. These works demonstrated significant enhancement in absorption (×100), when compared to previous IRAS measurements [84], attributed to the increased path length of light when coupled into propagating SPPs, which in turn, interact with the absorbing monolayers. Interestingly, the observed absorption enhancement is spectrally removed from the primary transmission (SPP) resonances of the gratings, though the authors argue that absorption is spectrally coincident with a number of broader, weaker, resonances. Further work investigated thin films of hexadecane on EOT gratings, and in particular the interaction of the CH₂ rocking absorption line with the primary EOT transmission peak. It was shown that absorption was enhanced by approximately a factor of 3 when compared to the same absorption peak for hexadecane on a ZnSe substrate [84].

A direct comparison of SERS and SEIRA (or SPEIRA) is difficult. While both rely on plasmonic enhancement (though in very different wavelength ranges), the enhancement is of fundamentally different phenomena (the nonlinear Raman shift for SERS and direct absorption for SEIRA), with different selection rules. While SERS can provide an enhancement of 10¹², compared to SEIRA’s 10–1000 enhancement, the signal being enhanced in SEIRA is much stronger than in SERS, making the resulting effectiveness of the two techniques comparable. Furthermore, the short-wavelength LSP leveraged in SERS provides strong coupling to molecules in the LSP near-field (monolayers) on metals such as Au, Ag, or Cu. The long-wavelength ‘plasmonic’ or antenna resonance used for S(P)EIRA is better able to probe thicker films, on a wider range of metals, including transition metals (which all have large negative permittivity in the mid-IR). Ultimately, many research teams use a variety of surface-enhanced sensing techniques, including both SERS and SEIRA [85].

The use of metal films capable of supporting SPPs or localized resonances for sensing in the mid-IR has clearly shown promise. However, unlike plasmon-enhanced sensing mechanisms at shorter wavelengths, which have been widely adopted as standard laboratory techniques (SERS) [86] or even commercial systems [87], mid-IR plasmonics-based sensing has not been broadly adopted outside the research lab, despite the value of the mid-IR as the molecular fingerprint region of the electromagnetic spectrum. This may be due, in part, to the large mode volume and weak confinement of the mid-IR SPP, which limits the interaction strength of the SPP mode with the molecules of interest. For this reason, there has been significant interest in developing plasmonic structures in the mid-IR which are capable of confining light to ultra-small mode volumes.

The enhanced absorption observed from SPEIRA measurements may, in some part be evidence of such an effect. The enhanced propagation lengths of the mid-IR SPPs on the EOT gratings (and thus enhanced interaction lengths with the adsorbed molecules) should be balanced by the weak overlap of the mode with the molecules of interest (due to the large dielectric penetration depth of the mid-IR SPP), limiting the achievable absorption enhancement for SPEIRA measurements. Taken together with the observed enhancement of absorption away from wavelengths corresponding to SPP-coupling on the metal mesh, these results may point to a more complex enhancement process than can be described by a simple path-length enhancement. Because the transmission process requires the coupling of light or SPP modes through the subwavelength aperture, the field profile around the grating structure is not a pure SPP mode, and can be quite strong in the aperture itself. Transmission through random apertures shows significantly weaker, but also broader, peaks than for the SPP-mediated transmission through periodic arrays of apertures [88]. Thus, it may be the field strength in the apertures that is responsible for the SPIERA effect, as well as the larger-than-expected tuning seen from voltage tunable plasmonic devices using EOT top gates in a metal-oxide-semiconductor structure [89].

Field enhancement in the mid-IR can also be achieved by use of micro-antenna-like structures. Here, again, the line between antenna and localized plasmonic resonances is somewhat blurred. Such structures do not exhibit resonances at exactly the wavelengths expected for pure antenna resonances of micro-scale structures fabricated from perfect electrical conductors (PECs) [39]. Instead, slight redshifts are observed in the structures’ resonances, resulting from the finite (negative) value of the metal permittivity. As such, these resonances are often referred to as ‘plasmonic’. However, the spectral positions of these resonances are strongly dependent on the antenna geometry (both shape and size), requiring the antenna dimensions to be on the order of the resonant wavelength. In this sense, such structures are dissimilar to the localized surface plasmon resonances (LSPR) observed in the visible and IR wavelength ranges for strongly subwavelength metallic particles, where spectral position of the resonances are shape dependent, but less directly dependent on size (assuming particle size <<λ) [90, and references therein].

Nonetheless, the benefits associated with the enhanced field strength of these structures should translate directly to similar systems with purely LSPR-like resonances. Micro-antenna arrays have been shown to strongly enhance (~10⁵) the spectral signature of proteins (shown in Figure 12) and single layers of octadecanethiol [91, 92]. In addition, metal-dielectric-metal structures in which the top metal layer is patterned into a periodic array of antennae have demonstrated distinct spectral signatures indicative of strong coupling between molecular resonances in the dielectric material and the geometric antenna resonances of the patterned top layer [93]. The above-described structures have shown potential for sensing applications, though they are limited to a minimum size of approximately λ/2.

Figure 12

(A) Scanning electron microscope image of a periodic array of nanoantenna. (B) Comparison of reflectivity for bare antenna array and antenna array coated with a thin film of silk fibroin. Copyright (2001) PNAS, from ref. [91], with permission of the author.

3.3 Detectors, sources, and waveguides

The use of plasmonic structures for improving semiconductor-based mid-IR detectors is fundamentally quite similar to SP-enhanced sensing in the mid-IR. In both cases the SP mode is used to enhance local field intensities and thus, interaction with either molecules of interest, or the device active region. Early proposals sought to use plasmonic surfaces in two slightly different ways: first, to couple to incident light and funnel this light to a subwavelength detector, for improved signal to noise ratio [94], and second, to couple light into SP/semiconductor modes, and enhance light/SPP interaction with the detector active region, using patterned top contacts are used [95].

The majority of experimental demonstrations of plasmon-enhanced mid-IR light detection have utilized the latter approach for integration with quantum dot-, quantum dots in a well-, and quantum well infrared photodetectors (QDIP, DWELL, and QWIPs, respectively) [96–99]. Figure 13 shows the results for a QDIP structure patterned with a top-contact EOT grating. In addition, recent work has demonstrated a plasmon-enhanced DWELL focal plane array camera with strongly enhanced absorption at λ=6 µm [100]. In each of the above detectors, incident light is coupled into EOT gratings above the detector structure, which simultaneously serve as top contacts for the detector device. The responsivity of the detectors is spectrally modulated by the momentum matching conditions for coupling to the grating modes, and at peak response can be significant, with enhancement factors as high as 30 reported [97].

Figure 13

Micrograph of metal photonic crystal (MPC) QDIP top contact (with expanded view) and (B) schematic of MPC-QDIP layer structure. (C) Spectral response of MPC-QDIP at 2.4 V (red) and -3 V (blue) applied bias compared to reference sample at the same biases (green, black) without the EOT top contact. (D) Normalized spectral response of MPC-QDIP structures at 4.4 and -4V and EOT periodicities of 2.5 and 3.6 µm. Copyright (2009) OSA, from ref. [91], with permission of the author.

Interestingly, one of the first uses of plasmonic materials in the mid-IR was not for detector enhancement, but as waveguiding structures in quantum cascade lasers. Traditionally, semiconductor lasers use semiconductor (dielectric) waveguides which confine the laser mode to the device active region by means of the difference in index of refraction between the core (active region) and cladding (dielectric waveguides). Such waveguides work well for the InGaAs/AlInAs QCL material system grown on InP substrates, but for GaAs/AlGaAs QCLs, the refractive index difference between laser core and cladding is small, and very thick waveguide layers are required to efficiently overlap the laser mode with the device active region (quantitatively given by Γ, the waveguide overlap factor). To overcome these problems, surface plasmon waveguides were implemented. In this case, the SPP mode is confined to the interface between the semiconductor and a gold top contact, with the first demonstration of such a laser achieving a Γ of 70% [101]. Later work demonstrated single-mode [102] and far IR [103] emission from QCLs with SP-waveguides, with Γ’s>80%. At the long wavelengths used in these lasers, the penetration depth of the SPP into the metal is relatively small, leading to low losses and high transverse-optical-mode confinement despite the high optical absorption of the metal.

Plasmonic waveguides, while widely used for THz QCLs (where the long emission wavelengths make dielectric waveguides impractical), are used in the mid-IR only for long-wavelength devices. In fact, recent work has demonstrated that unintentional coupling to SP modes the sidewalls of wet-etched, long-wavelength (λ~14 µm) QCLs is a source of waveguide loss [104].

There has recently been significant interest in intentional coupling of QCL emission to plasmonic structures, either for beam shaping, polarization control, or on-chip waveguiding. In the latter case, researchers have demonstrated the ability to couple QCL emission to SPP modes, either by use of a grating element to convert a QCL dielectric waveguide mode to a propagating SP [105] (shown in Figure 14), or by direct coupling of QCL facet emission to a planar SPP waveguide [106]. In addition to demonstrating a method for electrical injection of SPP modes, this work demonstrates the mid-IR’s utility as a test-bed for technologies at shorter wavelengths. Here, the mid-IR devices were used as a launching point for the proposal of a telecom SPP generator with potential applications for on-chip interconnect technology [107].

Figure 14

Schematics of QCL waveguides (A) without and (B) with grating coupler for efficient coupling of generated SPP modes to a flat passive SPP waveguide. AFM images of the fabricated devices (C) without and (D) with grating coupler and NSOM images of the same devices (E) without and (F) with grating coupler. In the case where a grating coupler is patterned into the metal layer, a clear SPP standing wave pattern can be observed, indicative of the interference between SPPs launched in the “generation” regions of the device and coupled via the grating into the “passive section”. Copyright (2010) Aps, from ref. [105], with permission of the author.

3.4 Beam shaping and steering

In addition to use as waveguides in, or coupled to, emitter devices, plasmonic structures in the mid-IR offer the potential for beam shaping forming applications. Early work at visible/IR wavelengths [35] demonstrated narrow-divergence beam steering from a subwavelength slit in a metal film, flanked by linear arrays of periodic corrugations. For such structures, light incident on the subwavelength slit can be directly transmitted, or alternatively, can couple to propagating SPs at the corrugated back-side of the metal film. These surface waves are then scattered from the back surface by the corrugations, and interfere in the far-field of the device structure to form the resulting beams.

In the mid-IR, such structures hold particular promise. Traditional edge-emitting QCLs emit highly divergent beams, diffraction-limited to angles θ ~ arcsin(λ/T) (with T being the thickness of the QCL active region, usually on the order of a few µm). As such, collection and collimation of the emitted light requires additional, external optics, as well as careful alignment of these optics. However, when plasmonic beam steering/shaping structures are patterned on the front facet of a QCL, the resulting beam can be narrowed (in the QCL growth direction) to values as small as ~2.4º [108, 109]. The work showing these effects sprung from earlier demonstrations of antennae patterned on QCL facets for strong near-field confinement of light at the QCL facet [110]. Similar structures fabricated on freestanding substrates showed the ability to steer incident QCL light by control of the incident light’s wavelength, or alternatively, the permittivity a dielectric material at the corrugated metal/dielectric interface [111]. Further work demonstrated that similar plasmonic facet patterning (either with a single, angled corrugation array, or with two orthogonal facet gratings) allows for control of the polarization state of light emitted from the QCL, giving angled linear, elliptical or circularly polarized emission, dependent on grating geometry, as shown in Figure 15 [112].

Figure 15

(A) Scanning electron micrographs of the facet of a QCL patterned with a plasmonic polarizer designed to give polarized emission 45° with respect to the vertical. (B) Far-field emission from the device shown in (A). (C) Measured (red circles) and calculated (black line) emission intensity as a function of polarization. Copyright (2009), AIP, from ref. [112], with permission of the author.

For these beam steering and shaping structures, the long propagation length of SPPs on metal films at longer wavelengths allow for interaction of the weakly-bound SPP mode with a large number of corrugations, giving rise to the opportunity to effectively control the evolution of the beams formed above the plasmonic grating. Careful design of non-periodic corrugations on the backside of the beam steering structure can effectively focus the transmitted light above the subwavelength slit, as shown in Figure 16 [37]. In such a structure, the surface is designed so that the light scattered from each corrugation interferes constructively at a common point above the surface, whose position can be controlled by the designer. In Figure 16, one can see both simulations and experimental results for such devices, designed to focus the transmitted light 250 and 500 µm above the central slit, respectively. In such structures, the long propagation length of the mid-IR SPP is used to the designer’s advantage and allows for the control of beam evolution in the vicinity of the plasmonic geometry.

Figure 16

(A, D) Numerical simulation of transmitted light intensity on back side of an engineering grating structure flanking a subwavelength slit (B, E) Simulation data from (A, D) smoothed to approximate experimental setup’s limited resolution. (C, F) Experimental measurement of field intensity on backside of the grating structure, for devices designed to focus 250 µm (A, B, C) and 500 µm above the sample surface. Reproduced from [37].

Recent work has also demonstrated mid-IR beam steering, of a sort, using localized resonances of V-shaped micro-antenna arrays on planar dielectric surfaces [113]. Here, the incident field interacts with the micro-antennas on the planar surface, introducing a phase shift to the transmitted light. If the antennas are designed to give equal scattering intensity and spatially-varying phase shifts, the angle of refraction for the antenna-coupled light can be controlled by the magnitude of the phase gradient along the surface. It has also been suggested that the phenomenon of beam steering by resonant gratings can be described in terms of diffraction theory [114]. The Au subwavelength antenna structures used here and in subsequent work [115] are essentially folded quarter-wave antenna, demonstrating the utility of noble-metal ‘plasmonic’ antenna structures with localized resonances for beam steering, shaping, and polarization control applications [116].

3.5 Thermal emission

Plasmonic devices such as the beam steering and shaping structures described above, the QCL and on-chip QCL-coupled waveguides, and the EOT gratings and islanded plasmonic surfaces used for sensing, must, at all wavelengths, contend with losses. These losses appear as a decaying field intensity with propagation of the SPP mode, or a spectral broadening of a localized resonance resulting from scattering or absorption, and in most cases adversely affect the desired performance of the plasmonic structure. However, the mid-IR holds a spectral position with the unique ability to link an object’s optical and thermal properties, providing the opportunity to use plasmonic structures for thermal emission/signature control applications. The thermal emission resulting from the finite temperature of any object takes the spectral form given by Planck’s law, modulated by the object’s surface emissivity (ε). Most materials have largely constant emissivity as a function of wavelength, which allows thermal imaging systems to identify the object’s temperature despite the limited spectral range of the system sensor. However, by integrating engineered absorbers on an object’s surface, one can alter the emissivity of the surface as a function of wavelength, so that when heated, the object selectively emits thermal radiation in preferred wavelength bands. These coatings can then be used either as narrow-band thermal light sources, or alternatively, can be used to change the thermal signature of an object when viewed with a thermal imaging system. As such, the losses which plague most plasmonic structures can, in the mid-IR, be harnessed to demonstrate a new class of selective thermal emitters.

As an example, a simple 1D grating structure patterned into commercially available steel (shown in Figure 17A) allows for the out-coupling of thermally excited SPPs to free space photons. Because the grating is 1D, SPPs can only out-couple in one polarization (TM-polarized), while emission in the orthogonal (TE) polarization remains blackbody-like, with ε<<1, and constant. When heated, such a structure shows a clear spectral selectivity in its thermal emission, when comparing TM to TE polarized emission, as shown in Figure 17B [117]. At resonance (λ~10 µm), SPP-enhanced, TM-polarized thermal emission is a factor of ~2.6 greater than TE-polarized emission, an effect which can be clearly observed in polarized thermal images of the sample surface (Figure 17C, D).

Figure 17

(A) Schematic and scanning electron micrograph of grating-patterned steel substrate. (B) Thermal emission of patterned steel for a variety of etch depths, compared to patterned glass. Here, the spectra shown represent the TM polarized emission (with background subtracted) normalized to the TE emission (background subtracted). (C) TE and (D) TM polarized thermal images of the selective emitter surface, taken with a FLIR BX320 thermal camera. Adapted from [117].

However, it is preferred that the absorption/emission resonance be angle-independent, which makes the use of SPP-based devices difficult, as momentum-matching is required to couple these modes to free space photons (whose momentum depends on emission angle). Thus, most examples of selective thermal emitters utilize some form of localized plasmonic, metamaterial, or photonic crystal resonance. One such example is arrays of nano-scale slots in a metal film, with depths on the order of λ/4, allowing the slots to support “organ-pipe” modes with distinct absorption and, when heated, thermal emission resonances [118]. Later work demonstrated that the spectral linewidth of the thermal emission from these structures can be significantly narrowed by tailoring the nano-slot array periodicity to efficiently outcouple the organ-pipe plasmon resonances to free-space. These engineered materials were then used as narrowband thermal light sources for sensing applications [119].

The ability to modify the emissivity of an object allows one, effectively, to adjust its apparent temperature when viewed by a thermal sensor, as evidenced in Figure 17D. While strongly selective in emission, the organ-pipe structures described above are polarization sensitive, and require numerous exacting and time-consuming fabrication steps, which are difficult to incorporate into low-cost, large area coating applications. Planar, patterned metallic structures can be used to achieve strong absorption at designed resonances. Plasmonic perfect absorbers (PAs) consisting of subwavelength metallic discs or lines separated from a ground plane by a thin dielectric layer, have been shown to give perfect absorption at telecom frequencies [120, 121]. At longer wavelengths, PAs and selective thermal emitters have been demonstrated with similar structures using metamaterial top layers [122, 123]. Alternatively, antenna-like structures can replace the lithographically challenging metamaterial resonators of the top metal layer, and perfect absorption, as well as strongly selective and angle-insensitive thermal emission can be achieved [124]. Figure 18 shows a schematic of such a structure, in addition to the experimental reflection and thermal emission spectra from the fabricated sample. Such meta-surfaces allow for strong enhancement of thermal emission (ε>0.9) at resonance, with weak emission at all other wavelengths (ε<0.2).

Figure 18

Experimental reflection and thermal emission from a mid-IR selective emitter structure, demonstrating strong spectral selectivity, with (ε>0.9) at resonance, and weak thermal emission at all other wavelengths (ε<0.2). Inset shows schematic of the device characterized.

3.6 Summary

Despite the significant progress in mid-IR selective thermal emitters, angle-insensitive and polarization-independent plasmonic perfect absorbers have yet to be demonstrated in the mid-IR. This is simply a result of the antenna-like nature of localized resonances in the mid-IR when using traditional metals. Thus the length scales required for localized resonators at these wavelengths will always be on the order of the wavelength of light. Likewise, for the many SPP-based geometries reviewed in this section, specifically the beam steering, waveguiding, and sensing structures, it can be argued that none are capable of true subwavelength mode confinement (at least in all three dimensions), due to the nature of traditional metals in the mid-IR. Reaching the nano-scale with mid-IR plasmonics therefore requires engineering the optical properties of plasmonic metals. Happily, this may not be as difficult as it might seem, as we will demonstrate in the subsequent section of this review, where we discuss how semiconductors can be engineered to behave in the mid-IR similarly to traditional metals in the visible and near-IR wavelength ranges, opening the door to bringing micron-scale wavelengths of mid-IR light down to the nano-scale.

4.1 Mid-IR graphene plasmonics

Graphene, a 2D monolayer honeycomb lattice of carbon atoms, was first experimentally demonstrated only recently [125–127], despite being an active field of theoretical investigation for decades. While much of the interest in graphene has focused on its unique electronic, mechanical, and thermal properties, there also has been growing interest in the optical properties of graphene. In particular, graphene has drawn attention as an alternative plasmonic material at long wavelengths due to its unusual band structure, high mobility, tunable carrier concentration, and the resulting predicted strong confinement and long propagation lengths of SPPs [128, 129]. A recent review of graphene plasmonics provides a comprehensive overview of the field [130], so here we limit our discussion to selected long-wavelength applications of graphene as an alternative plasmonic metal.

A combination of low dimensionality and unique band structure results in a graphene plasma frequency which scales as n^1/4, as opposed to the n^1/2 for typical 2D or bulk (3D) structures (Eq. 3). Extremely large confinement, characterized by the ratio of the free-space wavelength λ₀ to plasmon wavelength λ_p can be expected on graphene. The 2D carrier concentration of graphene can be varied over a wide range, using either electrostatic doping via a gate electrode, or by doping using molecular charge transfer [131], with carrier concentrations as high as 5×10¹³cm^-2 reported [132]. This ability to design, or electrostatically control, the 2D carrier concentration, and thus the plasma frequency, makes graphene an exciting new material for IR plasmonic applications, as well as for fundamental investigations of 2D plasmons. IR spectroscopy has proved to be a valuable tool for probing the unique bandstructure of graphene, as well as demonstrating voltage tuning of graphene’s Fermi energy [133, 134].

Geometry plays equally important role in tuning graphene’s optical response. SPPs propagating in graphene nanoribbons have been reported at terahertz frequencies [135]. By either changing the ribbon width or applying a gate voltage to the ribbons, the frequency of the SPP excitation can be tuned. Surface plasmons in graphene sheets deposited on an SiO₂/Si substrate have been imaged with scattering-type SNOM at λ₀=11.2 μm [136], revealing extremely high confinement factors (λ_o/λ_p>50), tunable by application of a gate voltage. Similar SNOM studies on tapered graphene nanostructures have imaged propagating plasmonic modes and demonstrated strong confinement (λ_o/λ_p>40) [137]. Moreover, propagation of the plasmonic mode can be effectively switched on and off by application of a gate voltage, opening intriguing possibilities for hybrid plasmonic/nano-electronic devices.

Coupling graphene and dielectric systems provides additional design elements in engineering the optical response. Localized plasmonic modes in nano-pillars with alternating graphene/insulator layers have shown plasmonic resonances in the far-infrared [138]. The spectral position of the resonance can be tuned by changing the number of layers in the stack. Though the field of graphene plasmonics is in its infancy, these initial results indicate that graphene could be a promising new material for infrared plasmonics.

4.2 Highly doped designer plasmonic metals

One of the primary benefits of graphene plasmonics is the design flexibility that comes with a material whose plasma frequency can be controlled, either by design (doping) or dynamically (electrostatic gating). Recently, significant design flexibility has also been achieved in a technologically mature class of materials: doped semiconductors.

Doped semiconductors offer an intriguing option for the development of ‘designer’ mid-IR plasmonic materials. The optical response of a doped semiconductor can be modeled as a Drude material, in the same manner as the classical metals described by Eq. (3),

where γ is the scattering rate for electrons in the semiconductor and ε_S is the undoped semiconductor’s relative permittivity. Here, the plasma frequency ω_p differs from that of Eq. (3) in two key aspects. First, the carrier density n (and charge polarity) can be controlled either during growth or during fabrication, by choice of dopant (n-type or p-type) and doping density. Second, the electron (hole) mass used in Eq. (3) must be replaced by a frequency dependent effective mass m^*(ω) which reflects the effects of the semiconductor’s bandstructure. Here, in contrast to the noble metal expression for ω_p, we normalize the ω_p of Eq. 11 with respect to the background permittivity, such that Re{ε(ω)}=0 for ω=ω_p. A combination of high doping and small effective masses can lead to mid-IR plasma frequencies for III/V and group IV semiconductors.

The concept of designer plasmonic metals is not unique to semiconductors. Recent work has demonstrated transparent conducting oxides (TCOs), such as indium tin oxide (ITO), and nitrides with plasma frequencies in the near-IR to visible region, including the technologically important telecom wavelength range [1]. TCOs can be grown using a variety of techniques, including sputtering and pulsed laser deposition (PLD). In PLD, the dopant concentration is controlled by ablating two separate targets, for example In₂O₃ and SnO₂ for ITO, with the appropriate number of pulses on each target. TCOs show significantly smaller losses in these wavelength ranges than the traditional plasmonic materials of gold and silver, as evidenced by their smaller imaginary dielectric constant shown in Figure 19B.

Figure 19

Real (A) and imaginary (B) part of the dielectric constant for nitride and TCO films as compared to gold and silver. Copyright (2011), OSA, from ref. [1], with permission of the author.

Typical doped TCOs demonstrate a plasma wavelength in the 1.5 μm range, which would result in extremely good confinement of plasmonic modes in the near IR, as opposed to the rather poor confinement seen in gold and silver. The plasma frequency of the TCOs can be tuned across a range of microns (a significant tuning range, given the short wavelengths of operation) by changing the doping of the material during growth [1].

Similarly impressive results have been demonstrated with nitride-based materials. Transition metal nitrides can be deposited by sputtering a pure metal target with a nitrogen-argon gas mixture. The film properties can be controlled by tuning the nitrogen:argon ratio. Though nitrides have higher losses than gold and silver, they are also easier to process using standard semiconductor techniques and can be grown epitaxially on substrates such as sapphire. This leads to lower losses from surface roughness in nanostructures than are seen in gold or silver, making TiN a good alternative to gold for structures such as metal-insulator-metal waveguides and sensors utilizing localized surface plasmon resonances [2]. In addition, extremely thin films of TiN grow continuously, unlike the percolating islands seen in gold films with thicknesses >10 nm. Smooth, non-islanded, films of these thicknesses are necessary for transformation optics applications, such as hyperbolic metamaterials, recently demonstrated with TCOs [139].

Traditional semiconductors provide a powerful potential material system for mid-IR designer metals with plasma wavelengths in the 3–30 µm range. Semiconductors have a number of clear advantages when compared to both traditional metals and the newer class of conducting oxide and nitride materials systems. First, epitaxial growth of semiconductors allows for single-crystal quality materials, atomic-scale precision in layer thicknesses, and accurate control of doping concentrations. Other semiconductor doping mechanisms, such ion-implantation or thermal diffusion of dopants do not offer similar precision in the vertical direction, but do offer the opportunity for lateral patterning of dopants. Moreover, traditional semiconductors are well understood material systems with an extensive existing processing and fabrication infrastructure. As a rule, these materials exhibit higher mobility (and thus smaller scattering rate) than their noble metal or transparent oxide counterparts [140]. Finally, the use of doped semiconductor designer metals allows for the potential integration of plasmonic structures with traditional semiconductor electronic and optoelectronic devices.

Doped semiconductors such as silicon and germanium, were among the first to be proposed as candidates for mid-IR plasmonic devices [141, 142]. Silicon has the obvious materials advantage of being perhaps the most well-understood of the two, with an extensive fabrication infrastructure available. Mid-IR plasmonic behavior has been realized both in ion-implanted silicon [143] as well as commercially-doped silicon wafers [144]. As shown in Figure 20, which gives ε of the doped Si for each of these two works, the plasma wavelength of the doped films can be tuned from approximately 4 μm to 11 μm using both n- and p-type material.

Figure 20

(A) Real (solid) and imaginary (dashed) parts of the permittivity for n-type (blue) and p-type (red) silicon. (B) Real and imaginary parts of the permittivity for silicon wafers doped n-type at 1×1020 cm^-3 (red) and 6×10¹⁹ cm^-3 (black). The use of doped Si allows for control of the material ω_p across a wide wavelength range. Copyright (2011), AIP, from refs. [143, 144], with permission of the authors.

Designer metals in the mid-IR have also been shown in III/V semiconductor materials. III/V materials have distinct advantages when compared to Si. First, they possess both the potential for extremely high doping concentrations, as well as small electron effective masses, leading to remarkable wavelength flexibility across the entire mid-IR wavelength range. Furthermore, the III-Vs have the advantage of bottom up growth by well-established epitaxial techniques such as molecular beam epitaxy (MBE) or metalorganic chemical vapor deposition (MOCVD). Use of MBE or MOCVD for growth of doped semiconductors allows for single-crystal growth, precision control of layer thicknesses, and potential integration of semiconductor metals with the complex heterostructures and nanostructures used for cutting edge optoelectronic devices such as quantum cascade lasers or quantum well, quantum dot, or type-II superlattice infrared photodetectors.

Doped InAs in particular, is an excellent candidate for a mid-IR plasmonic material [145]. It has an extremely small effective mass and can be doped up to 1×10²⁰ cm^-3[146, 147]. Experimentally, the plasma wavelength can be tuned from 5.5 μm to 17 μm (or longer), as shown in Figure 21A for InAs films grown with various doping and growth parameters. Interestingly, the narrow band-gap of InAs, which would normally give strong absorption at shorter wavelengths in the mid-IR for these materials, shifts to frequencies in the near-IR due to conduction band state filling (Burstein-Moss effect), making these designer metals transparent out to telecom wavelengths [148, 149].

Figure 21

(A) Real and (B) imaginary parts of the permittivity of heavily doped InAs films with various carrier concentrations, layer structures, and growth parameters.

These films can also show low losses, as indicated in Figure 21B. This figure plots the imaginary part of the dielectric constant as a function of frequency normalized to the plasma frequency, and allows materials with very different plasma frequencies to be compared. The best highly-doped InAs films have losses at their plasma frequency comparable to the TCOs. A comparison to gold and silver, at least near the plasma frequency, is difficult due to the large interband absorption seen in these metals near the plasma frequency, which distorts the imaginary components of the metal permittivity away from a simple Drude model.

The development of mid-IR metals capable of mimicking the optical properties of classical metals at shorter wavelengths opens the door to a new class of mid-IR plasmonic structures, where truly subwavelength optical mode confinement can be achieved. As an example, subwavelength structures from traditional plasmonic metals can only support the LSP resonances discussed earlier when |ε_m|~ε_d, limiting the LSP resonance frequency to shorter wavelengths. However, when such structures are fabricated with ‘metals’ designed to have |ε_m|~ε_d in the mid-IR, LSP resonances can be supported at longer wavelengths. Subwavelength puck structures fabricated from highly doped InAs films show a just such a resonance (Figure 22), in agreement with both numerical simulations, as well as an analytical model of the puck structure as a subwavelength metallic nanoparticle [3]. As would be expected, and as has been demonstrated for metallic nanoparticles at shorter wavelengths, the spectral location of the resonance does not depend on the size of the puck structure, indicating that the observed resonance is, in fact, a true LSP resonance.

Figure 22

Experimental transmission, simulated absorption, and modeled absorption for 1.7 μm-diameter pucks of heavily-doped InAs, showing a LSP resonance at 9 µm. Figure adapted from [3].

4.3 Designer metal ENZ- and meta-materials

The ability to grow plasmonic materials with designed optical properties opens the door to a wide range of materials and structures demonstrating unique optical phenomena. Key among these are epsilon near zero (ENZ) materials and hyperbolic metamaterials. The unique properties of ideal ENZ materials, from guiding fields around cloaked space to squeezing light through deep subwavelength openings, have their origin in the ability of this media to drastically increase the local wavelength of light [150–152]. The increase of the local wavelength causes the electric field to be homogeneous over areas that can be much larger than the vacuum wavelength of incident light. As a consequence, electromagnetic energy can relatively easily bend around the geometric obstacles formed by “conventional” materials in the ENZ matrix. In particular, such energy “tunneling” leads to the dramatic increase of light transmission through subwavelength openings filled by the ENZ material.

As a demonstration of this effect, sub-wavelength apertures were formed in Au over a highly doped InAsSb layer λ_ENZ~8 µm [4]. At ENZ, for TM polarized light (which is the only polarization which can couple to the waveguide formed by the subwavelength Au slit), a strong enhancement of transmitted light is observed, as shown in Figure 23B. In fact, when compared to transmission through undoped GaAs (Figure 23A), transmission is actually enhanced, indicating that the funneling effect is large enough to overcome the significant losses associated with the large carrier concentrations in the InAsSb.

Figure 23

Transmission of TM (red) and TE (blue) light through (A) a 1.0 μm slit fabricated in a Au layer on an undoped GaAs substrate and (B) a 0.9 μm metal slit fabricated in a Au layer on an epitaxially grown heavily-doped InAsSb ENZ layer. Note the strong transmission peak at λ_ENZ~8 µm for the TM polarized light in (B). Schematic representations of light transmission through a 1 µm gold slit on GaAs (top) and InAsSb (bottom) are shown between (A) and (B).

Highly doped semiconductor layers have also been utilized as plasmonic layers in optically thick hyperbolic metamaterial structures consisting of alternating layers of undoped and highly doped AlInAs/GaInAs. As mentioned earlier, the use of doped semiconductors as plasmonic materials allows for the integration of such material with traditional semiconductors. If grown by MBE (as is the case for the metamaterial work), precise control of layer thicknesses allows for exact control of the metal layers, while control of dopant incorporation allows for control of the metals’ optical properties.

The interleaved layers of doped In_0.53Ga_0.47As and undoped In_0.52Al_0.48As, lattice matched to an InP substrate, demonstrated negative refraction for TM-polarized light [5]. For these samples, the onset of negative refraction occurs at the plasma frequency (8.8 μm) and extends to near 11.8 μm. Negative refraction was demonstrated by shining light on the metamaterial while blocking part of the transmitted light with a blade. When the sample exhibits normal refraction, the light is bent away from the blade and more transmitted light is seen. When the sample exhibits negative refraction, the transmitted light is bent toward the blade, resulting in a reduction in the transmitted light intensity, shown in Figure 24. Observation of negative refraction validates the hyperbolic dispersion in these metamaterials, which could potentially be utilized for subwavelength control of radiation and for emission engineering [5, 153–158].

Figure 24

Ratio of transmission of a partially blocked beam to an unblocked beam for TM-polarized light through a hyperbolic metamaterial consisting of alternating layers of undoped (AlInAs) and heavily doped (InGaAs) semiconductor. The shift in the relative transmission intensity results from the negative refraction of light through the hyperbolic metamaterial layer. Figure from [5].

4.4 Summary

The development of mid-IR designer metals has the potential to dramatically affect the field of mid-IR plasmonics. These materials have the ability to effectively mimic the behavior of traditional short-wavelength plasmonic materials at mid-IR wavelengths. By designing metals with |ε_m|~ε_dat mid-IR frequencies, plasmonic structures with either tightly bound SPP modes or LSP resonances can be demonstrated, allowing for greater mode confinement, and in many cases, greater functionality for mid-IR plasmonic devices. However, it is not simply the ability to mimic the behavior of shorter wavelength structures that makes these doped semiconductor materials so powerful. The ability to tune the optical properties of the metal component of a plasmonic structure realizes a design parameter simply not available with traditional plasmonic metals at shorter wavelengths. Moreover, compatibility of these materials with both traditional and new, cutting edge, semiconductor electronic and optoelectronic devices gives, arguably, a much broader range of potential devices and applications to the mid-IR plasmonics researcher than his/her short-wavelength counterpart.