Introduction

The infrared (IR) spectral regime provides particularly pronounced molecular information due to the excitation of molecular vibrational, ro-vibrational, and rotational transitions. Generally, the MIR range is defined as the regime of electromagnetic (EM) waves with a wavelength extending from approx. 2 to 20 µm. Adjacent, the near-infrared (NIR) and the far-infrared (FIR) are extending towards the shorter wavelength (approx. 800 nm to 2 µm) and the longer wavelength (approx. 20 µm to 3 mm) regime, respectively. In the analytical MIR, it is common practice that wavenumber units (i.e., inverse wavelength) are used instead of the wavelength for direct scaling with the photon energy. Across disciplines (e.g., physics, electrical engineering, etc.) and in literature, the MIR is frequently further divided into the short-wave infrared (SWIR, 1.4–3 µm), the mid-wave infrared (MWIR, 3–8 µm), and the long-wave infrared (LWIR, 8–20 µm), respectively. When illuminating samples with MIR radiation, fundamental rotational and vibrational resonances within virtually any kind of molecule—with few exceptions—independent of molecular weight, complexity, dimension, physical status (i.e., gas, liquid, semi-solid, solid) or nature (i.e., organic and inorganic) are excited. Hence, the MIR spectrum of a sample is frequently referred to as the ‘molecular fingerprint’ enabling the detection and quantification of a wide range of relevant molecular compounds in environmental, health and medical, industrial, and security and defense applications [1,2,3].

Historically, most MIR studies have been performed using Fourier-transform infrared (FTIR) spectrometers, which use broadband thermal light sources and an interferometer for radiation modulation generating high-resolution reflectance, transmission or absorption spectra. Broadband MIR radiation is typically provided by a heated material (e.g., SiC filament) covering the entire MIR spectral range. However, due to the nature of broadband blackbody radiators, the energy density per wavelength remains limited. Additionally, such radiation sources are inefficient (i.e., in terms of optical output) and not compatible with applications that require either low power consumption or highly directional, narrowband, or coherent MIR radiation.

Nowadays, specific molecular transitions in the MIR regime may advantageously be excited via narrowband light sources such as quantum cascade—or interband cascade—lasers (QCLs, ICLs).

QCLs were first experimentally shown in 1994 [4], and are based on intersubband transitions between engineered conduction band states in complex semiconductor heterostructures instead of conventional electron–hole recombination within a semiconductor bandgap [5, 6]. These electronic transitions lead to efficient photon emission in the MIR range and enable a variety of spectroscopic target applications [7, 8]. However, QCLs are limited to wavelengths λ > 5 μm due to the common use of the InGaAs/InAlAs material system.

An alternative to QCLs are so-called interband cascade lasers (ICLs), which were proposed and realized around the same time as QCLs [9, 10]. ICLs are based on electron–hole recombination of type-ΙΙ interfaces, whereby as an example electrons of the InAs conduction band recombine with holes of the InGaSb valence band. With this design, it is possible to operate ICLs in continuous-wave (cw) mode at room temperature with low electrical power requirements (vs. QCLs) within a spectral window of 3–7 µm. However, these laser sources are complex and expensive to develop, and are therefore of limited utility for low-cost system development. Power-hungry QCLs likewise have limitations for miniaturized and/or low-cost systems.

In this review, we discuss latest MIR light sources above and beyond these nowadays established and proven laser systems, and provide an overview on alternative MIR emitters and their potential analytical utility in future low-cost, low-power, narrow-band MIR sensing and diagnostic technologies. It should be noted that indeed this review is of rather technical nature, however, it is important to summarize, compare, and understand advances of these emitters vs. laser technology for highlighting their importance in molecular analytical applications.

Advancements in MIR light sources beyond lasers

Thermal emitters

Thermal emitters are structures that emit electromagnetic radiation in the MIR range by heating to approx. 200–1400 K. The emitted surface radiation can be described via Planck’s law:

$${\text{E}}\left( {\lambda ,\,{\text{T}}} \right) = \; \in \left( {\lambda ,\,{\text{T}}} \right)\frac{{2\pi {\text{hc}}^{2} }}{{\lambda^{5} }}\frac{1}{{{\text{e}}^{{{\text{hc/}}\lambda {\text{kT}}}} - 1}},$$
(1)

where λ is the wavelength of the emitted light, T the temperature of the object, h Planck’s constant, c the speed of light, k the Boltzmann constant and ϵ (λ, T) the emissivity.

The emissivity ϵ (λ, T) describes the spectral emittance of a blackbody as a function of wavelength and temperature, whereby the limit ϵ (λ, T) = 1 represents the case of an ideal blackbody radiator. Typical blackbody emitters are the Nernst glower and the globar, whereby either a ceramic rod (e.g., SiC) or a tungsten filament, respectively, are resistively heated.

Even though thermal emitters are inexpensive, the emitted radiation—central wavelength and radiant intensity—can only be controlled by the temperature of the emitter element. Figure 1 shows the blackbody radiation of an ideal emitter and the temperature dependence of the emitted radiation in terms of wavelength. This dependence renders thermal emitters for specific applications in the fingerprint region (e.g., 6–15 µm) of limited utility, as for this spectral regime a thermal emitter would have to be operated at room temperature, and thus would have only low emissivity. To overcome these limits, one may refer to Kirchhoff’s law of radiation. Kirchhoff's law states that the emissivity of a blackbody radiator at a given temperature and wavelength is equal to the ratio of emissivity ϵ and absorptivity α at that same wavelength and temperature constant:

$$\in \left( {\lambda ,\,{\text{T}},\,{\uptheta }} \right) = {\upalpha }\left( {\lambda ,\,{\text{T}},\,{\uptheta }} \right) = 1 - {\text{R}} - {\text{T}},$$
(2)

where R is the reflectivity, T the transmissivity and θ the angle of incidence of the light. Thus, following from Kirchhoff's law an adequate absorber must also be an adequate blackbody emitter, which can be exploited for the development of advanced thermal emitters. Despite evident progress in light source technology, a large number of the systems reported to date are based on either a change of a specific property of a natural material or a temperature change, and are, therefore, limited by operation at high temperatures, a limited operating frequency, or a slow modulation speed.

Fig. 1
figure 1

Blackbody radiation for various temperatures in the range 100–2000 K

Full size image

In recent years, advances in nanofabrication and coating techniques have led to the development of deliberately design and tailorable surface structures (i.e., especially so-called metamaterials), which promise approaches for also improving MIR thermal radiation emission. By applying metamaterial structures, selective spectral losses are generated for controlling the emissivity, and consequently, the heat of the thermal emitter.

Customizing the dielectric constant and permeability of metamaterials by deliberately tailoring the geometry, dimension, and material composition along with periodic patterns leads to novel optical properties enabling the control of electromagnetic waves at the sub-wavelength level.

In the following, we will discuss in more detail the development of metamaterial-based thermal emitters and their potential applications.

Microelectromechanical system (MEMS)-based thermal emitters

MEMS were first introduced in the mid-1980s [11]. MEMS devices are usually structured from components at the 1–100 µm scale, and the dimensions of the entire resulting device is in the range of few micrometers up to several millimeters [12]. These components are usually fabricated from materials including polymers, metals, silicon—and other semiconductors—and ceramics, and are structured and fabricated using various lithographic, etching, and deposition methods [13,14,15].

Shortly after their fundamental introduction, the first MEMS thermal emitters for sensor applications were reported [16,17,18]. Parameswaran et al. [16] first reported the fabrication of polysilicon microbridges, sandwiched oxide microbridges, and sandwiched oxide cantilever beams using a twin-tub complementary metal–oxide–semiconductor (CMOS), and discussed their potential for sensor applications. In the following, Greenwald et al. [19] showed etching periodic structures onto a silicon surface coated with gold by varying the interstitial proportions and geometric patterns. Thereby, the central emission wavelength was controlled and narrowband heat emitters with selected wavelengths were produced. Further research has resulted in the use of advanced structuring techniques and novel materials for the fabrication of MEMS systems. For example, silicon-on-insulator (SOI) wafers (e.g., made of SiO2, AnSiO2, etc.) were used in MEMS fabrication, and in addition, silicon-based surfaces were no longer coated with metals but also doped with semiconductors (e.g., SnO2:Sb, poly-Si, PtSi, Boron-doped poly-Si, etc.) to ensure resistive heating and tailor the properties [13,14,15, 20,21,22,23].

Most recent research has focused on increasing the emissivity at high temperatures (> 300 °C) by tuning the MEMS structure using gold metamaterials on SiO2 [24, 25]. Again, it was shown that by varying the structure thickness and pattern spacing, the central emission wavelength and the intensity of the thermal radiation changes depending on these parameters. Furthermore, it has been shown by that by depositing graphene oxide (GO) using the CMOS process, the intensity of the emitted radiation may be further enhanced [26, 27]. Chen and Chen [26] as well as Li et al. [27] showed that by applying GO coatings at the thermal emitter surface, an increase of up to 150% of the signal can be achieved. Furthermore, Liu and Padilla demonstrated two distinctly different MEMS emitters with various geometric designs and central emission wavelengths (~ 5 and ~ 8.5 µm, respectively), which have spectral radiant energy densities of up to 2.0∙107 W/m3 at room temperature [28, 29]. Fig. 2 shows the temperature dependence and wavelength shift of spectral emissivity as a function of applied temperature for the above MEMS emitter.

Fig. 2
figure 2

Emitted power density of the thermochromic metamaterial. Power density per wave-length for different temperatures as indicated in the inset legend. Copyright © 2015, John Wiley and Sons. Reprinted and edited with permission from [28]

Full size image

Despite the significantly lower radiant energy density compared to thermal emitters operated at much higher temperatures, these advancements demonstrate the potential of novel metamaterial-based MEMS emitters at room temperature operation, which renders them ideally suited for integration into miniaturized sensing systems and diagnostics.

Plasmonic nanostructured arrays

Plasmonic nanostructured arrays take advantage of the fact that the electrons in these structures that want to restore the ground state are prevented from moving by the oscillation of the EM wave, and are instead stimulated to oscillate at the frequency of the EM radiation resulting in a shift of the dipole.

In 1998, Ebbesen and colleagues have for the first time been able to show transmission in the NIR resulting from perforated metallic films via plasmonic structures [30, 31]. For this purpose, an Ag film of thickness 0.2 µm or between 0.2 and 0.5 µm, respectively, was deposited onto a quartz surface by evaporation, and subsequently arrays of cylindrical holes were fabricated via focused ion beam (FIB) milling. Later, first plasmonic thermal emitters in the mid-infrared regime were also published. In contrast to Ebbesen and colleagues, here a metal/insulator/metal layer (e.g., Ag/SiO2/Ag, Au/SiO2/Au, nanoamorphous carbon:TiN/SiO2/nanoamorphous carbon:TiN) [32,33,34,35,36,37,38] or a graphene/insulator layer (e.g., graphene/SiO2, graphene/diamond-like carbon, SiNx) [39, 86, 87]. In these systems grown on GaSb substrate, the optical extraction efficiency limited by the large refractive index mismatch between the semiconductor and air has been addressed by placing the active structure between two DBRs. Figure 7 shows a schematic of the complete RCLED structure with top DBR/MQW/bottom DBR regions. Here, the first RCLED with a target wavelength of 4.2 µm consisted of lattice-matched AlAs0.08Sb0.92/GaSb DBR mirrors and a MQW InAs0.90Sb0.10 active region [86]. The second RCLED, on the other hand, had a target wavelength of 4.5 µm with AlAs0.08Sb0.92/GaSb DBR mirrors and MQW Al0.12In0.88As/InAs0.85Sb0.15 active region [87]. They could demonstrate for both RCLED systems that they have a significantly stronger electroluminescent as well as narrower emission spectrum compared to a reference MIR-LED without resonance cavity. However, they also saw a wavelength shift of the RCLEDs with increasing temperature.

Fig. 7
figure 7

Schematic of a MQW resonant cavity LED structure showing the details of the AlInAs/InAsSb MQW in the active region of the p-i-n diode within the AlAsSb/GaSb DBRs, which form the microcavity. © 2020 Optica Publishing Group. Reprinted and edited with the permission from [87].

Full size image

Another recently explored method to increase the efficiency and performance of LEDs is to introduce stages into LED structure. For their system, Schaefer et al. [88] deposited three samples, each with nine stages (1 \(\times\) 9 centered at the vibrational antinode, 3 \(\times\) 3 positioned at the node, and 3 \(\times\) 3 positioned at the vibrational antinode) in the GaSb/InAs active layer on a low-absorption substrate. They were able to show for their 640 \(\times\) 640 µm sized system with a target wavelength of 3.5 µm that the electrical properties and decoupling efficiencies were improved when all active stages were centered within a single antinode of the optical field. The resulting optimization of the voltage efficiency as well as the lower optical losses led to a maximum radiant efficiency of 0.7% and a power output of maximal 5.1 mW at 0.6 A drive current.

Delli et al. [89] used a different approach to fabricate the MIR-LED heterostructures on a silicon substrate. For fabrication, they used a three-level buffer layer InAs/GaSB/Si (top to bottom). On this buffer layer, they deposited a five-period AlSb/InAs dislocation filter superlattices (DFSL), where each individual superlattice period again consisted of five alternating layers of AlSb and InAs. The individual superlattice periods were separated with a spacer layer of InAs. Finally, they deposited the InAs/InAsSb MQW structure on top of the DFSL. With their novel structure, they were able to show emission in the spectral range of 3.5–4.2 µm. Similar to what Kuze's team showed [85], Delli and colleagues also saw a strong filtering effect from filamentary dislocations at the InAs/GaSb interface.

Montealegre et al. [90] introduced even more stages in their system. They report a cascaded W-SL structure with an inner part consisting of 16 stages. Each of the stages in turn consists of an SL AlAsSb/InAs/GaInSb/InAs emission layer and an n/p GaSb/AlInAsSb tunnel junction. In addition, they thinned and roughened the emission side of the structure to improve light outcoupling, resulting in randomization of light direction and reduction of internal reflection and light outcoupling. With their 400 \(\times\) 400 µm cascaded superlattice LED structure, they were able to demonstrate a power output of 6.8 mW at a wavelength of 4.2 µm in quasi-continuous wave mode (10 µs pulses) at room temperature.

Selected applications

IR spectroscopy as a generic analytical technique may particularly benefit from the discussed advanced MIR light sources. By default, IR spectroscopy is based on broadband thermal infrared radiation. For target application scenarios, these may be replaced by non-coherent yet better tailorable IR light sources such previously described. Quantitative light-matter (i.e., molecule) interactions in IR spectroscopy are described using Lambert–Beer’s law:

$$A_{{\uplambda }} = { }\ln \left( {\frac{{I_{0} }}{{I_{1} }}} \right) = \varepsilon_{{\uplambda }} \cdot {\text{c}} \cdot {\text{d}}$$
(3)

with \(A_{\lambda }\) the absorbance, I0 the intensity of the incident light of the source, \(I_{1}\) the measured intensity, \(\varepsilon_{\lambda }\) the absorption coefficient, c the concentration and d the length (i.e., absorption path length) of the irradiated sample.

The probably most common application of MIR spectroscopy in need for low-cost and efficient light sources is the detection of gaseous molecules in environmental analysis, process monitoring and biomedical diagnostics. In the infrared region, many prominent gases such as NH3 (2.1 μm), CH4 (2.35 μm and 3.3 μm), HCl (3.5 μm), N2O (3.9 μm and 4.5 μm), SO2 (4 μm), CO2 (4.2 μm), and CO (2.3 μm and 4.6 μm) have their molecule-specific vibrational, ro-vibrational, and rotational transitions just to name a few [91].

As previously described, currently many thermal MIR emitters as well as MIR-LEDs emit in a wavelength range of 4 ± 1 µm. Especially for MIR-LEDs, energy transitions of the semiconductor materials ≥ 6 µm remain a challenge, which is why to the best of our knowledge to date MIR-LEDs in the MIR range 6–20 µm have rarely been reported to date. Nevertheless, a sizeable number of gas measurements and sensors based on MEMS thermal emitters [92,93,94,95,96], plasmonic emitters [63], and a MIR gas sensor for acetone and ammonia at 8.26 and 10.6 µm, respectively, by **ng et al. [97]. These groups used 2D PC and a plasmonic nanostructures, respectively, to achieve their emitter. Also, these studies impressively demonstrated qualitative and quantitative analyses using novel MIR emitters, and document their utility as a viable alternative to more commonly used laser sources.

Additional applications of the presented MIR emitters/absorbers worth mentioning here are beyond analytical scenarios and go beyond the scope of the present review, e.g., among others in military for stealth technology and thermovoltaics [3, 104, 105], and for power generation scenarios [44, 56, 106, 107].

Conclusion and outlook

The mid-infrared spectral regime is the most information-rich part of the electromagnetic spectrum suitable for establishing sensing and diagnostic devices providing access to selective molecular absorption characteristics and multiple atmospheric transmission windows. Therefore, MIR devices—and the associated component technologies—are considered fundamental not only for chem/bio sensing, but also for imaging and communications applications. Key to both the fundamental and applied aspects of MIR technology are efficient radiation sources in the 2–20 µm wavelength range.

The present review discusses MIR light sources above and beyond devices resulting from the ‘laser revolution’ providing less intense yet tailorable emission characteristics with potentially low-cost per device, a tunable and confined emission range, and reduced power requirements during operation in miniaturized sensing and diagnostic systems. Table 1 gives an overview of the MIR light sources discussed in this review with their characteristics along with some prominent application examples.

Table 1 Overview of the discussed MIR light sources
Full size table

Of particular emphasis are recent developments in the field of MIR-LED radiation sources, which not only provide advantageous properties, but may in future be augmented by appropriate organic materials (i.e., OLEDs) and quantum dots (i.e., QDs) leading to even more advanced technologies. While OLEDs have already been thoroughly researched emitting in the NIR [108,109,110,111], to date MIR-OLEDs have not been shown. In a similar fashion, NIR-QD emitters have readily been published [112,113,114,115,116], while the first MIR-QD system reported by Briggs et al. [117] in 2020 signals that MIR-QD emitters are on the horizon. Last but not least, expanding these concepts up to wavelengths around 20 µm is still pending, yet promises a highly active research field benefitting MIR chem/bio sensors and diagnostics in the years to come.