There is interest in the continued miniaturization of light sources from both a scientific perspective of understanding the properties of such small light sources, and also from the utilization of such light sources. In particular, nanoscale lasers have been predicted and shown to have very high modulation speeds and low power,^[1] attributes that lend themselves well to future applications such as on-chip optical communications^[2] where low power, high speed, and small size are important. Furthermore, nanolasers are needed in the continued miniaturization of integrated photonic systems, which could be employed in diverse areas such as optical sensing systems or digital processing of optical information.^[3]

Of all the types of small cavities so far employed to form lasers or resonant cavity light emitting diodes (LEDs), metallic cavities are unique in offering both subwavelength confinement of the light and a small overall footprint for the device,^[4] as shown in Fig. 1. Furthermore, the field of plasmonics primarily involves using metal structures to guide, confine, and concentrate light into small regions often much smaller than the diffraction limit.^[5] These highly concentrated fields not only offer the possibility of further miniaturization of photonic devices, but also provide much stronger light–matter interaction due to the high optical field strengths produced.

Fig. 1.

Figure Option ViewDownloadNew Window

Fig. 1. (color online) Size comparison of metallic and plasmonic nanolasers to dielectric nanolasers. The electron microscopy pictures are scaled to the free space emission wavelength λ of each laser. (a) A photonic crystal laser, a type of small dielectric cavity laser;[66] (b) metallic nonplasmon mode laser;[7] (c) metallic propagating plasmon mode laser;[10] (d) localized plasmon mode laser;[9] and (e) another metallic nonplasmon mode laser.[12] The metal cavity lasers developed recently exhibit a dramatic reduction in size compared to dielectric cavity lasers.

A key issue common to both plasmonics and the use of any small metallic cavity is that of the high optical losses caused by the free electrons in the metal. It has been shown that these metallic losses can be compensated for through the use of an optical gain medium and clever design of the metallic cavity or waveguide.^[6–15] The optical loss compensation can occur while still achieving a small device size and subwavelength confinement of the light. Furthermore, coherent light emission has been confirmed in these metallic light emitters by studying the statistics of the emitted light.^[15,16] Many of these demonstrations of loss compensation and lasing in metallic/plasmonic devices have occurred at cryogenic temperatures and/or employed optical pumping of the gain medium.^[7–15] Operation at cryogenic temperatures has been shown to reduce the metal losses via reduced phonon scattering of the free electrons in the metal.^[7,10,17,18] Employing optical pumping allows a wider range of gain medium to be employed, and also avoids the losses and issues introduced with electrical pumping of the gain medium. These electrical pumping issues will be discussed in detail later.

While optical pumping and cryogenic temperatures have been demonstrated in many interesting devices, for most applications it is necessary to have devices that are electrically pumped and can operate at room temperature or higher. Electrical pumping is advantageous for many reasons including the following: electrical pumping avoids the requirement of a secondary light source for pumping, and allows direct electrical modulation of the nanoscale light source. Furthermore, electrical pumping permits gain material buried deeply inside metallic structures to be pumped. Finally, the energy for pumping can be directed efficiently into many possibly spatially separated nanoscale regions of gain material via wires.

Electrically pumped room temperature operation is also an important contemporary topic for the smallest of dielectric lasers and photonic crystal lasers.^[19,20] Typically in dielectric lasers, metal contacts must be kept a considerable distance from the dielectric cavity. Metal-based small lasers have an inherent advantage compared to dielectric cavity lasers when considering electrical pumping. Specifically, the metals used to form the cavity itself are available to transport the current to the gain medium.

Loss compensation for plasmonics in particular is an important issue because typically the losses are high when light is tightly confined in plasmonic waveguides or cavities, and light can only propagate a small distance before decaying.^[21] (There are plasmon mode waveguides that allow long propagation distances; however, they do not confine the light to small regions.^[21]) Widely employed integrated photonic systems typically have means of compensating for the losses in their waveguides and of providing integrated laser sources, for example, InP-based photonic-integrated circuits.^[3] Alternatively, the losses in the photonic system need to be sufficiently small so that a meaningful number of manipulations of an optical signal can be performed before it is excessively attenuated, for example, silicon on insulator photonic-integrated circuits.

Hence, loss compensation in plasmonic systems is important to permit more complex systems to be built and provide integrated sources of coherent plasmons. An electrically pumped means of loss compensation that can provide large numbers of coherent plasmon sources is particularly important.

The rest of this review will look at the following aspects of electrically pumped metallic and plasmonic nanolasers. Firstly, the critical issues that need to be addressed to obtain loss compensation and lasing in small metal structures are examined, and then the key elements required for electrical pumping that typically impact nanolaser design are identified. Examples will be given of some of the optically pumped metal cavity nanolasers and how they achieve lasing. Then, examples of electrically pumped devices will be given, and it will be shown how the requirements for electrical pumping increase the challenge of overcoming the metal losses. A comprehensive review of the latest electrically pumped devices (including those that may fall short of achieving lasing) will be given. There is also a considerable amount of literature providing only simulation or theory results about electrically pumped nanolasers. This simulation and theory literature will also be reviewed to see what is possible in electrically pumped nanolasers and what could be the future directions for their development.

The dielectric function, ε_m, of the metals employed for metal nanolasers and plasmonic waveguides is both negative and complex, i.e., ε_m = ε ′_m + iε″_m. Silver is the metal with the lowest losses at wavelengths of most interest for plasmonics and nanolasers, and thus is employed in many nanolaser demonstrations. However, even for silver, the loss due to ε″_m is significant (ε_m = –130 + i3.3 at a 1550 nm wavelength^[22]). The decay of light in a metallic cavity or along a metallic waveguide depends on how much the electric field of the mode overlaps the metal. In general, for optical modes that are tightly confined by the metal structures, propagation lengths and decay times are short.^[21] For less-confined optical modes, propagation lengths along waveguides can be in the order of a millimeter.^[21]

A nanolaser consists of two main components: a nanocavity or resonator, and an optical gain medium that overcomes the optical losses in the nanocavity. The nanocavity can be characterized via a photon lifetime τ_p, which gives the time the optical power in the cavity mode decays to 1/e of its initial value.^[23] τ_p is related to the cavity quality factor Q, resonant wavelength λ₀, and the speed of light c,^[23]

The semiconductor gain medium is characterized by the material gain G_a. The confinement factor Β is defined as the proportion of the laser mode energy that overlaps the optical gain medium.^[23]

The G_a required to overcome the cavity losses and so create a laser G_th, is related to the other cavity parameters^[23]

In Ref. [24], this tradeoff between metal losses and Β is explored to achieve a sufficiently low G_th so that lasing can be achieved at room temperature,^[11] even when higher optical loss metals such as aluminum are employed. In Ref. [11], a thick dielectric spacer between the metal and gain material is employed to reduce the overlap of the optical mode with the confining metal.

In Refs. [10] and[26], the surface plasmon modes typically involve just one surface of the metal or the metal stripe is very thin. Furthermore, the refractive index of the majority of the surrounding dielectric is low, which greatly reduces the amount of optical energy in the metal compared to having a high refractive index material adjacent to the metal.^[30] Typically, these plasmon modes are called long range surface plasmons,^[21] due to their significant propagation distance.

Another feature that occurs in both Refs. [11] and[10] that enables a large Β is the low refractive index of the dielectric material that completely surrounds the relatively high refractive index semiconductor gain material. In Ref. [11], SiO_x (n ∼ 1.5) and air (n ∼ 1) surround the gain material (n ∼ 3.5), which leads to the optical mode being strongly confined on the gain material.

The strategy of having the gain medium surrounded by a low index dielectric has been very successful in producing nanolasers;^[10–12] however, it requires that the gain medium is optically pumped. Typically, low index materials are insulators and no pathways exist to carry current to the semiconductor gain medium.

To move from optically to electrically pumped nanolasers, two key issues must be solved. First, there needs to be a way to efficiently inject electrons and holes into the semiconductor from metal electrodes. Typically, what is desired is to form low-resistance ohmic contacts between metal electrodes and a particular n- or p-doped semiconductor. Considerable effort over the last 50 years has gone into developing such low-resistance ohmic contacts to many types of semiconductor.^[31] In fact, the ability to efficiently inject electrons and holes into a particular semiconductor via electrodes determines whether or not it is suitable for use in laser diodes or energy-efficient LEDs. The second key issue is that there often needs to be semiconductor pathways from the electrode contact regions to the semiconductor gain material.

A schematic of an electrically pumped double heterostructure ridge waveguide is given in Fig. 2(a). Consider for the moment that the encapsulating silver in Fig. 2(a) is removed, and we just focus on illustrating solutions to the metal contact and pathway issues. The double heterostructure waveguide concept enabled the first electrically pumped room temperature lasers and is still used in some form in most electrically pumped lasers. The device in Fig. 2(a) employs the indium phosphide/indium gallium arsenide (InP/InGaAs) material system, which covers near infrared wavelengths. Heavily doped low bandgap energy semiconductors are typically employed to form the electrical contact regions. In the case of Fig. 2(a), a heavily n-doped InGaAs (bandgap ∼ 0.75 eV) on top of the ridge forms the n electrode contact. At the base, there is a heavily p-doped InGaAsP layer that forms the p electrode contact. The semiconductor gain region consists of undoped (intrinsic) InGaAs. To transport the electrons to the intrinsic InGaAs region from the n contact, there is a region of n-doped InP (bandgap ∼ 1.34 eV), and similarly to transport holes from the p contact there is a region of p-doped InP. Holes and electrons are trapped in the lower bandgap undoped InGaAs gain region where they can recombine creating optical gain. The InGaAs also has a slightly higher refractive index than the InP so the optical mode is typically centered on the undoped InGaAs region. To keep the optical mode away from the contact regions, the length of the InP regions employed to transport current to the gain region is often on the order of microns. The double heterostructure ridge waveguide solves the problem of having the optical mode centered on the gain region and separated from the contacts in a way that is easy to implement with epitaxy and standard processing techniques.

Fig. 2.

Figure Option ViewDownloadNew Window

Fig. 2. (color online) Different forms of small electrically pumped metallic and plasmonic lasers. (a) A semiconductor heterostructure encapsulated in silver.[8] This structure has been used in many electrically pumped metal lasers. With the silver removed, the structure is similar to many dielectric waveguide heterostructure lasers. (b) Plot of the magnitude of the electric field for a transverse magnetic 0 (gap plasmon) mode that propagates down the metal waveguide of (a). Red indicates the highest intensity and blue indicates the lowest intensity.[8] (c) The semiconductor core encapsulated in silver in (a) can have many shapes, round pillars, straight waveguides, or, as shown here, perturbations to form distributed feedback lasers;[36] the scale bar is 100 nm. (d) Another form of semiconductor heterostructure laser, but now having the mode travelling up and down the pillar, which is equipped with mirrors at each end.[39] (e) The structure of a far infrared laser that employs an MIM structure and circuit concepts to reduce the laser size well below the wavelength of the laser emission.[43]

The two requirements of contacts and transporting current from contacts to gain region create three key problems that the optically pumped metal cavity nanolasers do not suffer from.

Firstly, stable low-resistance electrical contacts to semiconductors typically involve a number of metals that have optical losses much higher than metals such as silver. These metals, examples of which are titanium and platinum, among other tasks provide adequate adhesion of the contact to the semiconductor and prevent diffusion into the semiconductor of other metals that form part of the contact. The key point is that the metal contact regions exhibit extremely high optical losses. The optical mode of the nanolaser must be kept away from the contact regions to avoid high cavity losses that cannot be compensated for by the gain medium, similar to a dielectric cavity laser. Some low-optical loss contact metallization schemes have been investigated^[32] and may play a role in the future.

Secondly, the refractive index contrast between the semiconductors, which transport the carriers from the contacts and the gain medium, is typically quite low. For example, the refractive index n of InP is 3.17 compared to 3.6 of InGaAs, at 1.55 microns wavelength. Most optically pumped nanolasers have dielectrics such as air or SiO₂ (n ∼ 1 to 1.5) adjacent to the semiconductor gain medium. The main result of the much lower index contrast is that it is harder to obtain a high Β in electrically pumped nanolasers. Additionally, the higher refractive index of the materials next to the metal results in a larger amount of optical mode energy in the metal, and therefore metal cavity losses are also increased.^[30]

Thirdly, resistive heating from both the electrical contacts and the semiconductors transporting the carriers to the gain medium will be detrimental to the operation of the nanolaser.

While there has been a wide variety of optically pumped metallic nanolaser structures, only a comparatively few forms of electrically pumped metallic nanolaser have been demonstrated. The first metallic nanolasers demonstrated were electrically pumped and based on the concepts of the double heterostructure ridge waveguide laser, as shown in Fig. 2(a). In Fig. 2(a), electrical contacts are separated from the InGaAs semiconductor gain medium by a few hundred nanometers of InP. As in conventional electrically pumped lasers, the InP regions transport the carriers from the electrode contacts to the gain medium. The pillar or ridge that forms the core of the nanolaser is encapsulated first in a thin dielectric layer, then a noble metal such as gold or silver. The thin dielectric layer prevents the noble metal layer short circuiting the pillar structure, and also stops diffusion of any noble metal into the semiconductor. The noble metal shell forms part of one electrical contact at the top. The optical cavity is formed by the presence of the higher index InGaAs and the metal shell. In the case where the etched semiconductor structure is a pillar,^[7] the wavelength of the resonant mode is at the cutoff wavelength of the waveguide formed by the InGaAs-filled metal pillar. The mode is localized at the center due to the fact that the cutoff wavelength in the InP region is smaller than that in the InGaAs region. Essentially, the light is bouncing between the vertical walls of the pillar.

Where the etched semiconductor region forms a ridge,^[8] light is guided in a metal–insulator–metal (MIM) waveguide. These particular MIM waveguide lasers employ the MIM waveguide concept explored in Ref. [33], but with the MIM structure now vertical instead of horizontal.^[33]

The concept shown in Fig. 2(a) maintains the electrical contacts at a sufficient distance from the optical mode to minimize their losses. Furthermore, a low-resistance path is provided from the contacts to the gain medium by the InP. However, localization of the mode on the InGaAs is much poorer than in optically pumped nanolasers due to the low index contrast between InP and InGaAs, particularly with nonideal pillar shapes. Improvements to the mode localization have been proposed and realized by making the InP regions thinner.^[34,35] Furthermore, the presence of high-index semiconductors everywhere increases the optical mode penetration in the metal and hence losses. In spite of these shortcomings, the etched pillar concept shown in Fig. 2(a) has been employed in a number of electrically pumped nanolasers. In particular, it has been used to demonstrate small plasmonic distributed feedback lasers^[36] and nonplasmon mode lasers that show room temperature lasing.^[37] It has also been the basis of a simulation study that showed that in theory it is possible to create very fast plasmon mode amplifiers.^[38]

Another form of etched structure that has similarities to the structure shown in Fig. 2(a) has also produced useful electrically pumped small lasers.^[39,40] The structure of this laser is shown in Fig. 2(d). The central core of the structure consists of a pillar with a central gain region coated with a thin dielectric layer and encapsulated in metal. Here, the optical mode travels up and down the pillar at wavelengths below the cutoff wavelength. Bragg or metal mirrors constitute the end sections of the pillar and confine the optical mode to the central gain region. The Bragg mirrors also keep the optical mode away from the electrode contacts and transport carriers to the gain region. Note that the structure has a relatively large height (several microns) due to the depth of the Bragg mirrors. Furthermore, the diameter of the pillar is also of the order of a micron. Nevertheless, reliable room temperature lasing has been achieved in these structures. A progression from using Bragg mirrors to a much smaller device size using only metal mirrors has been proposed.^[41]

A horizontal MIM structure has also been used to create mid and far infrared wavelength lasers (approximately > 3 microns).^[42] Here, the semiconductor gain medium is sandwiched between two pieces of metal, which form the electrical n and p-type contacts. Generally, a thin layer of titanium or other adhesive metal is used between the semiconductor and gold to assist adhesion of the metal to the semiconductor. Additionally, there are often specific doped semiconductor layers close to the metal to achieve a low contact resistance between the metal and semiconductor.^[42] For the longer wavelengths such devices operate at, the metal losses are low and the extra losses due to the titanium layer are acceptable. Furthermore, the semiconductor layer between the metal is typically several microns thick. The thickness of this layer permits the use of n- and p-doped contact layers in the semiconductor, while still having a significant proportion of the semiconductor as undoped gain medium.

In Ref. [43], a variation of the MIM structure for far infrared lasers involved introducing circuit concepts to create the resonant cavity. Here, sections of MIM with different aspect ratios are used to form inductor- and capacitor-like elements. It would be interesting to see how such resonators based on circuit concepts could be employed for much shorter wavelengths.

The dimensions of these far infrared wavelength MIM laser structures are far from nanoscale. Reducing this horizontal MIM structure down to nanoscale dimensions and shorter wavelengths while still achieving lasing is difficult due to the high optical losses of the metal contact regions close to the optical mode. Nevertheless, these long wavelength MIM lasers highlight an important advantage of electrically pumped devices: confining metal structures themselves can be employed as electrical connections.

There have been a number of demonstrations of light (or surface plasmon) emitters that employ such horizontal MIM structures, with nanoscale dimensions and shorter wavelengths. It is worthwhile looking at these structures considering that the number and variety of demonstrated electrically pumped metallic and plasmonic nanolasers is so small. As highlighted by a number of authors,^[44] nanoscale metal cavity surface plasmon emitting diodes (SPEDs) or LED sources may be useful sources. At the very least they can provide an interim solution for small plasmon sources until technology develops sufficiently for the challenges of metal cavity nanolasers to be overcome in a wide variety of configurations.

One of the first demonstrations of surface plasmon emission in an MIM structure was reported in Ref. [45] (Fig. 3(a)). Here, electrons tunneled through a thin (∼ 100 nm) aluminum oxide layer which was sandwiched between gold contacts. The aluminum dioxide layer contained silicon nanocrystals, which were electrically pumped when electrons tunneling between the gold contacts collided with the nanocrystals. The spontaneous emission from the silicon nanocrystals couples to the gap plasmon mode that propagates in the MIM waveguide.

Fig. 3.

Figure Option ViewDownloadNew Window

Fig. 3. (color online) Some of the LEDs and SPEDs that have been realized. (a) An MIM structure with silicon nanocrystals placed between gold electrodes.[45] (b) Another MIM waveguide, but with semiconductor emitters placed outside the waveguide. Plasmons are coupled into the waveguide via a slit.[46] (c) A quantum well semiconductor emitter inside an MIM waveguide[44] with the MIM layer orientation rotated 90 degrees with respect to those in (a) and (b). (d) Example of a metal-clad nanocavity LED coupled to a dielectric waveguide.[48]

Compound III–V semiconductors have also been employed for small surface plasmon emitters, forming SPEDs. In Ref. [46] (Fig. 3(b)), a quantum well emitter is placed in close proximity to a slot in an MIM waveguide. One of the diode electrodes was formed by one of the metal plates in the MIM waveguide. Spontaneous emission from the quantum well coupled to the gap plasmon mode inside the MIM waveguide to generate propagating gap plasmons.

Another demonstration of a nanoscale SPED was given in Ref. [44] (Fig. 3(c)). Here again a quantum well emitter was employed. However, the emitter was placed inside the MIM waveguide. Here, the SPED had truly nanoscale dimensions of 130 nm × 60 nm. Furthermore, a number of circuits formed with the MIM waveguides were demonstrated by employing the SPED source.

In Ref. [47], an electrically pumped quantum well emitter has was also placed next to a thin gold layer. The idea was to fully compensate for the metal losses for the long range surface plasmon supported by the gold layer. However, loss compensation was not achieved.

Finally, a recent demonstration of an LED involving a metal nanocavity was given in Ref. [48] (Fig. 3(d)). Here, the actual cavity structure is similar to the encapsulated heterostructure in Ref. [7]. Such structures as in Ref. [48], when correctly constructed, should in theory be able to provide efficient electrically pumped nanoscale laser sources coupled to integrated photonic waveguides.^[34,49]

There has been a considerable amount of theory and simulation articles published about metal and plasmonic cavity nanolasers. It is worthwhile briefly examining some of the literature pertinent to electrically pumped devices and in particular those focused on semiconductor gain medium, as shown in Fig. 4. Simulation results provide an idea of what could be possible in the future if technology to manufacture metal nanolasers is perfected, and also identify interesting concepts to pursue.

Fig. 4.

Figure Option ViewDownloadNew Window

Fig. 4. (color online) Some of the many small metallic and plasmonic lasers that have been proposed. (a) A metal-clad laser that couples its light efficiently to a waveguide.[34] (b) Simulation of the structure of (a) showing the magnitude of the electric field and that laser light couples well to the waveguide.[34] (c) A proposal for an extremely small “nanocoin” laser.[41] (d) Another proposal for a waveguide-coupled nanolaser, but now employing a surface plasmon mode.[50]

Some of the initial theory to look at semiconductors amplifying surface plasmons at visible and near infrared wavelengths was given in Ref. [30], showing it was indeed likely that metallic losses could be overcome. Further work by the same group showed that by using less-confined transverse electric (TE) modes and dielectric spacers, some metal-coated laser resonators could be formed with quite low threshold gain requirements that could be readily satisfied by semiconductors.^[24] Other modeling work looked at Fabry–Perot metal nanocavities.^[23] Again, it was shown in theory that with current semiconductor gain media it would be possible to obtain lasing in these cavities. However, all of the above-mentioned simulations did not incorporate structures to permit electrical pumping of the gain medium.

A number of later simulations also looked at coupling metal nanocavities to waveguides and examining their potential as laser sources for on-chip communications,^[34,49] as shown in Figs. 4(a) and 4(b). In particular, these simulation studies took into account structures for electrical pumping that kept electrical contacts distant from the optical mode. Furthermore, they showed that the light output from the nanolaser could be efficiently coupled to a photonic waveguide. As such, they could form useful laser sources in the near future. The modes employed in these cavities were TE modes, and thus while the cavities were small their dimensions were not below the diffraction limit. Another study looked at coupling a plasmonic nanocavity laser to a photonic waveguide,^[50] as shown in Fig. 4(d). However, electrical contacts were not separated from the plasmonic mode, and estimates of threshold gain were not directly given.

A number of simulation studies have highlighted that considerable improvements in the performance of metal cavity devices are possible by optimizing or modifying their structure. In Ref. [51], it is shown how a simple change in the shape of the ends of a Fabry–Perot cavity, constructed with the waveguide of Fig. 2(a), can significantly reduce metal losses. In Ref. [52], a different structure from that shown in Fig. 2(a) is presented, where the electrical contacts are separated from the optical mode. In particular, it shows that confinement of the optical mode and metal losses can be significantly improved compared to waveguides based on the structure shown in Fig. 2(a). Furthermore, it is shown that in terms of providing an amplifier with a small active region, it can have a figure of merit better than that of the best conceivable dielectric waveguide. The improvements are due to the use of low-index materials close to the metal and gain regions.

A number of theory and simulation studies have also examined the intrinsic modulation speed of both metal nanocavity lasers and LEDs.^[1,53,54] All studies suggest that the metallic nanocavities can provide useful and high-speed devices with distinct advantages in terms of size and speed over dielectric-based cavity devices.

Some studies have made broad dismissive statements about a particular single nanoparticle cavity^[55] or a simplified one-dimensional metal structure^[56,57] being used for devices. However, it is dangerous to generalize these results to all possible structures that could exist. The theory and simulation studies cited above already indicate that metallic and plasmonic nanolasers are worth pursuing, and they can have advantages in certain respects to dielectric cavity lasers. It is true, particularly for electrically pumped devices, that the current experimental device performance is far from what theory predicts. However, with improving manufacturing technology, device performance should improve to approach that of theory; additionally, new device structures with better performance may be found. Indeed, optically pumped devices are already showing interesting results, which indicates that plasmonic laser devices do indeed have particular performance advantages over dielectric cavity devices since their size goes below the diffraction limit.^[58]