1.1.

Device Size and Energy-Efficiency Challenge

The sizes of photonic devices are generally limited by the wavelengths involved. Thus, it becomes an important challenge to overcome wavelength limits or diffraction limits. The important questions to ask are whether, how, and to what degree we can break the diffraction limit to create ever smaller and better optoelectronic devices such as lasers. An important related issue is the energy efficiency of photonic devices (such as lasers, modulators, and switches) when used for information transmission,¹ in terms of joule per bit of information transmitted. These questions and their related challenges are important for realizing future integrated nanophotonic on-chip circuits.

1.2.

Wavelength or Bandgap Diversity Challenge

All semiconductor photonic devices, including lasers, are based on light–semiconductor interaction involving either absorption or emission of photons by semiconductors. Important spectral response (emission, refraction, or absorption) of any semiconductor is ultimately determined by its electronic bandgap and bandstructure. Many applications require (or can significantly benefit from) bandgaps that can be controllably tuned in a wide range, allowing bandgap diversity or flexibility, preferably on a single substrate or monolithically.² However, our ability to achieve the required diversity of bandgap is rather limited, primarily because of the lattice-matching required in typical planar epitaxial growth of high-quality semiconductors. Such lack of ability to produce the requisite bandgaps severely impedes technological progress in many applications, including displays, solid-state lighting, solar cells, detectors, and widely tunable lasers. Nanoscale semiconductors, such as nanowires,^2–4 perovskite platelets,⁵ and quantum dots, offer many potential benefits and have enabled many exciting developments, such as widely tunable emission in the entire visible spectrum, or white lasers from a single substrate, or a single monolithic semiconductor. Such nanomaterials must be fully and systematically explored to develop more mature devices.

1.3.

Integration Challenge

A well-recognized long-term challenge is the achievement of integrated photonics on a silicon platform. While most passive devices can be fabricated directly on an Si platform, light sources are still almost exclusively made of III-V materials. Thus, heterogeneous integration of III-V-based lasers, or gain materials onto Si-waveguides, has become a prevailing approach.^6–9 Although mismatch of the material properties between typical III-V semiconductors and Si has prevented direct growth of high quality III-V semiconductor films on Si using the typical planar growth techniques, recent work in nanomaterials has shown promise for direct monolithic growth of various nanomaterials on Si, such as III-V quantum dots^10–12 and nanowires.^13–21 Such a nanomaterials approach could be widely applied in the long run. However, achieving efficient lasers on a silicon substrate remains an important challenge.

This paper presents an analysis and summary of the aforementioned challenges in semiconductor laser research, with focus on one of the frontiers of the field, namely, semiconductor nanolasers. We focus on the long-term issues and fundamental challenges that are likely to remain unresolved for the foreseeable future, and those that will impact the field in profound ways. In particular, the size-energy-efficiency challenge is emphasized because of its importance and potential impact on the other two challenges and on the entire field of semiconductor photonics. In the following sections, the background of the size-energy-efficiency challenge is introduced, and the relevant progress made in the last decade or so is highlighted. The potential and shortcomings of each approach are analyzed. Finally, we present future perspectives in the resolution of these challenges and the possible impact on the field of semiconductor lasers and photonics.

3.1.

Vertical Cavity Surface Emitting Lasers

VCSELs, which are one of the most important types of microsize lasers, were invented²⁴ many years after the initial demonstration of conventional EELs. VCSELs offer a size reduction of at least an order of magnitude in total device volume compared to EELs (see Fig. 4). Currently, VCSELs are one of the most energy-efficient types of lasers that are deployed in practical applications. Even though VCSELs as energy efficient as 50 fJ/bit have been demonstrated in research laboratories,^25–27 those deployed in practice are much less energy-efficient. Typical VCSELs have diameters of $2\mu \mathrm{m}$ to $10\mu \mathrm{m}$ with a total DBR thickness of $5\mu \mathrm{m}$ to $10\mu \mathrm{m}$, resulting in a total device volume of 10 to 100 times the wavelength cubed, although the modal volumes are much smaller. Further significant reduction in sizes and energy usage are difficult to achieve because of the poor heat dissipation through thick DBR mirrors and the poor lateral confinement of the VCSEL structures. Thus, it is very difficult for VCSELs to meet the long-term requirements of the size and energy-efficiency challenge for on-chip applications, despite great improvements in recent years.²⁷

3.2.

Microdisk Lasers

Microdisk lasers were initially developed^28–30 in attempts to miniaturize semiconductor lasers for integrated photonics applications using a free-standing piece of semiconductor based on whispering-galley modes due to the high contrast of refractive indices between semiconductor and air. Such index contrast provides the strongest mode confinement realistically possible without the use of metals or thick DBRs. The unique structures allowed much smaller lasers than possible at the time. The initial devices had diameters between $3\mu \mathrm{m}$ and $5\mu \mathrm{m}$ with a thickness of approximately one micron, resulting in a total device volume of several times the wavelength cubed. Recent work³⁰ combining the photonic crystal (PC) structure with a microdisk allowed a total device volume of only a small fraction of a wavelength cubed, one of the smallest total volumes achieved (see Fig. 4). Significant progress has also been made recently in demonstrating micro- or nano-disk lasers using perovskites.^5,31 Even though perovskites show remarkable optoelectronic properties and may find many interesting applications elsewhere, their use for on-chip interconnects is hindered by several issues such as incompatibility with traditional microfabrication and the lack of perovskites that emit at near-infrared communication wavelengths (e.g., $>850\mathrm{nm}$).

3.3.

Photonic Crystal Lasers

Photonic crystal lasers^30,32–38 are promising candidates for use in energy-efficient applications because they have very small volumes of optical modes and represent a type of laser based on a very fruitful wave-analogy between electrons and photons.^39,40 These lasers can have very low thresholds as a result. Recently, direct modulation at 5.5 GHz at room temperature via optical pumping has been demonstrated.^33,37 The energy efficiency was estimated to be 13 fJ, with one of the lowest threshold pumping powers, estimated at $1.5\mu \mathrm{W}$. Although the defect modes that define the cavity in a PC laser can be extremely small, the overall sizes of PC structures can have diameter or side length on the order of $10\mu \mathrm{m}$, resulting in a total device volume that ranges from several to 100 times the wavelength cubed, as shown in Fig. 4. Such lasers are of great importance both for studying basic quantum optics of nanolasers and for potential use in integrated photonic applications because of the ability to achieve unprecedented tight three-dimensional (3-D) mode confinement. The total device sizes including the large periodic dielectric structures are very large, however.

3.4.

Nanowire or Nanopillar Lasers

Semiconductor nanowires or nanopillars in air provide one of the best semiconductor optical cavities via the large index contrast (similar to that of microdisk lasers). As with the microdisk lasers, the mode confinement is much better than in typical double-heterostructures, with the possibility of achieving a confinement factor of $>1$. Such nanowires with the two ends exposed to air provide a unique structure both for high-reflective laser cavity and a gain medium at the same time, an ideal combination for laser miniaturization.⁴¹ After the initial wave of laser demonstration in the UV and visible wavelength regime,^42–44 the first near IR lasing⁴⁵ was demonstrated using a single GaSb nanowire at the telecom wavelengths. Recently, several realizations of lasing were demonstrated in the short wavelength NIR regime.^13,46,47 To date, almost all of these demonstrations of nanowire lasing were achieved under optical pumping, with only one exception.⁴⁴ The ranges of volumes of nanowire lasers are currently comparable to those of microdisk lasers and PC lasers (see Fig. 4), with the total size as small as twice the wavelength cubed in the IR range.⁴⁵ However, further miniaturization of nanowires is possible. One additional advantage of nanowire lasers is the combined material and bandgap flexibility,² which allows realization of lasers at wavelengths that are difficult to achieve using either the microdisk approach or the PC approach. Many recent review articles^3,4,48–53 are available on nanowires and the nanowire-based lasers.

3.5.

Two-Dimensional Material Nanolasers

With the reduction of laser cavity size or volume, the cavity quality factor decreases. This decrease in cavity quality factor occurs for both pure dielectric cavities and metallic or plasmonic (see Sec. 4) cavities, albeit according to different scaling laws.⁵⁴ Therefore, it is important to constantly search for better gain materials that provide high optical gain within a small volume. In this regard, the newly emerging 2-D materials, such as transition metal dichalcogenides (TMDCs), show great promise. One of the key features of these 2-D materials is the extremely large exciton binding energy (generally one or two orders of magnitude larger than conventional semiconductors). This feature allows 2-D-based materials to provide high optical gain at room temperature or higher through exciton-related emissions before the Mott transition, rather than through the typical quasifree electron-hole plasma used in conventional semiconductor lasers. The possibility of excitonic gain is extremely appealing for low power applications since it requires much lower levels of carrier density than electron-hole plasma. Even for plasma gain, strong excitonic enhancement of optical gain will also be beneficial. Another advantage is the monolayer thickness of the 2-D materials, thereby providing the necessary flexibility for integration with different substrates, especially Si substrates, to address the integration challenge. The much larger tolerance to strain and other mismatched mechanical properties could be very valuable compared to conventional III-V materials for Si integration. The monolayer also provides the thinnest gain materials with an overall small volume, important for low input energy. Lasing demonstrations using 2-D materials have been reported in a microdisk cavity,⁵⁵ and in 2-D⁵⁶ and 1-D⁵⁷ PC cavities; such demonstrations have been achieved at low temperatures^55,56 and at room temperature for IR wavelengths that are Si transparent.⁵⁷ Strong and weak coupling to cavity modes^58–60 has also been studied. As mentioned, the monolayer thickness combined with a nanoscale cavity⁵⁷ leads to a small overall laser with a mode volume smaller than $0.5{(\lambda /n)}^3$ and a gain medium volume $\sim {(85\mathrm{nm})}^3$. All these characteristic parameters are quite attractive for energy efficient applications.

4.1.

Progress Overview

Plasmonic lasers or spasers^61–65 are the newest member of the laser family; such lasers can realize the smallest sizes of any lasers and thus can be potentially energy-efficient lasers when small size is essential. Plasmonic lasers utilize the coupling between photons and plasmon excitation at metal–dielectric interfaces to confine photons to the smallest possible spatial volumes. The pioneering proposal of spasers by Bergman and Stockman⁶¹ was to create a stimulated plasmonic source; it soon became clear that the unavoidable far-field coupling could be utilized to allow the stimulated plasmonic source to emit coherent laser light to serve as an ultrasmall laser. Plasmon polariton propagation was earlier studied for the purpose of prolonging the propagation length.^64,66,67 A plasmonic nanolaser design with the explicit purpose of achieving a miniaturized laser was proposed in the form of a semiconductor core in a silver metal shell.⁶² The design and simulation showed that, despite the expected high plasmonic loss, the high level of optical gain in the semiconductor core could overcome the associated high threshold loss. Afterward, a similar core–shell-structured laser was soon demonstrated using a traditional top-down lithographical fabrication based on InP, demonstrating the first metallic cavity core–shell nanolaser.⁶⁸ Although this demonstration was impressive in terms of fabrication quality down to such a small size, the operation modes were not plasmonic, but conventional dielectric modes.

One of the first demonstrations of plasmonic mode lasing⁶³ was reported soon afterward; such lasing, which is closer to the original proposal,⁶² involved the use of an InP/InGaP/InP core and silver shell, with the core having a rectangular cross section etched from an InP-InGaAs wafer with thickness varying between 80 and 340 nm. The device operated in the so-called plasmonic-gap mode, with a semiconductor core as thin as 80–90 nm. The entire optical thickness of the device, including the SiN insulating layer and the metal penetration, is $\sim 400\mathrm{nm}$, which is much smaller than the half-wavelength (670 nm) of the operating laser emission. This work represents the first laser with at least one dimension below the “diffraction limit,” a result that is impossible to achieve using a pure dielectric mode. Simultaneously, other forms of spaser or plasmonic nanolaser operation were demonstrated using an Au core coated by a dye-doped dielectric outer shell,⁶⁹ or a CdSe nanowire placed on a silver screen separated by a thin dielectric film.⁶⁵ Note that both Refs. 63 and 65 are based on the same plasmonic gap modes, even though the gap in Ref. 65 is much smaller. After these initial demonstrations of plasmonic nanolasers or spasers, the field has flourished, with a wide array of designs reported,^70–82 for which the plasmonic or metallic structures feature prominently as the mechanism for light confinement. Such a fundamental change, from the traditional dielectric/semiconductor cavities to metallic or plasmonic cavities, represents a potential paradigm shift in laser cavity design. These plasmonic cavity designs include metallic co-axial structures,^71,81 metallic nanopan structures,⁸⁰ metallic structures combined with VCSELs, metallic trench FP cavity, and metal-grating DFB structures.^74,75 Although plasmonic laser operation under optical pumping dominates the research literature, electrical injection devices^{54,63,68,77,83,84} have been studied from the very beginning, based on semiconductor wafer structures that are more compatible with the standard III-V fabrication technologies. Room temperature operation of metallic cavity lasers under electrical injection has also been demonstrated,^77,83 indicating potential applicability of such lasers for practical applications. Linewidth of plasmonic nanolaser is an interesting issue and has been studied recently.⁸⁵ Many good review articles in the literature provide more complete perspectives on plasmonic nanolasers from the perspectives of various groups.^{49,54,84,86–91}

4.2.

Benefits of Plasmonic Nanolasers

Several important advantages of plasmonic nanolasers that are relevant to size and energy efficiency are worth mentioning here. One advantage is the size or volume. When comparing the sizes or volumes of various lasers, it is important to note that there are three types of volume that are relevant: the volume of the active region, the modal volume, and the total volume of devices. The volume of the active region determines the total number of electron-hole pairs that need to be injected. For a given transparency carrier density, the smaller the volume of the active region is, the smaller the threshold current, and thus, the smaller the total electrical energy input (I-V product). For low energy operation, a small active region is preferred. The small modal volume often results in a large confinement factor if the optical modes enclose the active region. A large confinement factor corresponds to a large modal gain. The total device volume is often limited by the precious availability of real estate on chip for integrated photonics applications and is also related to the heat dissipation efficiency. Thus, all three volumes must be small for lasers to be used as on-chip light sources. An efficient small laser will preferably have all three volumes as small and as closely equal as possible. Plasmonic lasers can be designed to have the smallest values of all three volumes among all the proposed types of nanolaser.⁹² For example, PC lasers, which have small modal volumes, often have large volumes of the active region and large total device volumes. Currently, most of the plasmonic lasers demonstrated have not been miniaturized in all three dimensions and can be further optimized to further reduce sizes in one or two more dimensions.⁹²

Because energy efficiency is defined as the energy consumed per bit of information transmitted, energy efficiency benefits greatly from high-speed operation. Purcell enhancement (see Fig. 6) in plasmonic devices is the strongest among all the proposed structures. Such enhancement leads to higher laser modulation speed. Extremely high speed of operation was predicted (up to THz) theoretically.^90,93 Ultrafast experiments indicate enhanced gain switch and recover times.⁸² The realistically achievable speed is, however, much lower.^22,93–96 Modulation at hundreds of GHz appears to be realizable;^22,95 such modulation speed is already much higher than that of typical semiconductor lasers. Significant improvement in modulation speed is expected with further improvements in plasmonic laser design and fabrication.

Low lasing threshold is generally desired for low power consumption. This might lead people to believe that plasmonic or metallic cavity lasers are unsuited for energy-efficient applications. However, low threshold alone does not always translate into high energy efficiency. Many conventional lasers, such as VCSELs, have extremely low thresholds; however, their energy efficiency is too low for future integrated photonics applications, as mentioned earlier. Figure 3 shows that small sizes are important for low energy operation. However, when we demand both small size and low threshold, the situation is somewhat more complicated. Figure 5(b) shows a comparison of the $Q$-factor for modes with wavelength of $\sim 1500\mathrm{nm}$ for two modal laser cavities:⁸⁴ dielectric disk cavities with and without metallic shell. Clearly, a dielectric cavity without a metal shell displays a much more rapid decrease of $Q$ with the decrease of diameter than that with a metal shell; this difference is caused by the rapid increase of far-field radiative loss with a decrease in diameter, as mode confinement deteriorates. In contrast, the cavity with metal shell shows a much slower decrease compared to the cavity without a metal shell. The two curves cross over near 1000 nm, indicating an overall benefit in using a metallic shell cavity for smaller (submicron) devices, due to reduction of far-field radiative loss, despite the large metal loss. A trade-off clearly exists between far-field radiative loss and internal absorption loss. For a smaller device, the metallic cavity is an overall better choice for lower threshold and higher energy efficiency (Fig. 3), contrary to the intuitive arguments.

Fig. 5

Comparison of a dielectric cavity and a metallic cavity. (a) The dependence of the cavity $Q$ factor on its diameter for two cavities: a pure dielectric cavity and a dielectric cylinder with a metal shell (adapted from Ref. 84). (b) Laser performance comparison of a semiconductor pillar cavity [denoted with (D)] and a semiconductor pillar cavity with metal shell [denoted with (M)] (from the supporting information of Ref. 22): the laser output power (P) and temperature (T) are shown.

Heat dissipation is another important factor for high-density on-chip integration and thus must be considered when comparing various lasers. Many of the other proposed laser miniaturization solutions suffer from poor heat dissipation, such as nanowire lasers, PC lasers, and microdisk lasers, because of the long or poor heat conduction channels involved in these lasers. For example, the airgaps involved in most of the PC lasers prevent a more efficient downward heat dissipation to the substrate. Plasmonic lasers provide potential advantages in this respect via the close proximity of metal shell to active region^22,63 and much improved heat conduction by the metal. These advantages are illustrated in Fig. 5(b). As discussed in detail in Ref. 22 (see Supporting Information there), two semiconductor lasers of cylindrical pillar shape with and without the metallic shell were considered in numerical simulation. The two lasers have the same diameter of 920 nm. The $Q$-factor considered is 280 and 421 for nanolaser with and without metal shell, respectively. Note that intentional choice of somewhat larger size favors the nanolaser without metal shell because it has higher $Q$ and thus lower threshold. As seen in Fig. 5(b), the laser without the metal shell indeed turns on at lower pumping. However, at higher pumping levels, the poor heat dissipation leads to significant heating, with a temperature increase as high as 100 K above the substrate temperature. Overheating eventually leads to device shut off. In contrast, the laser with a metal shell shows only a modest temperature increase; as a result, it has much higher output power at high pumping levels. Ability to operate at high driving current is indeed an important advantage of plasmonic lasers compared to conventional semiconductor lasers.

Another counterintuitive result is the plasmonic enhancement of optical gain due to an unusual feature of confinement factor near plasmonic resonance,⁹⁷ resulting in slowing down of energy propagation. This slow energy propagation leads to a giant enhancement of optical gain,⁹⁸ leading to much higher optical gain than material gain. Such giant optical gain could lead to a significant decrease of laser threshold. However, the slowed-down energy propagation may affect the modulation speed when extreme high speed is desired. A detailed optimization of the trade-off between low threshold and high modulation speed is required to achieve optimal performance.