You are on page 1of 9

B36 J. Opt. Soc. Am. B / Vol. 27, No. 11 / November 2010 Martin T.

Hill

Status and prospects for metallic and plasmonic


nano-lasers [Invited]

Martin T. Hill
Department of Electrical Engineering, Technische Universiteit Eindhoven, Postbus 513,
5600 MB Eindhoven, The Netherlands (m.t.hill@ieee.org)

Received July 7, 2010; revised August 22, 2010; accepted August 24, 2010;
posted September 3, 2010 (Doc. ID 131276); published October 12, 2010
A remarkable miniaturization of lasers has occurred in just the past few years by employing metals to form the
laser resonator. From having minimum laser dimensions being at least several wavelengths of the light emit-
ted, many devices have been shown where the laser size is of a wavelength or less. Additionally some devices
show lasing in structures significantly smaller than the wavelength of light in several dimensions, and the
optical mode is far smaller than allowed by the diffraction limit. In this article we review what has been
achieved then look forward to what some of the directions development could take and where possible appli-
cations could lie. In particular we show that there are devices with an optical size slightly larger or near the
diffraction limit which could soon be employed in many applications requiring coherent light sources. Appli-
cation of devices with dimensions far below the diffraction limit is also on the horizon, but may take more time.
© 2010 Optical Society of America
OCIS codes: 250.5403, 250.5960.

1. INTRODUCTION into the metal, which is typically of the order of tens of


There has been a steady push to further miniaturize la- nanometers. Until recently many researchers doubted if
sers and other photonic devices, as there are benefits it would be possible to overcome the losses in plasmonic or
similar to those seen with the shrinking dimensions of metallic waveguides and cavities with the currently avail-
electronic components. A smaller laser requires less able optical gain materials. Unfortunately, the metal’s
power and can potentially be switched on and off faster. conduction electrons, which oscillate in synchronism with
An excellent example of this process is the Vertical Cavity the optical field, dissipate energy through collisions with
Surface Emitting Laser (VCSEL) [1]. This was the first la- the metal’s atomic lattice. This energy dissipation leads to
ser with a wavelength scale optical mode and it is being high optical losses, so SPPs can travel only short dis-
employed in an increasing number of applications. How- tances, and metal mirrors have higher losses than dielec-
ever, in the past decade, the main thrust of research in tric ones. However, a number of workers did examine the
nano-lasers has been into photonic crystal cavity lasers possibility of using gain in metallic waveguides [5–7].
[2]. Here the light is confined strongly in two dimensions However, in the past couple of years efforts to use met-
by what are effectively two-dimensional Bragg mirrors als to form the nano-laser resonator have allowed both
formed by making a pattern of holes in a high-refractive- the overall size of the laser to be reduced to smaller than
index semiconductor. In all of these cavities the refractive the wavelength of light, and also the optical mode dimen-
index differences of dielectric materials are used to con- sions to be reduced below the diffraction limit. The
fine the light. The smallest dimensions to which the opti- progress that has been made by many groups in just a few
cal mode of such cavities can be confined is related to the years has been remarkable. Some of these devices are
diffraction limit, and is of the order of one half of the coming close to being useful light sources, and it may be
wavelength of light in the dielectric material. Addition- only a few years before we see lasers based on metallic
ally, the dielectric mirror structure typically has to be nano-structures in applications. In this article we review
many wavelengths in size in order to be effective, imply- the progress that has been made in metallic and plas-
ing that the total laser size is still large compared to the monic nano-lasers. Next we look at some of the significant
wavelength of light. advantages that the use of metals brings to nano-lasers
In theory it is possible to confine light with the use of and how they may be employed in useful devices or to fur-
metals to form the laser resonator. The metal may form ther knowledge. Some of the most promising forms of
either strong compact mirrors, able to confine light to metal-based nano-lasers have semiconductor gain media
about the size of the diffraction limit, or the light may in- as their basis. Semiconductor gain media can be electri-
teract strongly with the free electrons in the metal, being cally pumped efficiently, which will be of major impor-
guided in the form of surface plasmon polaritons (SPPs) tance for applications. We will look at a couple of the ma-
at the interface between a metal (conductor) and a dielec- jor forms possible for electrically pumped devices and
tric (non-conducting) material such as air [3,4]. The mini- highlight some of the directions that research and devel-
mum size scales of structures employing this form of light opment into these devices could take.
confinement is related to the penetration depth of light Metallic or plasmonic nano-lasers will likely become

0740-3224/10/110B36-9/$15.00 © 2010 Optical Society of America


Martin T. Hill Vol. 27, No. 11 / November 2010 / J. Opt. Soc. Am. B B37

technologically significant, particularly when the laser evidence of reduced optical losses in the metals has been
and mode size can be reduced far below the diffraction reported at low temperatures [8,10] due to reduced pho-
limit. Briefly, some of the reasons for this significance are non scattering of free electrons. The light in the cavity
as follows: (1) There is a perceived need for small, fast, was tightly contained in a region with dimensions ap-
low-power lasers to cope with the increasing bandwidth of proximately one half the wavelength of light in the me-
communications inside and between integrated circuits. dium, and the mode was a hybrid electric (HE11) mode
(2) The ability to make arrays of coherent emitters with [11]. Further developments of this pillar concept were
sub-wavelength pitch may open up new applications. (3) shown in 2009 with lasing in metal–insulator–metal
These metallic devices could offer improved light emitters (MIM) waveguides with a core width approximately one
compared to dielectric cavity devices, particularly for the quarter the wavelength of light in the medium, or about
longer wavelengths. (4) Having ultra-high-speed low half the diffraction limit, and propagating a gap-plasmon
power lasers may open up the possibility for optical signal or transverse magnetic (TM) mode [12]. Furthermore
processing that is competitive with electronics in the room temperature lasing was also shown in slightly wider
high-performance region.(5) Finally these devices will waveguides that propagated a transverse electric (TE)
also provide important sources and amplifiers for sub- mode.
wavelength plasmonic circuits. A number of other groups also showed in 2009 nano-
lasers made from metal semiconductor structures lasing
in the near infrared. In particular a design emerged based
2. STATE OF THE ART FOR METALLIC/ on a metal/dielectric encapsulated semiconductor hetero-
PLASMONIC NANO-LASERS structure that showed room temperature operation for an
The first report of lasing in metallic nano-structures oc- optically pumped device that was less than a free space
curred in 2007 [8]. Here a round semiconductor pillar en- wavelength of light in all three dimensions [13]. This de-
capsulated in a dielectric and then gold was pumped elec- vice exploited a TE mode to obtain a higher cavity Q and
trically (Fig. 1(a)). The pillar had an overall diameter of lower lasing threshold. Others at the California Institute
approximately 260 nm and an InGaAs active region of Technology showed evidence of lasing in a microdisk
height of 300 nm. The presence of the InGaAs hetero- cavity having one side coated with metal [14]. Another
structure in the pillar and the metal formed a resonator group showed lasing at cryogenic temperatures in a small
with an optical mode trapped on the InGaAs gain medium patch of metal/semiconductor/metal sandwich named a
(Fig. 1(b)). At cryogenic temperatures it was possible to nano-patch laser [10]. Both TE and TM lasing modes were
obtain sufficient gain in the InGaAs to overcome losses in observed.
the resonator and achieve lasing at a wavelength of ap- Meanwhile, with a different approach and at far
proximately 1400 nm. In semiconductor lasers, operation shorter wavelengths, researchers at the University of
is often first demonstrated at low temperatures due to the California at Berkeley, the Lawrence Berkeley National
ease of obtaining higher optical gains [1,9]. Additionally Laboratory, and Peking University showed lasing in a sec-
tion of a hybrid dielectric/plasmonic waveguide [15]. The
waveguide is constructed from a flat metal substrate (sil-
ver) coated with a thin dielectric layer on which a thin
semiconductor nano-wire is placed. The optical mode
propagating in the silver/dielectric/semiconductor wave-
guide has a significant amount of its energy squeezed into
the thin dielectric gap between the nano-wire and the
metal substrate, leading to a highly localized mode which
has, according to calculations, an area that is significantly
below the diffraction limit. The tail of the propagating
mode overlaps with the semiconductor nano-wire, which
acts as the gain medium. The nano-wire forms a Fabry–
Perot cavity with the plasmonic modes resonating be-
tween its two ends (which are a few micrometers or more
apart). Using short optical pulses to pump the semicon-
Fig. 1. (Color online) A number of groups have made metallic or ductor nano-wire, lasing is shown in the Fabry-Perot cav-
plasmonic nano-lasers recently. a. One approach that has been
pursued by several groups is to encapsulate a semiconductor het- ity at low temperatures in the blue part of the visible
erostructure pillar in an insulator, then a low-optical-loss metal spectrum.
such as silver or aluminium. Such an approach is particularly However, probably the most remarkable nano-laser
suitable for electrical pumping of the laser. The InP / InGaAs het- was reported by researchers at Norfolk State University,
erostructure shown is used with an SiN insulator. Some light in
Purdue, and Cornell. They show lasing in a gold nanopar-
the optical mode escapes through the bottom of the pillar and the
substrate. b. The optical mode in such structures is trapped on ticle with a diameter of 14 nm that is surrounded by a
the InGaAs gain region in the center of the pillar due to refrac- silica shell containing dye molecules that exhibit optical
tive index differences. The contour lines show 兩E兩2 of the trapped gain (to give an overall diameter of 44 nm) [16]. The gold
optical mode in a slice through the pillar center, with color giving nano-particle that forms the optical cavity has a plasmon
intensity. The arrows show the magnitude and direction of the
time averaged Poynting vector for this slice of the mode, and
resonance at a wavelength of ⬃520 nm. When the dye
show that the energy of the mode is mostly being dissipated as molecules are pumped by short optical pulses, there is
heat in the metal sidewalls. sufficient gain created to overcome the losses in the gold,
B38 J. Opt. Soc. Am. B / Vol. 27, No. 11 / November 2010 Martin T. Hill

and a coherent source of highly localized SPPs is created. 4. ELECTRICALLY PUMPED


It is truly remarkable to demonstrate such a laser-like de- SEMICONDUCTOR DEVICES
vice with a cavity, gain medium, and optical mode that
In looking further at possible applications and develop-
are all an order of magnitude smaller than the wave-
ment directions for metallic/plasmonic lasers we will fo-
length of light in all three dimensions. This device may be
cus on electrically pumped devices, and in particular the
the first example of a spaser—surface plasmon amplifica-
encapsulated heterostructure devices. The reason for this
tion by stimulated emission of radiation [17].
is that these devices are perhaps the most promising for
As technology and designs improve there have been
electrical pumping, which will be important for many ap-
more demonstrations of devices, and in particular more
plications. Furthermore a number of groups have already
devices working at room temperature, which is important
successfully made lasers employing this concept or some
for applications. One device of note is an encapsulated pil-
variation on it [8,12,13,18]. We will look at the possibili-
lar device that also incorporates a Bragg mirror in the
ties of two broad classes of devices that have already been
base of the pillar [18]. This device, although still with a
demonstrated: (1) isolated nano-cavity devices that func-
large diameter of 2 ␮m (further size reduction is pos-
tion with light emission vertical to the device substrate
sible), does show continuous room temperature lasing
wafer, somewhat similar to the VCSEL; (2) devices that
with electrical pumping. Hence, some forms of metallic or
are made as part of a wave guiding system; in particular
plasmonic nano-lasers may soon be suitable for applica-
we look at the MIM waveguide systems [21,22].
tions.

A. Surface Emitting Type Devices


3. POTENTIAL ADVANTAGES A number of groups have shown isolated laser devices
The use of metals to construct all or part of the resonator that function by emitting their light through the sub-
for a nano-laser has a number of significant advantages strate of the device wafer [8,13,18].
compared to the use of dielectrics: Typically these devices have a circular pillar cross sec-
Electrical contacts and connections for pumping the tion, though other shapes are possible. The devices func-
gain medium can now form part of the laser itself instead tion by trapping the optical mode on the higher-index cen-
of being kept at a distance as in the case of most ter region of the pillar. Above and below this region there
dielectric-based nano-laser designs [19]. The metal cavity is no propagation, as the metal waveguide is in cutoff at
may also serve as part of a heat sink. Heating is one of the the operating frequency. In these devices the energy in
major limiting factors of nano-lasers [19]. Often, low- the optical mode is traveling back and forth between the
index materials that are poor heat conductors surround pillar sidewalls. Figure 1(b) shows the Poynting vector of
the dielectric cavity. In contrast, metals like silver have such a mode, and this movement of energy into the side-
many times higher thermal conductivity than the most walls can be seen. The modal energy is eventually dissi-
thermally conductive optical semiconductor materials, pated as heat in the metal. However, some of the energy
and orders of magnitude higher conductivity than most from the optical mode can escape to the substrate due to
low-index materials used to obtain good confinement in the finite length of the waveguide section under the gain
dielectric nano-lasers. This thermal advantage is particu- region.
larly strong for the case when the device is formed via an More recently, metal encapsulated devices that are
encapsulated heterostructure, as discussed later. Here, closer to a traditional VCSEL have been demonstrated.
the nano-scale gain medium is surrounded by a solid Here the confinement of the mode in the vertical direction
metal structure that can be made arbitrarily large. is aided by the use of Bragg mirrors and the silver top
In the cases mentioned in the introduction a number of contact [18].
widely varying approaches were used to create the cavi- To make a useful laser device the energy from the opti-
ties for the various devices. Furthermore the form and cal mode must be emitted into free space and collected for
confinement of the mode was widely varied, including the the particular application. To obtain the highest efficiency
ability to make resonator designs with dimensions below for the device it is desirable that the smallest proportion
the diffraction limit. This is a particular advantage of us- of the optical mode energy be dissipated as heat in the
ing metals. Compared to dielectric reflectors like Bragg metal, and the greatest proportion be emitted out of the
gratings or photonic crystals, metals can form strong cavity through the substrate. For the devices relying on
broadband reflectors with minimum dimensions of the or- trapped cutoff modes to obtain the vertical confinement,
der of a few tens of nanometers. Such reflectors can be there is a tradeoff between the maximum cavity quality
used in widely varying cavity or waveguide designs. Ad- factor Q and the amount of light emitted to the external
ditionally, in designs where the metal cavity forms part of environment, just as in most laser devices [23]. By mak-
a larger metal structure, higher optical intensities may be ing the length of the cutoff waveguide under the active
supported than with dielectric mirrors. layer shorter, more of the modal light can escape the cav-
The metal nano-laser’s small size, electrical connec- ity. Figure 2(a) shows the structure of a surface emitting
tions, powerful heatsinking, and, if desired, a suitably pillar device. This device operates with a TE01 resonant
short photon lifetime in the cavity allows a number of pos- mode, a plot of which is given in Fig. 2(b). Figure 2(c)
sibilities. In particular, the material can be pumped con- shows a plot of the cavity Q for various lengths of the InP
tinuously at very high pump densities, thus achieving waveguide below the InGaAs gain region h. The device
very high material gains and also possibly extremely high has a circular center gain region of InGaAs 440 nm in di-
modulation bandwidths [20]. ameter and 300 nm thick, with a silicon dioxide insula-
Martin T. Hill Vol. 27, No. 11 / November 2010 / J. Opt. Soc. Am. B B39

Fig. 2. (Color online) a. Slice through silver encapsulated semiconductor core pillar device showing details of its structure used in a
simulation. The height of the InGaAs core is 300 nm. The light gray region between the InGaAs core and silver is the SiO2. The substrate
below the pillar is InP. The height parameter h can be varied to modify the Q of the cavity, allowing more or less light from the cavity
mode to leak to the outside of the cavity. Typically electrical contacts to these pillar devices to pump the InGaAs gain region can be made
through the top of the pillar via the actual silver encapsulation, and another contact to the other side of the InGaAs gain region via the
InP substrate. b. FDTD simulation results showing 兩E兩2 plot of a slice through the pillar indicating a TE mode centered on the InGaAs
and leaking power to the substrate. c. Plot of quality factor of cavity versus length of InP stub under the InGaA shows tradeoff can be
made between Q and emission efficiency.

tion layer 80 nm thick. For a long InP waveguide section forced to lase far from its band-edge. Figure 3 shows the
the Q is approximately 1400. For an optimized choice of spectra from a series of round pillar devices that were fab-
InGaAs insulation layer thickness, maximum Qs of the ricated next to each other on the same wafer and that op-
order of 3000 are possible [13]. For shorter and shorter erate with a HE11 mode. Here the devices are very simi-
sections of waveguide the Q decreases, indicating that lar to those reported in [8], with the main difference being
more and more energy is escaping to the substrate. Room that the gold used for the cavity in [8] was replaced with
temperature lasing in devices with Qs of the order of just silver. The devices are operated at 78 K, with continuous
a few hundred has been shown [12,18]. Hence it is pos- electrical pumping of between 30 and 60 ␮A. The diam-
sible in such devices to make an intelligent tradeoff be- eter changes between each device by approximately
tween Q and confinement in order to obtain efficient gen- 20 nm. It can be seen that the nano-laser wavelength can
eration of lasing light. be widely varied while using the same substrate and gain
Coupling the emitted light into a narrow beam with the material.
desired direction, or perhaps some waveguide, is another This wavelength tuning could be used to advantage for
step that needs to be addressed and will certainly form an certain applications when compared to more conventional
important line of research. It can be seen from Fig. 2(b) VCSEL designs. In VCSELs the epitaxially grown Bragg
that the beam maxima propagate into the substrate at mirror may to a significant extent limit the range of las-
about 45° to the vertical. In principle a high-numerical- ing wavelengths. Additionally the metal used to contain
aperture lens should be able to capture most of the emit- light in these devices is effective over a wide wavelength
ted light. Optimization of the cavity shape for the TE01 or range from the blue in the visible to the far infrared. Ef-
other modes could possibly produce more directed beams.
Also, it has been shown that light from sub-wavelength
apertures can be emitted in narrow beams [24]. For the
waveguide type metallic devices discussed in the next
subsection, there are already schemes for efficient cou-
pling of light from metallic waveguides to dielectric
waveguides, and these will be mentioned there.
The TE01 mode pillar device described above is just one
example of how metallic/plasmonic nano-lasers could be
employed as efficient coherent light sources. As men-
tioned earlier, there is considerable freedom in the form,
size, and how metals are incorporated into a cavity, so
there will be many approaches to making efficiently emit-
ting nano-laser devices [25]. Another aspect is that these
cavities have been shown to have high Purcell factors
[8,13], which can dramatically increase the emission effi-
ciency for light emitting diodes (LEDs). It may be that for
some applications, lasing is not required, rather the en-
hanced spontaneous emission is a sufficient advantage
[26]. Fig. 3. (Color online) Three spectra from three separate pillar
In the devices that rely on a cutoff mode to confine the devices in a row. The devices operate with a HE11 mode and the
light vertically, the resonant wavelength is determined by diameter of the pillar differs by approximately 20 nm between
each device. The plot shows that single-mode lasing over a wide
the shape and size of the cavity. For small cavities, an ex- range of wavelengths can be achieved with devices on the same
tremely large free spectral range can be achieved, hence wafer and close to each other. The spectra have been offset by
only one mode can lase, and the gain medium can be 3000 counts from each other for clarity.
B40 J. Opt. Soc. Am. B / Vol. 27, No. 11 / November 2010 Martin T. Hill

fective Bragg mirrors on the other hand can be difficult to of the few metallic structures that permit true deep sub-
realize at all wavelengths, so for some wavelengths the wavelength confinement of light is the MIM waveguide
metal-confined device may be simpler to produce. structure [21,22]. Here the light can be confined inside an
Like VCSELS, these surface emitting metallic nano- arbitrarily small insulator layer between two metal slabs.
lasers can be formed into arrays of emitters. However Penetration into the metal on each side of the insulator is
they have a number of significant advantages. Dielectric- typically also limited to a few tens of nanometers [22], al-
based cavities cannot be formed into arrays of indepen- lowing in principle tightly packed waveguides [28].
dent devices spaced less than a few wavelengths apart. It There have been a number of experiments involving
has been suggested that metallic nano-lasers will permit placing optical emission and gain material inside MIM
arrays of independent devices with sub-wavelength spac- waveguide structures. The most obvious method to make
ing. Furthermore, in contrast to conventional VCSEL de- MIM waveguides is to deposit metal, then insulator, then
vices, the wavelength of the individual devices in an array metal onto a flat substrate. This construction method has
can be varied at will over a large range in one step at the been used to make surface plasmon emitting devices
time of manufacture just by changing the pillar size. [29,30], where the insulator between the metals has been
Some of the suggested applications for such arrays are for replaced by light emitting material. For far infrared
atom traps, sub-wavelength scanning [13], and inter-chip wavelengths this technique has also been successful in
communications [18]. making lasing devices [31] with semiconductor gain ma-
Another emerging application of VCSEL arrays are for terial. In all these devices the confining metal plates also
high-power laser sources [27]. Arrays of metallic nano- serve as electrical contacts. For far infrared devices the
structures may turn out to be superior to VCSELs in semiconductor layer thickness is several micrometers and
these applications, due to their inherent high quality allows for specially doped layers to give efficient electrical
heatsinking. contacts. However, for shorter wavelengths, where the
In summary, the combination of sub-wavelength size semiconductor layer thickness is reduced to a few tens of
and packing density, powerful control over wavelength, nanometers, providing both good contact layers and gain
and prospects for good heat sinking may make these types layers may be difficult.
of devices attractive for many applications of lasers and Our approach has been to place the MIM waveguide on
light emission. its side by etching long thin waveguide structures
through the epitaxially grown substrate. Then we evapo-
B. Waveguide Type Devices rate metal onto the sidewalls of the pillar. The refractive-
The surface emitting devices discussed above will mostly index difference between the InP and InGaAs isolates the
likely exploit higher-Q oscillatory modes inside the metal mode on the InGaAs gain region [32]. This configuration
cavities [11]. Such modes have minimum sizes close to the has the advantage that the electrical contacts that inject
diffraction limit 共␭ / 2 n兲. The unique property of metal the carriers into the gain InGaAs region are located away
cavities is that light can be confined to regions smaller from the tightly confined optical mode. Furthermore a
than the diffraction limit. The high intensity of the elec- thin insulation layer isolates the metal from the semicon-
tric field that occurs at the metal–dielectric interface can ductor gain medium (Fig. 4(a)).
lead to a modal volume much smaller than the diffraction We have constructed sections of these waveguides ter-
limit [4]. However, the optical field still extends distances minated in metal mirrors to form Fabry-Perot cavities.
of the order of a wavelength away from the interface, This is achieved simply by totally encapsulating a pillar
placing a limit on how closely devices can be packed. One that has a thin rectangular cross-section in silver. The di-

Fig. 4. (Color online) a. Electrically pumped MIM waveguide structure that has the usual flat MIM configuration turned on its side. The
propagating mode is confined to the InGaAs gain region in the center of the waveguide by index differences. The dark (red online) region
under the top n-contact is n-InGaAs employed to give low contact resistance. The gold/gray layers to the side and at the bottom of the
ridge are Au/ Ti layers employed to help bond the silver encapsulation to the semiconductor/dielectric core. The schematic shows a cut
through the ridge structure. In actual devices the ends of the ridge are also encapsulated in dielectric and silver of the same thickness
as the sidewalls. b. Plot of 兩E兩2 for the gap-plasmon mode taken horizontally across the waveguide structure for a 25 nm wide InGaAs
region and 5 nm thick SiN insulator. The lighter lines (blue online) are for the real refractive indices of InGaAs and SiN, darker (red
online) when the SiN has its refractive index equal to that of InGaAs. A significant proportion of the modal energy can travel in the
insulator, which can increase the gain requirements from the InGaAs. c. SEM photos of sections of waveguide cores; after being encap-
sulated in silver they will form Fabry–Perot cavities. The scale bar is 1 ␮m.
Martin T. Hill Vol. 27, No. 11 / November 2010 / J. Opt. Soc. Am. B B41

electric and silver layer thickness was the same all round dia with sufficient optical gain and sufficient overlap with
the pillar. Some light from the lasing mode could escape the lasing mode to overcome the optical losses. To place
through the substrate to show what was occurring inside the minimum requirements on the optical gain medium
the cavity. For improved light extraction the waveguide requires that the optical mode energy have the highest
could be cut to remove the silver encapsulation at the end overlap with the gain medium. Unfortunately as can be
of the cavity. With this concept we have shown lasing seen from Fig. 4(b), a significant amount of modal energy
based on the propagation of TM0 or gap-plasmon modes is contained in the dielectric layer between the semicon-
along the waveguide [12]. The width of the semiconductor ductor and the metal, reducing the proportion of the
core in these waveguides was as thin as ⬃90 nm. Electri- modal energy overlapping with the gain medium. As the
cal contacting to such thin waveguides is possible, since width of the semiconductor region is reduced and the
the small electrical contact at the top of the semiconduc- thickness of the insulator kept constant, this proportion is
tor core is connected to a much larger metallic structure. reduced. There will likely be a minimum practical thick-
This metallic structure includes the encapsulating silver ness of insulator needed to effectively isolate the metal
shell, Fig. 4(a). from the semiconductor and passivate the semiconductor
The insulator surrounding the semiconductor typically surface, typically of the order of a few nanometers. Given
has a lower refractive index n than the semiconductor. the finite maximum gain available from any optical gain
For example, InGaAs with n = 3.6 and silicon nitride with material, this reduced overlap due to the mode energy in
n = 2. The discontinuity in the refractive index of the MIM the insulator and the metal will place a limit on the
waveguide core causes a significant deviation from the smallest size of a device. The optical energy peak in the
mode shape of a TM0 mode [21] (Fig. 4(b)). However, as insulator does lead to a small calculated optical mode vol-
with MIM waveguides containing a homogeneous core the ume [15]. However it is not desirable when the goal is to
insulator part of the waveguide can be made arbitrarily reduce the physical laser size and in particular the gain
thin while still propagating a mode. region size to the smallest possible.
In the vertical direction the mode can also be strongly Calculations indicate that with widths of semiconduc-
confined. In the literature this has been shown to be pos- tor around 25 nm, an insulator thickness of 5 nm, silver
sible via simulations using either the change in refractive as the metal, and wavelengths around 1.5 microns, then
index or an indentation of the waveguide sidewalls an optical gain of ⬃2500 cm−1 is required [35]. This may
[32,33]. Effective-index arguments have been used to ex- be increased depending on how the mode overlaps the
plain such confinement. With the presence of the insulat- gain material in the vertical direction. In any case such
ing layer the effective index of the mode does not change gains are in theory possible from bulk semiconductor ma-
strongly in the pillar. Furthermore, with the indented terials, even at room temperature [9]. Higher optical
structures there are coupled surface plasmon modes on gains in semiconductors have been inferred in recent
the tops and bottoms of the metal indentations. Neverthe- nano-laser experiments [18,36] involving semiconductors.
less, numerical simulations of such structures show that If the miniaturization of the gain medium can be made to
the mode can still be strongly confined in the vertical di- the point where quantum confinement effects (typically if
rection, even to levels well below the diffraction limit. the dimensions of the gain medium are less than a few
A particularly strong vertical confinement is achieved if tens of nanometers) then potentially higher optical gains
a combination of index loading and indentation of the could be achieved of the order of 104 to 105 cm−1 [37]. Fig-
waveguide is employed [34], as shown in the schematic of ure 6(b) shows a plot of gain required for lasing versus In-
Fig. 5(a). Figure 5(b) shows a slice through a simulated GaAs core width for the structures relating to Figs. 5(b)
3D structure of such a waveguide. Here the InGaAs re- and 6(a). The crosses show the gain required if in the ver-
gion in the center is approximately 25 nm thick and the tical direction all the modal energy were contained in the
InP cladding is indented approximately 60 nm on each InGaAs layer. For a relatively thick InGaAs layer such a
side. The SiN layer is 5 nm thick all over the sidewalls.
The grid size used in the simulation is 5 nm, and this grid
discretization can be seen in Fig. 5(b). Figure 5(c) shows
the mode profile of such a waveguide calculated with 3D
FDTD for wavelengths of ⬃1500 nm with the back-to-
back-C-shaped areas in the center (red online) indicating
the highest intensity regions. It can be seen that in the
vertical direction the mode is tightly contained in the cen-
ter region. Figure 6(a) gives a plot of the proportion of
mode energy contained in the 45 nm thick InGaAs layer Fig. 5. (Color online) a. Structure of an active surface plasmon
as a function of the width at the waist of the waveguide. polariton gap waveguide. Lithography, dry etching, and selective
wet etching can be used to form the three-dimensional nano-
The InP cladding is indented by 60 nm on each side for all structure. Metal can be deposited by evaporation around the
the width points. It can be seen that for small waveguide form to complete the waveguide and provide a top electrical con-
waists the mode is strongly confined in the vertical direc- tact. b. Slice through simulated 3D structure with deviations
tion. Even for larger widths, the mode is still several from the idealized shape to take into account what can be
times more strongly confined than would be the case for a achieved with typical fabrication techniques. The InGaAs is
45 nm high and 25 nm wide. The InP cladding has an indenta-
TM0 mode in a wide slab waveguide without the indenta- tion of approximately 60 nm on each side. c. Plot of 兩E兩2 from
tion and metal sidewalls. FDTD simulation of such a structure showing tightly confined
To form a laser requires not only a cavity but gain me- mode.
B42 J. Opt. Soc. Am. B / Vol. 27, No. 11 / November 2010 Martin T. Hill

Fig. 6. (Color online) a. Simulated (3D FDTD) vertical confinement of the modal energy in the 45 nm thick InGaAs layer found by
looking at 兩E兩2 through the center of the waveguide pillar, versus InGaAs width. The straight line at the bottom (red online) is the con-
finement for a TM0 mode in such a InP / InGaAs heterostructure with a width of 2 ␮m, but without metal or sidewall indentations. b. The
crosses show the gain required if in the vertical direction all the modal energy were contained in the InGaAs layer. This is found by using
a mode solver and looking at the energy distribution across the waveguide. For a relatively thick InGaAs layer such a threshold gain
would be approached. The circles show the gain required if the vertical energy confinement shown in Fig. 6(a) is taken into account by
dividing the gain of the curve of crosses with the vertical energy confinement.

threshold gain would be approached. The circles show the gain region is approximately 18 nm thick, the waveguide
gain required if the vertical energy confinement shown in waist is approximately 25 nm, and the waveguide is
Fig. 6(a) were taken into account. It can be seen that for 20 ␮m long. The device was measured at 10 K. Increased
an InGaAs width of 25 nm and a thickness of 45 nm a light output and band filling can be observed with in-
gain of ⬃4000 cm−1 is required. creased electrical pumping, which indicates that the
In any case, there is in theory potential to reduce the semiconductor gain medium is functioning. Furthermore
size of such semiconductor laser devices down to a few the stronger light emission at the waveguide ends indi-
tens of nanometers in at least two dimensions. We have cates that a significant amount of the light emitted is go-
been developing technology to realize these very small de- ing into modes propagating along the waveguide [12].
vice sizes. In particular a number of issues need to be ad- However, we have yet to achieve lasing in these devices.
dressed. Roughness of the pillar sidewall needs to be If these MIM devices are to be of use in actual systems
small. We have managed to reduce the sidewall rough- then they will likely have to be coupled to either conven-
ness to the level of a few nanometers or less. The shape of tional dielectric or plasmonic passive waveguides. Fortu-
the sidewall needs to be indented consistently along the nately it has been shown in theory that indeed it is pos-
length of the pillar and the shape of the indentation con- sible to couple plasmons in MIM waveguides with high
trolled to allow consistent deposition of metals and dielec- efficiency to photons propagating in conventional dielec-
trics in the grooves formed. We have employed wet chemi- tric waveguides [38]. Furthermore the conversion can be
cal and plasma-assisted etching processes to achieve performed in a compact device. Such conversion is an im-
sufficient control of the sidewall shape. The conformal portant point for the deployment of plasmonic nano-
coating of the pillar with a thin dielectric layer is a par- lasers, as signals from the lasers will eventually need to
ticularly critical step. As mentioned before it is desirable be transported over distances of at least several millime-
to have the dielectric layer as thin as possible and also ters, and the best way to do this is with dielectric
with a high refractive index to minimize the energy con- waveguides or possibly waveguides supporting long-range
tained in it. However, the dielectric layer also must be surface plasmons [4].
sufficiently thick not to break down when a voltage is ap- Another aspect is precise frequency control in such
plied across the laser diode. It must not cause damage to waveguide lasers. By making the cavity small, as in the
the semiconductor surfaces when it is deposited, and fi- VCSEL-like lasers mentioned before, there is only one
nally needs to afford protection to the semiconductor sur-
face during subsequent cleaning and deposition steps in
the device manufacture. The deposition of high-quality
conformal silver layers on the complex sidewall shape is
probably the last critical process that needs to be ad-
dressed.
Figure 7 shows our progress in addressing these issues.
The cross section of a waveguide with deposited silver
given in Fig. 7(b) shows that it is possible to physically Fig. 7. (Color online) a. Actual fabricated semiconductor
InP / InGaAS semiconductor core. b. Cross section of a completed
construct such structures. We have also completed pro- waveguide that has been encapsulated in silver. The scale bars in
cessing of a number of devices with electrical contacts. both a. and b.) are 100 nm. c. Emission through substrate of a
Figure 7(c) shows a photo of the electro-luminescence waveguide similar to b. with active region width of ⬃25 nm and
seen through the substrate of a device where the InGaAs the waveguide 20 ␮m long.
Martin T. Hill Vol. 27, No. 11 / November 2010 / J. Opt. Soc. Am. B B43

A large number of groups have also looked at using


semiconductor gain media. Semiconductors bring the im-
portant technological advantages of electrical pumping
and the possibility of long device lifetimes. In this article
we have looked broadly at two possible tracks for
semiconductor-based metallic/plasmonic nano-lasers.
First we have considered devices where the optical mode
is confined to regions that are typically about the size of
the diffraction limit in the material, or slightly larger.
These devices are already showing operation at high tem-
peratures and also with electrical pumping. In theory
there is a significant excess of optical gain possible from
Fig. 8. (Color online) Semiconductor core of MIM waveguide the semiconductor, so that with correct design highly effi-
with modulated width used to form plasmonic DFB lasers as re-
cient laser sources could be made. Furthermore, the size
ported in [41]. Scale bar is 100 nm.
is still small enough that many of the benefits of metallic
devices are available, such as high Purcell effect, good
lasing mode. With sufficiently accurate control on the cav- heat sinking, and sub wavelength dimensions. It is pos-
ity dimensions, the wavelength can be precisely deter- sible that in a short time such devices will be used in
mined. However, for longer waveguide cavities there may many current and new applications for lasers.
be many lasing modes. One approach used successfully to The use of metals allows the unique possibility of re-
precisely control the lasing wavelength has been the use ducing the laser and optical mode size far below the dif-
of Bragg gratings to form distributed feedback (DFB) or fraction limit. In theory, given current semiconductor gain
distributed Bragg reflector (DBR) lasers [39]. It has also materials, it should be possible to overcome metal losses
been reported that the Bragg grating concept is possible in these extremely small devices, even at room tempera-
to implement in MIM waveguide structures [40] by sim- ture. At cryogenic temperatures the increased gain avail-
ply modulating the waveguide width. Such a concept can able from the semiconductor and the reduced metal losses
be implemented easily in our realization of the MIM will allow many interesting first demonstrations. We have
waveguide concept by simply modulating the waveguide looked at one particular example of these very small
width, which is defined in the lithography step. Indeed we structures, MIM waveguide devices. Having lasers that
have successfully created DFB lasers that employ a gap- operate within a wave guiding system is important as it
plasmon mode in our technology; the results are reported allows complex designs for the laser cavity, and a method
in [41]. In particular we can show single-mode lasing with to employ and direct the light from the laser.
controlled wavelength in a 100␮m long MIM waveguide. In summary if the challenges of electrical pumping,
Figure 8 shows a fabricated semiconductor core for a MIM high-temperature operation, and good device lifetime can
waveguide designed for a DFB laser that has a 1 / 4 wave- be met, then spasers and metallic or plasmonic nano la-
length phase shift in the center of the grating. sers will be widely employed in diverse areas including in-
If such nano-lasers as we describe above can be made to formation processing, communications, sensing, and pos-
work at room or higher temperatures then they could be sibly many other new applications.
applied as light sources for communications systems
based on light and plasmonics, which operate inside or be-
tween silicon chips [42]. Such devices will have significant ACKNOWLEDGMENTS
advantages such as low power and potentially high speed, This work was supported by the Netherlands Organiza-
in particular, for all-optical modulation. For electrical tion for Scientific Research (NWO)through the “NRC pho-
modulation there may be issues with the maximum speed tonics” grant, and by the Dutch governmentthrough the
possible [26]. Such devices could serve as a basis for high- MEMPHIS grant.
speed, low-power, high-complexity, all-optical signal pro-
cessing systems that could outperform electronics in high-
speed applications. REFERENCES AND NOTES
1. K. Iga, “Surface-emitting laser—Its birth and generation of
new optoelectronics field,” IEEE J. Sel. Top. Quantum Elec-
tron. 6, 1201–1215 (2000).
2. O. Painter, R. K. Lee, A. Scherer, A. Yariv, J. D. O’Brien, P.
5. CONCLUSION D. Dapkusand, and I. Kim, “Two-dimensional photonic
The last couple of years have seen a dramatic reduction in band-gap defect mode laser,” Science 284, 1819–1821
the size of the laser through the use of metallic resona- (1999).
3. H. A. Atwater, “The promise of plasmonics,” Sci. Am. 296,
tors. Probably the most exciting result has been the use of
38–45 (April, 2007).
dye gain media to demonstrate a laser that is an order of 4. W. L. Barnes, A. Dereux, and T. W. Ebbesen, “Surface Plas-
magnitude smaller than the free space wavelength of the mon subwavelength optics,” Nature (London) 424, 824–830
light it emits in all three dimensions [16]. However, the (2003).
use of a dye gain medium has significant drawbacks with 5. M. P. Nezhad, K. Tetz, and Y. Fainman, “Gain assisted
propagation of surface plasmon polaritons on planar metal-
regard to electrical pumping and device lifetimes. Such lic waveguides,” Opt. Express 12, 4072–4079 (2004).
dye-based devices may have some applications in sensing 6. S. A. Maier, “Gain-assisted propagation of electromagnetic
or biology. energy in subwavelength surface plasmon polariton gap
B44 J. Opt. Soc. Am. B / Vol. 27, No. 11 / November 2010 Martin T. Hill

waveguides,” Opt. Commun. 258, 295–299 (2006). 822 (2002).


7. A. Maslov and C. Z. Ning, “Size reduction of a semiconduc- 25. S.-H. Kim, Y.-H. Lee, J. Huang, and A. Scherer, “Unidirec-
tor nanowire laser using metal coating,” Proc. SPIE 6468, tional vertical emission from photonic crystal nanolaser,”
646801–646807 (2007). 11th International Conference on Transparent Optical Net-
8. M. T. Hill, Y.-S. Oei, B. Smalbrugge, Y. Zhu, T. de Vries, P. J. works ( ICTON ’09) (IEEE, 2009), paper Tu.C4.5.
van Veldhoven, F. W. M. van Otten, T. J. Eijkemans, J. P. 26. E. K. Lau, A. Lakhani, R. S. Tucker, and M. C. Wu, “En-
Turkiewicz, H. de Waardt, E. J. Geluk, S.-H. Kwon, Y.-H. hanced modulation bandwidth of nanocavity light emitting
Lee, R. Nötzel, and M. K. Smit, “Lasing in Metallic-Coated devices,” Opt. Express 17, 7790–7799 (2009).
Nanocavities,” Nat. Photonics 1, 589–594 (2007). 27. M. Miller, M. Grabherr, R. Jager, and K. J. Ebeling, “High-
9. G. P. Agrawal and N. K. Dutta, Semiconductor Lasers, 2nd power VCSEL arrays for emission in the watt regime at
ed. (Van Nostrand Reinhold, 1993). room temperature,” IEEE Photon. Technol. Lett. 13, 173–
10. K. Yu, A. Lakhani, and M. C. Wu, “Subwavelength metal- 175 (2001).
optic semiconductor nanopatch lasers,” Opt. Express 18, 28. R. Zia, M. D. Selker, P. B. Catrysse, and M. L. Brongersma,
8790 (2010). “Geometries and materials for subwavelength surface plas-
11. B. Prade and J. Y. Vinet, “Guided optical waves in fibers mon modes,” J. Opt. Soc. Am. A 21, 2442–2446 (2004).
with negative dielectric constant,” J. Lightwave Technol. 29. D. M. Koller, A. Hohenau, H. Ditlbacher, N. Galler, F. Reil,
12, 6–18 (1994). F. R. Aussenegg, A. Leitner, E. J. W. List, and J. R. Krenn,
12. M. T. Hill, M. Marell, E. S. P. Leong, B. Smalbrugge, Y. Zhu, “Organic plasmon-emitting diode,” Nat. Photonics 2, 684–
M. Sun, P. J. van Veldhoven, E. J. Geluk, F. Karouta, Y.-S. 687 (2008).
Oei, R. Nötzel, C.-Z. Ning, and M. K. Smit, “Lasing in 30. R. J. Walters, R. V. A. Van Loon, I. Brunets, J. Schmitz, and
metal-insulator-metal sub-wavelength plasmonic A. Polman, “A silicon-based electrical source of surface plas-
waveguides,” Opt. Express 17, 11107–11112 (2009). mon polaritons,” Nature Mater. 9, 21–25 (2009).
13. M. P. Nezhad, A. Simic, O. Bondaenko, B. Slutsky, A. 31. C. Walther, G. Scalari, M. I. Amanti, M. Beck, and J. Faist,
Mizrahi, L. Feng, V. Lomakin, and Y. Fainman, “Room- “Microcavity laser oscillating in a circuit based resonantor,”
temperature subwavelength metallo-dielectric lasers,” Na- Science 327, 1495–1497 (2010).
ture Photon. 4, 395–399 (2010). 32. F. Kusunoki, T. Yotsuya, J. Takahara, and T. Kobayashi,
14. R. Perahia, T. P. Mayer Alegre, A. H. Safavi-Naeini, and O. “Propagation properties of guided waves in index-guided
Painter, “Surface-plasmon mode hybridization in subwave- two-dimensional optical waveguide,” Appl. Phys. Lett. 86,
length microdisk lasers,” Appl. Phys. Lett. 95, 201114 211101 (2005).
(2009). 33. K. Tanaka and M. Tanaka, “Simulations of nanometric op-
15. R. F. Oulton, V. J. Sorger, T. Zentgraf, R.-M. Ma, C. Glad-
tical circuits based on surface plasmon polariton gap wave-
den, L. Dai, G. Bartal, and X. Zhang, “Plasmon lasers at
guide,” Appl. Phys. Lett. 82, 1158–1160 (2003).
deep subwavelength scale,” Nature (London) 461, 629–632
34. M. T. Hill, “Micro and nanolasers for digital photonics,” Pro-
(2009).
ceedings of the European Conference on Integrated Optics
16. M. A. Noginov, G. Zhu, A. M. Belgrave, R. Bakker, V. M.
(ECIO), Copenhagen (Technical University of Denmark,
Shalaev, E. E. Narimanov, S. Stout, E. Herz, T. Suteewong,
2007) (pp. WC0-64/67).
and U. Wiesner, “Demonstration of a spaser-based nanola-
35. M. T. Hill, “Metallic nano-cavity lasers at near infrared
ser,” Nature (London) 460, 1110–1112 (2009).
17. D. J. Bergman and M. I. Stockman, “Surface plasmon am- wavelengths,” Proc. SPIE 7394, 739409 (2009).
plification by stimulated emission of radiation: Quantum 36. D. J. Gargas, M. C. Moore, A. Ni, S.-W. Chang, Z. Zhang,
generation of coherent surface plasmons in nanosystems,” S.-L. Chuang, and P. Yang, “Whispering gallery mode lasing
Phys. Rev. Lett. 90, 027402 (2003). from zinc oxide hexagonal nanodisks,” ACS Nano 4, 3270–
18. C.-Y. Lu, S.-W. Chang, S. L. Chuang, T. D. Germann, and D. 3276 (2010).
Bimberg, “Metal-cavity surface-emitting microlaser at room 37. M. Asada, Y. Miyamoto, and Y. Suematsu, “Gain and the
temperature,” Appl. Phys. Lett. 96, 251101 (2010). threshold of three-dimensional quantum-box lasers,” IEEE
19. H. G. Park, S.-H. Kim, S.-H. Kwon, Y.-G. Ju, J.-K. Yang, J. Quantum Electron. QE-22, 1915–1921 (1986).
J.-H. Baek, S.-B. Kim, and Y.-H. Lee, “Electrically driven 38. P. Ginzburg, D. Arbel, and M. Orenstein, “Gap Plasmon po-
single-cell photonic crystal laser,” Science 305, 1444–1447 lariton structure for very efficient microscale-to-nanoscale
(2004). interfacing,” Opt. Lett. 31, 3288–3290 (2006).
20. H. Altug, D. Englund, and J. Vučković, “Ultrafast photonic 39. J. Buus, M. C. Amann, and D. J. Blumenthal, “Distributed
crystal nanocavity laser,” Nat. Phys. 2, 485–488 (2006). Feedback Lasers,” in Tunable Laser Diodes and Related Op-
21. B. Prade, J. Y. Vinet, and A. Mysyrowicz, “Guided optical tical Sources, 2nd ed., (Wiley, 2005), pp. 59–68 .
waves in planar heterostructures with negative dielectric 40. Z. Han, E. Forsberg, and S. He, “Surface plasmon Bragg
constant,” Phys. Rev. B 44, 13556–13572 (1991). gratings formed in metal-insulator-metal waveguides,”
22. J. A. Dionne, L. A. Sweatlock, H. A. Atwater, and A. Pol- Bull. Pol. Acad. Sci.: Biol. Sci. 19, 91–93 (2007).
man, “Plasmon slot waveguides: Towards chip-scale propa- 41. M. J. H. Marell, Dept. of Electrical Engineering, Technische
gation with subwavelength-scale localization,” Phys. Rev. B Universiteit Eindhoven, Postbus, 513, 5600 MB Eindhoven,
73, 035407 (2006). The Netherlands, is preparing a manuscript on gap-
23. S.-W. Chang, T.-R. Lin, and S. L. Chuang, “Theory of plas- plasmon mode DFB lasers in the near infrared wave-
monic Fabry-Perot nanolasers,” Opt. Express 18, 15039– lengths.
15053 (2010). 42. J. A. Dionne, L. A. Sweatlock, M. T. Sheldon, A. P. Alivisa-
24. H. J. Lezec, A. Degiron, E. Devaux, R. A. Linke, L. Martin- tos, and H. A. Atwater, “Silicon-based plasmonics for on-
Moreno, F. J. Garcia-Vidal, and T. W. Ebbesen, “Beaming chip photonics,” IEEE J. Sel. Top. Quantum Electron. 16,
light from a subwavelength aperture,” Science 297, 820– 295–305 (2010).

You might also like