1
Ultrafast ZnO nanowire lasers: nanoplasmonic acceleration of gain dynamics
at the surface plasmon polariton frequency
Themistoklis P. H. Sidiropoulos
1*
, Robert Röder², Sebastian Geburt²,
Ortwin Hess
1
, Stefan A. Maier
1
, Carsten Ronning², and Rupert F. Oulton
1*
1. Imperial College London, Prince Consort Road, UK-SW7 2BZ London, United Kingdom
2. University of Jena, Max-Wien-Platz 1, D-07743 Jena, Germany
*Contact: r.oulton@imperial.ac.uk
and t.sidiropoulos10@imperial.ac.uk
Light-matter interactions are inherently slow as the wavelengths of optical and electronic states
differ greatly. Surface plasmon polaritons, electromagnetic excitations at metal-dielectric
interfaces, have generated significant interest because their spatial scale is decoupled from the
vacuum wavelength, promising accelerated light-matter interactions. Meanwhile, the possibility of
accelerated dynamics in recently demonstrated surface plasmon lasers remains to be verified. In
this letter, we report the observation of <800 fs pulses from hybrid plasmonic zinc oxide (ZnO)
nanowire lasers. Operating at room temperature, ZnO excitons lie near the SPP frequency in such
silver-based plasmonic lasers, leading to accelerated spontaneous recombination, gain switching,
and gain recovery compared to conventional ZnO nanowire lasers. Surprisingly the laser dyanmics
can be as fast as gain thermalization in ZnO, which precludes lasing in the thinnest nanowires
(diameter < 120 nm). The capability to combine surface plasmon localization with ultrafast
amplification provides the means for generating extremely intense optical fields with applications
in sensing, non-linear optical switching, as well as in the physics of strong field phenomena.
Lasers that use metallic cavities have emerged recently as a new class of light source
1–3
. Plasmonic
lasers achieve optical confinement and feedback using surface plasmon polaritons (SPPs), quasi-
particles of photons and electrons at metal-dielectric interfaces, which can be amplified by suitable
optical gain media
4
. The high gain of inorganic crystalline semiconductors is typically necessary to
overcome fast electron scattering in metals (~10 fs), which leaves plasmonic lasers with high
parasitic cavity loss. Nevertheless, SPPs offer the capability to reduce optical mode sizes far below
the scale of the vacuum wavelength
3,5–8
leading to compact lasers that can generate extremely
focussed optical excitations on potentially ultrafast time scales
1,9
with applications in Raman
sensing
10–12
, non-linear frequency generation
13–15
, and non-linear optical switching
16
. Despite the
draw-back of loss, numerous plasmonic lasers have been reported recently with progress being
made towards reducing the laser threshold to a point where practical applications are viable. In
particular, several devices now operate at room temperature
7,17
and even under electrical
injection
18
. While the practical issues of these devices have seen progress, few experimental works
have studied their underlying limitations and capabilities. In terms of limitations, it is currently
unclear how much confinement is realistically sustainable. In plasmonics confinement is associated
2
with loss and accelerated recombination, which both affect laser threshold
19–21
. In particular, only a
few works have reported plasmonic laser action near the SP frequency where the electromagnetic
field equally shares energy with electron polarization maximizing both confinement and loss
6
. As for
capabilities, the expectation that plasmonic lasers are ultrafast amplifiers
9,19,22
due to optical
confinement and the corresponding Purcell effect
23–25
, has yet to be experimentally proven.
Certainly, accelerated gain dynamics are achievable
26
since the Purcell effect accelerates both
spontaneous and stimulated recombination rates due to the intrinsic relationship between Einstein’s
A and B coefficients. However, it is less clear what degree of accelerated recombination is
sustainable due the effect on threshold and the finite time scale of carrier thermalization in the
amplifying medium. In this article we address both of these questions. Firstly, we report the
demonstration of hybrid plasmonic
27
lasers at room temperature operating near the SPP frequency
by exploiting the ultraviolet gain spectrum of ZnO nanowires. Secondly, we report accelerated laser
dynamics of plasmonic lasers compared to conventional photonic ZnO nanowire lasers. Remarkably,
we have measured sub-picosecond plasmonic lasers pulses, within the ultrafast regime, where
temporal dynamics are generally only accessible by non-linear all-optical techniques
28
. Here we have
used a novel double-pump approach that exploits the non-linearity of the laser process to expose its
own dynamics.
The plasmonic lasers under investigation consist of individual ZnO nanowires placed on a 10 nm
thick lithium fluoride (LiF) spacer layer over a silver (Ag) substrate, as shown in Fig. 1a. The insulating
spacer layer affords optical confinement control and isolates ZnO excitations from quenching at the
metal surface
5
. The photonic lasers consist of ZnO nanowires from the same batch of nanowires but
placed on a Si/SiO
2
substrate
29
. The finite length of each nanowire (520 µm) defines a cavity;
laterally confined modes propagate backwards and forwards along the nanowire with feedback
arising from modal reflection at the end-facets. Under optical pumping, ZnO is capable of the
extremely strong optical gain necessary to achieve plasmonic lasing near its bandedge at 3.37 eV.
This originates mainly from the two lowest energy of three excitons (
3.309 eV,
3.315
eV and
3.355 eV) that have a sufficiently large binding energy (60 meV) to be stable at room
temperature
30
(Fig. 1b). Under weak optical pumping, spontaneous photoluminescence and gain
occur below the band-edge near 3.24 eV through either exciton-exciton scattering or via optical
phonon scattering (
72 meV)
31,32
. Meanwhile, under strong excitation above the Mott
density, carrier screening causes exciton dissociation into an electron hole plasma (EHP), which
together with band-gap renormalization, provides gain as far below the band-edge as ~3.19 eV
33
.
Even for photonic nanowire lasers, the EHP mechanism is required for sufficient gain to overcome
the high cavity loss
34,35
. Thus, we also expect the EHP mechanism to occur in plasmonic lasers.
3
The dispersion relations for the transverse modes of the plasmonic and photonic devices are both
influenced by the excitonic dispersion of ZnO. Figure 1b shows the dispersion relation of the
fundamental plasmonic transverse mode for a nanowire diameter of 130 nm. Since the SPP
frequencies of Silver-Air (
3.65 eV) and Silver-ZnO (
2.90 eV) encompass the
exciton energies of ZnO, excitons should couple strongly to SPPs (Fig. 1b). Indeed, calculations show
a surface plasmon frequency within the absorption spectrum of ZnO, where we have used a Brendel-
Bormann model for the permittivity of silver, which was fit to the data of Palik
36,37
. The large surface
plasmon wavenumber within ZnO’s gain spectrum suggests that strong optical mode confinement
occurs (see supplementary information). In previous plasmonic laser work, a hybrid plasmonic
mode
27
, shown in Fig. 1c, was identified through the absence of a mode cut-off and the
enhancement of spontaneous recombination
5
. In contrast, a photonic nanowire of 150 nm is
very close to cut-off within ZnO’s gain spectrum. Indeed, previous works have not observed photonic
ZnO nanowire lasers for 150 nm
29
. Both plasmonic and photonic laser can also support other
transverse modes for thicker nanowires, but these generally exhibit weaker confinement and
feedback (supplementary information)
29
. To our knowledge, all reported photonic ZnO lasers have
operated via the EHP mechanism with emission energies near 3.23 eV where polariton dispersion is
relatively weak
35
. In contrast, the plasmonic lasers in this work operate in the vicinity of the exciton
energies, near 3.30 eV.
Each nanowire was optically pumped at a wavelength of 355 nm with nominally 150 fs pulses at a
repetition rate of 800 kHz, chosen to avoid heating in the devices, but also to probe their ultrafast
dynamics. The spectra shown in Fig. 2a are representative of photonic and plasmonic lasers pumped
at twice their respective threshold energy densities (inset of Fig. 2a). In general, plasmonic lasers
show a suppressed super-linear light vs pump response near the laser transition compared to
photonic devices, which is characteristic of enhanced spontaneous recombination arising from mode
localization and reduced mode competition
2
. (Other examples of plasmonic lasers are shown in the
supplementary information). The difference in thresholds of 43 µJcm
-2
and 200 µJ cm
-2
for the
photonic and plasmonic laser, respectively, are reminiscent of the differences observed in previous
work
5
. The role of SPPs in the plasmonic lasers can also be clearly confirmed in their emission
polarisation: plasmonic lasers are polarized along the nanowire, which is consistent with the
dominant field components of hybrid SPP modes
27
, as shown in Fig. 2a. Above threshold photonic
devices lase in the EHP regime between 3.19 3.25 eV while plasmonic devices only lase above
3.25 eV (Fig. 2b). The blue shift occurs as the lossy plasmonic cavity requires considerably higher
gain, which can only be achieved nearer the exciton transition energies
38
. Moreover, a reduction in
nanowire diameter causes a further blue shift, even beyond the energies of the A and B excitons,
4
primarily due to higher loss for smaller diameter nanowires, but also due to state filling
39
. The
threshold behaviour, high gain and the laser mode polarisation strongly suggest the role of SPP
modes in the lasing action. Furthermore, we have measured plasmonic lasers with diameters down
to 120 nm where photonic lasing cannot occur
29
. Remarkably, we do not observe a significant
increase in plasmonic laser threshold with decreasing diameter, despite the apparent increases in
loss (supplementary information).
The stark differences between plasmonic and photonic light vs pump curves and spectral responses
suggest a modification of the laser process that both accelerates and redistributes spontaneous
recombination
2
. In order to examine the influence of confinement on stimulated emission dynamics,
ideally we should measure the temporal shape of light pulses emanating from these lasers. This
presents a significant challenge as plasmonic lasers are anticipated to operate in the sub picosecond
regime, which is at the limit of electrical detection techniques, such as streak cameras
26
. Generally,
non-linear all-optical techniques would be required
28
to characterize such short light pulses,
however, individual nanowires produce insufficient signal for such time resolved non-linear
spectroscopies. Here, we have explored the temporal dynamics of these nanowire-based lasers by
measuring their response to optical excitation with two energetically identical pump pulses
separated by a variable time delay, . The technique uses the inherent non-linearity of the
population inversion to reveal internal temporal dynamics that establishes clear bounds on the
characteristic laser response times.
Figure 3 illustrates the expected response of a hypothetical laser cavity incorporating a 3-level
electronic gain system to the double-pump pulse strategy (Methods). Here, the intensity of one
beam is sufficient to excite the laser, whereas the second beam is weaker and cannot induce lasing
on its own. The laser’s response as a function of time delay indicates spontaneous recombination
and gain recovery dynamics of the laser. The chosen parameters reproduce the observed responses
of the measured plasmonic lasers.
For < 0 the weak pulse initially excites carriers into the upper level, which rapidly thermalize to the
excited state and subsequently recombine spontaneously (Fig. 3a). This population is then further
excited by the strong pulse causing the system to lase. Prior to the arrival of the strong pulse the
excited state population exponentially decays at the spontaneous recombination rate, thus
increasing the laser output power with decreasing delay between the pulses, as shown in Fig. 3c.
Around the zero delay point we can expect intricate interference of the two input pulses
40
, which
only becomes apparent with varying time delay, as shown in Fig. 3c. For > 0 the strong pump pulse
initially creates an inversion, generating an output laser pulse that partially depletes the excited
5
state population (Fig. 3b). The weak pump pulse may now have a significant effect on the laser’s
response as the residual excited state population can facilitate lasing.
For > 0 we can further identify three situations. For a small time delay the excited state is still near
peak population and the weak pump pulse merely amplifies the output pulse. Note that the strong
pump pulse depletes the ground state and thus reduces the nanowire’s absorption of the weaker
pump pulse. For increased time delays, absorption of the weaker pump pulse grows and this leads to
the gradual formation of a second output pulse as well as the amplification of the first output pulse,
as shown in Fig. 3b. The two output pulses emerge on distinct time scales: time delays between
pump and output pulses,
and
, occur for the strong and weak pump pulses, respectively; and
the peak to peak separation of the two output pulses is
. These time scales are related to the
pump pulse delay by
. Note that since the initial pump pulse must create the
inversion, usually
. The double pump response reaches a maximum, denoted by
, when
the absorption of the second pump pulse recovers also marking the termination of the first output
pulse (Fig. 3c). Finally, for large time delays the residual excited population gradually depletes with
the spontaneous recombination time until a point where the second pump pulse no longer induces
lasing and merely creates incoherent emission.
This simple theoretical model predicts a number of characteristics that agree well with the measured
double-pump responses of the plasmonic and photonics lasers, shown in Fig. 4. To make a fair
comparison of the two devices, we chose similar normalized pump energy densities ranging between
once and twice the respective threshold values (Fig. 4c,d). Furthermore, since the responses are
non-linear, we also fixed the power ratio of the two pump pulses at 1/5
th
. The plasmonic laser
clearly shows accelerated spontaneous recombination, evident from the steeper exponential decay
for 0 compared with the photonic device. A quantitative assessment of the Purcell factor is
difficult since the photonic laser has an almost flat response for 0 and the plasmonic laser
exhibits a shallower non-linear light versus pump response than the photonic device, as shown in
Figs. 4c and 4d, respectively. The magnitude of the enhanced spontaneous recombination in the
plasmonic device is clearer for 0, where after-pulsing from the weaker pump pulse persists for
less than a 10
th
of the time scale with respect to the photonic device. Interestingly, the double pump
responses in Figs. 4a and 4b become faster with decreasing pump energy density, which is
counterintuitive. The reader should note that both pump pulses change intensity here: since the
ratio of pump intensities is fixed at 1/5
th
, a second output pulse becomes less sustainable for larger
time delays, leading to an apparently faster double-pump response. A more reliable indication of the
accelerated plasmonic laser dynamics is the peak response time,
. While the observed decay