diskk lasers review 2008n
TRANSCRIPT
-
8/19/2019 Diskk Lasers Review 2008n
1/11
Early View publication on www.interscience.wiley.com(issue and page numbers not yet assigned;citable using Digital Object Identifier – DOI)
Laser & Photon. Rev., 1–11 (2008) / DOI 10.1002/lpor.200710032 1
Abstract We describe recent progress in photonic crystal
nanocavity lasers with an emphasis on our recent results on
ultrafast pulse generation. These lasers produce pulses on the
picosecond scale, corresponding to only hundreds of optical
cycles. We describe laser dynamics in optically pumped single
cavities and in coupled cavity arrays, at low and room temper-
ature. Such ultrafast, efficient, and compact lasers show great
promise for applications in high-speed communications, infor-
mation processing, and on-chip optical interconnects.
© 2008 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
Ultrafast photonic crystal lasers Dirk Englund 1 ,* , Hatice Altug 2 , Bryan Ellis 1 , and Jelena Vuˇ ckovi´ c 1
1 Ginzton Laboratory, Stanford University, Stanford CA 943052 Electrical and Computer Engineering Department, Boston University, Boston MA 02215
Received: 13 November 2007, Revised: 17 March 2008, Accepted: 13 May 2008
Published online: 2 July 2008
Key words: photonic crystal; laser; Purcell effect; ultrafast
PACS: 42.55.Sa,42.55.Tv,42.50.Ct, 42.70.Qs
1. Introduction
The field of photonics is transitioning towards highly in-tegrated nanoscale devices. For the first time, researchersare able to integrate fast low-power optical components onsemiconductor chips, causing some to draw analogies to thesemiconductor electronics revolution. The most commer-
cially significant application of such integrated photonicsis optical communications. Because of low output pow-ers, devices would first find applications in short-distancecommunication, including high-speed local networks, andboard-to-board and chip-to-chip interconnects. Additionalapplications lie in biochemical sensing and data storage.
One of the most promising architectures for integratednanoscale devices is the planar photonic crystal (PC). In-plane confinement is achieved by distributed Bragg reflec-tion using periodic arrangements of holes, while out-of-plane confinement results from total internal reflection. Cav-
ities, defined by defects in the PC, can confine light nearthe ultimate volume limit λ/2n in all dimensions. Withoptimized local geometry, extremely high confinement ispossible, with quality factors on the order of 104 in ac-
tive [1–3] and 106 in passive structures [4]. Through thecombination of a high quality factor (Q) and small modevolume V m, such cavities can dramatically increase thevacuum Rabi energy, enabling cavity quantum electrody-namic effects such as enhanced spontaneous emission (SE)rate of embedded emitters [5]. This cavity Purcell effectlowers the lasing threshold through higher SE coupling
efficiency β , far in excess of 50% for even modest meanPurcell factor [3]. By contrast, β in Vertical Cavity Sur-face Emitting Lasers (VCSELs) is typically less than 0.1%.The Purcell effect can also increase the direct modulationspeed [6, 7]. The photonic on-chip design is ideally suitedfor the integration of different optical components; e.g.,light sources/detectors and multiplexers/demultiplexers foroptical communications.
In this paper, we describe recent progress on photoniccrystal nanocavity lasers. We will focus mainly on ultrafast
laser dynamics reported in recent work from our group,but will attempt to mention the major works in the fieldwhere appropriate. Previous studies investigated the las-ing dynamics indirectly in the frequency domain [7]; here,we instead focus on direct time-domain measurements. In
Corresponding author: e-mail: [email protected]
© 2008 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
-
8/19/2019 Diskk Lasers Review 2008n
2/11
2 D. Englund, H. Altug, et al.: Ultrafast photonic crystal lasers
Sect. 2.1, we begin with the designs for single-cavity andcoupled-cavity devices, then in Sect. 2.2 introduce a modelthat describes lasing action in our devices. In Sect. 3, we
discuss our recent work on optically pumped PC lasers,driven by multiple quantum wells (QWs). These structuresshow extremely fast lasing action due to fast relaxationdynamics. In the near-IR range, we demonstrate room tem-perature lasing with large-signal modulation response onpicosecond time-scales, i.e., with only several hundreds of electric-field oscillations per pulse. A surface passivationtechnique greatly improves the practicality of the PC laserby limiting nonradiative (NR) losses. Using a similar de-sign in the InP/InGaAsP material system, we demonstrate
large-signal modulation with pulse widths near 10ps in thetelecom band. In Sect. 4, we shift our attention from theQW to quantum dot (QD) gain medium. In typical cavities
with Q ≥ 1000 and QD density ≥ 100/µ m2, where thresh-old is determined by material properties, the QD activematerial reduces lasing threshold because of lower gainarea and surface recombination losses. We conclude with abrief discussion of electrical pumping.
2. Small-volume PC lasers
The small-volume, high-Q cavities enabled by photoniccrystals can decrease turn-on time and lasing threshold [6].This improvement results when the gain spectrum overlapswith the cavity resonance so that spontaneous emission
into the cavity mode exhibits higher spontaneous emissionrate and spontaneous emission coupling efficiency β . Theseeffects can decrease turn-on time and lasing threshold [6].In addition, microcavity lasers can be designed with verybroad modulation bandwidth because the relaxation oscilla-
tion can be shifted beyond the cavity cutoff frequency [8].
Above threshold, higher pump powers lead to fasterdecay due to increased stimulated emission rates. Smallmode volume PC cavities can be used to achieve large pho-ton densities and speed up this process. Compared to othertypes of lasers such as VCSELs, PC lasers offer lower driv-
ing power (Sec. 3), higher relaxation oscillation frequency(see Sec. 2.4), and potentially faster electrical modulation
speed because of the potential for lower device capacitanceand resistance.
2.1. PC nanocavity laser design
In designing the PC structure for fast lasing action, twoconsiderations are weighed: the Q value must be relativelylarge to achieve SE Purcell enhancement and hence highSE coupling efficiency; at the same time, the mode energyring-down time τ c = Q/ω should be small as it limitsthe laser’s response time. We choose Q ≈ 2 · 103, corre-sponding to τ = 1 ps at laser wavelength λ ≈ 1 µ m. This
value of Q is easily achieved with the the single-defectcavity shown in Fig. 1c, defined in a square-lattice photonic
(a) B
(c) (d)
z
x
y
x
y
(b) Bz
|E|2
Figure 1 (online color at: www.lpr-journal.org) Square-lattice
photonic crystal laser structures. (a) x-dipole-mode field pattern(out-of-plane magnetic field Bz). y−dipole is rotated by 90
◦.
(b) Quadrupole mode. (c) Single-defect cavity with electric field
intensity (inset). (d) Coupled cavity array structure in GaAs.
crystal. The quadrupole mode, shown in the inset, has aQ ∼ 2000 as predicted by Finite Difference Time Domain(FDTD) simulations [9].
Because of its small size, the single-defect PC laser has
the disadvantage that output power is low – on the order of a few µ W – and much of this power is lost due to a wide-angle emission profile. On the other hand, band edge PClasers, which operate in slow-group velocity regions of thePC dispersion, comprise a greater gain area, as was shownin the first PC laser demonstrations [10, 11]. They also of-fer greatly improved emission directionality [12]. However,
they entail other drawbacks, such as reduced lateral confine-ment [13]. A good compromise appears to be combiningthe strengths of the nanocavity and band edge lasers by ar-ranging single-cavity lasers into an array [9]. If the cavities
are sufficiently close, lasing can be achieved in a commonmode. This PC nanocavity array laser has far more direc-tional emission and larger active material than the singlelaser, while providing better lateral confinement than theband edge laser. The nanocavity array achieves very lowgroup velocity in any photonic crystal direction and a veryhigh density of electromagnetic states; in effect, the struc-ture is the two-dimensional analog of coupled resonatoroptical waveguides (CROWs) in photonic crystals [14, 15].Though coupled arrays of small numbers of VCSELs werepreviously investigated [16], coupling between individuallasers is difficult and requires a rather complicated fabri-
cation procedure. Photonic crystal nanocavity arrays allowprecise control of both the uniformity and the coupling.
© 2008 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.lpr-journal.org
-
8/19/2019 Diskk Lasers Review 2008n
3/11
Laser & Photon. Rev. (2008) 3
!"# !$#
Figure 2 (online color at: www.lpr-journal.org) Three-hole de-
fect cavity. (a) FDTD design with electric field intensity. (b) Fab-
ricated structure in GaAs with central InAs QD layer.
We investigated two-dimensional cavity arrays withmodes that couple equally in the different crystal directions
(monopole and quadrupole modes in the square lattice ormonopole and hexapole modes in the triangular lattice). Weprimarily consider PC cavity array structures in a square
lattice [9]; this choice was arbitrary and we would expectsimilar results in the triangular lattice. The coupled cav-ity array and its dipole and quadrupole field patterns areshown in Fig. 1a,b. In these modes, the in-plane electricfield components E x and E y , as well as out-of-plane Bz ,are maximized in the center of the slab. This is commonly
called the transverse electric (TE)-like mode.
For QD lasers, we explored higher Purcell factors with athree-hole defect cavity [17]. Though theoretically limitedto Q ∼ 120, 000 in this design, the fabricated structureshown in Fig. 2 has Q ∼ 3000, reduced by fabricationimperfections and material loss [18].
2.2. Rate equations
A simple rate equations model describes the lasing dynam-ics well. Material gain is averaged over the full mode as theQW or QD layers span the full structure. The mode holds p photons in a volume V m. The laser dynamics are mod-eled with three carrier levels: the excitonic ground state,the pump level carrier number nE (populated above theGaAs-bandgap using a laser with power Lin), and the QWlasing level carrier number nG (resonant with the lasingmode frequency). We then have [19]
d pdt
= g(nG) p + F mnGτ r
− pτ p
(1)
dnGdt
= nE τ E,f
− nG
F m + F PC
τ r+
1
τ PC,nr
− g(nG) p
dnE dt
= η Lin ω p
− nE
1
τ E,r+
1
τ E,nr+
1
τ E,f
In the top equation, the cavity photon number is drivenby the QW through stimulated emission (gain term g(nG) p– see [19]) and SE (at the resonant mode’s Purcell-enhanced
rate F m/τ r). The cavity loses photons at the cavity lossrate 1/τ p. The carrier number nG in the center equation is
pumped by carrier relaxation from the pump level popu-lation nE at rate 1/τ E,f . Besides pumping the cavity, nG
decays through NR channels at rate 1/τ PC,nr and PC leakymodes at rate F PC /τ r , where F PC ≈ 0.2 expresses SErate quenching inside the PC bandgap compared to the SE
rate 1/τ r in the bulk QW (following simulations in [5]).In the bottom equation, the nE level is pumped throughabove-band optical excitation with power Lin at frequencyω p(the first term) and decays through carrier relaxation tonG, NR recombination, and SE (second term).
In the following text, we will use these rate equationsto model the lasing action of single and coupled PC lasers,at both room and low temperature (∼ 10K) and containingQWs or QDs as gain material. We use a logarithmic gainmodel for QWs and a linear gain model for QDs. Underhigh pump power, the rate equations model would requiremodification to account for QD saturation [20].
2.3. Threshold
Solving Eqs. (1) in steady state gives the lasing thresholdpower, defined here as the power where the average photon
number p = 1 [8]. Assuming that most pump-level popu-lation drops into the lasing level (τ E,f τ E,r , τ E,nr), thethreshold is given by
Lin,th = ω pτ pη
nG,th
F PC
τ pτ r
+ τ pτ PC,nr
+ 1
.
In our QW- and QD-driven cavities, the threshold carriernumber in the active volume V a is approximated using thematerial’s transparency concentration, nG,th ≈ N trV a ≈1018 cm−3V mΓ , where the gain confinement factor Γ ≈0.16 approximates the cavity mode overlap with the QWregion. Furthermore, with τ p ∼ 1 ps, τ r ∼ 600 ps, andτ PC,nr ∼ 100 ps, it is easy to see that for our laser structuresthe first term in the brackets dominates, giving
Lin,th ≈ ω pη V aN th
F PC τ r
+ 1
τ PC,nr
(2)
The threshold is thus determined by the gain material’stransparency concentration N tr, radiative loss into non-lasermodes, and nonradiative recombination. In the QD-drivendevices, nonradiative recombination is reduced and theterm F PC /τ r dominates. The factor F PC indicates thatthreshold is reduced by suppression of SE into non-lasing
modes [5]. On the other hand, in QW-driven devices, thenonradiative term 1/τ PC,nr determines threshold. AlthoughEq. (2.3) was derived for steady-state, we find that it is alsoa good approximation for pulsed excitation.
2.4. Laser intensity modulation
Two modulation schemes are used in telecommunications:
small-signal and large-signal modulation [21, 22]. In small-signal modulation, the laser is driven at a constant above-
threshold pump power Lin,0 and modulated with a small sig-nal ∆Lin, resulting in differential changes ∆P = ∆( p/V m)
www.lpr-journal.org © 2008 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
-
8/19/2019 Diskk Lasers Review 2008n
4/11
4 D. Englund, H. Altug, et al.: Ultrafast photonic crystal lasers
and ∆N G = ∆(nG/V a) to the steady-state photon densityP 0 and lasing-level carrier densities N G,0. At low drivingpower above threshold, the differential output power∝ ∆P about the steady-state output P 0 is limited by the relaxationoscillation frequency [6],
ω2R = avgP 0
τ p+
β
τ pτ r/F m+
βN 0P 0τ PC,nrτ r/F m
(3)
Here β = F m/(F PC + F m), as described in [5], a is thedifferential gain, and vg the group velocity. In conventionallasers, only the first term in Eq. (3) is considered as β issmall [21]. Therefore, to increase bandwidth, P 0 and hencethe driving power is raised. The higher power can resultin thermal problems [23], though injection locking mayhelp in VCSELs [24]. Eq. (3) shows that in the high-β
laser, strong cavity effects help to increase ωR without theneed to increase pump power, opening a new pathway forincreasing laser modulation bandwidth [8].
In large-signal modulation, the rate equations predictthat the modulation rate is limited by the pump-level re-laxation time τ E,f and cavity response time τ p = Q/ω.An additional turn-on delay arises as spontaneous emission
builds the cavity field to the point when stimulated emissionbecomes dominant. This delay time is reduced in the high-Purcell regime through faster SE rate and higher β . Thisis seen from the turn-on behavior for laser cavities withdifferent Q in Fig. 3. Here the Purcell factor is calculated as
F m = ξ · 34π2 QV m/(λ/n)3, (4)
where the factor ξ accounts for spatial averaging and is esti-mated at ξ ≈ 0.18 from the measured Purcell rate enhance-ment of the coupled cavity array in Sect. 3.2. The figureshows that as F m is increased, the turn-on delay asymp-totically decreases to a value determined by the carrierrelaxation time and the pump power. The delay decreases
1 2 3 4 5 6
10
P (mW)
0
t(ps)
out
pump P
Q=1600, F = 81
Q=800, F = 41
Q=400, F = 20
Q=200, F = 10
Q=100, F = 5
0
10 0
0
10 0
100
10 0
in
m
m
m
m
m
Figure 3 (online color at: www.lpr-journal.org) Calculated lasing
power P (t) · (V m ωG/τ p) in response to a 3-ps pump pulse (top),for a range of Q. The turn-on delay drops with increasing Q. The
excitation carrier density is 3N tr per pulse for all plots, and pumpefficiency η = 1 in this idealized model.
with pulse energy, which is set here to excite a carrier con-centration of 3 ×N tr. As Q is increased from 100 to 1600,the lasing duration first decreases with faster SE rate, then
extends as it approaches the cavity ring-down time. De-pending on the driving conditions, the modulation rate canbe optimized with a Q that provides high Purcell factor butdoes not excessively slow the cavity response.
2.5. Rate equations model in FDTD
The three-level rate equations model does not account forspatial variations in the carrier concentration across the pho-
tonic crystal device. However, we have found that spatialeffects such as carrier transport from the pump spot to thegain region, or spatial hole burning effects [25], are impor-
tant in understanding lasing efficiency and time response.For that reason, we have developed a finite-difference timedomain model that includes carrier dynamics.
Material gain is implemented in FDTD by an effectiveconductivity σ, as in references [26, 27]. An auxiliary dif-ferential equation is used to describe the evolution of thecurrent density J . In turn, J is related to the carrier densityN G (assumed to be equal for holes and electrons in theintrinsic semiconductors considered here). The set of equa-tions obtained when J = σE is substituted into Maxwell’sequations and is then expressed in the time-domain anddiscretized, as described in [28]. The resulting nonlinearFDTD model allows calculation of the carrier drift into the
lasing mode, and is important for explaining the laser timeresponse measurements covered in Sect. 3.
3. Quantum well photonic crystal lasers
Quantum wells provide large gain when embedded in thecenter of the PC membrane, where the resonant TE-likemode has the maximum electric field energy density. Asingle QW in the PC slab center would see the highestelectric field and hence the highest gain overlap; however,
to optimize the laser current, it is often better to distributecarriers (or current) across several quantum wells [29]. AQW-driven PC nanocavity laser was first demonstrated with
four InGaAsP quantum wells [30] and was soon followedby other demonstrations employing between three and six
quantum wells [31, 32], all operating in the telecommunica-tions band.
3.1. GaAs/InGaAs structures
We first investigated time-domain characteristics of PCnanocavity lasers using a streak camera with a Hama-matsu N5716-03 streak tube. Since the detector responseis limited to wavelengths below 1 µ m, we fabricated PClasers emitting between 900–980 nm. These employ four
8-nm In0.2Ga0.8As QWs separated by 8-nm GaAs barri-ers (see illustration in Fig. 4). The top and bottom QWs
© 2008 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.lpr-journal.org
-
8/19/2019 Diskk Lasers Review 2008n
5/11
Laser & Photon. Rev. (2008) 5
cryos a
4 0 0 5 0 0 6 0 0 7 0 0 8 0 0 9 0 0 1 0 0 0
spectrometer
or OSA
streak camera
PBSHWP
objective
lens
Ti:Saph
pump
pulses
MQWs
substrate
Figure 4 (online color at: www.lpr-journal.org) Confocal micro-
scope setup (lens numerical aperture=0.65). The laser structures
are mounted in a cryostat, which is cooled for some measurements.
Emission is directed to either the streak camera, spectrometer with
cooled Si detector, or optical spectrum analyzer for IR spectra.
Inset: Illustration of coupled cavity array membrane.
are about 32 nm from the center and still see 89% of thecentral maximum field intensity. We use compressivelystrained QWs which have higher differential gain, lower
transparency carrier density N tr, and higher coupling to theTE-like-polarized cavity mode than unstrained QWs [21].We first consider lasers consisting of 172 nm-thick GaAsslabs patterned with 9× 9 arrays of coupled PC cavitiesin a square-lattice PC (Fig. 1d). The structures are fabri-cated by electron beam lithography in polymethyl methacry-
late (PMMA), followed by plasma-etch mask transfer andwet-etch removal of a sacrificial layer beneath the mem-brane. To reduce nonradiative (NR) surface recombinationon the large QW area exposed through PC patterning, thesample was passivated in a (NH4)S solution, which re-sulted in a 3.7-fold reduction in the lasing threshold [33].We found that surface passivation was critical in our sam-ples for room-temperature and continuous-wave (CW) low-
temperature operation.
The structures are pumped optically with 3-ps shortpulses at an 80 MHz repetition rate and a wavelength cen-tered at 750 nm using the confocal microscope as described
in [34] and shown in Fig. 4. High-resolution lasing spec-tra are measured with the spectrometer, while time re-sponse is obtained using a streak camera with 3-ps resolu-tion. At room temperature, the photoluminescence of theIn0.2Ga0.8As quantum wells peaks at 980 nm. For highergain and heat dissipation, we first evaluated cooled struc-tures [6].
The PC array laser in Fig. 1d supports a lasing mode atλmode = 950 nm at low temperature (LT) of 10K (Fig. 5a).
Because of fabrication imperfections, PC holes near theedges of the structure were slightly smaller and cavities
100 200 300 400
5
15
25
5 10 15
5
15
25
10 30 50
20
40
60
80
L (W)in
(d) 293K,pulsed
(e) 10K,CW
(c) 10K,pulsed
model
model model
L (W)inL (W)in
4
c o u n t s / s
1 0
.
2
c o u n t s / s
1 0
.
4
c o u n t s / s
1 0
.
930 950 9700
4
8
12
(nm)
I n t e n s i t y
( 1 0 0
0
c o u n t s )
Q=1520
(a) 10K,pulsed
5 10 15
2
6
10
14
10.4
cts
L (W)in
(f)10K, CW (struct 2)
model
coupled cavity
coupled cavity coupled cavity
coupled cavity single cavity
i n t e n s i t y
( a . u .
)
20 300
unpass-
ivated
passivated
10
(b) 10K,pulsed
0
L (W)in
Figure 5 (online color at: www.lpr-journal.org) QW-driven PC
lasing characteristics (passivated structures). (a) Coupled-cavity
array spectrum below threshold and at low temperature (10K).
The lasing mode consists of an estimated 7–9 cavities. (b) Low-
temperature lasing curve shows threshold reduction after passiva-
tion. (c,d) Low-and room-temperature lasing curves with pulsed
excitation (3.5-pulses at 80 MHz repetition, passivated structure).
(e,f) Continuous excitation lasing curves for coupled and single
cavity. Horizontal axes show average pump power. The fits are
by Eqs. (1).
showed a higher resonance wavelength. As a result, weobserved that coupled cavity modes existed only in a sub-set of the full array. From optical microscope images, weestimate that the lasing mode comprises only 7–9 cavities;the pump beam diameter was adjusted to this size. Fig. 5c
shows the lasing curve for pulsed excitation (3.5 ps at 13 nsrepetition), with an averaged threshold of 6.5 µ W(measuredin front of the objective lens). This corresponds to a largepeak pump power of ∼ 21mW.
The threshold power is much lower under continuouspumping at low temperature. Fig. 5e displays the lasingcurve of the passivated structure, indicating onset of lasingat only ∼ 9 µ W. For a single cavity, threshold is even lower,near 2 µ W, shown in Fig. 5f. This threshold and a similarlylow value recently reported with GaInAsP/InP QWs [35]are lower than in previous low-threshold QW lasers [36,37].
We believe that three main factors reduce threshold inCW operation. First, carrier radiative efficiency is higher in
steady-state lasing as stimulated emission outpaces othernonradiative recombination processes, which are more sig-
www.lpr-journal.org © 2008 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
-
8/19/2019 Diskk Lasers Review 2008n
6/11
6 D. Englund, H. Altug, et al.: Ultrafast photonic crystal lasers
0t(ps)
(a)
0 10 20 30 40
t(ps)2
46
8
10
1017
.
c o n c e n t r a
t i o n ( 1 / c m ) 3
NGNE
pumppump
efficient carrier
conversion
threshold
0 10 20 30 40 50 60
(b)
pump
c o u n t s ( n o r m a l i z e d ) RT LT
experimenttheory
P(t) 5.
laser response:
Figure 6 (online color at: www.lpr-journal.org) Laser time re-
sponse. (a) Experimental data shows response nearly following
the excitation pulse at room temperature; data at both temperatures
are acquired at 2× lasing threshold. (b) Illustration of pump inef-ficiency in pulsed operation. Pump energy is efficiently channeled
into the cavity mode only during lasing (shaded area under P (t)curve, amplified here 5× for visibility); much of the remaining
pump energy is wasted to SE and NR losses.
nificant in pulsed operation (see illustration in Fig. 6b).Second, for the same average pump power, the peak powerof the pulsed beam is thousands of times larger and resultsin a higher temperature of photoexcited carriers. The fasterdiffusion of the high-temperature carriers results in a largereffective pump spot (observed in photoluminescence) withlower gain overlap. Third, we estimate that CW operation
is made even more efficient by carrier drift into cavity. This
drift results from a carrier density gradient caused by spatial
hole burning in the cavity mode (see Fig. 10a). It is expectedto be insignificant for the higher-temperature pumping inpulsed operation [34].
These contributions are quantified by applying the ratemodel of Eqs. (1)(see fits in Fig. 5). All recombination ratesare estimated from time-resolved measurements on non-lasing structures. The model indicates that pump CW ab-sorption efficiency η ∼ 0.055 is far better than in pulsedoperation, where η = 1.3 · 10−3 [34]. The comparisonof CW and pulsed excitation regimes indicates that thereis significant room for improving pumping efficiency inpulsed mode at low temperature.
At room temperature (RT), threshold is higher. The las-ing curve in Fig. 5d indicates a lasing threshold of 68 µ Waverage power. The larger threshold results in part from a
higher transparency concentration, smaller optical gain [21],and larger NR surface recombination rate [33]. These ef-
fects are furthermore exaggerated by heating due to higherthreshold pump power. Because of larger thermal velocity
and diffusion, the above-mentioned carrier drift into the
lasing cavity will be reduced. The larger threshold causesheating in the suspended membrane structures that limitsthe maximum output power, as can be seen in the fall-off in Fig. 5d at ∼ 350 µ W. Because of this heating, weachieved only quasi-CW operation at RT. This required achopper wheel that provided 1 ms-long pulses at a 17 Hzrepetition rate. Heat dissipation can be greatly improvedin RT-operation by fabricating the PC laser structures ontop of low-index substrates such as sapphire or silicon ox-
ide [31,32,38–40], or by replacing QWs with QDs whichhave lower nonradiative loss and carrier transparency [41].We have also found that capping the photonic crystal mem-brane in PMMA improves heat dissipation by up to 20×,
based on measurements of the maximum pump power be-fore the structure is damaged. The capping method alsohelps prevent re-oxidation of passivated structures.
Because of faster carrier dynamics, RT operation results
in faster modulation speed. This is seen in Fig. 6a compar-ing RT and LT lasing response to 3.4-ps-long pump pulses(13 ns repetition). Both measurements were obtained withpump powers roughly 2× above threshold, correspondingto averaged pump powers of 13 µ W and 136 µ W at low-androom temperature, respectively. We measured significantlyfaster lasing response at room temperature, with the lasingpulses roughly following the 3.4-ps pump duration. Fre-quency chirp was less than the cavity linewidth up to ∼ 2×
threshold pump power.The speed-up results primarily because the intraband
relaxation time is shorter at RT; we measured τ E,f
-
8/19/2019 Diskk Lasers Review 2008n
7/11
Laser & Photon. Rev. (2008) 7
"
$ " %
&
(a) bulk
(b) PC, uncoupled
(c) cavity
unpassivated, =618ps
passivated,
=605ps
unpassivated
=34 ps
passivated
=142ps
! " # $ " % ! # &
! " # $ " % ! # &
' ()' )'' *)' +''' +()' +)'' +*)' ('''#,-%.
'/(
'/0
'/1
'/2
+
' () )' *) +'' +() +)' +*) (''#,-%.
'/(
'/0
'/1
'/2
+
' +' (' 3' 0' )'
'/(
'/0
'/1
'/2
+
#,-%.
passivated
=19ps
Figure 7 (online color at: www.lpr-journal.org) Microphotolumi-
nescence from bulk quantum well, PC (uncoupled to cavity array
mode), and non-lasing PC cavity array at 1/2 threshold power
(P in = 12 µ W before objective lens for original and 12 µ W forpassivated structures, pulse length 3.5 ps with 80 MHz repetition).
Measurements at 10K. Solid fits are by Eqs. (1); dashed fits show
exponential decay approximations.
nescence signal decay according to1
τ coupled= F m + F PC
τ r+
1
τ PC,nr
1
τ uncoupled= F PC τ r
+ 1
τ PC,nr
1
τ bulk =
1
τ r+
1
τ bulk,nr
(5)
From bulk measurements, we estimate the natural radia-
tive lifetime τ r ∼ 605 ps, assuming τ bulk,nr τ r. Eqs. (5)then give F m ≈ 28. Repeating these measurements for anunpassivated single-defect cavity gives a spatially averaged
F m ≈ 81 [6]. The high Purcell factor for single cavitiesis not surprising as they are expected to have a maximum
(a)
(b)(c)
Time (ps)
I n t e n s i t y ( a . u . )
P u m p l a s e r
P h C l a s e r
0
0.5
1.0
1.5
2.0
2.5
05 10 1 5 20 2 5
delay ~1.5 ps
w a v e l e n
g t h
Figure 8 (online color at: www.lpr-journal.org) Large-signal
lasing response in QW-driven PC laser. (a) Response to excitation
pulses at (i) 9 ± 0.5 and (ii) 15 ps. (b) Excitation pulse traincreated by etalon setup. Imperfect mirror arrangement causes
an exponential decrease in pulse power and only the first three
pulses exceed the photonic crystal lasing threshold. (c) Lasing
response delay.
F of 165 for the cavity with this set of Q and V m [5], in-dicating that spatial averaging over the mode reduced F mby ∼ 2×. Baba et al. previously estimated SE lifetime en-hancement exceeding 16 (detector response limited) forsimilar structures in GaInAsP PC nanocavities [25].
3.3. Delay time
As we indicated above, an important parameter in the large-signal modulation scheme is the delay time, which de-creases in high Purcell-factor cavities. We measured the de-
lay time at 100K (with 890-nm pump wavelength) as 1.5 ps(Fig. 8c). This delay time is nearly two orders of magnitudeshorter than in previous measurements for VCSELs [44].
3.4. Large-signal modulation
To further demonstrate high-speed characteristics, we di-rectly modulate single-defect cavity lasers at high speedsby pumping with a series of 170-fs pulses generated usinga Fabry-Perot etalon [6]. Fig. 8a,b shows the results fordirect modulation of a nanocavity at low temperature. Thismeasurement shows that in principle, large-signal modula-tion well in excess of 100 GHz is indeed possible. Faster
operation at room temperature is expected, but the etalonmeasurements were not repeated in the passivated structure.
www.lpr-journal.org © 2008 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
-
8/19/2019 Diskk Lasers Review 2008n
8/11
8 D. Englund, H. Altug, et al.: Ultrafast photonic crystal lasers
!"## !"!# !"$# !"%##
#&$
#&'
#&(
#&)
!
+,-.
/ , 0 1 , 2 3 0 4 + , 5 6 -
7 8 3 9 1 : .
/ , 0 1 , 2 3 0 4 + , 5 6 -
7 8 3 9 1 : .
!"#$%&'
# $# '# (##
#&$
#&'
#&(
#&)
!
;3,+
-
8/19/2019 Diskk Lasers Review 2008n
9/11
Laser & Photon. Rev. (2008) 9
I n t e n s i t
y
( a . u . )
! #$%&' (#)*%
(#)*%(#)*%
#+% #,%
#-% #.%
I n t e n s i t y
( a . u . )
/0)/1&2/'(*(3/415
pump
response
I n t e n s i t y
( a . u . )
I n t e n s i t y
( a . u . )
Figure 11 (online color at: www.lpr-journal.org) GaAs PC laser
with InAs QD gain. (a) Measured lasing curve and fit by Eq. (1).
(b) The measured turn-on delay between pump (first peak) and
laser response (second peak) is limited by carrier relaxation.
(c) Measured large-signal modulation speed increases with pump
power. (d) Corresponding fit by rate equations.
effect, while large β and thus higher efficiency and lowerthreshold are achieved. We investigated these aspects inthe 135 nm thick GaAs PC membrane shown in Fig. 2(a),containing a high-density (600 µ m−2) of InAs QDs. Theseself-assembled dots have shallow confinement and operateonly at cryogenic temperature with an emission wavelength
of 940 nm and inhomogeneous linewidth of 20 nm. Lower-energy QDs allow laser operation at room temperature inpulsed [47] and CW modes [41].
The y-polarized fundamental mode (Fig. 2(b)) is reso-nant in the structure near 920 nm, with cold-cavity Q ∼3000(τ p ∼ 1.5 ps). We measure a gradual onset of lasingnear 1 µ W, as shown in Fig. 11a. From fits to the lasingcurve, we estimate a SE coupling factor β ∼ 0.2. Streak camera measurements of the rise time of photolumines-cence from quantum dots in bulk GaAs indicate that the car-
rier relaxation time τ E,f ∼ 10 ps for a wide range of pumppowers. We also find that resonant pumping of higher-order
confined states of the QDs (such as p-level states) doesnot appreciably lower τ E,f . Because the carrier capturetime is longer than the cavity photon lifetime, it ultimatelydetermines the maximum modulation bandwidth. This iswhat we observe in Fig. 11b which shows a delay of 13.5 ps
(at five times threshold) and does not drop below 12 ps forhigher powers. Simulations with Eqs. (1) support this ob-servation as rise time is limited by the carrier capture time.In our cavity-QED-enhanced structure, the relaxation-timelimit is rapidly reached in the high-β case. In contrast, innon-PC quantum dot lasers not employing strong cavityeffects, far higher pump power is needed to reach this limit.
Once lasing is reached, stimulated emission causes fast
carrier recombination. We measured a decay time of 8.5 psat pump powers around five times the threshold (Fig. 11c).
For higher pump powers the laser response appears largelyunchanged, presumably due to carrier saturation. We againmodel the system with Eqs. (1), employing a linear gain
model and parameters given in [48]. Fig. 11d shows thesimulated laser response at various pump powers, demon-strating good agreement between theory and experiment.
The present work predicts that large-signal modulationin present PC lasers employing conventional self-assembled
In(Ga)As QDs is limited to ∼ 30 GHz due to relaxationdynamics. While there is also evidence to suggest thatcarrier relaxation and hence maximum modulation rateactually further slows at increased temperature [49], re-cent advances in QD growth can open the way to higherperformance. QDs driven through phonon-assisted tunnel-ing show very short relaxation time, with τ E,f ∼ 2 ps atroom temperature [50], and were recently demonstrated in
ridge waveguide lasers with 25 GHz small-signal modula-tion bandwidth [50]. This bandwidth may be significantly
improved using a PC laser cavity. In addition, reduction inthe inhomogenous linewidth broadening and reduction inhot-carrier effects and associated gain compression [51],will improve PC QD laser efficiency and speed. P-typedoping of quantum dots also promises to speed up car-rier dynamics [52].
5. Conclusions and future directions
Photonic crystal lasers provide unprecedented speed, reach-ing pulses on picosecond scales. They also show verylow threshold, lasing at only several microwatts of pumppower. Their planar design makes them ideal on-chip in-tegration. But for practical applications, PC lasers willneed to be pumped electrically. Many groups are currentlypursuing this goal, and electrically driven single-cavityPC lasers have been demonstrated in free-standing mem-branes [53, 54] and band-edge laser structures [55]. Forhigh-speed electrical modulation in extended structuressuch as band edge and nanocavity array lasers, it will beimportant that the structure be uniformly pumped, as thespatial modeling of carrier dynamics in Sect. 3.6 suggests.
A further challenge for any PC laser will be keeping RC time constants small, where C and R are the capacitance
and resistance of the laser. Compared to VCSELs, PC laserspromise far lower capacitance due to small a footprint andlower resistance because of thin intrinsic material betweenelectrodes. A promising recent step demonstrated time con-
stants below 10 ps using micron-scale contacts with sub-fFcapacitance [56]. With recent advances in integration, elec-trical pumping, and ultrafast operation, PC crystal laserspromise to fill a growing need for integrated, ultrafast opti-cal communication.
Acknowledgements This work was supported by the MARCO
IFC Center, NSF Grants ECS-0424080 and ECS-0421483, the
MURI Center (ARO/DTO Program No. DAAD19-03-1-0199), as
well as NDSEG & NSF Fellowships (D.E.) and Stanford GraduateFellowship (B.E.).
www.lpr-journal.org © 2008 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim
-
8/19/2019 Diskk Lasers Review 2008n
10/11
10 D. Englund, H. Altug, et al.: Ultrafast photonic crystal lasers
Dirk Englund is a graduate studentin applied physics at Stanford Uni-
versity. He received his Bachelor’sof Science degree in physics fromthe California Institute of Technol-ogy in 2002. He is a recipient of the NSF, NDSEG, Stanford Graduate
Mayfield, and U.S. Fulbright fellow-ships. His research focuses on quan-
tum photonic devices.
Hatice Altug received the B.S. de-gree in Physics from Bilkent Univer-sity, Turkey in 2000. She receivedthe M.S. and Ph.D. degree in electri-cal engineering and applied physicsfrom Stanford University in 2006.Currently, she is a Peter Paul Career
Development Professor in Electricaland Computer Engineering Depart-
ment at Boston University. Her research involves de-sign and implementation of high performance and ultra-compact nano-photonic devices and sensors includinglasers and all-photonic switches and their large-scale on-chip integration for communication and bio-sensing ap-plications.
Bryan Ellis was born in Denver, Col-orado in 1983. He received the B.S.Edegree in electrical engineering from
Princeton University in 2005. He iscurrently working towards a Ph.D.degree in electrical engineering from
Stanford University. His research in-terests include nanophotonic devicesemploying optical microcavities for
use in optical communications and optical intercon-nect technologies.
Jelena Vuckovic received the PhD de-gree from Caltech in 2002, and hasbeen working at Stanford Universityas a faculty since 2003. Her researchfocuses on nano- and quantum pho-tonic devices and circuits. She is anauthor of more than 60 publicationsin refereed journals, more than 70invited and plenary talks, five book
chapters, five issued and several pending U.S. patents,and a recipient of numerous awards, including the Of-fice of Naval Research Young Investigator Award andthe Frederick Terman Fellowship, given to the mostpromising young faculty in sciences and engineeringat Stanford.
References
[1] K. Nozaki, S. Kita, and T. Baba, Opt. Express 15(12), 7506–
7514 (2007).
[2] D. Englund, A. Faraon, I. Fushman, N. Stoltz, P. Petroff,
and J. Vučković, Nature 450(6), 857–61 (2007).
[3] S. Strauf, K. Hennessy, M.T. Rakher, Y. S. Choi,
A. Badolato, L. C. Andreani, E. L. Hu, P. M. Petroff, and
D. Bouwmeester, Physical Review Letters 96(12), 127404
(2006).
[4] B. S. Song, S. Noda, T. Asano, and Y. Akahane, Nature
Materials 4, 207 – 210 (2005).
[5] D. Englund, D. Fattal, E. Waks, G. Solomon, B. Zhang,
T. Nakaoka, Y. Arakawa, Y. Yamamoto, and J. Vučković,
Physical Review Letters 95(July), 013904 (2005).
[6] H. Altug, D. Englund, and J. Vučković, Nature Physics 2,
484–488 (2006).[7] T. Yoshie, M. Lončar, A. Scherer, and Y. Qiu, Applied
Physics Letters 84(18), 3543–3545 (2004).
[8] G. Bjork and Y. Yamamoto, IEEE Journal of Quantum
Electonics 27(11), 2386–96 (1991).
[9] H. Altug and J. Vuckovic, Applied Physics Letters 86(11),
111102 (2005).
[10] K. Inoue, M. Sasada, J. Kawamata, K. Sakoda, and J. Haus,
Japanese Journal of Applied Physics 38, L157–59 (1999).
[11] M. Meier, A. Mekis, A. Dodabalapur, A. Timko, R. E.
Slusher, J. D. Joannopoulos, and O. Nalamasu, Applied
Physics Letters 74(1), 7–9 (1999).
[12] J. Mouette, C. Seassal, X. Letartre, P. Rojo-Romeo, J. L.
Leclereq, P. Regreny, P. Viktorovitch, E. Jalaguier, R. Per-
reau, and H. Moriceau, Electronics Letters 39(6), 526–528(20 March 2003).
[13] X. Letartre, C. Monat, C. Seassal, and P. Viktorovitch, J.
Opt. Soc. Am. B 22(12), 2581–2595 (2005).
[14] K. Sakoda, K. Ohtaka, and T. Ueta, Opt. Express 4(12),
481–489 (1999).
[15] T. D. Happ, M. Kamp, A. Forchel, J. L. Gentner, and
L. Goldstein, Applied Physics Letters 82(1), 4–6 (2003).
[16] D. G. Deppe, J. P. van der Ziel, N. Chand, G. J. Zydzik, and
S. N. G. Chu, Applied Physics Letters 56(21), 2089–2091
(1990).
[17] Y. Akahane, T. Asano, B. S. Song, and S. Noda, Nature
425(October), 944–947 (2003).
[18] D. Englund, A. Faraon, B. Zhang, Y. Yamamoto, and
J. Vuckovic, Optics Express 15(April), 5550–8 (2007).[19] For InAs MQW, g(nG) = Γ g0vg ln(nG/V aN tr), V a =
active material volume, g0 ≈ 2400 /cm, group velocityvg = c/n, N tr ≈ 3 · 10
18 cm−3(from [21]); gain confine-
ment factor Γ ≈ 0.16; η : pump power absorption ratio;τ r: SE lifetime in unpatterned QW; τ PC,nr: NR lifetime inPC; τ E,f , τ E,r , τ E,nr: lifetimes of pump-level relaxation,SE, and NR transitions (resp). At RT, we model SE rate
as Rsp = BN 2
G in the intrinsic QW; the ratio in B forcavity and bulk emissions gives SE rate modification. For
InAs QDs at 10K, g(nG) = Γ g0(nG/V a − N tr), g0 =8.2 · 10−6cm3 /s, N tr = 3.2 · 10
17 cm−3, Γ = 0.028.[20] M. Grundmannand D. Bimberg, Phys. Rev. B 55(15), 9740–
9745 (1997).
[21] L. A. Coldren and S. W. Corzine, Diode Lasers and Pho-tonic Integrated Circuits (New York: Wiley, 1995).
© 2008 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim www.lpr-journal.org
-
8/19/2019 Diskk Lasers Review 2008n
11/11
Laser & Photon. Rev. (2008) 11
[22] C. H. Chang, L. Chrostowski, and C. J. Chang-Hasnain, J.
Sel. Top. Quantum Electron. 9, 1386–1393 (2003).[23] K. L. Lear and al et, Advances in Vertical Cavity Surface
Emitting Lasers in Trends in Optics and Photonics Series15, 6974 (1997).
[24] L. Chrostowski, X. Zhao, and C. Chang-Hasnain, Mi-
crowave Theory and Techniques, IEEE Transactions on
54(2), 788–796 (Feb.2006).[25] T. Baba, D. Sano, K. Nozaki, K. Inoshita, Y. Kuroki, and
F. Koyama, Applied Physics Letters 85(18), 3989–3991
(2004).[26] M. Kretschmann and A. A. Maradudin, J. Opt. Soc. Amer.
B, Opt. Phys. 21(January), 150158 (2004).[27] S. C. Hagness, R. M. Joseph, and A. Taflove, Radio Sci
31(4), 931941 (1996).[28] W. H. Pernice, F. P. Payne, and D. F. Gallagher, Opt. Ex-
press 15(18), 11433–11443 (2007).
[29] L. A. Coldren and S. W. Corzine, Diode Lasers and Pho-tonic Integrated Circuits (New York: Wiley, 1995).[30] O. Painter, R. Lee, A. Scherer, A. Yariv, J. O’Brien, P. Dap-
kus, and I. Kim, Science 284(June), 1819–1821 (1999).[31] J. K. Hwang, H. Y. Ryu, D. S. Song, I.Y. Han, H. W. Song,
H. K. Park, Y. H. Lee, and D. H. Jang, Applied Physics
Letters 76(21), 2982–2984 (2000).[32] C. Monat, C. Seassal, X. Letartre, P. Viktorovitch,
P. Regreny, M. Gendry, P. Rojo-Romeo, G. Hollinger,
E. Jalaguier, S. Pocas, and B. Aspar, Electronics Letters
37(12), 764–766 (7 Jun 2001).[33] D. Englund, H. Altug, and J. Vučković, Appl. Phys. Lett.
91(July), 071124 (2007).[34] D. Englund, H. Altug, I. Fushman, and J. Vučković, Appl.
Phys. Lett. 91(July), 071126 (2007).
[35] K. Nozaki, S. Kita, and T. Baba, Opt. Express 15(12), 7506–7514 (2007).
[36] M. Nomura, S. Iwamoto, and M. Nishioka, Appl. Phys.
Lett. 89(161111) (2006).[37] M. H. Shih, W. Kuang, A. Mock, M. Bagheri, E. H. Hwang,
J. O’Brien, and P. Dapkus, Appl. Phys. Lett. 89(101104)
(2006).[38] C. Monat, C. Seassal, X. Letartre, P. Regreny, M. Gendry,
P. R. Romeo, P. Viktorovitch, M. L. V. d’Yerville, D. Cas-
sagne, J. P. Albert, E. Jalaguier, S. Pocas, and B. Aspar,
Journal of Applied Physics 93(1), 23–31 (2003).[39] G. Vecchi, F. Raineri, I. Sagnes, A. Yacomotti, P. Monnier,
T. J. Karle, K. H. Lee, R. Braive, L. L. Gratiet, S. Guilet,
G. Beaudoin, A. Taneau, S. Bouchoule, A. Levenson, and
R. Raj, Opt. Express 15(12), 7551–7556 (2007).
[40] B. B. Bakir, C. Seassal, X. Letartre, P. Regreny, M. Gendry,
P. Viktorovitch, M. Zussy, L. D. Cioccio, and J. M. Fedeli,
Opt. Express 14(20), 9269–9276 (2006).
[41] M. Nomura, S. Iwamoto, K. Watanabe, N. Kumagai,Y. Nakata, S. Ishida, and Y. Arakawa, Opt. Express 14(13),
6308–6315 (2006).
[42] M. Asada, IEEE J. of Quant. Electron. 25(9), 2019–2026
(1989).
[43] J. Y. Duboz, E. Costard, E. Rosencher, P. Bois, J. Nagle,
J. M. Berset, D. Jaroszynski, and J. M. Ortega, Journal of
Applied Physics 77(12), 6492–6495 (1995).
[44] H. Deng, Q. Deng, and D. G. Deppe, Applied Physics Let-
ters 70(6), 741–743 (1997).[45] H. Altug and J. Vučković, Opt. Express 13(22), 8819–8828
(2005).
[46] D. Englund, H. Altug, and J. Vučković, to be published
(2008).
[47] T. Yoshie, O. Shchekin, H. Chen, D. Deppe, and A. Scherer,Electronics Letters 38(17), 967–968 (2002).
[48] B. Ellis, I. Fushman, D. Englund, B. Zhang, Y. Yamamoto,
and J. Vuckovic, Appl. Phys. Lett. 90(June), 151102 (2007).
[49] J. Urayama, T. B. Norris, H. Jiang, J. Singh, and P. Bhat-
tacharya, Applied Physics Letters 80(12), 2162–2164
(2002).
[50] Z. M. S Fathpour and P. Bhattacharya, Journal of Physics
D: Applied Physics 38(13), 2103–2111 (2005).
[51] D. R. Matthews, H. D. Summers, P. M. Smowton, and
M. Hopkinson, Applied Physics Letters 81(26), 4904–4906
(2002).
[52] D. Deppe, H. Huang, and O. Shchekin, Quantum Electron-
ics, IEEE Journal of 38(12), 1587–1593 (Dec 2002).
[53] H. Park, S. Kim, S. Kwon, Y. Ju, J. Yang, J. Baek, S. Kim,and Y. Lee, Science 305, p. 1444–1447 (2004).
[54] M. K. Seo, K. Y. Jeong, J. K. Yang, Y. H. Lee, H. G. Park,
and S. B. Kim, Applied Physics Letters 90(17), 171122
(2007).
[55] H. Matsubara, S. Yoshimoto, H. Saito, Y. Jianglin,
Y. Tanaka, and S. Noda, Science 319(5862), 445–447
(2008).
[56] R. Schmidt, U. Scholz, M. Vitzethum, R. Fix, C. Metzner,
P. Kailuweit, D. Reuter, A. Wieck, M. C. Hübner, S. Stufler,
A. Zrenner, S. Malzer, and G. H. Döhler, Applied Physics
Letters 88(12), 121115 (2006).
www.lpr-journal.org © 2008 by WILEY-VCH Verlag GmbH & Co. KGaA, Weinheim