Room temperature all-silicon photonic crystal nanocavity light emitting diode at sub-bandgap wavelengths
A. Shakoor, R. Lo Savio, P. Cardile, S. L. Portalupi, D. Gerace, K. Welna, S. Boninelli, G. Franzo, F. Priolo, T. F. Krauss, M. Galli, L. O Faolain
RRoom temperature all-silicon photonic crystalnanocavity light emitting diode at sub-bandgapwavelengths
A. Shakoor, R. Lo Savio, P. Cardile, S. L. Portalupi, D. Gerace, K. Welna, S. Boninelli, G. Franzò, F. Priolo, T. F. Krauss, M. Galli, and L.O’Faolain SUPA,
School of Physics and Astronomy, University of St. Andrews, Fife KY169SS, St. Andrews, United Kingdom Dipartimento di Fisica, Università di Pavia, via Bassi 6, 27100 Pavia, Italy CNR-IMM MATIS and Dipartimento di Fisica e Astronomia, Università di Catania,via S. Sofia 64, 95123 Catania, Italy* [email protected]: +44-1334463091
ABSTRACT
Silicon is now firmly established as a high performance photonic material. Its onlyweakness is the lack of a native electrically driven light emitter that operates CW atroom temperature, exhibits a narrow linewidth in the technologically important 1300-1600 nm wavelength window, is small and operates with low power consumption.Here, an electrically pumped all-silicon nano light source around 1300-1600 nm rangeis demonstrated at room temperature. Using hydrogen plasma treatment, nano-scaleoptically active defects are introduced into silicon, which then feed the photoniccrystal nanocavity to enahnce the electrically driven emission in a device via Purcell effect. A narrow (Δλ = 0.5 nm) emission line at 1515 nm wavelength with a power density of 0.4 mW/cm is observed, which represents the highest spectral powerdensity ever reported from any silicon emitter. A number of possible improvementsare also discussed, that make this scheme a very promising light source for opticalinterconnects and other important silicon photonics applications.Abstract Figure: Tunable electrically pumped silicon nanolight source attelecommunication wavelengths. KEYWORDS: silicon photonics, optically active defects, nano light sources,Photonic crystal cavity, Silicon light emission
1. Introduction
The absence of a cheap, efficient and electrically driven silicon based light emittercreates a significant barrier to the use of silicon photonics in low cost, high volumeapplications- a major issue for key fields such as optical interconnects and bio-sensing. Furthermore, micro- and nano- emitters are necessary for achieving a highintegration density and a high channel count. The addition of III-V materials throughbonding or epitaxy is currently the preferred solution giving efficient on-chip lasers[1], but the use of a costly material with complex processing makes mass manufacturechallenging. A true group-IV nano light emitter remains the ultimate solution thatwould allow the full potential of silicon photonics to be realized. Many approacheshave been employed to improve the luminescence of silicon, such as siliconnanocrystals that exhibit electroluminescence [2] and optically pumped gain [3] in thevisible and near-infrared wavelength ranges. Stimulated emission has been observedfrom silicon quantum wells [4] and the incorporation of rare earth dopants hasenabled optically pumped transparency, e.g. from erbium doped silicon nitride nano-cavities [5]. Highly doped, strained Germanium can be made to exhibit a directtransition with optically pumped lasing [6] and electroluminescence [7] reported.Raman lasers have been realised that produce very high output powers [8], however,there is no possibility of electrical driving. While these approaches have muchpotential, they do not combine all of the desired characteristics of a silicon lightsource, namely: electrical pumping, operation at sub-bandgap wavelengths, roomtemperature operation, small size and narrow emission linewidth.Optically active defects offer an alternative and particularly powerful approach toimproving the luminescence of native silicon. Such defects add in conservingmomentum between recombining carriers, problematic due to the indirect bandgap ofsilicon, and may create luminescence lines and bands including emission in theimportant telecommunication windows [9-11]. Examples for such defect-basedemission includes the formation of dislocation loops that have been used in broad areaelectrically driven LEDs operating near the silicon band-edge [12] or the creation of“A-centres” that have been used to demonstrate stimulated emission at cryogenictemperatures [13]. Here, we use hydrogen plasma treatment to incorporate nano-scaleoptically active defects into the silicon host and the high Purcell enhancementavailable with high Q-photonic crystal cavities to create an electrically driven all-silicon nano-LED, schematically shown in Fig. 1a, that demonstrates the desiredcharacteristics, listed above, of a silicon light source.
Figure 1. Schematic and mode profile of fabricated all- silicon PhC nanocavityLED . a , SEM micrograph of a silicon photonic crystal nano-LED. The doped regionsare shown schematically and extend well into the photonic crystal. Far-fieldoptimization gives a vertical, near Gaussian output beam. b , near-field and c , far-fieldintensity of the fundamental cavity mode, calculated by FDTD simulations of thefabricated device. A schematic view of the modified holes around the cavity is givenin the near-field plot b .
2. Photonic crystal nanocavity design
We use L3 photonic crystal (PhC) cavities in a suspended 220nm thick silicon slab,fabricated using electron beam lithography and reactive ion etching. The period was420 nm and the normalized radius (r/a) was 0.28. We have previously observed Q-values in excess of 100,000 in this system [14]. We then apply the band-foldingtechnique to improve the vertical light extraction by engineering a near Gaussian far-field emission of the fundamental cavity mode [15, 16]. Corresponding finite-difference time-domain (FDTD) simulations are shown in Fig. 1b-c. The details of thedesign and fabrication of PhC nanocavity used are given in Methods section.
3. Creating optically active defects
Our method of creating optically active defects proceeds by treating the samples ina hydrogen plasma [9, 10], which generates surface defects at which hydrogen mayget trapped [11, 17-18] or causes hydrogen to attach to pre-existing defects, i.e.created during reactive ion-etching of the PhC. To better understand the nature andthe location of the defects created by the hydrogen plasma treatment, we carried outtransmission electron microscopy (TEM) analysis of the hydrogen plasma treatedPhotonic Crystals and Czochralski silicon (Cz-Si). We performed both planar view(PV) and cross-sectional (CS) TEM measurements using a JEOL JEM 2010instrument operating at an acceleration voltage of 200 kV.Fig. 2a shows a PV TEM image of the region containing the PhC using bright fieldimaging. The black spots indicate the presence of extended defects induced by thehydrogen plasma treatment during reactive ion etching (RIE). Even though the defectsare present throughout the entire silicon surface, it is evident that their concentrationincreases towards the sidewalls of the holes. This effect is a consequence of theplasma treatment, during which, hydrogen impacts on all exposed surfaces andproduces extended damage. To better illustrate the structure of the resulting defectpopulation, a defocused off-Bragg CS view of the plasma treated Cz-Si substrate isshown in Fig. 2b.We notice different types of defects as a function of penetration depth. Near thesurface, concentrated within the first 10 nm, the defect population is dominated bynanobubbles, whose size is a few nm. Their exact nature is unknown, but they mostlikely consist of agglomerates of vacancies. Going further down, we find apreponderance of platelets (indicated by white arrows in Fig. 2b), whose meandiameter is about 10-15 nm, occupying the (100) plane (parallel to the surface) andthe {111} planes. Similar platelets were observed in refs [19-21]. A high resolutionimage of a few platelets is shown in the top left inset in Fig. 2b. Moreover, some ofthe dark traces, located between the two previously described regions, exhibit thetypical “coffee bean” shape indicative of dislocation loops. One of them is shown inthe top right inset of Fig. 2b. While the existence of these hydrogen plasma induceddefects has been described before [19-21], their luminescence properties are still notwell understood.
Figure 2. Transmission electron microscopy (TEM) images showing the defectsin silicon created by hydrogen plasma treatment . a , Planar view (PV) TEM imageof hydrogen plasma treated PhC. The defects concentration increases incorrespondence of the holes sidewalls. The inset shows a zoom-in region close to asingle hole, where some defects are indicated by arrows. b , Cross- sectional (CS)view of a plasma treated Cz-Si sample. White arrows indicate platelets and yellowarrows indicate extended defects. In the top left inset a high resolution image of theplatelets is given, while in the top right inset a typical coffee bean shape of adislocation loop is shown.
4. Luminescence of optically active defects within a PhC nanocavity
Fig.3 shows the photoluminescence (PL) spectra of hydrogen plasma treatedPhC nanocavities with different lattice constants, as compared with that of bulksilicon (Cz-Si) (see Methods section for details of the excitation scheme). Thephotoluminescence signal from the bulk silicon is very weak, while the treated PhCcavities show a background signal with strong, sharp peaks corresponding to thefundamental mode of each PhC cavity. The enhancement of background PL is due toincorporation of optically active defects by hydrogen plasma treatment while the peakreflects the enhancement due to the Purcell effect and the improved extraction due tothe nanocavity resonance and Photonic bandgap effects. Combining these effects, weobserve an overall 40000-fold (4 orders of magnitude) increase of the PL signal atroom temperature relative to bulk silicon. As demonstrated in Fig 3, the emission lineis easily tunable through the 1300-1600 nm wavelength range and also demonstratesthe robustness and repeatability of this method. Following refs. 22-24, we estimate aPurcell enhancement factor of ~10, which is responsible for an increase in theradiative recombination rate increase and subsequent suppression of thermalquenching.
Figure 3. Comparison of the photoluminescence of hydrogen treatednanocavities with Czochralski silicon.
Peaks correspond to the fundamental mode ofcavities with different lattice period, a. The emission line (PL of fundamental cavitymode) of the hydrogen plasma treated cavity is over four orders of magnitude higherthan bulk silicon over the entire range and tunable between 1300 and 1600 nm.We found that the key parameter is the duration of the plasma treatment,which impacts both on the total PL intensity and on the cavity quality factor. The PLemission is characterized by two broad bands centred around 1300 and 1500 nm (seeFig. 4a). In Fig. 4b we show the PL emission intensity, integrated between 1400 and1600 nm, for different treatment duration and for two different plasma compositions -pure hydrogen (20 sccm gas flow) and a hydrogen/argon mix (20 sccm and 4sccm gasflow for hydrogen and argon respectively). The signal from the samples treated withpure hydrogen is clearly stronger and quickly reaches the maximum value. The argonwas added to the plasma in order to elucidate the role of physical damage, which turnsout to inhibit the signal enhancement, especially for the longer treatment times; usingpure argon plasma gave a very weak PL signal (not shown), clearly suggesting thathydrogen plays the key role in the emission process [19].In addition, we also measured the cavity Q-factors using the resonant scatteringtechnique [14] and a typical spectrum is shown in Fig. 4c. The Q-factor follows asimilar trend as the PL intensity for different treatment duration, Fig. 4d, that is, anincrease for hydrogen-treated cavities, and a decrease when the plasma containedargon. This is also consistent with our observation that argon causes physical damageand introduces roughness to the silicon surface. Such roughness increases the opticalloss of the PhC cavity, thus reducing the overall Q-factor. In contrast, the hydrogenplasma treatment increased the Q-values. In fact, a close comparison of SEM pictures,not shown here, indicates that the sidewalls of the PhC cavities become smoother afterhydrogen plasma treatment, giving rise to a smoothing or “plasma polishing” effectthat can explain the observed increase in Q-value. Overall, we conclude that hydrogenhas the dual benefit of increasing both the PL signal and the Q-factor.0
Figure 4. The effect of plasma treatment duration on photoluminescence and thecavity Q-factor. a , Photoluminescence observed from hydrogen-treated silicon-on-insulator (SOI). b , Effect of treatment duration on the integrated PL intensity; the redcurve represents treatment with pure hydrogen plasma and the black curve showstreatment with a mixed hydrogen/argon plasma. c , Resonant scattering spectrum ofthe highest Q-cavity with the best-fit of the experimental data to a Fano lineshape (redline) [14]. d , Effect of different plasma conditions on the cavity Q factors.These results were achieved for samples treated after the photonic crystal etch step,i.e. in the presence of the etched holes as shown in Fig. 2a. We also performed theplasma treatment prior to the etching step and found a much weaker enhancement.1This reduced enhancement highlights the importance of the exposed surface area andthat the observed PL enhancement is mainly a surface effect. Clearly, the hydrogen ions only have very low energy (≈ 400 eV) and, consequently, a low penetrating power into the bulk material. In addition, hydrogen plasma treatment is known topassivate the silicon surface. We expect that it reduces carrier recombination atsurface defects and thus improves the efficiency of radiative recombination [25].
5. All-silicon nano- LED
Finally, we proceed to electroluminescent devices. Since photonic crystalfabrication is fully compatible with ULSI processes [26], we create pin junctionsusing multiple lithography and ion implantation steps providing a monolithic sourceof electron-hole pairs to feed our optically active defects. Here, we create fingers ofdoped regions, as marked in Fig. 1a, which extend into the photonic crystal, similar to[27]. This forces carriers to recombine in the cavity by virtue of the low resistivity ofthe two highly doped fingers. The conditions were carefully optimised to provide thebest possible current injection, and to finely control the position of the depletionregion with respect to the cavity region.Electroluminescence (EL) was generated by applying a forward bias across thejunction. Light was collected with the same experimental apparatus used for PLmeasurements, thus allowing a direct comparison between PL and EL emissionintensities. This comparison is shown in Fig. 5a, where we report the maximum powerspectral density expressed in pW/nm for the fundamental cavity mode for both PL andEL. The photoluminescence is recorded for a pump power of 0.8 mW and at anexcitation wavelength of 640 nm. The electroluminescence is recorded at an appliedvoltage of 3.5 V, with a current of 156.5 µA, thus consuming an electrical power of0.55 mW. Remarkably, the EL signal is more intense than the PL signal across the2entire spectral range, in contrast to that usually observed, and is a testament to thepotential of this system as an electrically driven source. The PL signal is lower thanEL for almost same input power due to the low absorption (~5%) of the thin siliconslab at the 640 nm excitation wavelength.Fig. 5 also shows an optical image of the device under zero voltage (Fig. 5b) andwith a 3.5 V bias applied (Fig. 5c). The latter image was captured with an infra-red(IR) camera and the bright emission spot is clearly visible as soon as the voltage isturned on. The Q-factor of the cavity for the electrical devices was 4000, while forbare PhC cavities (before device fabrication) higher Q-values were observed (see Fig.4c). This reduction is a consequence of imperfections introduced during the pin junction fabrication process as well as free carrier absorption.
Figure 5: Electroluminescence and Photoluminescence from the device . a ,Comparison of EL and PL from a PhC nanocavity (with integrated pin junction)treated with hydrogen plasma. b , Micrograph (top view) and c , filtered IR picture ofthe device showing strong electrically driven emission from the silicon PhC3nanocavity at room temperature (a low-pass filter with a cut-off at 1500 nm wasused). Integrating over the fundamental cavity mode, the emission power is 4pW(electrically driven). Integrating over the full wavelength range (1200-1600nm), wemeasure a total power of 45 pW. The actual power generated in the cavities is 10-100times higher than the measured output powers, considering the percentage of lightemitted vertically, collection efficiencies and losses in the setup. It should also benoted that only a fraction of the optical pump is absorbed in the top silicon layer,resulting in coincidentally similar values for the optical and electrical pump power.The spectral density of the emission is the key figure of merit for manyapplications such as Wavelength Division Multiplexing (WDM) for communications,multimodal operation for biosensing or spectral coherence for interferometry. In thefundamental cavity mode, our device gives a spectral density of 10 pW/nm (byconsidering the active area, 800µW/nm/cm ), which is by far the highest reportedvalue for any silicon-based electrically-driven nano-emitter (even without restrictingthe comparison to room temperature emission and telecom wavelengths- see table 1),thus encouraging further efforts for the realization of the first Si-based electricallypumped laser. The compatibility of these optically active defects with electricaldriving even allows the traditionally inefficient silicon emission to give a measuredoutput power that is comparable with that of most electrically driven III-V photoniccrystal-lasers and LEDs [28-30], where collected powers are reported to be in thepicowatt to nanowatt level, similar to the power levels reported here. Additionally,our nano-LED is easily tunable in the 1300-1600 nm range, by chnging the latticeperiod of the PhC lattice, as demonstrated in Fig 3.4Table 1. Comparison of different band-edge and sub-bandgap silicon light sources InjectionCurrent Operatingwavelength λ (nm)
Operatingtemperature(K) Emissionlinewidth
Δλ (nm)
Powerdensity*µW/cm PowerspectraldensityµW/nm/cm Wall plugefficiency**
CommentsSi:Dislocationloops [12] 50mA 1150 ~300 90 600 6.6 10 -4 Broad areadevice,Band-edge(structured) [31] 130mA 1150 ~300 50 300*** 6*** 10 -2 Broad areadevice,A-centers [13] optical 1280 <70 0.5 a.u a.u not reported Works onlyat cryogenictemperatures.W-centres [32] 2mA 1218 6 ~1 2.7 2.7 10 -7 Works onlyat cryogenictemperaturesErbium in SiN(membraned) [5] optical 1565 5.5 0.03 a.u a.u not reported Works onlyat cryogenictemperatures.Erbium in SiN(on Si) [33] 1.5A/cm Optical visible1550 ~300 100 a.u a.u not reportedBandedge(PhCs) [24,34] optical 1100 ~300 <0.5 a.u a.u not reported Band-edgeemissionThis work 156 µA 1200 to1600 ~300 0.5 400 800 10 -8 Sub-bandgap,roomtemperatureoperation,electricallypumped,small size.Suitable forinterconnectand othersiliconphotonicsapplications * Calculated using the emitting area** The wall plug efficiency is calculated by considering the measured outputpower.*** Output powers expressed in arbitrary units (a.u). Power calculated on thebasis of efficiencies and drive current, voltage and area.5As a consequence of the very narrow emission linewidth (0.5 nm), the powerefficiency of our LED is 0.7x10 -8 , (based on the collected power). This efficiencyshould hence be compared with those of other micrcon scale emitters with narrowlinewidths. In fact, our efficiencies are within an order of magnitude of thosereported- 1.6-3 x10 -8 [29] and 8x10 -8 [28]- for direct bandgap (III-V) lasers operatingbelow threshold, a remarkable result considering the indirect band gap of silicon.The device is temperature stable up to 350 o C, after which hydrogen starts todiffuse out of the silicon matrix, resulting in a reduction of the luminescence signal.For example, annealing the device at 500 o C for 1 hour reduced the luminescence by afactor of 4 to 5. This also indicates the importance of hydrogen in the emissionprocess. The physical defects created by the plasma treatment and decorated byhydrogen are therefore essential for the emission process.There is a very slight (0.1nm) red shift of the mode under active bias conditions,There may also be a masked blue shift due to the injected carriers, estimated to be lessthan 0.5nm [22,27]. This indicates that the temperature of the device increases by only10-20 o C, which is well below the temperature beyond which hydrogen outdiffusionoccurs (please see methods section for more details). Damage to the device or areduction in the EL level was also not observed under active bias conditions. Thus nodetrimental effect of the active bias conditions was apparent during this experiment.There is still considerable scope for improvement of our device, and it is realistic toexpect laser operation in due course. A key to improving the operation and deviceefficiency is to better understand the incorporation of defects into the silicon host. Thelow ion energies in the plasma treatment we currently use results in low penetrationdepths of the ions, resulting in defect formation close to the surface (see Fig. 2b),which limits the active volume available. The luminescence is also very broadband6indicating that most of the emitters are off resonance and do not contribute to theluminescence; therefore, only a tiny fraction of the defects actually emits into thecavity mode and experience PhC enhancement. Ion implantation, the preferredapproach for defect based photodetection [35], has much potential in this respect, as itgives access to a wide range of species and energies thus providing a large parameterspace for further optimization. The small mode volume of our device, thoughimportant for a high Purcell factor, is also a big limitation with respect to total outputpower. This can be addressed by using a coupled cavity configuration which is apromising route to combining high Purcell factors and large mode volumes [36, 37].Using different cavity designs, such as the H0-type [38], will also ensure single modeoperation across the entire gain bandwidth.We estimate that by maximizing the number of defects in the active volume, morethan a factor of 10 increase in luminescence is possible. In addition, a betterunderstanding of the nature of the defects may lead to a narrower emission bandwidth,potentially adding another order of magnitude. This will also lead to dramaticimprovements in the efficiency, as in the current device, due to the low number ofdefects, the bulk of the current passes through the device without recombining. Asuccessful combination of these effects thus makes our approach a very promisingroute towards an all-silicon electrically driven nano-laser.
6. Conclusions
In conclusion, we have enhanced luminescence from silicon by more than 4 ordersof magnitude by combination of hydrogen plasma treatment (up to 2 orders ofmagnitude), and both Purcell and extraction efficiency enhancement (300-400 times),and demonstrated an all-silicon nano-LED with the highest power spectral densityreported in silicon to date (see table 1). The Photonic Crystal cavity suppresses7thermal quenching of the hydrogen plasma induced optically active defects andimproves the emission rate, thereby making them a route to a practical light source.Our nano-LED operates at room temperature and in the technologically importantwavelength range of 1300-1600 nm. As our approach can provide narrow linewidthnano-LED anywhere in this range, it is ideally suited to Wavelength DivisionMultiplexing (WDM) like applications. Additionally, it provides a means to realisinghigh spectral purity light sources in a CMOS environment, key to making practicaldevices such as inexpensive biosensors.
Methods1. Design and fabrication of Photonic crystal nanocavity
L3 PhC nanocavities were realized by removing three holes from the -K direction ofa PhC pattern, characterized by a period of a = 420 nm and a normalized radius (r/a) =0.28. The two holes adjacent to the cavity were reduced in size and displaced laterallyto increase the Q-value of the cavity [39]. These parameters produce an L3 nanocavitywith a fundamental mode around 1.5 µm wavelength. To obtain the maximum out-coupling efficiency in the vertical direction, we applied a far-field optimizationtechnique whereby alternating holes around the cavity are enlarged to form a “secondorder grating”, which significantly increases the vertical out-coupling efficiency [16].Here, we used an enlargement of 15nm in hole radius to provide maximal vertical out-coupling efficiency. The far-field optimization reduces the Q-values of the cavity, but,similar to [22], the Q-values are, in fact, limited by free carrier absorption. A range ofPhC cavities were realized with period spanning from 370 to 450 nm in order todemonstrate that we are able to very precisely control the fundamental cavity mode inthe range 1300 - 1600 nm.8The fabrication of L3 PhC nanocavities was carried out by electron beamlithography and reactive ion etching. We used CHF /SF gas chemistry for etching thePhC holes. A free standing membrane was formed by removing the supporting siliconoxide layer using wet chemical etching by hydrofluoric acid.Importantly, for large scale integration, undercutting is not necessarily aprerequisite for high quality Photonic Crystals [40]. Oxide cladding may instead beemployed providing a very stable and robust device. Here, we use membraned devicesas they provide the greatest operating tolerances, which is useful for initialdemonstrations.
2. Ion implantation for pin junction
The n-type finger-like arm of the device was doped by ion implantation to alevel of 10 P/cm and it was separated from the corresponding p-type finger (dopedwith 10 B/cm ) by a distance of 500 nm. This separating region is slightly p-typedoped with ~10 B/cm (background doping of the as-bought SOI wafer). Alignmentbetween the steps is carefully carried out by means of electron beam lithography. Forthese conditions, the cavity region is fully depleted. Therefore, the intrinsic region isfilled with carriers during the operation that then recombine in an efficient way.Doped silicon is known to be a source of optical losses; however, doping has beenshown to be relatively insignificant in this system with Q-factors up to 40,000observed for carrier densities of 10 /cm [41].
3. Plasma treatment
The plasma treatment was performed in a parallel plate reactive ion etching systemwith a hydrogen and hydrogen/argon flow of 20 sccm and 24 sccm respectively, a9pressure of 1×10 -1 mbar, an RF power of 40 W, resulting in a DC bias of ≈ -400 V. This treatment step was carried out on the membraned PhCs after dopant and contactannealing was performed. Importantly, since the hydrogen plasma treatment is carriedout as a post process, it is compatible with silicon photonics device processing and isstable under operating temperature of silicon photonic devices. Furthermore, thedefects created by the treatment were found to be reasonably stable, with only amoderate decrease in the emission observed over 6 months.
4. Characterization
For characterization, we used room temperature confocal PL and the resonantscattering method to measure the PL and Q-values of the nanocavity modes,respectively. The cavities were excited with a CW diode laser emitting at 640nm. Thespot was focused to 1 µm in the centre of a cavity with a microscope objective (NA=0.8) and the emitted light was collected back with the same objective and fed to agrating spectrometer. Further details of these characterization methods are given in[14, 22].The emitted light from the electrically driven device was collected using the samePL setup. From Fig. 5, it is observed that the EL signal (also CW driving) has moreemission power compared to PL across the entire spectral range. This indicates thatthe injection of carriers is relatively efficient in this scheme. The PL signal is lowerthan EL for almost same input power due to the low absorption (~5%) of the thinsilicon slab for the 640nm excitation wavelength.In addition, we notice that the device heating is similar under both optical andelectrical pumping, and is estimated to be of the order of 1-2°C from the smallredshift (about 0.1 nm) of the cavity resonance as compared to very low power0excitation. This is consistent with previous literature works on silicon and GaAsmembrane PhC nanocavities under similar pumping conditions [42-43]. In fact,despite the very high power density in the cavity region under both electrical andoptical pumping (about 50 kW/cm ), a small temperature increase is observed due tothe very efficient heat dissipation occurring in membraned PhC nanocavities [43].Oxide cladding, see above, is expected to further improve heat dissipation. Authors Contributions
AS and LOF conceived the idea of plasma treatment of PhC cavities. AS developedthe plasma treatment process (under the guidance of TFK). AS, PC, SLP and LOFdesigned and fabricated the samples. RLS and MG carried out all the opticalmeasurements. PC and GF performed the ion implantation and SB performed theTEM analysis, guided by FP. DG and KW designed and modelled the PhC cavities.All authors contributed to discussions and analysis of data. AS, PC, MG and LOFwrote the manuscript, with contributions from the other co-authors. TFK, MG, LOFand FP coordinated and directed the work.
Acknowledgements
We acknowledge Lucio C. Andreani (Dept. of Physics, Univ. of Pavia) for usefuldiscussions and suggestions, and A. Liscidini (Dept. of Electronics, Univ. of Pavia)for providing wire bonding to the EL devices. This work was supported by Era-NETNanoSci LECSIN project coordinated by F. Priolo, by the Italian Ministry ofUniversity and Research, FIRB contract no. RBAP06L4S5 and the UK EPSRC UKSilicon Photonics project. The fabrication was carried out in the framework ofNanoPiX (see ).1
References [1] K. Tanabe, K. Watanabe and Y. Arakawa, Sci. Rep. , 349DOI:10.1038/srep00349 (2012).[2] A. Marconi, A. Anopchenko, M. Wang, G. Pucker, P. Bellutti and L. Pavesi, Appl.Phys. Lett , 221110 (2009).[3] L. Pavesi, L. Dal Negro, C. Mazzoleni, G. Franzò and F. Priolo, Nature , 440-444 (2000).[4] S. Saito, Y. Suwa, H. Arimoto, N. Sakuma, D. Hisamoto, H. Uchiyama, J.Yamamoto, T. Sakamizu, T. Mine, S. Kimura, T. Sugawara and M. Aoki, Appl. Phys.Lett , 241101 (2009).[5] Y. Gong, M. Makarova, S. Yerci, R. Li, M. J. Stevens, B. Baek, S.W. Nam, L. DalNegro and J. Vuckovic, Opt. Express , 13863-13873 (2010).[6] J. Liu, X. Sun, R. Camacho-Aguilera, L.C Kimerling and J. Michel, Opt Lett ,679-681 (2010).[7] X. Sun, J. Liu, L.C. Kimerling and J. Michel, Opt Lett , 1198-1200 (2009).[8] H. Rong, R. Jones, A. Liu, O. Cohen, D. Hak and A. Fang, Nature , 725-728(2005).[9] A. Henry, B. Monemar, J. Lindstrom, T.D. Bestwick, G.S. Oehrlein, J. Appl. Phys , 5597-5603 (1991).[10] J. Weber, Physica B: Condensed Matter , 201-217 (1991).[11] L.T. Canham, M.R. Dyball, W.Y. Leong, M.R. Houlton, A.G Cullis and P.Smith, Mat Sci and Eng: B , 41-45 (1989).[12] Wai Lek Ng, M. A. Lourenço, R.M. Gwilliam, S. Ledain, G. Shao and K.PHomewood, Nature 410, 192-194 (2001).2[13] S.G Cloutier, P.A Kossyrev and J. Xu Nat Materials , 887-891 (2005).[14] M. Galli, S.L. Portalupi, M. Belotti, L.C. Andreani, L. O’Faolain and T.F.Krauss, App. Phys. Lett , 071101 (2009).[15] N.-V.-Q. Tran, S. Combrié and A. De Rossi, Phys. Rev B , 1-4 (2009).[16] S.L Portalupi, M. Galli, C. Reardon, T.F. Krauss, L. O’Faolain, L.C Andreaniand D. Gerace, Opt. Express , 16064-16073 (2010).[17] M.O. Watanabe, M. Taguchi, K. Kanzaki, Y. Zohta, Jap. Journal of Appl. Phys , 281-286 (1983).[18] M. Hwang, D.K. Schroder, W.J. Biter, J. Appl. Phys , 4166-4169 (1987).[20] C. Ghica, L.C. Nistor, S. Vizireanu and G.J Dinescu, Phys. D: Appl. Phys ,033506 (2009).[22] R. Lo Savio, S.L. Portalupi, D. Gerace, A. Shakoor, T.F. Krauss, L. O’Faolain,L.C. Andreani, M. Galli, Appl. Phys. Lett , 201106 (2011).[23] A. Shakoor, R. Lo Savio, S. L. Portalupi, D. Gerace, L.C. Andreani, M. Galli,T.F. Krauss, L. O'Faolain, Physica B , 4027-4031 (2012)[24] N. Hauke, T. Zabel, K. Müller, M. Kaniber, A. Laucht, D. Bougeard, G.Abstreiter, J.J Finley and Y. Arakawa, New. J. Physics , 053005 (2010).[25] V. Kveder, M. Badylevich, E. Steinman, A. Izotov, M. Seibt and W. Schröter,Appl. Phys. Lett , 2106 (2004).[26] M. Settle, M. Salib, A. Michaeli, and T.F. Krauss, Opt. Express , 2440-2445( ).3[27] T. Tanabe, K. Nishiguchi, E. Kuramochi and M. Notomi Opt. Express , 22505-22513 (2009).[28] H.-G. Park, S.-H. Kim, S.-H. Kwon, Y.-G Ju, J.-K. Yang, J.-H. Baek, S.-B. Kimand Y.-H. Lee, Science , 1444-1447 (2004).[29] B. Ellis, M.A. Mayer, G. Shambat, T. Sarmiento, J. Harris, E.E. Haller and J. Vučković, Nat Photonics , 297-300 (2011).[30] G. Shambat, B. Ellis, A. Majumdar, J. Petykiewicz, M.A. Mayer, T. Sarmiento, J. Harris, E.E. Haller and J. Vučković, J. Nat. commun , 539, DOI:10.1038/ncomms1543 (2011).[31] M. A. Green, J. Zhao, A. Wang, P.J. Reece, M. Gal, Nature , 805-808 (2001)[32] J. Bao, M. Tabbal, T. Kim, S. Charnvanichborikarn, J. S. Williams, M. J. Aziz,F. Capasso, Optics Express , 6727-6733 (2007)[33] S. Yerci, R. Li, L. Dal Negro, Appl. Phys. Lett , 081109 (2010)[34] M. Fujita, B. Gelloz, N. Koshida, S. Noda, Appl. Phys. Lett , 121111 (2010)[35] A.P Knights, J.D.B Bradley, S.H. Gou and P.E.J Jessop, Vac. Sci. Technol. A, , 783-786 (2006). [36] H. Altug and J. Vucković, Opt. Express , 8819-8828 (2005). [37] H. Altug, D. Englund and J. Vučković, Nat Physics , 484-488 (2006).[38] H.-S. Ee, K.-Y. Jeong, M.-K. Seo, Y.-H. Lee and H.-G Park, Appl. Phys. Lett ,011104 (2008).[39] Y. Akahane, T. Asano, B.-S Song and S. Noda, Nature , 944-947 (2003). [40] S.-W. Jeon, J.-K. Han, B.-S Song and S. Noda, Opt. Express , 19361-19366(2010).[41] P. Cardile, G. Franzò, R. Lo Savio, M. Galli, T.F. Krauss, F. Priolo and L. O’Faolain, Appl. Phys. Lett , 203506 (2011).4[42] M. E. Kurdi, X. Checoury, S. David, T.P. Ngo, N. Zerounian, O. Kermarrec, Y.Campidelli and D. Bensahel, Opt Express 16, 8780-8791 (2008)[43] F. Intonti, N. Caselli, S. Vignolini, F. Riboli, S. Kumar, A. Rastelli, O.G.Schmidt, M. Francardi, A. Gerardino, L. Balet, L.H. Li, A. Fiore and M. Gurioli,Appl. Phys. Lett100