Thermal properties of InP-based quantum well photonic crystal nanobeam lasers heterogeneously integrated on silicon on insulator waveguides are studied. We show both numerically and experimentally the reduction of the thermal resistance of the III-V cavities by adjusting the composition of the layer which bonds the III-V materials to the silicon wafer and by adding an over-cladding on top of the cavities. Using a bonding layer made of benzocyclobutene and SiO2 and an over-cladding of MgF2, we found a decrease by a factor higher than 35 compared to air-suspended photonic crystal nanobeam cavities. Such optimized structures are demonstrated to operate under continuous wave pumping for several 10's of minutes despite the adverse effect of non-radiative surface recombination of carriers.
© 2014 Optical Society of America
Planar photonic crystals (PhCs) allow near-perfect control of the electromagnetic field confinement and propagation by associating total internal reflection with a photonic band gap. In particular, they enable the realization of possibly the best resonators at optical wavelengths - with record quality factor (Q) over volume ratios (V) . When combined with active materials, these resonators give nanolasers with ultra-small footprints  and low thresholds . The rapid expansion of research in this field has provided some outstanding demonstrations [4–6], but it is only in the last few years that the difficulties associated with PhCs have begun to be taken into account, particularly related to thermal management. Indeed, the excellent confinement of the electromagnetic field in these structures, which is based on alternating materials with very different refractive indices, usually goes hand in hand with poor thermal evacuation. It is, of course, at its best in PhCs fabricated in air-clad membranes but, unfortunately, air is also well known to be a remarkable thermal insulator. Even though this can be exploited for achieving efficient thermo-optical effects , it represents a major hurdle for obtaining stable CW lasers or switches working at high bit rate. Thermal management is hence one of the crucial issues to be tackled in nanophotonics to envisage the exploitation of PhC active nano-components for energy efficient applications such as optical interconnects.
To overcome this hurdle, several groups have explored the situation through two main channels, one based on the selective growth of materials with better thermal conductivity3, the other on the use of wafer bonding to heat-sinking substrates [8–10].
In this article, we investigate both numerically and experimentally, the thermal properties of InP-based nanobeam PhC lasers heterogeneously integrated to silicon-on-insulator (SOI) waveguides  with the view to managing the heat evacuation issues. As previously demonstrated, these structures offer a promising solution for future photonic circuits since they combine CMOS compatibility with power efficient ultra-compact nanolasers. Particular attention is devoted to exploring the impact on the heat sinking, of the composition of the layer which bonds the III-V nanocavity to the SOI circuitry as well as to that of the deposition of an over-cladding material.
In Section 2, we describe the finite element simulations performed on the different configurations of hybrid nanolasers which result in the determination of their thermal resistance (Rth). In Section 3, we detail the fabrication of the structures and the experimental characterization of their effective thermal resistance to be compared with the simulations. Finally in section 4 we discuss the CW operation of the nanolasers which were fabricated in consequence.
2. Numerical determination of the hybrid III-V-SOI nanolaser thermal resistance
The hybrid III-V/SOI PhC nanolasers structure under investigation is schematically represented on Fig. 1. It consists of a 2-optical-level architecture, one formed by passive low-loss SOI wire waveguides (220nm thick and 500nm wide) and the other by InP/InGaAsP PhC nanobeam cavities (~20µm long). The nanobeam cavities are made of a 285nm thick and 505nm wide InP based wire waveguide drilled with a single row of equally sized holes (radius 120nm). The InP-based stack embeds 4 InGaAsP strained quantum wells whose emission peaks at 1530nm. For simplicity, in the thermal simulations, the cavity is modelled as fully made of InGaAsP material which represents actually more than 2/3 of the InP-based stack composition. The distance between the hole is subtly varied to obtain a high Q cavity (Q>106 in simulations) even in the case where the cavity is over-clad. More details on the design of the cavities can be found in . The 2 levels are separated by a thin bonding layer of low refractive index to ensure good optical confinement and efficient evanescent wave coupling .
The numerical study carried out aims at determining the impact of the composition of the bonding layer and that of the addition of an over-cladding material on the heat sinking of the nanolasers. 3D finite element method (FEM) is used to determine the temperature distribution T(x,y,z) as a function of the thermal power injected in the cavity. In all calculations, the simulation window is 14μm high, 40μm long and 20µm wide. The cavity is placed at 1μm from the top of the simulated region as it was found, a posteriori, that most of the heat evacuates through the substrate and not through the top cladding above the cavity. We take into account the presence of the SOI waveguide and, more importantly, the 2μm-thick oxide layer which acts as a major thermal insulator in the structure. The SOI waveguide is surrounded by a 300nm layer of Benzocyclobutene (BCB) which is the polymer used to perform the wafer bonding. The total thickness of the bonding layer (i.e. the distance between the BOx and the III-V surfaces) is kept constant at 750 nm which is a typical value in our hybrid structures. A 5µm long heat source is placed in the centre of the PhC cavity to emulate the thermal power Ptherm produced by optical pumping using a laser focused on the sample (5µm spot size). Numerical thermal boundary conditions are set as follows:
- • On the lateral edges they are taken as insulating. Indeed, several preliminary tests have shown that a large fraction of the heat flow passes through the Si substrate due to the low thermal conductivities of the other surrounding materials composing the hybrid structure.
- • At the bottom of the simulation region (i.e. in the silicon substrate), they are set at a fixed temperature (room-temperature).
- • The top of the simulation region (i.e. at the top of the over cladding layer) is taken as insulating (tests with a convective air model gave similar results).
The refractive indices as well as the thermal constants of the different materials considered for the simulations are given in Table 1.
2.1 Impact of the bonding layer composition
We begin the analysis by comparing hybrid structures with no over-cladding layer (i.e. in air). The bonding layer is composed of two parts, a 450nm upper part, chosen to be either air, BCB or SiO2, and its lower part being 300nm of BCB in all cases (imposed by the wafer bonding technology). The choice of not using materials with a higher thermal conductivities arises from the necessity to maintain the refractive index of this layer as low as possible (close to 1.5) in order to obtain high Q factors. When it is chosen to be air, we are in the case of a suspended PhC nanobeam laser which is the most studied configuration in the literature [13–15]. Nanobeam cavities are surely the most compact PhC structures and seem very promising for their interfacing with regular wire waveguides  and for some specific applications such as optomechanics . However, from the point of view of thermal management, the use of PhC wire is not ideal as against a “regular” 2D PhC membrane. Poor heat evacuation is expected, mainly due to the presence of 4 thermal insulating “walls” around the cavity. In , it is shown that the thermal resistance (Rth = dT/dPtherm) of nanobeam cavities made of Si (5.5x105K.W −1) is more than 20 times greater than that of 2D PhC cavities made of the same material. In our system of materials, simulations give a thermal resistance for the suspended nanobeam cavity to be 7.2x105K.W−1, which is of the same order of magnitude. We plot, on Fig. 2, for the different configurations studied in this work, the temperature elevation (ΔT) obtained at the centre of the cavity as a function of the injected thermal power. The replacement of air as under-cladding material by BCB or SiO2 leads to a substantial decrease of Rth (2.4 times less for BCB and 6.5 times less for SiO2). As indicated in , this decrease is enabled by favouring the heat flow towards the substrate when the PhCs are bonded.
2.2 Impact of an over-cladding deposition
In order to increase the heat sinking, one way is to replace the air surrounding the upper and the side flanks of the cavity by depositing an over-cladding material with a higher thermal conductivity. So, we perform the simulations by adding a 1µm thick over-cladding layer composed of SiO2 or MgF2, on hybrid structures with a BCB/SiO2 bonding layer (lowest Rth). The results are also reported on Fig. 2.
With respect to a structure with an air over-cladding, Rth is reduced by a factor of 1.8 when the structure is encapsulated in SiO2 and by a factor 5.6 when it is with MgF2. We represent on Fig. 3, the spatial distribution T(x,y) of temperature in the median plane of the PhC nanobeam for hybrid structures with a BCB + SiO2 bonding layer and the different over-cladding materials (air, SiO2 and MgF2). Ptherm is here fixed to 100µW.
As the thermal conductivity of the over-cladding material is increased, the heat sinks more and more on the sides and above the nanobeam cavity which gives in turn, a lower temperature rise. Thus, MgF2 seems to be a very interesting material for our goal as it provides both low refractive index (n = 1.38) and large thermal conductivity relatively to SiO2 (k = 14 W/(K.m)). However, from the practical point of view, its use may be restricted as it is a rugged material, resistant to chemical etching. As an example, it is not possible to use it in the bonding layer as chemical etching is necessary during the processing of the structures.
In what follows, we study experimentally the thermal properties of hybrid nanolasers fabricated in the different configurations numerically investigated.
3. Fabrication and measurement of the effective thermal resistance of the hybrid nanolasers
The fabrication of our hybrid nanolasers is based on the adhesive bonding of the InP-based active heterostructure onto a SOI wafer processed with wire waveguides, using the planarising polymer BCB . The BCB is diluted with Mesithylene in order to obtain a 300nm or 750nm thick layer on the SOI wafer after spin coating at 5000rpm. For the samples with the bonding layer composed of BCB (300nm) and SiO2 (450nm), the InP wafer is sputtered with SiO2 prior to the bonding. After the substrate removal, the nanobeam cavities are patterned into the InP layer using inductively coupled plasma etching  through a Hydrogen Silsesquioxane (Fox 15) mask defined by electron beam lithography. A scanning electron microscope image of etched nanolasers aligned  to the SOI wires is shown on Fig. 4.
Then, when needed, the samples are encapsulated in a 1µm thick layer of SiO2 (sputtering) or MgF2 (thermal evaporation in a ion beam assisted deposition chamber).
The principle of the experiment is to optically probe the temperature elevation in the PhC nanolasers by monitoring the emission wavelength (λ(T)) above the laser threshold as a function of the pump power (P), the goal of this experiment being, of course, to retrieve the apparent thermal resistance. The direction and the amplitude of a wavelength shift observed in these samples can be directly attributed to an increase or a decrease of the refractive index. In a semiconductor quantum well laser, the emission frequency, ω(N, T), is given by Eq. (1) 
To perform the measurements, we use a micro-photoluminescence set-up where both the pump power and the sample temperature are controlled. The nanolasers are surface pumped by a laser diode at 1.18µm which is modulated to obtain 40ns long optical pulses with different duty-cycles to modify the amount of heat released in the structure. The temperature is controlled by a Peltier cell and a platinum temperature sensor is used to read-out the temperature of the holder. The emitted light is collected with a single mode fibre positioned on top of a grating coupler at one of the extremities of the SOI wire waveguide coupled to the nanolaser.
For each sample configuration, we proceed to the measurement of by first calibrating the emission wavelength as a function of the temperature of the sample by setting the experimental conditions so that the optical pumping induces negligible heating in the sample. To do so, the duty cycle of the pump laser diode is set at a low value of 1% such that the repetition time of the pulses (4µs) is longer than the usual thermal diffusion times observed in III-V PhC ~1µs . The results of an example of such calibration are shown on Fig. 5. By varying the current in the Peltier cell, we find that λ(T) varies linearly with T and ∂λ/∂T is close to 0.1nm/K for all the different samples configurations investigated.
This done, we measure the variation of the emission wavelength as a function of the pump power in a regime where the optical pumping induces the heating of the cavity. The pump laser duty cycle is increased to a higher value of 20% (200ns repetition time). The temperature of the sample is set in the range where the wavelength shift with temperature was previously calibrated. The emission wavelength is successively recorded as a function of the pump peak power in the cases where the pump is set in the low then in the high duty cycle regimes. The results for the hybrid structures with BCB/SiO2 bonding layer and MgF2 over-cladding are represented on Fig. 6.
Here, the threshold power is measured to be around 4mW and 5mW when the pump is set in the low and high duty cycle regimes respectively. We observe that the emission wavelength varies almost linearly below and above threshold but with a much smaller slope above threshold due to the quasi clamping of the carrier density. The two slopes above the threshold (∂λ/∂P)L and (∂λ/∂P)H are respectively the slopes in the low and high duty-cycle regimes. We obtain for this particular case, an increase of the slope for the high duty cycle regime compared to the low one from (∂λ/∂P)L = - 0.0031nm/mW to (∂λ/∂P)H = - 0.0018nm/mW. The apparent thermal resistance can now be easily determined by using the calibration measurement by calculating Eq. (2):Table 2.
The thermal resistance measured through this method is an apparent thermal resistance as P is the total optical pump power impinging onto the sample and not the thermal power that is generated in the cavity due to the pump absorption. Taking this into account, in order to compare these numbers with the FEM simulations, instead of trying to estimate the real thermal power generated in the material, we compare the relative increase or decrease of the apparent thermal resistivities obtained for the different types of hybrid structures.
First, by replacing the bonding layer fully made of BCB by a BCB + SiO2 bi-layer, is found to decrease from 1.55 K/mW to 0.56 K/mW. Thus, the ratio of the thermal resistivities (BCB/air over BCB + SiO2/air) is experimentally 1.55/0.56 ≈2.8. This is in total agreement with our numerical simulation results where the ratio 304/110 gives also ≈2.8. Then by adding an over-cladding layer on top of the structures where BCB + SiO2 is used as the bonding layer, we measure that is further reduced to 0.35 K/mW and 0.15 K/mW when SiO2 and MgF2 are employed respectively. This gives ratios of when the BCB + SiO2/air configuration is taken as reference, of 1.54 for the SiO2 over-cladding and of 3.6 for the MgF2 over-cladding. This is again in good agreement with what is obtained with the FEM simulations (ratios of 1.8 and 5.6). However, slight discrepancies are visible especially when MgF2 is used as over-cladding. We attribute this to two uncertainties concerning our MgF2 material, one due to the PhC hole filling during the deposition and the other due to its porosity, which can affect significantly the thermal conductivity.
4. Discussion on CW operation and conclusion
The outstanding role of the heat sinking of the different stacks studied in this work triggered us to test CW laser operation by pumping at 1180 nm, a hybrid PhC active nanobeam cavity bonded with BCB + SiO2 and encapsulated in 1 μm-thick MgF2 layer. Below the laser threshold, we notice that the emission intensity increases when the pump is increased, and does not diminish quickly as it is the case for non-encapsulated structures. By increasing the pump power further, we observed all the characteristic variations which usually accompany the laser emission (nonlinear increase of the emission power, decrease of the linewidth) as it is illustrated in Fig. 7.
Despite the proof of the improved thermal sinking, the behaviour of these hybrid PhC nanobeam lasers has not proven to be stable with time. Indeed, we observed that the emission slowly but surely decreases after a few tens of minutes. Figure 7 shows the modification of the emission curves after letting the cavity operate in CW at twice the threshold for one hour. Even though the laser still works, its threshold is multiplied by more than a factor 3 and keeps on increasing if we let the CW pump switched-on. Eventually, we observe that the emission almost completely vanishes by pumping for an even longer time. Note that this is due to a damaging of the sample as it is not reversible. The mechanism of damaging has not been investigated in this work, but we suspect the apparition of localised hot-spots in the QWs due to the enhanced non radiative surface recombination at the etched sidewalls of the PhC nanobeam cavities . Chemical surface passivation of the QWs can be used to reduce non radiative recombination of carriers and enable stable CW operation .
In conclusion, we have shown that substantial increase of the heat sinking in hybrid InP on SOI PhC nanobeam lasers can be obtained by partly replacing the BCB content of the bonding layer by SiO2, and by encapsulating them with SiO2 or MgF2. The exhaustive experimental study of the contribution of the different materials showed that the apparent thermal resistance decreases from 1.55 K/mW to 0.56 K/mW by replacing BCB bonding layer by a BCB + SiO2 bi-layer. A further reduction to 0.35 K/mW is obtained using SiO2 over cladding to attain a minimum of 0.15 K/mW when MgF2 over cladding is employed. This is again in good agreement with the results of the FEM simulations. All in all, in the best configuration, the thermal resistance in the hybrid structure is 35 times less than that of air-clad structure. Moreover, the association of wafer bonding and encapsulation renders these structures mechanically stable and imparts isolation from environmental perturbations which is essential for reliable devices. This a crucial progress not only allowed us to demonstrate CW operation of nanolasers but also error-free high wavelength conversion at 10Gbits/s .
We acknowledge for funding the FP7 European Projects COPERNICUS (249012) and PHOXTROT (318240) as well as the French ANR jeunes chercheurs PROWOC.
References and links
2. Y. Halioua, A. Bazin, P. Monnier, T. J. Karle, I. Sagnes, G. Roelkens, D. Van Thourhout, F. Raineri, and R. Raj, “III-V photonic crystal wire cavity laser on silicon wafer,” J. Opt. Soc. Am. B 27(10), 2146–2150 (2010). [CrossRef]
3. S. Matsuo, A. Shinya, T. Kakitsuka, K. Nozaki, T. Segawa, T. Sato, Y. Kawaguchi, and M. Notomi, “High speed ultracompact buried heterostructure photonic crystal laser with 13fJ of energy consumed per bit transmitted,” Nat. Photonics 4(9), 648–654 (2010). [CrossRef]
4. K. Nozaki, A. Shinya, S. Matsuo, Y. Suzaki, T. Segawa, T. Sato, Y. Kawaguchi, R. Takahashi, and M. Notomi, “Ultralow-power all-optical RAM based on nanocavities,” Nat. Photonics 6(4), 248–252 (2012). [CrossRef]
5. A. Bazin, K. Lengle, M. Gay, P. Monnier, L. Bramerie, R. Braive, G. Beaudoin, I. Sagnes, R. Raj, and F. Raineri, “Ultrafast all-optical switching and error-free 10Gbit/s wavelength conversion in hybrid InP-silicon on insulator nanocavities using surface quantum wells,” Appl. Phys. Lett. 104(1), 011102 (2014). [CrossRef]
7. H. M. H. Chong and R. M. De La Rue, “Tuning of photonic crystal waveguide microcavity by thermo-optic effect,” IEEE Photon. Technol. Lett. 16(6), 1528–1530 (2004). [CrossRef]
8. G. Vecchi, F. Raineri, I. Sagnes, A. Yacomotti, P. Monnier, T. J. Karle, K.-H. Lee, R. Braive, L. Le Gratiet, S. Guilet, G. Beaudoin, A. Taneau, S. Bouchoule, A. Levenson, and R. Raj, “Continuous-wave operation of photonic band-edge laser near 1.55 µm on silicon wafer,” Opt. Express 15(12), 7551–7556 (2007). [CrossRef] [PubMed]
9. J. Hwang, H. Ryu, D. Song, I. Han, H. Park, D. Jang, and Y. Lee, “Continuous room-temperature operation of optically pumped two-dimensional photonic crystal lasers at 1.6 µm,” IEEE Photon. Technol. Lett. 12(10), 1295–1297 (2000). [CrossRef]
10. J. R. Cao, Wan Kuang, Zhi-Jian Wei, Sang-Jun Choi, M. Haixia Yu, J. D. Bagheri, O’Brien, and P. D. Dapkus, “Sapphire-bonded photonic crystal microcavity lasers and their far-field radiation patterns,” IEEE Photon. Technol. Lett. 17(1), 4–6 (2005). [CrossRef]
11. Y. Halioua, A. Bazin, P. Monnier, T. J. Karle, G. Roelkens, I. Sagnes, R. Raj, and F. Raineri, “Hybrid III-V semiconductor/silicon nanolaser,” Opt. Express 19(10), 9221–9231 (2011). [CrossRef] [PubMed]
12. A. Bazin, R. Raj, and F. Raineri, “Design of silica encapsulated high-Q photonic crystal nanobeam,” J. Lightwave Technol. 32(5), 952–958 (2014). [CrossRef]
13. Y. Zhang, M. Khan, Y. Huang, J.-H. Ryou, P. Deotare, R. Dupuis, and M. Lončar, “Photonic crystal nanobeam lasers,” Appl. Phys. Lett. 97(5), 051104 (2010). [CrossRef]
16. Q. Quan, P. B. Deotare, and M. Loncar, “Photonic crystal nanobeam cavity strongly coupled to the feeding waveguide,” Appl. Phys. Lett. 96(20), 203102 (2010). [CrossRef]
18. L. D. Haret, T. Tanabe, E. Kuramochi, and M. Notomi, “Extremely low power optical bistability in silicon demonstrated using 1D photonic crystal nanocavity,” Opt. Express 17(23), 21108–21117 (2009). [CrossRef] [PubMed]
19. V. Moreau, G. Tessier, F. Raineri, M. Brunstein, A. Yacomotti, R. Raj, I. Sagnes, A. Levenson, and Y. De Wilde, “Transient thermoreflectance imaging of active photonic crystals,” Appl. Phys. Lett. 96(9), 091103 (2010). [CrossRef]
20. G. Roelkens, J. Brouckaert, D. Van Thourhout, R. Baets, R. Nötzel, and M. Smit, “Adhesive bonding of InP/InGaAsP dies to processed silicon-on-insulator wafers using dvs-bis-benzocyclobutene,” J. Electrochem. Soc. 153(12), G1015–G1019 (2006). [CrossRef]
21. S. Guilet, S. Bouchoule, C. Jany, C. S. Corr, and P. Chabert, “Optimization of a Cl2-H2 inductively coupled plasma etching process adapted to nonthermalized InP wafers for the realization of deep ridge heterostructures,” J. Vac. Sci. Technol. B 24(5), 2381–2387 (2006). [CrossRef]
22. T. J. Karle, Y. Halioua, F. Raineri, P. Monnier, R. Braive, L. Le Gratiet, G. Beaudoin, I. Sagnes, G. Roelkens, F. van Laere, D. Van Thourhout, and R. Raj, “Heterogeneous integration and precise alignment of InP-based photonic crystal lasers to complementary metal-oxide semiconductor fabricated silicon-on-insulator wire waveguides,” J. Appl. Phys. 107(6), 063103 (2010). [CrossRef]
23. L. A. Coldren, S. W. Corzine, and M. L. Masanovic, Diode Lasers and Photonic Integrated Circuit (John Wiley, 2012).
24. F. Raineri, C. Cojocaru, P. Monnier, A. Levenson, R. Raj, C. Seassal, X. Letartre, and P. Viktorovitch, “Ultrafast dynamics of the third-order nonlinear response in a two-dimensional InP-based photonic crystal,” Appl. Phys. Lett. 85(11), 1880 (2004). [CrossRef]
25. H. Oigawa, J.-F. Fan, Y. Nannichi, H. Sugahara, and M. Oshima, “Universal passivation effect of (NH4)2Sx treatment on the surface of III-V compound semiconductors,” J. Appl. Phys. 30(3A), L322–L325 (1991). [CrossRef]