Open Access
Add to BibSonomyAbstract
We have demonstrated an ultracompact buried heterostructure photonic crystal (PhC) laser, consisting of an InGaAsP-based active region (5.0 x 0.3 x 0.15 μm3) buried in an InP layer. By employing a buried heterostructure with an InP layer, we can greatly improve thermal resistance and carrier confinement. We therefore achieved a low threshold input power of 6.8 μW and a maximum output power in the output waveguide of −10.3 dBm by optical pumping. The output light is effectively coupled to the output waveguide with a high external differential quantum efficiency of 53%. We observed a clear eye opening for a 20-Gbit/s NRZ signal modulation with an absorbed input power of 175.2 μW, resulting in an energy cost of 8.76 fJ/bit. This is the smallest reported energy cost for any type of semiconductor laser.
©2011 Optical Society of America
1. Introduction
A photonic network on a silicon CMOS chip is a desired breakthrough technology for overcoming bandwidth and power consumption limits [1,2]. This is because over 50% of the total power consumption is used for the interconnects, with this fraction increasing with data rate [3]. Furthermore, multi-core processors based on network-on-chip technologies would be beneficial for increasing total throughput without increasing the power consumption [4,5]. Therefore, the introduction of a photonic network into a silicon CMOS chip is believed to be a potential solution, based on the realization of a high-bandwidth photonic data transport network from long-distance networks to board-to-board interconnections with low power consumption. If we are to construct a photonic network on a CMOS chip, the density requirement is a critical issue. In this context, directly modulated lasers with ultra-low energy costs are needed for constructing a photonic network on a silicon CMOS chip. The International Technology Roadmap for Semiconductors (ITRS) predicted that the on-chip clock rate will increase to 14.3 GHz by 2022 [6] and, in ref. 7, the required energy cost for the data transport of a single bit has been estimated at less than 10 fJ/bit. For example, when an on-chip transmitter is modulated at 20 Gbit/s, the input power should be less than 200 μW.
This extremely small energy requirement is difficult to achieve with previously developed laser structures. Recently, 25-Gbit/s directly modulated edge-emitting lasers have been developed for 100-Gbit Ethernet applications, in which the device consumed about 100 mW, corresponding to 4 pJ/bit [8,9]. Vertical-cavity surface emitting lasers (VCSELs) have a considerably lower threshold power, but the figure of merit is still around 286 fJ/bit at 35 Gbit/s [10]. Therefore, the reduction of the active volume is a critical issue if we are to achieve high-speed modulation with an extremely low energy cost. As a result, lasers based on photonic crystal (PhC) technologies have attracted much attention because the PhC cavity provides a small mode volume and a high Q-factor. Furthermore, PhC technologies enable us to obtain various functional devices with extremely small energy [11,12] that are also key devices for constructing photonic network on chip. This is because the photonic network based on wavelength division multiplexing (WDM) devices such as optical filters and switches is expected to reduce the total power consumption of CMOS [2]. Therefore, we propose to construct the photonic network chip based on photonic crystal technologies. However, it is difficult to achieve room temperature CW operation with a PhC nanocavity laser because the PhC employs an air-bridge structure, which increases the thermal resistivity of the device. There have been a few reports on CW operation using quantum dots [13] and an H0 cavity with a single quantum well (SQW) [14]. However, the output power was very limited. Furthermore, it is difficult to fabricate an input/output waveguide in a PhC slab when we use PhC lasers with an all-active structure.
To overcome these problems, we have already developed a buried heterostructure (BH) PhC nanocavity laser, which greatly reduces thermal resistivity and effectively confines both the carrier and photon in the cavity [15]. In this device, there is an InGaAsP-based active region inside the line-defect of the InP-PhC. We achieved a low threshold input power of 1.5 μW in an input waveguide, and high-speed modulation of 5.5 GHz with a small signal modulation. However, certain problems remain, namely the output waveguide is not integrated in the PhC slab and the maximum output power of 4 μW was still small in terms of detecting optical signals of more than 10 Gbit/s.
In this paper, we report the design and demonstration of the integration of an output waveguide with a PhC laser. The output waveguide is placed in the Γ-M direction of the PhC cavity, enabling us to achieve effective coupling between the cavity and the output waveguide. Furthermore, to increase the output power, we increased the number of quantum wells to 3, whereas the devices reported in ref. 10 have only a single quantum well. We achieved a maximum output power of 93.3 μW in the output line-defect waveguide and an external differential quantum efficiency of 53%. In terms of transmitting several tens of Gbit/s optical signals in photonic network chip, these values are quite essential. Furthermore, the dynamic modulation realized by optical pumping exhibits clear eye opening with a 20-Gbit/s NRZ signal, which gives an estimated energy consumption of 8.76 fJ/bit. This is the smallest energy yet reported for transmitting a single bit in any type of laser.
2. Device structure and fabrication
Figure 1 shows a schematic diagram of our proposed photonic network chip, which is hybrid integrated with multi-core processor. The photonic network chip consists of photonic switches and waveguides connecting between switches. Each photonic switch based on photonic crystal technologies consists of micro-lasers, photodetectors, switches and filters. Routers in multi-core processor control the photonic switches in the photonic network chip. In this device, there is no optical connection between electric and photonic planes. Therefore, proposed device can be fabricated by using micro-bump techniques [16].
Fig. 1 Illustration of photonic network chip integrated with multi-core processor. Each photonic switch based on photonic crystal technologies consists of micro-lasers, photodetectors, switches and filters.
Figure 2(a) shows the schematic diagram of BH-PhC nanocavity laser and Fig. 2(b) shows cross-sectional view of active region. We placed an extremely small buried active region in a straight-line defect waveguide in an InP-PhC slab. This is because a cavity based on a line-defect waveguide is expected to provide an extraordinarily high-Q (Q>1 million) nanocavity with a mode volume (V eff) of ~(λ/n)3 by implementing the slight local structural modulation of a line defect waveguide [17]. With a BH cavity, similar cavity modes can be created by introducing refractive index modulation, instead of structural modulation [18]. The straight-line defect waveguide also acts as an input waveguide because the buried InP layer is transparent for the pumping wavelength. Figure 2(c) shows a scanning electron micrograph (SEM) image of our fabricated BH-PhC laser with an air-bridge structure. The input waveguide consists of a 3-μm wide waveguide, a taper waveguide and the line defect of the PhC. Therefore, the pump light is input into the active region through a line defect waveguide and is effectively absorbed in the active region. On the other hand, the output waveguide is placed in an offset position with respect to the line defect including the active region, as shown in Fig. 2(c).
Fig. 2 (a) Schematic diagram of buried heterostructure photonic crystal nanocavity laser. An extremely small active region is buried in a straight-line defect waveguide in an InP-PhC slab. (b) Cross-sectional view of active region. (c) Scanning electron micrograph image of the fabricated BH-PhC laser with an air-bridge structure.
Figure 3 shows mode profiles calculated by the finite-difference time-domain (FDTD) method, where the active region is inside the line-defect waveguide. Since the refractive index of the InGaAsP-based active region (n eq = 2.77) is larger than that of the surrounding InP layer (n eq = 2.59), we can confine the light without shifting the position of the air holes as described above. Figure 3(a, b) is the calculated device structures without and with output waveguide, respectively. The field profile of the cavity mode is shown in Fig. 3(c) for the device without output waveguide. FDTD calculations show that it exhibits a high-Q (Q~1.1 x 106) cavity mode with Veff~0.25 μm3. As shown this, the mode field is extended to the Γ-M direction when the cavity has no output waveguide. Thus, effective coupling can be realized with the output waveguide when the output waveguide is placed in the Γ-M direction of the PhC cavity. Figure 3(d, e) shows the calculated mode profiles of the devices with output waveguide for offset 3a and 4a, respectively, where a is lattice constant. As shown in these figures, the coupling between cavity and output waveguide is increased with decreasing the offset. On the other hand, the cavity Q-factor was decreased from ~15,000 to ~2,900 with decreasing the offset. The calculated mode volume was ~0.26 μm3 for the device with offset 4a. The calculated mode volume was slightly increased to ~0.3 μm3 for the device with offset 3a because we calculated the mode volume including the output waveguide. However, compared with Fig. (c-e), the mode profiles of the active region are almost same. Thus, we seem that the mode volume of the cavity with output waveguide is the same for the cavity without output waveguide. When the offset was 2a, the cavity Q-factor was decreased to ~800 although the coupling to the output waveguide was increased. Thus, we choose the offsets of fabricated device for 3a and 4a. Furthermore, there is only one output port in this device and this is an important characteristic when constructing a large-scale photonic integrated circuit designed to reduce crosstalk. This is in sharp contrast to other types of laser structure such as the conventional edge emitting lasers and VCSELs, in which there are two output ports.
Fig. 3 Mode profiles calculated by the FDTD method. The active region is inside the line-defect waveguide (a) without an output waveguide and (b) with an output waveguide. Calculated mode profiles (c) without output waveguide, (d) with output waveguide for Woffset = 3a, and (e) with output waveguide for Woffset = 4a, where a is lattice constant.
Figure 4 shows a cross-sectional SEM image of the fabricated BH-PhC laser. The active region was slightly etched after cleaving the device to obtain the SEM image. The active region consists of 3 quantum wells with a 1.55-μm photoluminescence (PL) peak sandwiched between InGaAsP barrier layers with a 1.35-μm PL peak. The BH must be fabricated within the line-defect waveguide to prevent any etching of the active region during the wet chemical etching of the InGaAs sacrificial layer. The active region was defined with a SiO2 mask using electron beam (EB) lithography and etched with inductively coupled plasma reactive ion etching (ICP-RIE) and selective wet chemical etching techniques. Then, the InP layer was selectively regrown by using metal organic chemical vapor deposition (MOCVD). After removing the SiO2 mask, we deposited a 35-nm InP layer to obtain a flat surface [15]. Air holes were etched with ICP-RIE and air-bridge structure was fabricated by the selective wet chemical etching of an InGaAs sacrificial layer through air holes. The fabrication of this structure allowed us to greatly reduce the thermal resistance and realize the strong confinement of both photons and carriers in the active region, because the thermal conductivity and bandgap of the InP layer are larger than those of the InGaAsP layer [13]. The width of the 3 quantum wells was 300 nm. A flat surface and a smooth dry-etched air-hole surface were obtained, which is important for realizing a high-Q cavity.
Fig. 4 Cross-sectional SEM image of the fabricated BH-PhC laser. The active region was slightly etched after cleaving the device to obtain a SEM image. The active region, consisting of 3 quantum wells, was buried in an InP layer.
3. Device characteristics
3.1 Static characteristics
Figure 5 (a) shows the light-in versus light-out (LL) characteristic and the external differential quantum efficiency near the threshold input power range. We define the external differential quantum efficiency as a ratio of photon number between the pumping light in the input line defect waveguide and the output light collected in the output line defect waveguide. In this experiment, we used a 5-μm-long active region with a volume of 5.0 x 0.3 x 0.16 μm3. A 1310-nm DFB laser was employed for optical pumping. The pump light was input into the input line defect waveguide from the fiber by using a collimator lens, a 3-μm wide waveguide and a tapered waveguide. We measured the transmission power of the reference waveguide, which have same structure we used for input waveguide except for the reference waveguide has no active region. From this measurement, we estimated the total coupling loss from the fiber to the line-defect waveguide for 1.3-μm light was 10 dB. The facet of the output waveguide was coated with AR film whereas that of the input waveguide was as cleaved. Thus, the laser output light from the output line-defect waveguide was collected into a single-mode fiber with an 8.5-dB coupling loss. The device exhibited a clear kink at a threshold of 6.8 μW at room temperature (298 K) with CW operation. We achieved a high external differential quantum efficiency of 53%, as shown in the figure. This indicates that the generated carrier was effectively converted to photon because BH was confined both the carrier and photon in the wavelength-sized cavity. Furthermore, the pumping waveguide effectively inputs the pump light into the active region, and the output waveguide effectively collects the laser output.
Fig. 5 (a) Light-in versus light-out (LL) characteristics and external differential quantum efficiency of near threshold region on a linear scale for the device with offset 3a. (b) LL characteristics for the device with offsets 3a (red line) and 4a (blue line).
Figure 5(b) also shows the LL characteristics with a wide input range for devices with offset 3a (red line) and 4a (blue line). The threshold input power of the device with offset 4a was decreased to 3.2 μW compared with the device with offset 3a because the Q-factor at the threshold was increased from 14,300 to 41,000. However, compared with the device characteristics described in ref. 15, the threshold input power was increased from 1.5 to 3.2 μW due to the reduction of the Q-factor from 71,400 to 41,000 and the increase in the number of quantum wells from 1 to 3. The maximum output power was −16.5 and −10.3 dBm for the devices with offset 3a and 4a, respectively. The maximum output power for the device with offset 3a became 20 times that of the previous device by increasing the quantum well number from 1 to 3. To the best of our knowledge, the maximum output power and external differential quantum efficiency are the highest reported values for a PhC nanocavity laser for CW operation at room temperature. These results clearly indicate that the BH improves the carrier confinement and thermal conductivity of the laser. These characteristics are excellent for constructing the high-density photonic integrated circuit.
3.2 Dynamic characteristics
Finally, we describe experimental results for direct modulation with optical pumping. This is an important characteristic when we use the device to construct a photonic network chip on a CMOS chip. The laser relaxation oscillation frequency is given by
where vg is the group velocity, g is the differential gain, Np is the photon density, τp is the photon lifetime, ηi is the internal efficiency, P is the input power, Pth is the threshold input power and V is the mode volume. As shown by this equation, to obtain a high-speed modulation with low power consumption, it is important to reduce the mode volume because the modulation speed of the laser increases with the input power density and the power consumption decreases with decreasing input power. In addition, this equation also indicates that the laser must maintain a low threshold input power, a high internal efficiency, and a differential gain for high-injected powers if we are to obtain high-speed direct modulation. As an increase in temperature adversely affects these parameters, reducing the increase in the active region temperature is important when the laser operates at a high input power density. These characteristics are inherently difficult for nanocavity lasers to achieve without a BH [13,14]. On the other hand, in terms of the signal detection, we need appropriate laser output power. Since the output power of the laser is proportional to the active volume, we should carefully design the active volume. As described previous section, we achieved output power of −10.3 dBm with 5.0 x 0.3 x 0.15 μm3 active region. We seem that this power is enough to detect the several tens of Gbit/s signal when we consider several dB loss in waveguide connecting between photonic switches as shown in Fig. 1.Figure 6(a) shows the experimental setup. To measure the direct modulation response of the fabricated laser, the 1.31-μm pump light was modulated by using a LiNbO3 modulator from 35 to 208 μW with an NRZ signal with a pseudo-random bit sequence (PRBS) of 231-1. Figure 6 (b) and (c) shows the eye diagrams of 1.55-μm output light modulated by 15- and 20-Gbit/s NRZ signals, respectively. In these experiments, the output signal was observed by using an optical fiber amplifier and an optical filter. As shown in these figures, clear eye opening was observed. This indicates that we can obtain enough output power to detect the signal. In previous our report in ref. 15, the modulated signal with sine-wave shape was observed by using averaging function of oscilloscope. The energy cost for transferring a single bit was estimated to be 8.76 fJ by taking into consideration a confinement factor of 0.616 and an absorption coefficient of 6000 cm−1 for a 1.31-μm pump light when the device was operated with a 20-Gbit/s NRZ signal. This is the smallest reported energy cost for any type of semiconductor laser. When we modulated the device with 25-Gbit/s signal, we did not observe clear eye-opening.
Fig. 6 (a) Experimental setup for direct modulation. Eye diagrams for (b) 15 Gbit/s and (c) 20 Gbit/s NRZ signals.
4. Conclusion
We have proposed and demonstrated an ultracompact BH-PhC laser, integrating in-plane pumping and output waveguides. The device exhibits a low threshold input power of 6.8 μW and a high output power of −10.3 dBm in the output waveguide. A high external differential quantum efficiency of 53% is also achieved, which indicates that a buried InP layer can be used for constructing waveguides connecting various elements. The maximum output power and external differential quantum efficiency are the highest reported values for a PhC nanocavity laser with CW operation at room temperature. Furthermore, we observed clear eye opening with 15- and 20-Gbit/s modulations. The energy cost for transferring a single bit was estimated to be 8.76 fJ/bit. In the future, we believe an electrically driven BH PhC laser can be realized by employing the selective regrowth of p-type, n-type, and semi-insulating InP buried layer [19], or ion-implantation of dopants into a semi-insulating InP buried layer [20]. Both methods have demonstrated enough forward current for our BH PhC laser, with the maximum injection current required for the laser estimated to be less than 100 μA. Thus, the BH-PhC laser is a promising device for constructing a photonic network chip on a CMOS chip.
Acknowledgements
We thank T. Segawa, Y. Shouji, K. Ishibashi, and T. Aihara for fabricating and testing the device. Part of this work was supported by the National Institute of Information and Communications Technology (NICT).
References and links
1. Y. Vlasov, W. M. J. Green, and F. Xia, “High-throughput silicon nanophotonic wavelength-insensitive switch for on-chip optical networks,” Nat. Photonics 2(4), 242–246 (2008). [CrossRef]
2. A. Shacham, K. Bergman, and L. P. Carloni, “Photonic networks-on-chip for future generations of chip multiprocessors,” IEEE Trans. Comput. 57(9), 1246–1260 (2008). [CrossRef]
3. N. Magen, A. Kolodny, U. Weiser, and N. Shamir, “Interconnect-power dissipation in a microprocessor,” Proc. 2004 Int. Workshop System Level Interconnect Prediction, Session Interconnect Anal. SoCs Microprocess., 7–13 (2004).
4. L. Benini and G. De Micheli, “Networks on chips: a new SoC paradigm,” IEEE Computer 35, 70–78 (2002). [CrossRef]
5. S. Vangal, J. Howard, G. Ruhl, S. Dighe, H. Wilson, J. Tschanz, D. Finan, P. Iyer, A. Singh, T. Jacob, S. Jain, S. Venkataraman, Y. Hoskote, and N. Borkar, “An 80-Tile 1.28TFLOPS Network-on-Chip in 65nm CMOS,” IEEE International Solid-State Circuits Conference, 2007. ISSCC 2007. 98 – 589 (2007).
6. http://www.itrs.net/Links/2007ITRS/Home2007.htm.
7. D. A. B. Miller, “Device Requirements for Optical Interconnects to Silicon Chips,” Proc. IEEE 97, 1166–1185 (2009). [CrossRef]
8. T. Tadokoro, T. Yamanaka, F. Kano, H. Oohashi, Y. Kondo, and K. Kishi, “Operation of a 25-Gb/s direct modulation ridge waveguide MQW-DFB laser up to 85 C,” IEEE Photon. Technol. Lett. 21(16), 1154–1156 (2009). [CrossRef]
9. K. Otsubo, M. Matsuda, K. Takada, S. Okumura, M. Ekawa, H. Tanaka, S. Ide, K. Mori, and T. Yamamoto, “1.3-μm AlGaInAs multiple-quantum-well semi-insulating buried-heterostructure distributed-feedback lasers for high-speed direct modulation,” IEEE J. Sel. Top. Quantum Electron. 15, 687–693 (2009). [CrossRef]
10. Y.-C. Chang and L. A. Coldren, “Efficient, high-data-rate, tapered oxide-aperture vertical-cavity surface-emitting lasers,” IEEE J. Sel. Top. Quantum Electron. 15(3), 1–12 (2009).
11. K. Nozaki, T. Tanabe, A. Shinya, S. Matsuo, T. Sato, H. Taniyama, and M. Notomi, “Sub-femtojoule all-optical switching using a photonic-crystal nanocavity,” Nat. Photonics 4(7), 477–483 (2010). [CrossRef]
12. M. Notomi, A. Shinya, K. Nozaki, T. Tanabe, S. Matsuo, E. Kuramochi, T. Sato, H. Taniyama, and H. Sumikura, “Low power nanophotonic devices based on photonic crystals towards dense photonic network on chip,” to appear in IET Circ. Dev. Syst.
13. M. Nomura, S. Iwamoto, K. Watanabe, N. Kumagai, Y. Nakata, S. Ishida, and Y. Arakawa, “Room temperature continuous-wave lasing in photonic crystal nanocavity,” Opt. Express 14(13), 6308–6315 (2006). [CrossRef] [PubMed]
14. K. Nozaki, S. Kita, and T. Baba, “Room temperature continuous wave operation and controlled spontaneous emission in ultrasmall photonic crystal nanolaser,” Opt. Express 15(12), 7506–7514 (2007). [CrossRef] [PubMed]
15. 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 13 fJ of energy consumed per bit transmitted,” Nat. Photonics 4(9), 648–654 (2010). [CrossRef]
16. E. Higurashi, D. Chino, T. Suga, and R. Sawada, “Au–Au Surface-Activated Bonding and Its Application to Optical Microsensors With 3-D Structure,” IEEE J. Sel. Top. Quantum Electron. 15(5), 1500–1505 (2009). [CrossRef]
17. T. Tanabe, M. Notomi, E. Kuramochi, A. Shinya, and H. Taniyama, “Trapping and delaying photons for one nanosecond in an ultrasmall high-Q photonic-crystal nanocavity,” Nat. Photonics 1(1), 49–52 (2007). [CrossRef]
18. M. Notomi and H. Taniyama, “On-demand ultrahigh-Q cavity formation and photon pinning via dynamic waveguide tuning,” Opt. Express 16(23), 18657–18666 (2008). [CrossRef]
19. T. Okumura, M. Kurokawa, M. Shirao, D. Kondo, H. Ito, N. Nishiyama, T. Maruyama, and S. Arai, “Lateral current injection GaInAsP/InP laser on semi-insulating substrate for membrane-based photonic circuits,” Opt. Express 17(15), 12564–12570 (2009). [CrossRef] [PubMed]
20. C.M. Long, A.V. Giannopoulos, and K.D. Choquette, “Photonic Crystal Band Edge Diode light Emitters,” Proceedings of Indium Phosphide and Related Materials 2009 ThA2.3, 399–402 (2009).
