1. Introduction
InP-based Digital Optical Switch (DOS) has been demonstrated using integrated optics a long time ago [
1
J.F. Vinchant, M. Renaud, M. Erman, J.L. Peyre, P. Jarry, and P. Pagnod-Rossiaux, “InP digital optical switch: key-element for guided-wave photonic switching,” IEE Proceedings
140, 301–307 (1993)
]. The switching effect is based on a carrier induced refractive index change [
2
B.R. Bennet, R.A. Soref, and J.A. Del Alamo, “Carrier-induced change in refractive index of InP, GaAs and InGaAsP,” IEEE Journal of Quantum Electronic.
26, 113–122 (1990) [CrossRef]
]. The electrical component is basically a (direct biased) PIN diode which intrinsic layer is constituted by the optical waveguide epilayers. The optical component is a Y-junction. The current flow follows the electrodes geometry and the so-induced carrier induced index variation modifies the propagation within the Y-junction directing the light beam towards one output arm (the one which electrode is unbiased). Combining DOS structures in integrated optics in order to obtain a switch matrix leads to rather long devices. In this paper, we propose using nanophotonics for the realisation of the switching function that potentially offers the possibility decreasing the size of components. We use narrow deep-etched InP/InGaAsP waveguides [
3
M. Lesecq, S. Maricot, J.P Vilcot, and M. Beaugeois, “Passive photonic components using InP optical wire technology,” IET Optoelectronics (to be published)
]. A first attempt has been reported [
4
R. Grover, A. Ibrahim, S. Kanakaraju, L. Lucas, L.C Calhoun, and P.T Ho, “A tunable GaInAsP-InP Optical Microring Notch Filter,” IEEE Photon. Technol. Lett.
16, 467–69 (2004) [CrossRef]
]; it reports on the electrical tuning of a GaInAsP/InP microring notch filter. In this work, a localised electrode structure process has been set-up in order to precisely define the carrier injection zones.
2. Device presentation
The original aspect of our structure relies on the use of nanophotonics structures and, in particular, of narrow deep-etched waveguides. The DOS-like operation is still obtained using carrier induced index change obtained by current injection into a forward biased PIN diode structure.
The epilayer structure is grown on an n-doped substrate (
Fig. 1(a)). The waveguide is composed of a 0.3µm-thick InGaAsP quaternary material (cut-off wavelength λ
c=1.3µm) guiding layer and two 1.2µm-thick InP top and bottom cladding layers: these 3 layers realize the passive optical waveguide. A 0.1µm-thick p-doped InP layer and a 0.2µm-thick p-doped InGaAs top contact layers are added to improve the p-type contact resistance. These layers will be obviously removed all over the wafer except under electrodes; these parts of the switch can be denominated as the active part.
Optical waveguides are obtained by etching down ridges to 3µm. Their width is 0.8µm or 1µm. The lower cladding layer is intended to cut off light leakage by coupling into the substrate and the upper one to avoid excess loss due to light absorption by electrodes. The vertical confinement is “low” and is linked to the material index difference between the quaternary core (3.42) and the cladding (3.16) and is so comparable to the one obtained in “classical” integrated optics. But a high lateral confinement is obtained owing to the index difference higher than 2 between the semiconductor material and surrounding air. The fabrication details as well as the characterization of such passive micro-waveguides have been reported in [
3
M. Lesecq, S. Maricot, J.P Vilcot, and M. Beaugeois, “Passive photonic components using InP optical wire technology,” IET Optoelectronics (to be published)
].
The switch structure is based on two asymmetric Y-junctions laid out in a serial configuration (
Fig. 1(b)). Firstly, the transcription of “classical” integrated optics symmetrical DOS structures has been attempted and modelled. Due to the high lateral confinement, a simple Y-junction does not afford low crosstalk values. Although value around -12 dB can be achieved, trade-off getting simultaneously decrease of crosstalk and switching losses values is difficult to reach. In order to overcome this, we moved to an asymmetric Y-structure and the use of a serial arrangement of 2 elementary switches. As it can be noticed on
Fig. 1(b), 3 outputs are so obtained, the two farthest ones are the real switch outputs; the centre one is not used. Whatever the technology is, light always prefers going straight ahead; this is even truer considering these high lateral confinement micro-waveguides. The centre waveguide collects the unswitched light and so contributes to improve the crosstalk between the two other outputs. Y-junction and electrode designs have been determined using Optiwave © modelling tools. Electrodes are symbolized by index variation areas. The index variation that has been used in modelling is -5×10
-3 which corresponds to previously recorded index variation in such material structure at this wavelength [
5
M. Zegaoui, D. Decoster, J. Harari, J-P. Vilcot, F. Mollot, V. Magnin, and J. Chazelas, “Comparison between carried induced optical index, loss variations and carrier lifetime in GaInAsP/InP heterostructures for 1.55 µm DOS applications,” Electron. Lett.
41, 613–614 (2005) [CrossRef]
]. Electrode length is 800 µm and they are separated by a 1 µm gap.
In order to switch light on Output 1, both electrodes, #2a and #2b, shall be simultaneously current driven (see
Fig. 1(b)); identical operation exists for Output 2 and electrodes #1a and #1b. A common bondpad has been so designed in order to only get one electrical connection for each switching direction. Thanks to a smart bond pad report process [
6
G. Ulliac, S. Garidel, J.P. Vilcot, and P. Tilmant, “Air bridge interconnection and bond pad process for non-planar compound semiconductor devices,” Microelectron. Eng.
81, 53–58 (2005) [CrossRef]
], the tiny electrodes have been connected to the bond pads. The DC characterization of the so fabricated dual PIN diodes shows a serial resistance of 5.7 ohms and a direct voltage of 1.6V at 100 mA drive current.
Figure 2 shows a SEM (Scanning Electronic Microscope) view of the whole structure. The input waveguide can be distinguished on the upper left side, and the three output waveguides on the lower right side. After the active parts, low radius bents [
3
M. Lesecq, S. Maricot, J.P Vilcot, and M. Beaugeois, “Passive photonic components using InP optical wire technology,” IET Optoelectronics (to be published)
] are used to separate the outputs. The length of the active parts of the switch is 1.7mm (two 800µm-long Y-junctions and extra 100µm-long gap in between). The width is 200 µm which is almost linked to bondpad dimensions. Taking into account fibre access parts, overall length is close to 2.5 mm owing to the very small extra space needed for the separation of the different outputs. Lateral tapers are used at input and output waveguides to improve coupling efficiency with a lensed fibre [
3
M. Lesecq, S. Maricot, J.P Vilcot, and M. Beaugeois, “Passive photonic components using InP optical wire technology,” IET Optoelectronics (to be published)
].
Fig.1a. Schematic crossview of waveguide structure
Fig 1b. Schematic overall view of optical switch using deep etched waveguides
Fig.2. SEM view of a switch
3. Crosstalk measurement
Light injection is made using a 1.55µm-wavelength laser and a lensed fibre. Light polarization is TM (electrical component of optical field perpendicular to epitaxial layer planes). Although DOS switch can be considered to have low light polarization sensitivity, narrow deep-etched waveguides used here have high; in particular their propagation losses are higher in TE polarization state than in TM polarization state [
3
M. Lesecq, S. Maricot, J.P Vilcot, and M. Beaugeois, “Passive photonic components using InP optical wire technology,” IET Optoelectronics (to be published)
]. Output optical beams are collected either by an infra-red camera for near field analysis or by an as-cleaved fibre connected to an optical power meter for power ratio measurements. A current generator is used to launch the driving current into the switch active areas.
First, the near field analysis is performed in order to get a quick overview of switch behaviour.
Figures 3(a) and
3(b) show respectively passive and active states. Without current injection (passive state), output beam power of output1/direct output/output2 has been respectively measured to -6dB/0dB/-5dB taking direct output power as reference channel. Under current injection (50mA -25 mA per elementary Y-junction- for the case corresponding to the picture of
Fig. 3(b)), the light beam is effectively switched towards the unbiased waveguides. Qualitatively, we can appreciate the crosstalk but also the usefulness of the centre waveguide.
Then, fibre to fibre measurements are performed to evaluate the crosstalk. Output powers are recorded (
Fig. 4) versus the injected current and normalized in relation to the initial power, obtained without current injection (passive state of
Fig. 3(a)). Under a 100mA drive current (so 50mA per elementary Y-junction), a power decrease of 16dB at the unswitched output port (Output 1) and a power increase of 3dB at the switched output port (Output 2) are obtained. The crosstalk between the two outputs is so close to -19dB. The insertion loss of the switch (switched output power/launched input power) has been measured close to -30 dB; a large part of this loss being due to coupling ones although the presence of the lateral tapers.
Fig.3a. Output beams without current injection
Fig.3b. Output beams with a current injection of 50mA
Fig. 4. Normalised measured output power versus injected current
4. Results discussions
As it can be seen on
Fig. 4, the overall evolution of power repartition versus current injection within this switch is almost the same that the one recorded for “classical” integrated optics [
7
K. Blary, S. Dupont, J.P. Vilcot, F. Mollot, D. Decoster, and J. Chazelas, “DOS optical switch for microwave optical links based applications,” Electron. Lett.
38, 1697–1699 (2002) [CrossRef]
]. Nevertheless, the recording of the power issued by the direct output (not used) shows that crosstalk with the switched output is around -10dB. This shows the usefulness of the topology but also the intrinsic switching losses of such a structure. These first results also allowed evaluating a lateral spreading of carriers closed to 1µm. This value can now be taken into account to feedback on modelling and to optimize more deeply the switching region, in particular, the electrode positioning can be revised. This should lead to decrease the crosstalk and switching losses values as well as the driving current.
5. Conclusion
We realized the first optical switch based on narrow deep-etched waveguides in the InP material line. The switch used two asymmetric DOS-like junctions following each other. This arrangement has been defined since trade-off between crosstalk and switching losses values is very difficult to obtain on a single Y-junction using such high lateral confinement optical waveguiding structures. In counterpart, provided that the etching depth is high enough (so at least 3 µm for our structure) the performance of the switch only depends on dimension (lithography) issues that can be easily handled. An optical crosstalk close to -20dB is experimentally achieved for a current of 100mA.
Acknowledgments
This work has been partly supported under the ERDF funded Belgian-French Interreg III “PREMIO” project as well as the project # 02 60 65 092 from the “Délégation Générale pour l’Armement” of the French Ministry of Defence.
References and links
1. |
J.F. Vinchant, M. Renaud, M. Erman, J.L. Peyre, P. Jarry, and P. Pagnod-Rossiaux, “InP digital optical switch: key-element for guided-wave photonic switching,” IEE Proceedings
140, 301–307 (1993) |
2. |
B.R. Bennet, R.A. Soref, and J.A. Del Alamo, “Carrier-induced change in refractive index of InP, GaAs and InGaAsP,” IEEE Journal of Quantum Electronic.
26, 113–122 (1990) [CrossRef] |
3. |
M. Lesecq, S. Maricot, J.P Vilcot, and M. Beaugeois, “Passive photonic components using InP optical wire technology,” IET Optoelectronics (to be published) |
4. |
R. Grover, A. Ibrahim, S. Kanakaraju, L. Lucas, L.C Calhoun, and P.T Ho, “A tunable GaInAsP-InP Optical Microring Notch Filter,” IEEE Photon. Technol. Lett.
16, 467–69 (2004) [CrossRef] |
5. |
M. Zegaoui, D. Decoster, J. Harari, J-P. Vilcot, F. Mollot, V. Magnin, and J. Chazelas, “Comparison between carried induced optical index, loss variations and carrier lifetime in GaInAsP/InP heterostructures for 1.55 µm DOS applications,” Electron. Lett.
41, 613–614 (2005) [CrossRef] |
6. |
G. Ulliac, S. Garidel, J.P. Vilcot, and P. Tilmant, “Air bridge interconnection and bond pad process for non-planar compound semiconductor devices,” Microelectron. Eng.
81, 53–58 (2005) [CrossRef] |
7. |
K. Blary, S. Dupont, J.P. Vilcot, F. Mollot, D. Decoster, and J. Chazelas, “DOS optical switch for microwave optical links based applications,” Electron. Lett.
38, 1697–1699 (2002) [CrossRef] |