Open Access
Add to BibSonomyAbstract
An Indium Phosphide-based device, switched by telecommunication wavelength laser pulses capable of operating at microwave frequencies up to 15 GHz has been designed and fabricated. Initial results confirm that using high energy nitrogen ion implantation to create EL-2 type trapping levels produces a photocarrier recombination time of a few picoseconds. The ion size and mass selected produces uniform bulk point defects in an In0.53Ga0.47As light absorbing region leading to high photocurrent mobility not exhibited in heavy ion irradiated samples resulting in a reduced peak pulse power requirement to switch the device.
©2012 Optical Society of America
1. Introduction
The requirement for ever faster switches for the sampling and processing of waveforms up to frequencies of tens of gigahertz has led to the growth of research into ultrafast photo-switches and pulsed lasers. The attraction of this approach is the low jitter achievable with pulsed lasers and the elimination of the signal crosstalk which is a characteristic of high frequency electronic switching. The Auston photo-switch [1] has been the basis of many such devices with low temperature grown Gallium Arsenide (LT-GaAs) being the most established material due to its ultrafast trapping characteristics and low leakage current [2]. However, the requirement for 800 nm illumination to switch such devices results in the use of bulky and expensive Ti-Sapphire pulse lasers, which has generally limited the use of such devices in sampling application due to the large footprint and cost of such systems. The use of a switch operating at the optical communication wavelength of 1550 nm is an attractive alternative due to the availability of compact fibre and solid state mode locked lasers. The use of ultrafast LT-In0.53Ga0.47As for such switches has been difficult owing to the difficulty in obtaining consistent low dark current and reducing surface recombination defects [3]. Other approaches to obtaining ultrafast InGaAs have been used, including Iron doping [4], Erbium Arsenide superlattice [5] and heavy ion irradiated InGaAs [6] which have concentrated upon the generation and detection of terahertz radiation. However, high energy, lighter ion implantation of InGaAs has not previously been attempted. The advantage of this process is the choice of ion mass, control of ion implantation energy and ion fluence to engineer a suitable material for the sampling of microwave frequencies up to 20 GHz.
2. Design
2.1 Device structure
The device consists of a 25 μm x 25 μm square mesa structure of 300 nm thick InGaAs lattice matched to a Fe:InP semi-insulating substrate buffered by an 80 nm Fe:InP layer. To create the switch, an interdigitated electrode is deposited upon the mesa (shown in Fig. 1(a) ), which is incorporated within a 25um wide conductor rail which is in turn part of a 750 μm long coplanar waveguide (Fig. 1(b)). To enhance the ohmic contact between the electrodes and semiconductor, 50nm of n-doped InGaAs is used (with 50 nm n-InP etch stop layer.). To create the increased population of carrier traps for ultrafast switching (5 x 1017 cm−3), the devices are irradiated with Nitrogen ions with implantation energy chosen such that they come to rest well below the active device region deep within the substrate. Therefore none of the implanted ions play any role in the behaviour of the device only the resulting displacement of target Ga and As atoms in the target material being significant.
Fig. 1 Micrograph of (a) InGaAs mesa with interdigitated electrodes and (b) complete device with coplanar waveguide upon Fe:InP dielectric
The inert nature of nitrogen in Indium phosphide is also an advantage and it has been shown that there is negligible effect upon the substrate permittivity [7], which could otherwise be detrimental to the transmission characteristics of the planar waveguide.
2.2 Ion implantation
Ion implantation of semiconductors is well established for the doping of semiconductors, however this involves the shallow deposition of the appropriate donor and acceptor ions in regions of the device. However, high energy implantation involves the irradiation of the target wafer with ions of such energy that they come to rest in an amorphous region deep below (typically 4 um) the device. The displacement caused by the ions colliding with the Gallium and Arsenic atoms in the InGaAs layer produce the required vacant Gallium and interstitial Arsenide sites such that the trapping level EL-2 (see Fig. 2 ) is created [3] which provides the picosecond photocarrier recombination times required.
Unlike LT-GaAs and iron doped InGaAs, no annealing stage is required to overcome large lattice defects which contribute to poor photoconductance. In contrast to heavy ion deep implantation [6], the photocarrier mobility of the InGaAs layer is quite high (6,900 V cm−1 s−1) indicating a low peak laser power to create the on-state resistance.
3. Fabrication
3.1 Switch processing
The devices were fabricated by wet etch methods using well established reagents and photolithographic methods. The initial stage required the deposition of metallic alignment marks to ensure the precise alignment of the etched mesa and interdigital electrodes. This was followed by the selective etching using H2PO3: H2O2:8H2O to remove the InGaAs layers and H2PO3:HCl for removal of the InP layers. Removal of the ohmic contact layer; etch stop layer, and absorbing layer was performed so that a 400 nm mesa was created (Fig. 3(a)). Using the alignment marks, the electrode and planar waveguide pattern was deposited as 20 nm Chromium and 350 nm Gold metal layer using conventional lift-off processes (Fig. 3(b)). Finally, the electrodes were electrically isolated using selective etching of the n doped layers using the gold pattern to mask the n –doped ohmic layers to remain under the interdigitated fingers.
Fig. 3. (a) Detail of mesa structure showing InGaAs/n-InP/n-InGaAs layers (b) Interdigitated electrode resist pattern on the etched mesa.
3.2 Ion implantation
The ion implantation was performed at the Ion Beam Centre at the University of Surrey using the 2 MV ion implanter. The ions were implanted at an energy of 4 MeV and 1013 cm−2 fluence. During implantation, the wafer sample was heated to 200°C and the ion beam orientated to a 7 degree angle normal to the surface to reduce channeling effects. From simulation using the SRIM (Stopping range of ions in matter) software a trap concentration of 5 x 1017cm−3 was estimated, producing an average recombination time of around 1 picosecond.
4. Evaluation
4.1 Direct current and optoelectronic characteristics
Measurement of the dark current of the implanted switches provided an indication of the off state resistance and carrier mobility. From measurement of the device conductivity, the carrier mobility was shown to be around 6,900 cm2V−1s−1. Plots of the device dark current and resistance are shown in Fig. 4 . To investigate the effects of ion implantation upon the electrical and optoelectronic properties of the devices, measurements of the InGaAs absorbing layers were made before and after implantation. The pre-implanted device shows the ohmic nature of the metal semiconductor contacts with an off resistance of around 2,300Ω. After implantation, the device exhibited a higher resistance of 6,050Ω. The resistance between conductor and ground planes of the coplanar devices was measured to ensure that this did not contribute towards the value of the off state resistance. This was measured as more than 200 kΩ, which indicated that the leakage current across the waveguide was negligible.
In addition, the photocurrent was measured for increasing incident optical powers from an amplified CW 1550 nm source delivered to the active region by a lens ended fibre, with a modulated signal to prevent overheating the device under CW illumination. The results are shown in Fig. 5 on the next page. The plots in Fig. 5 give an indication of the required peak pulse powers needed to provide a value of the on state resistance of 50 Ohm which is the required on-state resistance. If it is considered that there are negligible substrate heating effects or non-linear behaviour at high pulse powers and role of hole transport is negligible in the ultrafast device, then the photocarrier concentration is proportional to the incident optical power (1).
By considering the background n-doping of 1.01 x 1015 cm−3 from dark current analysis and determining the carrier concentration while illuminated by 50 mW CW laser (nphoto = 4.17 x 1014 cm−3), then the estimated carrier concentration for 50 Ohm switch resistance is ntotal = 9 x 1016 cm−3 which requires an increase in carrier concentration of n = 8.89 x 1016 cm−3 which has been calculated as a peak optical pulse power of 11.2 W.Fig. 5 (a) Plot of Current vs. Bias voltage for dark and illuminated switch by 1550 nm laser. (b) Carrier concentration derived from switch conductance vs. incident laser (CW) power.
4.2 Temporal response
To measure the temporal response of the implanted switches, the devices were illuminated by an erbium fibre based 1,550 nm mode locked fibre laser. With optical amplification, the peak laser pulse power achieved was 1.3 Watts for a 2 ps wide pulse. The electrical signal of the biased switch was recorded on a high bandwidth sampling oscilloscope as shown in Fig. 6 . The recombination time was measured to be 4.4 ps which would enable sampling for an ideal waveguide with no leakage into the substrate in excess of 50 GHz. Prior to implantation, the recombination time had been measured at around 2.5 ns therefore a 500-fold decrease in the recombination time had been achieved. In addition, an estimation of the conductance increase of the switch at the peak voltage, provides another estimation of carrier concentration at the mode locked laser peak power of 1.3 W of around n = 9.04 x 1015 cm−3. Making the same assumptions discussed in Section 4.1 above, linear scaling of photocarrier concentration with peak laser power provides an estimation of peak pulse power of 12.8 W incident upon the switch which is similar to the value derived from the CW evaluation discussed previously.
4.3 Frequency response in the off state
As an indication of the switch operation over the range of its intended input frequencies, the frequency response of the device was measured from 1 GHz to 20 GHz. The results are shown in Fig. 7 .
The plot can be explained by considering the equivalent circuit of the device (Fig. 8 ). The device behaves as a high pass filter with the electrode gap producing the capacitive element while the photoconductor provides the resistive part. At lower frequencies the capacitative reactance of the interdigital electrode is high and device impedance is predominantly resistive, but as the frequency increases, the capacitive reactance falls reducing the device impedance. In the on state, the resistive element forms the transmission path.
The switch capacitance was measured as 8 fF.
4.4 Simple switching evaluation
An initial investigation of the devices switching capabilities was observed by the measurement of the microwave output while illuminated by a relatively low power (10 μW) low repetition rate (3.75 MHz) mode-locked fibre laser delivering a 127 mW peak power pulse with a 2 ps FWHM pulse width. The resulting side-bands indicated the switch was capable of switching up to 15 GHz. as shown in the Fig. 9 below.
Fig. 9 Output spectra of switch illuminated by 1550nm mode-locked laser with 3.75 MHz repetition rate showing modulation sidebands at 1GHz, 10 GHz and 15 GHz
The output spectra of the switches show that they are capable of switching frequencies up to 15 GHz even at the relatively low laser illumination power of 127 mW peak. As optimum switching (when Ron = 50 Ω) is expected to occur at a peak laser power of at least 11 W (as discussed in section 4.1), then the sidebands would be expected to be prominent at higher pulse powers and to be observed up to higher frequencies as the switch impedance would be predominantly resistive rather than effects of the capacitive reactance observed at lower laser powers.
5. Conclusion
The high energy nitrogen ion implantation of an InP based photoconductive switch has produced a device capable of switching high frequency signals up to 15 GHz. Using an higher power pulsed laser source and analogue to digital circuitry, sampling could be achievable for microwave waveforms of higher frequencies. The high photocarrier mobility shows that the optical peak pulse power theoretically required to switch the device will be lower than that required for LT-GaAs switches and hence the power requirement for such a sampling system would be much reduced. This is the first demonstration of an high energy nitrogen ion implanted Indium phosphide based photo-switch and these results indicate that it has several advantages over similar ultrafast switches.
References and links
1. D. Auston, “Picosecond optoelectronic switching and gating in silicon,” Appl. Phys. Lett. 26(3), 101–103 (1975). [CrossRef]
2. R. Urata, R. Takahashi, V. Sabnis, D. Miller, and J. Harris “High-speed Sample and Hold using Low Temperature Grown GaAs MSM switches for Photonic A/D Conversion,” CLEO 2001 Tech. Dig. 66–67, May 2001.
3. A. Kroktus and J.-L. Coutaz, “Non-stoichiometric semiconductor materials for terahertz optoelectronics applications,” Semicond. Sci. Technol. 20(7), S142–S150 (2005). [CrossRef]
4. T. Kimura, S. Yamamura, K. Koike, T. Morita, S. Yugo, and T. Kamiya, “Realization of Fast InGaAs Photoconductive Response by Ion Implantation and Annealing with No Degradation of Peak Responsivity,” Jpn. J. Appl. Phys. 29(7), 1270–1275 (1990). [CrossRef]
5. J. Bjarnason, T. Chan, A. Lee, E. Brown, D. Driscoll, M. Hanson, A. Gossard, and R. Muller, “ErAs:GaAs photomixer with two-decade tunability and 12uW peak output power,” Appl. Phys. Lett. 85(18), 3983–3985 (2004). [CrossRef]
6. J. Delagnes, P. Mounaix, H. Nemec, L. Fekete, F. Kadlec, P. Kuzel, M. Martin, and J. Mangeney, “High photocarrier mobility in ultrafast ion-irradiated In0.53Ga0.47As for terahertz applications,” J. Phys. D Appl. Phys. 42(19), 195103 (2009). [CrossRef]
7. E. P. Burr, M. Pantouvaki, A. J. Seeds, R. M. Gwilliam, S. M. Pinches, and C. C. Button, “Wavelength conversion of 1.53-microm-wavelength picosecond pulses in an ion-implanted multiple-quantum-well all-optical switch,” Opt. Lett. 28(6), 483–485 (2003). [CrossRef] [PubMed]
