Bistable Optically Controlled Semiconductor Switches in a Frequency-Agile RF Source

The processes of persistent photoconductivity followed by photoquenching have been demonstrated at megawatt power levels in copper-compensated, silicon-doped, semi-insulating gallium arsenide. These processes allow a photoconductive switch to be developed that can be closed by the application of one laser pulse ( = 1.06 m) and opened by the application of a second laser pulse with a wavelength equal to twice that of the first laser ( = 2.13 m). This switch is called the bistable optically controlled semiconductor switch (BOSS). The opening phase of the BOSS requires a sufficient concentration of recombination centers (RC) in the material for opening to occur in the subnanosecond regime. These RC’s are generated in the bulk GaAs material by fast-neutron irradiation ( 1 MeV). Neutron-irridated BOSS devices have been opened against a rising average electric field of about 36 kV/cm (18 kV) in a time less than 1 ns while operating at a repetition rate, within a two-pulse burst, of about 1 GHz. The ability to modify the frequency content of the electrical pulses, by varying the time separation, is demonstrated. Results demonstrating the operation of two BOSS devices imbedded in a frequency-agile RF source configuration are also discussed.


Bistable Optically Controlled Semiconductor
Switches in a Frequency-Agile RF Source

I. INTRODUCTION
A LMOST all photoconductive switches are closed either by direct excitation across the bandgap, the so-called linear mode, or by a very low energy trigger pulse from a laser diode, the so-called gain, lock-on, or avalanche mode.An alternative switching mechanism was proposed by Schoenbach et al. [1].This concept, called the bistable (or bulk) optically controlled semiconductor switch (BOSS), relies on persistent photoconductivity followed by photoquenching to provide both switch closing and opening, respectively.Persistent photoconductivity results from the excitation of electrons from the deep copper centers found in copper-compensated, silicondoped, semi-insulating GaAs (GaAs:Si:Cu).The small cross section for electron capture back into the Cu centers allows long conduction times after the first laser pulse is terminated.Photoquenching is accomplished by the application of a second laser pulse, of longer wavelength, which elevates electrons from the valence band back into the copper levels.This laser pulse floods the valence band with free holes which rapidly recombine with free electrons to quench the photocurrent over a time scale given by the electron-hole lifetime of the material.These processes allow a switch to be developed which can be closed by the application of one laser pulse m) and opened by the application of a second laser pulse with a wavelength about twice that of the turn-on laser.
Preliminary experimental results showed that the current through a BOSS switch could not be fully quenched by the application of a 140 ps (FWHM) 2.13 m laser pulse.A numerical solution of the semiconductor rate equations for copper-doped GaAs showed that the primary cause for incomplete photoquenching was that the concentration of the recombination centers (RC) was too low [2].As stated above, the opening transient is the result of a two-step process.The second step is controlled by the electron-hole recombination lifetime in the bulk material.If there is an insufficient RC concentration, the holes that were generated by the 2 m laser pulse would be retrapped into the copper centers before they could recombine with electrons in the conduction band.This would result in the switch remaining closed after the second laser pulse.Wang et al. [3] recently reported reduction of the minority-carrier lifetime by increasing the RC concentration in GaAs using fast-neutron irradiation.This work directed us toward the investigation of neutron damage for the purpose of RC enhancement in BOSS devices [4].The neutron source that was used in this work is Sandia National Laboratories' (SNL) SPR-III reactor which has an energy spectrum peaked at about 1 MeV.

II. SAMPLE PREPARATION
Low-resistivity, silicon-doped (n-type) GaAs can be made semi-insulating by the introduction of copper acceptor levels through a thermal-diffusion process [5].The GaAs material used in this investigation was originally doped with a silicon concentration of 2 10 cm which yielded a resistivity of about 7 10 cm.After the thermal diffusion step, at C for 6 h, the samples were polished on both sides to a mirror finish.The sample dimensions were roughly 10 5 0.5 mm.The p -i-n devices were manufactured by depositing an Au-Ge based metallization for the n-type contact and an Au-Zn based metallization for the p-type contact.The contacts resulted in a switch geometry that was 5 mm wide and separated by a 5 mm gap on the same side of the sample.After deposition, the contacts were annealed at 450 C for 5 min in at atmospheric pressure.Following the contact anneal, the samples were irradiated with fast neutrons to increase the RC concentration.Two sets of BOSS devices were neutron irradiated at two different fluences.The lower fluence was measured at 2.45 10 cm (sample ), while the higher fluence was measured at 3.93 10 cm (sample ) 1 MeV GaAs equivalent damage [6].The dc I-V characteristics U.S. Government work not protected by U.S. copyright of the samples indicated an increase in the switch resistance from about 3.2 to about 55 M for the lower fluence, and an increase from about 4.3 to about 273 M for the higher fluence.

III. SWITCHING EXPERIMENTS
The BOSS-switching experiments were conducted with a mode-locked Nd:YAG laser system (1.06 m), manufactured by Continuum Inc., that was equipped with an optical parametric generator (OPG) that served to double the wavelength (2.13 m).The laser system produced a Gaussian pulse with an FWHM of about 140 ps.A simple optical delay then was used to adjust the time between switch closure and when the switch was opened.Photoconductivity measurements were performed to evaluate the operation of the neutron-irradiated BOSS devices.The BOSS switches were embedded in a 50 transmission line (two-way transit time 8 ns) that was pulse charged with roughly a 40 ns FWHM voltage pulse generated by a Krytron switch as shown in Fig. 1.The maximum voltage applied to the BOSS devices was about 18 kV.The current through the device was measured by a 50 current-viewing resistor (CVR) placed after the switch in the 50 line.The current waveform was recorded by a Tektronix SCD5000 digitizer with a 3.0 GHz analog/digital bandwidth.

A. Lower Fluence Switching Results
Switching results illustrating the photocurrent for sample are shown in Fig. 2 for an applied voltage of 3.7 kV.The maximum voltage that was switched with this device was biasmodulator limited to about 18 kV.The switching behavior of sample did not change as the applied voltage was increased.In order to demonstrate the ability to control the pulsewidth of the electrical pulse delivered to the 50 CVR, Fig. 2 shows several current waveforms superimposed as the turn-on and turn-off laser pulse separation varied from 400 ps to 1 ns.The laser pulse energy for both the 1 and 2 m wavelengths was set at 4.5 mJ.The minimum switch resistance during the transient was measured to be The minimum pulsewidth achieved within this device was measured to be about 650 ps FWHM.A curve fit to the switch conductance during the opening phase, after it was extracted from the circuit load line, indicated a recombination time constant of 100 ps.
Fig. 2 illustrates that, as the time between the two laser pulses is increased, the switch conductivity decreases with time prior to the turn-off laser pulse as a result of the enhanced RC density.This effect ultimately will limit the switch on time of neutron-irradiated BOSS devices.Therefore, there appears to be a tradeoff between the maximum time that the switch will remain closed after the 1 m laser pulse and the RC density in the material.It should be noted that BOSS devices that were not irradiated with neutrons have demonstrated on times of several hundred nanoseconds [7].If we define switch efficiency in terms of the amount of energy transferred to the load for a given amount of turn-on energy, then increasing the RC density reduces the amount of energy transferred by the switch, resulting in a lower efficiency.
One of the goals of this research was to determine the minimum electrical pulsewidth that could be generated by a BOSS device, while still maintaining the ability to vary the pulsewidth by adjusting the time delay between the turnon and turn-off laser pulses.Fig. 3 illustrates the effect of further reducing the time between the two laser pulses from the minimum time separation shown in Fig. 2. The dashed line in Fig. 3 shows the smallest pulsewidth illustrated in Fig. 2. Fig. 3 shows that the pulsewidth actually increases as laser pulse separation decreases.The time-delay reduction between waveforms in Fig. 3 is roughly 100 ps.When the time between laser pulses is too short, the 2 m laser pulse is no longer able to completely quench the photocurrent.Eventually, when the two laser pulses are roughly coincident, the switch behaves as if there were no 2 m laser pulse at all.Note that the peak amplitude of the current was already load-line limited, so no increase is observed when the two pulses overlap.
The primary reason why the switch will not open properly when the laser pulses are too close together is that it takes a certain amount of time for the electron-hole plasma generated by the 1 m laser pulse to recombine.This, coupled with the fact that the dominant copper center requires a certain amount of time to fill with a sufficient quantity of holes to permit complete photoquenching, sets a lower limit on the time between the turn-on and turn-off laser pulses.The minimum electrical pulsewidth that can be generated by BOSS devices has been examined numerically using a rateequation model [8].Using known material parameters for GaAs from the literature, and assuming "reasonable" values for the recombination center and cross sections, the same qualitative behavior was predicted.

B. Higher Fluence Switching Results
Photoconductivity experiments also were conducted on sample , which was irradiated at a fluence of 3.93 10 cm .One drawback of an increased RC concentration is that the on-state conductivity will be reduced because electrons in the conduction band will recombine with holes in the valence band before those holes can be trapped in the center.This process reduces the number of holes that are trapped in the center which, in turn, reduces the available sites to receive electrons from the valence band during the turn- off laser pulse.However, a benefit can be derived if the RC concentration in the bulk material is made high enough to cause the switch to open without the need of the turn-off laser pulse.This effect is shown in Fig. 4 where two 1m laser pulses were used to close sample at a high repetition rate.For these experiments, the switch was only illuminated by two 1m laser pulses with a variable time delay between them.Sample opened rapidly even without the turn-off laser pulse because it was irradiated at a higher neutron fluence than sample , yielding a higher RC density in the bulk material of sample .
The purpose of this experiment was twofold.First, we wanted to see how the switch responded to the turn-on laser pulse, and second, we wanted to test the repetition rate capability of sample .The applied voltage for the waveform shown in Fig. 4 was about 16 kV.The minimum on-state resistance that was measured for sample was about 20 The average pulse width was measured to be about 340 ps.The time separation between the two laser pulses was varied from about 3.5 ns, corresponding to a repetition rate of roughly 285 MHz, down to less than 1 ns, corresponding to a repetition rate of greater than 1 GHz.These repetition rates are basically five orders of magnitude higher than any other type of highpower photoconductive switch.The maximum voltage that was switched with this device was 18 kV, yielding an average electric field of 36 kV/cm.
The most striking attribute of the current pulses in Fig. 4 is that the switch completely opened without the 2 m laser pulse, and with no indication of the device collapsing into a filamentary-current mode of conduction at any point in the switching cycle.This is significant since almost all previously reported photoconductive switching experiments on nonneutron-irradiated GaAs, including those performed on GaAs:Si:Cu material, exhibited a transition into filamentary conduction at average electric fields of kV/cm [9].It has been reported by Loubriel et al. [10], however, that chromedoped GaAs switches, that were irradiated at a neutron fluence of 5 10 cm , did not transition into a filamentary conduction mode until the applied average electric field was greater than 62 kV/cm.Assuming that similar results could be achieved with the 5 mm gap BOSS devices, this electric field would correspond to an operating voltage of roughly 30 kV.Ultimately, normal avalanche breakdown due to impact ionization would limit the operating voltage to about 50 kV.
Since one of the potential applications of a BOSS device is in an impulse or ultrawide-band (UWB) radar, we examined the spectral frequency content of the waveforms in Fig. 4 to determine the effects of pulse separation.The energy density spectrum was obtained using standard fast Fourier transform (FFT) techniques on 1024-point data sets that were zeropadded to 4096 points for better frequency resolution.The spectral variations of the pulses from Fig. 4 are shown in Fig. 5. Fig. 5 shows that both pulses had significant frequency content up to about 3 GHz, which was the bandwidth limit of the SCD5000 digitizer.The result of adjusting the delay between the two pulses can be seen as a change in the number and location of the nulls in the spectra.The details of the spectral shape depend directly on the shape of the time-domain waveform, and therefore might vary from device to device due to statistical variations in material parameters.However, the existence and location of the nulls result solely from the pulse separation, and should not be device dependent.

C. BOSS-Based Rf Source
The capability to dynamically vary the location of spectral nulls could be very useful in UWB radar applications, particularly if certain frequency bands are to be avoided.The primary goal of this research was to produce a wide-band, frequencyagile source that can radiate the RF energy with a broadband antenna.In order to maximize the radiative efficiency of the source, it is necessary to produce ac power, thereby reducing the nonradiating dc component of the waveform.The ability of the BOSS switch to open, as well as close, in the subnanosecond regime enables a new type of RF source to be developed that is capable of generating repetitive highpower microwave cycles of varying duration, depending on the relative delay between the turn-on and turn-off laser pulses.A source configuration that is capable of generating ac power with a real-time frequency agility is shown in Fig. 6, and is called the pulse-switch-out (PSO) generator.This source uses two BOSS switches that are embedded in oppositely charged 50 transmission lines which can generate single positive and negative half-cycles by first closing and opening each switch.Both switches then feed into a single 50 transmission line that leads to a matched load.
Experiments were conducted with the circuit shown in Fig. 6, at a bias voltage of about 9.5 kV, and with BOSS devices that were irradiated at the higher neutron fluence and, therefore, only illuminated with 1m laser pulses.Experiments are planned with devices that require both of the laser pulses.The time-domain waveforms for a time delay of about 500 ps and about 2 ns are shown in Fig. 7.We found that the pulsewidth of the postive half-cycle could be reduced, at the expense of the negative half-cycle peak voltage, by closing the second BOSS switch before the first one was completely open.The Fourier spectra for these two waveforms differ significantly, as shown in Fig. 8.As expected, the bipolar pulses significantly reduced the dc component in the spectra.In addition, the generation of nulls in the spectra increased the power spectral density at some of the lower frequencies.As before, the number and location of these nulls can be adjusted by varying the time delay between the two laser pulses.

D. Potential RF Waveforms
The laser system that is required to test the operation of the PSO generator at high repetition rates is not presently available.However, since the operation of the switch has been demonstrated at about 1 GHz in a burst mode, and since the operation of two BOSS devices in a PSO generator also has been demonstrated, it is reasonable to examine the effect of generating a short burst of RF cycles at megahertz repetition rates.To accomplish this, a potential RF output from a PSO generator was simulated by replicating the waveform shown in Fig. 7 for a laser pulse separation of 500 ps, and assuming a  time separation between RF cycles of 10 ns.The result of this data manipulation is shown in Fig. 9, where four RF cycles, at a 100 MHz repetition rate, are given with the total time window set at 80 ns.Fig. 9 is an illustration of the type of waveform that is expected to be generated by a PSO RF source once such a laser system becomes available.
The energy density spectra of the single 500 ps monocycle and the fabricated waveform shown in Fig. 9 are illustrated in Fig. 10.The spectrum for the single monocycle has a maximum amplitude of about 0.1 pJ/Hz.The spectrum of the  four-pulse burst has a similar spectral envelope, but develops lobes spaced at 100 MHz as a result of the 10 ns spacing between the RF cycles.The maximum spectral energy density of the four-pulse burst is about 1.5 pJ/Hz, 16 times greater than the single-pulse case.This is due to the simple increase in energy from the increase in the number of pulses, as well as the energy "compression" which occurs when the number of peaks and nulls increases.
The same type of results are obtained when we fabricate a four-pulse burst from the more widely spaced 2 ns monocycle pulse in Fig. 7, resulting in the waveform shown in Fig. 11.The spectra for the monocycle and four-pulse burst for this case are shown in Fig. 12.The maximum of the four-pulse burst spectrum is about 3.5 pJ/Hz, an increase by more than a factor of 2 over the case where the pulses were separated by 500 ps.These results indicate that the time delay between the positive and negative half-cycles dictates the shape of the envelope of the spectrum, while the cycle repetition rate within the burst dictates the placement of the various spectral peaks.The type of waveforms shown in Figs. 9 and 12, with their associated spectra shown in Figs.11 and 13, respectively, may have considerable advantages for use in UWB-radar and other high-power microwave applications.

IV. CONCLUSION
Experiments were performed to determine the effect of irradiating BOSS material with two different 1 MeV neutron fluences.For the lower fluence of 2.45 10 cm , the optically induced closing and opening effects were demonstrated at voltages up to 18 kV.Experimental data were presented which demonstrated the pulsewidth agility of the BOSS switch in the subnanosecond range.Results also were presented for the operation of BOSS devices that were irradiated at a fluence level of 3.93 10 cm These devices exhibited an ability to open, at voltages of up to 18 kV, without the need for the 2 m laser pulse.It was determined that this effect was due to the electron-hole recombination time being much faster than the time for hole trapping into the level.Two- pulse bursts were generated which demonstrated their ability to operate at repetition rates ranging from the hundreds of megahertz to 1 GHz.A BOSS switch that exhibits a selfopening effect may have substantial advantages over other irradiated photoconductive switches because the switching mechanism is a true bulk-conductivity effect.Most other shortlifetime photoconductive switches rely on lasers with photon energies greater than the bandgap of the semiconductor.These types of devices inherently operate using surface conduction rather than bulk conduction.A PSO generator, operating at high power levels, has been demonstrated for the first time.True frequency agility now has been made possible through the use of simple time-delay techniques.An added advantage of the PSO generator is that, by using a single mode-locked laser, the RF source will maintain phase coherence between the pulses within the burst and between the bursts themselves.
David C. Stoudt, Member, IEEE, Michael A. Richardson, and Frank E. Peterkin, Member, IEEE Abstract-The processes of persistent photoconductivity followed by photoquenching have been demonstrated at megawatt power levels in copper-compensated, silicon-doped, semi-insulating gallium arsenide.These processes allow a photoconductive switch to be developed that can be closed by the application of one laser pulse ( = 1.06 m) and opened by the application of a second laser pulse with a wavelength equal to twice that of the first laser ( = 2.13 m).This switch is called the bistable optically controlled semiconductor switch (BOSS).The opening phase of the BOSS requires a sufficient concentration of recombination centers (RC) in the material for opening to occur in the subnanosecond regime.These RC's are generated in the bulk GaAs material by fast-neutron irradiation ( 1 MeV).Neutron-irridated BOSS devices have been opened against a rising average electric field of about 36 kV/cm (18 kV) in a time less than 1 ns while operating at a repetition rate, within a two-pulse burst, of about 1 GHz.The ability to modify the frequency content of the electrical pulses, by varying the time separation, is demonstrated.Results demonstrating the operation of two BOSS devices imbedded in a frequency-agile RF source configuration are also discussed.

Fig. 1 .
Fig. 1.Experimental setup used for performing high-speed and high-power testing of BOSS devices.

Fig. 2 .
Fig. 2. Demonstration of pulse-width agility by increasing the time separation between turn-on and turn-off laser pulses from 400 ps to 1 ns.

Fig. 3 .
Fig. 3. Demonstration of the minimum electrical pulse width by showing the effect of decreasing the turn-on and turn-off laser pulse separation from 400 ps to 0 in 100 ps steps.The dashed curve corresponds to the shortest pulse in Fig. 2.

Fig. 4 .
Fig. 4. Demonstration of a 1 GHz (1 ns pulse separation) and a 290-MHz (3.4 ns pulse separation) repetition rate within a two-pulse burst at an applied voltage of 16 kV using sample B.

Fig. 7 .
Fig. 7. Demonstration of two bipolar pulses produced by a PSO generator for a 500 ps and a 2 ns delay between the 1 m laser pulses.

Fig. 9 .
Fig. 9. Potential output of a PSO generator constructed by replicating the 500 ps waveform shown in Fig. 7 at 10 ns intervals in a 4-pulse burst.

Fig. 11 .
Fig. 11.Potential output of a PSO generator constructed by replicating the 2.0 ns waveform shown in Fig. 7 at 10 ns intervals in a 4-pulse burst.