Effects of spot size on the operation mode of GaAs photoconductive semiconductor switch employing extrinsic photoconductivity

A coplanar GaAs PCSS was fabricated and used in our tests. The device structure is represented in figure 1(a). The dark electrical resistivity and the electron mobility of the SI-GaAs wafer are ~ 5×10⁷ Ω·cm and ~ 8500 cm²/(V·s), respectively. The wafer size is 14 mm×6 mm×1 mm. The Au/Ge/Ni (200 nm/100 nm/100 nm) multilayer alloys were deposited on the surface of the SI-GaAs wafer by electron-beam evaporation, and then good ohmic contact between the electrode and the wafer was formed through rapid thermal annealing. The horizontal gap between the contacts is 5 mm. To avoid the thermal damage, the conductive silver adhesive curing at the temperature below 100 °C, was adopted to weld the copper foil, linking the PCSS to the external circuit. Furthermore, the GaAs PCSS was immersed into the transparent transformer oil to prevent surface flashover.

Zoom In Zoom Out Reset image size

The test circuit shown in figure 1(b) is set up to carry out the on-state test of the GaAs PCSS. A ceramic capacitor of 140 pF is charged by an adjustable high-voltage (HV) DC power supply via a 10-MΩ resistor, which provides the bias voltage (V_b) for GaAs PCSS. The laser used for triggering the PCSS is a 1064-nm Nd: YAG nanosecond laser with the full width at half-maximum (FWHM) of ~ 2 ns. In the triggering light path, the laser passes through a beam expander and a square diaphragm in sequence, and then is divided into two beams by a beam splitter. One beam is used to trigger the GaAs PCSS, after passing through a filter, a convex lens and a mirror. The other is monitored by an energy meter to determine the real-time illuminating energy (O_E). In the experimental setup, the rectangular spot [in figure 1(c)] is adopted to illuminate the middle region of the GaAs PCSS. The spot size is measured by using an infrared sensor and a vernier caliper. The beam expander is added to make the spatial energy of the laser beam more uniform, to provide an approximately uniform spot for triggering the PCSS. The square diaphragm is used to adjust the spot size by controlling the uniform beam passing the aperture. Combining the square diaphragm and the displacement stage set below the PCSS, we can achieve the individual adjustment on the spot length (L_s) or width (W_s), under the case that O_E is fixed. In the test, once the PCSS biased at V_b is triggered, the 140-pF capacitor discharges to the equivalent load of 50 Ω through the PCSS, during which the outputted electric pulse is attenuated in sequence by a high-voltage attenuator (30 dB/2 GHz/50 Ω, Huaxiang/SHX MS2-8000) and a conventional coaxial attenuator (40 dB/4 GHz/50 Ω), then recorded by a 500-MHz oscilloscope with a 50-Ω channel. A high-voltage probe (Tektronix P6015A) is used to measure the charging voltage across the ceramic capacitor to determine the bias voltage applied to the PCSS.

Under the condition that the bias voltage, triggering energy and the W_s are fixed, the output electric pulses under different L_s are first tested. Figure 2 shows the typical output waveforms of L_s = 2 mm, 3 mm, and 4.5 mm, under the case of W_s = 2 mm, V_b = 5.1 kV and O_E = 103 μJ. It can be seen that the output waveforms cannot be simply classified as the outputs under the linear or nonlinear mode. On one hand, these waveforms present the characteristics of the linear mode, in which the rising edges are similar to that of the optical pulse. On the other hand, the apparent lock-on characteristics of the nonlinear mode also occur on the falling edge of these waveforms, to induce the wider pulse width than that of the optical pulse.

Zoom In Zoom Out Reset image size

More importantly, as the variation of spot length, the different evolution trends can be found in the output waveforms. When L_s increases from 2 mm to 4.5 mm, the falling edge of the output waveform uplifts gradually, forming a new peak (B), which will replace the previous peak (A) and dominate the increase of the output amplitude. The obvious double-peak characteristic can be observed under L_s = 3.0 mm, which indicates that the PCSS is operating in the transition from the linear mode to the nonlinear mode [20, 21], allowing us to analyze the evolution of the operation mode influenced by the variation of spot length. Here, the main part around peak A corresponds to the linear switching and the part around peak B corresponds to the nonlinear switching. From the evolution characteristics, we can find that the extension of the spot length along the current channel can enhance the development of nonlinear switching while restricting the development of linear switching.

Under the condition that the bias voltage, triggering energy and spot length are fixed, the output electric pulses under different spot widths are also tested. Figure 3 shows the typical output waveforms of W_s = 2 mm, 3 mm, and 4 mm, under the case of L_s = 1 mm, V_b = 4.5 kV and O_E = 100 μJ. The obvious double-peak characteristics can also be found in the output waveforms under W_s = 2 mm or 3 mm. In these outputs, the parts around the first peak (C) are identical to the waveform of the optical pulse, which can be determined as the linear output. Interestingly, different from the evolution trend shown in figure 2, as the increase of spot width, the linear part of the output waveforms increases significantly and dominates the increase of the output amplitude. Pushed up by the increased linear part, the nonlinear part (around the peak D) also presents a slight increase when W_s increases to 3 mm from 2 mm, but when W_s is further widened, the nonlinear part is weakened obviously due to the stronger linear switching.

Zoom In Zoom Out Reset image size

It can be determined that the expansion of the spot width can induce a more significant enhancement on the linear switching, during which the output amplitude of the linear part can be approximately expressed as [22], U_out = (qvR_L n_c V)/l, where q is the unit of electric charge, v is the saturated velocity of carrier, R_L is the load resistor, n_c is the optically generated excess carrier density, V is the effective volume of the current channel, and l is the distance between two electrodes. For the same switch, the values of l, q, and v are approximately given. Considering the optical absorption mechanism of GaAs at 1064 nm [23], a higher triggering energy can excite more photogenerated carriers, inducing a higher n_c V and U_out. Since the illuminating laser energy is kept as a constant in the experiment, we can infer that the increasing linear output with the increase of W_s is attributed to the more photogenerated carriers being assigned to participate in the linear switching.

Based on the above experimental results, it can be expected that a spot covering a large switching area can significantly enhance both linear and nonlinear switching. Thus, we compared the output waveforms under a small spot and a large spot approximately covering the whole switching area. The results are shown in figure 4. It can be seen that, compared to the small spot of 1×1 mm², the large spot of 4.5×4.5 mm² can induce a higher output. Here, a 1.9-fold improvement can be obtained, under V_b = 4.5 kV and O_E = 100 μJ. Correspondingly, the voltage conversion efficiency increases to 39.8% from 21%. In our test, further improvement in the output voltage can also be achieved if we adopt a higher bias voltage or a larger triggering energy. Moreover, it is easy to find, when the spot is broadened from 1×1 mm to 4.5×4.5 mm, both of the evolution characteristics caused by the size variations in L_s and W_s are integrated into the output waveforms, from which the significant increase can be observed on the linear part (along trail E) as well as the nonlinear part (along trail F). This evolution trend is consistent with our above expectation.

Zoom In Zoom Out Reset image size

It is worth noting that the above experimental results can also match the existing reports about the effect of the typical spots on the on-state performances of GaAs PCSS. According to our results, the quasi-linear mode reported in references [19] and [24] can be well demonstrated. In this mode, compared to a line-shaped spot across the electrodes, a spot covering the whole switching area is adopted to obtain the higher photocurrent and suppress the current filament appearing on the nonlinear switching, contributing to the higher photoelectric conversion efficiency and device lifetime. For the GaAs PCSS operating in the nonlinear mode, as reported in [17], compared to the long-striped spot perpendicular to the current channel, the focused spot with a narrower spot width can induce a higher nonlinear-mode output. Additionally, similar to the results presented in reference [16], our results demonstrate that the line-shaped spot across the electrodes and the spot covering the whole switching area are more effective for inducing a higher output than the point spot or partial line-shaped spot.

To gain insight into the physical mechanism of experimental results, the switching process of the GaAs PCSS operating in the transition from linear mode to nonlinear mode is first analyzed. Figure 5 depicts the physical diagrams of carrier dynamics during the switching. It is well known that the ultrafast switching of GaAs PCSS occurs on the filamentary current channel with a density of several MA/cm² and an upper limit for the diameters ranging from 50 to 300 μm [8]. Prior to the appearance of the current filament, the delay stage of the switching transient always exists in the PCSS, and it is mainly a process that produces electron-hole pairs (photogenerated carriers) [25]. As shown in figure 5(a), an initial distribution of carriers is produced upon the optical excitation, and is concentrated in an area covered by the spot. Under the force of the applied electric field, the carriers move separately toward the anode and cathode, forming the photogenerated current. The current has a linear relationship with the incident optical intensity, presenting the linear-switching characteristics.

Zoom In Zoom Out Reset image size

As the carriers are drifting toward the contacts to achieve the linear switching, the static high-field domains also gradually come into being near both the anode and cathode terminals, in which the peak field can reach over 200 kV/cm. The static domains play an essential role in injecting the ionization-produced carriers and generating the electron-hole plasma along a filamentary channel, as shown in figure 5(b). When the local electric field in the plasma region exceeds ~ 4 kV/cm, the classical Gunn-like domain instability starts to form the initial domains, then gradually converts the initial domains to the high-field, powerfully avalanche domains (> 250 kV/cm) due to the negative differential mobility (NDM) effect at the ultrahigh field. The formation and evolution of multiple powerfully avalanching domains inside the filamentary channel have been demonstrated to be the physical reason for the nonlinear switching of the avalanche GaAs PCSS [26‒28]. So, in figure 5(c), a co-existence state of the linear and nonlinear switching can be presented in the switching transient. Eventually, in figure 5(d), although the incident optical intensity is falling to weaken the linear switching, the nonlinear switching can still be continued until the applied bias voltage of the switch cannot sustain the voltage locking.

It is clear that the filamentary channel grows and develops on a fraction of the cross section of the total current channel. For the PCSS operating in the transitional state, its output waveforms will always integrate the characteristics of the linear and nonlinear modes, just as the double-peak phenomenon shown in figures 2 and 3. Correspondingly, the variation of these output waveforms can be attributed to the influence of the applied operating conditions on the nonlinear switching developing in the filamentary channel (active part) and the linear switching developing in the remaining channel (passive part) of the total current channel. Since the active and passive parts have a parallel relationship and the switching in the passive part always precedes that in the active part, the enhancement of linear switching can slow down the nonlinear switching. In contrast, the development of the nonlinear switching can restrain or cut off the linear switching. It is worth noting in our experiment, a thickened GaAs wafer with a thickness of 1 mm is adopted, and the extrinsic photoconductivity is employed by using the laser pulse with a wavelength of 1064 nm, together contributing to the total current channel with a larger cross section, which makes the transitional characteristic easier to be obtained.

It can be predicted that the increase of triggering energy mainly acts on the passive part to heighten the linear switching directly, although the triggering energy also accelerates the formation of avalanche domains in the active part to shorten the delay time of the nonlinear switching [26]. Different from the effect of the triggering energy, a higher bias voltage can induce the more drastic evolution of avalanche domains to enhance the avalanche multiplication [29, 30], thus strengthening the nonlinear switching in the active part, more significantly.

To verify the tight correlation between the output waveform and both the active and passive parts, under the condition that a spot approximately covering the whole switching area is used and the triggering energy is fixed at 100 μJ, the outputs are first tested at different bias voltages. As shown in figures 6(a)‒(c), influenced by the increasing bias voltages, the obvious uplift similar to that presented in figure 2 occurs on the falling edge of the output waveforms. In figure 6(c), the typical oscillation characteristic induced by the quenched-domain mode [23], has occurred on the tail of the output waveform. These evolution features mean that the increasing bias is boosting the switching in the active part toward the nonlinear mode. On the contrary, in figures 6(d)‒(f), influenced by the increasing triggering energy under V_b = 6.5 kV, the significant increase occurs on the parts induced by the linear switching, just like that shown in figure 3. The above results demonstrate the significant roles of the active and passive parts in determining the switching transient of GaAs PCSS.

Zoom In Zoom Out Reset image size

Based on the above physic of two equivalent channels, the effect of spot size on the operation mode can be explained as follows. Under the fixed illuminating energy, the increase of the spot width will increase the cross section of the total current channel, decreasing the concentration of photogenerated carriers on the total channel. Considering that the cross section of the active part is always constrained within the region with an equivalent diameter of 50‒300 μm (corresponding to the diameter of the current filament) for sustaining the high current density for the nonlinear switching, this concentration variation will directly reduce the photogenerated carriers distributing on the active part to weaken the nonlinear switching, while making more photogenerated carriers injected into the passive part to strengthen the linear switching. As for the spot length, the extension of this size can be considered mainly along both the active and passive parts, which does not significantly change the cross section of the active part or passive part, but strengthens the formation and evolution of the domains on the active part. Due to the parallel relationship between the active and passive parts, the "winner takes all" principle [28] plays an important role during the switching process. The strengthened switching in one channel will thus inevitably suppress the switching developing in the other channel.

The "two-channel" numerical model based on the theory of multiple avalanche domains has been demonstrated to be effective for interpreting the transitional switching transient of GaAs PCSS [20]. To further elucidate the inner domain dynamics influenced by the spot size, this predefined numerical model is introduced to simulate the GaAs PCSS. As shown in figure 7(a), the two equivalent channels of the GaAs PCSS are connected in parallel in the external circuit. The bias voltage is provided by a 0.14-nF capacitor C_s charged by a DC source via a 10-MΩ resistor R_s. At t = 50 ns, the 1064-nm laser pulse starts to trigger both parts, then the C_s discharges to the 50-Ω load. In the simulation, for higher computational efficiency, the coplanar structure of 30 μm×6 μm×6 μm is used to model the active part, and the bulk structure with a cross section of 30 μm×W_s and a length of 30 μm is used to model the passive part. The use of the scaling channels makes it difficult to directly compare the simulated outputs with the experimental outputs. However, the dimensions of the scaling structures are much larger than the domain dimension (submicron level to micron level), which can effectively support the comparative analysis of the domain evolutions under different spot sizes.

Zoom In Zoom Out Reset image size

Keeping the laser energy injecting the active and passive parts fixed, the typical currents through the load (I_L), the active part (I_a) and the passive part (I_p) are calculated, and shown in figure 7(b). It is noted that the I_p has a waveform similar to the optical pulse and the I_a presents the characteristic of the typical ultrafast switching after a significant delay, thus we can observe a superimposed characteristic of the linear and nonlinear switching on the I_L. As the L_s is extended from 9 μm to 18 μm, the significantly strengthened nonlinear part can be observed from the waveforms of I_L shown in figure 7(c), presenting an evolution trend similar to that shown in figure 2, and allowing us to compare the domain evolutions under different spot lengths.

Figure 7(d) shows the spatiotemporal field distributions at t = 50.7 ns. It can be seen that the peak fields of the static domains are 147 kV/cm near the anode and 334 kV/cm near the cathode under L_s = 9 μm, but these values reach 230 kV/cm and 401 kV/cm, respectively, when L_s is extended to 18 μm. In addition, an initial domain forming on the boundary of the static domain near the anode, has already been observed under L_s = 18 μm. The initial domains will further evolve into the avalanche domains due to the NDM effect, as shown in the field distribution at 51.4 ns. These results demonstrate that a larger L_s can contribute to the earlier formation of initial domains by increasing the peak field of static domains, then speeding up the formation of avalanche domains to bring forward and heighten the nonlinear switching.

As for the reason why a wider L_s can peak the amplitude of static domains, it is attributed to the extended electric field shielding area [26, 29]. The shielding area is formed by the separation of the photogenerated electrons and holes, which introduces a built-in electric field with a vector direction opposed to that of the applied bias field, and appears with the appearance of the static domains. As shown in figure 7(e), a wider triggering region significantly leads to a larger electric field shielding area. Relative to the original size (the horizontal gap between the contacts), the shielding area is extended to ~ 43% under L_s = 9 μm and ~ 62% under L_s = 18 μm. The extended shielding area will compress significantly the spatial electric field of the active part, peaking the static domains.

The opposite trend of domain evolution can be expected in the active part when the switching process is influenced by the increasing spot width. In figure 8(a), a high bias field of ~ 67 kV/cm and a large spot length of L_s = 26 μm are used to induce the higher nonlinear outputs relative to the linear outputs, Thus, the effect of the increasing spot width on I_L can be observed just from the switching stage of the output waveforms. Here, under the condition that the size of the active part and the illuminating total energy on both parts are fixed, the linear output under W_s = 1.8 mm presents a higher amplitude of 1.0 A than the amplitude of 0.5 A under W_s = 0.6 mm, or the amplitude of 0.8 A under W_s = 1.2 mm. Correspondingly, in figure 8(b), the apparent initial domain under W_s = 0.6 mm can be observed earlier than that under W_s = 1.8 mm, and the domain evolution under W_s = 0.6 mm is also faster than that under W_s = 1.2 mm. The larger width of the initial domain than that in figure 7(d) is attributed to the use of a larger spot length in the simulation. Because the domain evolution closely correlates with the carrier concentration in the current channel [29, 31], the slowing formation or evolution of domains also directly signifies the reduced photogenerated carrier concentration on the active part, when W_s is extended from 0.6 mm to 1.8 mm. Evidently, the above results demonstrate the explanation of the effect of the spot width on the operation mode.

Zoom In Zoom Out Reset image size

Based on the characteristics of domain evolutions, the effect of spot size on the on-state performances of GaAs PCSS can be further discussed. Firstly, accelerating the formation and maturation of avalanche domains has been considered an effective strategy to improve the triggering efficiency of the PCSS [25]. That indicates, for the avalanche GaAs PCSS triggered by the fixed illuminating energy, the lower jitter and on-state resistance can be obtained by adopting a line-shaped spot along the filament. Secondly, in our study, the increasing peak field of static domains can induce more ionization-produced carriers, then reducing the demand for the photogenerated carriers before the ultrafast switching. So, for the actual avalanche GaAs PCSS, a reduced laser energy requirement can be achieved by increasing the spot length along the filament channel. Thirdly, considering that the increased spot width can broaden the total current channel to slow the domain evolution, we can conclude that a thicker (~ mm level) GaAs PCSS employing the extrinsic photoconductivity is more beneficial for strengthening the quasi-linear mode, to achieve the higher output power and longer device lifetime.