Experimental investigations on the trigger characteristics of a cold-cathode plasma discharge switch with grid voltage control

DC power transmission technology is widely applied in long-distance power transmission, asynchronous grid interconnection, and renewable energy integration. As DC grid technology develops, the demand for fully controlled switching devices in power systems has grown urgent.¹ Fully controlled semiconductor switches, such as insulated gate bipolar transistors (IGBTs) and integrated gate-commutated thyristors (IGCTs), are becoming key components of DC grid systems due to their capability for precise and rapid switching control.^2,3 However, as the voltage level increases, numerous semiconductor switching devices must be connected in series to withstand the system voltage, which significantly raises manufacturing costs and complicates the control of semiconductor devices.^4,5 Gas discharge has been widely applied in the field of switching devices.^6–9 Exploring a fully controlled gas switch with high voltage withstand capability presents a potential solution to replace fully controllable power semiconductor devices, which can be applied in direct current (DC) power transmission.¹⁰ 

These cylindrical cold-cathode plasma discharge switches have an anode voltage withstand capability of 20–100 kV, but they exhibit a relatively high-conduction voltage drop of 450–500V (compared to semiconductor fully controlled devices).¹⁹ This results in significant heat generation during long-term operation in electrical power systems. Additionally, their long-term stable average current-carrying capacity is only 0.5–3 A,¹⁹ which still does not meet the current-carrying requirements for electrical system applications.

General electric (GE) conducted research types on planar-shaped cold-cathode plasma discharge switches and proposed their potential applications in voltage source converter-based high-voltage direct current (HVDC) transmission systems in 2016.¹⁰ Such plasma switches could replace the power electronic devices in voltage source converters. There have been reports on the prediction of gas high-voltage insulation properties,²⁰ magnetized cold-cathode plasma discharge modes,^21–23 and plasma sheath theory analysis.^24,25 However, there is limited published research on the triggering, current conduction, and interruption characteristics of the cold-cathode plasma discharge switch itself, which presents challenges for experimental investigations of such switches.

We developed a planar-shaped cold-cathode plasma discharge switch device and studied its trigger characteristics. The structure of the paper is as follows: Sec. II introduces the experimental platform and the specific parameters of the switch device; Sec. III presents the basic principle of operation with an illustrative pulse; Sec. IV experimentally explores the effects of source plasma discharge current, source grid–control grid distance, and different anode voltages on the switch trigger process; Sec. V provides an analysis and discussion of the phenomena observed in the experiments; and Sec. VI provides the conclusions.

In this equation, m_e represents the electron mass, e is the electron charge, and v, E, and B are the velocity, electric field, and magnetic field vectors, respectively. The Lorentz force causes the electron to gyrate. This significantly increases the path length of the electron and the probability of collision ionization; thus, the source plasma can be generated between the source grid and cathode. According to our previous optical diagnostic results,²⁷ the source plasma is primarily concentrated in the region where the annular magnetic field is strongest (20 mm < r < 25 mm). Both the source and control grids are made of thin molybdenum plates, with a grid thickness of 0.6 mm. The grids have a mesh-like structure with 0.6 mm apertures, as shown in Fig. 3(c).

The physical image of the experimental chamber is shown in Fig. 4(a). All the electrodes are placed inside a sealed chamber, with a pressure gauge providing real-time readings of the internal pressure. Helium was used as the gas medium in the subsequent experiments (in order to minimize the sputtering material loss rate) with the working pressure maintained at 6.5 Pa. Helium have low atomic mass and large ionization energy, compared with heavy metal atom.²³ Ions with high ionization energies are more likely to release electrons upon striking the cathode, contributing to current generation. In contrast, metal ions with lower ionization potentials are more likely to sputter the cathode due to their greater mass. To ensure gas purity, the chamber pressure was first reduced to below 0.1 Pa before introducing helium, after which the pressure was again lowered to below 0.1 Pa. This process was repeated three times. It should be noted that the experimental chamber has a relatively large volume and is exposed to an atmospheric environment externally. Once the vacuum pump stops operating, external air will enter the chamber, affecting the helium purity. To mitigate this issue, the vacuum pump remains working throughout the experiment, while helium gas is continuously introduced to maintain an operating pressure of 6.5 Pa. Figure 4(b) shows a stable source plasma discharge image, where the plasma's luminous region exhibits a ring-shaped distribution due to the magnetic field confinement. Figure 4(c) displays the spectrum image of the source plasma, with the prominent He spectrum. The spectral data were obtained using a spectrometer (Andor SR500i) with a 300 l/mm grating and an ICCD camera (iStar DH334T). The detailed optical setup can be found in our previous work.²⁷ Within the wavelength range of ∼395–∼700 nm, no significant spectral lines of N or O were observed (such as N I 575, N I 648, O I 615, and O I 645 nm), which would indicate leakage from external air.

We used a pre-charged RC circuit (100 Ω–300 μF) to supply the anode voltage and the discharge current. The 100 Ω ballast resistor is used to limit the conducting current for the experimental device protection. A DC power supply provided the energy to generate the source plasma, connected to the source grid via a 25 kΩ resistor. The control grid was grounded through a 50 kΩ resistor to keep the control grid voltage at 0 V. The positive +750 V bias voltage and negative −200 V bias voltage applied to the control grid were supplied by two pre-charged RC circuits, with thyristors controlling the conduction of the positively and negatively charged circuits. The 7 and 300 μF capacitors are each connected to a 7 Ω resistor to limit the current flowing through the control grid during the device's initiation and interruption phases. In the subsequent experiments, the conduction current, anode voltage, source grid voltage, and control grid voltage were recorded, as shown in Fig. 5.

With the electrode arrangement shown in Fig. 6, the current conduction and interruption test waveforms are presented in Fig. 7. Before applying the positive voltage to the control grid, a 1.9 kV voltage is applied to the anode. Then, the DC power source is switched on. The source plasma is stably discharging before the trigger process, with the source grid voltage maintained at 280 V. The control grid voltage keeps at 0 V, keeping the device in an off-state. After 3 s, SCR1 is triggered to apply a positive bias voltage to the control grid. It is important to note that in the waveform, t = 0 ms corresponds to the moment when the control grid voltage reaches 750 V. The detailed waveform of the trigger process is shown in Fig. 7(b). Under the influence of the positive control grid voltage, the source plasma passes through the grid apertures and reaches the anode. The device begins to conduct, and the anode current starts to rise from zero. At t = 0.067 ms, the control grid and source grid voltages drop, indicating the device is fully triggered and a plasma discharge channel of “anode–control grid–source grid–cathode” is formed inside the device. 1 ms after the device begins conducting, a −200 V negative voltage is applied to the control grid, forming a plasma sheath within the grid apertures. The current interruption is completed. The anode voltage recovers to 1.9 kV, and the source plasma returns to a stable discharge state, with the source grid voltage restoring to 270 V.

It is important to note that the source plasma plays a critical role in the trigger process of the device. Figure 8 shows the results under the same experimental conditions without the presence of source plasma. After applying the control grid conduction signal, the anode voltage remains at 1.9 kV, and the measured current is 0 A. This demonstrates that the trigger process in the cold-cathode plasma discharge switch relies on the charged particles provided by the source plasma.

In our experimental setup, the control grid was grounded through a 50 kΩ resistor (keeping the control grid voltage at 0 V before applying the +750 V trigger signal), effectively shielding the source plasma from the influence of the anode voltage. Figure 9 shows the experimental results with the control grid floating (the 50kΩ grounding resistor removed). When the device is in the off-state (t < 0 ms in Fig. 9), with an anode voltage of 2 kV and a generated source plasma, the floating control grid voltage rises to approximately 300 V under the influence of the anode electric field. Then, the device spontaneously switched on without applying the +750 V control grid trigger signal. During the current conduction process, the anode, control grid, and source grid voltages are approximately 100, 40, and 15 V, respectively. Therefore, grounding the control grid through a large resistor is necessary to maintain the experimental device's stability and controllability.

Based on the experimental platform introduced in Sec. II, an experimental study was conducted to investigate the effects of different source plasma discharge currents, “source grid–control grid” distances, and anode voltages on the trigger performance of the cold-cathode plasma discharge switch.

The source grid plasma provides a sufficient supply of ionized particles for the switch's trigger process. The experimental results with source plasma discharge currents of 11, 23.5, and 33 mA are shown in Fig. 10. Apart from the differences in source grid discharge currents, the experimental conditions are consistent with those described in Sec. III.

The trigger times corresponding to discharge currents of 11, 23.5, and 33 mA are 923, 631, and 67 μs, respectively. The trigger time decreases significantly as the discharge current increases. Based on previous optical diagnostic results,²⁷ ionized particles are primarily concentrated in the region where the annular magnetic field near the cathode surface is strongest. An increase in discharge current effectively raises the electron density and electron temperature.

The experimental results for the 33 mA condition have been detailed in Sec. II. It is worth noting that under the 11 and 23.5 mA conditions, the control grid voltage (black curve in the figure) shows a gradual increase before the switch is fully triggered, as illustrated in Figs. 10(a) and 10(b). This indicates that the control grid voltage was influenced by the anode voltage. A portion of the plasma that crossed the control grid did not immediately form a discharge path across the entire device. Instead, a weak discharge channel was formed within the anode–control grid gap. The switch only achieved fully conduction once the amount of source plasma crossing the control grid increased.

The investigated plasma switch contains three electrode gaps. Source plasma is generated within the “source grid–cathode” gap and moves through the “source grid (SG)–control grid (CG)” gap. The charged particle that passes through the control grid apertures is rapidly accelerated toward the anode by the strong electric field in the “anode–control grid” gap. The trigger time is determined by the time it takes for the source plasma to reach the anode. Therefore, the electrode arrangement within the device must be carefully designed, ensuring that the gap size is minimized while maintaining sufficient insulation strength.

In subsequent experiments, we reduced the SG-CG gap distance from 30 to 15 mm, while keeping the SG–cathode and anode–CG distances unchanged. The spatial distribution of the electrodes is shown in Fig. 11. The trigger characteristic test results are presented in Fig. 12.

By reducing the SG-CG gap to 15 mm, the trigger time decreased to 8.1 μs, an 88% reduction compared to the 67 μs trigger time with a 30 mm gap. This significant improvement demonstrates that adjusting the SG-CG distance can greatly enhance the trigger speed of the device. The faster trigger time resulting from the reduced gap distance is primarily due to the shorter travel path for the charged particles after crossing the source grid, allowing the discharge channel to form more quickly. A large amount of charged particles, after crossing the control grid, is immediately influenced by the anode voltage. This allows the formation of a stable and fully conductive discharge channel in a much shorter time. The shortened path reduces the delay in charged particle motion, which is key to the faster trigger of the switch.

Under the reduced electrode gap dimensions shown in Fig. 11, experimental results for anode voltages of 2.4 and 3.0 kV are illustrated in Fig. 13, with the 2.0 kV results detailed in Fig. 12 from Sec. IV B. Since the control grid is grounded through a 50 kΩ resistor, its voltage remains at 0 V before the pulse signal is applied, effectively shielding the source plasma from the influence of the anode voltage. The source plasma maintains stable discharge at different anode voltages. After applying the 750 V bias voltage to the control grid, the larger electric field between the anode and the control grid can further accelerate the charged particle motion passing through the control grid holes. The trigger times corresponding to anode voltages of 2.0, 2.4, and 3.0 kV are 8.1, 6.5, and 4.4 μs, respectively.

Figure 13(c) presents detailed current waveforms during the trigger process under three different anode voltages. The trigger moment at t = 0 μs and the complete triggered time (when the source grid voltage and control grid voltage drop, as illustrated in Fig. 7) are marked with dashed lines. Notably, under the 2.0 kV condition, from the moment the control grid trigger signal is applied, to the point where a stable conductive channel is fully established, the current gradually increases (shown by the orange curve). However, under the 2.4 and 3.0 kV conditions, a spike in the current at t = 5 μs is observed during the rise. This may be attributed to the stronger anode electric field providing more energy to the charged particles, which accelerates the ionization process, leading to a momentary increase in plasma conductivity. While ensuring the insulation performance of the device, a higher anode voltage may trigger more intense ionization, thereby enhancing the current-carrying capacity of the device.

Figure 14 shows the net current to the control grid during the whole working process (initiation, conduction, and interruption). The experimental setup for Fig. 14 is identical to that in Fig. 5. After the device has conducted current for 0.22 ms, a negative control grid voltage is applied to interrupt the current. A positive control grid current indicates that current is injected from an external control circuit into the device via the control grid.

In the initial conduction phase of the device, the 7 μF positive voltage capacitor injects a pulse current into the device, reaching a peak value of up to 139 A. Under the operating anode voltage of 3 kV, the conduction current (the measured anode current) is limited to ∼30 A by the 100 Ω ballast resistor. After applying the negative voltage pulse at t = 0.22 ms, approximately 8 A of current is transferred through the control grid to the negatively charged 300 μF capacitor.

Replacing the positively charged 7 μF capacitor with a 0.5 μF capacitor, while maintaining the pre-charge voltage at 750 V, significantly reduces the peak and duration of the control grid injection current, as shown in Fig. 15. The device can continue conducting even after the current injection process concludes. In addition, the experimental results in Fig. 9 also indicate that a stable discharge path can be generated even without the injected control grid pulse current. Hence, the capacitance of the positively charged capacitor can be appropriately reduced to decrease the injected pulse current on the condition. During the current interruption phase, part of the current can transfer to the negatively charged capacitor branch through the control grid. This undoubtedly reduces the burden for the plasma sheath (formed within the control grid apertures) to switch off the plasma conduction path. Increasing the capacitance of the negatively charged capacitor may be an option to enhance the device's interruption capability.

When the device is in the off-state, only the source plasma maintains a stable discharge within the system. This magnetically enhanced cold-cathode source plasma voltage drop ranges between 250 and 300 V. However, when a positive voltage is applied to the control grid and the device is fully conducting, the anode voltage decreases to approximately 100 V (as shown in Fig. 9). This voltage drop seems to be too low for the magnetically enhanced cold-cathode operation, which typically requires high-energy ions to be accelerated in the cathode fall region to produce a sufficient number of electrons to sustain the plasma.²⁸ Secondary electron emission may not be the only mechanism to release electrons from the cathode in our experimental conditions.

The cold-cathode plasma discharge switch in our study operates in a helium environment. However, when the device is not in operation (the vacuum pumping and helium injection are stopped), external air enters the internal chamber, which leads to the formation of an oxide layer on the surface of the internal brass electrodes. Under these conditions, thin-film field emission could become an important mechanism for plasma conduction. The positive ions adhere to the oxide film and create a strong electric field at the surface of the underlying metal, which will enhance the field emission of electrons from the cathode surface.^29,30 The cathode surface with removed oxide after multiple current conduction experiments is presented in Fig. 16. The oxide layer could be a possible explanation for the low-voltage operation. Similar experimental phenomena have also been reported by Sommerer et al.²³ They found that an oxide film on a molybdenum cathode is a key factor for the magnetized cold-cathode plasma in helium operating in a discharge mode with an 80 V plasma voltage. When the oxide film is removed, the plasma voltage increases to about 325 V. The low plasma voltages are correlated with high-speed videos showing a highly localized cathode spot. In our future experiments, it will still be necessary to verify through high-speed imaging whether similar cathode spot features can be observed, in order to further confirm the role of the cathode surface oxide film.

In our experimental results, the fastest trigger time is still relatively high (> 4 us) for practical applications. Therefore, it is crucial to reduce the device size and enhance its high-voltage withstand abilities. The cold-cathode plasma switches operate near the left side of the Paschen curve. According to the findings of Xu et al.,²⁰ when the pd value is between 0.4 and 0.6 Torr·cm, helium as a dielectric medium can withstand voltages ranging from 10 to 100 kV. To achieve rapid closure, it is preferable to operate the switch at the highest possible pressures, in order to achieve minimized electrode spacing. The cylindrical-shaped CROSSATRON switch,¹⁹ produced by Hughes Research Laboratories, can achieve a maximum closing speed of up to 3 × 10⁹–6 × 10⁹ A/s under an anode voltage range of 20–100 kV. In recent years, plasma generated in nanogaps has been applied in the field of switching, enabling ultrafast switching (picosecond level) of electric signals.^31,32 Currently, our experimental chamber structure remains relatively simple. However, further triggering time optimization of the planar-shaped cold-cathode plasma discharge switch device is achievable and has significant potential for improvement. This will require advancements in the manufacturing techniques of the device.

For cold-cathode plasma discharge switches, lifespan is an important indicator of operational performance. We used helium (with low atomic mass) as the gas medium to reduce electrode sputtering erosion. Brass electrodes were used in the above experimental tests. Refractory materials with high cohesive energy and strong sputtering resistance (such as W and Mo) are suitable candidates for cathode material selection. In the analysis in Sec. V B, we found that the formation of an oxide film on the cathode surface contributes to a relatively low plasma voltage drop, which can lead to a lower incoming ion energy. Materials prone to forming oxide films on the surface (such as Al, Fe, and Cr), should also be considered as potential options.

The purpose of our research on such cold-cathode plasma discharge switches is to enable their applications in HVDC transmission systems. General electric proposed that these switches could replace semiconductor power electronic devices in voltage source converters.¹⁰ However, the conduction voltage drop of such switches is still relatively high, and the current flowing through the switch during conduction may lead to significant conduction losses.

Another crucial component in HVDC systems is the HVDC circuit breaker. During normal operation, the circuit breaker carries the rated system current. When a fault occurs, it needs to quickly interrupt the fault current to ensure system safety. Figure 17 illustrates a DC interruption scheme. Due to the requirement to withstand the system's high voltage, a large number of power semiconductor devices need to be connected in series. During normal operation, the high-speed vacuum switch remains closed, carrying the system current. Once a fault occurs, the system fault current rapidly increases. The high-speed vacuum switch opens first, and an internal vacuum arc establishes an arc voltage. Under the influence of this arc voltage, the fault current is transferred to a branch consisting of multiple power semiconductor switches connected in series, where the fault current is interrupted by the semiconductor switches.

We aim for the cold-cathode plasma discharge switch to replace the large number of series-connected semiconductor power devices. However, this imposes high demands on the switch's current interrupting capability (in the kA range). Additionally, the conduction voltage drop of the cold-cathode plasma discharge switch remains relatively high (maintaining ∼100V in the aforementioned experiments). The voltage of the vacuum arc needs to be higher than the conduction voltage drops of the plasma switch in order to transfer the fault current. To address this issue, increasing the number of vacuum circuit breakers and applying an external transverse magnetic field to enhance the vacuum arc voltage could be considered,³⁴ as shown in Fig. 18.

Based on the experimental platform, the trigger characteristics of the cold-cathode plasma discharge switch were investigated. The experimental results demonstrate that optimizing the source plasma discharge current, electrode spacing, and anode voltage can significantly enhance the trigger performance of cold-cathode plasma discharge switches. The following conclusions can be drawn:

1. The source plasma provides a sufficient supply of charged particles, which is crucial for the switch's trigger. Before triggering, the source plasma maintains stable discharge between the source grid and the cathode. Higher source plasma discharge currents lead to a greater density of charged particles and enhancing ionization. As the discharge current increases, the switch-on time after applying the trigger voltage can be significantly shortened. Since the source plasma is mainly concentrated in the region where the magnetic field is strongest near the surface of the cathode, future optimizations could focus on increasing the radius of the magnetic field distribution and enhancing the peak magnetic field strength. By expanding the area of source plasma distribution and increasing the number of charged particles, the trigger time can be further reduced.

2. This type of plasma switch relies on the control grid voltage to regulate the plasma. The trigger process involves the charged particle motion from the source plasma toward the anode and the formation of a stable conductive channel. Therefore, the switch's trigger time is largely determined by the charged particle motion time. Reducing the SG-CG gap distance can significantly shorten the switch's trigger time. It is important to note that reducing the overall size of the device imposes higher demands on its insulation performance and gas sealing capabilities.

3. Increasing the anode voltage effectively enhances the electric field strength across the “anode–control grid” gap, accelerating the charged particle motion through the control grid apertures and into the “anode–control grid” region. A stronger anode electric field provides higher energy to the charged particles, promoting the ionization process and improving plasma conductivity. Within the limits of the device's insulation performance, a higher anode voltage could induce a more intense ionization process, enhancing the current-carrying capability of the device. However, it is crucial to shield the source grid plasma from the anode electric field to prevent breakdowns, especially when enduring higher anode voltages.

4. To apply such switches in the electrical power systems field, it is necessary to further reduce the device's conduction voltage drop. The experimental results show that the effect of the oxide film on the cathode surface could be an explanation for the low-voltage operation (∼100V). Furthermore, studies on the influence of cathode material on plasma voltage drop are needed. In addition, enhancing the current interruption capability of this switch is crucial. Relevant studies may involve plasma sheath and the physics of space–charge interactions, with related theoretical and experimental research yet to be conducted. In future studies, it is also important to improve the gas purity to remove any risks that could be applied to the experimental results.

This work was supported by the National Natural Science Foundation of China, Award/Contract No. U2241260, National Natural Science Foundation of China, Award/Contract No. U22B20121, National Natural Science Foundation of China, Award/Contract No. 52077170, and National Natural Science Foundation of China, Award/Contract No. 52025074.

The authors have no conflicts to disclose.

Yushi Zhang: Investigation (equal); Writing – original draft (lead). Hao Sun: Funding acquisition (equal); Writing – review & editing (lead). Tianxiao Liu: Investigation (equal). Yi Wu: Funding acquisition (equal). Chunping Niu: Funding acquisition (equal).

The data that support the findings of this study are available from the corresponding author upon reasonable request.