Effect of Cosmic Rays on the Failure Rate of Flexible Direct Current Converter Valves in High-Altitude Environment

Abstract

Aiming at the significant needs of flexible DC converter valve applications in high-altitude areas, we investigate the effect of atmospheric neutrons on the failure rate of key core power devices in three kinds of converter valves, namely IGBTs, thyristors, and diodes. The safe working voltage boundary of the devices is obtained, and the failure rate caused by atmospheric neutrons in the real working environment of the power devices is calculated according to the results of the ground-accelerated irradiation test of atmospheric neutrons and the atmospheric neutron environment under the actual working conditions. The test results show that the failure rate of IGBTs, thyristors, and diodes caused by atmospheric neutrons is greatly affected by the blocking voltage, and the larger the blocking voltage is, the higher the failure rate of the device is. The research results can provide a basis for the design of the operating voltage of key core power devices of flexible DC converter valves and guide the evaluation of the failure rate and engineering design of the electronic system of DC ultra-high-voltage power transmission and transformation converter stations in the high-altitude environment of the Qinghai-Tibet Plateau.

1. Introduction

In alignment with the national dual-carbon strategy, energy consumption in China is set to rapidly transition towards a green, low-carbon model. This shift will involve the large-scale harnessing of clean energy sources such as wind, solar, and hydroelectric power, particularly in the western regions of the country. Flexible direct current (DC) transmission technology is expected to play a crucial role in the efficient transmission of this new energy. However, the western region of China presents unique challenges due to its high altitude. Notably, there are currently no existing flexible DC transmission projects completed at altitudes above 2000 m, either domestically or internationally. Furthermore, there is a lack of data on cosmic ray atmospheric neutron tests conducted on the power devices of flexible DC converter valves [1,2,3,4,5,6,7]. The forthcoming results from these tests will be instrumental in advancing both domestic and international research in this field.

The flexible DC converter valve is the central component of the DC transmission system, utilizing key core power devices such as IGBT, thyristor, and diode valves to achieve AC/DC conversion. However, increasing altitude and latitude result in a rise in neutron flux, posing significant challenges. For instance, China’s western region, which is generally at a higher altitude, experiences an atmospheric neutron flux over 20 times that of Guangzhou. Operating power devices at high altitudes using the same standards as in the plains may exacerbate failure problems due to cosmic rays and atmospheric neutrons.

In areas of high cosmic ray and neutron radiation, the phenomenon of degradation or failure of individual electronic components or material particles due to impact by energetic particles (e.g., neutrons, protons, etc.) is known as single-event burnout (SEB). This phenomenon is particularly important in power electronic equipment at high altitudes, where the neutron flux increases significantly, which can lead to reliability and safety issues, directly impacting the safety and stability of the DC transmission system. To address this issue, accelerated irradiation tests can be conducted in a laboratory setting, where power devices are exposed to neutron flux for periods ranging from a few hours to decades [8,9,10,11]. The results of these tests can directly inform engineering design, making it highly efficient and effective in ensuring the reliability of power devices in high-altitude environments.

In 2013, Merlin et al. used the LiCAF neutron detector for high-latitude atmospheric neutron measurements [12], and in 2022 Akin et al. conducted a study of SiC power device failure due to atmospheric neutron irradiation, and in their results, the devices did not fail significantly at neutron fluxes less than 10¹⁰ n/cm² [13]. This research addresses the critical need for versatile DC converter valve implementation in high-altitude regions by investigating the impact of atmospheric neutrons on the failure frequency of essential power core devices in three types of converter valves: IGBTs, thyristors, and diodes. Using data from atmospheric neutron ground-accelerated irradiation tests and atmospheric neutron environment assessments, this study calculates the failure rate of these devices under actual working conditions. The safe working voltage limit for each device is determined, revealing that the failure rate of IGBTs and thyristors is significantly influenced by the blocking voltage [14,15,16]. Specifically, the failure rate increases proportionately with an increase in blocking voltage. These findings provide a foundation for establishing optimal operating voltages for critical power devices in flexible DC converter valves. Furthermore, they offer valuable insights for assessing failure rates and guiding the engineering design of electronic systems in DC ultra-high-voltage power transmission and transformation converter stations, particularly in the high-altitude environment of the Qinghai-Tibet Plateau.

2. Materials and Methods

The irradiation tests were carried out on the Atmospheric Neutron Irradiation Spectrometer (ANIS) test bed based on the China Scattered Neutron Source (CSNS); the altitude of Dongguan, where the platform is located, is about 0–50 m above sea level. The neutron energy range of the atmospheric neutron irradiation spectrometer test platform covers meV~GeV, and the shape of the neutron energy spectrum in its samples is close to that of the natural atmospheric neutron energy spectrum, but the neutron flux is increased by a factor of 1.65 × 10⁸ times compared with that at sea level so that it can accelerate the equivalent simulation of the effect of the natural atmospheric neutrons on the power devices. The comparison of the neutron energy spectrum of the test platform with that of the natural atmosphere is shown in Figure 1. The main parameters of the irradiated test source are shown in

Table 1. Atmospheric neutron irradiation test platform main parameter.

Parameter Name	Parameter Index
Energy range	meV~GeV
Neutron irradiation spot area	20 cm × 15 cm
Neutron flux at sample (Energy above 10 MeV)	4.7 × 105/n/cm2/s

. A list of the main equipment for the experiment can be found in

Table 2. List of major equipment and its model.

Equipment Name	Equipment Model
Atmospheric Neutron Irradiation Spectrometer	ANIS
High-voltage DC power supplies	DSP
High- and low-temperature box	NT125-55AS

, and the experimental site plan is shown in Figure 2.

In the process of the atmospheric neutron acceleration and irradiation test, it is necessary to bias the power device in the high-voltage blocking state and monitor its off-state current in real-time, and its test principle is shown in Figure 3. In the figure, from left to right, the scattering neutron source, the sample to be tested and the fixture, the multi-channel test platform, and the high-voltage DC power supply are shown in order. During the irradiation test, the samples and fixtures to be tested are placed in the irradiation room, and the test equipment (including the high-voltage DC power supply, the multi-channel test platform, the computer, etc.) and the testers are in the test room. The high-voltage power supply is connected to the multi-channel test platform and then connected to the sample to be tested through a cable of more than 30 m to supply power to it.

To enhance experimental efficiency, accelerated testing involves irradiating multiple devices simultaneously. During the test, all devices are connected in parallel and powered by a high-voltage source. If a device experiences a single-event burnout, it loses its blocking capability and exhibits a short-circuit characteristic. This failure can prevent the other parallel-connected devices from being properly energized. Therefore, it is necessary to disconnect the failed device branch from the power supply. In order to achieve the above purpose, a multi-channel test platform is used to realize parallel testing of multiple samples. The multi-channel test platform mainly consists of a control circuit board, current sensors, and relays. Each sample to be tested is connected to the same high-voltage power supply through the sensor and relay and then irradiated at the same time [17,18,19]. When a sample fails, the sensor detects a change in the corresponding current and sends a signal to the control circuit, which then controls the relay to disconnect from the high-voltage power supply, thus ensuring that the voltage is only applied to the non-failed sample.

During irradiation, a sample is considered to have failed if there is an instantaneous surge in off-state current and a loss of blocking capability. One failure event is recorded if a sudden surge in off-state current of 100 mA is detected during irradiation. After calculating the neutron flux at 29.79° N, 97.71° E, and 4000 m above sea level, the neutron flux above 10 MeV that may lead to the failure of the power device is 116 n/cm²/h. According to the requirement that the failure rate of the thyristor device is not higher than 50 FIT, i.e., 50 samples are failing in 1 × 10⁹ device hours, 5 chips are selected and irradiated at the same time. Galvanic irradiation, and then the total irradiation of each test chip, is approximately as follows:

116 n/cm²/h × 1 × 10⁹ h/(5 × 50) = 4.64 × 10⁸ n/cm²

Therefore, when the thyristor chip is used to carry out the test, the maximum neutron injection during the test is set at 4.64 × 10⁸ n/cm², and the irradiation test can be stopped even if there is no device failure when the cumulative neutron injection during the test reaches 4.64 × 10⁸ n/cm². The atmospheric neutron accelerated irradiation test is carried out under selected temperature and voltage conditions, and the test obtains the failure rate at different voltages and temperatures. Since the average temperature in Tibet is 3–17 °C, we conducted the experiment at a lower temperature of 5 °C to restore the temperature of the real application scenario; the irradiation test is carried out at a minimum of 5 different voltages for each sample, and the neutron irradiation injection and bias voltage chosen for the experiment are the result of a combination of experimental conditions and recommendations from the literature [13].

A matching fixture is designed for the test sample. The fixtures are used to hold the samples to be tested and to ensure that the neutrons are incident vertically on the samples. The fixture is also used to lead out the charging port of each sample. Among them, the IGBT chip and the diode chip share a common set of fixtures. The cross-sectional size of all the samples mounted in the fixture does not exceed the size of the neutron beam spot for the test (20 cm × 15 cm). Through laser positioning, it is ensured that the neutron beam spot can cover all the samples to be tested. Throughout the test, the samples to be tested were placed in a high- and low-temperature chamber, which was set at a constant temperature of 5 °C. The chamber had a 40 cm × 40 cm space for the sample to be tested. A 40 cm × 40 cm glass viewing port is left in the chamber, through which the neutrons irradiate the sample to be tested.

Record the number of samples r for which failure occurred during the irradiation test and record the total effective sample injection T_SUM (neutron injection with energy above 10 MeV):

$$T_{SUM}=\sum n_i$$

(1)

where i is the ith test sample, n_i is the neutron injection at the time of failure of the ith sample, or n_i is the cumulative neutron injection at the time of stopping the beam if no failure of this sample occurs. Thus, the accelerated test failure rate λ_ACC is as follows:

$${\lambda }_{ACC}=r/T_{SUM}$$

(2)

Based on the number of sample failures r recorded in the accelerated irradiation test, the corresponding confidence interval for the failure rate can be calculated as follows:

$$\left(2T_{SUM}{)}^{-1}{X}_{\alpha /2}^2\right(2r)<{\lambda }_{ACC}<(2T_{SUM}{)}^{-1}{X}_{1-\alpha /2}^2\left(2r\right)$$

(3)

where X² is the chi-square function with a degree of freedom of 2r; T_SUM is the total neutron injection of the effective sample. The confidence intervals are all calculated in this report according to the 95% confidence level, which corresponds to the 95% confidence level of α = 5%. The actual device failure rate λ_ACC is obtained by multiplying the failure rate λ_ACR obtained in the test by the average neutron flux Φ_n (energy above 10 MeV) in the actual environment:

$${\lambda }_{ACR}={\lambda }_{ACC}\times {\Phi }_n$$

(4)

Based on the results of the accelerated irradiation test, the device failure rates corresponding to 4000 m above sea level, 2000 m above sea level, and sea level are calculated in this work. In particular, the natural atmospheric neutron flux (Φn) above 10 MeV varies significantly with altitude. At an altitude of 4000 m, the flux is calculated to be 116 n/cm²/s. This value decreases to 32 n/cm²/s at an altitude of 2000 m. At sea level, the neutron flux further diminishes to 6.2 n/cm²/s. These calculations highlight the inverse relationship between altitude and neutron flux, with higher altitudes experiencing greater neutron activity.

3. Results and Discussions

Figure 3, showing the results for the IGBT chips, depicts the irradiation tests that were carried out under five voltage conditions, 2000 V, 2050 V, 2100 V, 2200 V, and 2300 V, respectively, at 5 °C. When the IGBT chips were operated at 2000 V, no failure was observed in the tests, and the failure rate at 4000 m above sea level was less than 0.96 FIT; when they were operated at 2050 V, the failure rate of a single chip was 1.96 FIT at 4000 m above sea level. The test results of the IGBT chip and the failure rate calculation results are shown in

Table 3. Statistics of IGBT chip failure due to atmospheric neutron irradiation at different bias voltages.

No.	Voltage (V)	Number of Samples	Number of Failures	Total Effective Neutron Injection (n/cm2)
1	2000	30	0	1.21 × 1011
2	2050	30	2	1.18 × 1011
3	2100	30	3	1.13 × 1011
4	2200	30	11	7.83 × 1010
5	2300	37	23	3.54 × 1010

, and the atmospheric neutron failure rate of the IGBT single chip at different altitudes is shown in Figure 4.

According to the cosmic ray failure rate model for power devices proposed by Zeller et al. [9], the SEB failure rate is exponentially related to the applied voltage. The parameters a and b in the failure model require experimental testing, as they vary across different device models.

$$\mathsf{\lambda }(T,V_{DC})=a_{\left(T\right)}{e}^{-{\displaystyle \frac{b_{\left(T\right)}}{V_{DC}}}}$$

(5)

In this context, λ represents the failure rate, and V_DC is the applied direct current voltage. The SEB failure caused by cosmic rays is inherently random. For a given voltage, temperature, and altitude, the failure rate remains constant throughout the device’s lifecycle. High-speed proton irradiation tests are used to simulate atmospheric cosmic ray exposure, allowing for a rapid determination of the thyristor SEB failure rate. To prevent unexpected results and enhance the accuracy of the failure rate experiments, 3–5 samples are tested under each set of conditions. The average failure rate is calculated to represent the failure rate of the wafer or device under those conditions. The formula for calculating the average failure rate FR is as follows:

$$FR={\displaystyle \frac{r}{(m-r)\times AF+{\sum }_{i=1}^{r}AFi}}\times 3.6\times {10}^{12}FIT$$

(6)

In this context, r represents the number of failures observed under a specific irradiation condition in the experiment, while m denotes the total number of device samples under the same condition; both r and m are measured in units. AF is the equivalent experimental time, calculated as the ratio of the total fluence of non-failed devices to the atmospheric neutron flux in the actual application environment. AFi is the ratio of the total fluence for the ith device during the experiment to the atmospheric neutron flux in the actual application environment, measured in seconds. The neutron irradiation test results were processed using the formula for calculating the average failure rate of devices, yielding the failure rate of the IGBT chip under neutron irradiation, as shown in

Table 4. Statistical results of chip failure rate of IGBT chips due to atmospheric neutron irradiation with altitude at different bias voltages.

	Voltage (V)	Failure Rate at 4000 m (FIT)	4000 m Failure Rate Confidence Interval (FIT, 95% Confidence)	Failure Rate at 2000 m (FIT)	2000 m Failure Rate Confidence Interval (FIT, 95% Confidence)	Failure Rate at Sea Level (FIT)	Sea Level Failure Rate Confidence Interval (FIT, 95% Confidence)
1	2000	<9.60 × 10−1	/	<2.65 × 10−1	/	<5.13 × 10−2	/
2	2050	1.96	2.38 × 10−1, 5.46	5.41 × 10−1	6.55 × 10−2, 1.51	1.05 × 10−1	1.27 × 10−2, 2.92 × 10−1
3	2100	3.08	6.35 × 10−1, 7.41	8.49 × 10−1	1.75 × 10−1, 2.04	1.64 × 10−1	3.39 × 10−2, 3.96 × 10−1
4	2200	16.3	8.13, 27.2	4.49	2.24, 7.51	8.71 × 10−1	4.35 × 10−1, 1.46
5	2300	75.4	47.8, 109	20.8	13.2, 30.1	4.03	2.56, 5.84

. In the experimental process, due to the device failure at a lower bias voltage, a single experiment time is too long; with regard to the device residual radiation hazards, taking into account the safety of experimental personnel and the reliability of the experimental equipment, when the bias voltage of the device does not fail, we will consider increasing the bias span for the experiment, making it easy to find the voltage value of the failed device. At the same time, in the vicinity of the voltage in which failure of the device occurred, we will then carry out an adjustment of the voltage up and down. At the same time, in the vicinity of the device failure voltage value, we will then conduct the experiment within the time interval, that is, consider increasing the bias span for the experiment, which is the reason why our test voltage is not exact according to the same step [20].

For the built-in diode chip of 4500 V/3000 IGBT, ten voltage conditions of 2100 V, 2150 V, 2200 V, 2250 V, 2300 V, 2700 V, 2800 V, 2900 V, 3000 V, and 3100 V were selected at 5 °C for the irradiation test. No failure was observed when the diode chip was operated below 2150 V. The failure rate of a single chip at 2200 V, corresponding to an altitude of 4000 m, was 8.58 FIT. At the same voltage of 2200 V, the failure rate of the diode chip was half that of the IGBT chip. The test results of the diode chip are recorded, and the failure rate is calculated as shown in

Table 5. Statistics of diode chips’ failure due to atmospheric neutron irradiation at different bias voltages.

No.	Voltage (V)	Number of Samples	Number of Failures	Total Effective Neutron Injection (n/cm2)
1	2100	15	0	5.94 × 1010
2	2150	15	0	5.92 × 1010
3	2200	15	3	4.05 × 1010
4	2250	15	4	1.44 × 1010
5	2300	15	10	2.78 × 1010
6	2700	15	12	5.46 × 109
7	2800	15	13	1.99 × 109
8	2900	15	14	8.32 × 108
9	3000	15	14	3.17 × 108
10	3100	15	13	1.33 × 108

. The calculated results of atmospheric neutron failure rates of built-in diode chips at different voltages are shown in

Table 6. Statistical results of chip failure rate of diode chips due to atmospheric neutron irradiation with altitude at different bias voltages.

	Voltage (V)	Failure Rate at 4000 m (FIT)	4000 m Failure Rate Confidence Interval (FIT, 95% Confidence)	Failure Rate at 2000 m (FIT)	2000 m Failure Rate Confidence Interval (FIT, 95% Confidence)	Failure Rate at Sea Level (FIT)	Sea Level Failure Rate Confidence Interval (FIT, 95% Confidence)
1	2100	<1.95	/	<5.39 × 10−1	/	<1.04 × 10−1	/
2	2150	<1.96	/	<5.40 × 10−1	/	<1.05 × 10−1	/
3	2200	8.58	1.77, 20.7	2.37	4.88 × 10−1, 5.70	4.59 × 10−1	9.46 × 10−2, 1.10
4	2250	32.3	8.79, 70.7	8.90	2.43, 19.5	1.72	4.70 × 10−1, 3.78
5	2300	41.7	20, 71.2	11.5	5.51, 19.6	2.23	1.07, 3.81
6	2700	255	132, 418	70.3	36.3, 115	13.6	7.04, 22.3
7	2800	759	404, 1220	2.09 × 102	1.12 × 102, 3.38 × 102	40.6	21.6, 65.4
8	2900	1.95 × 103	1.07 × 103, 3.10 × 103	5.38 × 102	2.94 × 102, 8.55 × 102	1.04 × 102	57.0, 1.66 × 102
9	3000	5.12 × 103	3.90 × 103, 8.14 × 103	1.41 × 103	1.08 × 103, 2.24 × 103	2.74 × 102	2.09 × 102, 4.35 × 102
10	3100	1.14 × 104	6.05 × 103, 1.83 × 104	3.14 × 103	1.67 × 103, 5.06 × 103	6.07 × 102	3.23 × 102, 9.80 × 102

. The curves of the atmospheric neutron failure rate of the diode single chip with different voltages at different altitudes are shown in Figure 5.

For the thyristor chip, six voltage conditions, 2000 V, 2100 V, 2150 V, 2250 V, 2300 V, and 2400 V, were selected at 5 °C for the irradiation test. No failure was observed when the thyristor chip was operated at 2000 V. The failure rate of a single chip at 2100 V, corresponding to an altitude of 4000 m above sea level, was 76.4 FIT. The test results of the thyristor chip and the calculation of the failure rate are shown in

Table 7. Statistics of thyristor chips’ failure due to atmospheric neutron irradiation at different bias voltages.

No.	Voltage (V)	Number of Samples	Number of Failures	Total Effective Neutron Injection (n/cm2)
1	2000	5	0	1.01 × 1010
2	2100	5	3	4.56 × 109
3	2150	5	5	3.71 × 109
4	2250	5	4	2.27 × 109
5	2300	5	5	1.96 × 107
6	2400	5	3	1.81 × 106

, and the calculated results of atmospheric neutron failure rates of bypass thyristor chips at different voltages are shown in

Table 8. Statistical results of chip failure rate of diode chips due to atmospheric neutron irradiation with altitude at different bias voltages.

	Voltage (V)	Failure Rate at 4000 m (FIT)	4000 m Failure Rate Confidence Interval (FIT, 95% Confidence)	Failure Rate at 2000 m (FIT)	2000 m Failure Rate Confidence Interval (FIT, 95% Confidence)	Failure Rate at Sea Level (FIT)	Sea Level Failure Rate Confidence Interval (FIT, 95% Confidence)
1	2000	<11.5	/	<3.18	/	<6.16 × 10−1	/
2	2100	76.4	15.7, 1.84 × 102	21.1	4.34, 50.7	4.08	8.42 × 10−1, 9.83
3	2150	1.56 × 102	50.8, 3.20 × 102	43.2	14.0, 88.4	8.36	2.72, 17.1
4	2250	2.04 × 102	55.7, 4.48 × 102	56.4	15.4, 1.24 × 102	10.9	2.98, 23.9
5	2300	2.95 × 104	9.59 × 103, 6.05 × 104	8.14 × 103	2.64 × 103, 1.67 × 104	1.58 × 103	5.12 × 102, 3.23 × 103
6	2400	1.92 × 105	1.02 × 106, 1.19 × 107	5.31 × 104	2.81 × 105, 3.28 × 106	1.03 × 104	5.44 × 104, 6.35 × 105

. The atmospheric neutron failure rate of the thyristor chip at different altitudes is shown in Figure 6.

When high-energy neutrons (≥10 MeV) strike power devices such as IGBTs, diodes, and thyristors, they cause ionizing recoils with silicon atoms in the substrate material. This interaction generates a large number of electrons and holes locally within the device, disrupting the electric field in the space charge region and leading to single-event burnout (SEB). This SEB, induced by high-energy neutrons, is a primary cause of device failure [21,22,23]. We conducted a failure analysis on IGBTs that experienced SEB, and the results are shown in Figure 7. Figure 7a presents the OBIRCH analysis of the damaged area, while Figure 7b shows the SEM morphology of the damaged region.

Based on the results of the accelerated irradiation test and the atmospheric neutron environment under the actual working conditions, the failure rate caused by atmospheric neutrons under the real working environment of the power devices is calculated. The test results show that the failure rate of IGBTs and thyristors caused by atmospheric neutrons is greatly affected by the blocking voltage [24]. The larger the blocking voltage, the higher the failure rate. For IGBT chips, when working at 2000 V, the failure rate at 4000 m above sea level is less than 0.96 FIT; when working at 2050 V, the failure rate of a single chip at 4000 m above sea level is 1.96 FIT; for a diode chip, when working at 2150 V, the failure rate at 4000 m above sea level is less than 1.96 FIT; when working at 2200 V, the failure rate at 2200 V is less than 1.96 FIT. For thyristor chips, the failure rate at 2000 V is less than 11.5 FIT at 4000 m. At 2100 V, the failure rate at 4000 m is 76.4 FIT.

4. Conclusions

This study simulates the single-event burnout (SEB) effects on power devices such as IGBTs, diodes, and thyristors in flexible DC converter valves caused by high-energy neutrons in the atmospheric environment using a neutron irradiation spectrometer. Accelerated tests were conducted using an equivalent total particle injection method, confirming the presence of SEB under neutron irradiation. The failure was characterized by a significant increase in off-state current, leading to the loss of the device’s blocking capability. Additionally, a method for calculating the average SEB failure rate of power devices like IGBTs, diodes, and thyristors was proposed. Results showed that the failure rate of these devices increases with applied voltage and is positively correlated with altitude, due to the increase in neutron flux at higher altitudes. These findings provide a reference for designing the operating voltage of thyristors in flexible DC converter valves, avoiding excessive redundancy and enhancing the reliability and cost-effectiveness of DC transmission projects.