A Method for Testing Defect Energy Levels of GaN Power Devices Based on Low-Frequency Noise

Abstract

The present invention relates to a method for testing defect energy levels of GaN power devices based on low-frequency noise. The method includes selecting low-frequency noise parameters to be measured for GaN power devices; determining the bias conditions for low-frequency noise testing of GaN power devices; obtaining the noise power spectral density of the parameters to be measured at different temperatures; extracting generation-recombination (G-R) noise at different temperatures; obtaining the peak frequencies of G-R noise at different temperatures; establishing an Arrhenius equation based on the peak center frequency and temperature of G-R noise; and extracting based on the best fit of the Arrhenius equation data E <subgt; a

Description

Technical Field

[0001] The present invention relates to the technical field of reliability analysis of semiconductor devices, and particularly to a method for testing defect energy levels of GaN power devices based on low-frequency noise. Background Art

[0002] Compared with traditional devices, GaN power devices have the advantages of high frequency, small size, high power density, high temperature resistance, etc. They are gradually replacing traditional devices in the fields of wireless communication and switching power supplies and will become the mainstream of future market applications.

[0003] GaN power devices also have the problem of reliability degradation. Moreover, new technologies are adopted in aspects such as materials, processes, and device structures, and the theoretical understanding of reliability degradation is relatively weak. The physical mechanisms of reliability degradation of traditional devices cannot solve the new problems of GaN power devices. Therefore, there is an urgent need for a defect test and characterization method suitable for GaN power devices to obtain relevant information such as defect energy levels, which is an important way to understand the reliability degradation mechanism of GaN power devices. The main mechanism of reliability degradation of GaN power devices is the inverse piezoelectric effect, which is different from the reliability degradation mechanism of traditional devices. That is, lattice defects are generated in the electric field environment, which are manifested as defect energy levels in the energy band and act as generation-recombination (G-R), trapping centers, leakage paths, etc., ultimately leading to device reliability degradation. Current research results prove that there is an inevitable connection between the reliability degradation of GaN power devices and the generation of lattice defects. However, the understanding of defect information in GaN power devices is still insufficient. Therefore, obtaining defect information has positive significance for the reliability prediction and establishment of acceleration methods of GaN power devices.

[0004] Low-frequency noise (LFN) and deep-level transient spectroscopy (DLTS) test technologies are currently the most mature semiconductor defect test and characterization methods. Through the test results under different temperature conditions, defect energy levels and related information can be analyzed. Among them, DLTS has higher constraints on device-type samples, while LFN is more applicable to device-type samples; LFN can not only obtain deep-level information but also obtain shallow-level defect information; in addition, the overall machine solution of the LFN test system is more mature.

[0005] Based on the LFN test technology with strong applicability and easy implementation, the present invention obtains defect energy level information of GaN power devices, which has positive significance for the theoretical understanding of the reliability of GaN power devices, degradation prediction, and establishment of acceleration evaluation methods. Summary of the Invention

[0006] The object of the present invention is to provide a method for testing the defect energy level of GaN power devices based on low-frequency noise. This method involves selecting the low-frequency noise (LFN) parameters to be measured for GaN power devices; determining the bias conditions for LFN testing of GaN power devices; obtaining the noise power spectral density (NPSD) of the parameters to be measured at different temperatures; extracting the generation-recombination (G-R) noise at different temperatures; obtaining the peak frequency of the G-R noise at different temperatures; establishing an Arrhenius equation based on the peak center frequency and temperature of the G-R noise; and extracting E based on the best fit of the Arrhenius equation data a The steps are completed. The method of the present invention has the following advantages compared with other methods: The LFN test system consists of general discrete instruments and does not require additional system customization, making it a relatively easy-to-implement test technology; the interface of this test technology is compatible with semiconductor parameter test systems, and the software and hardware can be integrated with semiconductor test systems; this test technology has stronger applicability and is generally applicable to commercial finished devices; the NPSD is the statistical result of a large number of noise signal sampling values, with lower test errors

[0007] A method for testing the defect energy level of GaN power devices based on low-frequency noise according to the present invention is carried out according to the following steps

[0008] a. Select the drain current I D as the analysis object of the low-frequency noise of GaN power devices

[0009] b. Determine the GaN power device and select the static operating point as the bias condition for low-frequency noise testing

[0010] c. Obtain the noise power spectral density of the parameters to be measured at different temperatures. Based on the 1 / f noise test system, obtain the noise power spectral density of I D at different temperatures, and the test frequency covers at least the range of 10 1 -10 5 Hz

[0011] d. Extract the generation-recombination G-R noise at different temperatures. First, exclude the influence of thermal noise in the noise power spectral density, and then extract the recombination G-R noise characteristics at different temperatures

[0012] e. Obtain the peak frequency of the G-R noise at different temperatures

[0013] f. Establish an Arrhenius equation based on the peak center frequency and temperature of the G-R noise

[0014]

[0015] where T is the Kelvin temperature, with the unit of K; τ is the G-R center emission rate; K is the Boltzmann constant, K≈1.380649×10-23 J / K; E a is the defect activation energy, with the unit of eV; f is the peak frequency of NPSD, with the unit of Hz;

[0016] g. Based on the Arrhenius equation data and the peak frequency of G-R noise at different temperatures, several groups of In(T 2 ·τ), 1 / KT data are obtained, and the slope of the best-fit straight line is E a .

[0017] A method for testing the defect energy level of a GaN power device based on low-frequency noise according to the present invention has the following innovative points:

[0018] E a is a microscopic parameter, and its relationship with macroscopic physical quantities can be established through the Arrhenius equation of formula (1). Currently, there are various observable physical quantities that can be used to establish the Arrhenius equation, such as the capacitance value of DLTS, the current pulse time of the thermally stimulated current method (TSC), and the current noise power spectral density (NPSD) value used in the present invention. The innovative point of the present invention is to obtain the current noise power spectral density (NPSD) at different temperatures based on the low-frequency noise (LFN) test method, establish the Arrhenius equation, and extract the defect E a , and has the following advantages compared with other methods: The low-frequency noise (LFN) test system consists of general discrete instruments, does not require additional system customization, and is a relatively easy-to-implement test technology; the interface of this test technology is compatible with the semiconductor parameter test system, and the software and hardware of the semiconductor test system can be integrated; this test technology has stronger applicability and is generally applicable to commercial finished devices; the noise power spectral density (NPSD) is the statistical result of a large number of noise signal sampling values, with lower test errors. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 is the flow chart of the method for testing the defect energy level of a GaN power device based on low-frequency noise (LFN) of the present invention;

[0020] Figure 2 is the DC bias schematic diagram of the low-frequency noise (LFN) test of the present invention, where 1 is the GaN power device, 2 is the source meter unit (SMU) channel 1, and 3 is the source meter unit SMU channel 2;

[0021] Figure 3 is the schematic diagram of the static operating point of the low-frequency noise (LFN) test of the GaN power device of the present invention;

[0022] Figure 4This is a schematic diagram of the LFN test principle of the present invention. Here, 1 is a GaN power device, 2 is the channel 1 of the source meter unit (SMU), 3 is the channel 2 of the source meter unit SMU, 4 is a variable-temperature sample chamber, 5 is a spectrum analyzer, 6 is a preamplifier, and 7 is a load and compensation circuit;

[0023] Figure 5 This is the LFN test bias condition of the test sample of the present invention, where (a) is the transfer characteristic curve and (b) is the output characteristic curve;

[0024] Figure 6 This is the relationship between NPS and frequency at different temperatures of the present invention;

[0025] Figure 7 This is the extraction result of Ea based on the best fit of the present invention.

[0026] A method for testing the defect energy level of a GaN power device based on low-frequency noise according to the present invention. In this method:

[0027] According to Figure 1 Step (1): Select the low-frequency noise (LFN) test parameters. Take the drain current I D As the low-frequency noise (LFN) test parameter, that is, obtain the noise power spectral density (NPSD) of I D And analyze it;

[0028] Figure 1 Step (2): Determine the LFN test bias condition, which is set in the constant voltage, constant current, and threshold Offset modes. For example, when working under the power amplifier condition, select I D Injecting constant current as the bias condition for the low-frequency noise (LFN) test. As shown in Figure 2 That is, Figure 2 The GaN power device (1) in is the sample to be tested, Figure 2 The (2) in is the channel 1 of the source meter unit (SMU). Apply the gate voltage to the sample to be tested. Figure 2 The (3) in is the channel 2 of the source meter unit (SMU). Inject constant Current into the GaN power device (1) sample to be tested. The DC operating point during the LFN test of the GaN power device (1) is as shown in Figure 3 Shown;

[0029] Figure 1 Step (3): Conduct the low-frequency noise (LFN) test at different temperatures. Establish the connection between the sample to be tested and the test system. As shown in Figure 4 Shown, place the Figure 4 GaN power device (1) sample to be tested in the Figure 4 Variable-temperature sample chamber (4) in, and according to Figure 2In (2), it is the source meter unit (SMU) channel 1, and (3) is the source meter unit (SMU) channel 2, which are respectively connected to Figure 4 In Figure 4 , (2) is the source meter unit (SMU) channel 1, and (3) is the source meter unit (SMU) channel 2, and they are respectively connected, and the bias conditions for the LFN test are set. At the same time, it is connected to Figure 4 the load and compensation circuit (7) in Figure 4 . The noise signal passes through Figure 4 the preamplifier for acquisition (6) and is output to Figure 4 the spectrum analyzer (5); select the test frequency range, which at least covers 10 1 -10 5 Hz frequency range to ensure that it covers the noise power spectral density (NPSD) frequency range at different temperatures; within the temperature range allowed by the device manual, determine the test temperature interval, such as 300 - 375K, and select several temperatures for the LFN test;

[0030] According to Figure 1 Step (4) to extract the G - R noise, and through step (3), obtain the test result of the low - frequency noise (LFN) of I D , that is, the noise power spectral density (NPSD) of I D , which contains main components such as G - R noise, 1 / f noise, and thermal noise. Among them, the G - R noise is related to Ea; first, remove the influence of thermal noise in the noise power spectral density (NPSD); then multiply the noise power spectral density (NPSD) by the corresponding frequency value f to obtain the noise power spectrum (NPS) of I D , and extract the G - R noise at each temperature according to this method;

[0031] According to Figure 1 Step (5) to obtain the noise peak frequencies at different temperatures. Based on the G - R noise at each temperature extracted in step (4), respectively read the G - R noise peak frequencies at each temperature. For example, there are n groups of temperature data. The G - R noise peak frequency at T1 temperature is f c1 , the G - R noise peak frequency at T2 temperature is f c2 , and the G - R noise peak frequency at T n temperature is f cn ;

[0032] According to Figure 1 Step (6) to establish the Arrhenius equation. Specifically, refer to formulas (2) - (7) to establish the Arrhenius equation:

[0033]

[0034] where τ is the G - R center emission rate; σ is the trap capture cross - section; <v>is the average thermal velocity; Nc is the effective density of states at the conduction band edge; g is the energy level degeneracy; E a is the defect activation energy, with the unit of eV; K is the Boltzmann constant, K≈1.380649×10 -23 J / K; T is the Kelvin temperature, with the unit of K; Nc and <v>It can be expressed by formula (3):

[0035]

[0036] M C is the number of equivalent conduction band minima; m* is the effective mass of carriers; h is Planck's constant, h≈6.6260755×10 -34 J·s; Substituting formula (3) into formula (2), 1 / τ can be expressed as:

[0037]

[0038] Γ is the temperature constant term, and let Γ be formula (5),

[0039]

[0040] Substituting formula (5) into formula (4), 1 / τ can be expressed as

[0041]

[0042] Converting formula (6) into logarithmic form, an Arrhenius equation in the following form is established

[0043]

[0044] f is the frequency, and the unit is Hz;

[0045] According to Figure 1 Extract E in step (7) a , and obtain the NPS peaks at n temperatures in step (5), that is, (f c1 , T1), (f c2 , T2)…(f cn , T n ), and substitute them into formula (7) to draw n points on the X-Y two-dimensional plane

[0046]

[0047] Use the best fitting method to fit the n points into a straight line, and E a is the slope of the straight line. Specific implementation method

[0048] Example

[0049] A method for testing the defect energy level of a GaN power device based on low-frequency noise according to the present invention is carried out according to the following steps:

[0050] a. Select the drain current I D as the analysis object of the low-frequency noise of the GaN power device;

[0051] b. Determine the GaN power device and select the quiescent operating point as the bias condition for low-frequency noise testing;

[0052] c. Obtain the noise power spectral density of the parameter to be measured at different temperatures. Based on the 1 / f noise test system, obtain the noise power spectral density of I at different temperatures. The test frequency should cover at least the range from 10 D to 1 10 5 Hz;

[0053] d. Extract the generation-recombination G-R noise at different temperatures. First, exclude the influence of thermal noise in the noise power spectral density, and then extract the characteristics of the recombination G-R noise at different temperatures;

[0054] e. Obtain the peak frequency of the G-R noise at different temperatures;

[0055] f. Establish the Arrhenius equation based on the peak center frequency and temperature of the G-R noise;

[0056]

[0057] where T is the Kelvin temperature in the unit of K; τ is the G-R center emission rate; K is the Boltzmann constant, K ≈ 1.380649×10 -23 J / K; E a is the defect activation energy in the unit of eV; f is the peak frequency of the NPSD in the unit of Hz;

[0058] g. Based on the data of the Arrhenius equation and the peak frequency of the G-R noise at different temperatures, obtain several groups of In(T2·τ) and 1 / KT data. The slope of the best-fit straight line is E a ;

[0059] Select a commercial GaN power device as the test object and conduct low-frequency noise (LFN) testing on I D ;

[0060] Set the bias condition for low-frequency noise (LFN) testing as I D = 10 μA, as shown by "○" in Figure 5 ;

[0061] Select the LFN test frequency range from 1 - 100 KHz and conduct low-frequency noise (LFN) testing at the temperatures of 293.15 K, 333.15 K, and 373.15 K;

[0062] The original data of the low-frequency noise (LFN) testing is V0. Exclude the influence of thermal noise through formulas (9) - (13):

[0063] V1 = V0 - G Preamp -10 log 10 RBW formula (9)

[0064] V0 is the original noise density in the LFN test, with the unit of dBV 2 / Hz; G Preamp is the preamplifier gain, with the unit of dBV 2 / Hz; RBW is the analysis bandwidth of the spectrum analyzer, with the unit of dBV 2 / Hz; V1 is the noise density after compensating the preamplifier and the spectrum analyzer, with the unit of dBV 2 / Hz;

[0065]

[0066] V2 is the transformed form of V1, with the unit of V / Hz 1 / 2 ;

[0067]

[0068] V3 is the voltage noise density excluding the influence of thermal noise, with the unit of V / Hz 1 / 2 ; K is the Boltzmann constant, K≈1.380649×10 -23 J / K; T is the Kelvin temperature, with the unit of K; Δf is the spectrum interval, with the unit of Hz; R LOAD is the test sample load, with the unit of Ω; R Preamp is the input impedance of the preamplifier, with the unit of Ω;

[0069]

[0070] I d_noise is the current noise density excluding the influence of thermal noise, with the unit of A / Hz 1 / 2 ; r ds is the leakage resistance of the small signal source of the test sample, with the unit of Ω;

[0071] NPSD = I d noise 2 formula (13)

[0072] The noise power spectral density (NPSD) is the noise power spectral density of I D with the unit of A 2 / Hz; The G-R noise of the noise power spectral density (NPSD) is extracted through the following formula (13);

[0073] NPS = NPSD × f formula (14)

[0074] NPS is the noise power spectrum of I D with the unit of A 2 ; f is the frequency, with the unit of Hz;

[0075] Determine the NPS peak frequencies at various temperatures, as Figure 6 indicated by "↓" in [reference], that is, the peak frequency at 293.15 K is 30 Hz, the peak frequency at 333.15 K is 1.18 KHz, and the peak frequency at 373.15 K is 5.905 KHz;

[0076] Substitute the temperature and the corresponding NPS peak frequencies into the Arrhenius equation in formula (1) to obtain three coordinate values on the two-dimensional plane. In this embodiment, the first point is (39.58, 6.12), the second point is (34.84, 2.71), and the third point is (31.1, 1.32);

[0077] Obtain the best-fit straight line for each Point, with the slope being 0.57, as Figure 7 shown, and obtain Ea = 0.57 eV.< / v> < / v>

Claims

1. A method for testing the defect energy level of a GaN power device based on low-frequency noise, characterized in that Proceed as follows: a. Select the drain current I D as the analysis object of the low-frequency noise of the GaN power device; b. Determine the GaN power device and select the quiescent operating point as the bias condition for low-frequency noise testing; c. Obtain the noise power spectral density of the parameter to be measured at different temperatures: Based on the 1 / f noise test system, obtain the noise power spectral density of I at different temperatures, and the test frequency covers at least the range of 10 D to 10 1 -10 5 Hz; d. Extract the generation-recombination G-R noise at different temperatures: First, eliminate the influence of thermal noise in the noise power spectral density, and then extract the recombination G-R noise characteristics at different temperatures; e. Obtain the peak frequency of G-R noise at different temperatures; f. Establish the Arrhenius equation based on the peak center frequency and temperature of G-R noise; Where T is the Kelvin temperature, with the unit of K; τ is the emission rate of G-R centers; K is the Boltzmann constant, K≈1.380649×10 -23 J / K; E a is the defect activation energy, with the unit of eV; f is the peak frequency of NPSD, with the unit of Hz, Γ is the temperature constant term, and σ is the trap capture cross section; g. Based on the Arrhenius equation data and the peak frequencies of G-R noise at different temperatures, several sets of In(T 2 ·τ) and 1 / KT data are obtained, and the slope of the best-fit straight line is E a ; Specifically, refer to Formula (2) - Formula (7) to establish the Arrhenius equation: where τ is the emission rate of the G-R center; σ is the trap capture cross-section; <v>is the average thermal velocity; Nc is the effective density of states at the conduction band edge; g is the energy level degeneracy; E a is the defect activation energy, with the unit of eV; K is the Boltzmann constant, K ≈ 1.380649×10 -23 J / K; T is the Kelvin temperature, with the unit of K; Nc and <v>It can be expressed by Formula (3):< / v> < / v> M C is the number of equivalent conduction band minima; m* is the effective mass of carriers; h is Planck's constant, h ≈ 6.6260755×10 - 34 J·s; Substitute Formula (3) into Formula (2), and 1 / τ can be expressed as: Γ is the temperature constant term. Let Γ be Formula (5). Substitute Formula (5) into Formula (4), and 1 / τ can be expressed as: Convert Formula (6) into logarithmic form to establish the Arrhenius equation in the following form: f is the frequency, and the unit is Hz; Extract E according to step g a , obtain the NPS peaks at n temperatures in step e, namely (f c1 , T1), (f c2 , T2) … (f cn , T n ), substitute them into formula (7), and draw n points on the X-Y two-dimensional plane The best fit method is used to fit n points into a straight line, and E a is the slope of the straight line.

Images