Evaluation of thermally activated defects behaviors in nitrogen-doped Czochralski silicon single crystals using deep level transient spectroscopy

Thermally activated defect behaviors in nitrogen (N)-doped Czochralski silicon (Cz-Si) single crystals were investigated using deep level transient spectroscopy and quasi-steady-state photoconductance to confirm the crystals’ applicability in insulated gate bipolar transistors (IGBTs). The thermally activated defects, which were probably N-vacancy complexes and degraded the minority carrier lifetime, were detected with extremely low densities in N-doped Cz-Si compared with N-rich floating zone Si single crystals after heat treatments at 500 °C, resulting in a high remaining value of minority carrier lifetime. The difference was assumed to come from whether vacancies were released in the Si matrix during heat treatment. For the Cz-Si, vacancies were assumed to be strongly bound with oxygen atoms with concentrations of 1017 atoms cm−3. Therefore, vacancies were not released during heat treatment, resulting in low remaining N-vacancy complex densities. N-doped Cz-Si are potential materials for IGBTs because of their low densities from thermally activated defects.


Introduction
Currently, power devices that can perform highly efficient power conversion and switching are widely used for automotive, smart grids, and other energy-saving applications. In particular, the role of insulated gate bipolar transistors (IGBTs) is becoming more important owing to their application in hybrid and electric vehicles. Additionally, silicon (Si) wafers used for IGBTs are required to be free from voids and oxide precipitates, which can degrade the gate-emitter breakdown voltage and bulk lifetime of minority carriers, respectively. 1) For IGBTs, Si wafers cut from single crystals with a diameter of 200 mm manufactured using the floating zone technique (FZ-Si) are typically used because they satisfy these requirements. Simultaneously, the transfer to 300 mm diameter wafers has started in Europe to improve productivity. 2,3) However, it might be difficult to enlarge crystal diameters using the FZ method, but using the Czochralski method, it is relatively easy to grow 300 mm diameter crystals. Therefore, it is necessary to develop a growth technique for 300 mm diameter Si single crystals manufactured using this method (Cz-Si) to replace FZ-Si in IGBTs. 4,5) Nitrogen (N)-doped Cz-Si single crystals with oxygen concentrations ([O I ]) below 3 × 10 17 atoms cm −3 (JEIDA) are potential materials for IGBTs. This is because N-doping for FZ-and Cz-Si suppresses void formation. [6][7][8][9][10][11][12] Moreover, no oxide precipitates are detected after heat treatments owing to their low [O I ] value. 13) Otherwise, N atoms form complexes easily with some impurities and point defects in Si single crystals such as oxygen atoms and vacancies. This is because these complexes are energetically favored as predicted using first-principles calculations. [14][15][16][17][18] Currently, N-vacancy related thermally activated defects with densities of 10 12 -10 13 cm −3 in FZ-Si wafers detected using a deep level transient spectroscopy (DLTS) technique are discussed. They behave as majority carrier traps in n-type Si wafers, resulting in reduction of the minority carrier lifetime after heat treatment at temperature ranges of 450°C-700°C. [19][20][21][22][23] However, such thermally activated defect behaviors in N-doped Cz-Si have not been reported. It is necessary to explore the defect behaviors in Ndoped Cz-Si to use as materials for IGBTs.
In this study, to confirm the utility of N-doped Cz-Si as materials for IGBTs, the thermally activated defect behaviors in N-doped Cz-Si were investigated using a DLTS technique. Minority carrier lifetime was also evaluated using quasisteady-state photoconductance (QSSPC) to investigate their effects on electrical properties. Cz-Si single crystals with N concentration of 1.5-5.0 × 10 14 atoms cm −3 were subjected to a 500°C heat treatment to generate thermally activated defects before DLTS and minority carrier lifetime evaluation.

Experimental methods
Cz-Si generally include oxygen atoms with a concentration of approximately 1 × 10 18 atoms cm −3 ; therefore, oxygen incorporation to Cz-Si can be suppressed under crystal growth conditions for low [O I ], such as using high gas flow rate, low crucible rotation rate, and an applied magnetic field. 24) By applying these growth conditions, N-doped Cz-Si single crystal ingots with diameters of 200 mm were grown in the 〈100〉 direction. N was doped by charging Si wafers with a Si nitride (Si 3 N 4 ) film using CVD to a Si melt in a quartz crucible for Cz-Si growth. N concentrations were estimated as 2.7-5.0 × 10 14 atoms cm −3 by considering a segregation coefficient of 0.0007 between the crystal and Si melt. 25) The [O I ] values measured using Fourier transform IR spectroscopy (FTIR) with a conversion factor of 3.14 × 10 17 cm −2 (JEIDA) were 2.0-2.4 × 10 17 atoms cm −3 . We prepared Ndoped FZ-Si to compare the defect behavior between Cz-Si and FZ-Si. In FZ-Si growth, N was doped to a Si melt via ambient N 2 gas. N concentration was 3.5 × 10 14 atoms cm −3 as evaluated by secondary ion mass spectrometry. These Ndoped Cz-and FZ-Si ingots were sliced perpendicularly to the crystal axis to an approximate thickness of 1 mm using a band saw. Eight samples were prepared from these ingots as listed in Table I. The surfaces of samples A and B listed in Table I were polished to commercial levels, and samples C-H were not polished but etched using a mixture of hydrofluoric acid (HF) and nitric acid (HNO 3 ) to remove damage caused by the slicing using a band saw. Subsequently, these samples were annealed at 500°C for 30 min under N 2 ambient conditions to generate thermally activated defects.
The concentrations of thermally activated defects that behave as majority carrier trap were measured by DLTS using SEMILAB DLS-1000 equipment with a frequency of 1 MHz measured using a capacitance meter. The Schottky contacts were formed by evaporating gold (Au) onto the sample surfaces. The contact area was sufficiently large to detect concentrations of 10 9 cm −3 . The Ohmic contacts were obtained by rubbing Ga on the rear surfaces. In the DLTS measurement, capacitance transients multiplied by a lock-in weighting function were measured with lock-in frequencies from 10 to 100 Hz. The reverse and pulse voltages were −6 V and −1 V for samples A to D and −2 V and 0 V for sample E, respectively. The detection areas were 7-15 μm from the Schottky contact for samples A to D and 2-6 μm for sample E. The voltage pulse width was 50 μs. The defect densities are proportional to the amplitude of the peak intensities of the DLTS spectra. The DLTS signal V 0 obtained using the lock-in weighting function is given by where ΔC is the capacitance change due to the reverse bias, T W is the time period of rock-in function, T d is the time interval between the rising edge of the bias pulse and the starting point of the lock-in reference signal, and τ is time constant of the capacitance transient. 26) The measured DLTS signal can be converted into trap densities N T according to the following equation where C is the capacitance under the quiescent reverse-biased condition, and N S is the dopant concentration. 26) The N S of each sample is determined from the capacitance voltage characteristics before DLTS evaluations and temperaturecorrected though DLTS evaluations (DLS-1000 measurement software for). ΔC is given from the DLTS signal using Eqs. (1) and (2). C is measured together with the DLTS signal at each DLTS measurement temperature.
To evaluate the electrical properties of as-grown and heattreated N-doped FZ-and Cz-Si, minority carrier lifetimes were measured using the QSSPC method. Sample surfaces were passivated by a methanol solution of quinhydrone (C 6 H 4 (OH) 2 ·C 6 H 4 O 2 , 0.05 mol l −1 ) following hydrofluoric acid treatments and ultrapure water rinse. Minority carrier lifetimes were measured using WCT-120TS equipment (Sinton Instruments). Figure 1 shows the current-voltage characteristic of the Schottky contact fabricated between the Au electrode and sample C. The Au electrode applied a positive voltage. The vertical axis is converted by logarithm of current (I). The values of ideality factor and barrier height of Schottky contact were 1.3 and 0.7 eV, respectively. Therefore, the Schottky characteristics were sufficiently displayed for the DLTS capacitance to be evaluated using the lock-in weighting function.  show the as-received and heat treated DLTS spectra of N-doped Cz-Si single crystals at 500°C for 30 min with a N concentration of 1.5 × 10 14 atoms cm −3 , respectively. These spectra were obtained with a lock-in frequency of 50 Hz. The DLTS signals were converted to majority carrier trap densities using Eqs. (1) and (2). The DLTS spectra include signals from built-in resistivity; thus, weak peaks detected in these DLTS evaluations removed the effect of built-in resistivity to obtain the trap densities. Therefore, trap densities were obtained from the difference between the peak intensity and base line. Base lines were defined by the following steps using the software equipped with DLS-1000. At first, the overall tilt comes from built-in resistivity which depends on evaluation temperature was corrected. Next, both end of the peak was connected by a straight line. The locations of the thermally activated defects for samples D and E were drawn using dotted chain lines which were labeled as E1 and E2. As shown in Figs. 2(a) and 2(b), there were no deep levels corresponding to E1 and E2 even in the heat-treated samples; therefore the defect concentrations seemed to be equal in the detectable range for DLTS (approximately 1 × 10 9 cm −3 , determined from 3 sigma of the white noise of the DLTS signal) regardless of heat treatment.

Results
Figures 2(c)-2(e) show the DLTS spectra of N-doped Cz-Si after heat treatment with N concentrations of 2.7 × 10 14 , 3.7 × 10 14 , and 5.0 × 10 14 atoms cm −3 , respectively. The deep levels of E1 and E2 were not detected at N concentrations of 2.7 × 10 14 atoms cm −3 , whereas E1, E2, or both were detected at a N concentration above 3.7 × 10 14 atoms cm −3 . The spectrum curve around E1 for sample C with N concentration of 2.7 × 10 14 atoms cm −3 did not come from E1 but from multi peaks which concentrations were below detective limit. Figure 3 shows the Arrhenius plot (e n /T on 1000/T, where e n is the emission rate of electrons, T is the DLTS measurement temperature) for E1 and E2 using sample D. The activation energy ΔE and apparent cross capture section σ app were 0.18 eV and 1.  V 2 O, and VO. 27,28) However, the activation energies obtained in DLTS evaluations are 0.23 and 0.43 eV for VV, 0.06 eV for VO 2 * , and 0.23 and 0.47 eV for V 2 O; therefore, they do not seem to be the origins of E1 and E2. The activation energy of VO is 0.18 eV and matched the result for E1; however, VO cannot survive heat treatments at 500°C.
The defect densities obtained from the peak intensities of the DLTS spectra are shown in Fig. 4. The black and white bars indicate the defect densities of Cz-Si and FZ-Si after heat treatment. 20) The as-received Cz-Si defect densities are shown using a dashed bar. The horizontal dashed lines with arrows show the detection limits of defect densities measured using the DLTS technique. The defect densities of Cz-Si were below 10 9 cm −3 at a N concentration range of 2.7-5.0 × 10 14 atoms cm −3 , lower than those of FZ-Si crystals with a concentration range of 10 12 -10 13 cm −3 . The pulse width in this study was 50 μs while the pulse width was 1 ms in the previous study, which describe the relationship between ΔC and pulse width of deep level for E1 and E2. 20) The impact of the difference in pulse width on ΔC is approximately 0% for E1 and 6% for E2. The observation that, the thermally activated defect densities are extremely low in N-doped Cz-Si compared with FZ-Si is not affected. Figure 5 shows the minority carrier lifetime evaluated by the QSSPC method. In the as-grown state, minority carrier lifetimes of both Cz-and FZ-Si exhibit high values. The minority carrier lifetime of heat-treated FZ-Si was significantly degraded to under 1000 μs. In the case of Cz-Si, the minority carrier lifetime remained sufficiently high for IGBT applications (approximately 5000 μs) even after heat treatment.
3.2. Discussion 3.2.1. Defect behavior models. As shown in Fig. 4, thermally activated defects which were detected using DLTS at the energy levels of E C − 0.18 and E C − 0.34 eV in the Si band gap were generated in the FZ-and Cz-Si single crystals after heat treatment at 500°C. The defect densities in Cz-Si are extremely low compared to those in FZ-Si crystals even at approximately equivalent N concentrations. The reasons for the difference in defect behaviors between the FZ-and Cz-Si crystals are discussed in this section.
It has been reported that thermally activated defects are related to N concentration and vacancy concentrations, suggesting that some N-vacancy complexes are candidates  for thermally activated defects. 19) Based on FTIR studies of an electron-irradiated N-doped FZ-Si single crystals, N 2 V 2 , located at 689 cm −1 in the FTIR spectrum, was identified as a suitable candidate for thermally activated defects, as its generation and disappearance behaviors were consistent with the DLTS results at 200°C-800°C heat treatments. [29][30][31] This suggests that the vacancies have a role in the generation of defects; however, the difference in the structure of vacancies between FZ-and Cz-Si have not been clarified because of their low concentration. Two models are proposed in this section using the vacancy structures of related complexes in as-grown FZ-and Cz-Si crystals.
(a) Small vacancy complexes in FZ-Si and VO 2 in the Cz-Si model Mullins et al. have reported that small vacancy complexes exist in as-grown FZ-Si single crystals which explains the thermally activated defect behaviors for generation after a 500°C heat treatment on the basis of DLTS evaluations. 20) The binding energy of vacancy complexes consisted of four to six vacancies that were estimated to be approximately 2.5 eV using first-principle calculations; 32) therefore, they were predicted to easily dissolve to single vacancies under a 500°C heat treatment resulting in the release of vacancies in the Si matrix during heat treatment. 20) Such released vacancies can form N-vacancy complexes because of their energetic affinity; [14][15][16][17] therefore, the concentration of N-vacancy complexes increases to detectable orders using DLTS evaluation in the models proposed by Mullins et al. 20) In general, vacancies in Cz-Si are considered to exist as VO 2 because of their high [O I ] values that are approximately 10 17 atoms cm −3 , 33) in contrast to FZ-Si with [O I ] values of approximately 1 × 10 16 atoms cm −3 . 29) VO 2 are stable under a 500°C heat treatment because they form at a relatively high temperature (around the void the formation temperature is approximately 1000°C-1100°C) during crystal growth; 34) thus vacancies are not released into the silicon matrix after heat treatment, and N-vacancy complexes are not detected, or are detected at extremely low concentration, using DLTS in Cz-Si crystals. The binding energies obtained using first-principle calculations were 3-5 eV for N 2 V 2 , [14][15][16][17] and 2.7 eV for VO 2 . 35) Although N-V complexes preferentially form with these binding energies, the reactions are affected by the binding energy and the concentration of point defects and impurities. It is assumed that the following reactions occur in the Ndoped Cz-Si: In equilibrium conditions, the following equations can be described based on the law of mass action: 33) ( ) where [X] is the concentration of complex X, A is preexponential factor, E X is the formation energy of complex X, k B is the Boltzmann constant, and T is the temperature during crystal growth. Although the value of A may vary in each equation, it is assumed to remain constant. The ratio of the concentration of VO 2 to N 2 V 2 is described as follows: We calculated the ratio at 1000 and 500°C; 1000°C is the temperature at which vacancy concentration is determined by the end of pair annihilation between vacancies and interstitial silicon during crystal growth, and 500°C is the heat treatment temperature.  [29][30][31] Electron irradiation introduces vacancies to the bulk of the Si single crystals resulting in detection of the vacancy related complexes which are undetectable in the as-grown state because of their low concentrations. 29) According to the study, after a 400°C -600°C heat treatment, N 2 V 2 which was a candidate for thermally activated defects were detected in Ndoped FZ-Si, whereas their concentrations were below the detection limit (approximately 10 12 cm −3 ) in Ndoped Cz-Si. 29) Comparing the results of this work, as shown in Fig. 4, the concentrations of thermally activated defects in Cz-Si were significantly lower than those in FZ-Si, which were reasonable with the N 2 V 2 behavior revealed using Inoue's FTIR studies. Inoue et al. proposed that VO which was detected in FZ-and Cz-Si just as electron irradiation had a role in the difference of N 2 V 2 behavior between FZ-and Cz-Si. 29) For FZ-Si, VO disappeared and new peaks originating from N 2 V 2 appeared after a 400°C-600°C heat treatment. It is considered that VO dissolves into vacancies and oxygen atoms because of low [O I ] values of FZ-Si and released vacancies attached to N, leading to the increment in the N 2 V 2 concentration after heat treatment. If Cz-Si is subjected to heat treatment at 400°C-600°C, VO disappears in a similar way to FZ-Si. However, any peaks of N-vacancy complexes such as N 2 V 2 are not detected and a new peak identified as VO n (n = 2-4) appears in the FTIR study, suggesting that the existing oxygen atoms diffuse in Cz-Si and attach to VO during heat treatment. 29) Although the existence of N 2 V 2 with a concentration below the detection limit of FTIR in N-doped Cz-Si is not discussed in Ref. 29, we speculate that the vacancies are not created in the Si matrix unlike FZ-Si and N 2 V 2 with low concentration in the as-grown state remaining undetectable by FTIR after heat treatment, but it can be detected by DLTS. 3.2.2. Utility of N-doped Cz-Si for IGBT materials. As shown in Fig. 4, thermally activated defects in N-doped Cz-Si single crystals with nitrogen concentration below 2.7 × 10 14 atoms cm −3 were not detected after heat treatment at 500°C for 30 min, but were detected with high densities of 10 12 -10 13 cm −3 in N-rich FZ-Si. 20) Furthermore, the minority carrier lifetime of Cz-Si remains sufficiently high after heat treatment. In the models mentioned in Sect. 3.2.1, the formation of the N-V complex in Cz-Si is suppressed because of competition with the formation of V-O complexes. V-O complexes also introduce deep level band gaps, such as the A-center. 36) However, electrically neutral VO 2 are formed by attaching oxygen atoms to VO complexes during heat treatment, thus deep levels introduced by heat treatments in N-doped Cz-Si are less than those in N-doped FZ-Si. Furthermore, the threshold of the N concentration for thermally activated defect detection is found to be approximately 3.7 × 10 14 atoms cm −3 , which is sufficiently high to obtaining as-grown defect-free Cz-Si crystals with low [O I ] values required for IGBT wafers. 12) Therefore, we believe that N-doped Cz-Si crystals with [O I ] values below 2.4 × 10 17 atoms cm −3 and N concentrations below 3.7 × 10 14 atoms cm −3 are potential materials for IGBTs. Regarding to the evaluation technique, N-vacancy complexes which are hardly detected because of their low concentration can be detected using the DLTS technique without the introduction of vacancies such as electron irradiation. Thus, the DLTS technique can be an effective method for the evaluation of as-grown N-doped Cz-Si single crystals to reveal defect behavior.

Conclusions
The thermally activated defect behaviors of N-doped Cz-Si single crystals were investigated using DLTS to confirm their utility for IGBT materials. Thermally activated defects which were likely N-vacancy complexes in N-doped Cz-Si were detected with significantly lower densities compared with N-rich FZ-Si single crystals after 500°C heat treatments. Meanwhile, minority carrier lifetimes were evaluated, and the Cz-Si lifetime value remained sufficiently high even after heat treatment. Two models were proposed to explain the difference in defect densities between Cz-and FZ-Si using the vacancy structures of as-grown FZ-and Cz-Si. In both models, the thermally activated defect densities were determined whether vacancies were released or not in the Si matrix during heat treatments. For Cz-Si, vacancies were assumed to be strongly bound by oxygen atoms included in the density (10 17 atoms cm −3 ). Therefore, vacancies were not released in the Si matrix during heat treatment, resulting in low remaining N-vacancy complex densities after heat treatment. N-doped Cz-Si single crystals are potential materials for IGBTs because of the low densities associated with their thermally activated defects.