GaN is a promising semiconductor for high power and high frequency devices such as pn diodes [1] and high electron mobility transistors (HEMTs). Moreover, GaN grown on silicon substrates allow the study of new technologies [2]. However, many defects can degrade the performance of the GaN based devices and occur in the form of native point defects, impurities and extended defects. Few defect states were found, such as nitrogen vacancies, Mg dopants or C impurities but they are not well explained. Therefore, in this work we investigated the defect characteristics of Mg-doped GaN-on-Si p+n diodes using the Deep-Level Transient Fourier Spectroscopy (DLTFS). This technique is based on the measurement of capacitance (C-DLTFS) or current (I-DLTFS) transients during charge release from filled traps. It is an electrical characterization technique used to determine and calculate the density N T, the energy position and the capture cross-section σ T of defect states.The p+n diodes were grown by MOCVD on Si substrates. The p+n structure was composed of a 400 nm thick Mg-doped p +-GaN layer at 6x1019 cm-3 and of a 750 nm thick Si-doped n-GaN layer at 3x1016 cm-3. The DLTFS measurements have been performed between 100 K and 425 K with a dynamic signal of frequency of 1 MHz for various reverse voltages U R and pulse voltages U P = -0.5 V for C-DLTFS and U P = 3 V for I-DLTFS. Capacitance-voltage measurements (not shown) evidenced a graded and non-linear p + n junction.The C-DLTFS spectrum (fig 1.a) shows two positive peaks (T1 and T2) respectively at 130 K and at 330 K and the I-DLTFS spectrum (fig 1.b) shows a positive peak at 390 K. For the peaks at high temperature on C-DLTFS and I-DLTFS spectra, average E A values of about 0.83 eV and 0.81 eV were extracted for both spectra respectively (fig 1.c). So, these two peaks are related to the same defect state. Besides, shifts to high temperature values (opposite as for the Poole-Frenkel effect) were observed on both spectra and this could indicate interface traps. Also, when increasing U P to positive values (not shown), the T1 amplitude increases while the T2 amplitude decreases [3]. This defect state might be associated with hole traps at the interface of the p+n junction on the p+ -side of the junction and are tentatively assigned to CN states [4]. For the peak at low temperature (T1), no temperature shift with U R was observed indicating bulk traps in the n-region and an activation energy (E A) of about 0.23 eV was extracted (fig 1.c). This defect is commonly assigned to dislocations related to VN defects or to VN-VG complexes. Moreover, Mg dopants (p-type) for this doping level have an E A value of about 0.25 eV very close to the T1 E A value. So, the graded and non-linear p + n junction could be explained by the Mg dopants diffusion into the n-type region along dislocations, as previously observed [5].[1] Y. Zhang et al., “High-Performance 500 V Quasi- and Fully-Vertical GaN-on-Si pn Diodes,” IEEE Electron Device Lett., vol. 38, no. 2, pp. 248–251, Feb. 2017.[2] M. Ishida et al., “GaN on Si Technologies for Power Switching Devices,” IEEE Trans. Electron Devices, vol. 60, no. 10, pp. 3053–3059, Oct. 2013.[3] Y. Lechaux et al., “Characterization of defect states in Mg-doped GaN-on-Si pn diodes using deep-level transient Fourier spectroscopy,” Semicond. Sci. Technol., vol. 36, no. 2, p. 024002, Dec. 2020.[4] T. Narita et al., “The origin of carbon-related carrier compensation in p-type GaN layers grown by MOVPE,” J. Appl. Phys., vol. 124, no. 21, p. 215701, Dec. 2018.[5] S. Usami et al., “Direct evidence of Mg diffusion through threading mixed dislocations in GaN p–n diodes and its effect on reverse leakage current,” Appl. Phys. Lett., vol. 114, no. 23, p. 232105, Jun. 2019. Figure 1