InGaAs photodiodes are very promising and have been widely used in short wavelength infrared（SWIR）detection ^［1-2］. The defects caused by the lattice mismatch between In_0.74Ga_0.26As and InP substrate is inevitable. In order to improve the performance of In_0.74Ga_0.26As photodiodes，it is necessary to optimize the epilayer and manufacturing processes of the InGaAs photodiodes. The dark current of the mesa In_0.74Ga_0.26As photodiodes is composed of body dark current and side dark current ^［3］. Passivation is one of the key processes of the mesa PIN photodiodes. Therefore，it is essential to optimize the passivation to reduce the dark current of the mesa In_0.74Al_0.26As（p）/In_0.74Ga_0.26As（i）/In_xAl_1-xAs（n）photodiodes.

As an outstanding thin film deposition technology，ALD has broad prospects in semiconductor device fabrication ^［4-6］. ALD involves two self-limiting surface reactions in which the growth substrate is exposed to alternating pulses of co-reactant and precursor ^［7-8］. Therefore，ALD can precisely control the thickness of thin film with an accuracy of an atomic layer. Moreover，films deposited by ALD are dense and of high quality. In our previous work，Al₂O₃ deposited by ALD has been proved to be an effective passivation film in In_0.74Ga_0.26As photodiodes ^［9-10］. However，the reason why the ALD-Al₂O₃ reduces the dark current of In_0.74Ga_0.26As photodiodes has not been explored. Therefore，it is necessary to study the interface state density between ALD-Al₂O₃ and In_0.74Al_0.26As layer.

In this paper，TEM and XPS measurements were performed to investigate the interface between two different dielectric films and In_0.74Al_0.26As layer. The two different dielectric films were ALD-Al₂O₃ and ICPCVD-SiN_x respectively. In addition，MIS capacitors have been prepared by using above two dielectric films to further quantitatively study the interface state density between dielectric film and In_0.74Al_0.26As layer.

To investigate the interface state density of SiN_x/In_0.74Al_0.26As and SiN_x/Al₂O₃/ In_0.74Al_0.26As，two series of MIS capacitor were prepared. The SiN_x or Al₂O₃ were deposited on the In_0.74Al_0.26As/In_0.74Ga_0.26As/In_xAl_1-xAs heterostructures that were grown on an InP substrate by gas source molecular beam epitaxy（MBE）^［11］. What’s more，the materials used to prepare the samples were cleaved from the same wafer. To remove surface native oxide layer，the wafer involved in this paper were treated with hydrofluoric acid buffer solution before film deposition.

TEM and XPS measurements were employed to investigate the interface between films and In_0.74Al_0.26As layer. The 20-nm-thick Al₂O₃ films were deposited on In_0.74Al_0.26As by 220 cycles of trimethylaluminum（TMA）and H₂O at 150°C for the TEM and XPS measurements. For comparison purpose，about 20-nm-thick SiN_x films were deposited on In_0.74Al_0.26As by ICPCVD of SiH₄ and N₂ at 75°C for the TEM and XPS measurements. Furthermore，bare In_0.74Al_0.26As wafer with native oxides was used as a control for XPS measurement. The insulator of sample A was deposited with 135-nm-thick SiN_x film by ICPCVD. The insulator of sample B is ICPCVD-SiN_x/ALD-Al₂O₃ bilayer，where 20-nm-thick Al₂O₃ was firstly deposited by ALD and 130-nm-thick SiN_x was then deposited by ICPCVD. Fig. 1（a）shows the sectional schematic of MIS capacitors. The dielectric film of sample A and sample B were SiN_x film and SiN_x/Al₂O₃ bilayer respectively. The photography of MIS capacitor was represented in Fig. 1（b）. The C-V characteristics of MIS capacitors at different frequencies were measured by Agilent B1500A Semiconductor Device Analyzer.

Pt was firstly deposited on the surface of In_0.74Al_0.26As wafer to protect the wafer from etching damage. Focused Ion Beam（FIB）was used to thin the sample，and then the thinned sample was subjected to TEM testing. The cross-section TEM images of ICPCVD-SiN_x/In_0.74Al_0.26As and ALD-Al₂O₃/In_0.74Al_0.26As structure were shown in Fig. 2（top）. In comparison with SiN_x/In_0.74Al_0.26As，a sharp transition from Al₂O₃ to In_0.74Al_0.26As can be observed. In addition，the cross-sectional composition information of Fig. 2（bottom）obtained by Energy Dispersive X-ray Spectroscopy（EDS）can also illustrate this. The transition from SiN_x to In_0.74Al_0.26As layer was ~5 nm，while the transition from Al₂O₃ to In_0.74Al_0.26As layer was ~3 nm.

In order to further study the interface state density between film and In_0.74Al_0.26As，XPS test combined with Ar⁺ sputtering（2 kV，20 μA）was utilized. XPS spectra were obtained with the Axis UltraDLD spectrometer（Kratos Analytical-A Shimadzu Group Company）with a monochromatic Al Kα source（hν=1486.6 eV）and a charge neutralization system. The spectra were taken when the vacuum of the analysis chamber was less than 5×10^–9 Torr. The electron energy analyzer works in the hybrid magnification mode，and the take-off angle for the analyzer relative to sample surface is 90°. The X-ray source power was set to 105 W（15 kV，7 mA）for high-resolution spectra acquisition. The pass energy of 40 eV were utilized for narrow scan spectra. The energy step size of 0.1 eV were chosen for narrow scan spectra. The C 1s peak of environmental pollution carbon adsorbed on the sample surface is used as the reference peak for spectral energy correction to complete the peak position calibration：set the C 1s peak position of the pollution carbon to 284.8 eV. Calculate the relative percentage of elements through the peak area of the element peak and the sensitivity factor of the instrument.

In3d_5/2 spectra obtained by XPS narrow scan was shown in Fig. 3. As shown in Fig. 3（a），there was a certain amount of In₂O₃ on the surface of the bare In_0.74Al_0.26As wafer. Compared with the surface of bare In_0.74Al_0.26As wafer，almost no In₂O₃ could be found in the bulk of In_0.74Al_0.26As［Fig. 3（b）］. The amount of In₂O₃ at the SiN_x/In_0.74Al_0.26As interface［Fig. 3（e）］was less than that of the surface of bare In_0.74Al_0.26As wafer. However，there was still a certain amount of In₂O₃ at the SiN_x/In_0.74Al_0.26As interface that couldn’t be ignored. Thus，ICPCVD-SiN_x film could reduce part of In₂O₃ at the interface between SiN_x and In_0.74Al_0.26As. The interface between Al₂O₃/In_0.74Al_0.26As［Fig. 3（h）］was equivalent to the bulk of In_0.74Al_0.26As，and there was almost no In₂O₃ could be found at the interface between Al₂O₃ and In_0.74Al_0.26As. Therefore，it can be concluded that ALD-Al₂O₃ can effectively suppress the In₂O₃ at the interface between film and In_0.74Al_0.26As.

Figure 4 shows the C-V curves of SiN_x/In_0.74Al_0.26As and SiN_x/Al₂O₃/In_0.74Al_0.26As MIS capacitors with bias voltage from 25 V to -25 V at 210 K. The C-V measurements of these capacitors was performed with frequency varying from 1 kHz to 1 MHz. Under the low frequency limit，the capacitance of MIS capacitor（C_LF）was equivalent to the parallel connection of semiconductor capacitance（C_s）and interface trap capacitance（C_it）and then the series connection of insulating layer capacitance（C_i）. Therefore，the C_LF can be expressed as：

However，under the high frequency limit，the interface trapped charge can not keep up with the change of high frequency. Therefore，the capacitance of MIS capacitor（C_HF）was equivalent to the series connection of semiconductor capacitance（C_s）and insulating layer capacitance（C_i）. Thus，the C_HF can be expressed as：

According to Eqs（1-2），the fast interface state density（D_it）can be expressed as（high-low frequency method）^［12-13］：

where A is the area of gate electrode.

In this paper，the C_LF was the capacitance of MIS capacitor under 1 kHz，while the C_HF was the capacitance of MIS capacitor under 1 MHz. The D_it values of sample A and B were 2.29×10¹³ cm^-2eV^-1 and 1.83×10¹² cm^-2eV^-1 respectively，which were calculated by high-low frequency method. Apparently，the calculated D_it of sample B is an order of magnitude smaller than that of sample A. Thus，ALD-Al₂O₃ effectively reduces the D_it between Al₂O₃ and In_0.74Al_0.26As. This can be explained by the growth of Al₂O₃ deposited by ALD. Firstly，an aluminum source is introduced to effectively neutralize the dangling bonds on the surface of In_0.74Al_0.26As，and then a water source is introduced to form a bond with aluminum. Therefore，the interface state density（such as In₂O₃）between the dielectric film and In_0.74Al_0.26As is effectively reduced.

Combining the XPS test and the C-V measurement results of the MIS capacitors，it can be concluded that ALD-Al₂O₃ effectively reduces the fast interface state density between the film and the In_0.74Al_0.26As. Therefore，the lower 1/f noise and dark current density characteristics of In_0.74Ga_0.26As photodiodes passivated by SiN_x/Al₂O₃ bilayer ^［9-10］are attributed to the lower fast interface state density between the film and the In_0.74Al_0.26As.

According to the study of the interface between dielectric film and In_0.74Al_0.26As，it is found that the interface between the ALD-Al₂O₃ and In_0.74Al_0.26As is sharper than that of ICPCVD-SiN_x and In_0.74Al_0.26As. In addition，ALD-Al₂O₃ effectively reduces the In₂O₃ at the interface between ALD-Al₂O₃ and In_0.74Al_0.26As. Furthermore，the C-V results of the MIS capacitors also indicate that the ALD-Al₂O₃ effectively reduces the fast interface state density of the dielectric film and In_0.74Al_0.26As. In summary，using ALD-Al₂O₃ as the passivation film of the In_0.74Ga_0.26As photodiodes can theoretically reduce the dark current of the In_0.74Ga_0.26As photodiodes. Furthermore，the device verification results have confirmed this statement.