1.
Introduction

Silicon nitride (SiN_x) has been widely used as the dielectric materials in the processing of optoelectronics and microelectronics semiconductor devices because of its excellent chemicophysics stability and superior optical properties such as high density, low leakage, anti-oxidation, high dielectric constant and remarkable masking ability^{[1, 2]}. Most of the III–V compound semiconductors have a relatively high interfacial density, and the interface defects will produce the centers of interfacial recombination, which provide leakage channels and significantly deteriorate the electrical properties of the devices. Many studies have shown that SiN_x is an outstanding candidate as a protection and passivation layer for III–V and other semiconductors^[3–5]. For the mesa-type infrared detectors, the side-surface leakage current is one of the key components of the dark current, which reduces the detection limit of the devices. The effects of surface passivated SiN_x deposition on the performances of InGaAs detectors have been investigated^[5–6]. SiN_x passivation has also been proved convincingly effective in reducing the surface recombination leakage current and improving the current gain of AlGaAs/GaAs and InP/InGaAs heterojunction bipolar transistors (HBTs)^[7–10].



To deposit SiN_x films, several kinds of technologies exist. Compared to conventional deposition technologies such as plasma enhanced chemical vapor deposition (PECVD) and low pressure chemical vapor deposition (LPCVD), inductively coupled plasma chemical vapor deposition (ICPCVD) shows many advantages, including higher plasma density, lower temperature, lower pressure, etc^{[1, 11]}. On account of the high sensitivity to energetic ion bombardment and deposition temperature for III–V compound semiconductors, the use of conventional PECVD technology is usually disadvantageous due to the requirement of higher deposition temperature (250–400 °C)^[1] and the presence of intensive ion bombardment. Chou et al.^[12] have developed the SiN_x passivation technique at a lower-temperature using high-density ICPCVD, which utilizes inductively coupled plasma to create high-density plasma (around 10¹² cm^–3) in the processing reactor. High density plasma allows the deposition of high quality SiN_x at lower temperatures below 150 ℃^[13]. Furthermore, the separation between deposition area and plasma generation area can reduce the surface damage deriving from energetic ion bombardment. Shi et al.^{[5, 14]} has comparably studied the mesa-type n-on-p In_0.83Ga_0.17As photodiodes passivated by PECVD and ICPCVD deposited SiN_x. They found the photodiodes passivated using ICPCVD show significantly reduced perimeter-dependent surface leakage current in contrast to those using PECVD. In recent years, the ICPCVD processing has been extensively investigated, while the effects of deposition temperature, ICP power, and chamber pressure on the quality of ICPCVD SiN_x films have been reported^[11]. However, due to the complicated deposition process, the investigations on the ICPCVD processing parameters and the optical properties of SiN_x films are still insufficient. In this work, we have investigated the effects of the flow rate of ammonia (NH₃) on the surface and optical properties of SiN_x films deposited by ICPCVD. The composition, growth rate and chemical bonding of the deposited films were characterized in detail.

2.
Experimental details

The SiN_x films were deposited on InP substrates using the SENTECH SI500D ICPCVD system. Before deposition, the InP substrates were rinsed with dilute hydrochloric acid for 5 s to remove the surface oxide layer, then washed with deionized water, dried with nitrogen and loaded into the deposition chamber immediately to keep a fresh surface. The SiN_x films were deposited at 130 °C for 15 min. During the deposition, the radio frequency (RF) power was 350 W, the flow rates of silane (SiH₄) and Ar were 135 and 145 sccm, respectively. To investigate the effects of NH₃, four samples with different flow rates of NH₃ from 3 to 12 sccm were deposited, and the detailed flow rates of NH₃ for various samples are listed in Table 1.



After deposition, the surface morphology and roughness of the SiN_x films were measured by using Multimode 8 atomic force microscopy (AFM) in tapping mode. The cross-section morphology and thickness of the SiN_x films were characterized by a JEOL 7800F scanning electron microscope (SEM). The etching rates of the SiN_x films were also characterized by using a buffered oxide etchant (BOE) solution which consisted of HF and NH₄F in the ratio of 1 : 6. The optical properties of the deposited films were investigated using an X-ray photoelectron spectroscopy (XPS) THERMO's ESCALAB 250 spectrometer using an Al Kα source. The Raman spectra were measured by a HORIBA Jobin Yvon HR 800 spectrometer, where a 514 nm argon ion laser with 3 mV output power was used as the excitation light source, and the light was converged onto the surface of the sample through a lens with the focused laser beam spot of about 1 μm.

3.
Results and discussion

AFM was carried out to characterize the surface morphologies of the SiN_x samples, and Fig. 1 shows the AFM images with a 5 × 5 μm² scanning area. The RMS roughness values are extracted and indicated. As the NH₃ flow rate increases from 3 to 6 sccm, the RMS roughness decreases from 1.53 to 0.74 nm. As the NH₃ flow rate further increases to 12 sccm, the RMS value increases to 1.29 nm.






onerror="this.onerror=null;this.src='http://www.jos.ac.cn/fileBDTXB/journal/article/jos/2018/8/PIC/17110012-1.jpg'"
class="figure_img" id="Figure1"/>



Download



Larger image


PowerPoint slide






Figure1.
(Color online) AFM images of the SiN_x samples with NH₃ flow rate of (a) 3 sccm, (b) 6 sccm, (c) 9.5 sccm, and (d) 12 sccm, respectively (scanning area: 5 × 5 μm²).




The SiN_x samples were characterized by cross-section SEM, and the typical image is shown in Fig. 2(a). In the image, the clear and smooth interface between the SiN_x film and the substrate is observed. The cross-section image also shows a dense surface and homogeneous film thickness. The film thicknesses and deposition rates are then extracted from the SEM images. Fig. 2(b) shows the RMS surface roughness and deposition rate as functions of the NH₃ flow rate. The film grown with NH₃ flow rate of 6 sccm has the lowest deposition rate of 11.3 nm/min. The deposition rate is increased to 16.4 nm/min for the NH₃ flow rate of 3 sccm. As the reactants have a very low content of NH₃, the deposition of amorphous silicon (α-Si) is the dominant process, causing a larger deposition rate. As the NH₃ flow rate increases to 9.5 and 12 sccm, the deposition rate is increased to 17.4 and 18.8 nm/min, respectively. The increase of NH₃ flow rate leads to more H atoms involved in bonding to form Si–H bonds and N–H bonds, making the SiN_x film become loose and grow faster. Generally, a dense dielectric film is preferred for the passivation layer of the III–V semiconductor to reduce the surface recombination effects. Therefore, a moderate deposition rate of SiN_x film is needed in the ICPCVD process.






onerror="this.onerror=null;this.src='http://www.jos.ac.cn/fileBDTXB/journal/article/jos/2018/8/PIC/17110012-2.jpg'"
class="figure_img" id="Figure2"/>



Download



Larger image


PowerPoint slide






Figure2.
(Color online) (a) Typical cross-section SEM image of the SiN_x samples (sample 1). (b) RMS and deposition rate of SiN_x samples as functions of NH₃ flow rate.




It is also observed from Fig. 2(b) that the surface roughness and deposition rate show a quite consistent trend. The RMS roughness is increased for the samples with larger deposition rates. In the process of film deposition, the surface adsorption atoms are buried by the subsequent atoms and become a “bulk atom”, these “bulk atoms” usually are not able to diffuse again. When the deposition proceeds slowly, the atom can be fully diffused and occupy the lower energy position before it becomes a “bulk atom”, so that the free energy of the film is minimal, and the film becomes smoother and less defective. On the contrary, as the deposition rate becomes too large, once the atoms do not have enough time to sufficiently diffuse, they will be covered by newly deposited atoms and become “bulk atoms”, these atoms are essentially in the initial deposition site and not the lowest energy position. This kind of process results in the film surface with more three-dimensional islands or column crystals and causes more rough surfaces.



To further characterize the structural properties, the SiN_x samples were etched using BOE solution and the etching rates were measured and listed in Table 1. The etching rate is almost zero for the film with NH₃ flow rate of 3 sccm. The α-Si is dominant in the sample when the NH₃ flow rate is low, while the Si-rich SiN_x:H with many Si-Si bonds have a very slow corrosion rate in HF solution. In the Si-rich SiN_x:H samples, many free Si atoms fill in the interspace of the Si-N network. This will increase the compactness of the SiN_x samples and cause a smaller corrosion rate in the BOE solution. When the NH₃ flow rate is increased, more H atoms are expected to be introduced in the film. Extra H atoms incorporated in the hydrogen-rich silicon nitride layer are Si₃N₄ network breakers, which also reduce the density of the layer. Both a larger amount of H and lower layer density are tending to increase the etching rate in the HF solution^{[13, 16]}. The BOE etching rate gradually increases from 0 for sample 1 to 35 nm/min for sample 3 as the NH₃ flow rate rises from 3 to 9.5 sccm, whereas it significantly increases from 35 to 230 nm/min as the NH₃ flow rate is further increased to 12 sccm for sample 4. This implies that many extra H atoms exist and the layer density is remarkably decreased in sample 4. When the NH₃ flow rate is between 6 and 9.5 sccm, the SiN_x films exhibit nice compactness with moderate etching rates.



The optical properties of the SiN_x samples were investigated by XPS and Raman measurements. The XPS measurements were carried out for the as-grown samples and those after being etched by Ar⁺ for 30 s. Fig. 3(a) and Fig. 3(b) show the XPS scanning spectra of sample 3 before and after Ar⁺ etching while other samples show similar spectra. The obvious peaks of O 1s, N 1s, C 1s, Si 2p and Si 2s core levels are observed in the XPS spectra and the elemental concentrations are estimated from the XPS measurements. Fig. 4 shows Si, C, N and O elemental concentrations in the SiN_x samples as functions of the NH₃ flow rate. For the as-grown samples, some O and C can be observed, which become negligible in the XPS measurements after Ar⁺ etching. The O and C elements are probably due to the contamination of atmospheric oxidation. The Si/N ratio has been calculated from the Si and N elemental concentrations in the XPS measurements after Ar⁺ etching and are listed in Table 1. As shown in Fig. 4(b) and Table 1, the Si concentration reduces and the N concentration increases as the NH₃ flow rate increases. The Si/N ratio is decreased from 11.3 for sample 1 to 0.9 for sample 4. Theoretically, the Si/N ratio is 0.75 for standard Si₃N₄. The deviation of the measured Si/N ratio to the theoretical one is probably due to the relatively large measurement error of the absolute elemental concentration in XPS. Nevertheless, the relative trend of the Si/N ratio for the samples with different NH₃ flow rates is obvious.






onerror="this.onerror=null;this.src='http://www.jos.ac.cn/fileBDTXB/journal/article/jos/2018/8/PIC/17110012-3.jpg'"
class="figure_img" id="Figure3"/>



Download



Larger image


PowerPoint slide






Figure3.
(Color online) XPS spectra of sample 3 (a) as-grown, (b) after Ar⁺ etching of 30 s.






onerror="this.onerror=null;this.src='http://www.jos.ac.cn/fileBDTXB/journal/article/jos/2018/8/PIC/17110012-4.jpg'"
class="figure_img" id="Figure4"/>



Download



Larger image


PowerPoint slide






Figure4.
(Color online) The concentration of Si, N, C and O in the SiN_x samples measured by XPS as functions of NH₃ flow rate (a) as-grown and (b) after Ar⁺ etching of 30 s.

Sample No.	NH3 flow rate (sccm)	BOE etching rate (nm/min)	Si/N ratio
Sample 1	3.0	0	11.3
Sample 2	6.0	10?20	1.5
Sample 3	9.5	35	1.0
Sample 4	12.0	230	0.9

Table1.
The etching rate and Si/N ratio of SiN_x samples with different NH₃ flow rates.



Table options
-->


Download as CSV

Sample No.	NH3 flow rate (sccm)	BOE etching rate (nm/min)	Si/N ratio
Sample 1	3.0	0	11.3
Sample 2	6.0	10?20	1.5
Sample 3	9.5	35	1.0
Sample 4	12.0	230	0.9

In the XPS spectra, the different binding energies correspond to different bonding states. Karcher et al.^[17] have shown that there are five different forms of Si and N bonding states, Si, Si₃N, SiN₂, SiN and Si₃N₄, which can be expressed as Si⁰, Si⁺¹, Si⁺², Si⁺³ and Si⁺⁴, respectively. The binding energy of 102.0 eV corresponds to Si⁺⁴, and the 99.6 eV corresponds to Si⁰, the binding energies between them correspond to the bonding states of Si⁰ to Si^+4[18]. Fig. 5 presents the change of Si 2p and N 1s core-level spectra in the four samples. It can be seen that the Si 2p core-level binding energy is at 99.7 eV for sample 1, close to the bonding state of Si⁰, which indicates the main components of α-Si in sample 1. The Si 2p binding energy shifts to about 101.5 eV as the NH₃ flow rate increases to 6 sccm, and keeps around 102.0 eV for higher NH₃ flow rates, indicating the bonding state of Si⁺⁴ for samples 2?4. In the α-Si crystal structure, each Si atom bonds with the other four nearest Si atoms. As NH₃ is added continuously, an increasing number of the homopolar Si–Si bonds will gradually be replaced by heteropolar Si–N bonds. The charge transferring from Si atom to the more electronegative N atom leaves a positive charge on the original Si atoms, which results in the Si core levels shifting towards higher binding energy^[17]. On the other hand, the N 1s core-level binding energy stabilizes at around 397.4?397.9 eV for the four SiN_x samples with various NH₃ flow rates, indicating that there is no significant change in the state of the N element: the N atoms mainly bond with Si atoms and form SiN_x.






onerror="this.onerror=null;this.src='http://www.jos.ac.cn/fileBDTXB/journal/article/jos/2018/8/PIC/17110012-5.jpg'"
class="figure_img" id="Figure5"/>



Download



Larger image


PowerPoint slide






Figure5.
(Color online) XPS spectra of Si 2p and N 1s core levels for the samples with NH₃ flow rate of (a) 3 sccm, (b) 6 sccm, (c) 9.5 sccm, and (d) 12 sccm, respectively.




Fig. 6 shows the Raman spectra of the SiN_x samples. It is observed that samples 1 and 2 show four obvious regions, which represent the Raman spectral characteristics of the α-Si, as follows. The Raman frequencies around 100?180 and 420?550 cm^?1 correspond to the transverse acoustic (TA) and transverse optical models of α-Si, respectively. The frequencies around 580?700 cm^?1 and 850?1050 cm^?1 probably correspond to the twofold longitudinal acoustic (2LA) and twofold transverse optical (2TO) models of α-Si^{[19, 20]}. The Raman frequency at 355 cm^?1 is the typical LO model of InP, which comes from InP substrate. As the NH₃ flow rate increases, the Raman peaks of α-Si decrease significantly and become quite weak for samples 3 and 4. The Raman results also indicate the main components of α-Si for the SiN_x samples deposited with low NH₃ flow rates and the decreased α-Si content for larger NH₃ flow rates.






onerror="this.onerror=null;this.src='http://www.jos.ac.cn/fileBDTXB/journal/article/jos/2018/8/PIC/17110012-6.jpg'"
class="figure_img" id="Figure6"/>



Download



Larger image


PowerPoint slide






Figure6.
(Color online) Raman spectra of various samples with NH₃ flow rate of 3 sccm (Sample 1), 6 sccm (Sample 2), 9.5 sccm (Sample 3) and 12 sccm (Sample 4), respectively.

4.
Conclusion

In conclusion, the effects of ICPCVD NH₃ flow rate on the surface and optical properties of SiN_x samples have been investigated. XPS and Raman measurements have revealed that α-Si is dominant in the samples with low NH₃ flow rates and the α-Si content is decreased for larger NH₃ flow rates. AFM and SEM results show that low NH₃ flow rates induce the increased surface roughness because of the dominant α-Si, whereas high NH₃ flow rates also enhance the surface roughness due to the higher deposition rate. Moreover, the bonding states of the Si and N atoms are investigated by using XPS analysis, which reveals that the Si atoms transfer from Si⁰ to Si⁺⁴ with the increased NH₃ flow rate.