Review—Silicon Nitride and Silicon Nitride-Rich Thin Film Technologies: State-of-the-Art Processing Technologies, Properties, and Applications

The Fourth Industrial Revolution, or Industry 4.0, is projected to have a radical impact on multiple industrial sectors, including the integrated circuitry (IC), optoelectronics, aircraft, energy, micro-electro-mechanical systems (MEMS), digital, medical, biological, chemical, and defense sectors.⁹⁰ Briefly, the Fourth Industrial Revolution builds on the Third Industrial Revolution (or the Digital Revolution) to develop innovations that unify the physical, biological, medical, and digital worlds.

The resulting effects are especially exemplified in the IC industry, which, while continuing its historic evolution towards higher performance processors and increased density memory chips, is placing greater emphasis on the development of hetero-devices and systems that will serve as essential technological drivers for a wide spectrum of industrial applications. Due to this transition, the introduction of new materials and their integration with currently used ones is projected to replace IC design and device scaling as the key enabler of improved device performance and larger density gains.^91,92

One such material system is the Si–N thin film material system in all its forms (SiN_x, SiN_x:H, SiN_x(C) and SiN_x:H(C)). Si–N continues to be the subject of intense R&D activity for applications in multiple industries. This activity focuses not only on examining how to extend Si–N's current usages, but also on exploring its promising role as a base platform the incorporation of new material systems, especially in the IC, optoelectronics, and solar cell industries.

Use of the Si–N thin film material system is projected to endure in a wide array of IC applications, including microprocessors, system-on-a-chip (SoCs), flash memory, and three-dimensional (3D) integrated systems. These include diffusion barriers, capping layers, and etch stops in traditional IC interconnect schemes, as well dielectrics in metal-insulator-metal capacitors and thin film transistors (TFTs). Similarly, SiN_x is expected to maintain its role as a permeation barrier, passivation layer, and encapsulation layer in optical, optoelectronic, and electroluminescent devices, such as flat panel displays, lighting, light-emitting devices (LEDs), organic LEDs (OLEDs), planar optical systems, and optical waveguides. In the green energy field, SiN_x plays an equally important and diverse role in solar cell applications, including passivation and cap layers. SiN_x also acts as an effective host matrix for Si nanocrystal light emitters. In biotechnology and medical fields, SiN_x is anticipated to sustain and expand its current role, particularly as a protective coating for in vivo and in vitro environments, including viewing windows for medical devices and insulating membranes for cell electroporation, as well as in biosensors for a variety of health-related applications.

Indeed, a review of the open literature demonstrates that the R&D interest in the Si–N thin film material system continues unabated. This review further shows a clear progression in SiN_x processing technologies from physical vapor deposition (PVD) to CVD and ALD methodologies, both in thermal and plasma-assisted modes. As expected, this shift has been primarily driven by the move toward more complex device structures with ever decreasing dimensions, leading to the need for much higher precision and tighter control in the formation of what is commonly referred to as "zero-thickness" SiN_x films with precisely-tuned composition and morphology.

A variety of CVD and ALD techniques have been explored for SiN_x deposition. In the case of CVD, these include thermal, hot wire (HW-CVD), radiofrequency (RF) capacitively coupled plasma-enhanced (PE-CVD), remote plasma-enhanced CVD, and inductively-coupled PE-CVD. In the case of ALD, they include thermal, plasma-assisted (PA-ALD), plasma-enhanced (PE-ALD), and remote plasma ALD.

Interestingly, only very recent work has reported on the development and application of SiN_x using what is commonly referred to as pulsed CVD. Pulsed CVD is the equivalent of ALD, except that the process is performed at a substrate temperature that leads to partial or complete decomposition of the pulsed precursor upon engagement with the substrate during every exposure cycle, as opposed to being restricted to the typical ALD physisorption or chemisorption reaction.⁹³ A co-reactant is subsequently introduced to complete the decomposition reaction and/or remove the reaction byproducts to ensure a clean film. What is more puzzling, no R&D has been reported on alternatives to ALD for near-zero-thickness SiN_x by other self-limiting processes, such as self-assembled monolayer (SAM) deposition and molecular layer deposition (MLD).⁹⁴

Table I lists physical, chemical, electrical, optical, and mechanical properties for CVD and ALD SiN_x as compiled from the open literature. It has been updated and expanded from Ref. 7 to include data from 2017 through 2019. The data presented clearly demonstrates that the various characteristics of CVD and ALD SiN_x are highly dependent upon the deposition technique employed and the resulting film texture and morphology, as well as the concentration and bonding configuration of H in the films. Wet etch rate (WER) has been included since it is increasingly considered a figure of merit for SiN_x films. While only recently reported data has been compiled, we note that the values and trends are generally consistent with earlier reports.⁹⁵ Bulk silicon nitride properties are included for reference.

Appendix A summarizes theoretical ALD modeling and mechanistic studies that are designed to examine the effects of Si precursor chemical structure and bonding configuration, N source reactivity in thermal and plasma environments, and substrate surface chemistry and pre-deposition treatment recipe on the adsorption and decomposition pathways of Si precursors and resulting SiN_x morphology, composition and properties. No similar CVD reports were published during the same time period. In particular:

• 1.  
  A. Dangerfield et al. (2017) employed a density functional theory (DFT) approach to perform first principle calculations simulating the reaction of DSBAS [di(s-butyl)aminosilane with hydroxyl-terminated oxidized Si surface under two scenarios: (i) a direct DSBAS pulse, and (ii) a trimethylaluminum (TMA) pulse followed by a DSBAS pulse. In the first scenario, the growing SiN_x film was represented as a Si₄N₆ cluster terminated by N−SiH₃ species; in the second scheme, N−Al(SiH₃)₂ configurations were used as SiN_x surface coverage. The calculations revealed that the presence of aluminum (Al) increases the adsorption density of molecular N₂H₄ onto the OH-terminated oxidized Si surface, resulting in the reactivity of N₂H₄ with DSBAS and leading to the inclusion of N in Si.⁶⁶
• 2.  
  X. Meng et al. (2018) also applied a density functional theory (DFT) protocol to examine the reactivity and growth rates of Si precursors on p-type Si(100) surfaces. They theorized that three bonding configuration considerations decrease the surface energy barrier to SiN_x nucleation and growth: (i) lower steric hindrance, (ii) reduced bond dissociation energies, and (iii) higher electron densities. They argued that the substitution of H atoms for Cl atoms would decrease the steric hindrance and allow the precursor molecule higher accessibility to reactive substrate surface sites with less physical intervention from the adjacent ligands. Partial substitution also increases the polarity of the molecule and enhances dipole−dipole interactions with substrate surface polar groups, which would (i) bring about higher electron density on the Cl atoms, thus improving the reactivity of the Si−Cl bonds, and (ii) generate another potential precursor adsorption and reaction pathway via Si−H bond cleavage. They cautioned, however, that increasing the number of H atoms in the precursor molecule could lead to higher H inclusion in the resulting SiN_x films due to the presence of less reactive Si−H groups from the parent precursor. In summary, their model predicted that the reactivity of disilane type precursors, such as hexachlorodisilane (HCDS, Si₂Cl₆), is higher than their monosilanes counterparts, with additional reactivity enhancement achieved by replacing a Cl atom in HCDS with a H atom to produce pentachlorodisilane (PCDS, HSi₂Cl₅).¹⁸
• 3.  
  Similarly, J.-M. Park et al. (2018) used density functional theory (DFT) simulations to obtain the lowest energy barrier to precursor adsorption and decomposition through geometry optimization of the substrate surface structure. They then used their findings to derive bond dissociation energy (BDE) values, reaction energy, and the energy barrier to precursor adsorption and reaction for the cyclosilazane type precursor 1,3-diisopropylamino-2,4-dimethylcyclosilazane (CSN-2). The calculations were carried out on an undercoordinated bare β-Si₃N₄ surface (>Si=N−) constructed to simulate an N₂-plasma-treated SiN_x surface. They explained that a more accurate approach to simulating the ALD process would be to model amorphous SiN_x with dangling bonds and H impurities. Such an approach would require an impractically high computing power, however, since a large unit cell with random structures is needed to represent amorphous materials.²⁷

Appendix B presents a summary of deposition parameters, post-processing treatments (if any), and key findings for SiN_x thermal, catalytic, and roll-to-roll CVD, as well as plasma-enhanced and mirror-plasma-enhanced CVD (PE-CVD). Table III and Appendix B present the following key experimental trends and major results:

• 1.  
  CVD and PE-CVD research continues in an effort to extend the use of SiN_x for traditional applications, such a

  • Stand-alone and embedded memory, due to its spin transport, magnetoresistance, and magnetic resonance properties.⁴⁷
  • Protective layers/etch stops for metal lines and polysilicon structures, stabilization layers, and active device structures in MEMS.²⁶
  • Antireflection coating and passivation layers in Si solar cells.^5,41,49
  • Protection/encapsulation layers, sacrificial layers, diffusion barriers, and functional structures, including top encapsulation layers in IC devices and three-dimensional (3D) devices.²²
  • Protection/encapsulation layers and spacers in DRAM.³²
  • Insulating layers in organic thin-film transistors and OLED.
• 2.  
  CVD and PE-CVD are also being explored for a myriad of new SiN_x applications including:

  • Host matrix for Si nanocrystals (Si-NCs) and quantum dots (QDs), deuterated SiN (SiN:D) waveguide in non-linear integrated photonics, low-loss SiN_x for SiN waveguide buses, and bottom and encapsulation layer for embedded QDs in a triple SiN_x/QD/SiN_x stack structure for on-chip quantum dot QD/SiN lasers in photonic integrated circuits (PICs) for applications in Si nanophotonics systems, photoluminescence, photovoltaics, data storage, and optoelectronics.^{36,40,45,50,51,54,57}
  • CMOS-compatible amorphous film with a bulk second-order nonlinearity through strong second-harmonic generation (SHG) for on-chip photonic devices.³⁴
  • Gold (Au) nanorods embedded in nano-structured SiN_x film to improve the reliability of MEMS switches with respect to dielectric charging phenomena in capacitive RF MEMS switches.⁴⁸
  • Single-layer SiN_x films on polyethyleneterephthalate as a moisture permeation barrier film for electronic devices on flexible substrates for incorporation into organic solar cells, QD displays, and organic light-emitting diode (OLED) displays.²³
  • Active light emitting Si-rich and N-rich layers for visible spectrum light emitting diodes (LEDs).⁴⁴
  • Passivation and dielectric layers in interdigitated back-contact (IBC) Si solar cells.⁴⁹
  • Waveguides in nonlinear frequency combs and supercontinuum generation, spectroscopy and sensing, and photonic devices for telecommunications.³⁸
  • Charge trap layers in 3D-NAND flash memory devices, gate spacers in FinFETs, stoppers in (CMP) in a self-aligned multiple patterning process when forming fins in FinFETs.^24,73
  • Tunable luminescent film for Si-based LEDs in Si based monolithic optoelectronic integration.⁵⁹
  • Membrane in microcapsules for transmission electron microscopy (TEM) analysis.⁶¹
  • a-SiN_x:H matrix as a host for light-emitting Si nanocrystals in a heterojunction p + in + diode for an electroformed light emitting memory device consisting of a glass/Cr/p⁺in⁺/ITO structure.⁵³
  • SiN_x and SiN_xO_y/SiN_x stacks as passivation layers for photodiodes in predictable quantum efficient detectors (PQEDs).⁴²
  • a-SiN_x thin film electrodes for lithium ion batteries.²¹
• 3.  
  PE-CVD SiN_x from inorganic Si sources, the reaction of silane (SiH₄) with an ammonia (NH₃) plasma as the chemical sources for Si and N, respectively, continues to be the primary SiN_x deposition methodology. Films are deposited in the substrate temperature range of 300 °C–400 °C, and their composition from Si-rich to stoichiometric to N-rich is typically controlled by adjusting R (SiH₄/NH₃ flow ratio) or (NH₃/SiH₄ flow ratio).^{26,32,33,39,45,49,53} H inclusion is dependent on the processing parameters employed, with values reported in the range of 20% to 40% for as-deposited SiN_x films. As expected, post-deposition thermal treatment led to a decrease in H content, although no quantitative data was provided. H effusion/release mechanisms for PE-CVD a-SiN_x from the reaction of SiH₄ and NH₃ were analyzed and quantified at different temperatures. The authors claim that these mechanisms can be represented as follows⁶²:

  $$\begin{eqnarray*}\begin{array}{ll}{\rm{For}}\,{\rm{H}}\,{\rm{diffusion}}: & \,:{\rm{Si}}+\mathrm{Si} \mbox{-} H\to \mathrm{Si} \mbox{-} \mathrm{Si}+{\rm{H}}\\ & \,:{\rm{Si}}+N \mbox{-} H\to \mathrm{Si} \mbox{-} N+{\rm{H}}\\ {\rm{For}}\,{{\rm{H}}}_{2}\,{\rm{surface}}\,{\rm{desorption}}: & \,2\mathrm{Si} \mbox{-} H\to 2{\rm{Si}}+{{\rm{H}}}_{2}\\ & \,2N \mbox{-} H\to 2{\rm{N}}+{{\rm{H}}}_{2}\end{array}\end{eqnarray*}$$

  The authors also reported that effusion from SiN_x films grown at low temperatures is generally associated with desorption of H₂ molecules through voids. However, in denser SiN_x structures grown at higher deposition temperatures, hydrogen is released in a non-bonded form through a diffusion mechanism.⁶²
• 4.  
  Efforts to lower H concentration in as-deposited inorganic PE-CVD SiN_x films include the work of J. Chiles et al. (2018), who explored the growth of PE-CVD SiN_x at 275 °C from deuterated silane (SiD₄) and NH₃ as Si and N sources, respectively. The authors argued that the use of deuterated silane (SiD₄) would eliminate high H content in the resulting PE-CVD films, and thus would minimize optical transmission losses from the N–H bonds overtone absorption peak located between 1500 and 1550 nm. The work produced high-quality deuterated SiN_x (SiN_x:D) with significantly lower interference from N–H bonds in the S&C telecommunications bands.³⁶
• 5.  
  For LP-CVD SiN_x from inorganic Si sources, a number of inorganic Si course precursors were examined. All LP-CVD processes require very high temperatures (>800 °C) to form SiN_x, which prevents their applicability to thermally fragile substrates. Examples include:

  • Dichlorosilane (DCS, SiH₂Cl₂), which in one report is reacted with NH₃ at 830 °C to grow stoichiometric SiN_1.33 for luminescence studies. The films are subsequently implanted with N and subjected to rapid thermal annealing (RTA) at 1200 °C for 3 min, resulting in increased luminescence in the blue-green region.^33,44,58
  • Pentachlorodisilane (PCDS, HCl₂SiSiCl₃), which in one instance is reacted with (NH₃ + H₂) at 800 °C–1400 °C to grow practically H-free and stoichiometric SiN_1.33. Films were amorphous up to 1000 °C and crystalline α-Si₃N₄ at 1200 °C and above.⁶⁰

  In either case, it is believed that, in comparison to SiCl₄, replacing Cl atoms with H atoms lowers the steric hindrance and enables higher accessibility of the precursor molecule to reactive substrate surface sites with less physical interference from adjacent ligands. The partial replacement of Cl atoms by H atoms also increases the polarity of the molecule to enhance dipole−dipole interaction with substrate surface polar groups and could simultaneously: (i) lead to higher electron density on the Cl atoms, thus enhancing the reactivity of the Si−Cl bonds, and (ii) generate an alternate precursor adsorption and reaction pathway via Si−H bond cleavage.⁶⁰ The reviewers also note that in the case of the disilanes, there is a tendency for these materials to develop explosive hydrolysis gels in practice,^96,97 with published and private industry communications suggesting that pentachlorodisilane has a higher propensity to form such dangerous species than hexachlorodisilane.
• 6.  
  In terms of non-traditional reactor geometries and novel SiN_x deposition processes, the following system configurations and architectures are noteworthy:

  • Catalytic CVD (Cat-CVD) reactor, also known as hot-wire CVD (HW-CVD). The basic strategy was to crack the gaseous reactants via catalytic decomposition reactions by passing them through a hot catalyzer located in close proximity to the substrates.⁶³ It is argued that this approach would enable film deposition at much lower temperatures than in conventional thermal and PE-CVD. In this specific embodiment, the (SiH₄ + NH₃) reactant mixture was passed over a tungsten (W) wire catalyzer that was maintained at 1800 °C prior to deposition on the substrate, which was heated to 100 °C, to yield SiN_x films. The reviewers note that catalytic CVD seems to be a misnomer, since it suggests the occurrence of a catalytic process that provides controlled pathways to reaction intermediates rather than thermally-induced precursor decomposition. In fact, the work does not describe a catalytic CVD technique in which the wire composition drives the actual deposition. For that reason, hot wire CVD would be a more accurate term.
  • Roll-to-Roll PE-CVD reactor. The reactor consisted of two major components: (i) a load-lock chamber to load, unwind, and rewind the flexible polyethylene terephthalate (PET) substrate, which included a cold trap for moisture removal and an RF generator for pre-deposition substrate cleaning; (ii) a deposition chamber made up of a central drum with controllable temperature and four linear microwave plasma sources, each equipped with a gas shower head for introducing the (SiH₄ + NH₃ + Ar + NF₃) reactant mixture. The system was employed to grow films with a composition ranging from SiN_0.81 for R (NH₃/SiH₄ flow ratio) = 2 to SiN_0.93 for R = 9, at substrate temperature below 45 °C.²³
  • Magnetic-mirror confined plasma source PE-CVD reactor (MPE-CVD). In this case, a Nd–Fe–B permanent magnet was applied to generate and focus the 5.85 GHz electron cyclotron resonance (ECR) magnetic mirror field. A primary benefit of this technique is that it minimizes plasma-induced ion-bombardment damage to substrate and film during processing. Using this system, the authors reported the deposition of SiN_x films at room temperature (RT) with a Si/N ratio ranging from 0.5 to ∼1.2 for R (SiH₄/NH₃ flow ratio) varying from 0.08 to 0.12. H concentration in the films decreased from 200 °C–400 °C with higher substrate T.³⁹
  • Laser-assisted CVD using an ultraviolet laser frequency to grow SiN_x films while minimizing c-Si and polyethylene terephthalate (PET) substrate damage.³⁵
  • PE-CVD reactor with a new showerhead geometry that replaces the air gap between the powered electrode (showerhead) and the ground electrode in a standard parallel plate plasma design with an aluminum (Al) plate. The plate was covered with a thick anodized coating containing concentric holes aligned with the showerhead holes. The resulting configuration was applied in PE-CVD SiN_x from hexamethyldisilazane (HMDSN, C₆H₁₉NSi₂) and Ar/N₂ mixture as carrier gas and co-reactant. The net effect was a reduction in H inclusion in the resulting SiN_x films compared to the standard parallel plate electrode architecture, as demonstrated by a significant increase in Si–N bond density as well as a more modest increase in N–H and Si–H bond densities.⁶⁴
• 7.  
  Other reported forms of the Si–N material system of interest grown by CVD or PE-CVD include:

  • PE-CVD of C-containing SiN_x using trimethylsilane (HSi(CH₃)₃) and NH₃ as Si and N sources, respectively, for use as a capping layer in IC Cu metallization schemes. The authors report the successful formation of a Si_29.3C_20.1N_21.7:H_28.9 phase through the addition of H₂ to the PE-CVD reaction, vs Si_27.3C_17.7N_20.9:H_34.1 prior to the inclusion of H₂ as co-reactant. The phase with the lower H content demonstrated improved robustness as a Cu capping layer. We note that these compositions would be more appropriately described as a silicon carbonitride (SiC_mN_n) phase. Film properties and structures can also be compared to earlier work reported for atmospheric pressure PE-CVD (AP-PECVD) from triethylsilane and N₂.⁹⁸
  • PE-CVD of a-SiN_x:O_y from 5% SiH₄ diluted in N₂ and (NH₃ + N₂) as co-reactants to form a-SiN_x, followed by in situ oxidation in O₂ plasma to form a-SiN_x:O_y for applications as an active layer in Si-based LEDs. The researchers state that the highest absolute photoluminescence quantum yield (PL QY) from tunable luminescent a-SiN_x:O_y is 8.38%, which was achieved at a PL peak energy of 2.55 eV.⁵⁹
  • PE-CVD of SiN_x and dual-layer SiN_xO_y/SiN_x stack as a capping layer in predictable quantum efficient detectors (PQEDs). Both types of capping layers displayed very good passivation properties post annealing, with the double layer stack exhibiting the best passivation, and yielding an effective carrier lifetime near 20 ms.⁴²

Appendix C presents a summary of deposition parameters, post processing treatments, if any, and key resulting findings for SiN_x thermal, plasma-enhanced, and remote plasma-enhanced ALD (PE-ALD). Table IV and Appendix C present the following key experimental trends and major results:

• 1.  
  ALD and PE-ALD are being explored primarily in efforts to extend the use of SiN_x in IC applications, such as:

  • Diffusion barriers, passivation layers, etch and chem-mechanical polishing (CMP) stops, spacers, dielectric layers, and hard masks in IC devices.^18,41,67,70
  • Gate spacers and sidewall spacers in DRAM.²⁸
  • Charge trap layers in 3D-NAND flash devices.⁷³
• 2.  
  ALD and PE-ALD are also being explored for a very limited number of new SiN_x applications, such as passivation/encapsulation layers in the form of Si–N nanostructures at the interface with GaAs to protect them against moisture, primarily in the form of hydroxyl groups.⁷²
• 3.  
  For PE-ALD from inorganic Si sources, the following Si source precursor are employed:

  • SiH₄ which was reacted with a NH₃ plasma to deposit SiN_x at substrate temperatures in the range of 200 °C–450 °C. This is surprising since it is contrary to the R&D trends from 2010–2016, wherein very few reports described the use of SiH₄ as Si source. In a few cases, the SiN_x layer grown by the SiH₄-based PE-ALD process was exploited as a baseline material to produce C-containing SiN_x (see paragraph 7).^68,72
  • A hydridohalosilane, namely, pentachlorodisilane (PCDS, HSi₂Cl₅), which was reacted with NH₃ or N₂/10%H₂ plasma at substrate temperatures in the range of 270 °C–360 °C. The SiN_x films deposited at 360 °C had low O content (∼2 at.%) and negligible Cl content (<1 at.%). The authors explain that the underlying surface reaction pathways were: (i) pre-deposition plasma nitridation to produce a substrate surface terminated with amine groups (−NH₂ and −NH−); (ii) PCDS was then pulsed in. It reacted preferably with −NH₂ groups and chemisorbed on substrate surface, releasing either H₂ or HCl; (iii) in the plasma cycle, Cl ligands and H atoms from adsorbed PCDS were removed by plasma reactive species, thereby returning the substrate surface to its coverage with reactive amine groups; and (iv) −NH− groups were embedded into SiN_x as the primary source of H bonds.¹⁸
  • Hexachlorodisilane (HCDS, Si₂Cl₆) was examined as replacement for SiH₄, and is reacted with NH₃ or N₂/10%H₂ plasma at substrate temperatures in the range of 270 °C–360 °C.¹⁸ The resulting films had similar composition and properties to those deposited using PCDS, expect that the latter exhibited higher reactivity and enhanced growth rates in comparison with HCDS.
  • Similarly, HCDS was reacted with a N₂/NH₃ plasma (flow ratio of 30/90) to yield Si_0.49N_0.48:O_0.03 at 360 °C.²⁴ Prior to deposition, surface nitridation was performed by pretreating the H-terminated Si substrate surface with NH₃ plasma. In addition, H content in the films was reduced through: (i) greater thermal activation by applying higher substrate temperatures, or (ii) higher density of plasma excited species by increasing plasma power.

  As discussed in Appendix A, Si₂Cl₆ has a reduced energy barrier compared to tetrachlorosilane (SiCl₄) for reactions with amine groups (−NH₂ and −NH−) terminated substrate surfaces. The reviewers also refer the reader to section 5 for a discussion of the effects of substituting H for Cl.
• 4.  
  Regarding co-reactants, NH₃ or N₂ continue to be the primary source of N in PE-ALD SiN_x (see Table IV). A number of reports employed anhydrous hydrazine (N₂H₄) instead, although the reader is cautioned that it suffers from high toxicity and chemical instability.⁶⁷ Surprisingly, no work was reported on any other N source, including t-butylhydrazine (C₄H₁₂N₂), which could exhibit an appreciably lower dissociation energy than N₂, thus making ALD SiN_x feasible at lower substrate temperatures than N₂.
• 5.  
  Very few reports were published on thermal ALD since 2016. The source chemistries employed are:

  • A halosilane: namely, hexachlorodisilane (Si₂Cl₆), was reacted with hydrazine at 285 °C substrate temperature. Prior to deposition, the substrates were exposed to 400 ml N₂H₄ at 285 °C to terminate the surface with NH_x groups. Post-deposition, samples were subjected to atomic H at a flow of 1800 liters under vacuum at 285 °C for 15 min, removing any residual Cl in the SiN_x film through the formation of an HCl type desorption byproducts.⁶⁷
  • An aminosilane, (di-s-butylamino)silane (DSBAS), ((s-Bu)₂N)SiH₃), which was reacted with hydrazine at 250 °C. Prior to deposition, a trimethylaluminum (TMA) pulse was applied to pre-treat the OH-terminated oxidized Si(111) with an Al monolayer.⁶⁶ The Al monolayer was applied to catalyze the reaction of N₂H₄ with -SiH₃ surface sites due to the lack of reactivity of N₂H₄ with the untreated Si–O–SiH₃ surface. However, the SiN_x film growth was limited to 1 nm. The TMA pulse was therefore repeated after every individual set of DSBAS/N₂H₄ cycles to deposit another Al monolayer and enable the growth of a thicker SiN_x film. The authors report that the most likely reactivity pathway is the insertion of N into the Al-Si bonds. The resulting Al-SiN_x interleaved film of varying atomic composition is strictly not a SiN_x phase, and will undoubtedly exhibit properties that cannot be equated to SiN_x.
• 6.  
  For PE-ALD from organic Si sources, the following Si precursors were employed:

  • Trimethylsilane ((CH₃)₃SiH), which was reacted with NH₃ plasma at substrate temperatures in the range 300 °C–450 °C.¹⁷ SiN_x film composition evolved from Si rich to stoichiometric Si₃N₄ with higher NH₃ gas flow and plasma power. Barrier properties against Cu diffusion for ALD SiN_x are equivalent to PE-CVD SiN_x, even though the ALD films exhibited a lower density.
  • Bis(dimethylaminomethylsilyl)trimethylsilylamine (DTDN₂-H₂), which was reacted with N₂ plasma at substrate temperatures in the range 100 °C–500 °C. The optimum process window was identified for substrate temperatures in the range of 250 °C–400 °C. SiN_x grown below 350 °C oxidized upon exposure to air due to low density, which was attributed to low deposition temperature. Near-stoichiometric SiN_1.35 is formed at 400 °C with ∼5% O.^25,41
  • 1,3-di-isopropylamino-2,4-dimethylcyclosilazane (CSN-2), was studied in two reports:
    • •  
      In the first report, it was reacted with a NH₃ plasma at substrate temperatures in the range 200 °C–500 °C. CSN-2 was selected because of its closed ring structure, which is conducive to good thermal stability (up to 450 °C) and high reactivity. A 3-step PE-ALD process was employed to improve film quality and step coverage on the lower sidewall of device structures: (i) CSN-2 pulsing step (ii) NH₃/N₂ plasma step to allow H radicals to remove CSN-2 ligands; and (iii) N₂ plasma step to remove surface H and activate precursor adsorption. The authors reported achieving the highest quality SiN_1.0 at 500 °C substrate temperature, with the inclusion of 7at% H and <2at% O.²⁷
    • •  
      In the second report, CSN-2 was employed in a three-step remote plasma ALD (RP-ALD process consisting of: (i) CSN-2 pulsing step, (ii) H₂ plasma pulsing step, and (iii) N₂ plasma pulsing step.²⁸ The authors argue for the use of H₂ plasma due to the fact that N radicals exhibit an exceptionally short lifetime vs H radicals and/or O radicals. As a result, recombination loss occurs, and ligands are not completely removed from the adsorbed precursor, leading to deterioration in film quality. The introduction of an H₂ plasma step prior to a N₂ plasma step led to efficient ligand removal from the substrate surface due to the long lifetime of the H radicals. The authors also explain that PE-ALD of SiN_x from aminosilane and NH₃ plasma produces low growth rates because the H- and NH_x-terminated surface is not undercoordinated, which prevents precursor adsorption. Alternatively, N₂ plasma can produce reactive undercoordinated surface sites with dangling bonds. As a result, PE-ALD using aminosilane and N₂ plasma leads to reasonable growth rates. The three-step RP-ALD process was reported to produce stoichiometric Si₃N₄ for substrate T ranging from 250 °C and 500 °C, with C and O content decreasing with higher substrate T from ∼4at% and 10at% to ∼0at% and ∼2at%, respectively.
  • Two aminosilanes: namely, bis(t-butylamino)silane (BTBAS), a dihydridosilane, and (di-s-butylamino)silane (DSBAS), a trihydridosilane, were selected to examine precursor and co-reactant (Ar or N₂ plasma) reactions with an H-Si(111) substrate surface to distinguish plasma-induced effects from surface chemical processes. To this end, DSBAS was used to analyze the reaction pathways and mechanisms induced by Ar plasma phenomena, while BTBAS was employed to study SiN_x PE-ALD growth in N₂ plasma. The authors report that: prior to Ar plasma exposure, the Si(111) surface is unreactive to DSBAS, even at temperatures as high as 300 °C; this implies that the formation of a reactive site alone is not sufficient, and the presence of neighboring H (Si−H) is needed to release the amino ligand and chemisorb −SiH₃ on active sites (probably Si dangling bonds or undercoordinated N). Exposure to Ar plasma caused desorption of H atoms from the H-Si surface, leaving Si dangling bonds that enable reaction with DSBAS. The authors also report that a pre-deposition 5 min exposure to N₂ plasma led to substrate surface saturation, both in terms of surface nitride layer thickness and N incorporation. Subsequent PE-ALD investigations at substrate temperatures of 100 °C and 200 °C and 175 and 275 W N₂ plasma power indicated that the SiN_x phase with the least C, O, and H inclusion is achieved at 200 °C and 175 W. Lower substrate temperature led to C contamination, while higher plasma power led to increased O inclusion.^70,73
• 7.  
  Other reported ALD-grown forms of the Si–N material system of interest consist of C-containing SiN_x films, as reported by R.A. Ovanesyan et al. (2017), in which PE-CVD SiN_x was used as a base platform in two- and three-step processes to grow the C-containing phase, as summarized below.⁶⁹

  • 2-step PE-ALD process: (i) PE-ALD SiN_x from the reaction of SiH₄ and NH₃ at 350 °C substrate temperature, followed by (ii) a PE-ALD SiCl₂(CH₃)₂ step at 475 °C substrate temperature. SiCl₂(CH₃)₂ was observed to interact with the –NH_x (x = 1, 2) terminated surface formed by NH₃ plasma exposure, with –CH₃ groups being removed from the precursor during its adsorption step, leading to only a small C concentration in the SiN_x films.
  • 3-step PE-ALD process: (i) same PE-ALD SiN_x process as above, followed by (ii) PE-ALD a-Si:H from the reaction of SiH₄ and H₂ plasma at 200 °C substrate temperature to form a -SiH_x (x.1, 2, 3)-terminated surface, then (iii) exposure of a-Si:H to CH₃I at 400 °C substrate temperature. CH₃I was observed to interact with surface –SiH_x (x = 1, 2, 3) to form surface –SiI_x (x = 1, 2, 3) groups with the release of CH₃.
  • 3-step PE-ALD process: (i) same PE-ALD SiN_x process as above, followed by (ii) PE-CVD SiN_x from the reaction of Si₂Cl₆ and Ar/NH₃ plasma at 350 °C substrate temperature, then (iii) exposure to Al(CH3)3 at 320 °C substrate temperature. Al(CH₃)₃ was observed to interact with the Cl-terminated Si surface to form–SiCH₃ species. Residual Al stayed on the surface in the form Al(CH₃)_x (x = 1, 2) groups that were not entirely released as reaction byproducts in the form of Al(CH₃ )_x Cl_3-x (x = 1, 2).

Due to the gradual shift in IC fabrication from PVD to CVD and ALD methodologies, few relevant reports on reactive magnetron sputtering (RMS) of SiN_x were published between 2017 and 2019. Interest in sputtering of the related material, SiAlON (in which Al and O atoms partially substitute Si and N, respectively) for luminescent applications has emerged but is not reviewed herein.^99,100 Table V and Appendix D present the following key experimental trends and major results:

• 1.  
  The purpose of the PVD SiN_x work reported within the 2017–2019 time period is to explore its use as:

  • A glazing layer, diffusion barrier, or corrosion resistance film within glass stacks for the glass industry.⁷⁴
  • An antireflection coating in solar cells and passivation layer/mask to protect from local oxidation of Si surfaces in optical waveguides.⁷⁵
• 2.  
  Electrochemical impedance spectroscopy (EIS) and transmission electron microscopy (TEM) analyses of glass/SiN_x/NiCr/Au/SiNx stacks was carried out to analyze through-porosity in SiN_x coatings. EIS and TEM findings indicate the existence of nanopores with a diameter ranging from 0.5 to 2.0 nm.⁷⁴
• 3.  
  As presented in Appendix D, the films deposited for anti-reflection coating applications are non-stoichiometric SiN_x, and exhibit a mixture of three types of defects: (i) optical inhomogeneity (as determined by the refractive index profile across the films); (ii) uniaxial anisotropy with a perpendicular optical axis to the boundaries; and (iii) random roughness of the upper boundaries.⁷⁵

The surface of SiN_x films, which can be defined as the topmost atomic layer of a SiN_x-rich film, has attracted increased attention for its influence on: (i) overall functional performance, such as enhanced capping and passivation properties; (ii) nanoscale engineering to enable transition to other material phases, such as coupling to biomaterials for specialized heterodevices. More specifically, the ultimate functional objectives dictate the actual techniques used to tailor the SiN_x surface in order to control its interfacial behavior in the formation of the subsequent layer of interest. These objectives can include adhesion, surface activation/suppression (in what is commonly referred to as area-specific deposition (ASD)), passivation, capping, or coupling to biological materials. Surface tailoring methods can be classified into two categories: aqueous and non-aqueous. Table VI and the discussion below present a survey of the purpose and application of both types of customized surface engineering techniques.

• 1.  
  Aqueous SiN_x surface modification techniques are being explored for SiN_x use as:

  • Protective coating for tribological components that require high reliability for automotive and aerospace applications.⁷⁶
  • Protection/encapsulation layers, sacrificial layers, diffusion barriers, and functional structures including top encapsulation layers in IC and MEMS/NEMS devices.⁷⁷
  • Interference films in a reflectometric interference spectroscopy (RIfS) sensor system for two new types of SiN_x biosensor chips: (i) SiN_x chip modified with lactose (carbohydrate); (ii) SiN_x chip with an antibody (protein). Potential applications include agriculture, biochemistry, medicine, health-care, food, and environment fields.⁷⁹
  • Single-nanometer-scale SiN_x pores for the detection, identification, and characterization of analytes for DNA sequencing and single-molecule nanoparticles measurements, biomolecules such as bacteria, protein, antibodies, nucleic acids, DNA, and RNA.⁸¹
  • Bioceramic coating for healing of soft and osseous tissue, inhibition of bacteria proliferation, and virus extermination for incorporation into orthopedics, dentistry, virology, agronomy, and environmental remediation.⁸⁵
  • Functionalized layers that serve as scaffolding for <100 nm-diameter nanofluidic ion-conducting channels (nanopores) for DNA sequencing, protein profiling, glycomics, and ionic circuit element construction.⁸⁶
• 2.  
  Non-aqueous (vapor phase) SiN_x surface modification techniques are being explored for the use of SiN_x as:

  • a-SiON:H as an encapsulation and protective layer of conducting polymers and chemically reactive electrodes against H₂O and O degradation.⁷⁸
  • Ultrathin nanoporous nanomembranes (NPN) coated with stable organic molecules to enhance functionality without reducing permeability and size selectivity for use in biological separations, cell culture, micropumps, and sensors.⁸⁰
  • A charge trap layer in 3D vertical NAND flash memory.⁸²
  • Biofunctionalized SiN_x to identify target molecules for applications in biosensors for medical diagnosis, drug development, food industry, and environmental monitoring.⁸³
  • A selective adhesion promoter, corrosion resistance cap, and protective layer against biofouling for applications in the glass industry, molecular electronics, and biotechnology, respectively.⁸⁴
• 3.  
  In terms of aqueous SiN_x functionalization protocols, the following surface treatments are noteworthy:

  • Various aqueous HF-etching solutions, aldehyde molecules, and tricholo-organosilanes were applied to examine the selective functionalization of the SiN_x surface. It was determined that HF etching results in an Si–F and Si–OH terminated surface with no selectivity. Alternatively, surface modification through SiN_x surface derivatization by aldehyde molecules followed by chlorosilanes achieved selectivity, which is associated with surface oxidation.¹⁰¹
  • A high pH (2 M KOH) solution was applied to two types of SiN_x biosensor chips: namely, a SiN_x chip modified with lactose (carbohydrate) and a SiN_x chip with an antibody (protein). This high pH modification of the SiN_x surface induced the formation of Si–OH surface groups which could then be coupled by organofunctional silanes to functional antibodies. One example of the latter was the use of a combined plant lectin and a toxin (ricin) to detect and screen for enterohemorrhagic Escherichia coli (EHEC) O157:H7 strain in the presence of possible interfering materials.⁷⁹
  • A 3-step chemical modification protocol was applied to achieve covalent-bonding of amphoteric poly-L-lysine (PLL) to the interior of SiN_x nanopore surfaces. The resulting structures were successfully applied in the detection of single-stranded DNA homopolymer translocation.⁸¹
  • Exposure of SiN_x to an acidic environment at low pH leads to a slow degradative oxidation of the SiN_x surface, primarily resulting in the release of NH₄⁺ ions. Alternatively, treatment at high pH ultimately produces NH₃ and H₂O according to the following equations:
    
    Both processes occurred at physiological pH (7.2–7.4), with residual silanol groups of the form Si−OH and basic secondary amines of the form (Si₂-NH) being generated on the SiN_x surface. This surface degradative process with elution of Si and NH₃ species was considered to be quite unique to SiN_x and is beneficial for repair and healing of osseous tissue while concurrently inhibiting bacterial and viral proliferation.⁸⁵
  • Controlled dielectric breakdown (CDB) was applied concurrently with hydrosilylation to functionalize SiN_x nanopores with various surface attachments ranging from acidic (Si−R–OH, Si–R–CO₂H) to basic (Si−R−NH₂), to nonionizable (Si−R−C₆H₃ (CF₃ )₂) to chemically control the nanopore size, surface charge polarity, and resulting chemical reactivity, as well as to modulate its conductance by varying the solution pH. The resulting structure was employed in the detection of a test molecule, λ -DNA.⁸⁶
• 4.  
  Non-aqueous surface modification techniques can be divided into two general categories: (1) plasma or ultraviolet light-based processes that are in accordance with generally accepted IC fabrication process flows; (2) organic derivatization techniques more usually associated with synthetic organic chemistry. The latter include, olefination, acid-base interactions, and self-assembled monolayers.

Plasma- or UV-based Surface Modification Techniques

• UV light-based vapor-phase carbene insertion into surface C–H bonds was performed to form <5 nm-thick molecular capping layers on ultrathin nanoporous SiN_x membranes. The latter were further chemically functionalized with additional monolayers of polyethylene glycol (PEG). Subsequent analysis showed that the resulting membranes: (i) maintained 80% of their original gas penetrability and 40% of their initial hydraulic permeability; (ii) were stable up to 48 h against water exposure; and (iii) inhibited nonspecific adsorption of immunoglobulin G (IgG) and bovine serum albumin (BSA) proteins.⁸⁰
• An inductively-coupled-plasma reactive ion etching (ICP-RIE) surface amination technique using H₂ was applied to convert SiN_x surface atoms to Si–NH₂. The resulting SiN_x surface was readily biofunctionalized with both protein and oligonucleotide through covalent immobilization. A demonstration of feasibility study was carried out to achieve: (i) 90% surface coverage employing N-5-azido-2-nitrobenzoyloxysuccinimide, a photoactivable amino reactive bifunctional crosslinker; (ii) immobilization and hybridization of single strand DNA (ssDNA) with its complemented strand.⁸³
• The application of O₂ plasma for conversion of the SiN surface to SiO_xN_y is well-known and could be part of a strategy for hydroxylating the SiN surface for reaction with organofunctional silanes to act as an improved adhesion promoter.¹⁰² More recently, this approach was expanded in a theoretical first-principles study to include the concept of combining O₂ and fluorocarbon plasmas to enable equivalent etch rates for both SiN and SiO₂ surfaces for 3D-NAND applications. The modeling work showed that once the SiN is converted to SiO_xN_y, the resulting etch rate more closely matches that of SiO₂ in the fluorocarbon regime.⁸²

Organic Surface Derivatization Techniques

Organic derivatization techniques utilize two primary protocols based on:

• Removal of surface N to expose Si–H sites, which would then react with olefin species. This reaction was proposed as a technique for providing resistance to oxidation as well as a mechanism for integration into bio-functional heterodevices.⁹²
• Reaction of organofunctional silane, phosphonic acid, or carboxylic acid type entities with hydroxyl (silanol) surface groups on Si–N, wherein O atoms replace Si atoms.⁹²

In this respect, it is recognized that the characterization of surface functionalization is often imprecise and consequently associated reports in the literature must be reviewed in this context.⁷⁷ On the other hand, derivatization of SiN_x surfaces by UV light-activated olefin addition to Si hydrides (formed through an HF etch) can be attributed to an exposed Si surface.⁸⁴

• 1.  
  The interaction of water with the SiN_x surfaces is also of great interest given the relevance to its role as a protective coating under extreme thermal, chemical, and mechanical environments. In this context, it was reported that the reaction of water with bulk SiN_x at neutral pH proceeded slowly. Early observations that both the coefficient of friction and wear rate of SiN_x were decreased in the presence of even small amounts of water or moisture have recently been attributed to tribochemical replacement of N by O in SiN_x, with the concomitant release of NH₃.⁷⁶ Alternatively, a-SiON:H was observed to degrade rapidly in the presence of water vapor, with the degradation occurring by reaction with H₂O through interstitial paths and nanodefects.⁷⁸

As illustrated in the work above, SiN_x biofunctionalization is typically a multi-step undertaking that consists of two sets of surface engineering processes. In the first set, the following reactions take place: (i) non-specific adsorption of a N rich poly(amino acid) or protein to the SiN_x surface; (ii) replacement of a topmost N with an O, usually in a hydroxylic form; (iii) reaction with a H bound to N (amidic protons). In the second set, an additional reagent is applied in an ensuing coupling reaction to bind the bio-functional compound to SiN_x. There do not appear to be any literature reports of biofunctionalization of the formal compound Si₃N₄.

Table II presents Si precursor names, chemical formulas, chemical structures, and CAS registration numbers, while Table VII lists Si– and N– bond dissociation energies for selected SiN_x source chemistries. Table VIII also summarizes relevant properties of CVD and ALD SiN_x precursors. All tables have been updated and expanded from Ref. 3.

A review of Table VIII and the published work presented here indicates that relatively few new Si source chemistries for CVD or ALD SiN_x have been introduced since 2016. The development of novel precursors that can deposit high quality SiN_x at low temperatures without the use of plasma or UV light activation remains an urgent objective. To this end, the general trend has been to explore a combined functionality in precursor design and synthesis. However, while the precursor must exhibit an intrinsic pathway for and be conducive to the deposition of SiN_x, the ability to affect a surface driven deposition in low energy environments continues to be elusive.

This challenge has perhaps serendipitously driven attention to precursors that can interact with substrate defects, either by substitution or with uncoordinated surface sites. Such chemistries include amino-functionals or iodo-substituted silane precursors in which the amine or the iodine can be displaced with a surface hydroxyl group. The most promising of these compounds are those in which 2 or more H atoms are bound to Si and appear to have the potential to interact with substrate surfaces that are H-rich or exhibit a high density of surface dangling bonds and low-coordination site openings.

As discussed in this review, these precursors include: di(s-butyl)aminosilane, diiodosilane, bis(dimethylaminomethylsilyl)trimethylsilylamine, and cyclic versions 1,3-di-isopropylamino-2,4-dimethylcyclosilazane and tris(isopropyl)cyclotrisilazane.¹⁰³ In particular, earlier work on iodosilanes is drawing revived interest for pulsed CVD, ALD, and PE-ALD applications.^104–106

In this supplement to our earlier review article on the SiN_x material system,⁷ the authors provide a survey of the most recent SiN_x work in the open literature, spanning mainly the years 2017 through 2019. The article summarizes key details pertaining to Si source precursors and co-reactant chemistries, theoretical modeling and mechanistic studies, CVD and ALD growth processes, current and new potential applications, and relevant physical, electrical, and optical properties. For the full potential of silicon nitride thin films to be achieved in integration into Ultra Large Scale Integration (ULSI) and heterodevice structures, effective control of the transition to other material phases must be realized. Consequently, this report introduces a new section that outlines novel research in the area of nanoscale engineering of the topmost atomic layer of a SiN_x-rich film to enable coupling to semiconductors and bio-materials for specialized heterodevice systems.

Moving forward, it is worth noting that, while the literature has focused on the Si–N system as a material that can be generated by a variety of methods for a plethora of applications, the descriptor silicon nitride should only be used as an aggregational term. The wide range of deposition processes used and continuously expanding set of usages make it difficult or impossible to provide useful universal figures of merit that adequately define or determine appropriateness for a given application. Silicon nitride should therefore be viewed as a poorly defined class of materials with significant variations in composition, structure, and morphology. Films quite frequently consist of differing combinations of Si, N, and H, often with non-trivial atomic concentrations of C and O and with significant compositional variations between film bulk and surface regions. It is also common for films to be amorphous, without any defined atomistic bonding or arrangement and with no specific information on H content or its impact on film properties. Despite these ambiguities, the progression of applications, beginning with bulk ceramics and structural composites and progressing to IC systems including dielectrics, passivation layers, and barrier films, then moving to active photonic, optical, and biomedical devices, has been astonishing.

However, the ability to grow high-quality near-zero thickness films in low energy environments without the need for plasma enhancement and its potential disadvantages, including excessive H inclusion, is inherent to the success of future SiN_x applications. This capability is necessary to ensure the successful integration of chemically and mechanically stable films into emerging heterodevice structures.

Equally important is the need to develop novel Si source precursors and/or innovative surface activation or treatment techniques to enable area specific deposition (ASD), either by selectively catalyzing or inhibiting film nucleation and growth. While progress has been made in this direction, real manufacturing-worthy breakthroughs have remained elusive. The key to ASD, at least in part, is the need to differentiate the SiN_x surface from other dielectrics—primarily SiO₂, as well as hydrogenated Si surfaces. This differentiation has so far been unattainable due to the fact that SiN_x surface oxide and/or hydride are either formed or exposed during the ASD processing steps. What is needed is discrete and differential bonding to a clean and highly reactive SiN_x surface. To this end, and while still difficult to envision in high-volume manufacturing (HVM) of IC devices, biomedical applications have realized a number of high-risk, high pay-off achievements that could provide some guidance, such as the application of nitrene or carbene "click chemistry" to pre-treat the SiN_x surface.

In any case, it is clear that work on silicon nitride materials will continue to be fertile and rewarding from the perspective of fundamental materials understanding, process technology, and eventually enabling new applications.