Research on Si/SiO₂ Interfaces Characteristics Under Service Conditions

Abstract

Si/SiO₂ interfaces, an important functional part of silicon-based devices, are the structures most likely to cause failure. Under external load in the service state, Si/SiO₂ interfaces can degrade in different forms, and they can change from an ideal symmetrical structure to an asymmetric structure with defects. To systematically analyze the Si/SiO₂ interface, the research methods of microstructure, including characterization and modeling, are first introduced. Then, the effects of irradiation, high field stress, mechanical stress, and high temperature on Si/SiO₂ interfaces are studied. Chemical bonds, conductive band structure, and interface roughness can be changed under high field and mechanical stress loads. In addition, defect initiation and impurity migration may occur due to irradiation and temperature loads, which can lead to the failure of devices. Under multiple types of loads, the degradation mechanisms are complex, and the interfaces become more sensitive, which makes investigations into interface degradation laws difficult. For improving the reliability of devices, a systematic analysis of the influence on Si/SiO₂ interfaces under complex loads is summarized.

1. Introduction

Si/SiO₂ interfaces, as key structures of semiconductor devices, are widely used in MOS, MEMS, capacitance, and FBAR-based resonators [1,2,3,4] due to their excellent dielectric properties and simple preparation process. Si/SiO₂ interfaces are composed of the Si layer, transition layer, and SiO₂ layer; suboxide, emitting silicon atoms, oxygen vacancies, and P_b defects are gathered in the transition layer. Interfaces used in the aerospace, nuclear industry, and other special industries must face loads such as irradiation [5,6], electric field, mechanical stress, and high temperature [7]. However, because of the special structure of the transition layer, Si/SiO₂ interfaces are more sensitive to external loads compared to other structures in the device. More seriously, the thickness of SiO₂ film has been reduced to nanometers with the iterative improvement of processes [8]; thinner film leads to more fragile interfaces and introduces challenges to the reliability of devices [9]. The reliability of the Si/SiO₂ interface largely depends on the manufacturing process, and the process method and parameters directly affect the quality of the interface. In the annealing process, the selection of temperature, time, and gas atmosphere greatly affect the Si/SiO₂ interface’s defect density and sensitivity to external loads. Thus, to improve the device stability under service conditions, it is necessary to research the Si/SiO₂ interface’s degradation mechanism under different loads. In this paper, the changes in defect density, electrical properties and interface morphology when the interface is subjected to different types of external loads are taken as the research focus.

Non-ideal effects were produced when Si/SiO₂ interfaces were subjected to external load. Irradiation is a common external load encountered at the interface, which can be classified as neutron irradiation, electron beam irradiation, heavy ion irradiation, X-ray, etc. During the silicon-based device production process, Si/SiO₂ can suffer from etching, ion implantation, and other processes. As devices enter service, interfaces may be exposed to radiation from sources such as cosmic rays and radioisotopes. In addition, in many studies, Si/SiO₂ is exposed to irradiation from particle accelerators. Depending on the irradiation type and dose, the Si/SiO₂ interface can face different degradation effects. Enhanced low dose rate sensitivity (ELDRS) is produced by irradiation load, which can greatly increase the defects in annealed interfaces with a low dose of irradiation [10]. Another effect is the total ionization dose effect (TID). The number of defects introduced by TID is related to the total irradiation dose, and atom migration occurs under large dose heavy ion irradiation [11,12]. Under strong electric field stress, dielectric layer degradation [13], breakdown [14], valence band deviation, and other adverse effects can influence the performance of Si/SiO₂ interfaces [15]. In addition, high temperature leads to increasing the interface’s overall energy, which causes the atoms to be more easily excited and causes a large number of defects, such as hanging bonds and oxygen vacancies [16,17]. The electrical properties of the interfaces are seriously affected by these interface defects [18,19], which leads to the accumulation of tunnel currents, traps, and charges [20,21]. Under mechanical stress, chemical bond length and angle may be distorted [22,23], and the interfaces may deviate from the steady state [24,25].

Research on interfaces’ degradation mechanism under multiple loads is of higher difficulty and complexity. Generally, the energy produced by eternal loads is transferred to Si/SiO₂ interfaces, which causes atoms or electrons to be excited and interface degradation. Bonds distorted by mechanical stress or electric field are easier to break by irradiation or high temperatures, generating increased defect density and property changes [26].

In this review, Si/SiO₂ interface characteristics under service conditions are researched. The types of loads that the Si/SiO₂ interfaces may encounter are systematically classified in Figure 1. The research methods of interface characterization, modeling, and simulation are introduced in Section 2. In Section 3, interfaces under single loads, including irradiation, electrical stress, mechanical stress, and high temperature, are analyzed. The change in interface properties under multiple loads is researched in Section 4. In Section 5, a summary and outlook are provided.

2. Interface Modeling and Characterization

2.1. Interface Modeling

Different from other abrupt interfaces, Si/SiO₂ interfaces have a transition layer with special properties. In the common α/SiO₂/Si interfaces, there is a 7 Å-thick transition layer between α/SiO₂ and crystal Si (c-Si). Using X-ray reflectance (XPR) characteristics, it has been discovered that the density of the transition layer is between 2.36 g/cm³ and 2.41 g/cm³ with a Si/O/Si bond angle of 136° [27]. Compared with α/SiO₂, the bond angle is more compressed, which means there are non-ideal structures. Through XPS characterization, it was found that there were different proportions of suboxides in different positions of the transition layer, and the proportions of suboxides were very sensitive to the preparation process. To analyze the structures and properties of interfaces, a variety of modeling and characterization methods have been proposed and are widely used. First-principle and molecular dynamics modeling are used to accurately describe the degradation behavior of Si/SiO₂ interfaces, analyze the changes in interface structure and energy band [28], and explore the initiation mechanism of interface defects [29].

2.1.1. Modeling Based on the First Principle

Generally, the idea of the first-principle model is to form a good connection between the surface of the SiO₂ structure and the Si structure, at which time there are a large number of defects caused by mismatching at the interfaces. The structure can be optimized using the first principle, and the relevant interface characteristics are calculated [30]. First principles calculation is a complex ab initio algorithm based on quantum mechanics, simplified by Born–Oppenheimer approximation and Hartree–Fock approximation.

Models based on the first principle have higher calculation accuracy. Multiple material properties such as charge density, band structure, interfacial state density, and dielectric constant can be calculated from the model. However, due to the amount of computation required being too large, models based on the first principle are suitable for small-scale modeling of dozens of atoms. Jamil et al. [31] used a model based on first principles to analyze the electric field-induced hydrogen attachment defect and the effect of this defect on the breakdown voltage of SiO₂. Jamil et al. used SeqQuest to perform first principles-based simulation calculations. They geometrically optimized a single hydrogen atom in a three-dimensional supercell containing 24 SiO₂ molecules, built a model of Si/SiO₂ interfaces, and performed 2D wafer calculations to determine the effect of the true electric field. Through calculation, it was found that at a high electric field, interstitial hydrogen atoms can combine to form a stably attached hydrogen dimer; at an electric field of 5 MV/cm, interstitial hydrogen could combine to form an attached hydrogen dimer. This dimer mainly appears on the Si/SiO₂ interfaces.

2.1.2. Modeling Based on Molecular Dynamics

Establishing a non-crystal SiO₂ structure with a good connection with Si is an important problem in the study of Si/SiO₂; molecular dynamics simulation is suitable for modeling the amorphous SiO₂ structure. The principle of molecular dynamics simulation is to extend Newton’s laws of motion in classical physics to microscopic systems, which can avoid solving very complex quantum mechanical equations. At the same time, molecular dynamics’ description of a microscopic object is statistically consistent with the simulation method using full quantum mechanics. Since the potential function applied in the simulation of molecular dynamics is an approximation of the quantum interaction between complex particles in the system, it does not take into account the fine interaction between complex particles or the state of the particles themselves, so the accuracy of the calculation is far lower than that of the first principles.

In recent years, the α/SiO₂/Si interface model built by combining molecular dynamics simulation with density functional theory optimization has attracted more and more attention. This method can be used to build a large α/SiO₂/Si interface model while maintaining high accuracy, and the computational cost is greatly reduced compared with first-principles simulation. The typical Si/SiO₂ interface model is completely coordinated, and the modeling can be based on the continuous random network (CRN) model to establish a fully relaxed Si/SiO₂ interface [32]. Khalilov et al. [33] established the c-Si/α/SiO₂ model using molecular dynamics to simulate and study the structure and properties of the interfaces of c-Si (100)|α-SiO₂ formed by thermal oxidation. The ReaxFF potential function is used in the simulation to describe the fracture and formation of the bond; the chemical and mechanical stresses of the interfaces are also considered. A high-temperature oxidation process is simulated by striking oxygen atoms or oxygen molecules onto the surface at high energy, resulting in the structure of silicon oxide caused by the impact of overheated (5 eV) oxygen. In the initial oxidation stage, the stress of the interfaces increases significantly and gradually becomes stable with the progress of oxidation.

2.2. Interfaces’ Characterization Method

2.2.1. EPR

The EPR phenomenon is based on the inherent magnetic moment generated by the spin of the electrons. In most systems, electrons appear in pairs with a magnetic moment of 0, but in some cases, such as interface defects, unpaired electrons appear, and their magnetic moments interact with the magnetic field. The magnetic moments of these electrons are parallel or antiparallel to the field vector, which defines the two energy levels [34]. Resonant frequency oscillation of the magnetic field can cause the transition between the energy levels. An EPR spectrometer absorbs and analyzes the energy generated by the transition and, according to the spectrum, can analyze the material composition of the sample surface.

2.2.2. XPS

X-ray photoelectron spectroscopy (XPS) is an interface characterization technique based on the photoelectric effect. It provides information on the elemental composition and content, chemical state, molecular structure, and chemical bond of various compounds for the study of electronic materials. In the analysis of electronic materials, it can provide not only general chemical information but also surface, small region, and depth distribution information. In addition, because the X-ray beam incident on the surface of the sample is a photon beam, it is very destructive to the sample.

The basic principle of XPS is the photoelectric effect. There are three possibilities for an X-ray photon: it passes through the atom and does not interact with the orbiting electron and nucleus; photons are scattered by the electron, which is Compton scattering. Photons interact with the electron in the atomic orbit so that the energy of the photon is completely transferred to the electron, which is the basic process of XPR. In the third case, electrons are emitted from atoms with photon energy (hv) and kinetic energy (E_k) with a difference in binding energy (E_b), and the atomic composition in the sample can be identified by detecting E_k [27]. A typical XPS experiment process is to place the sample on a suitable support in a high vacuum environment, as shown in Figure 2a. The light source emits monochromatic X-ray photons, which are directed at the sample. Some of the photons interact with atomic orbital electrons, and some of the photoelectrons generated are guided out of the sample. These are analyzed to measure the number of electrons with different kinetic energies and then processed by a computer to form a spectrum.

2.2.3. C-V Test

A capacitor–voltage (C-V) test is used to determine a variety of semiconductor parameters for different devices and structures, ranging from MOSCAP, MOSFET, and JFET to photovoltaic cells and MEMS devices. In research of Si/SiO₂ interfaces, the capacitance change of the interfaces can be obtained through the C-V test, and important parameters such as defect density, dielectric layer fixed charge, carrier concentration, and polarity of the interfaces can be analyzed [35]. However, the C-V test is a macroscopic technique that cannot give the spatial distribution of the obtained parameters and has certain limitations. In research of structure and electrical properties of Mn/SiO₂/p-Si structures, Ashery et al. [36] used X-ray diffraction (XRD), capacitance–voltage (C-V), and conductively–frequency (G/ω-V) measurements. The results show that the values of ε′, ε″, and tanδ have a positive correlation with temperature and bias, especially at low frequencies. At different frequencies, the permittivity (ε′), dielectric loss (ε″), and dielectric loss angle tangent (tanδ) of Mn/SiO₂/p-Si structures change with frequency.

2.2.4. DLTS

Deep-level transient spectroscopy (DLTS) is a method to characterize the deep-level and interfacial states of trace impurities and defects in semiconductors. DLTS studies the properties of defect centers through capacitance transients, which provide information on the level of impurities in the depletion zone by observing the capacitance transient that restores thermal balance after a disturbance of the system [37]. The basic principle of DLTS is that a trap in a space charge structure with a constant bias applies a periodic filling pulse at a constant temperature. These traps are in thermal equilibrium in the state of ionization and are filled with pulses. After each filling, the system can reach equilibrium by transmitting carrier heat into the corresponding energy band, and the emitted electrons are washed away by the electric field in the space charge region; the carrier emission rate at this temperature can be obtained. To solve the problem of consuming too much time to analyze the emissivity of a wide range of deep levels, DLTS adds the ability to set the rate window through the action of a time filter so that the measuring device gives an output only when the rate within the window is transient [38].

Figure 2. (a) Basic principles of the XPS experiment (reprinted from [39], Copyright 2020, with permission from AIP). (b) Image of Si/SiO₂ interfaces by TEM (image above reprinted from [40], Copyright 2022, with permission from MDPI); (image below reprinted from [41], Copyright 1988, with permission from APS).

Figure 2. (a) Basic principles of the XPS experiment (reprinted from [39], Copyright 2020, with permission from AIP). (b) Image of Si/SiO₂ interfaces by TEM (image above reprinted from [40], Copyright 2022, with permission from MDPI); (image below reprinted from [41], Copyright 1988, with permission from APS).

2.2.5. AFM

Atomic Force Microscope (AFM) is an analytical method that can be used to study the surface structure of solid materials, including insulators [42,43]. AFM analyzes the surface structure and properties of a substance by detecting the very weak interatomic interaction between the surface of the sample and a micro force-sensitive element, as shown in Figure 3a. When a pair of weak force-sensitive microcantilevers is fixed at one end, and a tiny probe (radius less than 10 nm) at the other end approaches the sample (probe-to-sample distance 0.2–10 nm), it can interact with the microcantilevers. The force can cause the cantilever to deform or change the motion state, as shown in Figure 3b. When scanning the sample, the sensor is used to detect these changes, and the force distribution information can be obtained so that the surface topography and structure information and surface roughness information can be obtained with nanometer resolution. The basic components of an AFM are a probe, a cantilever, a scanner, a laser, a data processor, and a photodetector.

2.2.6. TEM

The principle of the transmission electron microscope (TEM) is to gather the electron beam into a small beam spot through a series of coils and focus it on the surface of the sample (Figure 3c,d), scan the sample point by point, and install a detector under the sample to pick up the scattered electrons. TEM can be used to characterize the Si/SiO₂ interfaces, as shown in Figure 2b. The SiO₂ crack caused by the mismatch of the coefficient of thermal expansion (CTE) can be seen in the TEM image. The detector collects and processes the synchronization signal, matching the points in the sample to the points in the image [33,44]. Because of the different receiving angles, one or several signals can be collected at the same time during the experiment to obtain different images of the same position material. These images often contain different information about the material, which can complement the analysis of the material. At the same time, an X-ray energy spectrometer is placed above the sample, and the relevant composition information can be obtained along with the sample image.

Figure 3. (a) Left: bottom-up configuration. Right: top-down configuration (reprinted from [45], Copyright 2022, with permission from AIP). (b) Illustration of the cantilever oscillations induced by the thermal expansion cycle. Each time the sample absorbs the infrared light pulse, the thermal expansion kicks the tip and generates cantilever oscillations detected by the laser diode of the AFM (reprinted from [45], Copyright 2022, with permission from AIP). (c) Schematics of in situ TEM (reprinted from [46], Copyright 2020, with permission from WILEY-VCH). (d) Gatan 652 heating holder equipped with a tantalum furnace (reprinted from [47], Copyright 2021, with permission from Elsevier).

Figure 3. (a) Left: bottom-up configuration. Right: top-down configuration (reprinted from [45], Copyright 2022, with permission from AIP). (b) Illustration of the cantilever oscillations induced by the thermal expansion cycle. Each time the sample absorbs the infrared light pulse, the thermal expansion kicks the tip and generates cantilever oscillations detected by the laser diode of the AFM (reprinted from [45], Copyright 2022, with permission from AIP). (c) Schematics of in situ TEM (reprinted from [46], Copyright 2020, with permission from WILEY-VCH). (d) Gatan 652 heating holder equipped with a tantalum furnace (reprinted from [47], Copyright 2021, with permission from Elsevier).

2.3. Summary and Discussion

Modeling and characterization are essential methods for analyzing the evolution of Si/SiO₂ interfaces. First-principles modeling is employed to calculate various properties of the interfaces and is suitable for models with a scale of tens of atoms. Molecular dynamics models, on the other hand, are used to analyze larger interface models, with the number of atoms in the model reaching tens of thousands. Testing characterization is a more reliable research method. EPR is a magnetic resonance technique derived from the magnetic moment of unpaired electrons, which can be used to detect unpaired electrons in atoms or molecules qualitatively and quantitatively and explore the structural characteristics of their surrounding environment. To obtain information about the material distribution within the top 10 nm of the sample surface and the distribution of impurities or defects at concentrations as low as 1%, XPS can be selected, which relies on the photoelectric effect to excite inner-shell or valence electrons in atoms or molecules and can measure impurities at a concentration of 1%. C-V testing can provide the capacitance–voltage curve of the interfaces, allowing for the analysis of interfaces’ state density, but it does not provide specific information about the distribution of interface defects. For higher accuracy requirements, DLTS can characterize extremely low-concentration defects. DLTS uses the capacitance transients of semiconductor junctions to detect carrier emission processes from deep energy levels. AFM and TEM can observe the interface morphology directly. AFM is used to study the surface structure of solid materials, including insulators, providing morphological structure information of the Si/SiO₂ interfaces with a resolution of up to 0.1 nm. TEM obtains morphological information by focusing electrons on an ultra-thin sample, and compared to AFM, TEM can provide internal information about the sample with a resolution of up to 0.1 nm. Based on these modeling and characterization methods, research on Si/SiO₂ under load conditions can be conducted.

3. Interfaces Under Single Load Conditions

3.1. Interfaces Under Irradiation

Silicon-based semiconductor devices that serve in special industries, such as the aerospace and nuclear industry, must face the induction and promotion of irradiation on device degradation. Under irradiation, the Si/SiO₂ interfaces can suffer from ELDRS, and the TID effect can bring a large number of P_b centers to the interfaces. The structure of the interfaces can be changed under high-energy ionizing irradiation. These effects of irradiation can degrade the interfaces, cause atom migration, and affect the performance and reliability of the device.

3.1.1. Defects Related to H Atoms

A large number of point defects are inevitably introduced in the semiconductor oxidation process [48,49], and point defects play a key role in various physicochemical processes. The defects can cause effects like oxide charging [50], Fermi level pinning, defect passivation, electron transport through the defect state [51], and the charge states of moving impurities [52,53]. In the preparation process, many defects, such as hanging bonds, are introduced into the Si/SiO₂ interfaces, as shown in Figure 4d, which affect the performance of the interfaces. In order to eliminate this adverse effect, hydrogen gas annealing is usually carried out. H₂ diffusion in the interface (as shown in Figure 4f) causes the hanging bonds to react with hydrogen molecules to form Si/H bonds, and the Si hanging bonds are saturated. Saturation of hanging bonds by gas annealing can largely eliminate the influence of hanging bonds on interface properties; however, the annealing process introduces hydrogen defects in the interfaces. Hydrogen atoms involved in passivation in the interfaces can gradually dissociate and diffuse, and H atoms can also produce Si/H/Si, hydroxyl, and other defects [5], as shown in Figure 4e. Alkauskas et al. [43] used density functional theory (DFT) to study hydrogen defects at the Si/SiO₂ interfaces. It can be inferred that the average energy released by the hydrogen bridge defect during electron charging and discharging corresponds to 2.3 eV and 2.5 eV of PBE and PBE0, respectively.

The defects in annealed interfaces can increase significantly under low dose rate irradiation, which accelerates device degradation. This phenomenon is called ELDRS. The oxygen vacancy during annealing may split the H/H bond in a hydrogen molecule and combine with an H atom to form an E′_γ center, which can be expressed as Equation (1).

$$V_{o\gamma }+h^{+}\to E_{\gamma }^{\prime }$$

(1)

A similar reaction in Equation (2) can occur in the P_b center

$$H+P_b\to P_{b}H$$

(2)

In the case of low-dose irradiation, the irradiation energy is not insufficient to directly break the Si/O bond, Si/H bond, or other chemical bonds in Si/SiO₂ interfaces, but due to the H effect introduced in hydrogen gas annealing, the energy required to break the Si/H bond is reduced. When a free H is close to a saturated P_b center, an unstable Si/H/H bond may be formed, and the following reaction can form a P_b center and H₂ molecule, which can be expressed as Equation (3)

$$H+P_{b}H\leftrightarrow H-H-P_b\leftrightarrow H_2+P_b$$

(3)

As shown in Figure 5a and Equation (4), the activation energy required for this reaction is small and can occur in a low dose rate irradiation environment.

$$H-H-P_b\leftrightarrow H_2+P_b$$

(4)

The ELDRS phenomenon is closely related to the H atom, and the high concentration of H in the device can promote the degradation of the device under irradiation, and the defects in the SiO₂ layer and Si/SiO₂ interfaces also play a role. If energy was applied to the H atom, the energy of 2 eV could begin to dissociate; as this energy reaches 8 eV, almost all Si/H bonds can be dissociated. However, it requires at least 6 eV energy acting on Si atoms to dissociate the Si/H bond; to completely dissociate all of the Si/H bonds, the energy cannot be less than 18 eV. Such a big difference is because H is only connected to one Si atom, while Si atoms are connected to three other Si atoms. Energy applied to Si atoms may influence, not solely the Si/H bond, but also be absorbed by other Si atoms. In addition, under continuous ionizing irradiation at the Si/SiO₂ interfaces, E′_γ and P_b centers can be induced in the SiO₂ layer and the interfaces. It is found that in disordered SiO₂, irradiation-induced holes migrate by diffusion, and the diffusion coefficient decreases with the progress of irradiation. As a result, the formation rate of the E′_γ center is decreased, and the concentration of the E′_γ center increases slowly with the increase in dose. Through non-radiative recombination, the vibrational energy released by the E′_γ center can be transferred to nearby H2 and P_b centers, thereby stimulating their vibrational dynamics. This vibrational excitation reduces the activation energy of the reaction.

To analyze ELDRS more accurately, scholars have built models for simulation calculation. Zhaocan et al. [54] established a three-dimensional Si/SiO₂ model based on the first principle to simulate the effect of hydrogen on interface defects, which is more reasonable in describing the coupling of potential field and diffusion. For larger scale calculations, Sheikholeslam et al. [5] used LAMMPS for molecular dynamics-based simulations and established the Si<111>-SiO₂/Si<111> model shown in Figure 5b, which has a total of 17,818 atoms. As shown in Figure 5a,b, by using the PKA method, it was found that the H/Si bond is more likely to dissociation. Yu et al. [5] established a generalized model of the Si/SiO₂ interfaces under ion irradiation, remedying the defect of the traditional model at low dose rate concentration and deviation between the concentration of experimental observation. The formation of E′_γ centers is slowed down, and the conversion between E′_γ and P_b centers is reversible. In order to solve the problem that the interfaces are susceptible to irradiation after passivation, Yidan et al. [55] established four different interface models, including an interface model passivated by a hydrogen atom and an oxygen atom; in contrast, the oxygen atom passivation model has lower defect density. For the oxygen passivation model, the charge depletion region is around the silicon and the accumulation region is around the oxygen atom, while in the hydrogen passivation model, the electrons accumulate in the middle of the Si atom, forming a hanging bond in the middle of the Si atom, and the depletion region is on the SiO₂ side of the interface. H atoms can become ions under the action of an electric field, resulting in a large number of defects at the interfaces, while oxygen-passivated interfaces are much less likely to produce this phenomenon, which can effectively reduce the suspended bond density. After irradiation, there are a large number of defects, such as hanging bonds, that can be effectively eliminated by gas annealing.

3.1.2. Interfaces Under High Energy Irradiation

Under high energy, high flux irradiation, a large number of defects can be introduced in Si/SiO₂ interfaces, which cause a change in interface morphology. Defect density in Si/SiO₂ interfaces is positively correlated with the irradiation energy and flux; irradiation leads to the creation of E′ centers and P_b centers, which cause positive oxide charge accumulation. The higher density of this charge at lower doses is due to the higher density of the interfaces near the valence band edge in the interfacial trap distribution.

In order to study the specific mechanism of interface defects caused by ionizing irradiation of Si/SiO₂ interfaces, Capan et al. [56] studied radiation-induced interfacial defects. A sample of 50 nm oxide grown by thermal oxidation was placed in a TRIGA Mark II reactor [47,57], shown in Figure 6a, irradiated with 0.7 MeV neutrons [58]. By using high-frequency C-V measurement and DLTS, the interface defect density is shown in

Table 1. Comparison of defect density of different samples.

Sample	Neutron Fluence	$\Delta {\mathbf{E}}_{\mathsf{\alpha }}$ (eV)	${\mathbf{D}}_{\mathbf{i}\mathbf{t}}$ (eV−1 cm−2)
Sasprep	-	0.27 $\pm$ 0.01	$1.9 \times$ 1011
S-IR-1	1.6 $\times$ 1012	0.27 $\pm$ 0.02	$7.7 \times$ 1010
S-IR-2	$7.8 \times$ 1012	0.27 $\pm$ 0.01	$1.3 \times$ 1011
S-IR-3	1.6 $\times$ 1013	0.23 $\pm$ 0.01	$1.4 \times$ 1011

. There are P_b defects in the interfaces before irradiation, and the defect density increases with the increase in neutron flux. Ma et al. [59] studied the phenomenon of ultra-high dose TID at the Si/SiO₂ interfaces. N-FinFETs and P-FinFETs with channel lengths of 16–240 nm were used as samples in the experiment, as shown in Figure 6c. In order to restore the damage to the interface caused by irradiation, it is annealed at 100 °C for 24 h. In the process of irradiation, the channel length and type of FinFETs, the bias, and the irradiation dose all change the damage effect caused by irradiation. The ΔI_on-lin and ΔV_th changes in FinFETs with different channel lengths and bias after irradiation are shown in Figure 6d. The static DC response of the device in the linear region, before and after irradiation and after annealing, is measured at room temperature. N-FinFETs have stronger irradiation tolerance than P-FinFETs (as shown in Figure 6e–g), and the device with the longest channel shows more obvious degradation of the two FinFETs. The transconductance of the irradiated FinFET decreases obviously, and the threshold voltage degenerates slightly. Irradiation tests under different bias states show that FinFETs in the OFF state have the highest TID tolerance, and N-FinFETs in OFF, diode, and ON bias at a 500 Mrad dose are −2%, −8.6%, and −8%, respectively, which is much lower than P-FinFETs’ −16%, −25%, and −24.5%. This may be the electron–hole recombination enhancement under the condition of low interface bias.

Based on this, Halova et al. [60] used C-V and G-V characterization to study defects in MOS structures irradiated by 23 MeV high-energy electrons. It is observed that the interface trap shows obvious energy levels in the Si band gap, which were identified as donor-type P_b0 and P_b1 centers, whose concentrations depended on the irradiation dose. Additionally, Kim et al. [48] placed MOSFET with an oxide layer thickness of 30 nm under electron beam irradiation, then conducted gas annealing in an N₂ atmosphere and measured the shallow defect density results through ESR characterization, as shown in Figure 6b. The shallow defect density of irradiated and annealed samples is almost the same as that of unirradiated samples, which indicates that gas annealing can effectively reduce the shallow defect level. In research of high-flux heavy ion radiation [6], Yao et al. [61] studied the changes in the interface structure of Si/SiO₂ in 65 nm MOSFET. In the experiment, the sample was irradiated by Sn ions with an energy of 414 MeV [30]. The ion beam was applied to the surface of the device with a normal flux of 10¹² cm⁻². After irradiation, the sample was subjected to TEM imaging using FEI Tecnai G2 F2. It was found that there is a sub-layer near SiO₂ at the interfaces, which is composed of silicon suboxides, and Sn ions are found in the Si layer, which is caused by Sn penetrating the Si/SiO₂ interfaces.

Park et al. [62] used second harmonics to generate a signature B-doped Si/SiO₂ interface defect using a Mira Ti/sapphire laser to generate a p-polarized 800 nm beam and focus it on the sample to observe the intensity of the reflected SHG beam. It was found that in silicon wafers with high boron doping, the SHG measurements showed an initial significant reduction, indicating that boron doping formed a significant charge trap at the interfaces. As shown in Figure 7b, when the laser power is high, the oxygen trap is quickly occupied, resulting in an initial SHG reduction that cannot be observed. Heavy ion irradiation also leads to a decrease in the percentage of Si atoms in the interfaces. Before irradiation, the transition zone of the interfaces is very thin, and the percentage change of Si atoms is very drastic. After irradiation, however, the change becomes more continuous, and the average percentage of Si decreases from 50% to 30%. This is because, in the process of the Sn ion penetrating the interfaces, the Sn ion impacts the Si atom, destroying the original Si/O bond structure, making the excited Si move in the direction of the Sn ion, which leads to the appearance of extra oxide layer in the interfaces. Undoubtedly, a large number of interface defects will be introduced in this process. The oxygen atom concentration at the interface increases, the surface becomes rougher, clusters are formed, and the crack density increases. This is because the Si/O bond is broken by high-energy electron irradiation, and the dissociated free O atoms migrate through a defect in SiO₂, which may also be caused by irradiation [63,64].

After focusing irradiation, a large number of atoms can be observed to migrate deeper under the impact of irradiation, and excess Si atoms and O atoms are found in the excess Si zone, and the original clear interfaces begin to appear mixed. The SiO₂ layer is compressed, which also leads to the transfer of the interface region to the depth of the cell, which can be interpreted as the collision of high-energy ions with atoms in the SiO₂ layer, resulting in momentum transfer, resulting in the formation of O vacancy at the SiO₂ layer and O atom in the Si layer. Kaschieva et al. [65] used AFM to analyze the <100> oriented Si/SiO₂ interfaces after 20 MeV electron beam irradiation; the electron beam radiation can penetrate the entire silicon wafer. The AFM imaging results are shown in Figure 7a. Under the observation of AFM, it is found that the thickness of the oxide layer increases from 20 nm to 21.7 nm after irradiation, and the increase depends on the radiation dose and irradiation environment [68]. Sapkota et al. [7] irradiated the CeO₂/SiO₂/Si interfaces by Ar²⁺ ion irradiation with 1.6 MeV energy and TEM. After interface irradiation and annealing, a complex amorphous structure was formed at the CeO₂/SiO₂/Si interfaces. For this phenomenon, Fridlund et al. [66] established a Si/SiO₂/Si structure model for 25 KeV Ne⁺ irradiation based on molecular dynamics. As shown in Figure 7c,e, subjected to Ar²⁺ ion impact, Si atoms and O atoms are migrated, and a large proportion of non-ideal coordination numbers appear after irradiation.

In another study by Fridlund et al. [69,70], an extension to the MD algorithm called “adaptive moving environment (AME)” is presented to analyze Si/SiO₂ interfaces under high-fluence ion irradiation, a model established by AME is shown in Figure 8b. A cylindrical Si/SiO₂/Si structure is established to analyze Si atom migration, studying the dynamics related to the formation of the suboxide. The evolution process of the MD model during ion irradiation is shown in Figure 8a; due to atom mixing and Si atom migration, cylindrical height decreased rapidly, which provides verification of experiments by Xu et al. [71], as shown in Figure 8f,g. With the increase in irradiation flux, the concentration of Si atoms at Si/SiO₂ interfaces and SiO₂ layer is gradually increasing; atomic concentration curves are shown in Figure 8h,i. With the increase in flux, Si atomic content in the interface increased. The displacement of Si atoms during 6000 ions irradiation is shown in Figure 8j. A comparison of atomic displacement at different stages of irradiation is shown in Figure 8c–e. Most Si atoms at the top of the model move in a straight downward direction, while the Si atom near the bottom of the model has a random migration direction with a shorter distance. Sakaev et al. [72] studied the laser-induced Si/SiO₂ interface fracture process. The samples are SiO₂ film with thicknesses of 1, 2, and 4 μm deposited on Si by the PE-CVD process. After irradiation, the cracks spread from the edge of the interface cavity induced by the laser to the surface of the film responsible for the jet, and the small pieces break due to the tension reflection of the compression stress pulse on the surface of the film.

3.1.3. Effect of Irradiation on Interfacial Electrical Properties

At present, devices with Si and SiO₂ as the main structure are still the most common, including MOSFET, RAM, etc., and with the continuous reduction in the size of the device and the continuous improvement in the integration degree, the size of the Si/SiO₂ structure is also decreasing, and the thickness of the SiO₂ film is only a few nm. As a result, the Si/SiO₂ interfaces have a more obvious impact on device performance. Thinner SiO₂ films make the Si/SiO₂ interfaces more susceptible to direct irradiation and produce defects, which may be fatal to devices [73,74].

The quality of the Si/SiO₂ interface has an important impact on the reliability of memory [75,76]. For common DRAM devices, both ionizing radiation and non-ionizing radiation induce variable retention time (VRT) in DRAM units [77,78]. VRT is caused by defects present in the device, which can be either interface defects (such as the Si/SiO₂ interfaces) or bulk defects (such as defect clusters in silicon). Interface defects exist before radiation, and their density increases with the dose of ionizing radiation absorbed. Goiffon et al. [63] studied VRT in DRAM units under radiation. Two kinds of DDR3 LSDRAM chips with capacities of 2 GB and 4 GB were used in the experiment. A Xilinx KC705 evaluation kit was used for testing, and degradation caused by irradiation was evaluated according to DRAM retention time comparison before and after ⁶⁰Co irradiation, as shown in Figure 7d. The cells above the diagonal line have different maximum and minimum retention times, which can be considered as VRT cells. It can be found that irradiated DRAM has significantly more VRT cells.

The electrical characteristics of MOS devices can be significantly affected by ionizing irradiation. Defects in Si/SiO₂ can be introduced by irradiation, leading to a decline in the reliability of the devices and causing fluctuations in their electrical performance. The Si/O bonds at the SiO₂/Si interfaces can be readily broken by high-energy particles, facilitating the movement of oxygen atoms away from the lattice. As a result, V_O (vacancy–oxygen) defects and Si dangling bonds are left at the interfaces. Si hanging bonds are inherently interfacial states that trap electrons as deep states, which result in leakage current. Akchurin et al. [79] used different doses of neutrons and gamma rays to irradiate the samples and used C-V measurement to analyze interface properties. N_ox at the Si/SiO₂ interfaces changed from being dominated by N_f before irradiation to including deep donor (E_V + 0.65 eV) and deep acceptor type (E_C−0.60 eV) N_it, with a density comparable to that of N_f. Irradiation can affect the electron transport properties of the interfaces. Guanghao et al. [80], based on density functional theory, established a 72-atom Si/SiO₂ model (Figure 9b). In the SiO₂/Si interface structure, the formation energy of oxygen vacancy V_O is 2.23 eV, which is smaller than the formation energy of V_O in the SiO₂ part of 5.53 eV. The O atom at the interface acts as a barrier to electrons, which can prevent electrons from passing through. However, after irradiation, an oxygen vacancy is generated, and this barrier effect is weakened. The electrostatic potential difference at the Si/SiO₂ interfaces is shown in Figure 10a. It can be seen that the Si/SiO₂ interfaces with oxygen vacancy are more likely to generate leakage current.

The electrostatic potential of the SiO₂/Si interface structure along the Z-axis is without V_O, and b is with V_O. E_F is represented by a solid red line, and the vacuum degree of Si and SiO₂ is represented by a purple dotted line and a blue dotted line. More studies show that the leakage current of Si/SiO₂ interfaces is not only caused by oxygen vacancy but also contributed by the P_b center in the interfaces. Moxim et al. [66] used an HP 4155 A semiconductor parameter analyzer to conduct near-zero field magnetoresistance NZFMR and EDMR to characterize the irradiated Si/SiO₂ interfaces. The experimental process design is shown in Figure 9c. The interfaces after irradiation have a strong correspondence. The analysis of the high-frequency EDMR curve (Figure 10b) can be represented by 78%P_b0, 22%P_b1 or 77%P_b0, 21%P_b1, 2%E′ center curve and the P_b center has a strong contribution to the leakage current. Poehlsen et al. [82] studied the collection of charge carriers in p + n strip sensors near the Si/SiO₂ interfaces by using transient current techniques and focused nanosecond light pulses. It was found that the bias voltage and X-ray can affect the carrier collection of the sensor [83]. Koveshnikov et al. [84] studied the effect of electron beam irradiation on MOS capacitors manufactured on n substrate. It was found that the charge concentration induced by irradiation reached saturation with the increase in dose. Under 5 KeV and 15 KeV electron beam irradiation, according to the C-V curve, the experimental results at low dose and high dose are shown in Figure 11a,b. At room temperature, the negative charge induced by irradiation is relatively stable, while the positive charge shows a tendency to decrease near the interfaces. Zhang et al. [85] used a 12 KeV X-ray to irradiate SiO₂ films of different thicknesses growing in <100> and <111> directions. The samples were repeatedly annealed, characterized, and irradiated. The experimental results show that the interface defect and surface current density of <100>Si after irradiation is lower than that of <111> Si, and with the increase in the cumulative dose of X-rays, the interfacial defect density and surface current density tend to be saturated and decrease.

The deflection voltage at the interfaces has a significant influence on the defect density caused by irradiation. In the absence of interface bias, the electric field during irradiation is mainly the electric field generated by oxide charge, and the electric field is directed from the interfaces to the metal gate. For the positive bias voltage, the direction of the electric field is directed from the metal gate to the interfaces, which leads to more holes drifting toward the Si/SiO₂ interfaces, and the holes generated by irradiation are captured by the oxygen vacancy of the Si/SiO₂ interfaces, generating oxide charge. The threshold voltage (Vth) and subthreshold swing (S.S) degradation of transistors are also affected by irradiation. The interface state may have positive and negative charges. When the density of D_it is too high, the interface charge Q_it is increased, according to Equation (5).

$$\Delta V_{th}={\varphi }_{ms}+{\displaystyle \frac{\sqrt{4{\epsilon }_{s}kT{N}_A\ln \left(\raisebox{1ex}{$N_A$}\!\left/ \!\raisebox{-1ex}{$n_i$}\right.\right)}}{C_{ox}}}+2{\displaystyle \frac{kT}{q}}\ln \left({\displaystyle \raisebox{1ex}{$N_A$}\!\left/ \!\raisebox{-1ex}{$n_i$}\right.}\right)-{\displaystyle \frac{Q_f}{C_{ox}}}+{\displaystyle \frac{Q_{it}}{C_{ox}}}$$

(5)

Due to the action of the oxide trap charge, which can be considered the result of the joint action of the Si/SiO₂ interfaces charge and the gate oxide charge, which can be expressed as Equation (6).

$$\Delta V_{th}=\Delta V_{it}+\Delta V_{ot}=-{\displaystyle \frac{q\Delta N_{it}}{C_{ox}}}-{\displaystyle \frac{q\Delta N_{ot}}{C_{ox}}}$$

(6)

During the radiation process, electron–hole pairs are generated in the gate oxide, and some of the holes are captured and form positive charges in the Si/SiO₂ interfaces and the oxide layer, resulting in the shift of the threshold voltage. Ya-Hsiang et al. [70] studied the total dose effect of X-ray irradiation on low-temperature polycrystalline silicon (LTPS) thin film transistors (TFTs). In the experiment, X-rays with different intensities, time, and waveforms were used to irradiate the thin film transistor. The threshold voltage measured after irradiation is shown in Figure 11c,d. It can be seen that $\Delta {\mathrm{V}}_{\mathrm{t}\mathrm{h}}$ under all conditions gradually increases with the increase in time, and the slope of the curve will decrease when the intensity of the X-ray increases. Nikolskaya et al. [86] studied the photoluminescence (PL) properties [87] of silicon produced by the Si/SiO₂ system irradiated by Kr⁺ ions at 1235 nm wavelength. The researchers used Kr⁺ ions to irradiate a SiO₂/Si system on an n-Si (100) chip at an energy of 80 KeV at a dose of 5 × 10¹⁶ cm⁻². PL spectra were measured using a 405 nm laser to excite the sample, and PL spectra were measured at temperatures ranging from 9 K to 300 K. As the temperature increases, the electron–hole pair in the bound state decays or transforms into a free state. As a result, the charge carrier density in the 9 R-Si phase increases, leading to an increase in PL intensity with increasing temperature [88,89].

3.1.4. Summary and Discussion

When the Si/SiO₂ interfaces are irradiated, the structure and properties of the interfaces are changed, which may affect the function of the device using the Si/SiO₂ interfaces. Different defect forms, such as hydrogen defect, hanging bond, and oxide charge, can be produced, and the leakage current and threshold voltage of MOS can be affected in the device, making it fluctuate. Some research data are shown in

Table 2. Comparison of study data of Si/SiO₂ interfaces affected by irradiation.

Reference	Irridiantion Factor	Particle Energy (eV)	Dit (cm−2 eV−1)	Defects (cm−2)
[1]	Dose: 4 krad			Pb: 4.6 × 1010 E′γ: 5 × 1010
	Dose: 6 krad			Pb: 6.3 × 1010 E′γ: 6 × 1010
[49]	Flux: 7.8 × 1012	7 × 105	1.3 × 1011
	Flux: 1.6 $\times$ 1013	7 × 105	1.4 × 1011
[50]	Irradiated 30 s	2.3 × 107		Pb: 6.8 × 1012
	Irradiated 120 s	2.3 × 107		Pb: 1.48 × 1013
[48]	Unirradiated	1 × 104	0.76 × 1011
	N2 annealed	1 × 104	0.7 × 1011
[71]	Irradiation time 1 h	3 × 103	1.2 × 1011
[48]	Flux: 5 × 1015	3.5 × 104		9 × 1010
[44]	Dose: 0.1 krad			1.9 × 109

.

H₂ atmosphere annealing is often used in the preparation of Si/SiO₂ in order to passivate the defects introduced by the high-temperature oxidation process. However, the H/Si bond introduced is more likely to be broken by irradiation, resulting in the introduction of a large number of defects in ELDRS, which accelerates the aging of the device. For this phenomenon, several scholars have carried out research. Alkauskas et al. [43] used first-principles research to more accurately calculate the defect energy level of hydrogen bridge defects at Si/SiO₂ interfaces. Zhaocan et al. [44] established a three-dimensional model for the ELDRS phenomenon to simulate the role of H in the interface defect to describe the potential field more accurately and reasonably. Sheikholeslam et al. [5] conducted large-scale modeling and simulation and found that the Si/H bond was more likely to be dissociated by irradiation than the Si/O bond and Si/Si bond. The dissociated H atom induces the dissociation of other Si/H bonds, resulting in a large number of defects.

Ionizing irradiation can generate centers and P_b centers at the Si/SiO₂ interfaces. Yu et al. [1] found through research that these two defect centers can migrate through diffusion, and there is a reversible conversion between them. Capan et al. [49] analyzed the influence of neutron irradiation on the interface defect density, and Halova et al. [50] used C-V characteristic and G-V characteristic analysis to study the defects in the MOS structure irradiated by 23 MeV high-energy electrons. The results show that the interface damage caused by irradiation can be effectively recovered by N2 atmospheric annealing, and the interfaces passivated by N and O atoms have higher stability than H.

The influence of irradiation on the Si/SiO₂ interfaces also includes the change in interface structure. Yao et al. [51] used high-flux heavy ion irradiation on Si/SiO₂ interfaces. According to TEM, under the influence of irradiation, the oxide layer becomes thicker, the interfaces become thicker, and the boundary is not clear. This is caused by the motion of high-energy ions slamming Si atoms into each other. A similar study was also conducted by Kaschieva [54], who observed the Si/SiO₂ interfaces after electron beam irradiation through TEM and found that the interfaces’ roughness increased significantly and the interfaces’ oxygen content increased. High-energy irradiation can break many chemical bonds and promote recombination. Fridlund et al. [69] studied the effect of high-energy particle irradiation on the interfaces through molecular dynamics simulation. The simulation calculated that a large number of interface atoms on the surface were impacted by particles and migrated to the depth of the interfaces, introducing a large number of defects and changing the Si coordination number of the interfaces.

Interface defects and structure changes can influence device performance. Goiffon et al. [63] analyzed after irradiation effect of VRT in DRAM cells. After irradiation treatment, DRAM has more VRT units; this is due to irradiation being introduced to the interfaces of oxide charge and interface defects, which change its performance. Guanghao et al. [68] conducted a simulation calculation on the change in electron transport properties at the interfaces after irradiation and found that the Si/O bond in the interfaces is more easily destroyed, forming the oxygen vacancy V_O, missing the O atom that acts as an electron transport barrier, and the leakage current at the Si/SiO₂ interfaces is higher. However, it should be noted that the generation of leakage current is not only due to oxygen vacancy; the P_b defect center also contributes to the leakage current. Moxim’s [66] research shows that the generation of leakage current after irradiation is affected by various defects, among which the contribution to the leakage current is 77% P_b0, 21% P_b1, and 2% E′. Other researchers, such as Poehlsen et al. [69], used transient current calculations to test the effect of X-rays on sensor carrier collection, and Koveshnikov et al. [71] studied the effect of irradiation on MOS capacitors.

3.2. Interfaces Under Electrical Stress

3.2.1. Interfaces Breakdown

Dielectric breakdown refers to the process in which a dielectric material undergoes a partial transition from an insulating state to a conductive state under the action of an external high electric field. Dielectric breakdown is an important form of failure of MOS devices under high field stress [90,91]. The special properties of Si/SiO₂ interfaces affect the breakdown mechanism of dielectrics. An important characteristic of dielectric breakdown is dielectric breakdown-induced epitaxy (DBIE) [92], which can be observed by TEM.

In interfaces with thinner oxide layers, the breakdown mechanism has unique characteristics. In a thin oxide layer, when the electron is subjected to an electric field and its energy is higher than the valence band of 9 eV, shock ionization may occur. Because this mechanism requires a large enough potential difference, it is more likely to occur in a thicker film. In a film with a thickness of 1.4 nm, the trap generation mechanism is also not dominant because the thinnest data point measured is below the trap creation threshold. Usui et al. [8] analyzed the reversible breakdown of the Si/SiO₂ interfaces and proved that the F-N tunnel current plays a major role in the thin oxide layer. The dominant low position in this film is the anode injection mechanism. To verify this, Usui et al. conducted a breakdown experiment with a gold anode. The maximum breakdown voltage is in the film with a thickness of 4.4 nm and an electric field of 2.7 V/nm. As can be seen in Figure 12d, the curve of leakage current rising and falling with the electric field basically coincides with the onset of F-N tunneling and irreversible breakdown, indicating that the breakdown is reversible. This is because, under the action of a strong electric field, the SiO₂ conduction band is bent, the electron tunneling distance is reduced, and the F-N current begins to appear, which does not destroy the Si/O bond.

DBIE can be introduced with interface breakdown. The diffusion of Si atoms at the interfaces consists of two parts: (1) self-diffusion of Si atoms due to thermal gradient and acceleration of crystal defects, and (2) enhanced Si atom migration due to local high-density currents [93]. The process can be described by the following Equation (7) [9]:

$$\mathrm{a}=1, J_{Si}=C\left({\displaystyle \frac{D}{kT}}\right)Z^{\ast }e\rho J$$

(7)

where J_si is the diffusion density of Si atom, C is the concentration of Si atoms, and in the DBIE epitaxial structure, C = 1, D is the diffusivity of Si near the breakdown point, k is the Boltzmann constant and the number of effective charges, T is the temperature at the epitaxial growth location, e is the charge of the carrier and the resistivity at the epitaxial, and J is the current density at the epitaxial [79].

The resistivity of NMOS in the inversion channel is 2 times lower than that of PMOS under the same conditions, which results in a larger current density at the moment of breakdown, a higher local joule heat, and a higher density of defects. The higher current is more favorable to the electromigration of Si, causing more Si atoms to migrate. Because the epitaxial part of NMOS has higher current density and temperature, the diffusion coefficient D of Si is larger, and the DBIE size in NMOS is larger than that of PMOS for the above reasons. Figure 12a shows TEM images of the center and edge of the PMOS that were punctured, respectively, in which DBIE can be clearly seen. Pey et al. [94] studied the difference in DBIE size between NMOS and PMOS and obtained DBIE images through HRTEM. It can be seen that DBIE in NMOS is larger in size than that in PMOS. According to research, the ratio of DBIE size is above 2/1, and the arrow in Figure 12c marks the percolation path. Li et al. [79] observed the transistor breakdown through STEM, as shown in Figure 13a. As shown in Figure 12b, the breakdown locations’ oxide thickness is the same as in the unbroken region. Subsequently, electron energy loss spectral EELS analysis was performed at the breakdown location above DBIE in an attempt to analyze the chemical properties of the percolation path. The EELS analysis results are shown in Figure 13b,e. After the breakdown, the dielectric layer Si⁴⁺ decreased and the content of silicon nitrite increased, which was caused by hypoxia. Oxygen dissociates and erodes the Si/SiO₂ interfaces, causing the local atomic structure to rearrange and generating energy gaps inside the interfaces. Bonifacio et al. [9] used an EELS-enabled TEM (Figure 13c) to perform electrical application experiments on the biased gate electrode and found that oxygen vacancy permeates in the gate dielectric layer, which leads to the roughness of the interfaces and the reduction of the initial insulating layer thickness. In addition, the migration of cobalt across the polysilicon grain boundary was also observed, which was caused by the increase in current density and local joule heating. As shown in Figure 13d,e, under constant bias stress, permanent and temporary leak path formation occurs, which is due to the migration of oxygen vacancy. The formation of oxygen vacancy is due to oxygen migration driven by the electric field. Under increasing bias stress, more defect structures were observed, including interface roughness and metal ion migration. The formation of these defective structures leads to the destruction of the dielectric.

Figure 12. (a) DBIE image generated by PMOS breakdown obtained by HRTEM, top image near the channel center, bottom image near the channel edge (reprinted from [92], Copyright 2008, with permission from IEEE). (b) DBIE is marked, and the oxide area on top of the DBIE bump is displayed. The gate oxide is a 2.0 nm nitrided amorphous SiO₂ (reprinted from [94], Copyright 2003, with permission from AIP). (c) DBIE image generated by NMOS breakdown obtained by HRTEM, the arrow represents the percolation path of the leakage current (reprinted from [94], Copyright 2003, with permission from AIP). (d) Relation between electric field and leakage current density (reprinted from [13], Copyright 2013, with permission from Elsevier). (e) The background corrected 0 K edge spectra measured from bulk oxide and breakdown and non-breakdown gate oxides (reprinted from [13], Copyright 2013, with permission from Elsevier).

Figure 12. (a) DBIE image generated by PMOS breakdown obtained by HRTEM, top image near the channel center, bottom image near the channel edge (reprinted from [92], Copyright 2008, with permission from IEEE). (b) DBIE is marked, and the oxide area on top of the DBIE bump is displayed. The gate oxide is a 2.0 nm nitrided amorphous SiO₂ (reprinted from [94], Copyright 2003, with permission from AIP). (c) DBIE image generated by NMOS breakdown obtained by HRTEM, the arrow represents the percolation path of the leakage current (reprinted from [94], Copyright 2003, with permission from AIP). (d) Relation between electric field and leakage current density (reprinted from [13], Copyright 2013, with permission from Elsevier). (e) The background corrected 0 K edge spectra measured from bulk oxide and breakdown and non-breakdown gate oxides (reprinted from [13], Copyright 2013, with permission from Elsevier).

3.2.2. Defects Introduced by Electrical Stress

In MOS devices, gate insulation materials can be selected from a variety of options. With the pursuit of device performance improvements, such as Group III metal oxides and Group IVB metal oxides, high-k dielectric has a higher dielectric constant, generally higher than 3.9, which can increase the physical thickness of the gate insulation layer and reduce the leakage current. However, high-k dielectric may bond with the Si gate to form the Fermi pinning effect [95]. To combine the advantages of multiple dielectrics, pseudo binaries combine a variety of metal oxides in different proportions. SiO₂ is widely used because of its good dielectric properties, simple preparation process, and lower parasitic capacitance; the instability of Si/SiO₂ interfaces has always been an important factor limiting the performance and reliability of MOS devices. Under the environment of a strong electric field and strong current stress, the degradation and dielectric instability of SiO₂ films are significantly affected. In earlier studies, the roughness of the SiO₂ layer gradually decreases, and the interface roughness decreases with D_it because the local electric field at the projection of SiO₂ film is enhanced, while oxygen atoms are more likely to appear at the projection of Si substrate, which leads to faster interface degradation here. This oxidation process can be accompanied by the generation of a large number of hanging bonds, which increases the density of D_it and the thickness of the interfaces’ oxide layer.

Inoue et al. [96] used AFM to study the changes in Si/SiO₂ interface morphology in MOS capacitors caused by current stress. In the experiment, an electric field strength of 9 MV/cm was applied to an 18.5 nm thick SiO₂ thin film. Characterization revealed that the interface roughness decreases with the increase in electric field strength and duration. Moxim et al. [97] studied the defects of Si/SiO₂ interfaces under high field stress. Moxim et al. used a 126-cell MOSFET array with a gate oxide layer thickness of 7.5 nm for each MOSFET and used a Hewlett-Packard 4155 A semiconductor measuring instrument to generate high field stress. The sample was measured by EDMR under several conditions of gate voltage: V_g = −8, V_g = −9, V_g = +9, and V_g = +10. Under positive polarity bias voltage, P_b0/P_b1 = 6.7/1, and under negative polarity voltage, P_b0/P_b1 = 2.7/1. The NZFMR results showed that the high field stress induced hydrogen redistribution at and near the interfaces, and H was thought to contribute to the formation of additional E′ oxide defects. Poonet et al. [98] studied the properties of interfacial states in Al/SiO₂/nSi (111) structures using a photoactivated conductivity measurement method. The capture cross-section of the interface state for a few carriers is several orders of magnitude larger than that of the electron. The capture cross-section of the interface state has a strong dependence on the electric field intensity of the SiO₂ layer surface. The interfacial states with different energies show similar dependence on the variation of surface electric field intensity. It is indicated that the energy dependence of the interface state is actually attributed to the electric field difference during the charge exchange process between the interface state and the valence band. High field stress can increase the interface defects and introduce a large amount of charge in the SiO₂ layer. Paskaleva et al. [99] used C-V measurements to characterize the Si/SiO₂ interfaces under a high field. The experimental sample is a capacitor, and under high field stress, the charge is mainly generated by the behavior of the carrier injected into the oxide in the oxide volume.

For the interfacial chemical bond fracture caused by strong field stress, Bersuker et al. [10] studied the progressive breakdown characteristics of gate structures composed of high K media under the action of stress. The researchers performed I_d-V_g measurements, frequency-dependent charge pump (CP) measurements, SILC measurements, and CVS measurements on toroidal capacitors and transistors with a dielectric thickness of 3 nm, 5 nm, and 7 nm. Under the influence of electrical stress, some of the Si/O bonds in the SiO₂ layer break, leading to the overlapping of defect potentials and the formation of a carrier conduction path. This results in the creation of a leakage current path between defects. Tyaginov et al. [20] studied the degradation of hot carriers at the Si/SiO₂ interfaces, used an NMOS with an effective channel length of 44 nm for hot carrier degradation testing, and measured the change in leakage current. At lower stress voltage, the degradation of the device becomes less serious with the increase in temperature, while at higher V_ds and V_gs conditions, the degradation rate of the device becomes significantly faster.

Koharu et al. [100] used the DLTS measurement method to study the effect of DC bias stress on the high k/SiO₂/Si gate stack. Due to the sharp increase in the interface defect density caused by the bias stress, interface defect density and capacitance–voltage characteristics are changed significantly; the distribution is shown in Figure 14. The numerical simulation explains that both the fixed charge and the interface defect density increase due to the DC bias stress and cause the change in the local characteristics. Jung et al. [101] employed two processes, PE-CVD and PE-ALD, to prepare a Si/SiO₂ interface with a thickness of 25 nm at a temperature of 200 °C. Following deposition, the interface was annealed in N₂ for 10 min at 400 °C, after which the electrical properties of the interface were measured. After the C-V test and I-V test, the interface state density of the interfaces prepared by PE-CVD and PE-ALD was 6.12 × 10¹¹ and 5.01 × 10¹¹ cm⁻². The Si/SiO₂ interfaces were observed by AFM. The RMS of the interfaces prepared by PE-CVD and PE-ALD were 0.41 nm and 0.24 nm, respectively, and the interfaces prepared by PE-ALD were better in smoothness. Because the interfaces prepared by the PE-ALD process had lower interfacial state density and roughness, they have better electrical properties. When the electric field intensity was −1 MVcm⁻¹, the leakage current density of the two was 8.3 $\times$ 10⁻⁸ and 1.4 $\times$ 10⁻⁸ Acm⁻².

RTN is a kind of noise harmful to circuit performance and reliability, and with the continuous miniaturization of electronic devices, this damage becomes more and more obvious. The two-state fluctuation is confined to the interface trap, and its velocity and amplitude are related to the bias. Previous studies have indicated that the cause of RTN is the fluctuation in the number of charge carriers, and this occurs at interfaces with both interface charge traps and oxide traps (Si suspended bond and oxygen vacancy); it is necessary to improve the performance of RTN. In MOS devices, RTN is predominantly attributed to the tunneling occurring between the oxide traps and the channel. This tunneling results in the traps emitting and capturing charges to and from the channel, thereby causing variations in the carrier count. The amplitudes and rates of the random telegraph signal (RTS) produced by isolated traps at the Si/SiO₂ interfaces are influenced by bias. In other words, the parameters related to RTS are altered. Cowie et al. [102] used AFM to characterize random telegraph noise (RTN) at the Si/SiO₂ interfaces. RTN originates from carrier mobility fluctuations caused by interface traps. Fm-AFM Fd measurements show a significant increase in Fd at the defect location, which indicates an increase in scattering loss at the defect location. The generation mechanism of RTN is related to H passivation and inverse passivation at the P_b center of the interfaces, as shown in Equation (8).

$$P_b^{-}+h^{+}\leftrightarrow P_b^{+}$$

(8)

In the case of high electric fields, the Si/H bonds dissociate, which provides an explanation for the effect of bias on RTS.

3.2.3. Summary and Discussion

Under the action of a strong electric field, the Si/SiO₂ interfaces can degrade SiO₂ film, break down the interfaces, change the interface roughness, and increase the interface defect density. These phenomena have a great impact on the performance of MOS and other devices with Si/SiO₂ interfaces as the key structure. Representative data from some studies are shown in

Table 3. Comparison of study data of Si/SiO₂ interfaces affected by electric field.

Reference	Bias (V)	Time (s)	Leakage Current (nA)	DBIE (nm)
[67]	1		1.4 × 107
[42]	4.1	400	4 × 104	200
[80]	5.4			40
	5.4			5
[9]	5	2500	105
	10.5		225.8
[84]	15	30
[10]	5	104	500

. According to the research conducted by Inoue et al. [96], AFM observations have revealed that, after the application of a strong electric field, O atoms tend to migrate to the protrusions of the Si substrate. This migration leads to a reduction in interface roughness and an increase in defect density, which, in turn, degrades the interface quality. Concurrently, the external strong electric field can induce hot carrier degradation at the Si/SiO₂ interfaces, with the device degradation rate being more rapid at higher V_ds.

Si/SiO₂ interfaces under an electric field have a special breakdown mechanism. Usui et al. [8] studied the reversible and irreversible breakdown of the interfaces and found that there are three mechanisms, impact ionization, anode injection, and trap generation, which work together in the breakdown process of Si/SiO₂ interfaces. Each plays a leading role in the case of different electrode and bias voltage. When irreversible breakdown occurs at the Si/SiO₂ interfaces, DBIE can be detected by TEM characterization. Pey et al. [80] analyzed the difference in DBIE size between NMOS and PMOS at the time of breakdown, gave a mathematical model of local circuit density at the moment of breakdown, and found that PMOS had better anti-penetration robustness. Li et al. [79] analyzed and found that after breakdown by a strong electric field, the proportion of Si/SiO₂ on the Si/SiO₂ interfaces increased, and the joule pyrolysis from the Si/O bond after breakdown resulted in a large number of defects. In the study by Bonifacio et al. [9], the interfaces were broken down by applying a fixed bias stress. After the breakdown, interface roughness occurred, and the thickness of the SiO₂ layer was reduced, which was due to the penetration and migration of oxygen vacancies in the interfaces driven by the electric field, forming a defect structure. This mechanism can be used to explain the progressive breakdown of the dielectric under the action of a strong field.

The rise in interface defect density is a significant factor contributing to alterations in interface characteristics. Specifically, within the DC compression region of the Si/SiO₂ interfaces, there can be a considerable increase in defect density, which, in turn, leads to substantial changes in C-V (capacitance–voltage) characteristics. The density of interface states has a strong dependence on the electric field intensity, which essentially comes from the electric field difference existing in the charge exchange process between interface states and valence bands.

3.3. Interfaces Under Other Loads

3.3.1. Effect of Mechanical Stress

The presence of SiO₂ gate media leads to changes in the conduction band structure of the Si channel, especially the position of the off Γ-valley can shift down. Further analysis shows that the position change of the off Γ-valley is mainly affected by the interfacial stress and has an important effect on the transmission characteristics of the device. Jung et al. [101] studied the influence of interface stress of SiO₂ gate medium on device performance through first-principle calculation and established the structure model of SiO₂/Si/SiO₂, as shown in Figure 15a. Si/SiO₂ interface tensor and interface energy were obtained through first-principles calculation. Korkin et al. [103] studied the stress of Si/SiO₂ interfaces through modeling simulation calculation. The local density approximation (LDA) and GGA of DFT were adopted. The interface structure was established, and the Si/SiO₂ interfaces were calculated, designed, and optimized. The researchers believe that interfacial stress can be divided into chemical stress, mechanical stress, and electric polarization energy.

Si/SiO₂ interfaces play an important role in the reliability of MEMS devices. In MEMS devices, the main failure factors are cracking and fracture caused by dynamic load brought by mechanical stress. In order to obtain higher reliability of MEMS devices, Li et al. [106] proposed that the use of the deposition method to prepare dielectric materials can effectively improve the reliability of RFMEMS switches. Li et al. found that the mechanical properties of MEMS beams mainly depended on the quality of SiO₂. An important index to measure the quality of Si/SiO₂ interfaces is the residual stress, which is caused by the mismatch of CTE between the SiO₂ deposition layer and substrate. In addition, S, Ho et al. [107] studied the effect of deposition process parameters on interface reliability and found that the increase in the LF/HF power ratio makes the deposition film more compressed and does not reduce the deposition rate.

3.3.2. Effect of Temperature

In a study focusing on the electronic properties of Si/SiO₂ interfaces and their influence on the carrier mobility of Si devices, Liu et al. [107] established the interfaces model. In Figure 15b of the calculated results, the researchers found that LDOS at the Si/SiO₂ interfaces have universal characteristics [108]. The LDOS of the Si layer remained relatively stable from the Si side to the last Si interface layer, while the LDOS of the oxygen atom at the interfaces was significantly enhanced. They propose that the scattering mechanism at the interfaces may be either coherent reflection due to the distortion of the bond angle or incoherent phonon scattering due to the interface phonons. The tensile strain of the Si substrate enhances the hole mobility and decreases the electron mobility, while the compressive strain has the opposite effect. Kovacevic et al. [105] used the CASTEP (Cambridge Serial Total Energy Package) program for geometric optimization and energy calculations. The effects of silicon layer thickness on electronic structure and stress/strain were investigated. The average length, residual stress, dielectric constant, and band structure of the Si/Si bond were studied by analyzing the variation in silicon layer thickness. The author used the ReaxFF model and characterized the interfaces between Si [110] and [111] surfaces (Figure 15c) to establish a variety of interface structures. The stress sensitivity analysis in the table below shows that the stress of silicon is lower than that of silicon oxide under the same strain. The stress in the Si/SiO₂ system is mainly contained in the SiO₂ layer. In systems with a large mismatch of periodic constants, the bonds in the SiO₂ layer are parallel to the interfaces, and these bonds contribute greatly to the stress. For silicon-based devices, the interface mismatch can be reduced by applying high-density ultra-thin SiO₂ films [109,110].

In a system where the silicon layer is bent, the greatest stress is near the oxygen atoms that cause the bending. As is well known, the performance of silicon-based semiconductor devices can deteriorate under high-temperature conditions. The operation of these devices can be altered due to the multitude of defects introduced at elevated temperatures. In the case of the Si/SiO₂ interfaces, high temperatures can dramatically increase the defect density at the interfaces, raise the average energy of the interface atoms, facilitate the dissociation of Si/Si and Si/O bonds, and enhance the migration speed of silicon, oxygen atoms, and other impurities across the interfaces. These combined phenomena contribute to the degradation of the Si/SiO₂ interfaces at high temperatures.

Temperature can affect the Auger recombination of Si. The Auger recombination process is a three-particle process in which an electron recombines with a hole, transferring excess energy to a third free carrier. Ignoring the Coulomb enhancement effect, the Auger recombination rate can be described by Equation (9).

$$R_{Auger}=C_n{n}^{2}p+C_{p}n{p}^2$$

(9)

where C_n and Cp are the Auger coefficients of electrons and holes; n and p are the concentrations of electrons and holes. Under high injection conditions, excess carriers $\Delta \mathrm{n}=\Delta \mathrm{p}$, Auger lifetime can be expressed as Equation (10).

$${\displaystyle \frac{1}{{\tau }_{Auger}}}=C_a\Delta p^2$$

(10)

where C_a is the bipolar Auger coefficient. Wang et al. [111] analyzed the temperature dependence of Auger lifetime in crystalline silicon. Auger life and Ca under 5 × 10¹⁶ cm⁻³ injection conditions vary with temperature, as shown in Figure 16a,b. The bipolar Auger coefficient can be expressed as Equation (11).

$$C_a={\displaystyle \frac{1.1\times {10}^{-28}}{T-193}}+2.1\times {10}^{-33}T$$

(11)

The Auger lifetime is measured in the temperature range from 43 K to 473 K, and the temperature-dependent bipolar Auger coefficients are calculated. The results showed that between 243 K and 303 K, the Auger coefficient dropped and then remained roughly constant at higher temperatures.

In the case of high temperature, most of the trap charge is caused by oxygen vacancy, whose density rises rapidly with the increase in temperature. These vacancies were observed to expand laterally over time at high temperatures, regardless of the type of defect that serves as the nucleus. Four types of paramagnetic defects were observed, two of which were associated with oxide films and two with silicon substrates. By electron spin resonance linear and g factor analysis, the defects associated with oxide films are identified as E′ oxygen vacancy center and E excess Si/O vacancy complex-related defects, both of which involve trapped holes. Early studies on the response of SiO₂ interfaces to high temperatures show that more defects can be introduced into the interfaces at high temperatures. Nie et al. [112] tested the t_eff of Si interfaces at different temperatures using a newly developed experimental device, shown in Figure 16c, which consists of a silicon charge-coupled device camera and a laser that can be used to obtain PL images of Si. The experimental sample is P-type silicon, and a passivation film is deposited on a Si sheet by plasma enhanced chemical deposition system. It was found that the top of the Si wafer has a lower temperature sensitivity, which may be due to its higher impurity density and dislocation density. Devine et al. [113] studied the interfacial vacancy distribution of Si/SiO₂ at high temperatures and used scanning electron microscopy (SEM) and C-V measurement techniques to study the defect generation in the Si/SiO₂ structure during high-temperature annealing. Devine et al. used a vacuum ultraviolet light source to expose the sample through positron or ultraviolet radiation with a photon energy of about 10.4 eV. The decomposition reaction of SiO₂ begins at the site of crystal defects, resulting in the formation of vacancies.

The decomposition reaction of SiO₂ begins at the site of crystal defects, resulting in the formation of a vacancy in SiO₂. The I-V characteristics of the Si/SiO₂ interfaces are measured, the I-V characteristics are closer to a straight line with low curvature at lower temperature. Kumar et al. [16] studied the effects of temperature, bias, and frequency on the dielectric properties of Ni/SiO₂/pSi/Al MIS diodes. The phenomena of energy dissipation in media with different frequencies were studied. The complex permittivity of the diode is expressed as Equation (12) [114].

$$\mathrm{a}=1, {\epsilon }^{\ast }={\epsilon }^{\prime }-i{\epsilon }^{″}$$

(12)

At 1–3 V bias, the ${\epsilon }^{\prime }$ and ${\epsilon }^{″}$ dielectric dispersion in the frequency range of 10 kHz–1 MHz shows that the dispersion decreases with increasing frequency, and the early high-frequency region is almost frequency independent. The difference between ${\epsilon }^{\prime }$ and ${\epsilon }^{″}$ is due to the interaction of Si/SiO₂ interfacial states, which are sensitive to low frequencies, resulting in higher dielectric values in the low-frequency region.

Hofmann et al. [115] used SEM and I-V measurements to study the vacancy size distribution at Si/SiO₂ interfaces. During annealing, the decomposition of SiO₂ is introduced at crystal defect sites, resulting in the creation of vacancies within SiO₂, which have the potential to expand as annealing time is extended. Furthermore, it was noted that the availability of an adequate oxygen supply during the annealing process can avert the breakdown of the dielectric at low electric fields. When the oxygen pressure in the annealing environment reaches sufficient levels, a balance can be achieved between the decomposition and reoxidation reactions of SiO₂, thereby precluding the occurrence of low-field dielectric breakdown. The migration of impurities at the interfaces can be facilitated by high-temperature conditions. Celler et al. [116] used RBS characterization to study the unidirectional mass transport of As at the Si/SiO₂ interfaces at high temperatures and proposed a model of diffusion of As impurities at the interfaces to explain the existence of a stable and uniform SiO₂ layer until the melting temperature of Si. Khan et al. [15] studied the effect of temperature on NBTI in MOS devices. The degradation model of PMOS is described in Figure 16d. NBTI affects the threshold voltage and channel hole mobility of the transistor. The degradation of the threshold voltage is directly influenced by the interfacial traps that are accumulated due to NBTI at the Si/SiO₂ interfaces. When a hole recombines with an electron in the Si/H bond, resulting in only one electron remaining in the bond, the barrier for the dissociation of the Si/H bond is reduced [117]. The result of NBTI is an increase in threshold voltage, a decrease in saturation current, and a decrease in delay. The analysis shows that these effects increase exponentially with the scale of the technology and the increase in operating temperature.

The Si/H bond in the interfaces is most vulnerable to temperature action. The vacancy is close to the Si/H bond, and breaking one Si/H bond can form a hanging bond and an H atom, which has a net energy gain of 1.1 eV. According to Fick’s law, the H diffusion rate ${\mathrm{D}}_{\mathrm{H}}$ can be expressed as Equation (13).

$$D_H=D_{H0}\exp \left(-{\displaystyle \frac{E_a}{kT}}\right)$$

(13)

where ${\mathrm{D}}_{\mathrm{H}0}$ is the H diffusion rate of the temperature at ${\mathrm{T}}_{\mathrm{r}\mathrm{e}\mathrm{f}}$, ${\mathrm{E}}_{\mathrm{a}}$ is the activation energy, k is the Boltzmann constant, T is the temperature. The temperature-dependent N_IT, during negative stress, is given by Equation (14).

$$N_{IT}\left(T(t)\right)=(N_0{E}_{ox}\exp {({\displaystyle \frac{E_{ox}}{E_0}})}^{{\displaystyle \frac{2}{3}}}{({\displaystyle \frac{(k_H[T(t)])}{k_{H2}}})}^{{\displaystyle \frac{1}{2}}}\times {(D_{H2}[T(t)]t)}^{{\displaystyle \frac{1}{6}}}$$

(14)

where E_ox is the oxide field, E_o is the field acceleration constant, k_H and k_H2 are interconversion rates of H and H₂, respectively. The threshold voltage and delay of the transistor are affected due to the change. Khan et al. tested PMOS transistors with oxide layer thicknesses of 90 nm, 65 nm, and 45 nm at 25 °C, 75 °C, and 125 °C. A thinner oxide film thickness is more susceptible to degradation by NBTI, and higher temperatures tend to have greater delay changes and threshold voltage shifts in the same sample. It can be seen that the thicker SiO₂ layer provides higher reliability for the device.

3.3.3. Summary and Discussion

Macroscopic compression or stretching of the interfaces can be caused by mechanical stress at the interfaces. The length of chemical bonds and the bond angles can be changed by the microscopic stress on chemical bonds such as Si/Si, Si/O, and Si/H bonds at the interfaces. When these influences are applied to relatively unstable interfaces, the breakdown voltage and carrier transport characteristics of the interfaces can be changed. Jung obtained through the first-principle modeling calculation that under the action of interface stress, the conduction band structure of Si changes, and the position of the off Γ-valley shifts down.

Korkin also employed first-principles calculations based on DFT, which demonstrated that incoherent phonon scattering is generated at the interfaces as a result of the distortion of bond angles caused by stress. This phenomenon leads to a reduction in the electron mobility of the interfaces under tensile stress conditions. More studies show that Si is less sensitive to stress than SiO₂. According to the experimental results, it can be found that the threshold voltage and delay changes caused by temperature change are more significant in the transistor with a thinner oxide layer. The comparison of some current research data is given in

Table 4. Comparison of research data of Si/SiO₂ interfaces affected by temperature.

Reference	Sample	Temperature (K)	Other Conditions	Time (s)	${\mathbf{N}}_{\mathbf{it}}\left({\mathbf{cm}}^{-2}\right)$	$\mathbf{Diffusion}\mathbf{Depth}\left(\mathbf{nm}\right)$
[94]	Si + 45 nmSiO2	1173			6 $\times$ 1020
[97]	Si + 400 nmSiO2	1678	Injection As	3600		230
	Si + 400 nmSiO2 + 100 nmSi3N4	1678	Injection As	10,800		120
[15]	Si + 65 nmSiO2	398		105	6.91 × 1011
	Si + 65 nmSiO2	298		105	3.66 × 1011
[118]		77			6.7 × 1012
[119]		323			2.8 × 1012

.

This phenomenon is due to the defect of the interfaces under the action of temperature, which leads to the change in the properties of the device. High temperature can affect the recombination and transport of charge carriers at the Si/SiO₂ interfaces. Wang et al. analyzed that the Auger recombination process of charge carriers at the interfaces is affected by temperature. As the temperature increases, the Auger coefficient decreases and becomes stable at about 400 K. Shuai Nie et al. used PL images to analyze the relationship between Si interface temperature sensitivity and dislocation density. In the preparation process of Si/SiO₂, a high-temperature annealing process is usually adopted, and defects can be introduced due to the high temperature during annealing. Devine et al. measured the vacancy distribution at the Si/SiO₂ interfaces at high temperatures by SEM, in which oxygen vacancy is the main source of trap charge. The oxygen vacancy density increases rapidly with the increase in temperature. Hofmann et al. also analyzed the vacancy distribution in the process of high-temperature annealing. Through SEM and IV measurement methods, it was found that the decomposition of SiO₂ started at the crystal defect site. Celler et al. used RBS characterization to study the unidirectional mass transport of As at the Si/SiO₂ interfaces at high temperatures and established the mathematical model of As migration. Khan et al. studied the phenomenon of NBTI in MOS devices affected by temperature. After measuring samples with oxide layers of different thicknesses at different temperatures, it was found that threshold voltage increased and saturation current decreased because high temperatures introduced a large number of defects, especially in thin oxide layers; this phenomenon was more obvious.

4. Interface Characteristics Under Multiple Loads

4.1. Effect of Multiple Loads

In application environments, such as the nuclear industry and aerospace sectors, the Si/SiO₂ interfaces frequently encounter irradiation, high temperatures, stress, and strong electric fields. Subject to these multiple loads, the degradation mechanism of the Si/SiO₂ interfaces can become more intricate, and the rate of degradation may accelerate. Consequently, studying the interface characteristics under these multiple load conditions is of particular significance.

Si/Si bond length gradually decreased with the increase in thickness, and the interfaces’ residual stress changed from compressive stress to tensile stress. The interfaces’ stress has an impact on various characteristics of the interfaces. (1) The stress will change the Si/Si bond length and the bond residual stress of the silicon layer. With the increase in the stress, the static dielectric constant of the interfaces decreases. (2) The increase in stress will lead to a decrease in threshold voltage, as shown in Figure 17a; the change is more obvious under high doping concentration. In addition, the increase in stress also leads to a decrease in the subthreshold swing. (3) The results show that the device performance can be optimized by selecting an appropriate substrate thickness [120]. Lia et al. [121] established 9 Si layers with different thicknesses. The interfacial stress, tensile stress, or compressive stress can be controlled by the thickness of the Si layer without changing the thickness of the oxide layer. As the thickness of the Si layer changes from thin to thick, the interfacial stress gradually changes from negative stress to positive stress.

Corona charge has a positive effect on enhancing the stability of interface passivation at high temperatures. Nie et al. investigated the temperature dependence of the Si/SiO₂ interface recombination process by using temperature and the injection-dependent lifetime spectroscopy method [122]. The experimental results of T_eff of Sinton WCT-120 test samples are shown in Figure 17b,c. Klaassen et al.’s model was calibrated by comparing measurements in steady-state and transient modes [123]. In the sample after the corona charge, a bottom of T_eff appears at the position indicated by the arrow at 373 K, but with the rise in temperature, T_eff recovers the upward trend and reaches the peak. This phenomenon can be interpreted as the degradation of H passivation at high temperatures and the degradation of corona discharge charge at high temperatures. Through further experiments, it was found that the degradation of H passivation at high temperatures is not obvious. The Q_f of the five groups of samples is (1 ± 0.5) × 10¹⁰ qcm⁻².

In order to study the combined effects of strain, voltage, and irradiation load, Wu et al. [21] used a mechanical fixing device to bend the Si sheet that generates a thin oxide layer through thermal oxidation, as shown in Figure 17d. The I-V properties and the time it took for the film to break down were measured using AFM’s Pt probe, and the I-V properties were also re-measured using ⁶⁰Co radiation at a dose of 5 Mrad. According to the bending degree of the Si sheet in the fixed device, the compressive stress is measured according to Equation (15).

$$\sigma =6E\delta {\displaystyle \frac{h}{L^2}}$$

(15)

The tensile stress is calculated according to the Equation (16)

$$\mathrm{a}=1, \sigma =3E\delta {\displaystyle \frac{h}{L^2}}$$

(16)

The compressive stress in the central region of the compression sample is 825 MPa, and the compressive stress in the lateral region is 577 MPa, the stress in the central region of the tensile sample is 660 MPa, and the stress in the lateral region is 173 MPa. The I-V characteristics of compressed and stretched Si slices before irradiation are shown in Figure 17e. The leakage current in the center region 2 is larger and the breakdown voltage is smaller, and the breakdown voltage in the two sides region 13 is larger, which is close to the unstressed region. According to the CVS test of 8.8 V, it is found that the region with the greatest central stress can withstand the 8.8 V voltage for a shorter time without being broken down, regardless of the compressive strain or the tensile strain. This is because the stress in the central region is greater, resulting in an increase in the defect density in the central region. The samples irradiated by ⁶⁰Co show lower breakdown voltage and shorter breakdown time in region 2 of the I-V curve, which is due to the further increase in the defect level at the interfaces. Along the Weibull icon of cumulative failure distribution of breakdown time at different surface locations along the strain axis of the post-irradiation compressed strain oxide sample subjected to nanoscale CVS, the slope of the irradiated sample increases by 70%, which also indicates that irradiation reduces the penetration resistance of the sample.

The density of P_b1 defects gradually decreases as the thickness of the oxide layer increases, while the density of P_b0 defects remains the same, and this decay is faster at higher temperatures. The decrease in P_b1 defects is related to the interfacial stress. When the thickness of the oxide layer increases, the interfacial stress is released, the stress gradient decreases, and the defect density decreases. X-ray irradiation can introduce ${\mathrm{E}}_{\gamma }^{\prime }$ in the interfaces, and hole injection results in the formation of ${\mathrm{E}}_{\gamma }^{\prime }$ and ${\mathrm{E}}_{\delta }^{\prime }$ defects. After annealing, the sample has higher sensitivity to X-ray and hole injection, and the defect density in the interfaces increases greatly. The Si/SiO₂/Si structure after annealing was ground, and the top layer of Si was removed, leaving a certain thickness of oxide layer, and then hole injection was carried out to introduce defects, and the data were obtained after measurement. The E defect does not extend beyond 300 nm from the Si/SiO₂ boundary to the oxide layer. The high-temperature annealed oxide layer will introduce partial defects due to the diffusion of O atoms, resulting in O centers and complexes in the oxide, such as ${\mathrm{E}}_{\delta }^{\prime }$ defects. The O atoms released at the interfaces diffuse into the Si substrate, and this movement of the O atoms is due to differences in the different chemical potentials in the Si and oxide film, which induces O-related defects in the Si.

Jolly et al. [124] studied the properties of Si/SiO₂ interfaces by XPR and EPR and characterized the distribution of suboxide. Concentrations of Si¹⁺, Si²⁺, and Si³⁺ at different depths have little correlation with temperature; the samples thermally oxidized at 1000 °C and the samples thermally oxidized at 460–620 °C are basically the same. Devine et al. [94] irradiated the Si/SiO₂/Si interface structure prepared using CVD with X-rays at a dose of 26 Mrad. Some of the samples were injected with microwave-excited Kr plasma. After treatment, the internal structure of Si/SiO₂ interfaces was characterized by ESR and other means. It was found that high field stress and ion irradiation can affect the interface structure of Si/SiO₂. Paskaleva et al. [84] studied the defects caused by high field stress and ion irradiation. The Si/SiO₂ structure with a thickness of 13 nm was prepared by the thermal oxidation process of the experimental sample. In the experiment, the samples were exposed to N₂ plasma and subjected to high field action of grid substrate, and C-V measurement was performed quickly after treatment. The bridge oxygen vacancy is the most likely defect introduced by high field stress and plasma.

There is no Q_ot change when the gate is biased at 15 V and 30 s, and a positive charge of 4.8 $\times$ 10¹¹ cm⁻² is generated after 10 min, which is less than the charge at −15 V biased at 30 s. This is because when the positive bias is applied, the positively charged H⁺ is subjected to the electric field and migrates to the Si/SiO₂ interfaces. Under the action of high field stress, a positive charge threshold field with an intensity of 8.5 MV/cm is generated. After N ion treatment, gate lining voltage was applied to it, a C-V test was carried out, and Q_ot changes were obtained. Under a certain voltage, the neutralization process of Q_ot follows Equation (17)

$$\Delta Q_{ot}\left(t\right)=\Delta Q_{ot}(sat)\left(1-\exp \left(-At\right)\right)$$

(17)

In response to different biases and time, it can be found that these traps have different spatial and energy positions, and the speed of electron exchange with Si is slower when they are farther away from the interfaces. The plasma-treated sample’s Q_ot neutralization rate under negative bias was significantly higher than that under positive bias. Under the action of plasma treatment and high field stress, the mechanism of defect formation is the impact ionization of oxygen vacancy. XPS analysis shows that the Si/SiO₂ transition region is more sensitive to stress, and the oxides are an inducer of the interface trap.

4.2. Summary and Discussion

Under multiple loads, including stress, irradiation, and electric fields, the degradation of the interfaces is exacerbated. When subjected to a strong electric field, the leakage current at the interfaces primarily originates from voltage-induced V_O. Under the influence of stress, whether it be tensile or compressive stress, the bond lengths and angles at the interfaces can be altered, and V_O is more readily formed. Various types of loads usually promote each other and accelerate interface degradation. The evolution mechanism of Si/SiO₂ interfaces under multiple loads is relatively complex and has many influences on the interfaces. The data comparison of some studies on Si/SiO₂ interfaces under multi-load conditions is shown in

Table 5. Data comparison of Si/SiO₂ interfaces under multiple loads.

Reference	Stress (MPa)	Temperature (K)	Irradiation (Mrad)	Electrical Stress	Dit (cm−2 eV−1)	Breakdown Voltage (V)
[101]		400			1.3 × 1011
		400			4.5 × 1010
[21]	−825		5	Bias: 8.8 (V)		8.3
	+660		5	Bias: 8.8 (V)		8.5
	−825			Bias: 8.8 (V)		8.8
	+660			Bias: 8.8 (V)		9
[94]		1320	26		2.2 × 1012
[125]		413		Bias: −1.8 (V)	8.2 × 1010
		300		Bias: −2 (V)	1.6 × 1010
[126]		398		Electric field strength: 6.3 (MV/cm)	5 × 1012
[90]		373		Bias: −3.17 (V)	5.1 × 1011

.

The residual stress at the interfaces can distort the chemical bond at the Si/SiO₂ interfaces, and the stress acting on the chemical bond can make it easier to be dissociated. Breakdown is more likely to occur under the action of an electric field, and the leakage current can increase. By controlling the temperature of the interfaces and the corona charge time, Nie et al. [101] analyzed the changes in the interfaces’ T_eff and obtained the high-temperature stability of the interface corona discharge and the interface passivation.

In Si/SiO₂ subjected to mechanical stress, irradiation, and voltage at the same time, the leakage current generated by the Si/SiO₂ interfaces subjected to 825 MPa compressive stress and 660 MPa tensile stress under the same voltage is larger than that in the low-stress region, and the breakdown time is shorter under 8.8 V voltage. The leakage current of the strained sample increases further after ⁶⁰Co irradiation, and the high-strain region is more intolerant to irradiation than the low-strain region. Due to the effect of stress, the interfaces are damaged, and Si/SiO₂ becomes more sensitive. Devine et al. applied X-ray and plasma injection treatment to Si/SiO₂ interfaces, and the depth of the holes generated by X-ray irradiation and plasma injection did not exceed 300 nm, and the samples after gas annealing were more sensitive to irradiation and ion injection. Under the combined action of plasma injection and high field stress, the main mechanism of interfacial defects is oxygen vacancy collision ionization.

5. Conclusions

Si/SiO₂ interfaces, as the most used interfaces, constitute a critical structure in devices, which are highly vulnerable under various loads, including irradiation, electric fields, mechanical stress, and high temperatures. Therefore, it is essential to analyze the degradation mechanisms of Si/SiO₂ at the microscopic level. Under complex loads, changes in parameters such as the dielectric constant, resistivity, carrier mobility, and breakdown voltage of Si/SiO₂ interfaces are attributed to the occurrence of numerous defects at the interfaces under external loads, and the interfaces may reconfigure as the interfaces’ suboxide content increases. This paper discusses the research on the effects of external loads on Si/SiO₂ interfaces, analyzes the modeling and characterization methods, and summarizes the mechanisms and the degree of interface degradation under different loads, which can assist scholars in understanding the status and addressing the current issues. On this basis, some challenges and potential research directions are proposed as follows.

5.1. More Attention Should Be Paid to the Study of Si/SiO₂ Interfaces Under Multi-Load Conditions

Devices in service often withstand the combined effects of multiple types of loads, particularly in industries such as aerospace and nuclear applications. When the interfaces are subjected to several loads simultaneously, the degradation mechanism of the interfaces becomes more complex because the effects of multiple types of loads are not simply additive but have a mutual promotion. However, the research on the interface under multiple loads is not sufficient, and the damage mechanism of the interface under multiple loads is not clear. Thus, more research can focus on the coupling effect of thermal, mechanical stress, irradiation, electric field, and other external loads on the Si/SiO₂ interface to simulate the complex environment faced by the interface in real service. A variety of characterization techniques should be used to analyze the change in interface composition and degradation degree of the Si/SiO₂ interface and further analyze its degradation mechanism. On this basis, models for the degradation of Si/SiO₂ interface properties under multiple load conditions should be established to strengthen systematic research on Si/SiO₂ interfaces.

5.2. The Influence Mechanism of Electric Field on Interfaces Should Be Further Analyzed

Under high electric fields and strong voltages, electrical breakdown, morphological changes, and interface structure reconfiguration can occur on the Si/SiO₂ interfaces. The mechanisms of interface breakdown include impact ionization, anode injection, and trap generation. As the external electric field continues to increase, the breakdown can transition from reversible to irreversible. At the microscopic level, P_b defects, chemical bond breakage, and leakage current can be produced in the interfaces under high field stress. However, at present, more studies focus on the impact of Si/SiO₂ interface degradation on device performance parameters, such as DBIE, threshold voltage drift, and S.S. caused by electric field stress. At the same time, the microscopic recombination mechanism of the interface under the electric stress is missing. Thus, further research at the microscopic level should be conducted to target the specific structure of the Si/SiO₂ interfaces and analyze the atomic-level evolutionary process. Based on the study of interface degradation mechanisms, a more reliable theoretical model should be established.

5.3. Improving the Preparation Process to Enhance the Load Resistance of the Si/SiO₂ Interfaces

The preparation process plays a decisive role in the performance of Si/SiO₂ interfaces. By means of process control, the thickness of the oxide layer and the morphology of interfaces can be adjusted to manage interface stress and reduce the distortion of Si/O bonds, enhancing the stability of Si/SiO₂ interfaces. Meanwhile, the stability of the interfaces can be improved by an appropriate gas annealing process. However, the current research stops at the qualitative analysis, and the quantitative analysis of the effect of oxidation, annealing, and other process parameters on the reliability of the Si/SiO₂ interface is insufficient. Thus, future research can focus on the establishment of a database mapping process parameters and interface load resistance and realize the prediction model of the impact of process parameters on the Si/SiO₂ interface through intelligent algorithms, which is conducive to more efficient control of the reliability of the Si/SiO₂ interface through process parameters.