Why use hafnium oxide in a mos versus silicon dioxide?

Hafnium dioxide (HfO2) is an oxide with a relatively high dielectric constant. As a dielectric material, HfO2 is considered as an alternative field effect due to its high dielectric constant value (~20), large band gap (~5.5eV), and good stability on silicon substrates. Ideal material for conventional SiO2 dielectric layers in transistors. If the size of CMOS devices is lower than 1 μm, the technology of using silicon dioxide as the traditional gate dielectric will bring about a series of problems such as increased heat generation of the chip and loss of polysilicon. As the size of transistors shrinks, silicon dioxide dielectric requirements It must be thinner and thinner, but the value of the leakage current will increase sharply with the smaller thickness of the silicon dioxide dielectric due to the influence of quantum effects, so a more feasible material is urgently needed to replace silicon dioxide as the gate dielectric.

Hafnium dioxide is a ceramic material with wide bandgap and high dielectric constant. Recently, it has attracted extreme attention in the industry, especially in the field of microelectronics. It is likely to replace the core device metal oxide semiconductor field of silicon-based integrated circuits. The gate insulating layer silicon dioxide (SiO2) of the effect transistor (MOSFET) is used to solve the size limit problem of the development of the traditional SiO2/Si structure in the current MOSFET.

The core of the microelectronics industry is the CMOS integrated circuit, and its development level usually marks the development level of the entire microelectronics technology industry.

1. The necessity of shrinking MOS devices

In the past 40 years, CMOS technology has become the backbone of the semiconductor industry, and it has further contributed to the success of the semiconductor industry. From 1956 to 1996, the average growth rate of the semiconductor industry was 17%, while the average growth rate of other industries was only 8%. For a technology to be successful, three conditions must be met:

(1) Must provide a rapidly improving product performance;

(2) The price of new products must be reduced as much as possible in order to develop potential consumer groups;

(3) It must have new application potential and be able to develop new application fields.

Reductions in the size of CMOS devices can help the semiconductor industry achieve this goal.

The reduction in gate size of MOS transistors results in faster circuit switching. This greatly expands the application range of semiconductor products and improves the performance of products. Reductions in transistor size allow more transistors to be integrated on a chip. Therefore, the complexity and functionality of integrated circuits has been greatly increased while keeping circuit manufacturing costs low. Coupled with the use of larger diameter silicon wafers, chip costs are also greatly reduced.

The size reduction of MOS devices conforms to the law of proportional reduction. According to this law, the parameters of the device in the horizontal and vertical directions (such as channel length L, width W, gate dielectric layer thickness tox and source-drain junction depth Xj, etc.) and voltage are all scaled down by the same scaling factor X , while the substrate doping concentration Nb is increased by X times by this factor. At this time, the internal electric field of the device remains unchanged. Since the internal electric field remains unchanged, high electric field effects such as mobility reduction, impact ionization, and hot carrier effects do not occur.

In fact, in the process of scaling down the size of MOS devices, the power supply voltage does not decrease synchronously in the same proportion, which makes the internal electric field of the device stronger. When the gate dielectric thickness of the MOS device is reduced to about 2nm, the gate leakage current increases and the device cannot work normally. At the same time, when the channel length of the MOS transistor is reduced to below 01 um, the electric field intensity of the channel will exceed 1MV/cm. When the channel length is further reduced to the nm scale, the electric field will further increase, and the quantum effect under the strong electric field will affect the device performance, including the change of threshold voltage, the quantization of the inversion layer, the decrease of the effective gate capacitance and the leakage of the pn junction. The current increases and the mobility decreases.

2. The necessity of using high-k gate dielectric for MOS devices

A key factor for the success of the development of the silicon-based microelectronics industry is the excellent material and electrical properties of the gate dielectric material SiO2 that we have been using so far. This material actually exhibits several important properties as a gate insulating material:

(1) Amorphous SiO2 can be thermally grown on a silicon substrate, can precisely control the thickness and uniformity, and can form a low defect density and very stable interface layer with the silicon substrate. At the same time, these defect states and dangling bonds at the SiO2/Si interface can be passivated by post-annealing in an atmosphere with hydrogen.

(2) SiO2 exhibits excellent thermal and chemical stability, which is necessary for the manufacture of transistors, because annealing and oxidation are generally performed at high temperatures (above 1000C).

(3) SiO2 has a wide band gap (9eV), and has a large conduction band and valence band offset compared with Si. Therefore, it has excellent insulation properties, and the breakdown electric field reaches 13MV/cm.

These properties determine that SiO2 is very good as MOSFET gate insulating material. But when the SiO2 thickness is less than 3nm, due to the quantum tunneling effect, carriers can flow through this ultra-thin machine medium. Electricity WKB approximation shows that the probability of tunneling increases exponentially with the decrease of SiO2 thickness. For 1nm thick SiO2, when VoX is 1V, the leakage current density exceeds 100A/cm2. The requirement of ITRS for leakage current is that for high-performance logic circuit applications, the leakage current density should be less than 1A/cm2, and for low-power logic circuit applications, the leakage current density should be less than 1mA/cm2. Therefore, the thickness of 2.2-2.5nm SiO2 is the application limit of low-power logic circuits, and the thickness of 1.4-1.6nm SiO2 is the application limit of high-performance logic circuits. Therefore, it is impossible for SiO2 to be applied to the process of 80nm and below. Even if the silicon nitride oxide technology is used now, 12nm is the limit of the use of silicon nitride oxide, and it can only be extended to the 70nm process. In any case, it is problematic to further reduce the thickness of SiO2 as a gate insulating material (from the material science point of view, the lower limit of SiO2 thickness is 7A, and if the thickness is less than this thickness, there will be no complete bulk bandgap structure).

Another issue related to SiO2 thickness is reliability. When the MOSFET in the integrated circuit is working, the charge flows through the device to cause defects in the SiO2 gate dielectric layer and the SiO2/Si interface. When the critical defect density is reached, the gate dielectric layer breaks down, causing the device to fail under the action of electrical stress. Assuming Breakdown occurs via leakage paths between defects, and Degraeve et al. found that the breakdown versus time of ultrathin SiO2 layers can be replicated well with the leakage method. According to ITRS reliability requirements, the results obtained from this method study show that the limit of SiO2 thickness is about 22nm at room temperature and about 2.8nm at 1500C. Therefore, the limiting thickness of SiO2 is about 2.2nm. Below this thickness, SiO2 is not suitable as a gate dielectric.

Application of Hafnium Oxide in MOS

The field where hafnium-based oxides were first studied and applied was high-K/metal gate technology in advanced CMOS technology. Starting from the 45nm technology node, especially the high-K/metal gate technology commonly used in advanced CMOS technology below 28nm uses hafnium-based oxide as the core material, which was evaluated by the proponent of Moore's Law as the biggest technological revolution since the invention of CMOS technology. Since then, the new type of resistive memory (RRAM) technology represented by hafnium-based oxides is becoming a new generation of transformative integrated circuit technology due to its low-voltage, low-power consumption, high-density integration characteristics and new functions such as storage and computing integration. One of the candidates has been extensively studied. Recently, a new type of ferroelectric effect has been discovered in the hafnium oxide material system. This new type of ferroelectric effect not only maintains the good characteristics of traditional ferroelectricity, but also overcomes the incompatibility with CMOS technology and the size effect of traditional ferroelectric materials. It is difficult to achieve high-density integration and other shortcomings, and it has shown broad application prospects in the application of new neuromorphic devices and systems.

Hafnium-based oxide material properties

The rich physical effects and excellent device performance exhibited by hafnium-based oxide materials are closely related to the crystal structure, electronic and energy band structure properties of the material. Typical hafnium-based oxides such as HfO2 and ZrO2, when combined to form crystals, Hf and O are covalently bonded to form p and anti-p bond structures, as shown in the figure below. The interaction between its valence electrons presents a strong correlation feature. HfO2 and other hafnium-based oxides have a monoclinic stable structure at room temperature and pressure.

Physical effects of hafnium-based oxides and their device applications

1. High K dielectric properties and application of high K/metal gate technology

The so-called high-K dielectric properties refer to some materials called high-K dielectrics, mainly metal oxides, whose dielectric constant is higher than that of SiO 2 . Using high-K dielectric materials instead of traditional SiO 2 in CMOS devices can effectively reduce gate leakage current, so high-K dielectric materials are an inevitable choice to promote the further development of CMOS technology. With its high dielectric constant and stable and excellent chemical and physical properties, hafnium-based oxides were first applied in high-K/metal gate technology to overcome the high gate leakage current caused by SiO 2 gate dielectrics. However, due to the incompatibility of material properties between the polysilicon gate and the high-K gate dielectric, the channel mobility in CMOS devices is significantly reduced and the work function cannot be modulated. Therefore, the hafnium-based oxide high-K/metal gate combination is used The structural replacement of the polysilicon gate/SiO 2 gate structure in traditional CMOS devices has become an inevitable choice for advanced CMOS technology.

In high-K/metal gate technology, in order to optimize device performance, under the guidance of theory, people have carried out research on the modulation and optimization of hafnium-based oxide gate dielectric and metal gate structure by adopting technical methods such as doping and interface engineering. , and achieved expected effects in terms of gate leakage current, threshold voltage modulation, reliability improvement, etc.

2. Resistive switching effect and application of resistive switching devices

The resistive effect refers to the phenomenon that the resistance of some dielectric materials changes under the control of an external electric field. This resistance change can still be maintained after the electric field is withdrawn, and there will be different resistance changes under different external electric fields. response. Devices based on the resistive switching phenomenon are called resistive switching devices. Among them, the resistance of the device can be changed between high and low resistance states.

Since the resistive properties can realize functions such as non-volatile information storage, memristive synapses, and storage-computing fusion, it has potential and broad application prospects in the fields of neuromorphic computing and artificial intelligence.

Using the theoretical method that different doping can modulate the formation energy of oxygen vacancies as a guide, the effective control of the distribution of oxygen vacancies in resistive switching devices can be designed and realized. At the same time, the theory also points out that the resistive switching characteristics of resistive switching devices are related to the distribution of oxygen vacancies, so that the effective regulation of the performance characteristics of resistive switching devices can be realized. Using appropriate doping, combined with optimization of device structure and operation mode, the performance of resistive switching devices can be effectively regulated, which provides an effective technical approach for the design optimization of new resistive switching memristive devices and neuromorphic devices.

3. Ferroelectric effect and application of ferroelectric devices

The so-called ferroelectric effect is a phenomenon in which dielectric materials have spontaneous polarization, that is, in the absence of an external electric field, the dielectric material has macroscopic polarization, with two or more spontaneous polarization states in different directions, and different poles under the action of an external electric field The chemical states can be converted into each other. The typical characteristic of the ferroelectric effect is that the relationship between the polarization intensity and the voltage presents hysteresis characteristics, that is, the electric hysteresis loop.

It should be pointed out that in the study of the ferroelectric effect in hafnium-based oxides, doping has a significant modulation effect on the ferroelectric properties of the device. Although the physical origin of the ferroelectric effect in hafnium-based oxides is still under investigation, studies have shown that its ferroelectric properties are related to the properties of oxygen vacancies. A series of studies have shown that doping can modulate the properties of oxygen vacancies, so the modulation effect of doping on the ferroelectric properties of hafnium-based oxides may be related to the modulation effect of doping on the properties of oxygen vacancies.

Since hafnium-based oxide ferroelectric devices can be prepared by mainstream CMOS technology to achieve high-density 3D integration; at the same time, based on ferroelectric materials, devices with different structures can be constructed, such as ferroelectric memory FeRAM, ferroelectric field effect transistor Fe-FET, Ferroelectric negative capacitance transistor NCFET, ferroelectric tunnel junction FTJ, etc., realize different functional characteristics, which provide a new way for the research and development of brain-like neuromorphic computing and other transformative technologies.

Summarize

Hafnium-based oxide materials have attracted people's attention and widespread attention due to their rich physical effects, stable material properties, and excellent device characteristics, especially the ability to achieve device performance modulation through technical means such as doping and interface engineering. Research. Although after a series of basic and applied research, people have conducted extensive research on the material properties of hafnium-based oxides and their associated physical effects and device characteristics. Dielectric, resistive, ferroelectric and other properties have a significant impact, and the properties of oxygen vacancies can be modulated by methods such as doping and interface engineering, thereby realizing the modulation of device performance, but people are concerned about the inherent properties of hafnium-based oxide materials. The understanding of basic scientific issues such as the characteristics of oxygen vacancies, and the characteristics of oxygen vacancy defects is still far from enough. Therefore, it is very important to carry out basic research on the properties and characteristics of hafnium-based oxide materials and the oxygen vacancy defects formed in them, as well as the microscopic physical origin of related physical effects.