General introduction

Silicon technology based nonvolatile memories (NVM) possess the feature of retaining information without any consumption of power, they can maintain the function of a computer hard drive, storing a few bytes up to a few Gigabytes of code and/or data. Prominent examples of applications enabled by NVM are cellular phones, MP3 players, digital cameras, and microcontrollers.[3] (See Figure 1.1) A nonvolatile memory (NVM) device is a MOS transistor that has a source, a drain, an access or a control gate, and a floating gate. The schematic cross-section of a floating gate memory device shows in Figure 1.2. It is structurally different from a standard MOSFET in its floating gate, which is electrically isolated, or "floating".[4]

Nonvolatile memories are subdivided into two main classes: floating gate and charge-trapping. The different series of nonvolatile memory families can be qualitatively compared in terms of flexibility and cost. Flexibility means the possibility to be programmed and erased many times on the system with minimal granularity (whole chip, page, byte, bit); cost means process complexity and in particular of silicon occupancy, i.e., density or, in simpler words, cell size.

In floating gate memory devices, charge or data is stored in the floating gate and

is retained when the power is removed. All floating gate memories have the same generic cell structure. Floating gate devices are typically used in EPROM (Electrically Programmable Read Only Memory) and EEPROM's (Electrically Erasable and Programmable Read Only Memory). Charge-trapping devices were invented in 1967 [5] and were the first electrically alterable semiconductor devices. In charge-trapping memory devices, charge or data is stored in the discrete nitride traps and is also retained when the power is removed. Charge-trapping devices are typically used in MNOS (Metal Nitride Oxide Silicon) [6], SNOS (Silicon Nitride Oxide Semiconductor) [7], and SONOS (Silicon Oxide Nitride Oxide Semiconductor) [8].

The semiconductor industry has experienced many changes since flash memory first appeared in the early 1980s. The growth of consumer electronics market urges the demand of flash memory and helps to make it a prominent segment within the semiconductor industry. These concerns proved the flash memory market began to grow in the early 1990s. Broadly speaking, flash memory ideally suits the consumer electronics market, because it bestows upon electronic devices two qualities that the market demands: mobility and miniaturization. Because of its small, reliable, and nonvolatile properties, numerous applications not practicable with traditional data storage technology are emerging. Flash memory brings mobility and miniaturization to electronics products, two defining features of most consumer electronics products today.[9] These concerns proved to be unfounded as the flash memory market began to grow in the early 1990s (Table 1.1).

The flash memory devices are increasing interest for portable electronic productions because of its high data retention, low cost, and low power consumption characteristics.[10, 11] The basic concepts and the functionality of a floating gate (FG)

applying an external force: the FG stores charge. However, the conventional FG-NVM suffers from charge loss problem as the feature size of the device continues to shrink.[12] A discrete nanocrystal (NC) memory was then proposed as a replacement of the conventional FG memory.[13] The NC memory is expected to preserve the trapped charge efficiently due to the discrete charge storage node, while also demonstrate excellent features such as fast program/erase speeds, low programming potentials, and high endurance.[14] The chart of non-volatile memory history shows in Figure 1.3. Thanks to this characteristic, the nonvolatile memories offer the system a different opportunity and cover a wide range of applications, from consumer and automotive to computer and communication.

Figure 1.1 Main nonvolatile memory applications.

Figure 1.2 Schematic cross-section of a floating gate memory device.

Table 1.1 The rise of the flash memory market

Figure 1.3 Non-volatile memory history.

1.2 Overview of flash memory devices

The Flash memories were commercially introduced in the early 1990s and since that time they have been able to follow the Moore’s law, and the scaling rules imposed by the market. The cell size shrinkage of flash memory from 1992 to 2002 was shown in Figure 1.4. It turns out that the cell size decreases 30 times for each decade, closely following the scaling of the DRAM, today still considered as the reference memory technology that sets the pace to the technology node evolution.[3]

To address the scaling limitation of the insulators surrounding the floating gate,

"thin film storage" (TFS) memories have been developed. Floating gate structure has been widely used for nonvolatile memory application. The schematic cross section of a floating-gate memory device is shown in Figure 1.5. Instead of using floating gates, charges are stored in a thin insulating film which contains storage sites such as traps or small silicon crystals. The charges cannot move easily from site to site, and therefore a single oxide defect does not lead to complete charge loss.[15]

The study of flash memory in the recent 30 years was focused on the charge storage layer, as shown in Figure 1.6. The nitride layer is sandwiched between a thin bottom oxide and a blocking oxide.[16] SONOS devices trap charge in a nitride layer instead of using a poly-silicon floating gate. Within the nitride layer, electrons and holes can be stored in localized traps, with negligible lateral conduction. However, SONOS-type flash memories have several drawbacks such as shallow trap energy level, erase saturation and vertical stored charge migration.[17] The programming speed and operating voltage problem can be solved by reducing the tunnel oxide thickness, while seriously degrades the retention capability of the memory[18].

Due to the drawbacks of the conventional FG device prone to failure of the FG

structure and long-term nonvolatile was introduced. As shown in Figure 1.7,

“nanocrystal” refers to a crystalline structure whose dimensions are small enough to the nanometer scale that its electronic properties begin to resemble those of an atom or molecule rather than those of the bulk crystal. Nanocrystal memories can achieve better reliability and higher bit density than conventional non-volatile memories and thus, have been drawing much attention. In a nanocrystal flash memory device, charge is stored in discrete, mutually isolated, and crystallized nanocrystals or dots.

Each dot typically stores only a handful of electrons; collectively the charges stored in these dots control the channel conductivity of the memory transistor. The electrons stored in the nanocrystal directly above the defect will be affected since the nanocrystals are separated from each other by the gate oxide dielectric.[19] Given a large number of nanocrystals, the cell is immune to local defects of the tunnel oxide.

Many efforts have been made to improve the performance of the NC based nonvolatile memory.[20, 21]

Figure 1.4 DRAM and Flash cell size reduction versus year. The scaling has been of about a factor 30 in ten years.

Figure 1.5 Schematics of floating gate and thin-film storage-based embedded nonvolatile memory bit cells. Depending on the charge stored inside the gate dielectric of a MOS field-effect transistor, the threshold voltage can be very high (off state) or so low that a read voltage applied to the poly-silicon gate can turn on the transistor (on state).

Source: Motorola Inc.

Figure 1.6 Operating principle of traditional NVM based on charge storage on floating gates or in nitride traps (SONOS), embedded into the gate oxide of a MOSFET.

Figure 1.7 Schematic showing that one oxide defect can discharge the entire floating gate, whereas if localized charge is stored, only partial discharge occurs.

1.3 Sol-gel technology derived hybrid materials

The first, although incidental, observation of sol-gel process dates back to 1846.

It covered the hydrolysis and poly-condensation of silicic acid under humidity, which progressed to the point of a silicate glass formation. An extension of the chemical principles involved was shown in 1969-1971, pointing out the importance of reactions of several metal alkoxides in solution under formation of metal (I)–oxygen–metal (II) bonds. This made the possibility of production of defined multi-component oxide glasses, glass-ceramics and crystalline layers. This research effort into the involved basic chemistry followed the path used in the study of reactions of metal-organic compounds.[22] Sol–gel processing has been widely used because it permits fabrication of oxide materials under mild conditions and with a wide range of adjustable experimental parameters. To give a very brief definition, a sol–gel procedure encompasses any process that involves polymerization of soluble precursor molecules to afford a polymeric material, via the intermediate formation of a colloidal sol phase. [23]

Sol-gel process contains fractal geometry and percolation theory in physics, hydrolysis and poly-condensation mechanism in chemistry, sintering and structural relaxation in ceramics, and so on. Figure1.8 illustrates the sol-gel process cycle. The sol-gel process allows synthesizing ceramic materials of high purity and homogeneity by means of preparation techniques different from the traditional process of fusion of oxides. This process occurs in liquid solution of organometallic precursors, by means of hydrolysis and condensation reactions, lead to the formation of a new phase (Sol).

This method was derived from the known chemistry of reaction of a metal halide with a metal alkoxide acting as the oxygen donor. In this reaction, the condensation between M–X and M–O–R forms the M–O–M bridge with the elimination of alkyl

halide.[23, 24] Film-forming oxides are primarily Group 3 through Group 8 elements of the periodic table, such as Al, In, Si, Zr, Ti, Hf, Sn, Pb, Ta, Fe, Ni and several rare earth elements. They can exist as amorphous or (mostly) crystalline layers.[22]

Sol-gel processing can be divided into three main steps, that is, hydrolysis, condensation, and gelation (polymerization), as shown in Figure 1.9. The objective of these reactions is to generate the metal-oxygen-metal (M–O–M) bonds in the reacting solution that make up the oxide material. At hydrolysis step (hydrolysis), metal halides or metal alkoxides react with water to pass into hydroxo metals (≣M–OH) in sol-gel process. Then, two hydroxo metals (≣M–OH) reacts to an oxygen single bridge bond (–O–) between these two metal centers and an byproduct (water), which is called condensation step. Finally, gelation step proceeds continuous to condense numbers of hydroxo metals and link together to form a metal oxide network. The reaction of sol-gel solution and substrate surface are illustrated in Figure 1.10.

Figure 1.8 The sol-gel process cycle.

Figure 1.9 Basic steps of a typical sol–gel process.

1.4 Phase separation for spinodal decomposition

1.4.1 Spinodal decomposition

Spinodal decomposition is a method by which a mixture of two or more materials can separate into distinct regions with different material concentrations. This method differs from nucleation in that phase separation due to spinodal decomposition occurs throughout the material. With the advent of techniques capable of producing metastable crystalline phases with a grain size in the range of 10–100 nm (examples of such processes would include rapid solidification, laser processing and mechanical milling), there is considerable interest in phase transformations in such materials.[25]

The phenomenon of phase separation in glasses was discovered a long time ago.

As early as 1904, Guertler [26] observed immiscibility in alkali borate melts. In general, there are two principal mechanisms of phase separation by nucleation and growth or alternatively by spinodal decomposition. The morphology of the resulting phases is different. If we wish to study phase separation in the unstable region, we must choose systems in which it is possible to bring a homogeneous solution through the metastable region without significant phase separation. Thus we should choose a system in which phase is slow compared with the time while it takes to change the temperature of the sample. The speed of phase separation is related to the diffusion coefficient, which is small for solids and viscous liquids, such as glass forming solutions.[27]

Fick’s first law is used in steady-state diffusion, which is empirical in that it assumes that the diffusion flux is proportional to a concentration gradient. It would be more reasonable to assume that diffusion occurs in order to minimize the free energy so that the flux should be driven by a gradient of free energy:

so that

where the (positive) proportionality constant MA is known as the mobility of A; J is the diffusion flux; D is the diffusion coefficient or diffusivity; C is the concentration.

In this equation, the diffusion coefficient is related to the mobility by comparison with Fick’s first law. [28]

If then the diffusion coefficient is positive and the chemical potential gradient is along the same direction as the concentration gradient. However, if then the diffusion will occur against a concentration gradient.

The diffusion coefficient will be zero when

Figure 1.11 illustrates the miscibility gap defined by curve ad which is obtained by applying the common–tangent construction. Any homogeneous solution cooled into the miscibility gap will tend to decompose into A–rich and B–rich regions with a net reduction in the free energy. The spinodal decomposition of region III is defined by the locus of the points of inflexion on the free energy diagram as a function of temperature. Homogeneous solutions which are cooled within the chemical spinodal can in principle become unstable to infinitesimal perturbations in the chemical composition, leading to the development of A–rich and B–rich regions. Note that if a homogeneous solution is cooled into the region between the chemical spinodal and the miscibility gap, then large composition fluctuations are needed before phase separation can occur; this happens by a process known as nucleation and growth. [29]

Kingon and Maria were first to explain the observed microstructures with the presence of a liquid miscibility gap in the ZrO2–SiO2 system (Figure 1.12).[30] The reliable features in HfO2–SiO2 phase diagram coincide with features of the ZrO2–SiO2

system, this is the crystalline silicate. Metastable phase diagrams and phase hierarchies are used widely in rapidly solidified alloys and can be applied to amorphous materials synthesized from precursors.[31] As shown in Figure 1.13, the situation is more complicated for the ZrO2–SiO2 system (and most likely for HfO2–SiO2) due to the presence of the liquid miscibility gap and the thermodynamically required spinodal.

1.4.2 Comparison with nucleation and spinodal decomposition

An initially homogeneous solution develops fluctuations of chemical composition when the condition fell into the spinodal region. During nucleation and growth, there is a sharp interface between the parent and product crystals; furthermore, the precipitate at all stages of its existence has the required equilibrium composition.

Spinodal decomposition involves uphill diffusion whereas diffusion is always down a concentration gradient for nucleation. The growth of nucleation type and spinodal decomposition type are illustrated in Table 1.2.[28]

Figure 1.11 The miscibility gap and the chemical spinodal.

T

Region I

II III

Chemical spinodal

Figure 1.12 The pseudo-binary ZrO2–SiO2 phase diagram.[31]

Figure 1.13 Metastable phase diagram with the extensions of the liquid miscibility gap and spinodal (dashed), the glass transition temperature (dotted line), and T0

Table 1.2 Comparison of nucleation and spinodal decomposition

Nucleation Spinodal decomposition

Invariance of second-phase composition with time

Continuous variation of both extremes in composition with time

Interface between phases is always same degree of sharpness during growth

Interface between phases initially is very diffuse, eventually sharpens

Tendency for random distribution of particle sizes and position in matrix

Regularity of second-phase distribution in size and position characterized by a geometric spacing

Tendency for separation of second-phase spherical particles with low connectivity

Tendency for separation of second phase, non-spherical particles with high connectivity.

1.5 Motivation

The only stored charges at the nanocrystals adjacent to the defect leak through the tunneling dielectric, compared to huge charge loss of conventional Flash memory due to the lateral charge transport. This signifies that the NC memory has the ability to alleviate the scaling limitations of conventional Flash memory as well as extended the retention time.[32] Various materials have been used to form NCs, such as silicon [33, 34], germanium (Ge) [35] and metal [36], as the charge storage layer for nonvolatile memories.

In this thesis, we proposed a novel sol-gel spin coating method to fabricate high-κ material combined thin film, and annealing for driving the sol-gel film transformed into NCs. The sol-gel solution provided colloidal solvents or precursor compounds when metal halides were hydrolyzed under controlled conditions in a beaker. In the sol-gel reaction, hydrolysis, condensation, and polymerization steps occurred to form metal-oxide networks in the colloid liquid. The most interesting feature of sol-gel processing in the solution was its ability to synthesize new types of materials that were known as “inorganic-organic hybrids.”[24] The film formation with sol-gel spin coating method was a simpler than ALD, PVD, or CVD technologies due to its cheaper precursors and tools. In addition, the film can be fabricated in the normal pressure environment instead of high vacuum system.

The crystallization of transferring the charge trapping thin film into the NC phase during thermal annealing was dependent on the composition of sol-gel solution, preparation solvent, and annealing temperature. The formation of coexisting hafnium silicate and zirconium silicate NC memory had been previously published.[37, 38]

However, the effect of annealing temperature that controlled the formation of NC, degree of crystallization, interfacial energy, and charge retention in the sol-gel derived

memory was still unclear. The formation mechanism to explain the growth of nanocrystal was also still unclear. In the thesis, we had clarified the questionable point with physical and electrical characteristics studies.

Finally, we utilized the best NC formation condition to prepare charge trapping layer for a memory device. We expected the sol-gel derived nanocrystal memory devices with superior electrical characteristics in terms of large memory windows, high write/erase speed, long retention time, good disturbs, and excellent endurance performance.

Chapter 2

Nanocrystal Memory Device principles and operations

2.1 Fundamentals of the NC memory

Nonvolatile memory devices based on the charge storage in discrete charge traps such as semiconductor and metal nanocrystals (NCs) have recently attracted much attention in view of their potential application for high-density nonvolatile memory (NVM) devices.[39] The use of discrete charge-trap of nanocrystals offers an advantage of preventing lateral charge movement, thereby enhances the data retention characteristic.[40] The nonvolatile metal nanocrystal memory devices were extensively investigated over semiconductor nanocrystals because of several advantages, such as stronger coupling with the conduction channel, higher density of states (transport perspectives) than semiconductor (i.e., better charge storage), and a
wide range of available work functions (faster programming speed and better data
retention).[6]

2.2 NC Memory characteristics

2.2.1 Basic program/erase principle

The basic principle of programming the memory is that electrons or holes inject into the charge trapping layer and hence cause the threshold voltage shift. Figure 2.1(a)

shows the cross-section scheme of the nanocrystal memory device. The electrons/holes from channel are injected through tunneling oxide due to the horizontal field of controlling gate to Si-substrate, and therefore trapped into the nanocrystall. Two major mechanisms has been proposed to explain the program of the flash memory, that is, Fowler-Nordheim (FN) tunneling and channel hot-electron injection (CHEI). The energy bands of the nanocrystal memory device at program and erase are illustrated in Figure 2.1(b) and 2.1(c), respectively. During the program process, a positive gate voltage is applied to inject channel inversion-layered electrons into the nanocrystals. During the erase process, a negative gate bias is applied to cause the accumulation layered holes to inject into the nanocrystal and recombine with the trapped electrons..[8] The program and erase process result in the shifting of the threshold voltage, which is proportional to the quantities of trapped charges.

Program

The CHEI mechanism is generally used in flash memories, where a lateral

The CHEI mechanism is generally used in flash memories, where a lateral