CVD dielectrics are mainly used for making interconnections and shallow trench isolation [8]. With device dimensions shrinking and the number of transistors increasing, one layer of metal is no longer enough to connect all transistors, thus, two or more metal layers are employed for making the interconnections.
For IC chips with two metal layers, three metal CVD dielectric layers are needed: pre-metal deposition (PMD), the dielectrics between the two metal layers, or inter-metal dielectrics (IMD), and passivation film deposition.
As device dimensions continue shrinking to sub-micron, more metal layers are needed for making the interconnections. The CMOS IC devices below, with seven layers of metal, needed eight CVD dielectric layers, one pre-metal deposition, six inter-metal depositions and one final passivation film.
But for dynamic random access memory (DRAM), there is only three metal layers are employed for making interconnections in even technology goes to 70nm node generation and beyond.
2.3 Thin film phenomena
Having introduced the phenomena that occur during the growth of thin films, let us explore the properties of thin films, from a structural view-point, we will examine the way of physical, mechanical, and electrical properties, that which are affected by the film structure and microstructure. We will also examine deviations in material properties on going from bulk to thin film.
- 21 -
2.3.1 Physical properties of thin films.
Density or specific gravity of thin films can be determined through weight gain measurements using microbalances. We often observe an increasing in density in thin films with increasing thickness, with the bulk density as the upper limit. This phenomenon has been attributed to the smaller grain size and the increased grain boundary area, which is normally less dense than the grain itself [9]. More often, thinner films tend to contain microscopic voids that decrease as a percentage of the film volume as the film grows thicker. The tendencies that lead to crystallographic perfection generally lead to an increase in film density. Density of the film often correlates to resistivity and other film properties affected by the presence of voids.
Surface roughness arisen from the random nature of nucleation and coalescence [10]. Deviation from the average thickness Δt for films grown at relatively low temperature and at limited surface mobility can be modeled according to a poisson distribution.
Δt ∝ √t
In characterizing roughness, both Δt and the periodicity of the peaks and valleys need to be accounted for. Various optical scattering and surface profile-metric techniques have been developed to characterize roughness.
Another contributor to surface roughness is the presence of surface grooves.
Since higher temperatures often results in large grooves, a direct relationship can be observed between roughness and temperature of growth. In this
- 22 -
situation, measurement of surface roughness yields a direct measuring of the film grain size distribution. Surface roughness acts as an excellent indicator of surface contamination. Often deviation in the film roughness can be directly attributed to leak in the deposition systems and the presence of gaseous impurities such as oxygen [11].
2.3.2 Mechanical properties of thin films
2.3.2.1 Adhesion
The adhesion of grown and deposited films used in ULSI processing must be excellent. If the films lift from the substrate device failure can be result, and thus poor adhesion will represent a potential reliability problem.
Adhesion is also strongly affected by the cleanliness of the substrate.
Contamination generally results in poor adhesion, as does an adsorbed gas layer. Cleaning the substrate prior to deposition is therefore important to insure film adhesion capability. Substrate surface roughness can also affect adhesion [12], for example, increased roughness may promote adhesion because the substrate exhibits more to surface area than to flat surface, and mechanical interlocking between the film and the substrate may also occur. Excessive roughness, on the other hand, results in coating defects, which may promote adhesion failure.
It is highly advantageous to include a layer of a strong oxide-forming element between the oxide substrate and the metallization. This is particularly
- 23 -
true for gold-based metallization, where a chromium layer can be used to serve as an intermediate “adhesion layer” [13] [14]. The intrinsic stress in thin film is generally not sufficient to result in de-lamination, unless the film is extremely thick. More often, high stress results in the cracking of films.
2.3.2.2 Stress in thin film.
Nearly all films are found to be in a state of internal stress, regardless of the means by which they have been produced. The stress may be compressive or tensile: Compressive stressed films would like to expand parallel to the substrate surface, and in the extreme, film in compressive stress will buckle up on the substrate. Films in tensile stress on the other hand, would like to contract parallel to the substrate, and may crack if their elastic limits are exceeded. In general, the stress in thin film is in the range of 108 to 5X1010 dynes/cm2.
The total stress Q, in a film is the sum: a) any external stress Qext, on the film, perhaps from another film; b) the thermal stress Qth: and c) the intrinsic stress Qint. The total stress is written as:
Q = Qext + Qth + Qint
The thermal stress is easily to model, it arises from the difference in thermal expansion between the film and the substrate. During cooling from growth temperature, the film assumed to be tensile stress state if the film wants
- 24 -
to contract more than the substrate will allow. Conversely, the film assumes a compressive stress state if the substrate contracts more than the film wants to.
Intrinsic stress reflects the film structure in ways not yet completely understood [15]. It has been observed that the intrinsic stress in a film depends on thickness, deposition rate, deposition temperature, ambient pressure, method of film preparation, and type of substrate used, among other parameters.
Measurement of stress in the thin film can often be accomplished by measuring the curvature of the substrate, either as a disk or as a strip.
Interference rings, laser holography, traveling microscopes, and optical curvature measurement techniques have been used to measure curvature of disks and deflection of the strip.
Figure 2.5 (a) Tensile stress causes concave bending, and
(b) Compressive stress causes convex bending of substrate.
2.3.3 Electrical properties of thin films.
The electrical conductivity of a material is due to the motion of charge carriers through the lattice under the influence of applied electric fields [16].
Here I will introduce the properties of dielectric film.
- 25 -
As a class of materials, dielectrics exhibit large energy gaps in their band structure, with few free electrons to participate in electrical conduction. Band gaps in dielectrics can be on the order of a few electron volts. The important characteristics of the dielectric constant that affect its usefulness in microelectronic application are the dielectric constant, the breakdown strength, and the dielectric loss. Dielectric constant or permittivity is a measure of the amount of electrical charge a material can withstand at a given electrical field strength, not to be confused with dielectric strength. For a nonmagnetic, non-absorbing material, the dielectric constant is the square of the index of refraction. A vacuum, the perfect dielectric, has a dielectric constant of unity.
The capacitance C in farads of a parallel-plate capacitor shown in figure 2.6 with surface area A and a dielectric of thickness t in centimeters is given by
C = εoεΑ/4 t = 8.85 x 10-14εΑ/t
A high dielectric constant is required to obtain high values of capacitance for a storage capacitor in a DRAM. Interconnect applications require low capacitance between adjacent metal lines. Even though smaller thickness can result in a high value of the capacitance, dielectric strength or breakdown strength is a measurement of the resistance of the dielectric to electrical breakdown under the influence of strong fields [17]. The structural integrity of insulator, the presence of pinholes and metallic contaminants reduce the dielectric strength.
- 26 -
Dielectric loss is a measure of frictional loss, dissipated as heat, in the presence of a varying electric field. The loss occurs because electric polarization in a dielectric is unable to follow the electric field.
Figure 2.6 A parallel-plate capacitor.
2.3.4 Special properties requirements for microelectronics
Even though there are some introduction of general properties of thin films, such as mechanical characteristics and electron transport, there are certain unique requirements for thin films in microelectronics. These requirements are extensions of the properties discussed in the proceeding sections:
2.3.4.1 Conformability and step coverage
- 27 -
Conformability of a thin film refers to its capability to exactly reproduce surface topography of the under substrate. Conditions during growth and subsequent annealing, along with intrinsic properties of the materials determine conformality [18]. In some cases, geometrical constraints of the substrate topography preclude conformality. These concepts are illustrated in figure 2.7. The need for conformality arises because microelectronic processing proceeds by successively depositing and patterning features on thin films. If successive films do not follow the patterns created on the previous layers, voids in deposited layers begin to form. Etching these layers may result in stringers. These can lead to electrical shorts and opens, or to the failure caused by trapped material in the voids.
Figure 2.7 Step coverage and the related terms.
Arrival Angle and surface mobility can contribute significantly to conformality. However, even though surface mobility is a necessary condition for good step coverage, it is far from being sufficient. Conformality over a right-angled step is termed step coverage, the largest arriving angle it does at
- 28 -
the overhang of the step corner [19]. For the same step height, the narrower the gap, the smaller the arriving angle, and the harder it is to fill.
At lower pressure, the mean free path of the particle is longer, so it can affectively reduce the arriving angle at the step corner and improve sidewall step coverage.
Figure 2.8 Arrival angle and surface mobility.
2.3.4.2 Planarity
A related film properties to conformality is palnarity. Lithographic imagers are used in microelectronic manufacturing to pattern very small features on the substrate, these image tools have limited depths of focus and hence require that each successive layer is sufficiently planar [20]. Planarity, can be achieved by many means, some of which are unrelated to CVD.
However, one of the techniques used in planarization is the thermal flowing of doped oxide glasses deposited by CVD under elevated temperature. Figure 2.9 shown a cross-sectional electron micrograph of a thermally deposited
- 29 -
phosphosilicate glass deposited by CVD before and after reflow at a temperature of 800℃. Notice the improvement in the planarity.
Figure 2.9 Reflow achieves planarity in doped oxide films.
2.3.5 Chemical reaction kinetics
Chemical kinetics is the study of the rate and the mechanism by which one chemical species is converted to another. The rate of a reaction is the mass in moles of a product species produced or reactant species consumed in unit time. It can be expressed as
R = Moles with component formed / Mass of solid x time
By the term mechanism it can be account for all the individual collision processes involving the reactant and product atoms that result in the overall rate.
- 30 -
2.3.5.1 Temperature dependence of rate
The rate constant k relates to overall rate of the reaction to the concentration dependent terms. It has been found to be a strong function of temperature and is well represented empirically by the Arrhenius Law [21].
k=A exp (-Ea / kT)
A is the collision frequency term and Ea is the activation energy. The terms collision frequency and activation energy arise out of the concept of an activated complex.
Figure 2.10 Activation energy of the: the change in energy of a reacting species as the reaction proceeds forward the energy needed for the reactants.
AB* is called the activated complex and is in equilibrium with A and B.
The energetic of the reaction are shown in figure 2.10. The activated complex is formed by the collision of A and B molecules and is incapable energetically of existing itself. C is formed out of the decomposition of AB*. The rate of
A, B
- 31 -
formation of C is only dependent on thee concentration of AB* and its rate of decomposition.
2.3.5.2 Surface reaction and mass transfer controlled growths
Because the deposition process includes force convection, boundary-layer diffusion, surface absorption, decomposition, surface diffusion, and incorporation, there are several variables to be controlled. Temperature, pressure, flow rate, position, and reaction ratio all are important factors for high-quality films. The industry has optimized these conditions to improve the film properties.
Since the aforementioned steps for a CVD process are sequential, the one that occurs at the slowest rate will determine the deposition rate. The rate-determining steps can be grouped into gas-phase processes and surface processes [22]. For the gas-phase process, the concern is the rate at which gases impinge on the substrate. This model considers the rate at which gases cross the boundary layer that separates the bulk regions of flowing gas and substrate surface. Such transport processes occur by gas-phase diffusion, which is proportional to the diffusivity of the gas and the concentration gradient across the boundary layer. The rate of mass transport is only relatively weakly influenced by the deposition temperature.
On the other hand, at low temperature the surface reaction rate is reduced, and eventually the arrival rate of reactants exceeds the rate at which they are consumed by the surface reaction process. Under such conditions, the deposition rate is surface reaction rate limited. Thus, at high temperature, the deposition is usually mass transport limited, while at low temperature it is
- 32 -
surface reaction rate limited, as shown in figure 2.11. In actual processes the temperature at which the deposition condition moves from one of these growth regimes to the other depends on the activation energy of the reaction and the gas flow conditions in the reactor. In processes that are under surface-reaction-rate-limited conditions, the deposition temperature is an important parameter [23]. That is, uniform deposition rates throughout a reactor require conditions that maintain a constant reaction rate. This, in turn, implied that a constant temperature must also exist everywhere at all wafers.
On the other hand, under such conditions the rate at which reactant arrive at the surface is not so important, because their concentrations do not limit the growth rate. Thus, it is not so critical that a reactor be designed to supply an equal flux of reactants to all locations of a wafer surface.
Figure 2.11 The deposition rate Rg is a rapid varying function of temp.(T) in the surface-reaction-limited regime of operation, whereas it changes only slowly with temperature in the mass-transport-limited regime (high temperature).
- 33 -
In deposition processes that are mass-transport-limited, the temperature control is not so critical. The mass transport process, which limits the growth rate, is only weakly dependent on the temperature [24]. On the other hand, it is very important that the same concentration of reactants be presented in the bulk gas regions adjacent to all locations of a wafer, because the arrival rates of the reactants are directly proportional to the concentration gradient in the bulk gas. Thus to ensure that films are uniform across a wafer, reactors operated in the mass-transport-limited regime must be designed so that all the locations of the wafer surface and all the wafers in a run are supplied with an equal flux of reactant species.
2.3.6 Plasma fundamental
Plasma processes are widely used in the semiconductor industry for etch, CVD, PVD, and photo-resist strip. Plasma is a quasi-neutral gas of charged and neutral particles which exhibits collective behavior.
2.3.6.1 Physical characteristics of plasma
Any body of gas typically contains three species: neutral atoms or molecules, ions and electrons. Their relative concentrations (for example, the degree of ionization) are considerably different in plasma. The concentration of ions in the atmosphere is negligibly small. In typical glow discharge plasma used in CVD processes, ionic concentrations are of the order of 1010 per cm3. The electron concentration is the same as the ion concentration, so the overall plasma is electrically neutral.
- 34 -
From fundamental kinetic theory, the kinetic energy of a particle is given by
ε = ½mν2
An understanding of the interactions between the three types of species and the energetic involved in these processes is essential for us to make use of plasmas in producing thin films. Energy transfer into plasma and between the species, diffusion phenomena within the plasma and chemical reactions in the presence of charge species are some of the processes that depend on collisions between the three species.
Electron concentration 1010 to 1011 per cm3 Ion concentration 1010 to 1011 per cm3 Electron temperature 3-5eV
Ion temperature 0.05eV Electron velocity 107 cm /s Ion velocity 104 cm /s Table 2.1 Properties of a typical glow discharge.
2.3.6.2 Plasma chemistry
In a field-free space, charged particles behave the same way as neutrals, and their behavior can be treated similarly using the kinetic theory of gases
- 35 -
[25]. Collisions between particles can be written in terns of a mean free path λ given by
λ=1 / √2 nσ
n is the number density of the particles and σ is the collision cross section.
There are many kinds of inelastic collisions happening simultaneously in plasma. Three of them are the most important in CVD process, ionization, excitation-relation and dissociation.
Ionization in the plasma can occur through many mechanisms; the simplest is electron capture. For neutral species having high electron affinity, the following reaction occurs readily:
F + e- → F-
Electron capture contributes significantly to electron loss in halogen-containing plasma. An important mechanism for the maintenance of a glow discharge is the production of ions through electron impact. For instance, collision between an energetic electron and a xenon atom produces a xenon ion and another electron.
Xe + e- → Xe+ + 2 e-
- 36 -
The two electrons are now accelerated by the potential gradient in a sheath to ionize more neutral, starting a chain reaction. The electrons have to posses a higher energy than the ionization potential of the neutral (~12eV for xenon) for this process to occur.
Excitation-Relaxation is an extreme case of the various excited states that an atom or a molecule can reach on electron impact, similar to ionization.
There are electron energy thresholds equal to the energy of the first excited state, and they need to be exceeded for the excitation to occur.
e- + A → A* + e-
A* → A +h ν
where h is plank constant, and ν is the frequency of the glow light. Different atoms or molecules have different frequencies, that which is why different gases have different glow colors. Oxygen glow is grayish blue, nitrogen glow is pink, and fluorine glow is red, etc.
When a molecule dissociates upon electron impact either its constituent atoms or further ionized products. Of more relevance to CVD is the formation of radicals in the plasma [26]. When an electron collides with a molecule, it can break the chemical bond and generate free radicals which are molecular fragments with unpaired electrons:
- 37 -
Collision Byproducts Energy of Formation
e-+ SiH4 SiH2+ H2+e- 2.2eV SiH3+ H +e- 4.0eV Si + 2H2 +e- 4.2eV SiH+ H2+H + e- 5.7eV SiH*+ 2H +e- 8.9eV Si*+ 2H2 +2e- 9.5eV SiH2++ H2+2e- 11.9eV SiH3++ H +2e- 12.32eV Si++ 2H2+2e- 13.6eV SiH+ + H2+H + 2e- 15.3eV
e-+ SiH4 SiH2+ H2+e- 2.2eV SiH3+ H +e- 4.0eV Si + 2H2 +e- 4.2eV SiH+ H2+H + e- 5.7eV SiH*+ 2H +e- 8.9eV Si*+ 2H2 +2e- 9.5eV SiH2++ H2+2e- 11.9eV SiH3++ H +2e- 12.32eV Si++ 2H2+2e- 13.6eV SiH+ + H2+H + 2e- 15.3eV