Introduction

Chemical vapor deposition (CVD) is a material deposition method, where the deposition is based on chemical reactions. As comparison to physical vapor deposition (PVD), where the driving force for deposition is physical force, CVD is inherently a completely different mehod. In CVD, reactive precursors are fed to a sealed chamber where the precursors react either in the gas phase,^[1] on the substrate surface, or both in the gas phase and on the surface.^[2] The reactions in the gas phase result in products that are either already the desired material that adheres to the substrate, or to reactive intermediate products that cause chemical reactions with the substrate surface.^[2] The result in both of these cases is film growth on the substrate. The CVD process is illustrated in figure 1.

Figure 1. Principle of chemical vapor deposition. Plasma Electronic's image was used as source of inspiration.^[3]Figure: Samuel Rantataro.

The reactor type in CVD can be a single-wafer reactor or a batch wafer reactor.^[4] Naturally it is easier to optimize the process for a smaller batch size, but single-wafer reactors have a much larger throughput time^[1] and therefore batch reactors are favored if the process is capable for them.

CVD is mainly used to grow compound films and ultra pure elemental films, that would be very difficult or expensive to grow by PVD methods, e.g. epitaxial silicon films. Another very popular use for CVD is to grow silicon oxides and nitrides, which would be impossible to grow thermally because of thermal limitations of previously applied materials.^[1] Also, if deep trenches or holes are desired to be conformally filled, e.g. in the case of interlayer connections of transistors, CVD is the right method.^[4]

In CVD, it is possible to control both the conformality and the uniformity of the deposited thin film by controlling the reaction kinetics (Chapter 2 Kinetics) to suit the needs of the resulting product.^[4] However, atomic scale conformality is not possible with CVD.

The process flow of CVD

The general process flow for thermal CVD (Fig. 1) consists of 5 basic steps:^[4] 

1. Flushing the chamber with known inert gas, e.g. Argon or Nitrogen
2. Heating the chamber to the desired temperature and evacuating the chamber to desired pressure
3. Introducing the reactive precursor to the chamber
4. Chemical reactions and deposition takes place
5. Flushing the chamber with known inert gas

As chemical reactions greatly depend on temperature following the Arrhenius equation shown in equation 1,^[5] the deposition rate of CVD is therefore exponentially proportional to the temperature of the reactive components.^[4] This is more discussed in the Chapter 3.2. The process flow of Plasma-enhanced CVD (PECVD) differs from thermal CVD because in PECVD the required energy is generated by a plasma, not by the ambient temperature. \[ k_s = Ae^{-\frac{E_q}{RT}}\mathrm{\;\;\;\;\;\;\;\;\;\;(1)} \]

   (Eq. 1),^[5] where k_s is the reaction rate, A is a pre-exponential constant that is unique for each reaction, e is natural exponential function, E_a is the activation energy for reaction, R is the universal gas constant and T is temperature.

Since chemical reactions require at least the amount of E_a to take place, this energy has to be produced somehow. Generally the energy is produced either by heating the whole ambient temperature, heating the substrate locally by a laser,^[2] or by applying an electrical discharge to the gaseous media in chamber.^[1][2] Whereas in plasma-enhanced CVD (PECVD) the total energy translated to temperature of singular particles in plasma may be very high, the ambient and substrate temperature still remains relatively low because PECVD takes place in reduced pressure (= reduced number or particles colliding to surfaces and creating heat).^[1] This enables chemical reactions to take place, and therefore enables the deposition of films in lower temperatures.

The more volalite a material is, the more readily the material vaporizes at a given temperature.^[6] Byproducts are desired to be volatile for them to desorp from the substrate surface after reaction has taken place. However, volatility is also needed when liquid or solid precursors are used to enable deposition in lower temperatures.^[2] These are few of the main reasons to use metalorganic precursors in CVD,^[1]  which has been given its own name as Metalorganic CVD (MOCVD).

Precursors

Even though most CVD processes use gaseous precursors,^[4] liquid or solid precursors can also be used.^[2][4] Both liquid and solid precursors are transferred to the chamber by the help of a carrier gas, but while liquid precursors maintain their liquid phase and are transferred as vapors,^[4][7]solid precursors are sublimed.^[7]

Same film can be typically produced with multiple different precursors, as can be seen by looking at the chemical reactions 1. and 2. below. In choosing the right precursor many different aspects play a role:

• What is the temperature required for the deposition process to take place?
• Does the precursor have a possibility to cause unwanted reactions with material on the substrate, or is the chamber material suitable for the precursor?
• What are the byproducts? How easily are they removed? Do they react with the substrate?
• What is the cost of precursor?
• How toxic is the precursor?

As the world is full of different chemical compounds, the total number of CVD precursors is very large. If the reader is more interested in certain precursors and their resulting film composition, there is plenty of information found in the literature. Therefore only a couple of the most used precursors with their chemical reaction is shown here: \[ \mathrm{SiCl_4 (g) + 2H_2 (g) \rightleftharpoons Si (s) + 4HCl (g)}\text{ (1)} \] ^[1]

\[ \mathrm{SiCl_2 H_2 (g) \rightarrow SiCl_2 (g) + H_2 (g)}\text{ (2a)} \] \[ \mathrm{SiCl_2 (g) + H_2 (g) \rightarrow Si (s) + HCl (surface)}\text{ (2b)} \] ^[8] \[ \mathrm{SiH_4 (g) + O_2 (g) \rightleftharpoons SiO_2 (s) + 2H_2 (g)}\text{ (3)} \] ^[1]

\[ \mathrm{3SiCl_2 H_2 (g) + 4NH_3 (g) \rightleftharpoons Si_3 N_4 (s) + 6H_2 (g) + 6HCl (g)}\text{ (4)} \] ^[1] \[ \mathrm{2TiCl_4 (g) + N_2 + 4H_2 (g) \rightleftharpoons 2TiN (s) + 8HCl (g)}\text{ (5)} \] ^[1]

Generally the CVD precursors and reaction byproducts are very nasty substances that are either toxic, flammable, pyrophoric, or corrosive,^[1] or even all of these. Even though here only the simplest precursors are shown as examples, it is good to remember that there is plenty of very complex precursors with intriguing properties available, which is the case with metalorganic precursors.

Kinetics

The deposition rate of CVD varies depending on the chemical reaction's rate and the availability of reagents. If the system would have sufficient amount of energy for the reaction to take place but lacks the reagents, the deposition rate is then mass-transport controlled: this can be thought as of when the reagent diffuses into the reaction site, the reaction begins instantly. However, if the site where chemical reaction takes place is crowded with sufficient number of reagents, then the deposition rate is reaction rate controlled: this can be thought as of sufficient number of reagents, but there isn't enough energy for the reaction to happen faster.

The CVD process is always either mass-transport controlled or reaction rate controlled. Generally the process is desired to be reaction rate controlled, because in this case the conformality and uniformity of the growing film is better.

Mass transfer controlling

Mass-transport coefficient h_g has dependence on pressure and concentration difference between gas phase and substrate surface.^[4] However, typically the concentration in the gas phase is tuned for the chemical reactions themselves. Therefore to change the mass-transport coefficient, pressure of the system must be changed. Equation 2 is the well known Fick's law for diffusion, which states that flux (J) is equal to diffusivity (D) times the concentration gradient on direction x. Equation 3 describes the flux of reagents from gas phase to surface, where h_g is mass-transport coefficient. Equation 4 on the other hand describes the consumption of reagents during the chemical reaction. In equilibrium equation 3 is equal to equation 4. \[ J = -D\frac{\partial C}{\partial x}\mathrm{ (Eq. 2)} \] \[ J_{gas-to-surface} = h_g (C_g - C_s )\mathrm{ (Eq. 3)} \]

^[4] \[ J_{surface-reaction} = k_s C_s\mathrm{ (Eq. 4)} \]

^[4]

This Fick's law is modified into equation 5, where the concentration gradient is determined over the boundary layer thickness (δ). As we see from the equation 5, the mass-transport coefficient h_g must therefore be diffusivity divided by boundary layer thickness, as equation 6 states. Note that here the negative sign describes the directionality. Concentration in gas phase is denoted as C_g and concentration on the surface is denoted as C_s.  \[ -D\frac{\partial C}{\partial x} = -D\frac{(C_g - C_s)}{\delta} = h_g (C_g - C_s)\mathrm{ (Eq. 5)} \] \[ h_g = -\frac{D}{\delta}\mathrm{ (Eq. 6)} \] Changing pressure affects to both the diffusivity (equation 7) and boundary layer thickness (equation 8). Sharp students should notice now that decreasing pressure increases diffusivity but at the same time also increases the boundary layer thickness, which both change the mass-transport coefficient in equation 6 into different directions. However, because the pressure is squared in Equation 8, the change of diffusivity dominates the pressure's effect on mass-transport coefficient: a decrease in pressure leads to larger mass-transport coefficient. \[ D\propto \frac{T^{\frac{3}{2}}}{p}\mathrm{ (Eq. 7)} \]

^[4] \[ \delta = \sqrt{\frac{\eta L}{\nu \rho}} = \sqrt{\frac{\eta L}{\nu}*\frac{RT}{Mp}}\mathrm{ (Eq. 8)} \]

^[4]

where T is the temperature, p is the pressure, η is the fluid viscosity, L is the characteristic dimension of the system, v is the fluid velocity (flow velocity of reaction gases in figure 1), ρ is the fluid density and M is the molar mass of the fluid.^[4]

Reaction rate controlling

Reaction rate coefficient k_s is dependent on multiple different factors, but if we assume that the inflow of reagents, the reagents themselves, and system pressure to be constant, we are left with changing the temperature. Since reaction rate is exponentially dependent on temperature as seen from equation 1, and the mass-transport coefficient has a linear dependency to temperature, changing the temperature plays a larger role for reaction rate than mass-transport. Therefore we can say that the CVD process is reaction rate controlled in small temperatures, and mass-transport controlled in high temperatures. This can be seen in figure 2.

Figure 2. Illustration of deposition rate limitation zones and temperature dependence on growth rate. Figure: Samuel Rantataro (inspired by the original image by Franssila^[4]).

Varying parameters

Temperature

In previous chapters the temperature's effect to the reaction rate has been discussed, but substrate temperature has also other effects relating to the film quality.^[4] Note that now we are discussing about substrate and ambient temperature, translating to the temperature of the deposited film and not solely the energy of singular particles as for example in plasma.

Since temperature can be translated into kinetic energy, the surface diffusion is increased by increasing the system's temperature. This promotes the diffusion of adhered particles to find the most energetically favorable locations, where the binding energy of adatoms is larger compared to planar surface,^[1] such as vacancies, step and kink sites (see figure 3). The natural result of this is better conformality and uniformity, and smaller number of vacancies and other defects.

Figure 3. Illustration of surface sites. The image is from Terrace ledge kink model's Wikipedia page^[9] and it was modified for clarification. Figure: Samuel Rantataro.

Deposition in low temperature leads to less crystalline, sometimes even amorphous films.^[1] This is due to inhibited adatom diffusion, which results to the rate of arriving film growth precursors being larger than the surface diffusion.^[2] As a consequence, grain growth takes place locally nearby the deposition sites instead of in the thermodynamically favourable locations. Smaller grain size and less oriented structure arises.

Pressure

The base pressure of the CVD chamber before introducing precursors can vary from atmospheric pressure (AP) (1 atm = 760 torr = 7,60•10² torr) to ultra-high vacuum (UHV) (10^-6 to 10^-9 torr). The ultra-high vacuum might be needed to get rid of all the residual gases inside the chamber that can have an effect to the chemical reactions or the film quality. The pressure in the system during deposition can vary from atmospheric pressure to 10^-3 torr, with the latter being the UHV system.^[4]

Whereas APCVD generally operates in mass-transport limited zone, changing the system to lower pressure can change the operation point to reaction rate limited zone. If batch processes are desired, the system's operating point cannot be in the mass-transport limited zone, because this would lead into uneven layer thicknesses not only between the wafers, but also wafer-wise. This is illustrated in figure 4.

Figure 4. Effect of pressure to CVD process' operation point and resulting deposited film uniformity wafer-wise, and between wafers. Introduction to Microfabrication, 2nd ed. figure 34.4^[4] was used as source of inspiration.Figure: Samuel Rantataro.

As a conclusion from chapter 2 (Kinetics), decreasing the pressure increases the mass-transport coefficience. Therefore, the reactive gases and depositing materials are more uniformly distributed all over the substrate surface and more conformal growth is expected.

Film properties

Uniformity

Film uniformity means that the film thickness on each planar location on the substrate is similar, which is illustrated in figure 5: The more similar the film thickness in each location (1 to 9, grooves or no grooves) is, the more uniform the film is. Uniformity is generally expressed as noting the actual mathematical non-uniformity all over the wafer: if the average difference of film thickness is 5 %, this is expressed as film uniformity of 5 %. For a film to be uniform, sidewalls don't need to be covered.

Figure 5. Uniformity over the wafer in different locations, and uniformity along surface topography. Figure: Samuel Rantataro.

As film thickness is determined by film growth rate, a uniform reaction rate and number of reactants results in a uniform film. The temperature gradient over the wafer should be uniform, and mass-transport coefficient should be high for a uniform film growth to take place.^[2][4] Single wafer reactors provide a better uniformity compared to batch wafer reactors.^[4]

Conformality

Conformality has a little bit stricter rules compared to uniformity. For a film to be conformal, also the sidewalls have to be covered with equal thickness as can be seen from figure 6. For structures with small aspect-ratio, meaning height:width or depth:width ratios, this is generally not a problem. On the other hand, for simultaneously deep and narrow structures the conformality can be very difficult to achieve.^[4]

To achieve good conformality, the same principles as for uniformity apply: by increasing mass-transport coefficient compared to reaction rate coefficient, better conformality is expected. Changing the precursor can also have a significant effect, since metal-organic precursors are known for their capability to provide very conformal films.^[1][4] However, even though the deposition itself would be conformal, there is no guarantee that the film quality on sidewalls would be equal to those on planar surfaces.^[4]

Figure 6. Illustration of a conformal film. Figure: Samuel Rantataro.

Film stress

Film stresses arise to grown film if its lattice parameter, or coefficient of thermal expansion (CTE) differs much of the substrate's values.^[4] These two stresses accumulate to the film-substrate interface, and must be taken into account when designing the device and choosing materials. Stresses that occur inside the film are called intrinsic stresses. Voids and impurity atoms inside the film generally generate intrinsic stress. Stress relaxation steps such as annealing, where defects and impurities can anneal out, can be done to minimize the intrinsic stresses.^[4]

Film stresses can bring difficulties to wafer handling because the stresses bow the substrate wafer.^[4] Since CVD deposition is generally taking place on both sides of the wafer, films with large stresses might be needed to be removed from the backside of the wafer after CVD processing.

Adhesion

Adhesion between substrate and deposited film must be taken into account to be certain that the film actually can stick to the substrate, and that the adhesion is adequate to withstand the stresses arising to film-substrate interface. Adhesion failure can be seen in many different ways, e.g. as peeling off or film flaking. For example noble metals are known to not adhere well to inert, well-used substrates such as glass and silicon dioxide,^[10]and therefore an intermediate thin film, e.g. chromium or titanium, is grown to improve the adhesion between the substrate and the desired film.^[4][11] The composition of the adhesion improving film is known to have good adhesion to both the substrate and the desired film, hence making the desired film to stay on the substrate surface.

Quick comparison to PVD and ALD

Since this is only a quick comparison, plenty of generalization is done to PVD and ALD process parameters and film quality. The properties that are compared here are pressure, temperature, uniformity and conformality.

PVD

In PVD, the deposition mechanism is based on mechanical energy, and therefore any scattering from residual gases has a great effect on film deposition. Because of this, PVD generally requires a higher vacuum than CVD. The temperature on the other hand plays a smaller role in PVD, and therefore smaller temperatures are allowed. What must be noted in PVD however, is that the substrate temperature has an effect to film quality and as in CVD, elevated substrate temperatures result in better film quality.^[1][4]

Another key concept arising from the mechanical energy nature of PVD is directionality of deposition. This has its own limitation to conformal, or even uniformal film growth because of shadowing effects of surface topography.^[4] Since in PVD the deposition can be thought to be a beam of atoms or molecules, isotropic growth of films is very difficult to achieve.^[1][4] Some very complex systems have been introduced to produce a conformal film with PVD, where magnetic fields control the flight path of evaporated ionized target material.^[1]  However, whereas even CVD can struggle with conformal coating of high aspect-ratio structures, for sputtering this effect is even more pronounced.

Atomic Layer Deposition (ALD)

The main idea in Atomic layer deposition (ALD) is summarized elsewhere and will not be discussed here. ALD enables superior conformality compared to any other film growth method, even into very high-aspect ratio structures. The downside of ALD is the slow deposition rate and the expensiveness of the process.^[2] The temperature window for ALD is rather wide and lowering the process temperature doesn't sacrifice film quality as is the case with CVD.^[2]

References