next up previous contents
Next: 2.6 Electrical Test Up: 2. The Processing Chain Previous: 2.4 Mask Generation

Subsections



2.5 Fabrication

2.5.1 Description of Semiconductor Manufacturing Processes

In the following subsections an overview over the different process steps, a wafer undergoes during its fabrication in the clean-room, is given. A semiconductor manufacturing process differs markedly from other processes. In many other types of processing plants, the material being processed moves through the plant in a fairly simple, straightforward, and well-integrated manner. Despite the fact that the processing flow of this material is straightforward and linear, a flow chart depicting the process will usually be quite complicated. Contrast this with a semiconductor manufacturing process, which can be described very easily with a linear processing flow chart, but whose work-in-process (WIP) moving through the plant will follow complex paths, crisscrossing back and forth in intricate patterns. During wafer processing - i.e. in the semiconductor fabrication clean-room - the integrated circuitry is formed at the surface of the single crystal silicon wafer by numerous repetitive micro-lithographic, deposition, diffusion, and etching steps, until it is finished. During this processing, depending on the complexity of the technology, a set of about fifteen up to more than thirty-five separate wafer processing cycles (which form modules like gate module, LDD module, metal module and so on), including the associated lithography step, were performed. An expanded flow chart of one of these cycles appears as shown in Figure 2.8.

Figure 2.8: Expanded flow chart of the wafer fabrication cycle comprising one alignment step and the associated processing

\includegraphics[width=1.5\textwidth]{figures/wafer_cycle.2.ps}

Here it can be seen that the wafer will iterate through this inner circle as many times as there are masks(alignments)2.1 for adding new circuitry.

2.5.2 Lithography

Micro-lithography is the process of defining useful shapes on the surface of a semiconductor wafer. Typically this consists of a patterned exposure into some sort of photosensitive material already deposited on the wafer. A variety of processes that directly pattern the wafer are possible, such as directly writing on the wafer with an electron beam, or nanoimprinting structures with stamps, but at this time none is in use for high volume semiconductor manufacturing.
The imaging basics of optical lithography have been outlined in Section 2.4 already. Figure 2.5.2 shows the principal components of the illumination system of a typical i-line stepper [43] practical for structure sizes down to 350nm.

Figure 2.9: Schematic view of the main components of the illumination system of a lithography tool
\begin{figure}\par
\centering
\par
\input{figures/lithography_system.pstex_t}
\par
{}
\end{figure}
A detailed description of the lithography process and its modeling basics can be found in [44].

2.5.3 Ion Implantation

Ion implantation is a process whereby energetic dopant ions are made to impinge on a silicon or other target, resulting in the penetration of these ions below the target surface and thereby giving rise to controlled, predictable dopant distributions. Low implant energy produces dopant distributions near the surface such as are required for MOS source and drain regions, or bipolar emitter regions. High energies produce deeply implanted dopant profiles as required for CMOS retrograde wells and buried layers. The provision of wafer cooling during implant allows the use of photo resist masks to laterally control the location of dopant regions, an inevitable feature for the production of CMOS devices. Similarly, topographical features of the device, such as a gate stack, may be used to impose additional masked regions to the implant, thereby allowing for the production of cost- and yield-effective, self-aligned doping regions. A detailed description of the physics of this process step can be found in, e.g., [45],[46],[47],[48]. The general implanter schematic is shown in Figure 2.10.
Figure 2.10: General schematic of an ion implantation equipment
\begin{figure}\centering
\par
\input{figures/implanter_principle.pstex_t}
\par\par
\end{figure}

2.5.4 Diffusion

Diffusion is a key task of semiconductor wafer processing. Although dopants are generally introduced into a wafer by ion implantation, rather than thermally in a furnace, there is unavoidable diffusion of the dopants during any high temperature process step. The models in this area can be categorized into two major approaches, namely, the continuum theory of Fick's diffusion equation and the atomistic theory. The continuum theory requires the solution of Fick's diffusion equation, generally with constant values for the diffusion coefficient and is adequate for low dopant concentrations. When the doping concentrations are high, the diffusion profiles may exhibit anomalous diffusion behavior and a simple form of Fick's law cannot be applied, because the diffusion coefficient becomes concentration dependent. The picture then becomes considerably more complicated and requires an atomistic approach which studies the interactions between native point defects (vacancies and interstitials) and dopant atoms. The underlying idea behind all this is that the dopant atoms mostly dissolve substitutionally in the lattice. Only through interactions with native point defects are the dopant atoms able to jump form one site to another, effecting long-range diffusion.
Due to the agitation of the lattice by phonons some of the defects can wander throughout the lattice. For a simple cubic lattice this diffusion of defects can be understood by considering the jump process between two adjacent (100) planes, 1 and 2. If the lattice planes contain $ n_1$ and $ n_2$ defects per unit surface area, respectively, and the jump rate in either direction is given by $ \nu$ then the number of defects per unit surface area jumping form plane 1 to 2 in time $ dt$ is $ J_1dt=n_1 \nu dt$. For the same jump probability in either direction, the net flux of dopants from plane 1 to 2 $ J_{12}$, can be written as

$\displaystyle J_{12}=J_1-J_2=(n_1-n_2)\nu$ (2.6)

For a small lattice constant $ a$ and assuming that the number of defects changes slowly with distance x (continuum approximation), the above flux can be written as

$\displaystyle J=-\nu a \frac{\partial n}{\partial x}$ (2.7)

The defect concentration per unit volume $ C=(n / a)$, (2.7) becomes

$\displaystyle J = -a^2 \nu \frac{\partial C}{\partial x} = - D \frac{\partial C}{\partial x},$ (2.8)

where $ D = a^2 \nu$ is the diffusion coefficient or diffusivity. The above equation is Fick's first law of diffusion.

$\displaystyle D = D_0 \exp{-\frac{Q}{k_B T}}$ (2.9)

where $ Q$ is the activation energy. $ Q$ is just $ H_m^X$, and $ D_0$ is proportional to the Debye frequency.
A more detailed description of the theory of diffusion in semiconductors can be found in [49],[50] and [51].

2.5.5 Film Deposition

A variety of materials can be deposited by chemical vapor deposition (CVD). Some typical materials are silicon nitride ($ Si_3N_4$), silicon dioxide ($ SiO_2$), TEOS ( $ Si(OC_2H_5)_4$), polycrystalline silicon, and various metals. Depending on the particular deposition method, the temperature varies from about $ 300 ^\circ C$ up to $ 900 ^\circ C$. Additionally the pressure range may vary significantly which differentiates atmospheric pressure chemical vapor deposition (APCVD) from low pressure chemical vapor deposition (LPCVD). A typical characteristic of the deposition process is the deposition rate [nm/min]. Film deposition is used to deposit other materials on top of the silicon wafer. These materials are necessary to build functional parts of the devices (PMOS, NMOS,Bipolar Transistor) and their interconnects in the integrated circuit.
For simulation, deposition is performed by geometry operations, where the deposition rate may vary locally. Details about different simulation approaches for deposition can be found in [52], [53] and [54].

2.5.6 Etching

Etching can be subdivided into two main categories, isotropic and anisotropic etching.

2.5.6.1 Isotropic Etching

This category describes etching rates which are independent of direction. Isotropic etching is usually performed by means of wet chemistry and the wafer is immersed into a reactive solution. The etchant species diffuses towards the wafer surface, dissolution of the specific material takes place, and the generated products separate in turn by diffusion from the surface [55]. The advantages of wet etching are the possible high selectivity for specific materials and the low damage to the substrate.

2.5.6.2 Anisotropic Etching

A typical anisotropic method is reactive ion etching (RIE) [56]. Here ions are accelerated through a Chlorine- or Fluorine-based-plasma towards the wafer surface. The Chlorine (or Fluorine) penetration into the silicon surface is strongly enhanced by the ion bombardment. Since vertical surfaces are less exposed to the ions, a large ratio in the etching rates can be obtained. Furthermore (especially for deep trench etching), passivation of the etched sidewalls is obtained by polymerization of some components of the etching chemistry at the sidewalls. Thus only the bottom of the advancing etching front which is exposed to the ion bombardment is free of these polymers and, therefore, a strong anisotropy of the etched structure can be obtained. One drawbacks of the method is the possible damage to the silicon substrate due to the high energy of the argon ions. Furthermore, the excitation of the plasma field with an RF electromagnetic field induces potentials in the interconnect wires contacting e.g CMOS transistor gates and subsequently damaging the gate oxide of the CMOS transistors (antenna effect).

Footnotes

... masks(alignments)2.1
One has to carefully distinguish between the terms ``mask'' and ``alignment''. A mask is the physical reticle for the illumination process. An alignment is the group of steps performed in a lithography track as shown in Figure 2.8 in the box. The number of masks and alignments is normally not equal, because certain reticles may be used for more than one alignment.

next up previous contents
Next: 2.6 Electrical Test Up: 2. The Processing Chain Previous: 2.4 Mask Generation

R. Minixhofer: Integrating Technology Simulation into the Semiconductor Manufacturing Environment