Semiconductor Fabrication: From Sand to Chip · Module 6
Doping by Diffusion and Ion Implantation
Doping places the right atoms in the right places. This module covers the two ways to do it and how the crystal is repaired afterward.
What you will learn
- How dopants move through silicon by diffusion
- How ion implantation controls the depth and the number of dopant atoms
- Why implanted wafers must be annealed
- Where each doping step is used in a transistor
In plain language
Module 1 showed that a few dopant atoms turn silicon n-type or p-type. A chip needs those atoms only in chosen regions and at chosen depths, such as the source and the drain of each transistor.
The older method is diffusion. The wafer is heated with a source of dopant, and the atoms soak into the silicon the way dye soaks into cloth. The modern method is ion implantation. A machine gives dopant atoms an electric charge, speeds them up with a high voltage, and fires them into the wafer. The voltage sets how deep they go, and counting the charge that arrives sets exactly how many there are.
The impact knocks silicon atoms out of place. The wafer is therefore heated briefly afterward, which heals the crystal and lets the dopant atoms settle into positions where they can supply carriers.
Core lesson
Diffusion
Atoms in a hot solid move at random, and the net flow runs from where there are many to where there are few. The movement speeds up steeply with temperature, so dopant diffusion in silicon is carried out between 900 and 1,100 °C. The common dopants are boron for p-type and phosphorus and arsenic for n-type. Arsenic is a heavy atom that moves slowly, which makes it useful for shallow n-type layers.
Diffusion as a doping method has limits. The concentration is always highest at the surface, the dopant spreads sideways under the edge of the mask, and small amounts are hard to control. Diffusion still matters in every process, because each hot step after doping makes the dopants move again. Engineers call the total heat a wafer receives its thermal budget and keep it as small as they can.
Ion implantation
An implanter turns a gas that contains the dopant into ions. A magnet bends the ion beam so that only ions of the right mass pass through, which keeps the beam pure. The ions are then accelerated to an energy between less than a thousand and several million electron volts and scanned across the wafer. The tool measures the beam current, so the number of ions per square centimeter, called the dose, is known exactly.
Photoresist or oxide blocks the ions wherever doping is not wanted. In the open areas the ions come to rest at a range of depths around an average depth. The resulting profile has a peak below the surface, unlike a diffused profile.
ANNLab, University of Arizona
Figure 1. A typical implanted profile before and after annealing, with illustrative values. Annealing lowers the peak and spreads the dopant deeper.
Silicon is a crystal with open rows between its atoms. An ion that enters along a row can travel much deeper than intended, an effect called channeling. Fabs avoid it by tilting the wafer a few degrees or by disordering the surface first.
Damage and annealing
Each ion displaces thousands of silicon atoms before it stops, and a heavy dose can destroy the crystal order in a surface layer completely. Right after implantation, most dopant atoms also sit between lattice sites, where they supply no carriers.
Annealing fixes both problems. The wafer is heated so that the silicon rebuilds its crystal and the dopants move onto lattice sites, a change called activation. The challenge is to do this without letting the dopants diffuse far. Rapid thermal annealing heats a single wafer to about 1,000 °C for a few seconds. Newer flash and laser methods heat only the surface for a few thousandths of a second.
Where doping is used
- Wells are deep, lightly doped regions that give each type of transistor the right background.
- A light implant under the gate adjusts the threshold voltage.
- Very shallow, low-energy implants form the edges of the source and the drain next to the channel.
- Heavy implants form the main source and drain regions, which must have low resistance.
In the newest transistors, much of the source and drain doping is added while those regions are grown as a crystal film, a method covered in Module 8.
Checking the result
The quickest check is the sheet resistance of the doped layer, measured with four small probes pressed onto the surface. A more detailed method removes the surface atom by atom and counts the dopant atoms at each depth.
Going deeper
These sections add the equations behind the core lesson, each with a worked example. They assume first-year algebra and can be skipped without losing the thread of the course.
To a first approximation the implanted dopant follows a Gaussian curve centered at the projected range Rp, with a spread ΔRp. For a dose Q, the peak concentration is given below.
Npeak = Q / (2.5 ΔRp)
Worked example
A dose of 1 × 1014 ions per cm2 is implanted with ΔRp = 0.055 µm, which is 5.5 × 10−6 cm. The peak concentration is 1014 / (2.5 × 5.5 × 10−6) = 7 × 1018 atoms per cm3. This is the solid curve in Figure 1.
The typical distance a dopant moves in a time t is the diffusion length, √(D t). The diffusivity D rises steeply with temperature.
D = D0 e−Ea / kT
Worked example
For boron in silicon, D0 is about 0.76 cm2/s and Ea is about 3.46 eV. At 1,000 °C, which is 1,273 K, this gives D of about 1.5 × 10−14 cm2/s. In a 10 second anneal the diffusion length is √(1.5 × 10−13) cm, or about 4 nm. In one hour at the same temperature it is about 74 nm. Short anneals are the only way to keep shallow layers shallow.
A doped layer is described by its sheet resistance Rs, in ohms per square. A strip of the layer has a resistance equal to Rs times its length divided by its width.
Rs = 1 / (q μ Q)
Worked example
A layer holds 1 × 1015 active donors per cm2, and the electron mobility in such heavily doped silicon is about 100 cm2/V·s. Then Rs = 1 / (1.6 × 10−19 × 100 × 1015) = 63 ohms per square. A strip ten times longer than it is wide has a resistance of 630 ohms.
Check your understanding
Try each question before opening the answer. The last question uses the Going deeper material.
The ion energy sets the depth. The dose, found by measuring the beam current over time, sets the number of atoms per square centimeter.
The anneal repairs the crystal damage caused by the ions, and it moves the dopant atoms onto lattice sites where they become electrically active.
Dopants diffuse whenever the wafer is hot. A short anneal activates the dopants and repairs the crystal while letting them move only a few nanometers.
Tilting stops ions from traveling along the open rows of the crystal, which would carry them much deeper than intended.
Npeak = 5 × 1013 / (2.5 × 4 × 10−6) = 5 × 1018 atoms per cm3.
Key terms
- Diffusion. The movement of atoms from regions of high concentration to regions of low concentration
- Ion implantation. Doping by firing charged dopant atoms into the wafer
- Dose. The number of implanted ions per square centimeter
- Projected range. The average depth at which implanted ions come to rest
- Channeling. The deep travel of ions along open rows in the crystal
- Annealing. Heating the wafer to repair damage and activate dopants
- Activation. The move of dopant atoms onto lattice sites, where they supply carriers
- Thermal budget. The total amount of heat a wafer receives during processing
- Sheet resistance. The resistance of a thin layer, in ohms per square
In the fab
Implant engineers and technicians run some of the largest machines in the fab. The source gases are toxic, so gas safety and monitoring are a major part of the work.
Further reading
J. D. Plummer and P. B. Griffin, Integrated Circuit Fabrication: Science and Technology, Cambridge University Press. The authors offer a prepublication version that is free to read on the Stanford book site (opens in a new tab).
Previous: Module 5Course homeNext: Module 7, Etching
Created by Prof. Zafer Mutlu, ANNLab, University of Arizona.