Furnace Anneal vs RTA vs Laser Anneal: Time Scale, Diffusion, Activation and Uses
After ion implantation, wafers are annealed to repair crystal damage and to place dopants on lattice sites so that they become electrically active (activation). The longer the wafer stays hot, however, the more the dopants diffuse and the deeper the junctions become. A furnace anneal heats many wafers together in a furnace for tens of minutes to hours. RTA (rapid thermal annealing) heats one wafer at a time with lamps for only a few seconds, and spike annealing, which barely holds at the peak temperature, belongs to the same family. Laser and flash lamp annealing heat only a thin layer near the wafer surface for a millisecond or less. Shorter heating means less diffusion, so as transistors have shrunk, processes have moved toward shorter anneals.
| Aspect | Furnace anneal | RTA / spike anneal | Laser / flash lamp anneal |
|---|---|---|---|
| Heating time | Tens of minutes to hours | Seconds to tens of seconds; in a spike anneal, the time near the peak temperature is roughly 1 to 2 seconds | A millisecond or less: typically around a millisecond for flash lamps, and milliseconds down to nanoseconds for lasers, depending on the method |
| Ramp rate | Slow, typically a few to around ten-odd °C per minute | Fast, typically tens to hundreds of °C per second | Only the surface heats extremely fast, then cools quickly as heat flows into the wafer |
| Processing unit | Many wafers at once (batch) | One wafer at a time (single wafer) | One wafer at a time; flash lamps heat the whole surface at once, lasers scan a beam across it |
| Heated region | All wafers in the furnace reach a nearly uniform temperature | The whole wafer thickness reaches nearly the same temperature | Only a thin layer near the surface reaches the peak temperature; the bulk stays cooler (at the preheat temperature for flash lamps) and acts as a heat sink |
| Thermal budget | Large | Moderate, far smaller than a furnace anneal | Very small |
| Dopant diffusion | Large; junctions get deeper and transient enhanced diffusion (TED) has a strong effect | Suppressed; shallow junctions are possible, though some diffusion remains | Almost none; the as-implanted profile is largely preserved |
| Activation | Roughly capped by solid solubility at that temperature; raising the temperature also increases diffusion | High activation, since high temperatures can be used | High activation from very high temperature for a very short time; metastable activation above equilibrium solid solubility is possible, but part of it tends to deactivate in later thermal steps |
| Repair of crystal defects | Easy to repair with the long time available | Repairs most defects | Defects may remain because the time is so short; often combined with a spike anneal |
| Typical uses | Well drive-in, film densification, relatively low-temperature forming gas anneals and similar; furnaces of the same type are also used for thermal oxidation and LPCVD | Source/drain activation, silicide formation, thin oxide growth and nitridation | Activation of shallow source/drain and extension regions in advanced logic (combined with spike anneals), and steps where lower layers must not be heated (such as the backside of IGBTs) |
Figures are typical orders of magnitude and vary by process generation and product.