Rapid Thermal Processor

1. Theory

• Definition: A rapid thermal processor heats one wafer with tungsten-halogen or arc lamps at tens to hundreds of K/s, holds it briefly or not at all, and cools it radiatively, giving a far smaller thermal budget than a furnace (spike patent; Moslehi patent).
• Spike anneal: The wafer is ramped to about 1050 ${}^{\circ }\mathrm{C}$ and immediately cooled, with a residence time of about 1 $\mathrm{s}$ within 50 K of the peak (spike patent).
• Radiative cooling: With radiation only, the cool-down rate follows emission from both faces, proportional to $T^4$ (computed from the Stefan-Boltzmann law).
• Uniformity: Radial temperature differences of a few kelvin cause sheet-resistance variation and, above about 900 ${}^{\circ }\mathrm{C}$, slip (Moslehi patent).

2. Working Principle

• Heat: Lamp arrays irradiate the wafer through a quartz or sapphire window in ${\mathrm{N}}_2$ with a controlled trace of ${\mathrm{O}}_2$ (spike patent; Moslehi patent).
• Measure: Pyrometers or infrared probes read wafer temperature and feed multi-zone lamp control (Moslehi patent; Pyrometers page).
• Spike: Ramp at 150 to 250 $K/s$ to the peak, then cool as fast as possible (spike patent).

3. Applications

• Dopant activation: Spike anneals for source, drain and extensions (spike patent; Ramachandran et al.).
• Pre-amorphized junctions: Spike anneals of 200 to 500 eV boron after germanium pre-amorphization (spike patent; Ramachandran et al.).
• Thin oxides: Rapid thermal oxidation in the same lamp chambers (Rapid Thermal Oxidation System page).

Thermal Cycle

Name	Formula	Related Formulae	Variables/Constants
Wafer energy balance	$\rho c_{p}d\frac{dT}{dt}={\alpha }_{abs}{q}_{lamp}-2\epsilon {\sigma }_{SB}\left(T^4-T_w^4\right)$	${\frac{dT}{dt}|}_{cool}\approx -\frac{2\epsilon {\sigma }_{SB}{T}^4}{\rho c_{p}d}$	$\begin{array}{r}\rho ,c_p,d:silicondensity,heatcapacity\\ andwaferthickness\end{array}$ ${\alpha }_{abs}:absorbedfractionoflampflux{q}_{lamp}$ $\epsilon :emissivity$ ${\sigma }_{SB}:Stefan-Boltzmannconstant$ $\begin{array}{r}{T}_w:chamberwalltemperature;two-sided\\ Stefan-Boltzmannbalancewith{\sigma }_{SB}from\\ CODATA(computed),\rho =2329kg/m^3(Ioffe),\\ c_p=983J/(kgK)at1300KfromJANAF\\ 27.614J/(Kmol)andamolarmassof28.08\\ g/molfromIoffedensityandHadleyatom\\ density(computed)\end{array}$ $\begin{array}{r}(assumed:775\mu \mathrm{m}wafer,\epsilon =0.7;computed:\\ naturalcoolingof137K/sat1050{}^{\circ }C,62\\ K/sat800{}^{\circ }C,and59K/sat1050{}^{\circ }Cif\epsilon \\ =0.3)\end{array}$
Spike residence time	$t_R=t\left(T>T_p-50\mathrm{K}\right)$	$t_R\approx \frac{50}{{\dot{T}}_{up}}+\frac{50}{{\dot{T}}_{down}}$	$\begin{array}{r}{t}_R:timewithin50Kofthepeak{T}_p;spike\\ patent:standardspike1.6\mathrm{s},improved\\ spike0.9\mathrm{s}(about40percentsharper),\\ ramp-up50to400K/s(preferablyatleast\\ 180,claimat250),cool-down50to400\\ K/s(preferably90),soakbelow1\mathrm{s}ifany\end{array}$ $\begin{array}{r}(computed:250K/supand90K/sdowngive\\ t_R=0.76\mathrm{s};thestandard1.6\mathrm{s}implies\\ about62K/saverageramps)\end{array}$
Ramp-equivalent diffusion time	$t_{eq}\approx \frac{k_B{T}_p^2}{E_a\dot{T}}$	$L_D=\sqrt{D\left(T_p\right)\left(t_{eq,up}+t_{eq,down}\right)}$	$\begin{array}{r}{E}_a:activationenergyofthediffusing\\ species\end{array}$ $\begin{array}{r}{L}_D:intrinsicdiffusionlength;Laplace\\ estimateoftheeffective-timeintegral\\ (computed;FlashLampAnnealingSystem\\ page);Ramachandranetal.:improvedspike\\ givesmorethan35percentshorter\\ intrinsicdiffusionlengthandmorethan\\ 10percentbetter{R}_S-x_{j}trade-off\end{array}$ $\begin{array}{r}(computedwithboron3.5\mathrm{eV}at1050{}^{\circ }C:\\ 0.17\mathrm{s}upat250K/sand0.48\mathrm{s}downat90\\ K/s,L_D=1.7\mathrm{nm}withoutTED)\end{array}$

Process Window

Name	Formula	Related Formulae	Variables/Constants
Spike process window	$T_p=1000\text{to}1100^{\circ }\mathrm{C}\left(1050\text{typical}\right)$	$O_2=10^4\text{to}2.5\times 10^4\mathrm{ppm},p=5\text{to}100\mathrm{Torr}$	$\begin{array}{r}spikepatent:peaksbelow1000{}^{\circ }Cgave\\ littlediffusion,O_{2}typically1500to\\ 75000ppm(10000to15000ppmbeforethe\\ peakinclaim10),purgebelow5or50\\ ppm,5to10slmcarriergas,boron200to\\ 500\mathrm{eV}at1\times 10^{15}to2\times 10^{15}\mathrm{c}{\mathrm{m}}^{-2}withGe\\ pre-amorphization,x_{j}at1\times 10^{18}\mathrm{c}{\mathrm{m}}^{-3}bySIMS\end{array}$ $\begin{array}{r}(myreading:alittleoxygengrowsathin\\ oxidecapthatlimitsboronout-diffusion\\ andevaporation)\end{array}$
Implant dose unit	$Q=10^{15}\text{to}2\times 10^{15}\mathrm{c}{\mathrm{m}}^{-2}$	$\text{(written as atoms/cc in Ramachandran et al.)}$	$\begin{array}{r}Q:arealdose;spikepatent:atoms/c{m}^2;\\ Ramachandranetal.:thesameimplants\\ writtenas1E15to2E15atoms/cc\end{array}$ $\begin{array}{r}(myreading:adoseisanarealdensity,\\ soatoms/c{m}^{2}iscorrect)\end{array}$
Ultra-shallow junction targets	$x_j<17\mathrm{nm},R_S<760\Omega /\square \left(65\mathrm{nm}\text{node}\right)$	$R_S=542\Omega /\square \left(\text{Ge PAI},B500\mathrm{eV}\right),615\Omega /\square \left(B{F}_{2}2.2\mathrm{keV}\right)$	$\begin{array}{r}Thompsonetal.:millisecondmicrowave\\ annealat110GHzwitha1MWgyrotron,\\ 300000{}^{\circ }C/srampto1300{}^{\circ }C,about2\mathrm{ms}\\ above800{}^{\circ }C\end{array}$ $\begin{array}{r}(myreading:spikeRTPreachesitslimit\\ nearthesetargets,whichiswhy\\ millisecondannealsfollowed)\end{array}$

Uniformity and Sensing

Name	Formula	Related Formulae	Variables/Constants
Radial uniformity limit	$\left|\Delta T\right|<5\mathrm{K}\left(\text{preferably}<2\mathrm{K}\right)$	$T_{edge}<T_{centre}\text{by a few to a few tens of K}$	$\begin{array}{r}\Delta T:radialtemperaturedifference;\\ Moslehipatent:slipespeciallysevere\\ above900{}^{\circ }C,operatingrange250to1250\\ {}^{\circ }C,real-timemulti-zonecontroltokeep\\ gradientsbelowabout2K\end{array}$ $\begin{array}{r}(myreading:thewaferedgeradiatesfrom\\ bothfacesanditsrim,soitrunscold\\ withoutedgecompensation)\end{array}$
Thermal stress	${\sigma }_{th}\approx E{\alpha }_T\Delta T$	$\text{slip severe above}900^{\circ }\mathrm{C}$	$E:Youn{g}^{\prime }smodulus$ $\begin{array}{r}{\alpha }_T:thermalexpansion;Ioffe:E=130GPa\\ along[100]at300K;Swenson:{\alpha }_{T}about\\ 4.41\times 10^{-6}perKat1000K;Moslehipatent:\\ slipparticularlysevereabove900{}^{\circ }C\end{array}$ $\begin{array}{r}(computed:5Kgivesabout2.9MPa;Eis\\ takenatits300Kvalue)\end{array}$
Optical temperature sensing	$T\le 600^{\circ }\mathrm{C}:\text{silicon partly transparent}$	${\lambda }_{probe}\approx 5.3\mu \mathrm{m}$	$\begin{array}{r}Moslehipatent:waferpartlytransparent\\ at600{}^{\circ }Candbelow,coherentinfrared\\ probeatabout5.3\mu \mathrm{m}(COlaser5.3to5.7\\ \mu \mathrm{m}),sapphirewindowwithabouttwenty\\ timesthethermalconductivityofquartz,\\ edgering1/16to1/8inch,\\ tungsten-halogenorxenon-arclamps\end{array}$ $\begin{array}{r}(myreading:low-temperaturetransparency\\ iswhypyrometrybelow600{}^{\circ }Cneeds\\ specialwavelengthsorcontactsensors;\\ seethePyrometerspage)\end{array}$

Important Questions

Ques: What is a spike anneal and how is its sharpness measured?

A spike anneal ramps the wafer to a peak of about 1000 to 1100 ${}^{\circ }\mathrm{C}$ and cools it at once, with no soak or less than 1 $\mathrm{s}$ (spike patent). Sharpness is measured as the residence time $t_R$ within 50 K of the peak. The patent’s standard spike has $t_R$ = 1.6 $\mathrm{s}$ and the improved one 0.9 $\mathrm{s}$, about 40 per cent sharper. For linear ramps, $t_R$ is about $50/{\dot{T}}_{up}+50/{\dot{T}}_{down}$, so its preferred 250 $K/s$ up and 90 $K/s$ down give 0.76 $\mathrm{s}$, close to the improved value (computed). The 1.6 $\mathrm{s}$ standard spike implies average ramps of only about 62 $K/s$ near the peak (computed). Ramachandran et al. report that the improved profile shortened the intrinsic diffusion length by more than 35 per cent and improved the sheet-resistance and junction-depth trade-off by more than 10 per cent. Reference: US patent 6,897,131, ‘Advances in spike anneal processes for ultra shallow junctions’ (USPTO full text; Applied Materials Centura Radiance chamber named; assignee and inventors not shown in the fetched text) (fetch-tool summary); B. Ramachandran, R. Boas and S. Ramamurthy, ‘Solutions for Ultra Shallow Junctions - Improvements in spike anneal’, Applied Materials, Electrochemical Society meeting abstract MA 201-0736 (fetch-tool summary).

Ques: Can a lamp RTP really cool at 400 K per second?

Not by radiation alone. A free wafer cools at about $2\epsilon \sigma T^4/(\rho c_{p}d)$, which for an assumed 775 $\mu \mathrm{m}$ wafer with emissivity 0.7, CODATA’s Stefan-Boltzmann constant and JANAF heat capacity is about 137 $K/s$ at 1050 ${}^{\circ }\mathrm{C}$ and only 62 $K/s$ at 800 ${}^{\circ }\mathrm{C}$ (computed). The spike patent quotes cool-down rates of 50 to 400 $K/s$ but prefers 90 $K/s$. The 90 $K/s$ figure is more likely the realistic radiative rate, since it lies below the computed limit, while rates near 400 $K/s$ would need an extra cooling path or a thinner, more emissive wafer (my reading). Because cooling slows as temperature falls, the down-ramp dominates the thermal budget: at 1050 ${}^{\circ }\mathrm{C}$ it contributes 0.48 $\mathrm{s}$ of equivalent soak against 0.17 $\mathrm{s}$ for a 250 $K/s$ up-ramp (computed). Reference: US patent 6,897,131, ‘Advances in spike anneal processes for ultra shallow junctions’ (USPTO full text; Applied Materials Centura Radiance chamber named; assignee and inventors not shown in the fetched text) (fetch-tool summary); NIST, CODATA 2018 fundamental physical constants wallet card: Stefan-Boltzmann constant 5.670374419e-8 W m-2 K-4, Avogadro constant 6.02214076e23 per mol (fetch-tool summary); NIST-JANAF Thermochemical Tables, Si(cr), table Si-002: Cp = 26.338 J/(K mol) at 1000 K and 27.614 J/(K mol) at 1300 K (fetch-tool summary); Ioffe Institute, NSM archive, ‘Si - Silicon: Mechanical properties’: density 2.329 g/cm3, Young’s modulus 13.0e11 dyn/cm2 along [100] (fetch-tool summary); P. Hadley, TU Graz ‘Silicon’ materials page: 4.995e22 atoms/cm3 (fetch-tool summary).

Ques: Is the spike-anneal implant dose given per square centimetre or per cubic centimetre?

The two Applied Materials sources describe the same boron implants of 200 to 500 $\mathrm{eV}$ with germanium pre-amorphization, but the spike patent gives the dose as $10^{15}$ to $2\times 10^{15}$ atoms/$\mathrm{c}{\mathrm{m}}^2$, while Ramachandran et al. write 1E15 to 2E15 atoms/cc. A dose is an areal density, the number of ions per unit wafer area, so atoms/$\mathrm{c}{\mathrm{m}}^2$ is the correct unit and the abstract’s atoms/cc is a typographical error (my reading). With this dose, the junction depth is defined at $10^{18}$ $\mathrm{c}{\mathrm{m}}^{-3}$ by SIMS. The patent also adds 10000 to 25000 ppm of oxygen during the spike, which likely grows a thin capping oxide that limits boron loss (my reading). Reference: US patent 6,897,131, ‘Advances in spike anneal processes for ultra shallow junctions’ (USPTO full text; Applied Materials Centura Radiance chamber named; assignee and inventors not shown in the fetched text) (fetch-tool summary); B. Ramachandran, R. Boas and S. Ramamurthy, ‘Solutions for Ultra Shallow Junctions - Improvements in spike anneal’, Applied Materials, Electrochemical Society meeting abstract MA 201-0736 (fetch-tool summary).

Ques: Why must RTP keep radial temperature differences below a few kelvin?

Because thickness, activation and slip all depend steeply on temperature. Moslehi’s patent states that gradients should stay below plus or minus 5 K and preferably below 2 K, that the wafer edge can run a few to a few tens of kelvin cooler than the centre, and that slip is especially severe above 900 ${}^{\circ }\mathrm{C}$. Thermal stress is roughly $E{\alpha }_T\Delta T$, about 2.9 $\mathrm{MPa}$ for 5 K using Ioffe’s 130 GPa modulus and Swenson’s $4.41\times 10^{-6}$ per K expansion (computed), and the patent finds slip most severe above 900 ${}^{\circ }\mathrm{C}$. Multi-zone lamp control with real-time temperature feedback addresses this. Below about 600 ${}^{\circ }\mathrm{C}$ silicon is partly transparent, so the patent uses a 5.3 $\mu \mathrm{m}$ infrared probe instead of simple pyrometry (Moslehi patent; Pyrometers page). Reference: M. M. Moslehi, ‘Method and apparatus for real-time wafer temperature uniformity control and slip-free heating in lamp heated single-wafer rapid thermal processing systems’, US patent 4,891,499, Texas Instruments (filed September 1988) (fetch-tool summary); Ioffe Institute, NSM archive, ‘Si - Silicon: Mechanical properties’: density 2.329 g/cm3, Young’s modulus 13.0e11 dyn/cm2 along [100] (fetch-tool summary); C. A. Swenson, ‘Recommended Values for the Thermal Expansivity of Silicon from 0 to 1000 K’, J. Phys. Chem. Ref. Data 12, 179 (1983): about 4.41e-6 per K at 1000 K, extrapolated above 850 K (fetch-tool summary); Pyrometers and Wafer Temperature Sensors page of this vault.

Ques: Why did annealing move beyond spike RTP?

Because junction targets outran the shortest spike. For the 65 $\mathrm{nm}$ node, Thompson et al. cite junction depths under 17 $\mathrm{nm}$ with sheet resistance below 760 ohm per square. Even a sharp spike gives an intrinsic boron diffusion length of about 1.7 $\mathrm{nm}$ (computed), and transient enhanced diffusion adds much more (Furnace Anneal System page). Thompson’s millisecond microwave anneal ramped at 300000 ${}^{\circ }\mathrm{C}$/s to 1300 ${}^{\circ }\mathrm{C}$ and spent about 2 $\mathrm{ms}$ above 800 ${}^{\circ }\mathrm{C}$, giving 542 ohm per square for germanium-amorphized 500 $\mathrm{eV}$ boron and 615 ohm per square for 2.2 keV $\mathrm{B}{\mathrm{F}}_2$, both inside the target. Higher peak temperature improves activation while millisecond times limit diffusion, which is the principle of the laser and flash anneals on the next pages (my reading). Reference: K. Thompson, J. H. Booske, J. Lohr, L. Ives, Y. Gorelov, K. Kajiwara and M. Alvarez, ‘Millisecond Microwave Annealing of Ultra-Shallow Boron Doped Silicon’, University of Wisconsin-Madison, Calabazas Creek Research and General Atomics (fetch-tool summary); Furnace Anneal System page of this vault.

References

1. “Advances in spike anneal processes for ultra shallow junctions,” US Patent 6,897,131. Available: https://image-ppubs.uspto.gov/dirsearch-public/print/downloadPdf/6897131
2. B. Ramachandran, R. Boas and S. Ramamurthy, “Solutions for Ultra Shallow Junctions - Improvements in spike anneal,” Electrochemical Society Meeting Abstracts, MA 201, abstract 0736. Available: https://www.electrochem.org/dl/ma/201/pdfs/0736.pdf
3. National Institute of Standards and Technology, “Fundamental Physical Constants” (CODATA 2018 wallet card). Available: https://www.physics.nist.gov/cuu/pdf/wallet_2018.pdf
4. NIST-JANAF Thermochemical Tables, “Silicon (Si), crystal,” table Si-002. Available: https://janaf.nist.gov/tables/Si-002.html
5. Ioffe Physico-Technical Institute, “Si - Silicon: Mechanical properties,” New Semiconductor Materials. Available: https://www.ioffe.ru/SVA/NSM/Semicond/Si/mechanic.html
6. P. Hadley, “Silicon,” Microelectronics and Micromechanics course materials, TU Graz. Available: https://lampz.tugraz.at/~hadley/memm/materials/silicon/silicon.php
7. C. A. Swenson, “Recommended Values for the Thermal Expansivity of Silicon from 0 to 1000 K,” J. Phys. Chem. Ref. Data, vol. 12, p. 179, 1983. Available: https://srd.nist.gov/jpcrdreprint/1.555681.pdf
8. M. M. Moslehi, “Method and apparatus for real-time wafer temperature uniformity control and slip-free heating in lamp heated single-wafer rapid thermal processing systems,” US Patent 4,891,499, Texas Instruments, 1990. Available: https://patents.google.com/patent/US4891499
9. Pyrometers and Wafer Temperature Sensors page of this vault.
10. K. Thompson, J. H. Booske, J. Lohr, L. Ives, Y. Gorelov, K. Kajiwara and M. Alvarez, “Millisecond Microwave Annealing of Ultra-Shallow Boron Doped Silicon.” Available: https://folk.ntnu.no/skoge/prost/proceedings/aiche-2004/pdffiles/papers/book_054.pdf
11. Furnace Anneal System page of this vault.

Related pages in this vault: Physical Constants, Furnace Anneal System, Rapid Thermal Oxidation System, Pyrometers and Wafer Temperature Sensors, Beamline Ion Implanter, Short Channel MOSFET, Laser Annealing System, Flash Lamp Annealing System.