1. © 2000 by Prentice Hall Upper Saddle River NJ ION IMPLANTATION Dr. Wanda Wosik ECE 6466, F2012 Chapter 8

2. © 2000 by Prentice Hall Upper Saddle River NJ Basic Concepts Ion implantation is the dominant method of doping used today. In spite of creating enormous lattice damage it is favored because: Large range of doses - 10 11 to 10 16 /cm 2 Extremely accurate dose control Essential for MOS V T control Buried (retrograde) profiles are possible Low temperature process Wide choice of masking materials There are also some significant disadvantages: Damage to crystal. Anomalous transiently enhanced diffusion (TED). upon annealing this damage. Charging of insulating layers.

3. © 2000 by Prentice Hall Upper Saddle River NJ A. Implant Profiles At its heart ion implantation is a random process. High energy ions (1-1000keV) bombard the substrate and lose energy through nuclear collisions and electronic drag forces. Profiles can often be described by a Gaussian distribution, with a projected range and standard deviation. (200keV implants shown.) (1) (2)or where Q is the dose in ions cm -2 and is measured by the integrated beam current. Heavy atoms have smaller projected range and smaller spread = struggle  R p Doses 1*10 12 cm -2 to 1*10 16 cm -2 used in MOS ICs

4. © 2000 by Prentice Hall Upper Saddle River NJ Ranges and standard deviation ∆Rp of dopants in randomly oriented silicon. Energy Dependence R p and  R p for dopants in Si.

5. © 2000 by Prentice Hall Upper Saddle River NJ Monte Carlo simulations of the random trajectories of a group of ions implanted at a spot on the wafer show the 3-D spatial distribution of the ions. (1000 phosphorus ions at 35 keV.) Side view (below) shows Rp and ∆Rp while the beam direction view shows the lateral straggle. R p =50 nm,  R p =20 nm Lateral struggle  R| 3D Distribution of P Implanted to Si

6. © 2000 by Prentice Hall Upper Saddle River NJ The two-dimensional distribution is often assumed to be composed of just the product of the vertical and lateral distributions. (3) Now consider what happens at a mask edge - if the mask is thick enough to block the implant, the lateral profile under the mask is determined by the lateral straggle. (35keV and 120keV As implants at the edge of a poly gate from Alvis et al.) (Reprinted with permission of J. Vac. Science and Technology.) The description of the profile at the mask edge is given by a sum of point response Gaussian functions, which leads to an error function distribution under the mask. (See class notes on diffusion for a similar analysis.) Lateral Implantation - Consequences for Devices

7. © 2000 by Prentice Hall Upper Saddle River NJ B. Masking Implants How thick does a mask have to be? For masking, (4) Calculating the required mask thickness, (5) The dose that penetrates the mask is given by (6) Dose that penetrates the mask Depends on mask material Lateral struggle important in small devices

8. Masking Layer in Ion Implantation Photoresist, oxide mask Lateral struggle important in small devices Dose that penetrates the mask To stop ions: Poly thickness

9. Masking Efficiency Mask edges tapered – thickness not large enough Tilted implantation ( “ halo ” ) – use numerical calculations ( ex. to decrease short channel effects in small devices) Shadowing effect  rotate or implant at 0 Deg.

10. © 2000 by Prentice Hall Upper Saddle River NJ C. Profile Evolution During Annealing Comparing Eqn. (1) with the Gaussian profile from the last set of notes, we see that ∆Rp is equivalent to. Thus (7) The only other profile we can calculate analytically is when the implanted Gaussian is shallow enough that it can be treated as a delta function and the subsequent anneal can be treated as a one-sided Gaussian. (Recall example in Chapter 7 notes.) (8) or

11. Implantation Followed by Annealing  Function rediffused Annealing requires additional Dt terms added to C(x)  Cp , depth , C(x) remains Gaussian. Backscattering of light atoms. C(x) is Gaussian only near the peak.

12. © 2000 by Prentice Hall Upper Saddle River NJ Real implanted profiles are more complex. Light ions backscatter to skew the profile up. Heavy ions scatter deeper. 4 moment descriptions of these profiles are often used (with tabulated values for these moments). Range: Std. Dev: Skewness: Kurtosis: (9) (10) (11) (12) Real structures may be even more complicated because mask edges or implants are not vertical. Arbitrary Distribution of Dopants Pearson ’ s model good for amorphous (&fine grain poly-) silicon or for rotation and tilting that makes Si look like amorphous materials.

13. © 2000 by Prentice Hall Upper Saddle River NJ D. Implants in Real Silicon - Channeling At least until it is damaged by the implant, Si is a crystalline material. Channeling can produce unexpectedly deep profiles. Screen oxides and tilting/rotating the wafer can minimize but not eliminate these effects. (7˚ tilt is common.) Sometimes a dual Pearson profile description is useful. Note that the channeling decreases in the high dose implant (green curve) because damage blocks the channels.

14. Channeling Effect As two profiles Dual-Pearson model gives the main profile and the channeled part. Dependence on dose: damage by higher doses decreases channeling. No channeling for As @ high doses Parameters are tabulated (for simulators). Include scattering in multiple layers (also masks ’ edges). IMPORTANT in small devices! Screen oxide decreases channeling. But watch for O knock-out. Channeling not forward scattering c-Si, B

15. © 2000 by Prentice Hall Upper Saddle River NJ Channeling P implantation at 4- keV and low dose Q<10 13 cm -2 8° 0°

16. Two – Dimensional Distributions Near the mask edge 2D distributions  calculated by MC model should be the best – verification difficult due to measuring problems. Phenomenological description of processes is insufficient for small devices. Atomistic view in scattering Thin oxide Verification through SIMS Poly-Si

17. Manufacturing Methods and Equipment Mass Analysis Ion velocity Mass Selection AsH 3 PH 3 BF 2 in 15% H 2, all very toxic m  r Gives mass separation Integrate the current to determine the dose For low E implant no acceleration Neutral ions can be implanted (w/o deflection=center) but will not be measured in Dose (use trap) Ion beam heating T increases - keep it below 200 °C Centrifugal force Lorentz force B ++, B +, F +, BF, BF 2 +

18. High Energy Implants Applications in fabrication of: wells (multiple implants give correct profiles ex. uniform or retrograde), buried oxides, buried layers (MeV, large doses)! - replaces highly doped substrate with epi-layers CMOS U EB U BE 0.7V p-n-p n-p-n Decrease of R sub - less latch-up Thyristor structure In latch-up Future IC fabrication: implantation at high energy becomes more important - reduction of processing steps

19. Ultralow Energy Implants Required by shallow junctions in VLSI circuits (50 eV- B) - ions will land softly as in MBE Extraction of ions from a plasma source ~ 30keV Options: Lowering the extraction voltage V out  the space charge limited current limits the dose (Child’s Law) J  V 1/2 ext d -2 ex. J 2keV =1/4J 5keV Extraction at the final energy used in the newest implantors but not for high doses due to self limitation due to sputtering at the surface. Now 250 eV available,; 50 eV to come Deceleration (decel mode) more neutrals formed and implanted deeper that ions (doping nonuniformities) Transient Enhanced Diffusion (TED) present in the low energy Ion Implantation and {311} defects.

20. Models and Simulations Rutherford(1911) -  (He) backscattered due to collision with a + nucleus. Bohr- the nuclear energy loss due to + atoms cores and electronic loss due to free electrons decrease Many contributors. Lindhard, Scharff and Schiott (1963) (LSS)

21. © 2000 by Prentice Hall Upper Saddle River NJ Modeling of Range Statistics The total energy loss during an ion trajectory is given by the sum of nuclear and electronic losses (these can be treated independently). (13) (14) A. Nuclear Stopping An incident ion scatters off the core charge on an atomic nucleus, modeled to first order by a screened Coulomb scattering potential. (15) This potential is integrated along the path of the ion to calculate the scattering angle. (Look-up tables are often used in practice.) S n (E) in Eqn. (14) can be approximated as shown below where Z 1, m 1 = ion and Z 2, m 2 = substrate. (16) Nuclear stopping power The range Scattering potential Role of electrons in screening Thomas Fermi model Energy transferred Head-on collision (max energy transferred) Z 2, m 2 Computers used to find R Elastic collisions

22. © 2000 by Prentice Hall Upper Saddle River NJ B. Non-Local and Local Electronic Stopping Drag force caused by charged ion in "sea" of electrons (non-local electronic stopping). Collisions with electrons around atoms transfer momentum and result in local electronic stopping. (17) To first order, where C. Total Stopping Power The critical energy E c when the nuclear and electronic stopping are equal is B: ≈ 17keV P: ≈ 150keV As, Sb : > 500keV Thus at high energies, electronic stopping dominates; at low energy, nuclear stopping dominates. Energy loss w/o the trajectory change Nonlocal Local Inelastic Collisions with electrons  momentum transfer, small change of the trajectory.

23. © 2000 by Prentice Hall Upper Saddle River NJ Damage Production E d =Displacement energy (for a Frenkel pair)  15eV  large damage induced by Ion Implantation 30 keV As  Rp  25mm E decreases to E d so that ions stop. n= Number of displaced Si atoms  Dose – large damage!  Si Consider a 30keV arsenic ion, which has a range of 25 nm, traversing roughly 100 atomic planes.

24. Time for the ion to stop 1 ion  primary damage: defect clusters, dopant-defect complexes, I and V Increment in damage more recombination for heavy ions since damage is less dispersed than for light ions: B-0.1, P-0.4, As-0.6, BF2-0.6. Damage in Implantation Time Scale in Implantation Molecular dynamics simulation of a 5keV Boron ion implanted into silicon [de la Rubia]. Note that some of the damage anneals out between 0.5 and 6 psec (point defects recombining). Damage evolution (atomic interaction) stabilization @ lower concentrations due to local recombination Damage related to dose and energy Fraction of recombined defects (displaced atoms) Damage accumulates in subsequent cascades and depends on existing N -local defects

25. Damage in Implantation Including Amorphization Damage is mainly due to nuclear energy losses : for B @ Rp. For As – mostly everywhere in the Dopant profile.  - Si forms @ large doses and spread wider with the increasing Q.  - Si forms @ low T of II (LN 2 ), @ RT or higher – recombination  (in-situ annealing)  - Si is buried Preamorphization eliminates the channeling effect Cross sectional TEM images of amorphous layer formation with increasing implant dose (300keV Si -> Si) [Rozgonyi] Note that a buried amorphous layer forms first and a substantially higher dose is needed before the amorphous layer extends all the way to the surface. These ideas suggest preamorphizing the substrate with a Si (or Ge) implant to prevent channeling when dopants are later implanted. Critical energy for amorphization E c ~f(10 21 keV)/cm 3 (f=0.1-0.5 Si) For 100keV As implantation D=6x10 13 cm -2 Si used for amorphization

26. © 2000 by Prentice Hall Upper Saddle River NJ Damage Annealing - Solid Phase Epitaxy If the substrate is amorphous, it can regrow by SPE. In the SPE region, all damage is repaired and dopants are activated onto substitutional sites. Cross sectional TEM images of amorphous layer regrowth at 525˚C, from a 200keV, 6e15 cm -2 Sb implant. In the tail region, the material is not amorphized. Damage beyond the a/c interface can nucleate stable, secondary defects and cause transient enhanced diffusion (TED).

27. Damage Annealing (more) Formation of End-of-Range (EOR) defects @ a/c interface in Si  large damage after II @ the C-Si side but below the threshold for amorphization. Loops R= 10 nm grow to 20 nm in 1000 °C Solid Phase Epitaxy Furnace 850 °C RTP 1000 °C 5 min 60 min 960 min 1 sec 60 sec 400 sec  1000 °C gives stable dislocation loops 1100 °C/60 sec may be enough to remove the dislocation loops. Loops in P-N junctions  leakage Optimize annealing: Short time, high T to limit dopant diffusion but remove defects Optimize I 2 : LN 2 Ge 4*10 14 cm -2 RT- 5*10 14 cm -2 produces a-Si @ RT, EOR @ 100 nm depth   =25 nm, 10 10 cm -2 @ 900 °C/15 min @ LN 2 NO EOR! {311}&loops Heating by I-beam - defects harder to be remove

28. © 2000 by Prentice Hall Upper Saddle River NJ Damage Annealing - “+1” Model Goals: Remove primary damage created by the implant and activate the dopants. Restore silicon lattice to its perfect crystalline state. Restore the electron and hole mobility. Do this without appreciable dopant redistribution. In regions where SPE does not take place (not amorphized), damage is removed by point defect recombination. Clusters of I recombine = dissolve @ the surface Bulk and surface recombination take place on a short time scale. "+1" I excess remains. These I coalesce into {311} defects which are stable for longer periods. {311} defects anneal out in sec to min at moderate temperatures (800 - 1000˚C) but eject I  TED. Primary defects start to anneal at 400 °C  all damage must be annealed with only +1 atom remaining. (+1 model) Frenkel pairs @ 900 °C, 5 sec  10 11 cm -2 of {311}; not long=10 nm rods  then dissolve if below critical size or else grow  dislocation loops (stable) = extrinsic e. i. Si I planes on {111}  = secondary defects. (difficult to remove) After 10 -2 s only “I” Fast

29. Solid State Epitaxy Regrowth from the C-Si acting as a seed (as in crystal growth from melt) @ 600 deg C, 50 nm/min 20 nm/min 2 nm/min 2.3 eV is for Si-Si bond breaking Dopants are active =substitutional position with very little diffusion. But high T might be needed for EOR annealing. Time Regrowth rate No defects=no diffusion enhancements Regrowth 10x larger for highly doped regions

30. © 2000 by Prentice Hall Upper Saddle River NJ Dopant Activation When the substrate is amorphous, SPE provides an ideal way of repairing the damage and activating dopants (except that EOR damage may remain). At lower implant doses, activation is much more complex because stable defects form.

31. Dopant Activation – No Premorphization Low T Annealing is enough for low doses – low primary damage can be easily annealed. High doses – damage below amorphization  secondary defects = difficult to anneal and requires high T  950-1050 ° C. 1.High initial activation, full activation is fast @ low T, 2.Low initial activation, traps anneal out, I compete with B for substitutional sites, I –B complexes 3.More damage so activation decreases with dose maintaining the same behavior. (1) (2) (3) Full activation Doses below amorphization High doses - high T required which causes more diffusion - in small devices unacceptable Amorphization improves activation @ low T leading to 100% @ high T Note: very high doses may result in low activation (25%) Secondary defects from Carriers’ mobility increases with damage anneal Increasing dose