PARTICLE CONTROL AT 100 nm NODE STATUS WORKSHOP: PARTICLES IN PLASMAS Mark J. Kushner University of Illinois Department of Electrical and Computer Engineering Urbana, IL 61801 mjk@uiuc.edu December 1998 SEM129803
AGENDA Introduction to particles in plasmas Particle transport, trapping and contamination in high plasma density tools Small particles are different than big particles Concluding remarks SEM129810 University of Illinois Optical and Discharge Physics
FORCES ON DUST PARTICLES IN PLASMAS AND TRAPPING Dust particle trapping in plasma tools occurs in regions where the major electrical and mechanical forces are nearly balanced. PLASMA POTENTIAL NOZZLE - - - Ion Drag Electrostatic ELECTROSTATIC THERMO- PHORETIC ELECTROSTATIC FLUID DRAG ION DRAG GRAVITY (PLASMA) HEATED ELECTRODE In RIE tools, trapping often occurs near the edge of the sheath where ion drag and electrostatic forces balance. SEM1298M11
LASER LIGHT SCATTERING OF TRAPPED PARTICLES IN AN RIE TOOL SEM981201A! Photograph provided by Dr. Gary Selwyn.
SCHEMATIC OF 3-D DUST TRANSPORT SIMULATION DTS-3D obtains plasma parameters from HPEM-3D. With those properties, the equations of motion of dust particles, subject to mechanical and electrical forces, are advanced. HPEM-3D σ(r,z,θ) E(r,z,θ) MAGNETO- STATICS MODULE CIRCUIT MODULE I,V ELECTRO- MAGNETICS MODULE E(r,z,θ) B(r,z,θ) ELECTRON ENERGY EQUATION / BOLTZMANN MODULE Te(r,z,θ), S(r,z,θ), µ(r,z,θ) ELECTRON TRANSPORT (CONTINUITY) ION TRANSPORT (CONTINUITY, MOMENTUM) NEUTRAL TRANSPORT (CONTINUITY, MOMENTUM) POISSON SOLUTION AMBIPOLAR TRANSPORT/ SHEATH SURFACE KINETICS σ(r,z,θ) N(r,z,θ) φ(r,z,θ) P(r,z,θ) DTS-3D THERMOPHORESIS GRAVITY E(r,z,θ), Φ(r,z,θ), N(r,z,θ) ELECTROSTATIC FLUID-DRAG ION-DRAG AVS97M17
PARTICLES IN HIGH PLASMA DENSITY, LOW PRESSURE TOOLS Plasma tools for the 100 nm node will most likely be: High plasma density (> 10 11 cm -3 ) Low gas pressure (< 10s mtorr) Low wafer bias (< 100 V) As a result.. Particle nucleation in the gas phase is likely not a major source of particles compared to particles from surfaces. Plasma forces which scale with the plasma density (ion drag) will be dominant (which happen to be expulsionary) Particles which "find" themselves in the plasma will likely be driven to surfaces. SEM129808 University of Illinois Optical and Discharge Physics
CONFINING AND EXPULSIONARY PLASMA FORCES Forces which retard particles from contaminating surfaces are largely produced by the sheath electric fields. In this ICP reactor, a negative bias on the substrate produces a larger retarding force to particles being accelerated through the sheaths. The off axis maximum in ion flux produces asymmetric expulsionary forces. SEM1298M12 Plasma Potential Ar/Cl2-70/30, 10 mtorr, 300W, 150 V, 150 sccm Ion Density
PARTICLE LOCATIONS AND CONTAMINATION SITES 0.045 um particles are randomly distributed in the reactor. Sheath electric fields and ion draf forces trap particles above or divert particles across the wafer. The off axis maximum in ion flux, producing a local maximum in the expulsionary force, pushes particles through the sheath. POSSIBLE PARTICLE PATH SEM1298M13 Particle Locations Ar/Cl2 = 70/30, 10 mtorr, 300W, 150 V, 150 sccm Contamination Sites
WAFER SHIELDING BIAS The wafer can be "shielded" from negatively charged by a substrate bias. Smaller particles typically require smaller shielding biases, so for otherwise constant conditions, smaller particles are less likely than intermediately sized particles to be driven to the wafer by plasma forces. 200 180 ON-WAFER SHIELDED 160 140 120 Ar/Cl 2 =70/30, 10 mtorr, 300 W, 150 sccm 100 80 SEM129802 60 0.01 0.1 1 RADIUS (µm) University of Illinois Optical and Discharge Physics
SCHEMATIC OF ICP REACTOR FOR DUST PARTICLE TRANSPORT: NOZZLES AND LOAD LOCK BAY Due to the large ion drag forces, it is not common to have particle traps in ICP reactors. Structures which generate large ambipolar fields and low ion fluxes can, however, produce conditions which form traps for dust particles. COIL NOZZLE WINDOW LOAD LOCK BAY FOCUS RING SUBSTRATE NOZZLE FOCUS RING SUBSTRATE LOAD LOCK BAY POSITION (cm) AVS97M01
ION DENSITY: NOZZLES RESULT IN LOCALLY REDUCED FLUXES Due to the ion sources being high in the reactor, ion fluxes to the substrate are shadowed by the nozzles, resulting in scalloping on the substrate. Protruding grounded metal structures, recombination surfaces, produce locally smaller ion fluxes. This reduces the expulsion force for particles. ABOVE NOZZLE BELOW WINDOW 20 0 20 POSITION Ion Density: Max = 3.8(11) cm-3 Ar, 10 mtorr, 600 W, 0 V bias AVS97M03
PLASMA POTENTIAL: NOZZLES RESULT IN ENHANCED AMBIPOLAR FIELDS The plasma potential peaks in the center of the reactor. Protruding grounded metal structures produce locally large potential gradients. This increases the trapping force. BELOW NOZZLE BELOW WINDOW 20 0 20 POSITION Plasma Potential: Max = 15.4 V Ar, 10 mtorr, 600 W, 0 V bias AVS97M02
SYMMETRICAL ICP REACTOR In a symmetrical ICP reactor of moderate power (and without a bias), the ion drag forces are sufficiently large to prevent significant particle trapping. Trapping may only occur in the peripheral corners of the tool. Top view of particle locations Cut-away view showing trapping in corners. AVS97M18 Ar, 10 mtorr, 200 W, 200 nm
PARTICLE LOCATIONS vs TIME 25 ms 75 ms 125 ms 175 ms With randomly distributed particles, initial losses to walls occur quickly (< 20 ms). Remaining particles migrate to traps with low continuing losses to walls and substrate. Ar, 500 nm, 100 W Top View AVS97M07-RZ9A3-6
LOW POWER PARTICLE TRAPPING IN ICP TOOL: 500 nm, 100 W Particle traps are formed by perturbations in plasma potential and ion flux in the vicinity of the nozzles. Particles not trapped slowly drift towards the pump port. Top view of particle positions AVS97M04-RZ9A1 Ar, 10 mtorr, 100 W
LOW POWER PARTICLE TRAPPING IN ICP TOOL: 500 nm, 100 W Particles are closely aligned with the nozzles (above and below) with a broader trap (opposite the pump port) between the nozzles. AVS97M05-RZ9A2 Ar, 10 mtorr, 100 W
PARTICLE TRAPS vs POWER 100 W 200 W 500 W Due to the dominance of ion drag forces, geometrically induced trapping diminishes with increasing power. These effects are therefore most important during startup/shutdown cycles. Ar, 500 nm, 10 mtorr, Top View AVS97M08-RZ9A1,9G1,9H1
PARTICLE TRAPS vs PARTICLE SIZE 500 nm 200 nm 100 nm Trapping of small particles is more isolated. Small particles respond more (on a relative basis) to perturbations in plasma potential than ion flux, and so are more closely geometrically associated with the nozzles. Ar, 100 W, 10 mtorr, Top View AVS97M10-RZ9A1,9E1,9F1
TRAPS FOR 200 nm PARTICLES Smaller particles respond (on a relative basis) more quickly to perturbations in electrostatic potential. Traps therefore are more closely aligned with the nozzles. AVS97M11-RZ9E2 Ar, 100 W, 10 mtorr, 200 nm
SOLENOID ICP REACTOR The demonstration solenoid ICP reactor has 3 turns and asymmetric pumping. The coil feeds are on the same side of the reactor. 18.5 FEED SHOWERHEAD NOZZLE QUARTZ COIL FEED PUMP PORT FOCUS RING SUBSTRATE WAFER 0 20 10 0 10 20 RADIUS (cm) FEED COIL FOCUS RING QUARTZ WAFER / SUBSTRATE FEED PUMP PORT -20 0 20 POSITION (cm) -20 0 20 POSITION (cm) AVS97M12
POWER DEPOSITION IN 3-TURN SOLENOID REACTOR Transmission-line effects resulting from poor impedance matching of the coils can produce azimuthally varying power deposition. Power deposition: Max = 0.92 W-cm-3 (2 decades) 13.56 MHz, 100 nf termination, Cl2, 10 mtorr, 400 W AVS97M13-ICP_3D15B_POW2,POW.HDF023,26,29
PARTICLE TRAPS IN 3-TURN SOLENOID REACTOR Oblique view Transmission line effects which produce asymmetric power deposition may ultimately produce particle traps. Trapping occurs in the lower power volumes of the tool. This is particularly problemmatic for remote plasma devices where there are large gradients in plasma density. Blue: positions 0.075 s after release Red: traps 0.1 s after release. Side View 13.56 MHz, 100 nf termination, Cl2, 10 mtorr, 200 W, 500 nm AVS97M15-ICP_3D15B_DOT,DOT1
HDP TOOL: PARTICLES IN GAS FLOW In high plasma density tools, particles are likely not generated in the gas phase, and therefore are introduced in the plasma by other means. A particle entrained in the input gas flow may have sufficient kinetic energy to overcome ion-drag forces pushing the other way. PLASMA NOZZLE Particles introduced into the plasma in this manner will have unique trajectories depending on initial speed and direction. Where particles go depend on where they started from. ION DRAG KINETIC ENERGY OF 1 um PARTICLE = 0.4 v2(cm/s) ev FOR GAS FLOW OF 300 cm/s AT NOZZLE SEM1298M01 KE = 34 kev
HIGH PLASMA DENSITY GENERIC ICP The injection of particles through nozzles will be investigated using a generic inductively-coupled-plasma (ICP) reactor. The injection speed will be varied to determine the propensity for contaminating surfaces. 12 COILS DIELECTRIC WINDOW 8 NOZZLES 4 PUMP FOCUS RING 0 PORT POWERED SUBSTRATE 17 8.5 0 8.5 17 RADIUS (cm) SEM1298m02
Ar+ Density GENERIC ICP: TYPICAL PLASMA PROPERTIES As a worst case scenario, low ICP power without a biased substrate will be simulated. Plasma Potential Ion densities, here 2 x 1010 cm-3, could be an order of magnitude higher in some processes. Ar, 15 mtorr, 300 W, 80 sccm SEM1298M03
HDP TOOL: PARTICLES ENTRAINED IN GAS FLOW 200 cm/s Particles are entrained in the inlet gas flow and enter with a speed of 200 cm/s. The particles have insufficient momenta to overcome ion drag forces, and deposit on surfaces near the nozzles. Particle trajectories/locations Particle contamination sites SEM1298M04 0.75 um (radius) Argon, 15 mtorr, 300 W
HDP TOOL: PARTICLES ENTRAINED IN GAS FLOW 500 cm/s Particles are entrained in the inlet gas flow and enter with a speed of 500 cm/s. The particles deposit on surfaces further from the nozzle, but are still confined to the top of the reactor. Particle trajectories/locations Particle contamination sites SEM1298M05 0.75 um (radius) Argon, 15 mtorr, 300 W
HDP TOOL: PARTICLES ENTRAINED IN GAS FLOW 1000 cm/s Particles are entrained in the inlet gas flow and enter with a speed of 1000 cm/s. The particles have sufficient initial momenta to access the entire reactor. Here, particles deposit on the wafer. With an rf bias, the sheath voltage would likely be sufficient to shield the wafer from contamination. Particle trajectories/locations Particle contamination sites SEM1298M06 0.75 um (radius) Argon, 15 mtorr, 300 W
HDP TOOL: RANDOMLY GENERATED PARTICLES If particles are generated in the gas phase ( randomly appearing), they can access different locations. Those generated deep in the reactor obtain sufficient inertia from plasma forces to skip over traps and collect on surfaces. A particle which grows (or appears ) in a trap lacks the inertia to escape and will remain in the trap. Particle trajectories/locations Particle contamination sites SEM1298M07 0.75 um (radius) Argon, 15 mtorr, 300 W
DESIGN CONSIDERATION FOR HDP TOOLS HDP tools in which the substrate is remote from the plasma source, or there are nooks and crannies, have regions with low ion densities. HDP tools having these conditions are, from particle transport considerations, little different from low plasma density tools. Since expulsionary forces scale with plasma density (and trapping forces do not) the propensity to form traps increases. SEM1298M08 Large volume HDP with side-arm Ar, 15 mtorr
PARTICLE LOCATIONS IN ICP TOOL WITH LOW PLASMA DENSITY REGION ICP tools which have low plasma density regions behave similarly to RIE tools. These regions, where ion drag forces are low, have trapping sites. If these regions are also where gas flow is low, they may be nurseries for particle growth. SEM1298M09
SMALL PARTICLES IN ADVANCED PLASMA TOOLS.. The NTRS cites critical defect sizes as 50 nm (0.05 µm) at the 100 nm node. The "story to date" sounds encouraging with respect circumventing contamination by small particles in plasma tools. The "gotcha" is. Plasma tools at the 100 nm node will necessarily operate at low electron temperatures to prevent charging damage of gate oxide (< 2 nm) and to prevent notching. There are always more particles of smaller sizes than larger sizes. Smaller T e and many more smaller particles implies less charge per particle. Ref: T. Matsoukas, JAP 77, 4287 (1995) SEM129804 University of Illinois Optical and Discharge Physics
CHARGE FLUCTUATIONS ON SMALL PARTICLES When sufficiently small (< 10 nm) the statistical fluctuations of charges on particles can result in the particles either having no charge or actually being charged positively! SEM129805 Ref: T. Matsoukas, JAP 77, 4287 (1998) The end result is that particles which had previously been shielded from the wafer are now accelerated into the wafer! University of Illinois Optical and Discharge Physics
TRANSIENT PARTICLE CHARGING As particles transport through the reactor, their charge statistically fluctuates in time, resulting (if small enough) in a distribution of positive, negative and neutral particles. Ref: Choi and Kushner, Trans. Plasma Sci. 22, 138 (1994) As a result, the smaller neutral and positively charged particles "rain" down on the wafer. SEM129806 University of Illinois Optical and Discharge Physics
PARTICLE GROWTH MODEL Particles growth starts in by nucleation between neutral or neutral-charged clusters (0.1 - many nm). Larger clusters begin to charge and take on characteristics of "particles", and grow by charged cluster-neutral radical reacations. Radicals Clusters Netural or Singly Charged Clusters Small Clusters (0.1 - many nm) + - Growth by Neutral radicals or positive ions Negatively Charged Particles Large Clusters (many -10s nm) - - - - - - - - - + Rapid growth requires: Large radical densities (high powers), long residence time. SEM1298M14
PARTICLES TO THE WAFER IN A PECVD REACTOR In a PECVD reactor where nucleation rates are large, the flux of small neutral and positively charged clusters can be large. Ref: H. Hwang, PhD Thesis, 1996 Ar/SiH 4, 100 mtorr, 42 W, 0-150 sccm, 6 s nucleation time Large charged clusters are trapped and do not reach the wafer SEM129807 University of Illinois Optical and Discharge Physics
CONCLUDING REMARKS The "small particle" regime in plasma tools has largely not been theoretically, computationally or experimentally investigated. The major differences between the large and small particle regimes are: Charge distribution of particles and charge fluctuations. Growth mechanisms are much more sensitive to chemistry. Current CFM remediation techniques may not apply since they rely on "large particle" effects (fluid drag, wafer shield, thermophoresis) Sematech and SRC have abandoned (or have announced dates for abandonment ) of support for investigations into the fundamental of particles in plasmas and their modeling..no investment, no progress. SEM129809 University of Illinois Optical and Discharge Physics
MODELING OF DUST PARTICLE GENERATION AND TRANSPORT IN PLASMA TOOLS Simulation of dust particle transport is at the top of a modeling pyramid whose base consists of a comprehensive plasma equipment model. DUST PARTICLE TRANSPORT MICROSCOPIC TRANSPORT PARAMETERS: DUST CHARGING, ION-DUST CROSS SECTIONS, MICROSCOPIC GROWTH PARAMETERS: STICKING COEFFICIENTS NUCLEATION / SPUTTERING RATES, PLASMA EQUIPMENT MODEL: PLASMA POTENTIAL (ELECTROSTATIC FORCES), ION FLUXES (ION DRAG FORCES), ADVECTIVE FLOW FIELD (FLUID DRAG FORCES), TEMPERATURE FIELD (THERMOPHORETIC FORCES) SEM1298M10