Time-resolved diagnostics of pulsed plasmas etching processes 10/05/2000 Time-resolved diagnostics of pulsed plasmas etching processes G. Cunge Laboratoire des Technologies de la Microélectronique CNRS (France)

Plasma etching principles: example of Si etching in Cl2/O2: 10/05/2000 Si Mask SiOClx layer Cl SiCl4 Cl2/O2 plasma Cl2 + e-  SiO2 The Si surface is etched by Cl atoms  formation of SiCl4 volatile etch products (process strongly assisted by energetic ions bombardment) Plasma etching can be Anisotropic (vertical direction only) The ion bombardment is directional + Formation of passivation layers on the sidewalls Plasma etching can be Selective (Si is etched but SiO2 is not)

Time resolved diagnostics in reactive plasmas: challenges 10/05/2000 The time-resolution needed depends strongly on the application: 1) Real time monitoring of reactive plasmas 100 ms time resolution is good enough to monitor the variation of radical’s densities in real time. Challenge: highly sensitive diagnostic techniques are needed ! 2) Measurement of radical’s sticking coefficient  on surfaces (pulsed plasmas) We monitor the decay of radicals’ densities in the afterglow: kloss=kwall+kgas The radicals’ density must be measured with a time resolution < 100 µs Challenge: fast acquisition system synchronized with plasma pulses

 Fast acquisition systems synchronized with plasma pulses Time-resolved measurements in pulsed plasmas: basic principles and apparatus Plasma diagnostics provide a signal (typically a current). Goal: capture signal variations during one pulsing period (100 µs at 10 kHz) Modulation Plasma Potentiel V p RF power Detector + amplifier Trigger IT(t) Analog/Digital or Photon counter  Fast acquisition systems synchronized with plasma pulses CW signal (photodiode, Langmuir probes….etc) Pulses (Mass spec, PMT…etc) A/D converter (16 bit, 1 Ms/s) Multichannel analyzer

Experimental set-up 10/05/2000 10/05/2000 We are using an industrial ICP reactor from AMAT designed to etch 300 mm wafers but modified to host plasma and surface diagnostics LED for UVAS (Cl2, SiClX) Diode Laser (T) Mass Spectrometer (Cl, SiCl3) Plasma reactor (ICP) Reactor connected to XPS (by robotized vacuum transfer chamber)  chemical composition of the wafer surface + reactor wall coatings XPS 5

Detection of not optically accessible species by modulated beam threshold ionization Mass Spectrometry

Mass spectrometry: principle Plasma radicals Energy analyzer (0 - 150 eV) Mass Filter (0 - 500 amu) Ionization chamber (e- 0-70 eV) Chaneltron + counting system

Mass spectrometry: principle Plasma radicals Energy analyzer (0 - 150 eV) Mass Filter (0 - 500 amu) Ionization chamber (e- 0-70 eV) Issue: remove background Chaneltron + counting system

The issues associated with MS: beam and background to turbo pump (80 l.s-1) e- 5-70 eV Ar plasma to energy filter, mass filter and ion counting Ar Ar+ Ar atoms enter the MS head through an orifice and are ionized by an electron beam. Hypothesis: the Ar+ signal intensity is proportional to the Ar density in the plasma However, a large part of the signal originates from the residual pressure of Ar in the MS head  this background signal must be subtracted to capture the beam component  modulated beam MS

Introduction: Modulated Beam MS Why : to subtract the background signal MS ionization chamber plasma to energy filter to mass filter to ion counting to pump to pump H. Singh et.al., JVST A 17(5), 1999 + Measurement of beam+background and background alternatively Need of multiple pumping stages to reduce the background Hypothesis : the beam does not influence the background

Modulated beam Mass Spectrometry The tuning fork resonant (10Hz) chopper + differential pumping (Hiden) e-

Modulated beam Mass Spectrometry The tuning fork resonant (10Hz) chopper + differential pumping e-

Modulated beam Mass Spectrometry The tuning fork resonant (10Hz) chopper + differential pumping e- Chopper closed: beam blocked, measure background

Modulated beam Mass Spectrometry The tuning fork resonant (10Hz) chopper + differential pumping e- Chopper closed: beam blocked, measure background Chopper open: measure beam+background

Modulated beam Mass Spectrometry Example in O2 plasma (cw) finish to close finish to open start to open start to close O beam O backgrd 1 chopping period The tuning fork resonant (10Hz) chopper + differential pumping e- Chopper closed: beam blocked, measure background Chopper open: measure beam+background

Modulated beam Mass Spectrometry Example in O2 plasma (cw) finish to close finish to open start to open start to close O beam O backgrd O2 beam O2 backgrd The tuning fork resonant (10Hz) chopper + differential pumping e- The nature (reactive or stable species) of the beam influences strongly the background level.

Beam influence on the background ? (2/2) Background level vs. pressure finish to close start to close O2 beam O2 backgrd The background is mostly due to the beam, so it is crucial to estimate the background as soon as the beam is shut down. How can we measure those curves I(t) during the choping periode ?

Time resolved measurements with Hidden internal boxcar The issue of ghost pulses (improvement : the dead band time) Boxcar (Hiden Analytical) RF power Ions are detected by a channeltron  each ion is detected as a burst of e- current  the output signal is the number of burst averaged over a time window (dwell). chopper driver ghost pulse matching box channeltron MS

Time resolved measurements with Hidden internal boxcar The issue of ghost pulses (improvement : the dead band time) Dead Band Time Boxcar (Hiden Analytical) RF power chopper driver ghost pulse matching box channeltron MS Ghost pulses are corrected with the introduction of a dead band time.

Time resolved measurements with Hidden internal boxcar (Hiden Analytical) RF power chopper driver matching box channeltron MS For time-resolved measurements, (periodic) the system use a boxcar, which integrate the signal over a time-window (gate width). This signal is accumulated over several periods to improve S/R ratio (trigger is needed). The gate is then shifted to sample the signal at a later time in the cycle. time trigger Mesure a t Mesure a t+T Mesure a t+2T

Time resolved measurements with Hidden internal boxcar (Hiden Analytical) RF power chopper driver matching box channeltron MS For time-resolved measurements, (periodic) the system use a boxcar, which integrate the signal over a time-window (gate width). This signal is accumulated over several plasma pulses to improve S/R ratio (trigger is needed). The gate is then shifted to sample the signal at a later time in the cycle. time trigger Mesure a t Mesure a t+T Mesure a t+2T Issue: long measurement / huge loss of information !

pulse & delay generator Improvement : replace boxcar by the pulse counter time trigger Mesure a t pulse & delay generator Boxcar (Hiden Analytical) Pulse counter Pulse counter  signal/noise420 RF power chopper driver matching box Boxcar  signal/noise20 MS Thanks to the pulse counter, signal to noise ratio is strongly enhanced. Thus, fast synchronized acquisitions are achievable.

Modulated beam Mass Spectrometry Example in O2 plasma (cw) finish to close finish to open start to open start to close O beam O backgrd 1 chopping period The tuning fork resonant (10Hz) chopper + differential pumping e- With the multichannel pulse counter, we capture the entire signal during each chopping period  Excellent S/N ratio

Time-resolved measurements in pulsed plasma by MS Plasma pulses must be synchronized with chopper oscillations  plasma pulsing frequency > chopper frequency chopper plasma 30 Hz 10 Hz

Time-resolved measurements in pulsed plasma by MS Plasma pulses must be synchronized with chopper oscillations  plasma pulsing frequency > chopper frequency chopper plasma 30 Hz 10 Hz The chopper signal at 10 Hz is used to trigger a burst generator that produces pulses at N x 10 Hz to pulse the plasma

Cl atoms detection in pulsed discharge Plasma pulses must be synchronized with chopper oscillations  plasma pulsing frequency > chopper frequency chopper plasma The multichannel pulse counter capture the signal during the entire chopping period

Cl atoms detection in pulsed discharge Plasma pulses must be synchronized with chopper oscillations  plasma pulsing frequency > chopper frequency chopper plasma The multichannel pulse counter capture the signal during the entire chopping period Calibration on absolute density by comparison with Ar signal:

Summary What about the ions ? By combining diagnostics, we can monitor the densities of almost all the radicals even in complex chemistries (here SiCl4/Cl2) SiClX ClX What about the ions ?

Time resolved ion flux measurement in pulsed plasmas with a capacitively-coupled rf planar probe

Well known technique introduced by Braithwaite et al in 1996 Principle (in CW plasma) Well known technique introduced by Braithwaite et al in 1996 Principle: fed a planar probe with by RF bursts through a blocking capacitor Planar Probe (1 cm2) Blocking capacitor Pulsed rf Plasma (CW) -40 V Vprobe RF ON: capacitor charges (DC self bias about-40 V) RF OFF: e- can’t reach the probe polarized at -40V  Capacitor discharges by collecting ions Bohm flux Measurement of the capacitor discharge’s current in OFF period  ion flux

Current (ion flux) measurement in CW plasmas Potentiel V Plasma We use a direct current measurement system through a 1 k serial resistor (Booth et al, Rev. Sci. Instrum. 71, 2722) - RF ON: RF signal propagates through diodes - RF OFF: capacitor discharge’s current flow through resistor and is measured by the A/D (triggered by probe pulses) A/D Trig LP filter V I Blocking capacitor Ions only Ions + e- Pulsed RF generator

Current (ion flux) measurement in pulsed plasmas Issue: several plasma pulses are needed to charge blocking capacitor to - 40 V  Probe pulsing frequency < Plasma pulsing frequency (but synchronized) A/D Plasma Potentiel V p LP filter Freq. divider Trig  Use a frequency divider triggered by the plasma pulses to trigger probe pulses at f/10

Current (ion flux) measurement in pulsed plasmas Issue: several plasma pulses are needed to charge blocking capacitor to - 40 V  Probe pulsing frequency < Plasma pulsing frequency (but synchronized) Probe: 200 Hz Plasma: 2 kHz voltage (V) Probe RF Time variations of ion flux during one plasma pulse A/D Plasma Potentiel V p LP filter Freq. divider Trig

Typical pulsing frequency; 1-20 kHz Experimental set-up 10/05/2000 10/05/2000 We are using an industrial ICP reactor from AMAT designed to etch 300 mm wafers but modified to host plasma and surface diagnostics Mass Spectrometer (Cl, Br) Pulsed Plasma reactor (ICP): DPS LED for UVAS (Cl2) D2 lamp For VUVAS (HBr, Br2) System equipped with pulsed RF generators / Operation at constant peak power Typical pulsing frequency; 1-20 kHz 34

Results and discussion (1): Impact of the pulsing parameters on the radicals densities

Cl and Cl2 kinetics in Cl2 plasmas pulsed at low frequency 15 Hz Cl2 discharge Plasma ON: e- + Cl2  Cl + Cl + e- Plasma OFF (low Te): Cl + wall  Cl2 RF ON Cl Cl2 Wall

From low to high frequency pulsing 15 Hz Cl2 discharge Important conclusions: Reactive radicals (Cl) are : - Produced only during ON period - Lost continuously on the reactor walls Timescale for significant radical density variations: several milliseconds For pulsing frequencies > 1 kHz (T< 1 ms) the radical density is not anymore modulated during the pulse

From low to high frequency pulsing 15 Hz Cl2 discharge Important conclusions: Reactive radicals (Cl) are : - Produced only during ON period - Lost continuously on the reactor walls Timescale for significant radical density variations: several milliseconds 1 kHz Cl2 discharge For pulsing frequencies > 1 kHz (T< 1 ms) the radical density is not anymore modulated during the pulse RF OFF  Only time-averaged values will be considered in the following

Impact of the pulsing frequency and duty cycle on the Cl density Cl atoms density (Cl2 plasma / 20 mTorr / 800 W) The duty-cycle has a strong impact on the Cl atoms density The pulsing frequency has no effect ! Pulsed plasmas with small duty cycle = reduced chemical reactivity (fragmentation of feedstock gas by electron impact collisions is reduced)

When duty cycle increases  Longer ON period  higher Cl density Impact of the pulsing frequency and duty cycle on the parent gas fragmentation ~ 0.1 eV 3 eV Te RF ON RF OFF Radical loss Radical production Cl e- Cl2 Walls Plasma Potentiel V p RF power Cl atoms are produced only in the ON period but are lost continuously on the reactor walls When duty cycle increases  Longer ON period  higher Cl density

When duty cycle increases  Longer ON period  higher Cl density Impact of the pulsing frequency and duty cycle on the parent gas fragmentation RF ON RF OFF RF power Te Plasma 3 eV Plasma Potentiel V ~ 0.1 eV p Cl e- Cl2 Radical production Walls Cl Cl2 Walls Cl Cl2 Radical loss Radical loss Cl atoms are produced only in the ON period but are lost continuously on the reactor walls When duty cycle increases  Longer ON period  higher Cl density When frequency increases total ON time (averaged over many cycles) is unchanged  no effect on Cl density The ratio of atomic to molecular densities (Cl/Cl2) can be controlled by the duty cycle

Results and discussion (2): Impact of pulsing parameters on the ion flux

Impact of the duty cycle on the ion flux: (1) Electropositive plasma He plasma 10 mTorr / 700 W / 1kHz In the afterglow, ion flux decay is due to Te drop (UBohm ) and ambipolar losses (ni ) Rise and fall times of the ion flux  10 µs << pulsing period (1000 µs) In ON period the ion flux  reaches steady state and is high even at 10%DC

Impact of the duty cycle on the ion flux: (1) Electronegative plasma Cl2/SiCl4 plasma 10 mTorr / 700 W / 1kHz Rise time and decay time of the ion flux are longer than in He (attributed to negative ion formation in the afterglow)  In ON period, the ion flux strongly depends on the duty cycle and remains very small below 25 % DC Important conclusion for the design of pulsed etching process

Results and discussion (3): Impact of pulsing parameters on the IEDF

Without bias power, the ions are accelerated towards the wafer by Vp. Time-averaged IEDF in pulsed ICP plasmas (no bias) IEDF are recorded at wafer surface by a SEMION multigrid analyzer from IMPEDANS (see D.Gahan’s presentation) Ions from ON period Ions from OFF period RF ON RF OFF Te Vp 3 eV 15 V Te ~ 0.2 eV Vp ~ 1 V Without bias power, the ions are accelerated towards the wafer by Vp. In pulsed plasmas Vp is rapidly modulated between 15 V (ON period) and 1 V (OFF period)  Bimodal IEDF In a plasma pulsed with a small duty cycle the wafer is bombarded mostly by low energy ions  key point to reduce damages

IEDF in synchronous pulsed plasmas (with bias power) Cl2 plasma 1000 W / 50 Wbias CW or pulsed at 1 kHz Ions from ON period (ICP and bias) Bi-modal IEDF (corresponding to ions from ON and OFF periods) The ion energy in the ON period increases rapidly when the DC is decreased

IEDF in synchronous pulsed plasmas (with bias power) Cl2 plasma 1000 W / 50 Wbias CW or pulsed at 1 kHz Ions from ON period (ICP and bias) At 20 % DC the ions have 5 times the energy of those in the CW plasma for the same bias power ! Bi-modal IEDF (corresponding to ions from ON and OFF periods) The ion energy in the ON period increases rapidly when the DC is decreased

IEDF in synchronous pulsed plasmas (with bias power) Key point: Pbias (W)  Ii x Vbias When Ii drops at constant power, Vbias increases  Ion energy increases

IEDF in synchronous pulsed plasmas (with bias power) Key point: Pbias (W)  Ii x Vbias When Ii drops at constant power, Vbias increases  Ion energy increases Ion flux decreases Ion energy increases In synch. pulsed plasma at low duty cycle the wafer is bombarded by very energetic ions

Summary and conclusions RF power Plasma Impact of pulsing on plasma chemistry Duty cyle  Radical flux control  New domains of plasma chemistry Drawback: reduced polymerisation

Summary and conclusions Plasma RF power RF power (ion acceleration) Accessible cw IEDF Duty cyle  Radical flux control  New domains of plasma chemistry CW Plasmas: Restricted ion energy range

Summary and conclusions RF power Plasma Accessible cw IEDF Duty cyle  Radical flux control  New domains of plasma chemistry Pulsed plasma:  Access low energy  reduced damages

Summary and conclusions Plasma RF power RF power (ion acceleration) IEDF CW 1 kHz 20% nm Pulsed Accessible cw Duty cyle  Radical flux control  New domains of plasma chemistry Synchronised pulsing:  Access high energy  Improved profile control

Summary and conclusions Plasma RF power RF power (ion acceleration) IEDF CW 1 kHz 20% nm Pulsed Accessible cw By pulsing the plasma, the range of operation of ICPs can be broaden from Downstream etching conditions to CCP etching conditions

Impact of the pulsing frequency and duty cycle on the Cl2 density 20 mTorr Cl2 at 800 W The duty-cycle has a strong impact on the Cl2 density The pulsing frequency has no effect ! CW Cl2 residence time in the reactor (500 ms) >> pulsing period (< 1 ms) Each Cl2 molecule experiences many plasma pulses Cl2 dissociation fraction is driven by the total rf ON time during Cl2 lifetime This total ON time is fixed by the duty cycle and not by the pulsing frequency

Pulsed ICP plasmas: Timescales RF OFF RF ON Plasma Potentiel V p RF power The timescales for significant variations of ion flux and neutral flux are different: ne Ar plasma pulsed 10 kHz (Kushner et al) SiCl4/Cl2 plasma pulsed 100 Hz Gives hope to control independently ion and neutral fluxes

In ON period the ion flux is  independent of the pulsing frequency Impact of the duty cycle on the ion flux (1 kHz): (1) electropositive plasma He plasma 10 mTorr / 700 W Rise and fall times of the ion flux  10 µs << pulsing period (1000 µs) In ON period the ion flux  reaches steady state and is high even at 10%DC In ON period the ion flux is  independent of the pulsing frequency Note: in afterglow, ion flux decay rate is due to ambipolar losses (ni ) and Te decay (UBohm )

pulse & delay generator Time resolved measurements with Hidden internal boxcar pulse & delay generator Boxcar (Hiden Analytical) RF power chopper driver matching box Ions are detected by a chaneltron  each ion is detected as a burst of e- current  the output signal is the number of burst averaged over a time window (dwell). channeltron MS For time-resolved measurements, (periodic) the system use a boxcar, which integrate the signal over a time-window (gate width). This signal is accumulated over several plasma pulses to improve S/R ratio (trigger is needed). The gate is then shifted to sample the signal at a later time in the cycle. time trigger Mesure a t Mesure a t+T Mesure a t+2T