Control of Ion Energy in a Capacitively Coupled
Reactive Ion Etcher
H. M. Park , C. Garvin, D. S. Grimard and J. W. Grizzle
Electronics Manufacturing and Control Systems Lab oratory,
Dept. of Electrical Engineering and Computer Science, UniversityofMichigan,
Ann Arb or, MI 48109-212 2, USA
ABSTRACT
The energy of ions b ombarding the wafer is prop ortional to the p o-
tential di erence b etween the plasma and the p owered electro de in
Reactive Ion Etching (RIE) systems. This work seeks to control the
ion energy without altering the applied RF p ower or the chamb er pres-
sure since these variables are closely tied to other imp ortantquanti-
ties, such as reactivechemical sp ecies concentrations in the plasma and
wafer etch uniformity.Avariable resistor placed in parallel with the
blo cking capacitor allows the plasma self bias voltage (V )tobear-
bias
bitrarily varied b etween its nominal value and zero. Optical emission
sp ectroscopy for a CF plasma reveals that the nominal plasma chem-
4
ical concentrations do not change under this control metho d. The use
of a Langmuir prob e to measure the plasma p otential shows that the
ion energy changes by approximately one half of the change in V .
bias
The p otential uses of this ion energy control technique to plasma self
bias voltage regulation, etch selectivity and plasma cleaning of cham-
ber walls are demonstrated. A p otential drawback, namely decreased
plasma stability, is also indicated.
Author to whom corresp ondence should b e addressed. email: [email protected] 1
1 Intro duction and Background
Reactive Ion Etching (RIE) is the dominant etching pro cess for the transfer
of ne features from masks to wafers. It is well known that ion b ombardment plays
an imp ortant role in the creation of anisotropic pro les, etch rate enhancement and
surface damage [1{3] and hence signi cant research has b een dedicated to the control
of the ion b ombardment energy in RIE systems [4{8]. In standard capacitively
coupled RIE systems, the conventional approach to ion energy control has b een
through chamb er pressure or applied RF p ower. The drawbacks are that changing
power also alters the chemical concentrations in the plasma and changing pressure
will a ect the uniformity of the etching pro cess across the wafer [9, 10]. In part,
to overcome these drawbacks, new typ es of plasma sources have b een develop ed
so that ion energy can b e controlled indep endently of the chemical concentration
and chamb er pressure. These include Electron Cyclotron Resonance (ECR), helicon,
helical resonator, Transformer Coupled Plasma (TCP), and Inductive Plasma Source
(IPS), which op erate in low pressure and high plasma density range [11{14].
This pap er is exclusively concerned with ion energy control in classical capac-
itively coupled RIE systems. Figure 1 is a schematic of a typical parallel plate RIE
system along with a corresp onding electrical p otential from ano de to catho de. In an
asymmetric RIE system (smaller p owered electro de area than grounded electro de
area), the presence of a blo cking capacitor leads to a large negative plasma self bias
voltage (V ) b eing induced on the p owered electro de. The time-averaged plasma
bias
p otential (V ) can b e estimated as [15]
p
V + V
rf bias
V = ; (1)
p
2 2
where V is the p eak amplitude of the applied voltage of the RF generator p ower.
rf
Ions b ombard the wafer surface due to the p otential di erence b etween the plasma
and the p owered electro de, and this p otential di erence de nes the average ion
energy (E ) in a collisionless sheath via
ion
E = q (V V ); (2)
ion p bias
where q is the charge of an ion [16]. The plasma p otential (V ) is an equip otential
p
across the bulk plasma. In comparison to V , V is rather small in magnitude, and
bias p
therefore, V is traditionally used to represent ion energy.
bias
Several di erenttechniques can b e found in the literature for the control of ion
energy to increase etch rate [4, 5], or selectivity [6, 7], to pro duce a p ositively tap ered
etch pro le [8] or to minimize radiation damage [17]. The most common approach
for the control of ion energy has b een DC-biasing of the p owered electro de. Nagy [4]
increased the ion energy by adding a negative DC bias to the p owered electro de,
resulting in increased etch rate of p olysilicon and SiO . In his exp eriment, the
2
plasma p otential was slightly decreased with the addition of the negative DC bias;
however, the ion energy, V V , increased prop ortionally to the applied DC bias
p bias
voltage after the applied DC voltage exceeds a few tens of volts. Tai et al. used a
zener dio de tied to the blo cking capacitor to control the ion energy [6]. Tai achieved
a 50% selectivity improvementofSiO to HPR-206 photoresist by increasing (i.e.
2
decreasing its magnitude) of the V from -470 V to -250 V in CHF plasma.
bias 3
The metho d used in this pap er to control ion b ombardment energy is related
to [6]. As shown in Fig. 1, in a conventional RIE system, the net Direct Current
(DC) through the p owered electro de must b e zero due to the DC blo cking capacitor. 3
However, if a variable resistor is connected to the p owered electro de as shown in
Fig. 2, then the net DC is no longer forced to b e zero. Weobserved that the V
bias
develop ed at the p owered electro de increases (i.e. decreases in magnitude) as the
DC settles to a nonzero value. The result is that the V can b e controlled by
bias
regulating the amountofDCandthus can b e manipulated byavariable resistor.
From the de nition of ion energy (2), an increase of V by the use of a variable
bias
resistor will decrease the ion energy if the V is not varied during the control of the
p
variable resistor.
To b etter understand the e ects of this variable resistor on ion energy,itis
helpful to analyze two extreme cases illustrated in Fig. 3: a) If the variable DC
resistance is set to 1, then the system reduces to a conventional RIE system with
nom nom nom
V = V and E = q (V V ); b) If the DC resistance is zero, then the
bias ion
bias p bias
system turns into a DC coupled RIE system with V = 0 and E = q V . In the
bias ion p
latter case, the higher mobility of the electrons leads to electron current dominating
the DC through the electro des. The higher eux of electrons than ions from the
plasma will lead to an increase in the plasma p otential. This is illustrated in Fig. 3,
where the plasma p otential in a DC coupled RIE system is much higher than in a
capacitively coupled system, and its lowest value do es not approach zero [15, 18].
Therefore, the increase in the plasma p otential acts as a limiting factor in ion energy
control by this metho d. The amountofchange in ion energy that can b e obtained
by this metho d is investigated in Section 4.1.
Two applications of ion energy control by the use of a variable resistor are
explored in this pap er. In the rst one, the selectivity of p olysilicon with resp ect to
SiO is investigated. It is known that the etch rate of p olysilicon dep ends more on
2
the concentration of the reactivechemical sp ecies (chemical etching) than the ion 4
energy (physical etching); in contrast to this, the etch rate of SiO dep ends more
2
strongly on the ion energy than the chemical reaction [16, 19, 20]. Therefore, in order
to increase the selectivity of p olysilicon to SiO , the ion energy should b e decreased
2
while the reacting chemical concentration is either increased or kept constant. This
work is presented in Section 4.2. Section 4.3 investigates an application of the ion
energy control technique to an ion enhanced cleaning pro cess. Polymer buildup
on the etching chamber wall is regarded as a signi cant disturbance to the etching
pro cess and has b een correlated to reduced device p erformance and yield [21{23]. By
increasing the plasma p otential, the ion b ombardment energy toward the grounded
wall will b e increased, p ossibly enhancing a plasma cleaning pro cess.
2 Exp erimental Setup
The exp eriment to measure the plasma p otential with a Langmuir prob e and
the selectivity enhancement exp erimentwere p erformed on a Gaseous Electronics
Conference (GEC) reference cell that is describ ed in detail in [24, 25]. The schematic
of the exp erimental apparatus is shown in Fig. 4. The GEC is an RIE system having
parallel, four-inch diameter, one-inch spaced, water-co oled stainless steel electro des,
housed in a stainless steel chamb er. Vacuum is established bya mechanical pump
and a turb o pump connected to the b ottom of the chamb er and pressure is controlled
through a butter y throttle valve.
Pressure is measured byanMKStyp e 127A baratron capacitive manometer.
RF p ower at 13.56 MHz is supplied by an ENI ACG-5 p ower supply to the b ottom
electro de through an ENI MW-5 matching network. A Werlatone C1373 directional 5
coupler placed b etween the matching network and the p owered electro de is used to
+
measure forward (V ) and reverse voltage (V ) from which the p ower dep osited
into the plasma can b e determined as
+ 2 2
jV j jV j
P = P P = (3)
deposited f or w ar d reverse
Z
where P is the actual p ower dep osited in the RIE system, P is the
deposited f or w ar d
power supplied to the RIE system bythepower supply, P is the re ected
reverse
power from the RIE system due to the imp edance mismatch, and Z is a constant
characteristic imp edance, which is 50 in our exp eriment [26]. The upp er electro de
and chamber wall are grounded. The ow rate of plasma feed gas is controlled by
an MKS mass owcontroller and distributed through a showerhead arrangementin
the upp er electro de.
In order to mitigate any RF uctuation when measuring the plasma p otential
relative to ground with a Langmuir prob e, a tuned prob e was constructed so that
the imp edance of the plasma to prob e sheath is small in comparison to the prob e to
ground imp edance at 13.56 MHz. The prob e to ground imp edance was increased by
a pair of inductors which are self-resonant at 13.56 MHz along with the prob e's par-
asitic capacitance. The cylindrical prob e tip is made of tungsten and is placed at the
midp ointbetween the two electro des. The prob e is controlled by an automated data
acquisition system which is connected to a computer. A more detailed description
can b e found in [27{29]. The selectivity exp erimentwas p erformed in a CF plasma
4
with 4% pre-mixed O on unpatterned quarter slices of four-inch p olysilicon and
2
SiO wafers. The wafers slices were placed on the p owered electro de and etched at
2
the same time. Film thicknesses b efore and after etching were determined by o -line
ellipsometry measurements. The etchratewas calculated by dividing the thickness 6
of the lm removed by the etch time. A National Instrument's Data Acquisition
(DAQ) card and LabVIEW software were used to acquire data and control the gen-
erator p ower, gas ow rate, pressure and the variable resistor at a 4 Hz sampling
frequency.
The uorine concentration measurement and ion enhanced cleaning exp eri-
ments were p erformed on an Applied Materials 8300 (AME-8300) RIE system, de-
scrib ed in detail in [30, 31]. The AME-8300 is a hexo de typ e RIE system, where
wafers are placed vertically on a hexagonal p owered electro de and an aluminum b ell
jar acts as the grounded electro de. The AME-8300 is equipp ed with an Optical
Emission Sp ectroscopy (OES) system to monitor the intensity of the uorine 703.7
nm emission line (I ) and the intensity of the argon 750.4 nm emission line (I ).
F Ar
These emissions are used to estimate uorine concentration as [F ]= k (I =I ) P ,
F Ar
where k is a prop ortionality constantand P is pressure. This RIE has also b een
equipp ed with a sp ectral re ectometry sensor system for determining a real-time
lm thickness and etchrate. Both of these systems are describ ed in more detail
in [32].
3 Variable Resistor Implementation and
Control
The variable resistor illustrated in Fig. 5 was implementedby an enhancement
mo de p-MOSFET (breakdown voltage: 500 V). In order to minimize RF p ower
dissipation through the variable resistor and to minimize RF \propagation noise,"
an L-C lter was employed. The lter provides 30 dB attenuation at 13.56 MHz. 7
Typically, the magnitude of V is larger than 100 V, and thus, the enhancement
bias
mo de p-MOSFET is op erating in its saturation region. Therefore, the drain-to-
source resistance (R ) of the MOSFET is
DS
V V
DS DS
R = = ; (4)
DS
2
I k (V V )
DS GS T
where V is the drain to source voltage, I is the drain to source current, k is
DS DS
a device dep endent constant, V is the gate bias voltage, and V is the threshold
GS T
voltage. The total resistance of the variable resistor (R )is
VAR
(R + R ) [(R + R )(R + R )+R R ]
2 3 1 DS 2 3 1 DS
R = : (5)
VAR
(R + R )(R + R + R +2 R )+R R
2 3 1 2 3 DS 1 DS
With the assumption of R (R + R )and R (R + R ), (5) can b e approx-
DS 2 3 1 2 3
imated as
V
bias
R ' R + R ' : (6)
VAR 1 DS
2
k (V V )
GS T
Therefore, when V is ab ove the saturation voltage, the variable resistance can
DS
b e controlled by the gate bias voltage (V ). In this study, R =1 K , R =40 M
GS 1 2
and R =1 M were used, whichyielda variable resistance range of 1 K { 20.5 M
3
for R . V can b e measured through the resistivevoltage divider, R and R ,
VAR bias 2 3
which has a ratio 40:1.
Atwo-input two-output, decentralized controller was designed to regulate V
bias
and the dep osited p ower in the plasma to desired, constant setp oints with zero
steady state error. As describ ed in the exp erimental setup, the dep osited p ower was 8
determined from a directional coupler placed near the p owered electro de and V
bias
was measured by the voltage divider of the variable resistor circuit describ ed ab ove.
On the basis of these measurements, the controller adjusts the gate bias voltage of
the p-MOSFET (i.e. regulating the variable resistor) and the RF generator p ower
in order to control V and dep osited p ower to desired levels. Figure 6 shows an
bias
actual run of the controller on the GEC, where the desired setp oints for V and
bias
dep osited p ower were varied one at a time. The exp erimentwas p erformed on the
GEC at 30 sccm CF with 4% pre-mixed O and the pressure was held constantat
4 2
4.0 Pa with the standard MKS pressure controller. Due to the pro cess interaction
between dep osited p ower and V , it follows that an increase in dep osited p ower
bias
causes V to decrease (i.e. increase in the magnitude of V ). The gate bias
bias bias
voltage regulating V reacts to comp ensate for the V error by decreasing the
bias bias
resistance. The other interaction, an increase of V changing the dep osited p ower,
bias
is seen to b e less signi cant. This controller was used in all of the exp eriments
rep orted in the next section.
4 Exp erimental Results
4.1 V and Ion Energy vs. Plasma Potential
bias
In order to determine the relativevalue of the plasma p otential as a function of
the controlled V , Langmuir prob e measurements were p erformed on the GEC.
bias
Figure 7 is a plot of measured plasma p otential as a function of V under several
bias
op erating conditions. As describ ed in the published literature, the plasma p otential
increases slightly with increased dep osited p ower under xed pressure due to the 9
increased electron temp erature [18] and as illustrated by Fig. 7 the plasma p otential
shows a relatively large dep endence on the pressure under xed dep osited p ower. As
previously discussed in Section 1, the plasma p otential is increased with an increase
nom
)underV of V .Thechange in plasma p otential from its nominal value (V
bias bias
p
control is seen in Fig. 7 to b e a nearly linear function of the variation in V from its
bias
nom
nominal value (V ). From (1) and Fig. 7, the plasma p otential under ion energy
bias
control is mo di ed from (1) to
nom nom
V + V
bias rf
V + m(P ) V ; (7)
p bias
2
where,
V plasma p otential
p
nom
V nominal V without ion energy control
bias
bias
nom
V nominal V without ion energy control
rf
rf
nom
V V V
bias bias
bias
m(P ) pressure dep endent slop e
From the Fig. 7, the slop e m(P ) is calculated as 0.5 for 33.3 Paand 0.6 for 4.0
Pa. In this pap er, we simply express the plasma p otential under ion energy control
as (8) for simplicity and ease of manipulation.
nom nom
V + V
V
bias
bias rf
+ ; (8) V
p
2 2
Figure 8 shows the comparison of two mo dels of the plasma p otential: mo del I
based on (1); and mo del I I based on (8). Since the plasma p otential is increased
by approximately one half of the V change, it follows that the ion energy (E ),
bias ion 10
which is prop ortional to the p otential di erence b etween the plasma and the p owered
electro de, is changed by ab out one half of the V change.
bias