22nd International Symposium on Plasma Chemistry
July 5-10, 2015; Antwerp, Belgium
Reduction of plasma induced damage of porous low-k materials using a
cryogenic etching process
T. Tillocher1, F. Leroy1, L. Zhang2, P. Lefaucheux1, K. Yatsuda3, K. Maekawa4, J. F. de Marneffe2, M. Baklanov2 and
R. Dussart1
1
GREMI, CNRS/Université d’Orléans, Orléans, France
2
IMEC, Leuven, Belgium
3
Tokyo Electron, Tokyo, Japan
4
Tokyo Electron, Albany, NY, USA
Abstract: Cryogenic etching was applied to porous organosilicate glasses (OSG). Such
materials are important low-k candidates for advanced interconnects and their integration is
very challenging because of plasma induced damage. Two chemistries (SF 6 -based and
C 4 F 8 -based) have demonstrated a promising capability of low damage etching of this type
of material.
Keywords: low-k, damage, cryogenic etching, plasma, SF 6 , C 4 F 8
1. General
Porous Organosilicate Glasses (OSG) are low-k
materials of interest for advanced interconnects, where the
reduction of the RC delay and power consumption are
critical points to be addressed. However, due to a
significant porosity and the pore size, integration of such
materials is limited by Plasma Induced Damage (PID). It
has been demonstrated that porous OSG are very sensitive
to radicals, ions and photons. In addition, interconnected
pores ease diffusion of radicals, which further increase
PID. They contribute to carbon depletion and material
structure modification which eventually leads to the
increase of the k-value and leakage current. Several
solutions have been proposed to reduce PID: late porogen
removal approach [1], post-damage repair process [2] and
post-porosity plasma protection (P4)[3].
We propose in this paper a different approach, relying
on substrate cooling at cryogenic temperature. We show
that PID has been greatly reduced under such conditions
with both SF 6 -based and C 4 F 8 -based chemistries.
2. Experimental
An ICP etching tool with a diffusion chamber has been
used for all experiments presented in this paper. The
substrate holder, located at the base of the diffusion
chamber, is cooled with a liquid nitrogen circulation.
Temperature can be regulated between -150°C and 40°C.
Helium backside cooling enables good thermal contact
between the chuck and the wafer.
The material used in this work was OSG with k=2.2
spin-on deposited on (100) silicon wafers. Porosity is
37% and pore radius is 1.4 nm. Samples were 4x4 cm²
coupons glued on SiO 2 carrier wafer. Substrates were
introduced in the reactor via a loadlock. After etch,
substrates were kept in the loadlock under N 2 atmosphere
for warming-up without moisture uptake.
O-8-1
Thickness and refractive index evolution were
monitored in-process by in-situ ellipsometry. Post-plasma
damage was evaluated using an FTIR in transmission
mode. The concept of “Equivalent Damage Layer” (EDL)
was introduced to quantify the carbon depletion.
In parallel to etch experiments, desorption mass
spectrometry was performed to analyse desorbed species
during the warming-up of etched low-k surfaces. This
helps to figure out which mechanisms are involved in the
protection process of the material.
3. Cryogenic etching with SF 6 -based chemistry
This chemistry is directly inspired from the standard
cryogenic process used for silicon deep etching. It is
based on a SF 6 /O 2 plasma interacting with a silicon wafer
cooled down to a low temperature of typically -100°C. At
this very low temperature, a SiO x F y passivation layer
forms on the silicon sidewalls avoiding lateral etching [4].
A similar effect on low-k material should help to decrease
PID.
Etch tests were performed in pure SF 6 with or without
self-bias voltage (-120V) applied to the chuck. The first
case simulates what would happen at feature bottom while
the second represents the sidewalls. This helps to
anticipate the behaviour of processes on feature etching.
SF 6 flow was 50 sccm, pressure 3 Pa and source power
500W. Both etch rate and EDL are represented for 120°C, -50°C 20°C in Fig. 1.
These results clearly show a significant reduction of the
EDL at -120°C both with and without self-bias voltage.
The EDL is much higher at -50°C and 20°C than at 120°C, even without bias, which shows that only a very
low temperature provides a real benefit. The role of the
source power has been investigated between 200W and
1500W (not represented). Both etch rate and EDL
increase with source power. 500W power corresponds to
1
an optimum where the EDL is the lowest and the etch rate
is close to zero without any ion bombardment.
60
160
50
140
120
40
EDL (nm)
Etch rate (nm/min)
180
100
30
80
60
ER
20
40
10
20
0
ER (nm/min)
Bias -120C
Bias -50C
Bias +20C
74,8
120,6
155,9
8
53
44
EDL (nm)
0
OSG 2.2 with bias (~-120V)
60
160
50
140
120
40
EDL
100
EDL (nm)
Etch rate (nm/min)
180
30
80
60
20
40
10
20
0
WOBias -120C
WOBias -50C
WOBias +20C
ER (nm/min)
2,5
14,1
2,9
EDL (nm)
2,6
36
25
0
Fig. 2. Refractive index after etch and after annealing a
function of substrate temperature
After a post-etch annealing (350°C, under N 2 , 15 min),
the refractive index comes back to its initial value. This
clearly assists desorption of the trapped etch products,
without any additional damage, as confirmed with FTIR.
Desorption mass spectrometry was performed on 8x8
cm² OSG samples glued on a SiO 2 carrier wafer. The
low-k material was etched for 2 min with a similar
process than that described previously. Then, the
temperature was gradually increased from -120°C to
20°C. The desorbed species were detected by a mass
spectrometer mounted on the diffusion chamber. This
revealed in particular a significant desorption of SiF 4 at 60°C. In silicon cryoetching, this is related to passivation,
which suggests that a passivation layer may also be
present on the low-k material. Desorption of C x F y
molecules is also observed at -50°C, which shows that
fluorocarbon species may play a role in the protection
mechanism of the low-k material. Zhang et al. proposed
that etch by-products, like SiOF and alkyl alcohols
condense in the pores and prevent diffusion of free
radicals in the bulk material [5]. The observed desorption
of Si- and C-containing species can sustain such a
hypothesis.
4. Cryogenic etching with C 4 F 8 -based chemistry
As C x F y molecules desorb from porous OSG material
during warming-up, injection of fluorocarbon gas before
the etch process might help to further limit active radical
penetration.
At -120°C, C 4 F 8 gas effectively condenses in the pores
of the low-k layer, as evidenced from the increase of the
material refractive index by in-situ ellipsometry. In Fig. 3,
the refractive index of the layer is plotted as a function of
C 4 F 8 pressure.
OSG 2.2 without bias (~9V)
In-situ ellipsometry has revealed that the refractive
index of the OSG increases as the temperature decreases,
suggesting that part of etch by-products remains in the
film. (see Fig. 2).
1,36
1.35
1,32
1.29
RI
1,3
1,28
1,26
1,42
1,4
1,38
1,36
1,34
Absorption
Condensation
1,32
Desorption
Desorption
1,3
1,34
1.27
1.26
1.24
1.25
1.24
1.25
20 sccm C4F8, -120°C
1,44
Refractive Index
Fig. 1. OSG 2.2 etch rate and EDL as function of the
substrate temperature and self-bias voltage with SF 6
0
1
2
3
Pressure (Pa)
4
5
Fig. 3. Refractive index of porous OSG as a function of
C 4 F 8 gas pressure during condensation and desorption
1,24
1,22
1,2
1,18
20
-40
Temperature oC
2
-80
-120
The refractive index increases, during condensation,
from its initial value (1.32) to a maximum at 0.5 Pa (1.42)
and then, remains stable. A hysteresis is observed
between condensation and desorption: the refractive index
decreases for pressure below 0.5 Pa.
O-8-2
Once C 4 F 8 is fully condensed into the pores, OSG is
etched with SF 6 or a C 4 F 8 /SF 6 plasma mixture (total
pressure 3Pa, 500W source power, -135V self-bias
voltage, -120°C).
The resulting etch rate and EDL are plotted for both
etch chemistries and with or without self-bias voltage in
Fig.4.
EDL before annealing
EDL after
after annealing
annea
EDL before annealing
EDL after annealing
removed. Finally, there is almost no EDL with SF 6 /C 4 F 8
chemistry with or without self-bias voltage (EDL < 5
nm).
5. Conclusion
Plasma Induced Damage, which is a major issue in
etching of porous low-k materials, can be greatly reduced
with a cryogenic process. A passivation layer is believed
to form into the pores with SF 6 plasma at cryogenic
temperature of the substrate. This prevents carbon
depletion in the bulk material.
C 4 F 8 can be added before and during the plasma etch. It
condenses into the pores and prevents the diffusion of
radicals responsible for PID. This helps to further reduce
carbon depletion.
Post-etch annealing should be performed to fully desorb
etch by-products or condensates and recover surface state
close to pristine state.
Consequently, cryogenic etching can be considered as a
low damaging process of interest for porous low-k
materials.
6. References
[1] V. Jousseaume, L. Favennec, A. Zenasni and G.
Passemard, Appl. Phys. Lett., 88, 182908 (2006)
[2] Y. S. Mor, T. C. Chang, P.T. Liu, T. M. Tsai, C. W.
Chen, T. S. Yan, C. J. Chu, W. F. Wu, F. M. Pan, W. Lur
and S. M. Sze , J. Vac. Sci. Technol. B, 20(4), 1334
(2002)
[3] T. Frot, W. Volksen, S. Purushothaman, R. L. Bruce,
T. Magbitang, D. C. Miller, V. R. Deline and G. Dubois,
Adv. Funct. Mater., 22, 3043 (2012)
[4] R. Dussart, T. Tillocher, P. Lefaucheux and M.
Boufnichel, J. Phys. D, 47, 123001 (2014)
[5] L. Zhang, R. Ljazouli, P. Lefaucheux, T. Tillocher, R.
Dussart, Y. A. Mankelevich, J.-F. de Marneffe, S. de
Gendt and M. Baklanov, J. of Solid State Sc. And
Technol., 2(6), N131 (2013)
Fig. 4. OSG 2.2 etch rate and EDL versus substrate
temperature and self-bias voltage (-135V and 0V) with
C 4 F 8 condensation followed by an SF 6 /C 4 F 8 etch plasma
Both EDL and etch rate are lower with SF 6 /C 4 F 8
plasma whatever the self-bias voltage is. This shows the
importance of a sufficient C 4 F 8 partial pressure (above
0.5 Pa) during etching in order to keep the fluorocarbon
gas condensed into the material and hence observe the
protection effect. Consequently, with a C 4 F 8
condensation step prior to etch, the process becomes less
damaging.
After annealing, the EDL is greatly reduced in all cases
because a fluorocarbon layer, which deposits during
etching and is comprised in the evaluation of the EDL, is
O-8-1
3