22nd International Symposium on Plasma Chemistry
July 5-10, 2015; Antwerp, Belgium
Atmospheric pressure plasma enhanced spatial atomic layer deposition
Y. Creyghton, F. van den Bruele, A. Illiberi, F. Roozeboom and P. Poodt
Solliance/TNO, Eindhoven, The Netherlands
Abstract: A SDBD plasma source has been integrated in a spatial ALD process. Plasma
process settings have been established for homogeneous growth of ZnO and SiO 2 . For
conductive/patterned substrates, an alternative source geometry is proposed. The
homogeneity and reach of oxidative reactive plasma species for both sources have been
evaluated using amorphous carbon oxidation under static treatment conditions.
Keywords: SDBD, spatial ALD, ZnO, SiO 2 , amorphous carbon
1. The PE-S-ALD process
Atomic Layer Deposition is based on sequential and
self-limiting surface reactions between substrates and
gaseous precursors. Each deposition cycle consists of two
sub-monolayer deposition steps, one after each other,
where during each step the substrate is exposed to a
reactive precursor or a co-reactant. In plasma enhanced
ALD the co-reactant consists of plasma reactive species
such as electronic excited states, radicals and ions.
Spatial Atomic Layer Deposition (S-ALD) is an
emerging ALD process technology which allows to apply
ALD in a continuous mode and substantially reduce the
total time of a cycle. In S-ALD substrates are passing
along a series of spatially separated gas injectors [1].
Compared to time-sequenced ALD, a very large gain in
deposition rate (up to 100x) can be achieved because
time-consuming sequential injection and pumping of
gases are avoided, see Figure 1.
Fig. 1. Schematic of PE-S-ALD process for metal oxides.
S-ALD can be used for high throughput processing of
polymer foils. However, for various types of polymers the
normal operating temperature of thermal ALD is too high
and plasma enhanced ALD may be used to extend
applicability to lower temperatures. Apart from
broadening the ALD temperature window, PE-S-ALD
also widens the pallet of materials that can be deposited.
Metal-organic precursors for ALD of silica, titaniumoxides, titanium-nitrides, and pure metals require a radical
type co-reactant [2].
O-20-7
The laboratory S-ALD set-up at TNO uses a rotating
substrate holder, placed below a fixed reactor head, which
contains all the openings for N 2 purging, precursor
injection and gas exhausts (Fig. 2). N 2 gas bearing is used
to maintain a distance of ~20 µm between the substrate
and the main surface of the reactor head. The head has
been equipped with a Surface Dielectric Barrier
Discharge (SDBD) source. The SDBD plasma has a
mixed plasma structure with a high number density of
micro-discharges and glow discharges. Theoretically the
thickness of the plasma region is not more than 20 µm
[3]. The head-substrate distance in both the precursor
injection and plasma zone is at ~200 µm as standard
value. The height of plasma source relative to the
substrate has been increased using insert plates.
Fig. 2. Top view of the (fixed) reactor head with
precursor injector (45x10 mm) and plasma source (30x5
mm).
Standard electrical operating condition of the SDBD
plasma used in this contribution are: repetitive and
alternating unipolar pulsed voltage with amplitude V p =5
kV, pulse width = 5 µs, pulse repetition rate = 58 kHz,
power dissipation ≈20 W (for 30x5 mm2). The rotating
ALD reactor is placed in an oven which is kept at
constant temperature. In the electrode configuration
shown in Fig.1, reactive plasma species transport to the
substrate is mainly diffusional while the gas flow through
1
the plasma zone depends on the drag force of the moving
substrate. Though this situation does not lead to optimum
control of the plasma energy density (plasma power / gas
flow), positive results for many material systems led us
suspend replacement of the plasma source by alternative
configurations. Source replacement has been planned, e.g.
for improved control of gas flow and plasma induced
heating. The exposure time of the precursor is given as
τ=w/(2πfr) where w is the width of the precursor injection
zone and f the rotation frequency. In the mid-region, r=37
mm, w=10 mm, f=1 Hz, τ=43 ms. The exposure times of
the precursor injection and plasma treatment cannot be
varied independently.
2. ZnO deposition
ZnO deposition on Si using Diethylzinc (DEZ)
precursor has been used as a test case to determine the
ALD process conditions for homogeneous deposition
(Figs. 3 and 4). The growth per cycle (GPC) as a function
of exposure time does not reach a limit.
Nevertheless, saturation of GPC is practically obtained
when dosing the DEZ precursor flow. Increasing the 10%
O 2 /N 2 flow, saturation is reached under fairly low flow
conditions of 10 sccm. In order to check to potential
influence of the plasma energy density we have changed
the pulsed voltage amplitude and the pulse repetition rate.
Both electrical parameters can be varied independently. A
minimum voltage for reproducible and homogeneous
deposition appears to be ~3.5 kV. However, the GPC is
not influenced by changing the pulsed voltage amplitude
in the range 3.5-7.0 kV. Likewise increasing the plasma
power density by raising the frequency from 58 kHz to
100 kHz does not have a significant effect on GPC. An
explanation for the unsaturated GPC as function of
exposure time is not yet available. Heat generation by the
‘non-thermal’ plasma it-self may play a role here, as the
temperature in the plasma zone will increase with
exposure time. Temperatures in the plasma zone have not
been determined. High temperature may lead to a
chemical vapour deposition (CVD) growth component.
Fig. 3. Growth per cycle (GPC) of ZnO as a function of
DEZ exposure time and thickness as a function of the
number of cycles. Oven temperature is 125 °C.
Fig. 4. Growth per cycle (GPC) of ZnO as a function of
DEZ and O 2 precursor flows. Oven temperature is 125
°C.
This suggests that surface limited saturation
characteristic for ALD processes is not fully reached.
3. SiO 2 deposition in deep trenches
2
O-20-6
ALD is the technology of choice for deposition of
conformal thin layers following complex threedimensional substrate morphologies. Using the actual
configuration for PE-S-ALD as shown in Fig. 2, we
investigated plasma-assisted atmospheric spatial ALD of
SiO 2 using the H 2 Si(N(C 2 H 5 ) 2 ) 2 precursor (SAM.24, Air
Liquide) at low temperature (50° C). Though the stepconformality of 70% is not optimum, one can expect that
this preliminary result can be further improved.
Considering that at atmospheric pressure reactive nitrogen
and oxygen plasma species have life-times ranging from
the sub-µs to ms range, and recombination losses at sidesurfaces of trenches are significant, the shown degree of
conformality may be rather surprising. A possible
mechanism for radical reactive species availability in
deep-trenches proceeds via indirect generation of radicals
by less-reactive (thus more diffusing) intermediate plasma
products such as metastable N 2 (A) for example. The
details of mechanisms are not easy to elucidate since
plasma-induced N 2 -O 2 chemistry involves a large number
of reactions [4].
(5a)
4. Alternative plasma source geometries
For improved control of gas flow and temperature
through the plasma zone, alternative source geometries
are being investigated. For example two sources as shown
in Figs. 1 and 2 can be used at both sides of a central gas
injection slit. In a second example, the plasma generation
is remote, as shown in Fig. 6. The geometry of this
‘proximity plasma jet’ allows for generating SDBD
plasma in very close vicinity of the substrate without
electric field interaction in the plasma with conductive or
capacitive charging dielectric thin films and substrates
which may lead to inhomogeneous deposition and film
damage [5].
Not shown in the figure is a regular series of ridges on
the dielectric which permit a homogeneous distributed
surface discharge in thin gas channels with 1.5 mm width
and 0.2 mm thickness. Local gas velocities reach 4-10 m/s
at moderate gas flow rates (2-5 slm for a 30 mm long slit).
As a result of the high velocity a sufficiently fast transfer
of plasma reactive species through the exit slit towards
the substrate is realised.
(5b)
(5d)
(5c)
Fig. 5. Cross-section SEM images of Si trenches with
aspect ratio ~20:1 lined with an SiO 2 film deposited
during 259 cycles of ~100 ms (50°C, 0.4 Hz, 10%
O 2 /N 2 ).
O-20-6
(5a) Overall picture of the 4.85 µm trenched
structure, (5b) SiO2 top layer ≈ 35 nm, (5c) midrange layer ≈ 25 nm, (5d) bottom layer ≈ 25 nm.
3
Fig. 6. The proximity plasma jet source
The effectiveness and homogeneity of the flux of
oxidative plasma species has been measured using
oxidation of a thin layer of amorphous carbon. Layers of
~20 nm carbon are first magnetron sputtered on precleaned, 1.5 mm thick, 1 inch wide glass samples. The
light transmission of the samples is measured using a
photo-scanner and used as a measure of oxidative etching.
By subtracting images before and after plasma treatment,
a maximum sensitivity for this method is achieved. The
different plasma sources are positioned at variable
distance above the (not moving) carbon samples. Etch
patterns show the reach and spreading of the reactive
plasma flux. An example of an etch pattern obtained with
the proximity plasma jet source, is shown in Fig. 7.
Fig. 7. Etch pattern after 2 minutes treatment with the
proximity plasma jet source. The source-substrate
distance is ~1 mm, air flow 5 slm.
CR (nm/min)
variations >0.2 mm may occur. Using pure N 2 in the
plasma jet, the nitrogen reactive species being mixed with
downstream air, the cleaning rate shows a maximum. This
maximum is clearly associated with the indirect formation
of reactive oxygen species by excited nitrogen species.
Though the number of data is limited, trends are clearly
visible.
35
blanket
30
Jet N2 (day 4)
25
Jet air (day 2)
20
15
10
5
0
0
1
2
3
Distance (mm)
Fig. 8. Etch rates (CR) as a function of substrate-source
distance (air/N 2 flow rate is 5 slm, accuracy of distance
position is +/- 0.1 mm)
This paper shows a selection of results obtained with a
unique combination of atmospheric plasma generation
based on surface dielectric barrier discharges and spatial
ALD. Applied and fundamental research activities are
being continued and can be updated during ISPC22.
5. References
[1] P. Poodt, A. Lankhorst, F. Roozeboom, C. Spee, D.
Maas, and A. Vermeer, Adv. Mater., 22, 3564, (2010).
[2] H.B. Profijt, S.E. Potts, M.C.M. van de Sanden, and
W.M.M. Kessels, J. Vac. Sc. Technol. A, 29 (2011).
[3] T. Unfer and J.P. Boeuf, J. Phys. D: Appl. Phys. 42
(2009).
[4] M. Simek et al., J. Phys. D: Appl. Phys. 43 (2010)
[5] Y. Creyghton, P. Poodt, M. Simor, and F.
Roozeboom, patent application EP 14173878.1, filed June
25 (2014)
Carbon etching rates have been measured for both the
parallel (blanket) type plasma source and proximity
plasma jet source as a function of distance. The results
shown in Fig. 8, clearly shown the diffusion limited
extend of plasma reactivity to approximately 0.4 mm
from the surface plasma of the blanket source. Using the
proximity plasma jet source, the plasma reactivity at a
distance of 1 mm is still significant. This result can be
relevant for scaling-up of PE-S-ALD in sheet-to-sheet or
roll-to-roll applications where substrate position
4
O-20-6