Unlocking the potential of voltage control for high rate zirconium and

hafnium oxide deposition by reactive magnetron sputtering

Martynas Audronis
a
, Allan Matthews
a
, Ke stutis Ju

skevi

cius
b,
*
, Ramutis Drazdys
b
a
Department of Materials Science and Engineering, The University of Shefeld, Sir Robert Hadeld Building, Mappin Street, Shefeld S1 3JD, UK
b
Optical Coatings Laboratory, Centre for Physical Sciences and Technology, Savanoriu Ave. 231, LT-02300 Vilnius, Lithuania
a r t i c l e i n f o
Article history:
Received 5 February 2014
Received in revised form
23 April 2014
Accepted 25 April 2014
Keywords:
Zirconium oxide
Hafnium oxide
Reactive magnetron sputtering
a b s t r a c t
Target voltage (V
T
) feedback-based closed-loop gas ow control is one of the most cost-effective and
well-known methods to regulate reactive oxide thin lm sputtering processes. However, it can be
difcult to perform due to target voltage non-linearity, changes in signal direction and drifts in the case
of Zr and Hf sputtering in an Ar/O
2
ambient. We show in this paper that the use of the ratio of V
T
over the
O* (oxygen plasma emission, or alternatively oxygen partial pressure) as a feedback signal in a closed-
loop control, enables stable operation anywhere in the transition region (the region between metal and
fully poisoned sputtering target states) as well as extending into the forbidden parts of it. The result of
using V
T
over the O* ratio as the main feedback signal is a stable process exhibiting constant sputter
target state and partial pressure of oxygen as well as high deposition rate (e.g. 6.4

A/s for ZrO

2
and 9.2

A/s
for HfO
2
at D
TeS
20 cm by pulse-DC sputtering). Such a process is able to deliver ZrO
2
and HfO
2
coatings
that exhibit a low extinction coefcient in the near UV (0.0002 for HfO
2
and 0.001 for ZrO
2
at 350 nm)
and a high refractive index (2.05 for HfO
2
and 2.16 for ZrO
2
at 1000 nm), without any ion/plasma
assistance, at low substrate temperature (20e50

C) and no nitrogen addition.
2014 Elsevier Ltd. All rights reserved.
1. Introduction
Zirconium and hafnium oxide (ZrO
2
and HfO
2
) layers are used
widely in electronic and precision optical component layer stacks.
Many established techniques to deposit these layers exist and are
often adequate in terms of lms electrical and optical properties.
Rapid developments in the high power UV laser industry, however,
exert new pressures on optical components. Optical characteristics
and the Laser Induced Damage Threshold (LIDT) of components has
become the main limitation to the application of UV lasers [1]. ZrO
2
and HfO
2
exhibit a large transparent range from near-UV (above
350 nm) to mid-IR (w8000 nm), a high LIDT as well as good optical
and mechanical properties [2,3] and are therefore considered to be
the most important high index oxides for use in optical components
operating in the UV (e.g. high LIDT mirrors, polarizers, antireective
coatings and even chirped mirrors for group velocity dispersion
compensation in ultra-short laser pulse technology). The furthering
of the application of single compound layers in UV optical
components will involve mixtures, such as ZrO
2
/SiO
2
, HfO
2
/SiO
2
and HfO
2
/Al
2
O
3
[4e6].
Reactive pulse-DC sputtering is a well-established PVD tech-
nology for depositing oxide layers [7e16]. Literature surveys
(below) reveal however that ZrO
2
and HfO
2
are often not produced
in optimal conditions, in terms of process stability, deposition rate
and resultant lm properties.
Active feedback reactive sputtering process control is an
enabling technology that is relatively well-understood and has now
been practiced for decades [17]. High rate high quality reactive
sputter deposition of mixed oxides is even possible when operating
two processes simultaneously (co-sputtering), both in the transi-
tion region.
In 1997 Sproul et al. [11] reported the use of a closed-loop
process control system, based on a mass spectrometer, to deposit
zirconia and hafnia lms.
Other published work (e.g. papers by Venkataraj et al. [18] and
Zhao et al. [19]) reported production of ZrO
2
or HfO
2
lms while
operating the sputter target in or very near the fully poisoned
[20,21] target state, the disadvantages of which are well known.
In 2002, the addition of nitrogen was reported to provide im-
provements in process stability, arcing-avoidance and deposition
* Corresponding author. Tel.: 370 60405132.
E-mail addresses: m.audronis@yahoo.co.uk (M. Audronis),
kestutisjuskevicius18@gmail.com (K. Juskevicius).
Contents lists available at ScienceDirect
Vacuum
j ournal homepage: www. el sevi er. com/ l ocat e/ vacuum
http://dx.doi.org/10.1016/j.vacuum.2014.04.025
0042-207X/ 2014 Elsevier Ltd. All rights reserved.
Vacuum 107 (2014) 159e163
rate when operating in the fully poisoned state without closed-
loop control [22]. This was later conrmed by other groups [13,14].
Pervak et al. in their 2007 paper [12] on mediumfrequency (MF)
reactive HfO
2
sputter deposition said the lm quality was found to
degrade at a high rate. A good lm quality was realized at a rate of
around 0.3 nm/s. This was done using a system with a lambda
sensor for oxygen monitoring. Lambda sensors, however, have
many disadvantages, such as a slow response time, need for con-
stant air reference, need for heating, repeatability issues as well as
ability to work well only close to fully poisoned target state [20].
Target voltage (V
T
) feedback-based closed-loop gas ow control
(often termed Impedance Control) [23] is one of the most cost-
effective and well-known methods to regulate reactive oxide lm
sputtering processes (i.e. sputtering from a metal target). However,
this technique is difcult to performwith Zr and Hf oxide processes
due to target voltage non-linearity, changes in signal direction (i.e.
bidirectional signal changes caused by unidirectional changes in
gas ow rate) and drifts that are a known characteristics of reactive
Zr and Hf sputtering in an Ar/O
2
ambient [24]. Also, the use of V
T
-
based control allows reliable operation only on one of the slopes in
the transition zone and no stable operation is possible at the point
where V
T
peaks.
Optical Emission Spectrometry based process control (often
termed Plasma Emission Monitoring or PEM) [16] and mass
spectrometry-based process control [10] are able to x the partial
pressure of reactive gas or ux of sputtered metal species, but they
are not able to maintain the Hf or Zr sputtering target state con-
stant, which means the process control works only partially.
Also, it is not clear if nitrogen addition is needed when active
feedback process control is used.
The aim of this study was therefore to develop a simple coating
method that enables: i) quick and easy reactive Zr and Hf oxide
sputtering process stabilization anywhere in the transition zone,
ultimately providing stable processes in the most unstable point e
i.e. at max V
T
and ii) precise control of coating deposition processes
in a way that both target state (using Impedance Control) and
partial pressure of oxygen (using PEM) are constant during coating
deposition runs e thus the choice of V
T
and O* as feedback signals.
We demonstrate in this paper that the developed technology based
on the use of V
T
over the O* (oxygen plasma emission at 777 nm)
ratio as a feedback signal for closed-loop reactive gas ow control
(without nitrogen) allows stable and reproducible ZrO
2
and HfO
2
coating processes to be realized. Operation anywhere in the tran-
sition region (the region between metal and fully poisoned
sputtering target states) and extending into the forbidden parts, as
described further in the paper, is possible. Such processes can yield
routinely very high deposition rates (6.4

A/s for ZrO

2
and 9.2

A/s for
HfO
2
at D
TeS
20 cm) and excellent optical properties in the near
UV.
2. Experimental
Fig. 1 shows schematically the deposition chamber setup. A K.J.
Lesker system equipped with a four-inch diameter circular Torus
sputter cathode was used. This cathode was tted with either Zr or
Hf (purity e 99.95%) target, each 6 mm thick, 101.6 mm diameter
and non-bonded. The target-to-substrate distance (D
TeS
) was
200 2 mm. The system was initially pumped down with a cryo-
genic pump (Cryo-Torr 8F series from Helix Technology Corpora-
tion; 1500 L/s air) to a base pressure below 5 10
7
Torr
(6.7 10
5
Pa). The ambient (Ar/O
2
or Ar/O
2
N
2
) pressure during
sputtering runs was always between 2.4 mTorr and 2.6 mTorr. Prior
to depositing HfO
2
and ZrO
2
lms, the target was pre-sputtered for
3 min in pure Ar

(pressure 2.39 mTorr) to remove any surface

oxides. All process gasses used for deposition were >99.999% pure
and injected through MKS Instruments MassFlo1179A mass ow
controllers (MFCs). The Ar ow rate was kept constant at 20 stan-
dard cm
3
(sccm).
The substrate carousel was rotated at 15 rpm to improve the
uniformity of the deposited lms. Pulsed-DC power during growth
was maintained at 800 W (9.87 W/cm
2
average power density on
the target). The pulsed-DC generator (Advanced Energy Pinnacle
Plus) was operated at 100 kHz, 2 ms off time resulting in a 80% duty
cycle. The analog 0e10 V signal provided in the User port on the
generator was used as one of the feedback signals for process
control.
Prior to depositing the lms, the target was pre-sputtered for
3 min in a pure Ar discharge (working pressure 2.39 mTorr) to
remove any surface oxides.
Double-side polished fused silica (FS) substrates, 25.4 mm
diameter and 1 mm thick, were used to deposit thin layers with a
range of physical thickness (290e770 nm). The substrates were
maintained at room temperature (20e50

C) and no bias was
applied to the substrates during deposition. No additional ion or
plasma assistance was used.
A Nova Fabrica Ltd. FloTron X active feedback control system
was used to monitor the reactive sputtering processes and control
reactive gas inlet valves. Processes were carriedout inbothAr/O
2
and
Ar/O
2
N
2
ambients. Three process parameters were monitored: i)
Zr* and Hf* optical emission (spectral line at 351 nm and 298.3 nm),
ii) O* emission (spectral line at 777 nm) and target voltage V
T
(an
indicator of glowdischarge impedance). The process operating point
is selected by the user and a closed-loop controller then maintains it
xed at that set-point throughout the deposition run.
Transmittance spectra, from which lm refractive index (n) and
extinction coefcient (k) can be obtained, were measured in the
200 < l < 1300 nm wavelength range using a Lambda 950 (Perkin
Elmer) spectrophotometer. The transmittance data was analyzed
and optical constants were determined using OptiLayer Ltd.
OptiCharsoftware. Uncertainties of n and k determination are up
to 0.5%.
Fig. 1. A schematic diagram showing the process chamber used for Zr and Hf oxide
thin lm deposition by reactive magnetron sputtering.
M. Audronis et al. / Vacuum 107 (2014) 159e163 160
Deposition rates were calculated as the ratio of d
phys
/t; where
d
phys
e thin lm physical thickness determined by OptiChar, t e
deposition time. Standard deviation of ZrO
2
and HfO
2
deposition
rates from 5 to 4 processes respectively were evaluated resulting
rate measurement errors of 0.12

A/s for ZrO
2
and 0.15

A/s for
HfO
2
.
Variables investigated in this study were the operating set-point
in the reactive process window and nitrogen ow. The nitrogen
amount used was calculated by taking the ratio of N
2
to O
2
ows
and the values were varied from 0% to 37.5%.
3. Results and discussion
Fig. 2a shows an example of reactive gas ow ramp-up and
ramp-down experiments and the corresponding variation of the
monitored parameters. Classical responses [20] of Zr* and O* sensor
signals are observed. V
T
behavior is also as expected. The boundary
values (arbitrary, presented in %) of V
T
during the ramp are w76%
(maximum) and 7% (minimum). These values could be obtained
consistently by many consecutive O
2
ow ramp experiments.
When N
2
is introduced (Fig. 2b) along with O
2
the process behavior
changes only slightly and is essentially the same, within the limits
of experimental variation. Processes for both Zr and Hf sputter
targets appeared to be essentially identical, thus indicating essen-
tially very similar racetrack surface oxidation chemistry and
physics.
Normally, the maximum and minimum boundary values of the
V
T
signal would correspond to the process boundaries, beyond
which deposition processes yielding good quality stoichiometric
lms cannot be produced. These boundaries are established by
using industry standard reactive gas ramp routines (see for
example Refs. [20,25]). Behind those boundaries are inaccessible
zones often termed forbidden. However, our extensive experi-
ments (monitoring V
T
) have led to a discovery of the fact that, for Zr
and Hf, operation beyond the top prospective maximum value is
not only possible, but can yield a higher deposition rate process and
coatings with superior properties.
Fig. 3 shows an example of data logged during an experiment
demonstrating this phenomenon. Here reactive Zr sputtering in a
O
2
/Ar atmosphere is controlled by the closed-loop controller using
V
T
/O* as a feedback signal. The target is oxidized initially (period
from 0 to 40 s) after which the set-point is increased gradually to
move the process up the transition region. The V
T
during this
experiment has reached a signicantly higher maximum value
(w85%), as compared to the anticipated maximum76% indicated by
the O
2
ow experiment alone (as shown in Fig. 2a).
The full extent of the phenomenon can be seen in Fig. 4, where
feedback signals are shown during one of the deposition runs. The
HfO
2
process was controlled using a dual V
T
/O* sensor, for which
the set-point was 52%. The V
T
signal was 93% during the majority of
the run (whereas the boundary limit found by an O
2
ramp experi-
ment was 83%).
Such operation, i.e. at max V
T
value above the boundary limit,
was found to yield consistently the best processes (in terms of
deposition rate) and coating optical properties. The highest depo-
sition rates e (6.4 0.12)

A/s for ZrO

2
and (9.2 0.15)

A/s for HfO

2
at D
TeS
20 cm are obtained most readily when using the V
T
/O*
feedback signal. Although, as such, the use of ratios of two feedback
signals is a common practice in both industry and academia (e.g.
see paper by Dillon et al. [26]) no reports concerning deposition of
ZrO
2
and HfO
2
were found by the authors of this paper.
The introduction of nitrogen (up to 37.5% in respect to oxygen)
into both ZrO
2
and HfO
2
processes results in a decrease of more
than 30% in deposition rate, as compared to deposition with oxygen
only. Arcing is not an issue when operating in the transition region
due to the use of pulsed-DC power. Hence, the addition of nitrogen
appears not to be necessary in this respect either.
Results of this study as well as those published previously [24]
indicate that Zr and Hf reactive oxide sputtering processes exhibit
complex V
T
behavior. This behavior is essentially governed by
changes in the Ion Induced Secondary Electron Emission (ISEE)
[24]. The ISEE depends strongly on the nature of compounds
formed in the racetrack, as well as on coverage, impacting ion en-
ergy and other specic plasma parameters.
If a hypothesis proposed by Depla et al. [24] is correct, then data
for Zr and Hf V
T
behavior means that Zr and Hf target poisoning
Fig. 2. Sensor (V
T
, O* (777 nm), Zr* (351 nm) and V
T
/O*) and actuator (O
2
MFC ow rate) signals during oxygen (a) and oxygen nitrogen (b) ow ramp experiments.
Fig. 3. ZreAr/O
2
process data showing the set-point for the V
T
/O* ratio increasing as a
function of time from 10% (fully poisoned mode) to over 50% (near metal zone).
M. Audronis et al. / Vacuum 107 (2014) 159e163 161
processes (and possibly also for Ti) undergo several transitions (not
just one), such as metal-to-suboxide and then suboxide-to-oxide
(the suboxide could equally be a different phase oxide). These
transitions in the racetrack chemical composition show up clearly
in the V
T
behavior, but there is much less evidence of them in op-
tical emission, i.e. O* and Zr* (the same would be expected for O
2
p.
p.) signals (Figs. 2a, b and 3).
We found that the higher the V
T
signal is during the deposition
employing V
T
/O* ratio-based active feedback control, the better is
the optical quality and the higher is the deposition rate of coatings.
Remarkably, the V
T
can be pushed beyond the limits that process
characterization data suggests, i.e. into the forbidden states (Figs. 3
and 4). This corresponds to operation on the edge between two
transitions as discussed above. The target chemical composition
and state are unknown and were out of scope of the present study.
Experiments, similar to those conducted for reactive HIPIMS of Ti
[25] are highly desirable.
Our results, as well as previouslypublishedwork [13,14,18,22], all
indicate that, for the purpose of high rate reactive sputtering of ZrO
2
coatings, the addition of nitrogen can provide some benets when
operating only in a fully poisoned regime. The effects of adding
nitrogen to a feedback controlled ZrO
2
or HfO
2
process are mostly
adverse in terms of both deposition rate and optical lmproperties.
The highest deposition rate is obtained when sputtering with oxy-
gen only and the process is controlled using the V
T
/O* approach.
To illustrate the signicance of the results of this study Fig. 5a
and b compare optical properties of the best ZrO
2
lms produced
employing four different techniques: i) reactive sputtering in O
2
atmosphere using the V
T
/O*-based active feedback gas owcontrol
(at 9.87 W/cm
2
), ii) reactive sputtering in O
2
N
2
atmosphere using
the V
T
/O*-based active feedback gas ow control (at 9.87 W/cm
2
),
iii) reactive sputtering using direct gas owcontrol (at 3.08 W/cm
2
)
and iv) Ion Beam Sputtering (IBS) (optimized process).
The V
T
/O* ratio-based technique is not only an order of magni-
tude faster than IBS, but is also able to deliver lms that exhibit
higher n and lower k, hence lower absorption losses in the UV e
characteristics so important for laser technology.
The D
TeS
20 cm used in this study could be reduced to
w10 cm, which would allow an increase in the deposition rate by a
factor of 3e4. The V
T
/O*-based process control approach is also a
practical solution against target lifetime-related process drifts.
Oxide thin lm growth parameters at the substrate can be
vitally important and need to be discussed. Although accurate
analysis of lmgrowth conditions was out of the scope of this paper
it is clear that process technology developed in this study resulted
in optimal lm growth conditions. Higher V
T
will have produced
higher energy sputtered neutrals arriving at the substrate. We also
hypothesize that operation in the forbidden zone of the transition,
through V
T
and different racetrack surface chemical and phase
compositions, may have affected favorably composition and energy
distributions of charged species (e.g. negative ion generation and
their impact) arriving at the substrate. This needs to be veried by
means of mass and energy resolved mass spectrometry.
4. Conclusions
The V
T
/O* ratio, when used as a feedback signal in closed-loop
control of reactive Zr and Hf oxide sputtering process, guarantees
process stabilization anywhere in the transition region as well as
extending into the forbidden part of it. This yields a high rate
deposition process at room temperature with no additional plasma
assistance that is able to deliver lms exhibiting a low extinction
coefcient in the UV and high refractive index. Scaling this tech-
nology to large area is straightforward.
If just one feedback signal (e.g. O* or Zr*) is used (as in con-
ventional P.E.M.), V
T
is allowed to drift (thus no constant process)
and therefore maximum performance, in terms of stability, depo-
sition rate and optical lm properties, is not obtained. The highest
deposition rates e (6.4 0.12)

A/s for ZrO
2
and (9.2 0.15)

A/s for
HfO
2
at D
TeS
20 cm are obtained most readily when using the V
T
/
O* feedback signal.
Higher V
T
will have produced higher energy sputtered neutrals
arriving at the substrate. We hypothesize that operation in the
forbidden zone of the transition region, through V
T
and different
racetrack surface chemical/phase composition, may have favorably
affected composition and energy distributions of charged species
(e.g. negative ions) arriving at the substrate.
Fig. 5. Refractive index (a) and extinction coefcient (b) of ZrO
2
samples produced at
room temperature by V
T
/O* ratio-based feedback controlled reactive sputtering (black
lines; two shown e represent the range within which all the dispersions of samples
produced by V
T
/O* control method fall.), conventional constant ow rate controlled
reactive sputtering (dash line), 25% nitrogen addition (dash-dot line) and IBS (gray
line).
Fig. 4. An example of a HfO
2
deposition run controlled using the V
T
/O* ratio. Note the
V
T
is w93%, i.e. 10% higher than the boundary value (83%) identied.
M. Audronis et al. / Vacuum 107 (2014) 159e163 162
References
[1] Abromavicius G, Buzelis R, Drazdys R, Juskevicius K, Kicas S, Tolenis T, et al.
Lith J Phys 2011;51:303e8.
[2] Abromavicius G, Buzelis R, Drazdys R, Melninkaitis A, Sirutkaitis V. Proc SPIE;
2007:67200Y.
[3] Melninkaitis A, Miksys D, Sirutkaitis V, Abromavicius G, Buzelis R, Drazdys R.
Proc SPIE 2006;6596:65961I.
[4] Stenzel O, Wilbrandt S, Schrmann M, Kaiser N, Ehlers H, Mende M, et al. Appl
Opt 2011;50:C69e74.
[5] Stenzel O, Schrmann M, Wilbrandt S, Kaiser N, Tnnermann A, Mende M,
et al. Proc SPIE 2011;8168. 81681W1e10.
[6] Bruns S, Verghl M, Werner O, Wallendorf T. Thin Solid Films 2012;520:
4122e6.
[7] Audronis M, Matthews A, Leyland A. J Phys D Appl Phys 2008;41.
[8] Scherer M, Schmitt J, Latz R, Schanz M. In: 38th National Symposium of the
American Vacuum Society, AVS, Seattle, Washington (USA); 1992. pp. 1772e6.
[9] Schiller S, Goedicke K, Reschke J, Kirchhoff V, Schneider S, Milde F. Surf Coat
Technol 1993;61:331e7.
[10] Wong MS, Chia WJ, Yashar P, Schneider JM, Sproul WD, Barnett SA. Surf Coat
Technol 1996;86e87(Part 1):381e7.
[11] Sproul WD, GrahamME, Wong MS, Rudnik PJ. Surf Coat Technol 1997;89:10e5.
[12] Pervak V, Krausz F, Apolonski A. Thin Solid Films 2007;515:7984e9.
[13] Sarakinos K, Music D, Mrz S, To Baben M, Jiang K, Nahif F, et al. J Appl Phys
2010;108.
[14] Severin D, Kappertz O, Kubart T, Nyberg T, Berg S, Pug A, et al. Appl Phys Lett
2006;88:1e3.
[15] Severin D, Sarakinos K, Kappertz O, Pug A, Wuttig M. J Appl Phys 2008;103:
083306e11.
[16] Schiller S, Heisig U, Korndrfer C, Beister G, Reschke J, Steinfelder K, et al. Surf
Coat Technol 1987;33:405e23.
[17] Voevodin AA, Stevenson P, Rebholz C, Schneider JM, Matthews A. Vacuum
1995;46:723e9.
[18] Venkataraj S, Kappertz O, Weis H, Drese R, Jayavel R, Wuttig M. J Appl Phys
2002;92:3599e607.
[19] Zhao S, Ma F, Song Z, Xu K. Opt Mater 2008;30:910e5.
[20] Audronis M, Bellido-Gonzalez V, Daniel B. Surf Coat Technol 2010;204:2159e
64.
[21] Hollands E, Campbell DS. J Mater Sci 1968;3:544e52.
[22] Venkataraj S, Kappertz O, Jayavel R, Wuttig M. J Appl Phys 2002;92:2461e6.
[23] McMahon R, Afnito J, Parsons RR. J Vac Sci Technol 1981;20:376e8.
[24] Depla D, Heirwegh S, Mahieu S, Haemers J, De Gryse R. J Appl Phys 2007;101:
013301e9.
[25] Audronis M, Abrasonis G, Munnik F, Heller R, Chapon P, Bellido-Gonzalez V.
J Phys D Appl Phys 2012;45.
[26] Dillon RO, Le K, Ianno N. Thin Solid Films 2001;398e399:10e6.
M. Audronis et al. / Vacuum 107 (2014) 159e163 163