View
Online
Export
Citation
CrossMark
RESEARCH ARTICLE
|
SEPTEMBER 26 2023
Isotropic plasma-thermal atomic layer etching of
superconducting titanium nitride films using sequential
exposures of molecular oxygen and SF
6
/H
2
plasma
Azmain A. Hossain
;
Haozhe W
ang
;
David S. Catherall
;
Martin Leung
;
Harm C. M. Knoops
;
James R. Renzas
;
Austin J. Minnich
J. V
ac. Sci. T
echnol. A
41, 062601 (2023)
https://doi.org/10.1
116/6.0002965
Articles Y
ou May Be Interested In
Solid state ef
fects in the electronic structure of T
iCl
4
studied by XPS
J. Chem. Phys.
(July 1983)
Isotropic plasma-thermal atomic layer etching of aluminum nitride using SF
6
plasma and Al(CH
3
)
3
J. V
ac. Sci. T
echnol. A
(May 2023)
Sharp improvement of flashover strength from composite micro-textured surfaces
J. Appl. Phys.
(September 2017)
18 October 2023 18:13:37
Isotropic plasma-thermal atomic layer etching of
superconducting titanium nitride films using
sequential exposures of molecular oxygen and
SF
6
/H
2
plasma
Cite as: J. Vac. Sci. Technol. A
41
, 062601 (2023);
doi: 10.1116/6.0002965
View Online
Export Citation
CrossMar
k
Submitted: 13 July 2023 · Accepted: 10 August 2023 ·
Published Online: 26 September 2023
Azmain A. Hossain,
1
Haozhe Wang,
1
David S. Catherall,
1
Martin Leung,
2
Harm C. M. Knoops,
3,4
James R. Renzas,
3
and Austin J. Minnich
1
,
a)
AFFILIATIONS
1
Division of Engineering and Applied Science, California Institute of Technology, Pasadena, California 91125
2
Division of Natural Sciences, Pasadena City College, Pasadena, California 91106
3
Oxford Instruments Plasma Technology, North End, Bristol BS49 4AP, United Kingdom
4
Department of Applied Physics, Eindhoven University of Technology, P.O. Box 513, 5600MB Eindhoven, The Netherlands
a)
Author to whom correspondence should be addressed:
aminnich@caltech.edu
ABSTRACT
Microwave loss in superconducting TiN films is attributed to two-level systems in various interfaces arising in part from oxidation and
microfabrication-induced damage. Atomic layer etching (ALE) is an emerging subtractive fabrication method which is capable of etching
with angstrom-scale etch depth control and potentially less damage. However, while ALE processes for TiN have been reported, they either
employ HF vapor, incurring practical complications, or the etch rate lacks the desired control. Furthermore, the superconducting character-
istics of the etched films have not been characterized. Here, we report an isotropic plasma-thermal TiN ALE process consisting of sequential
exposures to molecular oxygen and an SF
6
/H
2
plasma. For certain ratios of SF
6
:H
2
flow rates, we observe selective etching of TiO
2
over TiN,
enabling self-limiting etching within a cycle. Etch rates were measured to vary from 1.1 Å/cycle at 150
C to 3.2 Å/cycle at 350
C using
ex situ
ellipsometry. We demonstrate that the superconducting critical temperature of the etched film does not decrease beyond that
expected from the decrease in film thickness, highlighting the low-damage nature of the process. These findings have relevance for applica-
tions of TiN in microwave kinetic inductance detectors and superconducting qubits.
Published under an exclusive license by the AVS.
https://doi.org/10.1116/6.0002965
I. INTRODUCTION
Titanium nitride (TiN) is a superconducting metal of interest
for microelectronics and superconducting quantum devices. Its
high kinetic inductance, low microwave loss, and high absorption
coefficient in the infrared and optical frequencies make it a promis-
ing material for single photon detectors,
1
,
2
ultrasensitive current
detectors,
3
quantum-limited parametric amplifiers,
4
and qubits.
5
,
6
Superconducting microwave resonators based on TiN routinely
exhibit internal quality factors Q
i
.
10
6
.
2
,
6
,
7
TiN is also used for
microelectronics applications in which it is employed as a copper
diffusion barrier and metal gate electrode.
8
–
10
In many of these
applications, imperfections at film interfaces are the primary
limitation to figures of merit for various devices; for instance, the
quality factor of superconducting microresonators is presently
thought to be limited by microwave surface loss associated with
two-level systems (TLS) in various interfaces.
11
–
13
Subtractive nano-
fabrication methods based on typical wet or dry etching processes
are unsuitable for mitigating TLS density in these devices due to
the lack of angstrom-scale precision in etching and the sub-surface
damage they induce.
14
–
16
Atomic layer etching (ALE) is an emerging subtractive nanofab-
rication process with potential to overcome these limitations.
17
–
19
Early forms of ALE focused on directional etching.
20
,
21
Directional
ALE is based on surface modification by adsorption of reactive
ARTICLE
pubs.aip.org/avs/jva
J. Vac. Sci. Technol. A
41
(6) Nov/Dec 2023; doi: 10.1116/6.0002965
41,
062601-1
Published under an exclusive license by the AVS
18 October 2023 18:13:37
species, and subsequent sputtering of the modified surface with
ions or neutral atoms of low energy exceeding only the sputtering
threshold of the modified surface.
22
,
23
Isotropic thermal ALE pro-
cesses have also been developed recently using sequential, self-
limiting surface chemical reactions.
24
In thermal ALE, the mate-
rial surface is modified to form a nonvolatile layer that can then
be removed by a selective mechanism, such as temperature
cycling,
25
,
26
ligand-exchange transmetalation reactions,
24
,
27
or
others.
18
Isotropic thermal and plasma ALE processes have now
been reported for various dielectrics and semiconductors, includ-
ing Al
2
O
3
,
28
,
29
SiO
2
,
30
,
31
AlN,
32
–
34
InGaAs,
35
,
36
and others.
18
,
37
–
39
Surface smoothing of etched surfaces using ALE has also been
reported for various metals and semiconductors.
28
,
36
,
40
,
41
For TiN, ALE processes based on fluorination and
ligand-exchange with Sn(acac)
2
, trimethylaluminum (TMA),
dimethylaluminum chloride, and SiCl
4
did not lead to etching.
42
When fluorinated, TiN retains its 3+ oxidation state, yielding TiF
3
.
TiF
3
either formed nonvolatile ligand-exchange products or did not
react with the precursors, and hence, no etching occurred. This dif-
ficulty was overcome by first converting the Ti to the 4+ oxidation
state with exposure to ozone or H
2
O
2
, which upon fluorination
using HF produced volatile TiF
4
.
43
A conceptually similar process
has also been reported using O
2
plasma and CF
4
plasma.
44
Despite these advances, limitations remain. The use of HF
vapor incurs practical complications. The process of
Ref. 44
based
on O
2
plasma and CF
4
requires a heating and cooling step per
cycle, which can lead to impractical time per cycle on conventional
plasma tools. Additionally, the recipe achieves nm/cycle etch rates,
which lacks the desired angstrom-scale control and low-damage char-
acteristics. Previous reports did not examine the effects of ALE on the
superconducting properties of the samples. Identifying alternate reac-
tants to HF vapor while maintaining angstrom-level precision over
the thickness and ensuring that superconducting properties are not
degraded, all remain topics of interest for TiN ALE.
Here, we report the isotropic atomic layer etching of TiN
using sequential exposures of O
2
gas and SF
6
/H
2
plasma. The
process is based on the selective etching of TiO
2
over TiN for
certain ratios of SF
6
:H
2
. The observed etch rates varied from 1.1 up
to 3.2 Å/cycle for temperatures between 150
C and 350
C, respec-
tively, as measured using
ex situ
ellipsometry. The etched surface
was found to exhibit a
40% decrease in surface roughness. The
superconducting transition temperature was unaffected by ALE
beyond the expected change due to the decrease in film thickness,
highlighting the low-damage nature of the process. Our findings
indicate the potential of ALE in the processing of TiN for super-
conducting quantum electronics and microelectronics applications.
II. EXPERIMENT
The plasma-thermal ALE process of this work is illustrated in
Fig. 1
. An exposure of molecular oxygen was used to oxidize the
surface of TiN to TiO
2
, followed by a purge. Next, a mixture of SF
6
and H
2
gas was introduced into the chamber and ignited to form
SF
6
/H
2
plasma. After this exposure, the reactor was again purged
to complete the cycle. The use of SF
6
/H
2
plasma was motivated by
noting that HF does not etch TiN, but fluorine radicals will sponta-
neously etch TiN.
43
,
45
Studies on SiN and Si etching using hydro-
gen and fluorine-containing plasma have shown that the plasma
formed by the mixture yields different products at different plasma
concentration ratios, including HF molecules at high hydrogen
concentrations (H
2
to F-containing gas flow rate ratio
*
2).
46
–
48
We, therefore, expected to observe an effect similar to that reported
in
Refs. 46
–
48
, in which fluorine radicals were found to combine with
hydrogen radicals via multiple pathways to form vibrationally excited
HF with negligible F radical concentration.
46
If this process did occur,
the HF formed
in situ
could then react with the film and selectively
etch TiO
2
over TiN, with minimal spontaneous etching from F radi-
cals at sufficiently high H
2
concentrations. The etch selectivity of TiO
2
over TiN is due to the differing oxidation states of Ti in each com-
pound, as previously observed in
Ref. 43
.TheformationofHFinthe
SF
6
/H
2
plasma is referred to as
“
in situ
HF
”
throughout the paper. An
SF
6
gas was used in this work because of its successful use in previous
work on the isotropic ALE of alumina and aluminum nitride.
33
,
49
We investigated this approach to ALE of TiN using an Oxford
Instruments FlexAL atomic layer deposition (ALD) system with an
inductively coupled plasma source, as described in
Refs. 50
and
51
.
The substrate table temperature varied between 150
C and 350
C,
as measured by the FlexAL substrate table thermometer. The
minimum temperature in our study was restricted to 150
C by the
tool. The sample was placed on a silicon carrier wafer, which sits
on the substrate table, which may cause a difference between the
true sample temperature and the table temperature. Prior to intro-
ducing the sample into the chamber for etching, the chamber walls
and the carrier wafer were conditioned by coating with 50 nm of
Al
2
O
3
using 300 cycles of Al
2
O
3
ALD.
51
Alumina was selected as it
does not form volatile fluoride species on exposure to SF
6
plasma. For
FIG. 1.
Schematic of the TiN ALE process involving exposures to molecular oxygen to oxidize the surface (O
2
, blue dots), followed by SF
6
/H
2
plasma (green dots) to
produce volatile etch products.
ARTICLE
pubs.aip.org/avs/jva
J. Vac. Sci. Technol. A
41
(6) Nov/Dec 2023; doi: 10.1116/6.0002965
41,
062601-2
Published under an exclusive license by the AVS
18 October 2023 18:13:37
TiN ALE, the sample was first exposed to 50 sccm O
2
and 50 sccm Ar
gas for 2 s at 100 mTorr pressure, followed by a 10 s purge. Next, a
mixture of 20 sccm H
2
and 4 sccm SF
6
was stabilized at 100 mTorr
for 5 s before striking the plasma at 100 W for 10 s. The excess reac-
tants were purged for 10 s before repeating the cycle. The recipe
resulted in a total time of
40 s per cycle. Before the sample was
moved to the loadlock, the chamber was pumped down for 60 s. The
sample was additionally held in the loadlock for 2 h to cool down
before exposure to air to reduce oxygen diffusion into the sample.
The film thickness before and after etching was measured by
ex situ
spectroscopic ellipsometry (J.A. Woolam M2000) at 60
and
70
from 370 to 1000 nm. Thickness was determined using 5
points on a 5
5mm
2
square array. Subsequently, the data were fit
using a Lorentz model to obtain the thickness of the samples.
43
,
52
The thickness and uncertainty values are the average and standard
deviation of the five points, respectively. XPS analysis was per-
formed using a Kratos Axis Ultra x-ray photoelectron spectrometer
using a monochromatic Al K
α
source. Depth profiling was per-
formed using an Ar ion beam with a 60 s interval for each cycle.
The estimated milling depth was calculated based on the initial and
final film thickness measured by
ex situ
ellipsometry and assuming
a constant ion milling rate. The XPS data were analyzed in
CASA-XPS from Casa Software Ltd. We adopt universal Tougaard
background and subpeak fitting routines from
Refs. 53
and
54
.
The film surface topography was characterized using a Bruker
Dimension Icon atomic force microscope (AFM) over a 0
:
25
0
:
25
μ
m
2
area. The raw height maps collected on the AFM were
processed by removing tilt via a linear plane-fit. The surface rough-
ness and the power spectral density (PSD) were computed from the
plane-fit height maps using procedures outlined in the previous lit-
erature.
41
,
55
The PSD provides a quantitative measure of the lateral
distance over which the surface profile varies in terms of spatial fre-
quencies.
55
,
56
The PSD was calculated by taking the absolute square
of the normalized 1D-discrete Fourier transform of each row and
column from the plane-fit AFM scan. The transformed data were
then averaged to produce a single PSD curve. Reported roughness
values were found to vary by <7% over three spots on each film.
Electrical resistivity measurements were performed on a
Quantum Design DynaCool Physical Property Measurement
System (PPMS). The TiN films were connected to the PPMS
sample holder by four aluminum wires, wirebonded with a
Westbond 7476D Wire Bonder. The film resistivity (
ρ
) was mea-
sured using a four-point setup.
7
The resistivity was measured from
6 to 1.7 K, and the data were used to calculate the superconducting
critical temperature (
T
c
) of the films.
The samples consisted of 50 and 60 nm thick TiN films on
high resistivity Si (100) wafers (
.
20 k
Ω
cm, UniversityWafer) pre-
pared using ALD with the same FlexAL system. The ALD process
consisted of sequential half-cycles of exposure to tetrakis(dimethy-
lamino)titanium (TDMAT) and nitrogen plasma with a 20 W DC
bias at 350
C, similar to the procedure reported in
Refs. 7
and
57
.
The resistance at 6 K and
T
c
of a 60 nm thick ALD TiN film were
measured to be 210
μ
Ω
cm and 3
:
22
+
0
:
06 K, respectively; these
values are comparable to those reported for other TiN films made
using TDMAT.
7
,
57
,
58
The chemical composition of the deposited
films is described in
Sec. III E
. The titania (TiO
2
) films used for
demonstrating etch selectivity in
Sec. III A
were made by oxidizing
TiN samples under an oxygen plasma for 5 min at 300
C, yielding
a 5 nm thick TiO
2
film on top of the TiN film. The thicknesses of
the TiO
2
films were measured using
ex situ
ellipsometry.
III. RESULTS
A. Selective etching with SF
6
/H
2
plasma
We begin by examining the etch rate of TiO
2
and TiN films
for various SF
6
:H
2
flow rate ratios,
η
.
Figure 2(a)
shows the etch
FIG. 2.
(a) Etch rate of TiO
2
(red circles) and TiN (blue squares) vs the SF
6
:H
2
flow rate ratio. The green shaded area represents the flow rate ratios for which
selective etching of TiO
2
over TiN was achieved. The vertical dashed black line
at a ratio of 0.2 represents the ratio used in the ALE experiments. (b) TiN thick-
ness change vs number of cycles with exposure only to O
2
gas (red triangles),
in situ
HF (green squares), full ALE process at 200
C (purple circles), and
300
C (blue diamonds). The dashed lines are guides to the eye.
ARTICLE
pubs.aip.org/avs/jva
J. Vac. Sci. Technol. A
41
(6) Nov/Dec 2023; doi: 10.1116/6.0002965
41,
062601-3
Published under an exclusive license by the AVS
18 October 2023 18:13:37
rates of TiN and TiO
2
versus
η
at 300
C. For
η
&
0
:
05, negligible
etching of either film is observed. At
η
¼
0, we measure an etch
rate of
0
:
03 Å/cycle. This value is within the measurement
error of the ellipsometer, and as such, we do not attribute physi-
cal significance to the negative value. The other negative etch
rates correspond to an increase in the thickness of the film,
which we assume to be growth of nonvolatile TiF
3
.For
η
0
:
1,
we observe spontaneous etching of TiO
2
, with the etch rate
monotonically increasing with
η
. For TiN, we observe no etching
for
η
0
:
2, but for
η
0
:
25, etching occurs. We attribute these
observations to the formation of
in situ
HF along with negligible
fluorine radical concentration for 0
:
05
,
η
0
:
2. For
η
0
:
25,
the concentration of F radicals becomes sufficient to spontane-
ously etch the TiN, leading to increasing etch rates for both
films. From our measurements, we find that 0
:
1
η
0
:
2
achieves selective etching of TiO
2
over TiN. To obtain the
highest etch selectivity of TiO
2
over TiN, we select
η
¼
0
:
2for
our experiments. This 1:5 ratio of SF
6
:H
2
plasma is used
throughout the rest of the paper.
B. TiN ALE using O
2
and
in situ
HF exposures
Figure 3(b)
shows the thickness change of TiN versus
number of cycles for both half-cycles and for the full ALE recipe
at 200
Cand300
C. For the half-cycles, the thickness change
was measured after exposure to only molecular oxygen or
in situ
HF. No etching was observed for either half-cycle. In contrast, we
observe a decrease in the thickness with increasing number of
cycles when using both steps. The etch rate is calculated by divid-
ing the total thickness change by the number of cycles, giving
values of 2
:
4
+
0
:
16 Å/cycle at 200
Cand3
:
2
+
0
:
10 Å/cycle at
300
C.
We further examine the effect of temperature on the etch rate.
Figure 3(a)
shows the EPC versus table temperature ranging from
150
C to 350
C. The etch rates are calculated from the thickness
change over 100 cycles. We find that the etch rate increases from
1.1 Å/cycle at 150
C to 3.2 Å/cycle at 300
C. In analogy to other
works,
18
,
43
,
44
we attribute the etch rate increase with temperature to
the higher diffusion rates at higher temperatures in the oxidation
step, leading to thicker oxides, which are etched at each step. We
also observe a constant etch rate from 300
C to 350
C, similar to
what is reported in Fig. 7 of
Ref. 43
.
We also explored the self-limiting nature of the process
by measuring the saturation curves of each half-cycle. For each
saturation curve, the temperature is set to 300
C, and the purge
times and one half-cycle time are fixed while the other is varied.
In
Fig. 3(b)
, the
in situ
HF step is fixed at 10 s, while the etch rate
is measured versus the oxygen exposure time. The etch rate is
observed to saturate to
3 Å/cycle above 2 s, which is consistent
with the self-limiting nature of the oxidation step. In
Fig. 3(c)
, the
oxidation step is fixed at 2 s, while the etch rate is measured versus
in situ
HF exposure time. The etch rate saturates to
3 Å/cycle
above 10 s, which is consistent with the selectivity of the
in situ
HF
to etch TiO
2
and terminate on the TiN.
C. Characterization of film composition
We next characterize the chemical composition of the TiN
films before and after ALE using XPS. In
Fig. 4
, we show the core
levels of Ti2p, N1s, O1s, C1s, and F1s. For the Ti2p XPS spectra in
Fig. 4(a)
, we observe five components. Each component is a
doublet consisting of a 2p
3
=
2
and 2p
1
=
2
subpeak. We observe sub-
peaks corresponding to Ti
–
C (454.9 and 460.4 eV),
59
–
61
Ti
–
N
(455.1 and 460.8 eV),
62
–
64
Ti
–
ON (456.5 and 462.3 eV),
62
–
64
Ti
–
O
(458.5 and 464.2 eV),
62
–
64
and Ti
–
F (459.4 and 465.6 eV).
65
,
66
In
Fig. 4(b)
, we report the N1s spectra with two subpeaks at 397.1 and
398.9 eV, belonging to N
–
Ti and N
–
O bonds, respectively.
62
–
64
In
Fig. 4(c)
, we report the O1s spectra with two subpeaks at 530.4 eV
and 532.2 eV, corresponding to O
–
Ti and O
–
N bonds,
respectively.
62
–
64
In
Fig. 4(d)
, we report the F1s spectra with two
subpeaks at 684.9 and 690.3 eV, corresponding to F
–
Ti and F
–
C
bonds, respectively.
65
–
67
FIG. 3.
(a) TiN ALE etch per cycle (EPC) vs substrate table temperature. (b) EPC vs O
2
gas exposure time with
in situ
HF exposure time fixed at 10 s at 300
C. (c) EPC
vs
in situ
HF time with O
2
exposure time fixed at 2 s at 300
C. The etch rates are observed to saturate with exposure time, demonstrating the self-limiting nature of the
ALE process. The dashed lines are guides to the eye.
ARTICLE
pubs.aip.org/avs/jva
J. Vac. Sci. Technol. A
41
(6) Nov/Dec 2023; doi: 10.1116/6.0002965
41,
062601-4
Published under an exclusive license by the AVS
18 October 2023 18:13:37
We observe that the Ti2p spectra are dominated by oxides and
oxynitrides, consistent with the presence of a native oxide on
TiN.
58
,
63
After ALE [bottom panels of
Figs. 4(a)
–
4(c)
], an increase
in the magnitude of the Ti
–
N and N
–
Ti peaks is observed along
with an overall decrease in the O1s peak magnitude. The decreased
O1s signal implies reduced native oxide concentration after ALE, as
has been observed in ALE of other materials.
33
,
68
,
69
The F1s spectra
for the original sample may be attributed to contamination from
using the same chamber for deposition and etching, which is con-
sistent with the reduced magnitude of the F1s peak in the original
sample compared to that in the ALE-treated sample [bottom panel
of
Fig. 4(d)
].
We also performed depth-profiling XPS to determine the
atomic concentrations on the surface and in the bulk. In
Fig. 5
,we
show the atomic concentrations of Ti, N, F, C, and O versus sput-
tering time and estimated depth in the original and ALE-treated
films. In the original sample [
Fig. 5(a)
], the atomic concentrations
on the surface are 31.9% (Ti), 37.6% (N), 16.1% (O), 12.0% (C),
and 2.4% (F). After 120 s Ar milling (
3
:
5 nm), the atomic concen-
trations plateau to their bulk values of 48.6% (Ti), 42.3% (N), 6.1%
(O), 1.9% (C), and 1.1% (F). The carbon and oxygen levels are con-
sistent with other reported ALD TiN films made using
TDMAT.
58
,
70
,
71
The carbon signal on the surface is observed to be
predominantly C
–
O and C
–
H bonds, expected from adventitious
FIG. 4.
Surface XPS spectra showing (a) Ti2p, (b) N1s, (c) O1s, and (d) F1s spectra. The spectra are shown for (top) original and (bottom) etched TiN films. The m
ea-
sured (gray dots) and fit spectra (black lines) intensities are reported in arbitrary units (a.u.) against the binding energy on the
x
axis. The
y
axis scale is identical between
panels within each subfigure.
ARTICLE
pubs.aip.org/avs/jva
J. Vac. Sci. Technol. A
41
(6) Nov/Dec 2023; doi: 10.1116/6.0002965
41,
062601-5
Published under an exclusive license by the AVS
18 October 2023 18:13:37