Wafer-Scale
MgB
2
Superconducting
Devices
Changsub
Kim,
*
Christina
Bell, Jake M. Evans, Jonathan
Greenfield,
Emma Batson,
Karl K. Berggren,
Nathan
S. Lewis, and Daniel P. Cunnane
*
Cite This:
ACS Nano
2024,
18, 27782−27792
Read
Online
ACCESS
Metrics
& More
Article
Recommendations
*
sı
Supporting
Information
ABSTRACT:
Progress
in superconducting
device
and detector
technologies
over
the past
decade
has realized
practical
applications
in quantum
computers,
detectors
for far-infrared
telescopes,
and optical
communications.
Superconducting
thin-
film materials,
however,
have
remained
largely
unchanged,
with
aluminum
still being
the material
of choice
for superconducting
qubits
and niobium
compounds
for high-frequency/high
kinetic
inductance
devices.
Magnesium
diboride
(MgB
2
), known
for its
highest
transition
temperature
(
T
c
= 39 K) among
metallic
superconductors,
is a viable
material
for elevated
temperature
and higher
frequency
superconducting
devices
moving
toward
THz frequencies.
However,
difficulty
in synthesizing
wafer-scale
thin films
has prevented
implementation
of MgB
2
devices
into the application
base of superconducting
electronics.
Here,
we
report
ultrasmooth
(<0.5
nm root-mean-square
roughness)
and uniform
MgB
2
thin (<100
nm) films
over 100 mm in diameter
and present
prototype
devices
fabricated
with these
films
demonstrating
key superconducting
properties
including
an internal
quality
factor
over 10
4
at 4.5 K and high tunable
kinetic
inductance
in the order
of tens of pH/sq
in a 40 nm thick
film. This
advancement
will enable
development
of elevated
temperature,
high-frequency
superconducting
quantum
circuits,
and
devices.
KEYWORDS:
MgB
2
, kinetic
inductance,
superconducting
devices,
wafer-scale,
thin films,
high frequency,
high-T
c
The
quantum
and nonlinear
nature
of superconductors
has
been
of scientific
interest
since
the discovery
of super-
conductivity.
Many
applications
of superconducting
phenom-
ena
using
thin-film
nano-
and
microdevices
have
shown
unparalleled
sensitivity
for both
power
1
−
4
and
coherent
detectors,
5,6
quantum-limited
amplification,
7,8
and
computa-
tion
(i.e.,
quantum
supremacy).
9
Current
state-of-the-art
superconducting
devices
are based
on tried
and
tested
elemental
aluminum
or niobium
thin films
due to the ease
of
deposition
and fabrication.
More
recently,
research
has taken
advantage
of more
novel
compounds
and doped
materials
like
TiN,
10
NbTiN,
11
Mn-doped
Al,
12
and granular
aluminum
(gr-
Al)
13
for high
nonlinear
kinetic
inductance
and to tune
the
critical
temperature
for pair-breaking
applications.
However,
because
of their
low-transition
temperatures
(
T
c
, 1.20
K for Al
and
9.26
K for Nb),
devices
operate
not
only
at low
temperatures
but also at low frequencies
of <90 GHz
for Al
and <700
GHz
for Nb from
their
small
superconducting
gaps,
Δ
= 1.764
k
B
T
c
, according
to the Bardeen
−
Cooper
−
Schrieffer
(BCS)
theory.
Using
higher
T
c
films
can allow
higher
temperature
operation,
higher
frequency
operation,
or a
combination
of the two to better
suit operational
needs
and
resilience
against
external
factors
and noise
.
MgB
2
has the
highest
bulk
T
c
of 39 K among
metallic
superconductors
14
and
is as high
as 41.8
K in thin films
by inducing
tensile
strain.
15
There
exist
two
superconducting
gaps,
with
the interaction
parameters
between
these
gaps
dependent
on film quality
and
orientation.
16
Superconducting
properties,
such
as density
of
states
and penetration
depth,
will fall somewhere
between
the
BCS
model
predictions
for the two
independent
gaps
�
Δ
π
∼
2.2
meV
and
Δ
σ
∼
7 meV
17
�
enabling
device
operations
above
1 THz.
18
MgB
2
thin-film
thermodynamics,
19,20
deposition,
15,21
−
23
and
fabrication
24
have
been
studied
extensively
since
the discovery
of superconductivity
in the compound,
and some
promising
prototypes
have
been
demonstrated,
25
−
32
but
practical
applications
have
not caught
on due
to lack
of scalability,
poor
reproducibility
of films,
and fabrication
immaturity
of the
material.
The
macroscopic
film properties
that would
enable
wider
adoption
of the material
include
large-scale
uniformity
and
roughness
below
1 nm root-mean-square
(rms),
while
Received:
August
11, 2024
Revised:
August
27, 2024
Accepted:
September
4, 2024
Published:
September
24,
2024
Article
www.acsnano.org
© 2024
California
Institute
of
Technology,
Gov\u2019t
sponsorship
acknowledged.
Published
by American
Chemical
Society
27782
https://doi.org/10.1021/acsnano.4c11001
ACS Nano
2024,
18, 27782
−
27792
This article is licensed under CC-BY-NC-ND 4.0
Downloaded via CALIFORNIA INST OF TECHNOLOGY on October 25, 2024 at 19:26:42 (UTC).
See https://pubs.acs.org/sharingguidelines for options on how to legitimately share published articles.
maintaining
good
superconducting
properties,
such
as high
T
c
and
J
c
. Furthermore,
deposition
on silicon
wafers
would
enable
direct
integration
of the material
into
the state-of-the-art
processes
and technologies
developed
by the semiconductor
industry,
such
as photolithography,
plasma
etch,
and
pack-
aging,
on a large
scale.
Wafer-scale
deposition
of MgB
2
by
reactive
evaporation
had been
reported
back
in 2006,
but there
are no follow-up
studies
since,
likely
due to its thickness
(500
nm),
roughness
(1
−
5
nm rms),
and lack of uniformity
around
the rotating
axis at the center
of the wafer.
33
Utilization
of a
commonly
available
PVD
technique
such
as sputtering
provides
good
uniformity.
While
many
groups
have
demon-
strated
the capability
to sputter
MgB
2
, none
have
matured
into
successful
technologies,
likely
due to difficulty
in fabricating
devices
from
these
films
or even
difficulty
in achieving
reproducible
films.
Here,
we report
large-scale
MgB
2
thin
films
on 100 mm diameter
Si substrates
with
T
c,0
over
32 K,
roughness
below
0.5 nm rms,
and
T
c,0
wafer
uniformity
of
94.50%.
T
c,0
can be as high
as 37 K, approaching
bulk
values,
if
we relax
the expectations
for film
roughness.
We further
developed
standardized
processes
for MgB
2
nano-
and
microdevice
fabrication
and
demonstrate
superconducting
resonators
with
Q
i
over
10
4
at 4.5 K,
J
c
of 10 MA/cm
2
at 4.2
K, and kinetic
inductance
which
can be tuned
from
moderate
to high
levels,
meeting
a strict
criterion
to realistically
achieve
mature
fabrication
capabilities.
RESULTS
AND
DISCUSSION
MgB
2
Thin-Film
Fabrication,
Properties,
and Charac-
terization.
The overall
flow
of our MgB
2
thin-film
fabrication
process
utilizing
magnetron
sputtering
is illustrated
in Figure
1a, and
the resulting
superconducting
MgB
2
thin
films
on
silicon
and
sapphire
wafers
are shown
in Figure
1b,c,
respectively.
Magnetron
sputtering
is widely
available,
easily
scalable,
and produces
uniform
films,
but in situ sputtered
MgB
2
films
resulted
in low
T
c
23,34
from
oxidation,
small
grain
size, and/or
off-stoichiometry
due to a high
vapor
pressure
and
low sticking
coefficient
of magnesium
at elevated
temperatures
over
200
°
C, as well as contamination.
Postannealing
of these
films
shows
improvements
in
T
c
but at the cost of roughness
(>10
nm rms).
Room-temperature
deposition
results
in a
uniform
distribution
of magnesium
but requires
a postanneal-
ing process.
Annealing
magnesium
−
boron
composite
film in
vacuum
resulted
in evaporation
of magnesium,
rough
surface,
and transition
temperature
around
6 K. Magnesium
evapo-
ration
is prevented
by capping
the composite
film with
a thin
(tens
of nm)
layer
of high
melting
temperature
material
such
as tantalum
or boron.
Tantalum
does
not
react
with
magnesium
or boron
at typical
annealing
temperatures
below
800
°
C but cracks
above
700
°
C and needs
to be removed
for
easily
measuring
superconducting
properties.
Boron,
a
dielectric
material
with
a high
melting
point
of 2076
°
C,
serves
as a better
capping
layer.
Surface
boron
oxide
has a low
melting
point
of 450
°
C and provides
a crack-free,
viscous
capping
layer.
35
There
are two potential
byproducts
between
MgB
2
and boron:
MgB
4
and MgB
7
. Both
have
slightly
higher
formation
energies
compared
to MgB
2
,
36
so their
formation
would
be minimal
and would
not affect
the measurement
of
superconducting
properties
even
if a thin
layer
of the
byproducts
is formed,
because
they
are dielectric
materials
unlike
metallic
MgB
2
. We have
developed
and demonstrated
fabrication
maturity
in removing
these
capping
layers
for
device
development
and optimization.
In our work,
we tried
many
deposition
conditions,
and mostly
through
optimizing
the roughness
of the film,
we chose
a codeposited
Mg
−
B
precursor
film at room
temperature.
Sputtering
of boron
is very
challenging,
and
a high
melting
point
leads
to a very
low
deposition
rate.
Given
the propensity
for oxidation,
we
optimized
the boron
sputtering
for a maximum
rate and then
tuned
the magnesium
sputtering
conditions
and
annealing
process
to achieve
optimized
films.
In order
to achieve
smooth
films,
the as-deposited,
preannealed
Mg
−
B
composite
film
Figure
1. Schematic
illustration
of superconducting
MgB
2
thin-film
fabrication
process
flow.
(a) Left:
magnesium
and boron
are cosputtered
onto
a rotating
substrate
to a desired
thickness
(e.g.,
50 nm).
A small
substrate
bias (e.g.,
15 W) is applied
to achieve
a smooth
surface.
Center:
a thin boron
capping
layer
is deposited
on top of the cosputtered
film.
Right:
the wafer
sample
is annealed
at around
600
°
C in a
100%
nitrogen
environment
for 2
−
10
min in a rapid
thermal
processor.
(b) The final product
is a superconducting
thin MgB
2
film with a
boron
cap on a substrate.
(c) MgB
2
thin film on a 100 mm diameter
single-crystal
silicon
substrate
with a thin (30 nm) silicon
nitride
buffer
layer.
(d) MgB
2
thin film on a 100 mm diameter
single-crystal
sapphire
substrate.
ACS Nano
www.acsnano.org
Article
https://doi.org/10.1021/acsnano.4c11001
ACS Nano
2024,
18, 27782
−
27792
27783
must
be as smooth
as possible.
A small
amount
of RF substrate
bias (15 W) during
deposition
reduced
the surface
roughness
from
1.74
nm to 0.34
to 0.476
nm (rms).
Increasing
the bias
beyond
this point
gave
denser
films
but etched
Mg or heated
the surface
too much
to achieve
controllable
stoichiometry
in a
cosputtered
film.
We
saw
indications
that
even
this
low
substrate
bias provided
enough
energy
to induce
some
reaction
between
Mg and B as we started
to see transitions
in as-
sputtered
films
before
any postannealing
step,
though
T
c,0
was
below
liquid
helium
temperatures.
A rapid
thermal
processor
(RTP)
provides
good
thermal
uniformity
across
large
wafers
and was used
for postannealing.
Because
magnesium
vapor
pressure
jumps
around
its melting
point
of 650
°
C, it is critical
to keep
the annealing
temperature
Figure
2. DC superconducting
properties
of MgB
2
thin
films.
(a) Resistivity
versus
temperature
plot
of a MgB
2
thin
film showing
superconducting
transition
with
T
c,0
= 32 K and (b) shown
across
a wider
temperature
range
from
4.2 to 300 K. (c) Critical
current
density
(
J
c
) of two superconducting
MgB
2
thin films
at different
temperatures
(
J
c
> 10
7
MA
·
cm
−
2
at 4.2 K) showing
reproducibility
of the films.
(d)
Resistivity
versus
temperature
plot of MgB
2
film with
T
c,0
= 37.2 K, highest
ever reported
for sputtered
MgB
2
film, and (e) shown
across
a
wider
temperature
range
from
4.2 to 300 K.
Figure
3. Morphological
characterizations
of postannealed
MgB
2
thin films.
(a) Atomic
force
microscopy
of the MgB
2
thin film with
T
c,0
of
32 K (Figure
2a) and the surface
roughness
of 0.476
nm rms. (b) High-angle
annular
dark-field
(HAADF)
STEM
image
of 40 nm thick
superconducting
MgB
2
thin film with a 30 nm boron
cap layer
on a high-resistivity
silicon
wafer
with a 30 nm silicon
nitride
buffer
layer,
showing
sharp
interfaces.
(c) Deviation
from
the median
magnesium
to boron
ratio
in 40 nm thick
boron-rich
MgB
2
film samples
on the
silicon
nitride
buffer
layer
(red)
and sapphire
(blue)
analyzed
by depth-profile
X-ray
photoelectron
spectroscopy.
The magnesium
to boron
ratio
stays
within
10% of the median
for samples
on silicon
nitride.
Significant
migration
of magnesium
from
the MgB
2
layer
to sapphire
results
in huge
deviation
of
−
30%
from
the median
ratio
near
the interface.
The range
of deviation
from
the median
Mg:B
ratio
for the
sample
on silicon
nitride
is shaded
in gray.
ACS Nano
www.acsnano.org
Article
https://doi.org/10.1021/acsnano.4c11001
ACS Nano
2024,
18, 27782
−
27792
27784
below
this
temperature
to avoid
any
potential
surface
roughening
caused
by evaporated
magnesium
that has yet to
react
with
boron.
By optimizing
the annealing
condition,
we
measured
T
c,0
of 32 K (Figure
2a,b)
for devices
from
MgB
2
thin
films
with
0.476
nm rms roughness
(Figure
3a) and a
circular
area of 100 mm diameter.
We further
achieved
T
c,0
of
37.2
K, shown
in Figure
2c,d,
highest
ever
reported
for
sputtered
film,
by inducing
large
grains,
thus sacrificing
surface
roughness,
from
postannealing
interchanging
magnesium
and
boron
layers
of stoichiometric
ratio
(2 boron
to 1 magnesium
atomic
ratio),
similar
to previous
attempts,
37
−
41
but with
an
additional
boron
capping
layer
on top. The massive
migration
of magnesium
into
the interfacing
boron
layers
led to large
MgB
2
grains
and high
T
c
, but at the same
time,
it resulted
in
large
voids
and rough
surfaces.
Certain
applications
may
take
advantage
of these
films,
particularly
when
devices
have
a very
small
active
area.
Sapphire
substrates
have
often
been
used
for growing
MgB
2
films
because
both
have
hexagonal
crystal
structure
and lattice
mismatch
is less than
0.1%
with
30
°
rotation.
42
However,
oxygen
readily
diffuses
from
sapphire
to MgB
2
because
magnesium
has
lower
oxidation
enthalpy
compared
to
aluminum
43
and
forms
an interdiffusion
layer.
18
This
is
especially
detrimental
for thin
films
under
100
nm.
An
alternative
substrate
is hexagonal
SiC,
especially
for in situ
growth
of MgB
2
, because
of the close
lattice
match.
In fact, the
highest
T
c
ever
measured
for MgB
2
is from
highly
textured
MgB
2
thin
film
slightly
strained
by SiC.
15
However,
SiC is
more
expensive
than
Si, and polycrystalline
MgB
2
films
from
the postannealing
process
would
not be able to take advantage
of the lattice
match.
Because
magnesium
reacts
with
silicon
to
form
Mg
2
Si, a thin inert
buffer
layer
was used
to prevent
direct
contact
between
MgB
2
and
Si. As was
demonstrated
previously,
33
silicon
nitride
proved
to be unreactive
against
magnesium,
and LPCVD
nitride
as thin
as 30 nm has been
successfully
tested
to be sufficiently
thick
enough
to serve
as a
good
buffer,
as shown
in the STEM
image
in Figure
3b. In
theory,
the thickness
is limited
to only
the threshold
where
no
pinholes
are left exposing
Si, mostly
dominated
by the initial
substrate
roughness.
The
LPCVD
nitride
deposition
on Si
wafers
is a highly
commercialized
process,
44
not requiring
the
development
of an independent
recipe
specific
to MgB
2
thin
films.
Early
in our work,
we found
that identical
deposition
and
annealing
recipes
resulted
in a higher
critical
temperature
on a
silicon
nitride
buffer
than
on sapphire.
Later,
it was confirmed
through
X-ray
photoelectron
spectroscopy
(XPS)
that
magnesium
from
the MgB
2
layer
migrated
to the sapphire
substrate
and resulted
in large
deviation
in stoichiometry
of the
film near
film/substrate
interface
(Figure
3c).
MgB
2
film
uniformity
is confirmed
by mapping
sheet
resistance
measurements
by eddy
current
on insulating
wafers,
such
as high-resistivity
silicon
or sapphire,
as well as measuring
T
c,0
at 15 different
points
across
a wafer.
As-deposited
Mg
−
B
composite
films
and post-rapid-annealed
MgB
2
thin
films
of
two
different
thicknesses
(50
and
100
nm
as-deposited,
reduced
by 20%
to 40 and 80 nm postannealed)
on 100 mm
Figure
4. Eddy
current
maps
of (a) 50 nm thick
as-deposited
Mg
−
B
composite
film with 85 W RF power
on magnesium
target,
(b) 40 nm
thick
postannealed
MgB
2
film after
annealing
at 600
°
C for 10 min,
(c) 100 nm thick
as-deposited
Mg
−
B
composite
film with
95 W RF
power
on magnesium
target,
and (d) 80 nm thick
postannealed
MgB
2
film after annealing
at 585
−
590
°
C for 2 min followed
by 600
−
615
°
C
for 4 min.
Sheet
resistances
from
eddy
current
measurements
directly
correspond
to magnesium
(as-deposited)
and MgB
2
(postannealed)
distributions.
The sheet
resistance
wafer
uniformities
(% of wafer
within
1
−
σ
of the average
sheet
resistance)
are between
96.49%
and
97.73%.
(e)
T
c,0
distribution
across
the 100 mm diameter
postannealed
MgB
2
film measured
at 15 different
points.
The
T
c,0
wafer
uniformity
is 94.50%,
calculated
by 1
−
(
T
c,0 max
−
T
c,0 min
)/(2
×
T
c,0 average
) as the number
of sample
points
is much
smaller
than
30.
ACS Nano
www.acsnano.org
Article
https://doi.org/10.1021/acsnano.4c11001
ACS Nano
2024,
18, 27782
−
27792
27785