of 32
1
Supplementary
Information
for
An
implantable
piezoelectric
ultrasound
stimulator
(ImPULS) for
deep
brain
activation
Jason
F. Hou
1
†, Md
Osman
Goni
Nayeem
1
†,
Kian
A. Caplan
2
, Evan
A. Ruesch
3
, Albit
Caban
-Murillo
3
,
Ernesto
Criado
-Hidalgo
4
, Sarah
B. Ornellas
1
, Brandon
Williams
5
, Ayeilla
A. Pearce
2
, Huseyin
E.
Dagdeviren
6
, Michelle
Surets
3
, John
A. White
5
, Mikhail
G. Shapiro
4
, Fan
Wang
2
, Steve
Ramirez
3
, Canan
Dagdeviren
1
*
Affiliations:
1
Media Lab,
Massachusetts
Institute
of
Technology, Cambridge,
MA
02139, United
States.
2
Department
of Brain
and
Cognitive
Sciences,
McGovern
Institute
for
Brain
Research,
Massachusetts
Institute
of
Technology, Cambridge, MA
02139, USA.
3
Department
of Psychological
and
Brain
Sciences,
The
Center
for
Systems
Neuroscience,
Boston
University, Boston,
02215,
MA,
USA.
4
Division
of Chemistry
and
Chemical
Engineering,
California
Institute
of Technology,
Pasadena,
CA,
91125, USA
5
Center
for
Systems
Neuroscience,
Neurophotonics
Center,
Department
of Biomedical
Engineering,
Boston University,
610 Commonwealth
Ave.,
Boston,
MA
02215, USA
6
Department
of
Neurosurgery, Faculty
of
Medicine,
Istanbul
University,
Istanbul,
34093, Turkey.
†These
authors
contributed
equally:
Jason
F.
Hou,
Md
Osman
Goni
Nayeem
*Corresponding
author. Email:
canand@media.mit.edu
This
supplementary
information
contains:
Supplementary
Note
1 -5, Supplementary
Figures
1-
23, Supplementary
Video
1, and Supplementary
References
1-
32.
2
Supplementary
Note
1:
Rationale for
choosing
KNN
as piezoelectric
material
for
ImPULS
We
chose
KNN
over
PZT as
the
piezoelectric
material
for
ImpULS
fabrication.
This
is because i)
KNN
is
biocompatible
ii) has
piezoelectric properties comparable or
greater
than
that
of
PZT (
SI ref
1-3
). A
comparison
of the d
33
value
and
Curie
temperature
of
different
piezoelectric
materials
(including
PZT)
in
the
thin
film
and bulk form
is given
below:
Material
d
33
(pC/N)
T
c
(
°
C)
Thin
Films
Bulk
PZT
100
-
150
(SI
ref
4)
390
-
510
(SI
ref
12)
300
400
KNN
74
-
128
(SI
ref
5&6)
300
-
690
(SI
ref
13)
350
BaTiO
3
100
(SI
ref
7)
250
-
500
(SI
ref
14)
130
ZnO
12.7
(SI
ref
8)
9.93
(SI
ref
15)
6
Doped
ZnO
(Transition
Metal)
128
(SI
ref
9)
110
(SI
ref
16)
280
-
500
(SI
ref
18)
AlN
6
(SI
ref
10)
-
1150
PVDF
-
TrFE
28
(SI
ref
11)
20
-
30
(SI
ref
17)
110
From
the comparison,
it is clear
that
KNN
has
comparable
piezoelectric
performance
to the
most
widely
used
piezoelectric
material,
PZT.
KNN
also
possesses superior
biocompatibility
with
regard
to the
cytotoxicity
of its breakdown byproducts
compared
to PZT (
SI
ref
19
).
Furthermore,
its high
Curie
temperature
enables
advanced
fabrication
techniques
to create
device
architectures
such
as the
piezoelectric
micromachined
ultrasound
transducer
(pMUT),
which
greatly
enhances
the
piezoelectric properties of
the
device without poling
or chemical
modification.
Indeed,
piezoelectric
devices
with
KNN have
been
fabricated,
implanted,
and evaluated
for
biocompatibility
(
SI
ref
20
).
In sum, other
piezoelectric
ceramics
and
polymers
have
biocompatible properties,
such
as
BaTiO
3
, ZnO,
and
PVDF,
however,
they
either
have
inferior
electromechanical
efficiency
and/or
thermal
processability
compared
to KNN.
These
design
parameters
make
KNN
the
best
material
choice
over
PZT,
especially
for
microfabricated
implantable
neurostimulation devices
with
the pMUT
architecture.
Supplementary
Note
2:
Flexible
SU
-8 based
pMUT
vs rigid
Si
-based
pMUT
SU
-8
is a
common
material
used
for
biocompatible
implanted
devices
due
to its low
stiffness (Young
s
Modulus, 100x
less
than
Si),
photo-
patternability
, and
encapsulation
ability
(SI
refs
21,
22
). These
factors
are important for
implanted
devices,
as previous
studies
reported
that,
due
to its flexibility,
the
mechanical
3
damage
caused
by SU
- 8 needles
in the
rat's
brain
during
its insertion
is lower
than
that
caused
by rigid
Si
needles
(SI
ref
23). In chronic
studies,
SU
-8 devices
outperformed
Si devices
in minimizing
tissue
damage
from
mechanical
mismatch
(SI
ref
24).
The
frequency
-dependence
of an edge
-clamped
circular
pMUT
can
be modeled
with
f
= (
훼훼
/2
휋휋
r
2
)
√(
D
E
/
h
) where
D
E
=
Eh
3
/(12(1-
ʋ
2
), where
D
E
is the
flexural
rigidity,
훼훼
is the
resonance
mode
constant,
h
is the
diaphragm
thickness,
E
is the
effective
Young
s modulus,
ʋ
is the
Poisson
s ratio,
is the
effective
density
of the
diaphragm.
As
observed,
to maintain
a sub-
MHz
resonance
frequency
with
Si,
the
diaphragm
thickness
must
be reduced
at the
expense
of insulation
or piezo
thickness;
or the
diaphragm
radius
must
be increased
significantly,
which
decreases
the
spatial
resolution
of the
device.
The
precedent
literature
serves
as the
basis
to make
an informed
decision
of fabricating
flexible
pMUT
using
SU
-8 instead
of using
rigid
Si.
We
performed
a COMSOL
multiphysics
simulation
for
resonance
frequency
and
acoustic
pressure
of a Si pMUT
and
compared
it with
the
results
of our
flexible
pMUT
as
shown
in supplementary
fig.
5.
As
seen
from
the
simulation,
for
the
same
device
dimension,
the
SU
-8-
based
pMUT
shows
a dominant
resonance
frequency
at 545
kHz
whereas
Si-based
pMUT
has
resonance
at
1350
kHz.
It is well
established
in the
literature
that
robust
stimulation
in the
brain
occurs
at a sub
-MHz
frequency
range
ideally
around
500
kHz
(refs
15,
22,
23,
24
in the
manuscript)
which
helped
in choosing
SU
-8 based
neurostimulator
design.
In addition,
as seen
in Supplementary
Fig
5b,
the
acoustic
pressure
exerted
by
both
probes
at their
resonance
frequency
is almost
similar,
yet
SU
-8- based
probes
have
the
obvious
advantage
of mechanical
flexibility
as discussed
above.
Indeed,
it
is possible
to reduce
the
resonance
frequency
of a pMUT
structure
simply
by
increasing
the
diameter
of the
cavity
(supplementary
fig.
4),
however,
to reduce
the
resonance
frequency
of Si-based
pMUT
to ~500
kHz
needs
significant
increase
in cavity
size
thereby
overall
device
size
(SI
ref
25). Considering
all this
, we
have
decided
to
fabricate
flexible
SU
-8 based
pMUT
instead
of rigid
Si
-based
pMUT.
Supplementary
Note
3:
Potential
mechanisms of
action
of
Ultrasound Neuromodulation
The
ability
of ultrasound waves
to activate
neural
cells of
various
types
has
been
demonstrated
in several
past
works
with
explorations
into
mechanisms encompassing
the
activation
of
mechano
-sensitive
PIEZO
and TRP
channels, demonstrated
in vitro
(refs
11,
14
in manuscript)
and
in vivo
with
transcranial
focused
ultrasound
(ref 14
in manuscript
and
SI
ref
26
). These
channels
exist
in both
neurons
and
astrocytes,
but
studies
have
shown
that
different
neural
cell
types
might
respond differently
to
US
stimulation.
Zhu
et al.
(2023)
demonstrated that knocking out
the
highly
mechano-
sensitive
PIEZO
channels
in neurons
resulted
in the
loss
of
US
modulation
sensitivity
while
knocking out
PIEZO
channels
in
astrocytes
did
not
(SI
ref
13). Lee
et al.
(2023)
and Oh
et al.
(2019)
demonstrated
that
TRPA
-channels
from
astrocytes can
be
ultrasonically
activated
and
are
sufficient
to indirectly
excite neurons
via
glutamate
release
(Ref
14
in
manuscript
and SI
ref
27).
Therefore, we
hypothesize
that
ultrasound can
activate
both neurons
and
4
astrocytes,
although likely
employing
different
mechanisms.
Indeed, the
ultrasound
frequency
and
stimulation
parameters
can
potentially
be
tuned
to achieve
a degree of
cell
selectivity
as has
been
demonstrated
by
other
groups
(ref
61 in
the
manuscript). Genetic
or
pharmacological
studies
that
disable
mechanosensitive
ion
channels
simultaneously
affect
other
physiological
processes that
maintain
cell
or
organism
viability
(SI
ref
28).
With
the
development
of
new
sonogenetic
tools
that
enhance
mechanosensitivity
without
disabling channel
activity
(SI ref
29)
and
that
maintain
a diverse
environment
of mechanosensitive cells
(SI ref
30),
the
future
direction
of
this
work
would
include
studies
on
how
the
neuronal
activation
is produced
by
the
ImPULS.
In our
study,
the
calcium
indicator
GCaMP7F
labels
primary
excitatory
neurons,
and
we
demonstrated
activation
of excitatory
cells
when
the
transducer
is placed
adjacent
to the
neuron
bodies
of granule
cells
in
the
dentate
gyrus
(supplementary
fig.
18).
It is important
to note,
however,
that
its expression
does
not
imply
the
lack
of activation
of astrocytes/glial
cells.
We
achieved
robust
stimulation
using
an ultrasound
driving
protocol
that
is known
to excite
neurons
(refs
23,
60
in manuscript),
but
the
mechanism
of action
can
be
partially
driven
by
indirect
excitation
via
astrocytes
gliotransmitters
release.
In the
scope
of this
work,
we
have
demonstrated
the
ability
of a new
implantable
and
spatially
precise
device
to cause
neuron
excitation,
and
we
envision
that
further
studies
can
help
elucidate
the
mechanism
of
action
of
this
activation
in the
various
regions
we
tested,
whether
via
direct
or
indirect
stimulation.
Supplementary
Note
4:
Rationale
for
Ultrasound parameters
The
driving
frequency
of 500
kHz
in water
is well
described
in the
literature
as an appropriate
frequency
for
transcranial
neuromodulation
in mice
brains
(refs
15,
22,
23
and
24
in manuscript),
as it presents
a good
tradeoff
between
skull
transmission
(more
efficient
in ranges
below
1 MHz)
and
spatial
selectivity
(focus
size
decreases with
frequency increase).
Although our
proposed
implantable
device
bypasses
the
skull
and
is focused
in a volume
<100
μm
3
, we
chose
to design
a device
with
500
kHz
resonant
frequency
as
it could
reproduce
or explore
similar
established
protocols
for
neuron
activation.
Ye
et al. (2016)
investigated
the
frequency
dependence
of ultrasound
neuromodulation
in the
mouse
brain,
and
in the
range
of 0.3
- 2.9
MHz,
demonstrated
that
the
activation
success
rate
was
nearly
flat
at lower
frequencies
but,
at higher
frequencies,
higher
spatial
peak
intensities
were
necessary
to attain
comparable
success
rates
when
contrasted
with
lower
ones
(ref 16
in manuscript).
The
Pulse
Repetition
Frequency
dependency
on
ultrasound
neuromodulation
was
explored
by
Manuel
et
al. (2020)
, who
demonstrated
that
pulsed
stimulation
was
more
efficient
than
continuous
wave
stimulation
(ref
62
in the
manuscript).
Among
pulsed
US,
the
parameter
space
that
led
to the
biggest
activation
of
neurons
(quantified
by
calcium
imaging)
had
PRF
of 1500
Hz,
center
frequency
of 500
kHz,
acoustic
pressure
of
100 kPa,
and burst
duty
cycle
of
60%,
which
are
very
similar
to our
stimulation
parameters.
5
Pressures
close
to 100
kPa
have
been
shown
to activate
neural
circuitry,
as exemplified
by Tufail
et al.
(2010)
(ref
22
in the
manuscript),
who
tested
pulsed
driving
frequencies
in the
0.25-
0.5
MHz
range
with
max
pressure
of 97
kPa
and
published
a separate
protocols
paper
(ref
21
in the
manuscript).
Notably,
they
found
that
their
pulsed
protocol
with
PRFs
in the
2.5
kHz
range
was
sufficient
to activate
robust
stimulation
in hippocampal
circuits.
In other
studies
by
Yoo
et al. (2022)
(ref
11
in the
manuscript)
in cortical
neurons,
pressures
exceeding
150
kPa
were
needed
to reliably
drive
neural
activation
but
subthreshold
pressures
still
had modulatory effects.
Taken
together,
we
can
see
there
are
a number
of parameter
investigations
for
the
ultrasonic
modulation
of
neuronal
tissue,
but
no
consensus
on
which
optimal
parameters
should
be used.
Recommendations
for
successful
parameters
of ultrasound
neuromodulation
encompass
a range,
and
our
chosen
parameters
are
contained
in the
recommendations
of both
Blackmore
et al.
s (2019)
review
and
Tufail
et al.
s (2011)
protocol.
Often
the
choice
of
parameters
is limited
by the
availability
of commercial
ultrasound
probes
or
the
manufacturing
capabilities
of research
facilities.
One
advantage
of our
chosen
method
of pMUT
fabrication
is the
possibility
of creating
an ultrasound
element
or an array
of elements
in many
sizes
and
center
frequencies,
a flexibility
that
will
allow
for
the
continuation
of parameter
investigations
in various
ultrasound
modulation
applications.
Therefore,
for
the
current
work
investigating
neuron
excitation
in mice,
our
ultrasound
driving
parameters
of 500 kHz,
10-50%
duty
cycle,
1500
Hz
pulse
repetition
frequency
are
consistent
with
literature
and
corroborated
our
findings
of ImPULS
s ability
to elicit
a modulatory
response.
Supplementary
Note
5:
COMSOL
simulation
parameters
The
geometric
parameters
used
in the
COMSOL
simulation
model:
Layer
Material
Radius
(μm)
Thickness
(μm)
Passivation
layer
SU
-
8
70
0.5
Top
electrode
Au
30
0.25
Piezoelectric
layer
KNN
50
1
Bottom
electrode
Pt
50
0.25
Membrane
layer
SU
-
8
70
1
Cavity
-
50
20
Backing
layer
SU
-
8
70
20
The
properties
of
materials
used
in the
COMSOL
simulation
model:
6
Properties
SU
-
8
Au
Pt
KNN
Density(Kg/m
3
)
1190
19300
21450
4000
Young’s
modulus
(GPa)
4.02
70
168
65
Poisson’s
ratio
0.22
0.44
0.38
0.3
Piezoelectric
coefficient
(C/m
2
)
12
Relative
permittivity
1500
7
Supplementary
Figure
1 | Schematic
illustration
of
microfabrication
of
the
ImPULS.
8
Supplementary
Figure
2 | Comparison
of transfer
printing
using
two
different
type
s of anchor
layer
patterning.
a,
Time
d- etch
transfer
method
b,
Single
-layer
anchor
transfer
method
c,
Microscopic
image
of
time
-etch
transferred
pattern.
This
type
of transfer
creates
crack
s on
the
bottom
Pt electrode.
Scale
bar,
100
μm.
d,
Microscopic
image
of Single
-layer
anchor
transfer
pattern.
No
crack
forms
on
bottom
electrode
during
transfer.
This
method
is less
time
sensitive,
therefore
improving
the
yield
significantly.
Scale
bar,
100 μm.
9
Supplementary
Figure
3 | a
,
Microscopic
image
of microfabricated
array
with
9 piezoelectric
elements.
Scale,
200
μm.
b
, Laser
doppler
vibrometer
(LDV)
response
of 9 elements
under
a periodic
chirp
excitation
in the
0 - 2 MHz
range,
showing
the
displacement
of the
unique
elements
is in phase.
c
, Average
frequency
response of
the
array
elements,
showing
resonant
frequency close
to
500 kHz.
10
Supplementary
Figure
4 | P-E hysteresis
loop
and
piezoelectric
displacement
butterfly
loop
of KNN
measured
on
donor
wafer.
11
Supplementary
Figure
5 | Shift
in resonance
frequency
in different
mediums
(air,
water
and
0.6%
agarose
gel).
Negligible
shift
in
frequency
is observed
in
water
versus
agarose
gel
medium.
N = 4.
12
Supplementary
Figure
6 | Effect of
cavity size on
resonance
frequency
and displacement
of
device a,
Resonance frequency for
varied
cavity
diameters comparing
simulated
and
experimental
results.
b,
Measured
displacement
with
LDV of
devices with
varied
cavity
size
and applied
voltages
.
13
Supplementary
Figure
7 | Experimental
setup
for
displacement
measurement
using
laser
doppler
vibrometer
(LDV).
The
test
device
was
mounted
on
a 3- axis
stage
for
positioning
under
microscope
for
LDV measurement.
14
Supplementary
Figure
8 | Three
representative
stages
of membrane
vibration
upon
application
of
sinusoidal
signal
at 10V
(p-p) at
fundamental
resonance frequency.
15
Supplementary
Figure
9 | COMSOL
Multiphysics
simulation
of Si-based
and SU
-8- based
pMUT.
a
,
Displacement
vs frequency
of Si and
SU
-8- based
pMUT
showing
their
resonance
behavior.
b
, Simulated
acoustic
pressure
profile
showing a
spherical
pressure
distribution.