Content uploaded by Priya S. Balasubramanian
Author content
All content in this area was uploaded by Priya S. Balasubramanian on Jun 02, 2023
Content may be subject to copyright.
iScience
Article
GHz ultrasonic sensor for ionic content with high
sensitivity and localization
Priya S.
Balasubramanian,
Amit Lal
psb79@cornell.edu
Highlights
Apulse-echobulk
acoustic wave GHz
ultrasonic sensor for ionic
content is developed
GHz ultrasonic wavefronts
allow for high spatial and
temporal resolution
The sensor is a label-free
and sensitive method to
detect acoustic
impedance
Both ionic content and
dynamicionicfluxcanbe
measured as a function of
time
Balasubramanian & Lal,
iScience 26, 106907
June 16, 2023 ª2023 The
Authors.
https://doi.org/10.1016/
j.isci.2023.106907
ll
OPEN ACCESS
iScience
Article
GHz ultrasonic sensor for ionic content
with high sensitivity and localization
Priya S. Balasubramanian
1,2,
*and Amit Lal
1
SUMMARY
Sensing the ionic content of a solution at high spatial and temporal resolution and
sensitivity is a challenge in nanosensing. This paper describes a comprehensive
investigation of the possibility of GHz ultrasound acoustic impedance sensors
to sense the content of an ionic aqueous medium. At the 1.55 GHz ultrasonic fre-
quency used in this study, the micron-scale wavelength and the decay lengths in
liquid result in a highly localized sense volume with the added potential for high
temporal resolution and sensitivity. The amplitude of the back reflected pulse is
related to the acoustic impedance of the medium and a function of ionic species
concentration of the KCl, NaCl, and CaCl
2
solutions used in this study. A concen-
tration sensitivity as high as 1 mM and concentration detection range of 0 to 3 M
was achieved. These bulk acoustic wave pulse-echo acoustic impedance sensors
can also be used to record dynamic ionic flux.
INTRODUCTION
Developing high-sensitivity ionic content sensors for dynamic environments remains incredibly chal-
lenging. Rising perspectives in improving spatial resolution, temporal resolution, and detection sensitivity
have become significant aspects to consider in sensor development. Contemporary sensors utilize either
optical or electrical techniques, using microscale arrays or solubilized nanosensors to interrogate ionic so-
lutions.
1–6
However, both of these techniques suffer from electrode degradation over the long term, and
potential effects such as phototoxicity can hinder the broad use of these techniques. As such, new sensing
techniques for dynamic nanoscale phenomena such as ionic flux bring forth new possibilities to meet a
ubiquitous need.
Sensing the nanoscale dynamics of ions in flux is important in monitoring biochemical and physiological
parameters. Ionic flux is critical in signaling, conduction, and contractility in various cell and tissue line-
ages.
7–11
Furthermore, ionic and molecular components of sweat are essential for identifying the effects
of body stress.
12–14
Recently, wearable sensors have been in demand to monitor sweat, and similar systems
can be devised to monitor internal and external body fluids and secretions.
15–17
The transport of ions is also
crucialinmanymoderndevicessuchasbatteries,fuelcells,andthecorrosionofmetals.Someprocessing
techniques, such as electroplating, depend on ionic transport and can benefit from monitoring for process
optimization. Experimentally measuring the diffusion of ions in the presence of electrostatic fields can be
coupled with many recent theoretical models of this process further to enhance the various downstream
applications of this research.
18
Previous efforts in ionic sensing
While many approaches to measuring ionic behavior utilize electrical potentials, a non-electric measure-
ment can be useful. For example, electrodes can corrode over time and require additional processing
and electrical connections to the fluid. Molecular sensors suffer from the necessity to introduce a new nano-
particle into the system, and in biological applications, this leads to low lifetime due to metabolic effects
and delocalization. Furthermore, several of these molecular sensors depend on photonic stimulation for
readout and, as a result, introduce thermal and photonic degrading effects to the system of interest.
Nonelectrical and nonmolecular interventions hold promise in sensor longevity and minimal interference
with the sensing medium. Ultrasonic sensors can enable ionic measurement without having any electrodes
on the sensing side.
19–22
Furthermore, when designed to be acoustic impedance sensors, they can sense
changes in medium properties with high sensitivity and minimal ultrasonic energy transmission. As such,
these sensors offer minimal interference due to input energy and heightened longevity compared to
1
School of Electrical and
Computer Engineering,
Cornell University, Ithaca, NY
14853, USA
2
Lead contact
*Correspondence:
psb79@cornell.edu
https://doi.org/10.1016/j.isci.
2023.106907
iScience 26, 106907, June 16, 2023 ª2023 The Authors.
This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
1
ll
OPEN ACCESS
electrodes and molecular sensors, which suffer from the development of capacitive layers, corrosion, mo-
lecular degradation, and delocalization through various mechanisms. While molecular sensors may have
high specificity for a specific ion of interest, it is also possible to differentiate between ionic species with
ultrasonic sensors using various analog and computational techniques. These techniques are surveyed
with relevant work cited in the below section.
Ion-selective electrode (ISE) systems have a charge concentration nonlinearity. The well-known Nernst equa-
tion can model the ionic activity in a solution, which can be theoretically related to concentration only under
certain operating principles.
5,6,23–28
The practical limit of ionic concentration resolution is the Nernstian slope,
set at 59.2 mV per decade of current density at room temperature and 61 mV/dec at body temperature. As the
voltage readout from the electrode is absolute, this is the often-cited measurement sensitivity.
5,29
This mea-
surement is also subject to inherent electrode noise floor and signal recording limitations. Arrays of ISEs can
improve the theoretical Nernstian slope response limit.
5
Ion-selective membrane-based electrochemical
transistors can accomplish super-Nernstian sensitivity at upward of 80 mV/dec, with temporal resolution at
1 s. There is still area for improvement to accomplish sub-millisecond temporal resolutions.
29
Optical measurements of ionic content include spectroscopy-based measurements, where high sensitivity
is traded for real-time measurements.
1,2,30–32
Other measurement techniques involve using optical signals
from dynamically binding molecular dyes, which canprovide sensitivity at the nano to micromolar level and
further rapid sub-millisecond response times. The optical response with the dyes can be nonlinear across
the range of concentrations and potentially phototoxic, thus incompatible with long-term monitoring.
Furthermore, temporal and spatial resolutions are highly influenced by background noise, nonspecific
binding, and autofluorescence.
33
While optical techniques can exhibit high dynamic range, there is usually
high sensitivity to low concentration in the micromolar range but limitations to sensitivity past this. Binding
kinetics can also be measured with ISEs and optical techniques using chip-scale technologies. The accuracy
is subject to membrane and sensor lifetime, binding sterics, and the associated signal-to-noise.
34,35
Furthermore, these techniques are more challenging to translate to clinical and medical settings.
Dielectric constant detection through microwave sensors is another technique that allows for high-sensi-
tivity measurements using microscale and potentially array-based slot and metallic resonators.
3,4,36–43
Sub-millimolar resolutions have been recorded with differential mode complementary split-ring reso-
nator-based sensors at low concentration range.
37,44,45
However, the resolution provided by these micro-
wave resonators is not just dependent on the sensor design and analyte of interest but also the packaging
around the sense volume. This makes these sensors susceptible to variations in packaging and boundary
conditions over time. High-density, miniaturized, and channel-free microwave sensors are still in progress
toward meeting the growing needs of the field of nanosensing.
Acoustic impedance sensors measure the impedance (Z=rvac ), where the density and the velocity can be
time-varying and include the real and the complex components.
19,22,46–52
Acoustic impedance is sensitive
to the density, speed of sound, and attenuation characteristics of the medium of interest. Acoustic imped-
ance measurement sensors are useful in comparing different materials and dynamically changing material
properties.
19,21,46–49,53–55
Changes of 20% volume fraction and sub 100 mM concentration changes have
been recorded sensitivities for these techniques.
19,21,46–49,53–55
The acoustic impedance may vary due to
changes in density and elasticity within the sample. Acoustic impedance sensors vary by the specific acous-
tic waves (SAW, bulk, shear, etc.), the operating resonance frequencies, and the method of interfacing to
the sample through coupling layers. Acoustic impedance can be sensed with high resolution using analog
and digital processing techniques and has various biological and material applications. The applications of
acoustic impedance measurements include sensing dynamic ionic flux in biological and nonbiological me-
dia, sensing structural properties of tissue, scar tissue formation, electrolyte sensing, sensing pressure dif-
ferentials, material flaw characterization, polymerization characterization, and quantification of reaction
rates of microscale reactors.
56,57
Acoustic impedance sensors
The acoustic impedance of a lossy medium is represented as follows,
Z=rcð1jalÞ(Equation 1)
ll
OPEN ACCESS
2iScience 26, 106907, June 16, 2023
iScienc
e
Article
where ris the density of the interface medium, ais the attenuation parameter of the medium, lis the wave-
length of the acoustic wave, and cis the speed of sound in the medium of interest.
49
Apart from the effect
on the acoustic impedance itself, the attenuation of the medium also defines the decay length over which
the acoustic wave decays exponentially. The decay length determines the sensing depth of the liquid, as
the acoustic field does not penetrate significantly beyond 2–3 decay lengths. Acoustic impedance varies
ideally with a density as shown in the above equation; however, there are expected deviations due to
the potential dependency of aon density.
One variety of acoustic impedance sensors utilizes bulk acoustic waves (BAW sensors) with waves gener-
ated by thickness-shear or longitudinal bulk mode transducers.
56,57
Surface acoustic waves (SAW) may
be generated by interdigitated transducers and used for sensing applications.
58,59
SAW wave penetration
into surface liquids can be highly localized in-depth to an order of wavelength at the surface. SAW trans-
ducers can consume considerable surface area, and the interdigitated fingers typically are on the same side
as the interdigital transducers where the liquid is to be sensed. The liquids and the electrical connections
on one side can complicate the packaging as the electrical conne ctions have to be electrically isolated from
the fluidic interfaces. BAW devices can launch waves into the liquid, and the ener gyt ransmitted is governed
by the transmission and reflection characteristics determined by the acoustic impedance. The transducers
canbeplacedontheoppositesideofthesensorsurface.Inthiscase,thewavestravelthroughthesub-
strate to the fluid-silicon interface. The BAW wave penetration in the liquid is dependent on the acoustic
loss in the liquid. At kHz–MHz frequencies, the absorption depth is large, and the reflecting boundary con-
ditions away from the solid-liquid interface typically define the depth resolution. For example, if operating
at MHz frequencies, the depth of loss is tens of centimeters, and hence any package boundary condition
determines the sensing volume. At GHz frequencies, the absorption depth is small, enabling a more
straightforward measurement of the ultrasonic impedance without the concern of packaging determining
boundary conditions. Thus, wavefront confinement is a function of the frequency both axially and laterally,
defined by both diffraction limits of acoustic beams and medium attenuation properties.
GHz ultrasonic acoustic impedance sensors – Background and operating principle
GHz ultrahigh frequency bulk acoustic waves accomplish high confinement - tens of mmfor the axial reso-
lution and down to 1 mmin lateral resolution in sense volume in aqueous medium. Furthermore, the high-
frequency GHz regime allows for higher sample rates inherent in the high-frequency stimulus signal and
lower in-band noise. At GHz frequencies, ultrasonic waves in water have wavelengths in the 1–5 mmin fluids
and decay within 25 mmin liquids, thus lending to higher spatial resolutions than lower frequency sonic
wavefronts. Given the tens of micron scale of attenuation depth, 1.5 GHz frequency scales well to a sense
volume not saturated with signal solely from double-layer formation at the silicon-liquid interface. We pre-
sent in this study one of the first chip-scale CMOS-compatible GHz ultrasonic acoustic impedance sensors
for ionic content. There have been approaches in which GHz US has been used for mass sensing. Here,
added molecules on a surface result in a change in the resonance frequency. This mass sensing technique,
used in the resonant mode of the device, is not compatible with pulse transmit-receive readout.
60–62
Pico-
second ultrasonics has been used to investigate the propagation of ultrahi gh frequency ultrasonic waves. In
this approach, a laser pulse is used as a pump, and an optical probe beam can be used to measure the dis-
placements and drive the ultrasonic motion.
63
This method is not amenable to miniaturized, implantable,
or handheld devices due to the size and cost of lasers and the space typically needed for optical beam
manipulation. Furthermore, it is more challenging to configure in a flexible array format that lends more
effectively to sensing systems. In addition, compared to high-sensitivity microwave resonator-based sen-
sors, GHz ultrasonic chip-scale transmit-receive sensors used in this study for acoustic impedance measure-
ments do not require restricted chambers for the interface with the analyte. CMOS-integrated GHz trans-
ducers can create compact devices, enabling low cost and miniature implementation.
64–68
Furthermore,
this technique is label-free, and does not require surface functionalization to capture the analyte of interest
as it simply relies on the acoustic impedance measurement. Representative of this technology, this paper
reports measurements with GHz ultrasonic bulk waves generated by a 70 mmsquare aluminum nitride, or
AlN thin-film transducers fabricated on a silicon wafer. The wavefronts generated by thetransducers on one
side travel through the silicon substrate and undergo a reflection from the opposing surface, which is in
contact with a liquid medium of interest. The wave packet reflects and travels through the silicon substrate,
and is then sensed by the same transducer that actuated it. The sensed wave packet intensity is propor-
tional to the ultrasonic reflection coefficient, attenuation and diffraction losses within the silicon transmis-
sion medium, and the electromechanical coupling coefficient. The reduction in the reflection coefficient is
ll
OPEN ACCESS
iScience 26, 106907, June 16, 2023 3
iScienc
e
Article
effectively a result of the amount of ultrasonic wave energy that entered the specimen, which is directly
related to ionic content and species. The reflection coefficient between a pulse arriving from material 1 (sil-
icon) into material 2 (liquid) is
G=Z2Z1
Z2+Z1
(Equation 2)
where Z
1
is the impedance of the silicon substrate, while Z
2
is the impedance of the fluidic medium of in-
terest. Due to the GHz frequency of the ultrasonic waves, the maximum resolution is on the order of a wave-
length in the medium of interest using a phased array operation. The carrier frequency of 1.5 GHz has
approximately 9 and 1 mm wavelength in silicon and water. In this work, the beam width defines the lateral
resolution attainable as a function of the ultrasonic carrier frequency. Furthermore, GHz wave packets
decayswiftlyintheaxialdirection.Theattenuationisgovernedbyanexponentialdrop-offinsignalinten-
sity as
I=I0eaDx(Equation 3)
where I0is the intensity at the silicon/water interface and ais a frequency-dependent loss factor. The
following power law may approximate the attenuation dependence on frequency,
aðfÞ=a0+a1jfjg(Equation 4)
where a
0
is often set to 0, and 0 <g%2. These parameters are measured experimentally, and the expres-
sion derived following the theory of time causal Hooke’s Law under power law absorption behavior of the
wave-medium interaction.
69–71
As described in the study by Szabo and Wu,
71
this model is valid for fre-
quencies up to 10
12
Hz. For GHz frequencies, acoustic attenuation is approximated in aqueous media,
as used in this paper, to be 0.134 dB
mmGHz
2and is quadratically related to frequency for the frequency range
of interest. At 1.55 GHz, the decay length for 3-decibel loss length is 10.7 microns. This provides an axial
localization within tens of microns. As a point of reference, the use of BAW acoustic impedance sensors
for sensing changes in ionic content has most often cited a tested resolution of 1%–5% by weight in min-
imum, which translates to 100s of mM rather than 1 mM in resolution tested.
35,72–79
Thus, this study cites
sensitivities that far exceed most all reported in the literature for a similar sensing modality.
This study shows that using the principles of acoustic impedance sensing, as governed by Equations 1 and
2, one can sense the temporal activities of dynamically changing medium in addition to the static counter-
parts through GHz bulk acoustic wave transmit-receive reflectometry. The thickness mode ultrasonic trans-
mit-receive transducer operating at GHz frequencies provides high localization of sensing and sensitivity.
The advantage of higher frequency ultrasonics is improved temporal resolution and heightened lateral and
axial resolutions. This paper is the first study to use a chip-integrated array of GHz ultrasound transducers
that can sense aqueous, time-varying media.
This study describes the techniques used to characterize and operate the transducer. The experimental
design was used to test the transducer’s sensitivity to concentration in both static and dynamic settings.
Information regarding sensor spatial resolution is detailed. The results describe the sensitivity todetection
for static and dynamic concentrations and further the influence of heat on dissolution.
RESULTS
Transducer design and characterization
Devices were fabricated at the Institute of Microelectronics, IME A*STAR, as part of the I-ARPA TIC pro-
gram. The transducers used in this study are identical to previously published devices.
66,80,81
Identically
fabricated and aligned transducers are on both sides of the 750 micron thick silicon wafer. Like experi-
mental setups in the study by Kuo et al., Abdelmejeed et al., and Balasubramanian et al.,
66,68,80–82
the
identically fabricated surfaces allow for proper alignment of medium and transducer, especially in the
experiments of salt dissolution described in the following sections.
The device layout and design is detailed in Figure 1. Both surfaces are identically aligned and fabricated to
allow for the alignment of the transmit-receive pulse. In addition to the PECVD-deposited SiO
2
,thethick-
ness mode resonance of the AlN films lends to a resonance frequency with an approximate center fre-
quency of 1.5 GHz. For the sensing experiments, the first echo’s return signal is of primary interest in
ll
OPEN ACCESS
4iScience 26, 106907, June 16, 2023
iScienc
e
Article
measuring the medium’s ultrasonic properties. The magnitude of the first echo return signal is used for the
analysis. The chip-scale device is depicted in Figure 2, at two different magnifications.
Transducer motion in air is measured using a Polytec UHF-120 and a topical motion scan over the backside
displays a diffraction pattern, as shown in Figure 3.
Frequency response of transducer
Data collected from the Polytec UHF interferometer under continuous-wave input to the transducer using
input single-frequency waves swept across the depicted frequency range, and the transducer’s pulse-echo
response using the first echo signal is shown in Figure 4. The loss of energy into the silicon substrate
through diffraction lowers the resonator’s quality factor, leading to a response over a wider bandwidth.
Figure 1. Device cross-section
Device rationale and layout with the liquid medium is shown.
ll
OPEN ACCESS
iScience 26, 106907, June 16, 2023 5
iScienc
e
Article
A2mmthickness AlN results in a thickness mode resonance, which could be seen in the KLM model as an
open circuit resonance frequency defined by f0=nsALN =2tALN,wheretALN is the thickness of the AlN trans-
ducer. Models of the added load impedance of the silicon backing, molybdenum electrode, and oxide
layers lower the actual resonance frequency and the quality factor, increasing the effective bandwidth of
the transducer. The actual frequency response is complex, and its full description is beyond the scope
of this paper. A more accurate acquisition of frequency response would require an automatic electrical
impedance matching network on an integrated chip, including the GHz transducer, as the electric imped-
ance of the device varies as a function of frequency. Several example systems are available in the literature
and are needed to provide the most accurate devi ce characterizations.
22,52,83,84
In this paper, given that the
operating frequency range is broad, an optimal sense frequency was chosen based on maximizing the
pulse-echo return for operation.
Analog signal acquisition of ultrasonic reflections
The signal processed and used to collect information on the acoustic impedance of the medium is the first
echo of the pulsed transmit-receive echoes. The echoes are shown in Figure S1. Since this echo represents
a discontinuous signal and the amplitude of the echo is what is related to the acoustic impedance of the
medium, signal processing in the analog domain was employed to get heightened sensitivity and contin-
uous sensing of the medium of interest by extracting the amplitude of the first echo.
The analog processing scheme is depicted in Figure 5. A qualitative description of the signal trajectory
through the analog processing technique is described in this paragraph. The RF switch allows the contin-
uous wave GHz signal to be sent to the transducer input using a pulse mode at 1 MHz. The signal is then
returned from the transducer as a pulse-echo return and amplified using the next amplifier. This increases
the signal-to-noise ratio and brings the output to the range of voltages for the envelope detector. The en-
velope detector generates an envelope of the RF signal through a low-pass filter. The sample-and-hold cir-
cuitry obtains the value of the signal at the first return amplitude where the signal is triggered. This signal
valueisusedtochargeacapacitor,whichthenholdsthisvaluethroughatriggersignalforagivenamount
of time until the next pulse is triggered. The output is then fed into a differential amplifier, which is refer-
enced to a control threshold voltage, set to be the signal value for the reference aqueous medium with no
ionic content supplemented, as shown in Figure 5. The final box is t he sampling from the oscilloscope of the
signal post-amplification, which shows the voltage related to the concentration through the acoustic
impedance of the medium of interest. More details are provided in the STAR Methods section.
Proposed mechanism of sensing
This sensor operates through the use of a pulse-echo signal that is generated by the backside transducer,
travels through the silicon as shown in Figure 1, and then hits the chip-liquid boundary, where the wave
partially transmits and partially reflects. The amplitude of the reflected signal is then sensed by the same
transmit transducer, with the first echo used to preserve high signal-to-noise. Alternative configurations,
including a different transmit and receive transducer, are possible, with continuous wave and pulse
Figure 2. Device image
Chip-scale device used for sensing photographed both at visible scale lengths (left) and high resolution through reflected
light microscopy (right).
ll
OPEN ACCESS
6iScience 26, 106907, June 16, 2023
iScienc
e
Article
mode operation. The amplitude of the returned signal is related to the transmission-reflection coefficient,
and thus related to the analyte properties such as density and concentration. As such, we can predict
changinganalytepropertieswithchangestotheamplitudeofthereturnsignal.Therawpulse-echosignal
is shown in Figure S1.
Input electrical impedance change due to interface medium is not a substantial contribution
Thematerialonthetransducer’sbacksideisthatair,water,orsaltsolutionmightchangetheeffectiveelec-
trical impedance for the AlN transducer or change the frequency response. The conductivity of the silicon
wafer separating the front and backside of the chip could allow for the backing to influence the transducer
resonance and frequency response. The first and second return experiments in the preceding section have
differing sensitivities. The second return signal having a heightened sensitivity suggests that the transmis-
sion of the acoustic wave into the interface medium is dominant in the signal changes observed. However,
as one cannot rule out the potential of the frequency response itself being altered or possible diffraction-
associated power loss as the wave packet travels through the silicon, the RF coupled input signal was also
tested and compared for air backed, deionized water backed, and 3 M NaCl backed GHz ultrasonic arrays.
The RF coupled signal amplitude was 136.26 G10.13 mV, 136.78 G10.01 mV, and 135.95 G10.01 mV,
respectively, with no statistically significant difference. Thus, it is assumed that the impedance is not varying
greatly as a function of the interface medium. In the future, this phenomenon canbe characterized precisely
by electrical impedance matching networks.
High concentration range ionic sensing
The concentrations tested span an extensive range, almost to the solubility limit in deionized water, from
0 M to up to 3 M. The average return signal and deviation is recorded across concentrations in Figure 6 with
concentrations resolvable as plotted in the figure. Calling on the relationship expressed in Equations 1 and
2, there are still substantial deviations from the theory to be considered. This can be attributed to the
following factors. First, there is a formation of an ionic layer at the surface interface of the chip and solution,
which substantially complicates the calculation of acoustic impedance and also creates concentration de-
pendencies of the aterm. Furthermore, as shown in Figure 6 in the theoretical data and the right side of the
figure, the relationship between concentration and density is not as expected across the entire range of
concentrations.
77,85,86,87
These effects, particularly the ionic layer formation, will substantially influence
the system behavior and attributed to the observed curves shown in Figure 6. Of these quantities that
have been investigated in the literature most extensively is the nonlinearity of the density and concentra-
tion relationship.
49,85,86
In Figure 5, the subpanels to the figure’s right depict the density and concentration
relationship for the entire concentration range for the ionic species of interest. Notably, the saturation of
density at higher concentrations follows with the results of the acoustic impedance sensor, which is further
Figure 3. Transducer characterization for a 70 mmsquare AlN transducer
The diffraction pattern obtained from Polytec UHF-120 scanning of the amplitude excited on the backside is shown.
ll
OPEN ACCESS
iScience 26, 106907, June 16, 2023 7
iScienc
e
Article
compounded by the other nonlinear effects described previously. Various features such as theoretical de-
viations of the sensing medium and lower signal-to-noise due to with more acoustic power transmission
into the sensing medium will influence the signal-to-noise across concentration values. Furthermore, it is
shown in the literature that the acoustic impedance of an aqueous-solid interface is a function of the dis-
tance from the interface, as the density and solvent properties vary due to interfacial interactions. This ef-
fect is also characterized in the study by Mante et al. (2014) for hydrophilic surfaces.
65
Thereisanexpecta-
tion that surfaces ranging from highly hydrophobic to highly hydrophilic will have varying changes in the
apparent density and acoustic impedance of solvent at the interface of interest.
Concentration sensing at 1 mM sensitivity
Figure 7 shows sodium chloride is prepared at various concentrations ranging from 0 to 5 mM, using a cali-
bratedhigh-precisionsmallvolumepipettor.TheresultingcurveisshowninFigure 7. A 1 mM resolution is
achievable across most of the range, at a significance level of a= 0.05. Five acquisitions were recorded per
concentration, with at least 200 returns averaged per acquisition. Standard deviation is depicted as error
bars in the Figure 1. A 1 mM sensitivity is a significant landmarkforionicresolutionsensingthatweaim
to utilize for biological and other sensing applications.
19,88
This was achieved here across most of the sam-
ples, though it remains to be determined whether the large-range concentration sensing will maintain the
resolution throughout. At the saturation level of 3M for NaCl, data suggest that a 10 mM concentration
sensitivity is achievable; further research will provide details and improvements to the signal processing
to obtain higher resolution.
Ionic compound dissolution rate
Sensing the rate of change of concentration is introduced to show that a reasonable temporal resolution
maybeachievedwithtimecoursedatathatmatchestheexpectedtheoreticalbehaviorofthesystem.The
rate of solubility, or dissolution rate, is given theoretically by dC
dt =DA
LV ðCSCÞ, where C is concentration,
C
s
is solution saturation concentration, V is volume, A is the surface area of solute, L is length or radius to the
measurement point, and D is diffusion coefficient.
89
The solution for this equation is C=Cs1eDA
LV t:
While model fitting poses added complexities due to the voltage-concentration nonlinearities that are
beyond the scope of this study, the theoretical dissolution properties qualitatively agree with the results
obtained by the sensing system. The ionic salts used were powdered into fine granules before addition
at once into 100 mLdeionized water. However, differences in the addition of salt will influence A and
thus the dissolution rate. The salt introduction was performed at the same spatial location, which is
Figure 4. Frequency response of device
A graphical depiction of the displacement amplitude across the frequency range of interest of the transducer is shown.
Error bars depict standard deviation.
ll
OPEN ACCESS
8iScience 26, 106907, June 16, 2023
iScienc
e
Article
0.5 cm from the active transducer center displacement onthebacksideofthechipafterthesaltisgroundto
small particulates using a mortar and pestle. The introduction of the salt into the solvent was immediate
and performed using a 6.3-inch tapered edge scoop-type chemical-grade spatula, gated by a flat end,
tapered spatula, and the salt is released within the solvent with an immediate release of the gated spatula.
The results of the dissolution of each of the ionic solutes tested are shown in Figure 8, for a total of 20 s.
Further time points from data not shown suggest that a steady state is accomplished following the first
10 s of solute introduction.
Effect of temperature on dissolution rate
To further illustrate the sensing capacity of this system, the transducer is contacted with a thermal element
that heats the sample and is calibrated with a thermocouple. Three different temperatures were tested,
room temperature 24C, 55C, and 76C. The dissolution time course is plotted in Figure 9. The gradient
of the dissolution rate is shown for a part of the process in Figure 9 intherightsubplot.Itcanbeseenthat,
as the temperature is increased, the dissolution rate is also increased through the influence of the dissolu-
tion coefficient and saturation concentration (solubility) changing as a result of higher temperatures of the
solution. To ascertain that heating is not causing evaporation that will lead to concentration changes in the
sample volume, the highest temperature assessed in the experimental section is applied for 100 s to both
deionized water and NaCl solution. Neither sample experienced significant changes in the signal return.
Thus, when both deionized water and ionic solutions are subject to 76C heating, there is no change in
signal return, indicating that evaporation is not a problem within the time frame data are acquired.
DISCUSSION
This study is the first comprehensive characterization of chip-scale GHz ultrasonic acoustic impedance sen-
sors used to understand and sense constant and time-varying properties of ionic aqueous solutions.
Furthermore, it is one of the first BAW pulse-echo ultrasonic impedance sensors to detect dynamic ionic
flux, in addition with one of the highest reported sensitivities at 1 mM. The use of BAW acoustic impedance
sensors for ionic content sensing has concentration sensitivities ranging between 1% and 5%, which trans-
lates to 100s of mM depending on the particular salt of interest.
35,72–79
Our current results not only sense
dynamic flux but also report a record resolution of 1 mM, one of the lowest reported sensitivities in the liter-
ature. The nonlinearity across the concentration range of the pulse return signal is derived from the
Figure 5. Analog processing of signal
The circuit diagram used to process the signal in the analog domain prior to sampling and storing data in the oscilloscope is depicted, with each component
outlined and identified.
ll
OPEN ACCESS
iScience 26, 106907, June 16, 2023 9
iScienc
e
Article
medium properties and the solid-liquid interface.
85,86
Furthermore, this speaks for the potential non-ho-
mogeneity of the concentration gradient from the surface of the chip outward, primarily well characterized
in studies regarding charges, gradients, and ion distributions of solutions at the solid-liquid interface.
65
Figure 6. Concentration range testing
The concentration of ionic solution for three different ionic species is tested across a broad concentration range up to values close to the room temperature
solubility limit. Error bars depict standard deviation. Data from transducers shown on the left, and theoretical data for density and concentration relationship
shown on the right from cited research.
77,85,86,87
ll
OPEN ACCESS
10 iScience 26, 106907, June 16, 2023
iScienc
e
Article
Evidence is shown that the nonlinearity can be at least partly attributed to the saturation of the density-con-
centration relation at high concentrations from theoretical values in Figure 6 (right side panels). This fasci-
nating phenomenon can be better characterized with ionic solutions once a broad range of transducers
with differing ranges of resonance frequency are developed to distinguish the effects of density-attenua-
tion-concentration and surface interface alterations in addition to ionic double layers in deviations of
acoustic impedance measures from theoretical models. Furthermore, the shear stress on the surface and
vortical mixing could disrupt structured interface effects and allow for the decoupling of the system nonlin-
earity from the spatial inhomogeneity that could be contributing to this phenomenon. Attenuation charac-
terization will be critical in the endeavor to fit results to a theoretical model. There is also the potential
involvement of the medium influencing the wavefront i tself, which could alter the wave diffraction and inter-
ference, thus leading to alterations in the response that could induce further unexpected nonlinearities.
Other computational models, experimental evidence, and theoretical simulations will provide more infor-
mation toward deciphering the nonlinear effects. Noise propagation techniques will allow for the deriva-
tion of signal-to-noise as a function of concentration and analog processing components. Notably, the
analog processing and system itself could benefit from alternative setups, such as interferometry, through
a phase-locked secondary transducer or IQ demodulation setup as in the study by Abdelmejeed et al.
64
It is
also worthwhile to note that the accompanying analog processing could be utilized in any differ ential mode
power sensing application. We utilize the amplitude of the first return with respect to a DC reference in dif-
ferential mode, but this can also be extended to use in any differential mode amplitude-based signal
sensing modality. This technique requires either power, amplitude, or a similar quantity to be derived
from the original signal rather than operating on a differential mode on the actual resonance frequency
of the signal. This method is more subject to error as a result of phase noise and jitter, especially in the
GHz regime.
This study is an exciting insight into the scaling and physics of GHz ultrasound and its interactions with ionic
solutions. It also has many biosensing and chemical engineering applications, ranging from measuring sin-
gle-cell and ion channel ionic flux to microreactor engineering, to materials characterization. This paper’s
results can guide the development of compact CMOS-integrated GHz ultrasonic devices for applications in
wearables, batteries, and new biochemical experiments where electrodes directly into the liquid are not
possible. The unique localization of GHz ultrasound, as seen in Figure 3, allows for a spatial resolution un-
met by other lower frequency ultrasonic devices. Furthermore, the GHz frequency itself lends to a high tem-
poral resolution capability. This study provides insight into the next generation of acoustic sensors.
Limitations of the study
One setback of the dissolution experiments is the inability to introduce the ionic solid to the solvent as a
delta function in time. In experimental settings, the solute introduction is not instantaneous, and the rate of
Return Voltage in mV
8
6
4
2
0
Concentration in mM
012345
Figure 7. Sensor sensitivity characterization
Transducer return signal versus ionic concentration of NaCl at 1-mM resolution. Error bars depict standard deviation.
ll
OPEN ACCESS
iScience 26, 106907, June 16, 2023 11
iScienc
e
Article
addition is challenging to control precisely. Microscale release reservoirs are needed for future
model-based evaluation, which will be a fabrication challenge given the need to compound electronics
and circuitry necessitated for controlled release in addition to the ultrasonic input on one chip. Further-
more, the surface area of the introduced solute is challenging to control and ascertain. Precision delivery
and characterizations of solutes using microfluidics and automated XYZ stages can be developed for future
experiments. These modifications will also inform future thermal experiments to investigate the thermal
effects on the dissolution of these ionic solutes in an aqueous solution. Another potential area of improve-
ment is in the analog processing circuitry. There are deviations in the return signal across concentrations
and for each concentration tested due to the drift of DC reference sources into the differential amplifier
and jitter from other pulse and trigger signals that influence gain, stability, and variability in results. This
can be improved through modulation or interferometry-based techniques that allow for more accuracy
and stability. Furthermore, integrated circuitry and sensor layouts of these chip-scale devices will signifi-
cantly enhance signal-to-noise.
Conclusions
This study presents the measurements of BAW GHz ultrasound pulse-echo amplitude changes to sense
ionic concentration changes in a liquid medium on the sensor surface. Ionic content sensitivity down to
1 mM is achievable with this system, along with a lateral spatial resolution of less than 50 mmand axial res-
olution of less than 30 mm, limited by the diffraction of the transducer through silicon and the design of the
transducer. Further design changes, beam steering, and focusing will improve the lateral resolution to the
plausible 1 mmdiffraction limit. Temporal resolution necessary to detect the dissolution of ionic species
and the changes in dissolution rate due to temperature increase is achievable under the current system
design. A 1 mM sensitivity is recorded for this BAW pulse-echo ultrasonic acoustic impedance sensor
format. Future directions include automated analog circuit-based pulse-echo processing and fluid sam-
pling device fabrication by integrating the transducers with CMOS as in the study by Abdelmejeed
et al.,
68
theory validation of GHz ultrasonic attenuation across different medium properties, and further un-
derstanding of sensing nonlinearities and system deviations. Applications include sensing dynamic ionic
flux as relevant to neural action potential, transport in tissue and vascular flow, biomedical sensor develop-
ment, materials characterization, and chemical reaction sensing.
STAR+METHODS
Detailed methods are provided in the online version of this paper and include the following:
dKEY RESOURCES TABLE
dRESOURCE AVAILABILITY
Figure 8. Dissolution sensing
Ionic solute dissolution in aqueous solution (deionizied water) is tested and recorded for three different ionic compounds.
ll
OPEN ACCESS
12 iScience 26, 106907, June 16, 2023
iScienc
e
Article
BLead contact
BMaterials availability
BData and code availability
dMETHOD DETAILS
dQUANTIFICATION AND STATISTICAL ANALYSIS
SUPPLEMENTAL INFORMATION
Supplemental information can be found online at https://doi.org/10.1016/j.isci.2023.106907.
ACKNOWLEDGMENTS
Equipment and materials were previously purchased and utilized from support by the National Science
Foundation under Grant No. 1744271 and Trusted Integrated Chips Grant. The authors acknowledge
the expertise and contributions of the SonicMEMS lab members, especially Dr. Justin Kuo and Dr. Mam-
douh Abdelmejeed, in the pioneering work toward GHz US MEMS t ransducers and accompanying circuitry.
The authors would like to acknowledge and thank the Institute of Microelectronics in Singapore (IME
Singapore A*STAR) for fabrication of chips in collaboration with SonicMEMS members. The authors would
like to thank Dr. Edwin Kan for shared lab space and equipment. The authors thank and acknowledge the
shared funding from Dr. Ankur Singh and Dr. Chris Xu. The authors would like to thank Kasra Kakavand, Jack
Danieli, Eric Lawrence, and many others from the Applications Team at Polytec GmbH for arrangingfor Pol-
ytec equipment performed at the Medically Advanced Devices Lab at the Center for Medical Devices in the
Department of Mechanical and Aerospace Engineering, University of California San Diego thanks to the
generous accommodations of Dr. James Friend, Gopesh Tilvawala, William Connacher, Ann Huang, Aditya
Vasan, Eric Lawrence, Jiyang Mei, Shuai Zhang, and Naiqing Zhang.
AUTHOR CONTRIBUTIONS
P.S.B. designed the study, conducted the experiments, collected and processed data, and wrote the pa-
per. A.L. provided supervision and advisement, designed the study, and wrote and edited the paper.
DECLARATION OF INTERESTS
A.L. is the cofounder of Geegah, which is commercializing GHz ultrasonic transducers and imagers.
Various patents related to GHz ultrasound transducers and imagers have been submitted and are at
differing stages of patenting process across the world.
Figure 9. Sensing dissolution at varying temperatures
A thermal heating element is used to interface with the transducer, changes in temperature of the sample volume are induced. Dissolution rate is changed
for NaCl in water (left) with the rate of change of each curve shown for select time points (right) over a limited time range.
ll
OPEN ACCESS
iScience 26, 106907, June 16, 2023 13
iScienc
e
Article
INCLUSION AND DIVERSITY
We support inclusive, diverse, and equitable conduct of research.
Received: December 29, 2022
Revised: May 2, 2023
Accepted: May 12, 2023
Published: May 19, 2023
REFERENCES
1. Yi, Z., Liu, J., Liu, B., Guo, H., Wu, Q., Shi, J.,
and He, X. (2022). Optical microfiber sensor
for detection of Ni2+ions based on ion
imprinting technology. Analyst 147, 358–365.
https://doi.org/10.1039/d1an01982a.
2. Miao, Z., Chu, Y., Wang, L., Zhu, W., and
Wang, D. (2022). Nonlinear optical and ion
sensor properties of novel molecules
conjugated by click chemistry. Polymers 14,
1516–1610. https://doi.org/10.3390/
polym14081516.
3. Rochman, J., Xie, T., Bartholomew, J.G.,
Schwab, K.C., and Faraon, A. (2023).
Microwave-to-optical transduction with
erbium ions coupled to planar photonic and
superconducting resonators. Nat. Commun.
14, 1153. https://doi.org/10.1038/s41467-
023-36799-0.
4. Mansour, R., Rioual, S., Lescop, B., Talbot, P.,
Abboud, M., Farah, W., and Tanne
´, G. (2020).
Development of a resonant microwave
sensor for sediment density characterization.
Sensors 20, 1058. https://doi.org/10.3390/
s20041058.
5. Wang, C., Yuan, H., Duan, Z., and Xiao, D.
(2017). Integrated multi-ISE arrays with
improved sensitivity, accuracy and precision.
Sci. Rep. 7, 44771–44810. https://doi.org/10.
1038/srep44771.
6. Han, T., Mattinen, U., and Bobacka, J. (2019).
Improving the sensitivity of solid-contact ion-
selective electrodes by using coulometric
signal transduction. ACS Sens. 4, 900–906.
https://doi.org/10.1021/acssensors.8b01649.
7. Pedersen, T.H., Nielsen, O.B., Lamb, G.D.,
and Stephenson, D.G. (2004). Intracellular
acidosis enhances the excitability of working
muscle. Science 305, 1144–1147. https://doi.
org/10.1126/science.1101141.
8. Cosofret, V.V., ErdOsy, M., Johnson, T.A.,
Buck, R.P., Ash, R.B., and Neuman, M.R.
(1995). Microfabricated sensor arrays
sensitive to pH and K+ for ionic distribution
measurements in the beating heart. Anal.
Chem. 67, 1647–1653. https://doi.org/10.
1021/ac00106a001.
9. Zhang, M., Tang, Z., Liu, X., and Van der
Spiegel, J. (2020). Electronic neural
interfaces. Nat. Electron. 3, 191–200. https://
doi.org/10.1038/s41928-020-0390-3.
10. Rai, P., Oh, S., Shyamkumar, P., Ramasamy,
M., Harbaugh, R.E., and Varadan, V.K. (2014).
Nano- bio- textile sensors with mobile
wireless platform for wearable health
monitoring of neurological and
cardiovascular disorders. J. Electrochem.
Soc. 161, B3116–B3150. https://doi.org/10.
1149/2.012402jes.
11. Sugawara, M., Kojima, K., Sazawa, H., and
Umezawa, Y. (1987). Ion-channel sensors.
Anal. Chem. 59, 2842–2846.
12. Buono, M.J., Ball, K.D., and Kolkhorst, F.W.
(2007). Sodium ion concentration vs. sweat
rate relationship in humans. J. Appl. Physiol.
103, 990–994. https://doi.org/10.1152/
japplphysiol.00015.2007.
13. Montain, S.J., Cheuvront, S.N., and Lukaski,
H.C. (2007). Sweat mineral-element
responses during 7 h of exercise-heat stress.
Int. J. Sport Nutr. Exerc. Metabol. 17,
574–582. https://doi.org/10.1123/ijsnem.17.
6.574.
14. Baker, L.B., Ungaro, C.T., Sopen
˜a, B.C.,
Nuccio, R.P., Reimel, A.J., Carter, J.M.,
Stofan, J.R., and Barnes, K.A. (2018). Body
map of regional vs. whole body sweating rate
and sweat electrolyte concentrations in men
and women during moderate exercise-heat
stress. J. Appl. Physiol. 124, 1304–1318.
https://doi.org/10.1152/japplphysiol.
00867.2017.
15. Jose, M., Oudebrouckx, G., Bormans, S.,
Veske, P., Thoelen, R., and Deferme, W.
(2021). Monitoring body fluids in textiles:
combining impedance and thermal principles
in a printed, wearable, and washable sensor.
ACS Sens. 6, 896–907. https://doi.org/10.
1021/acssensors.0c02037.
16. Chung, M., Fortunato, G., and Radacsi, N.
(2019). Wearable flexible sweat sensors for
healthcare monitoring: a review. J. R. Soc.
Interface 16, 20190217. https://doi.org/10.
1098/rsif.2019.0217.
17. Bandodkar, A.J., Ghaffari, R., and Rogers, J.A.
(2020). Don’t sweat it: the quest for wearable
stress sensors. Matter 2, 795–797. https://doi.
org/10.1016/j.matt.2020.03.004.
18. Del Rio
´, J.A., and Whitaker, S. (2016).
Diffusion of charged species in liquids. Sci.
Rep. 6, 1–11. https://doi.org/10.1038/
srep35211.
19. Antlinger, H., Clara, S., Beigelbeck, R.,
Cerimovic, S., Keplinger, F., and Jakoby, B.
(2012). Sensing the characteristic acoustic
impedance of a fluid utilizing acoustic
pressure waves. Sens. Actuators. A Phys. 186,
94–99. https://doi.org/10.1016/j.sna.2012.
02.050.
20. A
´lvarez-Arenas, T.G., and Camacho, J. (2019).
Air-coupled and resonant pulse-echo
ultrasonic technique. Sensors 19, 2221.
https://doi.org/10.3390/s19102221.
21. Kudryashov, E., Smyth, C., Duffy, G., and
Buckin, V. (2000). Ultrasonic high-resolution
longitudinal and shear wave measurements in
food colloids: monitoring of gelation
processes and detection of pathogens. Prog.
Colloid Polym. Sci. 115, 287–294. https://doi.
org/10.1007/3-540-46545-6_56.
22. Rathod, V.T. (2019). A review of electric
impedance matching techniques for
piezoelectric sensors. actuators and
transducers. Electron. 8, 169. https://doi.org/
10.3390/electronics8020169.
23. Fan, Y., Qian, X., Wang, X., Funk, T., Herman,
B., McCutcheon, J.R., and Li, B. (2022).
Enhancing long-term accuracy and durability
of wastewater monitoring using
electrosprayed ultra-thin solid-state ion
selective membrane sensors. J. Memb. Sci.
643, 119997. https://doi.org/10.1016/j.
memsci.2021.119997.
24. Huang, M.R., and Li, X.G. (2022). Highly
sensing and transducing materials for
potentiometric ion sensors with versatile
applicability. Prog. Mater. Sci. 125, 100885.
https://doi.org/10.1016/j.pmatsci.2021.
100885.
25. Pietrzak, K., Krstulovi
c, N., Bla
zeka, D., Car, J.,
Malinowski, S., and Wardak, C. (2022). Metal
oxide nanoparticles as solid contact in ion-
selective electrodes sensitive to potassium
ions. Talanta 243, 123335. https://doi.org/10.
1016/j.talanta.2022.123335.
26. Furukawa, K., Mizuno, K., Azuma, M.,
Yoshida, Y., Den, H., Iyoda, M., Nagao, S.,
and Tsujimori, Y. (2022). Reliability of an ion-
selective electrode as a simple diagnostic
tool for mastitis. J. Hum. Lactation 38,
262–269. https://doi.org/10.1177/
08903344221075050.
27. Briggs, A., and Kolosov, O. (2009). Acoustic
Microscopy, Second Edition (OUP Oxford).
28. Xu, W., Liao, X., Xu, W., Zhao, K., Yao, G., and
Wu, Q. (2023). Ion selective and water
resistant cellulose nanofiber/MXene
membrane enabled cycling Zn anode at high
currents. Adv. Energy Mater. 13, 1–11. https://
doi.org/10.1002/aenm.202300283.
29. Han, S., Yamamoto, S., Polyravas, A.G., and
Malliaras, G.G. (2020). Microfabricated ion-
selective transistors with fast and super-
nernstian response. Adv. Mater. 32,
e2004790–e2004798. https://doi.org/10.
1002/adma.202004790.
ll
OPEN ACCESS
14 iScience 26, 106907, June 16, 2023
iScienc
e
Article
30. Sasaki, Y., Lyu, X., Tang, W., Wu, H., and
Minami, T. (2022). Supramolecular optical
sensor arrays for on-site analytical devices.
J. Photochem. Photobiol. C Photochem. Rev.
51, 100475. https://doi.org/10.1016/j.
jphotochemrev.2021.100475.
31. Miele, E., Dose, W.M., Manyakin, I., Frosz,
M.H., Ruff, Z., De Volder, M.F.L., Grey, C.P.,
Baumberg, J.J., and Euser, T.G. (2022).
Hollow-core optical fibre sensors for
operando Raman spectroscopy investigation
of Li-ion battery liquid electrolytes. Nat.
Commun. 13, 1651–1710. https://doi.org/10.
1038/s41467-022-29330-4.
32. Robinson, K.J., Soda, Y., and Bakker, E.
(2022). Recent improvements to the
selectivity of extraction-based optical ion
sensors. Chem. Commun. 58, 4279–4287.
https://doi.org/10.1039/d1cc06636f.
33. Li, E.S., and Saha, M.S. (2021). Optimizing
calcium detection methods in animal
systems: a sandbox for synthetic biology.
Biomolecules 11, 343–435. https://doi.org/
10.3390/biom11030343.
34. Fuller, C.W., Padayatti, P.S., Abderrahim, H.,
Adamiak, L., Alagar, N.,
Ananthapadmanabhan, N., Baek, J., Chinni,
S., Choi, C., Delaney, K.J., et al. (2022).
Molecular electronics sensors on a scalable
semiconductor chip: a platform for single-
molecule measurement of binding kinetics
and enzyme activity. Proc. Natl. Acad. Sci.
USA 119, e2112812119. https://doi.org/10.
1073/pnas.2112812119.
35. Kimoto, A., and Kanazawa, A. (2009). A new
plane type sensor for concentration
measurement in liquid. Procedia Chem. 1,
140–143. https://doi.org/10.1016/j.proche.
2009.07.035.
36. Kazemi, N., Abdolrazzaghi, M., and Musilek,
P. (2021). Comparative analysis of machine
learning techniques for temperature
compensation in microwave sensors. IEEE
Trans. Microw. Theor. Tech. 69, 4223–4236.
https://doi.org/10.1109/TMTT.2021.3081119.
37. Abdolrazzaghi, M., Nayyeri, V., and Martin, F.
(2022). Techniques to improve the
performance of planar microwave sensors: a
review and recent developments. Sensors 22,
6946–6957. https://doi.org/10.3390/
s22186946.
38. Abdolrazzaghi, M., Daneshmand, M., and
Iyer, A.K. (2018). Strongly enhanced sensitivity
in planar microwave sensors based on
metamaterial coupling. IEEE Trans. Microw.
Theor. Tech. 66, 1843–1855. https://doi.org/
10.1109/TMTT.2018.2791942.
39. Moradpour, M., Hosseini, E., Jain, M.C.,
Narang, R., Tanguy, N., and Zarifi, M.H.
(2021). Patterned PEDOT:PSS-enabled
organic planar microwave resonator sensors.
Appl. Mater. Today 24, 101106. https://doi.
org/10.1016/j.apmt.2021.101106.
40. Liang, Y., Ma, M., Zhang, F., Liu, F., Lu, T., Liu,
Z., and Li, Y. (2021). Wireless microfluidic
sensor for metal ion detection in water. ACS
Omega 6, 9302–9309. https://doi.org/10.
1021/acsomega.1c00941.
41. Abbasi, Z., Niazi, H., Abdolrazzaghi, M.,
Chen, W., and Daneshmand, M. (2020).
Monitoring pH level using high-resolution
microwave sensor for mitigation of stress
corrosion cracking in steel pipelines. IEEE
Sensor. J. 20, 7033–7043. https://doi.org/10.
1109/JSEN.2020.2978086.
42. Banerjee, A., Azad, P., and Akhtar, M.J.
(2023). Detection of sodium ion imbalance in
human body fluids using an improved RF
sensor. IEEE J. Electromagn. RF Microw.
Med. Biol. 7, 73–81. https://doi.org/10.1109/
JERM.2022.3218156.
43. Baghelani, M., Abbasi, Z., Daneshmand, M.,
and Light, P.E. (2020). Non-invasive
continuous-time glucose monitoring system
using a chipless printable sensor based on
split ring microwave resonators. Sci. Rep. 10,
12980–13015. https://doi.org/10.1038/
s41598-020-69547-1.
44. Zhang, K., Amineh, R.K., Dong, Z., and
Nadler, D. (2019). A microwave sensor array
for water quality testing. In IEEE 20th Wirel.
Microw. Technol. Conf. WAMICON, 2019,
pp. 19–22. https://doi.org/10.1109/
WAMICON.2019.8765458.
45. Harrison, L., Ravan, M., Zhang, K., and
Amineh, R.K. (2021). Identification of
materials using a microwave sensor array and
machine learning. In Int. Appl. Comput.
Electromagn. Soc. Symp. ACES, 2021.
https://doi.org/10.1109/ACES53325.2021.
00166.
46. Zore
˛bski, E., Geppert-Rybczy
nska, M., and
Zore
˛bski, M. (2013). Acoustics as a tool for
better characterization of ionic liquids: a
comparative study of 1-alkyl-3-
methylimidazolium bis[(trifluoromethyl)
sulfonyl] imide room-temperature ionic
liquids. J. Phys. Chem. B 117, 3867–3876.
https://doi.org/10.1021/jp400662w.
47. Han, D.G., Seo, H.C., Cho, S., and Choi, J.W.
(2019). Measurements of normal incidence
reflection loss as a function of temperature at
the water-castor oil interface. Sensors 19,
3289. https://doi.org/10.3390/s19153289.
48. Praharaj, M.K. Mishra S. (2015). Study of
Acoustic and Thrmodynamic Parametersfor
Different Ratios of Aqueous Sodium
Chloride and Potassium Chloride Solution At
and About the Normal Human Body
Temperature. Int. J. Sci. Res. 2319–7064.
https://doi.org/10.13140/2.1.1798.1768.
49. Dukhin, A.S., and Goetz, P.J. (2017).
Characterization of Liquids, Dispersions,
Emulsions, and Porous Materials Using
Ultrasound (Elsevier).
50. Kuznetsova, I., Zaitsev, B., Krasnopolskaya, L.,
Teplykh, A., Semyonov, A., Avtonomova, A.,
Ziangirova, M., Smirnov, A., and Kolesov, V.
(2020). Influence of humidity on the acoustic
properties of mushroom mycelium films used
as sensitive layers for acoustic humidity
sensors. Sensors 20, 2711–2715. https://doi.
org/10.3390/s20092711.
51. Manwar, R., Saint-Martin, L., and Avanaki, K.
(2022). Couplants in acoustic biosensing
systems. Chemosensors 10.https://doi.org/
10.3390/chemosensors10050181.
52. Mohammadabadi, A., and Dugnani, R. (2019).
Design and evaluation of a novel low acoustic
impedance-based PZT transducer for
detecting the near-surface defects. Int. J.
Eng. Technol. Innov. 9, 196–211.
53. Nassar, G., Nongaillard, B., and Noel, Y.
(2001). Monitoring of milk gelation using a
low-frequency ultrasonic technique. J. Food
Eng. 48, 351–359. https://doi.org/10.1016/
S0260-8774(00)00178-3.
54. Camara, V.C., Laux, D., and Arnould, O.
(2010). Enhanced multiple ultrasonic shear
reflection method for the determination of
high frequency viscoelastic properties.
Ultrasonics 50, 710–715. https://doi.org/10.
1016/j.ultras.2010.02.007.
55. Frohn, A., Dick, H.B., Fritzen, C.P.,
Breitenbach, M., and Thiel, H.J. (2000).
Ultrasonic transmission in viscoelastic
substances. J. Cataract Refract. Surg. 26,
282–286. https://doi.org/10.1016/S0886-
3350(99)00361-2.
56. Ali, W.R., and Prasad, M. (2020). Piezoelectric
MEMS based acoustic sensors: a review.
Sensors Actuators, A Phys. 301, 111756.
https://doi.org/10.1016/j.sna.2019.111756.
57. Thompson, M., Kiplingt, A.L., and Duncan-
hewitts, W.C. (1991). Thickness-shear-mode
Acoustic Wave, 116.
58. Fox, C.G., and Alder, J.F. (1989). Surface
acoustic wave sensors for atmospheric gas
monitoring a review. Analyst 114, 997–1004.
https://doi.org/10.1039/AN9891400997.
59. Wu, T.T., Tang, H.T., Chen, Y.Y., and Liu, P.L.
(2005). Analysis and design of focused
interdigital transducers. IEEE Trans. Ultrason.
Ferroelectrics Freq. Control 52, 1384–1392.
https://doi.org/10.1109/TUFFC.2005.
1509798.
60. Rey-Mermet, S., Lanz, R., and Muralt, P.
(2006). Bulk acoustic wave resonator
operating at 8 GHz for gravimetric sensing of
organic films. Sensor. Actuator. B Chem. 114,
681–686. https://doi.org/10.1016/j.snb.2005.
04.047.
61. Rughoobur, G., Demiguel-Ramos, M.,
Escolano, J.M., Iborra, E., and Flewitt, A.J.
(2017). Gravimetric sensors operating at 1.1
GHz based on inclined c-axis ZnO grown on
textured Al electrodes. Sci. Rep. 7, 1367–
1369. https://doi.org/10.1038/s41598-017-
01545-2.
62. Adhikari, S., and Chowdhury, R. (2012).
Zeptogram sensing from gigahertz vibration:
graphene based nanosensor. Phys. E Low-
Dimensional Syst. Nanostructures 44, 1528–
1534. https://doi.org/10.1016/j.physe.2012.
03.021.
63. Matsuda, O., Larciprete, M.C., Li Voti, R., and
Wright, O.B. (2015). Fundamentals of
picosecond laser ultrasonics. Ultrasonics 56,
3–20. https://doi.org/10.1016/j.ultras.2014.
06.005.
64. Abdelmejeed, M., Davaji, B., and Lal, A.
(2019). Towards digitally controlled ultrasonic
IQ modulator. IEEE Int. Ultrason. Symp. IUS
ll
OPEN ACCESS
iScience 26, 106907, June 16, 2023 15
iScienc
e
Article
2019-Octob, 1967–1970. https://doi.org/10.
1109/ULTSYM.2019.8925880.
65. Mante, P.A., Chen, C.C., Wen, Y.C., Chen,
H.Y., Yang, S.C., Huang, Y.R., Chen, I.J.,
Chen, Y.W., Gusev, V., Chen, M.J., et al.
(2014). Probing hydrophilic interface of solid/
liquid-water by nanoultrasonics. Sci. Rep. 4,
6249–6256. https://doi.org/10.1038/
srep06249.
66. Kuo, J.C., Hoople, J.T., Abdelmejeed, M.,
Abdel-Moneum, M., and Lal, A. (2017). 64-
Pixel solid state CMOS compatible ultrasonic
fingerprint reader. Proc. - IEEE Int. Conf.
Micro Electro Mech. Syst. (MEMS), 9–12.
https://doi.org/10.1109/MEMSYS.2017.
7863326.
67. Hoople, J., Kuo, J., Abdel-Moneum, M., and
Lal, A. (2015). Chipscale GHz ultrasonic
channels for fingerprint scanning. IEEE Int.
Ultrason. Symp. IUS 2015, 3–6. https://doi.
org/10.1109/ULTSYM.2015.0027.
68. Abdelmejeed, M., Ravi, A., Liu, Y., Kuo, J.,
Sharma, J., Merugu, S., Singh, N., and Lal, A.
(2019). Monolithic 180nm CMOS controlled
GHz ultrasonic impedance sensing and
imaging. Tech. Dig. - Int. Electron Devices
Meet (IEDM 2019-Decem), pp. 2019–2022.
https://doi.org/10.1109/IEDM19573.2019.
8993623.
69. Chen, W., and Holm, S. (2003). Modified
Szabo’s wave equation models for lossy
media obeying frequency power law.
J. Acoust. Soc. Am. 114, 2570–2574. https://
doi.org/10.1121/1.1621392.
70. Szabo, T.L. (1994). Time domain wave
equations for lossy media obeying a
frequency power law. J. Acoust. Soc. Am. 96,
491–500. https://doi.org/10.1121/1.410434.
71. Szabo, T.L., and Wu, J. (2000). A model for
longitudinal and shear wave propagation in
viscoelastic media. J. Acoust. Soc. Am. 107,
2437–2446. https://doi.org/10.1121/1.
428630.
72. Puttmer, A., Hauptmann, P., and Henning, B.
(2000). Ultrasonic density sensor for liquids.
IEEE Trans. Ultrason. Ferroelectrics Freq.
Control 47, 85–92. https://doi.org/10.1109/
58.818751.
73. Hauptmann, P., Hoppe, N., and Pu
¨ttmer, A.
(2002). Application of ultrasonic sensors in the
process industry. Meas. Sci. Technol. 13,
R73–R83. https://doi.org/10.1088/0957-0233/
13/8/201.
74. Gao, W., Liu, W., Hu, Y., and Wang, J. (2021).
A novel NaCl concentration detection
method based on ultrasonic impedance
method. IEEE Trans. Ultrason. Ferroelectrics
Freq. Control 68, 3226–3233. https://doi.org/
10.1109/TUFFC.2021.3083773.
75. Gunawan, A.I., Dewantara, B.S.B., Santoso,
T.B., Wicaksono, I.D., Prastika, E.B., and
Prianto, C.E. (2017). Characterizing acoustic
impedance of several saline solution utilizing
range finder acoustic sensor. In Proc. IES-ETA
2017 - Int. Electron. Symp. Eng. Technol.
Appl. 2017-Decem, pp. 212–216. https://doi.
org/10.1109/ELECSYM.2017.8240405.
76. Ravichandran, S., and Ramanathan, K. (2010).
Ultrasonic investigation of ion-solvent
interactions in aqueous solutions of metal
ions in Polyvinyl alcohol solution 3, 375–384.
77. Praharaj, M.K. (2020). Molecular interaction IN
IONIC liquids at different temperatures. Int.
Res. J. Eng. Technol.
78. Kumar Praharaj, M., and Mishra, S. (2020).
Ultrasonic and ionic study of aqueous KCL
through walden plot. Int. Res. J. Eng.
Technol.
79. Dzida, M., Zore
˛bski, E., Zore
˛bski, M.,
_
Zarska,
M., Geppert-Rybczy
nska, M., Chora
˛_
zewski,
M., Jacquemin, J., and Cibulka, I. (2017).
Speed of sound and ultrasound absorption in
ionic liquids. Chem. Rev. 117, 3883–3929.
https://doi.org/10.1021/acs.chemrev.
5b00733.
80. Balasubramanian, P.S., Singh, A., Xu, C., and
Lal, A. (2020). GHz ultrasonic chip-scale
device induces ion channel stimulation in
human neural cells. Sci. Rep. 10, 3075–3120.
https://doi.org/10.1038/s41598-020-58133-0.
81. Balasubramanian, P.S., and Lal, A. (2018).
Cellular localization and dosage regulation of
neural stimulation enabled by 1.05 GHz
ultrasonics. IEEE Int. Ultrason. Symp. IUS
2018-Octob.https://doi.org/10.1109/
ULTSYM.2018.8579899.
82. Balasubramanian, P.S., and Lal, A. (2022). GHz
ultrasound and electrode chip-scale arrays
stimulate and influence morphology of
human neural cells. IEEE Trans. Ultrason.
Ferroelectrics Freq. Control 69, 1898–1909.
https://doi.org/10.1109/tuffc.2022.3152427.
83. Rathod, V.T. (2020). A review of acoustic
impedance matching techniques for
piezoelectric sensors and transducers.
Sensors 20, 4051–4065. https://doi.org/10.
3390/s20144051.
84. Yang, Y., Wei, X., Zhang, L., and Yao, W.
(2017). The effect of electrical impedance
matching on the electromechanical
characteristics of sandwiched piezoelectric
ultrasonic transducers. Sensors 17, 2832.
https://doi.org/10.3390/s17122832.
85. Grigorasx, C.G., Muntianu, G., and Gavril
a, L.
(2016). Mathematical modelling of CaCl2
aqueous solutions thermophysical
properties. Sci. Study Res. Chem. Chem. Eng.
Biotechnol. Food Ind. 17, 417–426.
86. Zhang, H.L., and Han, S.J. (1996). Viscosity
and density of water+sodium
chloride+potassium chloride solutions at
298.15 K. J. Chem. Eng. Data 41, 516–520.
https://doi.org/10.1021/je9501402.
87. Chung, C.W., Popovics, J.S., and Struble, L.J.
(2010). Using ultrasonic wave reflection to
measure solution properties. Ultrason.
Sonochem. 17, 266–272. https://doi.org/10.
1016/j.ultsonch.2009.07.004.
88. Dalmont, J.P. (2001). Acoustic impedance
measurement, Part I: a review. J. Sound Vib.
243, 427–439. https://doi.org/10.1006/jsvi.
2000.3428.
89. Dokoumetzidis, A., and Macheras, P. (2006). A
century of dissolution research: from noyes
and whitney to the biopharmaceutics
classification system. Int. J. Pharm. 321, 1–11.
https://doi.org/10.1016/j.ijpharm.2006.
07.011.
ll
OPEN ACCESS
16 iScience 26, 106907, June 16, 2023
iScienc
e
Article
STAR+METHODS
KEY RESOURCES TABLE
RESOURCE AVAILABILITY
Lead contact
Further information and requests for resources and reagents should be directed to and will be fulfilled by
the lead contact, Priya S. Balasubramanian (psb79@cornell.edu).
Materials availability
This study did not generate new unique reagents.
Data and code availability
dNo standardized datatypes are reported in this paper. All data generated in this study is available in the
figures section and will be further provided by the lead contact, Priya S. Balasubramanian (psb79@
cornell.edu) upon request.
dNo original code is generated by this study. All processing is detailed in the STAR Methods section.
dAny additional information required to reanalyze the data reported in this paper is available from the
lead contact, Priya S. Balasubramanian (psb79@cornell.edu)uponrequest.
METHOD DETAILS
For the characterization of the transducer, including frequency response, this paragraph describes the
detailed methods used. This displacement data was collected from 1 to 2 GHz at frequency interval steps
of 0.01 GHz to validate the transducer resonance frequency. The resonance was characterized by obtaining
the displacement at the backside of the transducer when driven by a continuous wave RF signal. The
displacement was sampled at a 22 mmdiameter of the displacement maxima. Each sample within the
area was averaged over ten pulse-echo received amplitudes. The raw data of the frequency response
was denoised with a second-order LOESS fit that uses a moving window that is 0.06 times the data size.
This frequency response is depicted in Figure 4.
For the analog processing, this paragraph describes the methods used in detail. Figure 5 depicts the
analog processing electronics used to sense and amplify the amplitude of the first echo amplitude over
time. The first return is commonly used as it has the largest amplitude. In contrast, the following echo re-
turns will have signal loss from attenuation, reflection-transmission, and diffraction, though potentially
more sensitivity to reflection transmission due to multiple traversals of the wave packet through the silicon
cavity. This phenomenon is depicted in Figure S2, where the sensitivity of the signal return to sensing con-
centration for the first and second echo return as a result of acoustic impedance change is graphed. Given
that the amplitude of the first echo is the largest, this study’s interest is to obtain the acoustic impedance-
dependent first return amplitude, as governed by Equa tions 1 and 2. To obtainthis value and allow for high-
resolution sensing of the first return amplitude, tracking and amplification of the signal was employed. The
signal from the transducer was amplified (ZX60- 2411BM-S+) and rectified, and the envelope of the move-
ment was extracted (LTC5564IJD). Any DC charge or voltage held by the transducer was removed through
a DC block high-pass filter and using a phase-matching pulse, the first echo envelope peak was tracked
through a sample-and-hold circuit (SHM-43MC-C). This output was fed through a differential amplifier
REAGENT or RESOURCE SOURCE IDENTIFIER
Software and algorithms
Polytec PSV and ScanViewer Polytec GmbH https://www.polytec.com/us/vibrometry/products/software
MATLAB R2021b Mathworks https://www.mathworks.com/products/matlab.html
Other
Polytec UHF-120 Polytec GmbH https://www.polytec.com/us
ll
OPEN ACCESS
iScience 26, 106907, June 16, 2023 17
iScienc
e
Article
(AD8250). The second input to this amplifier was a DC signal referenced to the value of the analog return
signal obtained for a chosen reference liquid. The reference liquid was chemical-grade deionized water for
the sensing experiments in the following sections. The output of the differential amplifier was fed through
an analog low pass filter (EF110) and sampled by the RTO-1024 10 GS/s oscilloscope. The signal was pro-
cessed post-acquisition digitally with a local moving weighted regression across 1% of the total sample size
for dynamic sensing outputs and averaged across multiple repeat trials for steady-state sensing across the
acquisition window.
For the ionic content sensing and data processing, this paragraph describes in detail the methods used. In
this study, Aluminum Nitride piezoelectric transducers generate and detect 100ns wide 1.55 GHz ultrasonic
wave packets at 1 MHz repetition. The experimental setup is shown in Figures 1 and S3,withthesiliconchip
in contact with the sensing medium of interest. The fluid laye r was a 100 mLthick aqueous solution of varying
concentrations and ionic species. A PDMS gasket controlled the volume of this sample. First, different ionic
solutions were prepared and tested with the same deionized water reference calibration of the sensing sys-
tem. The reference value was set to the 0 mV reference for a deionized water backing. Thus the signal ob-
tained from the oscilloscope following differential amplification was VO=VSHOS VSHOW.HereVO,VSHOS ,
VSHOW are the output voltage, the return voltage with the sample, and the return voltage with the deionized
water sample. All amplifications were post-processed to be calibrated to a 3 M NaCl solution with a cali-
bration constant to compensate for any phase error for each new acquisition. Ionic soluti ons were prepared
using serial dilutions of a stock solution with calibrated pipettors. Sodium Chloride (NaCl), PotassiumChlo-
ride (KCl), and Calcium Chloride (CaCl
2
) preparations were made, given their biological significance in the
human body as central ions used for neural, cardiac, and muscular signal conduction and of further impor-
tance in materials and environmental applications. The stock solutions were prepared with a 100 30.001g
analytical balance with maximum tested concentrations at or under 90% solubility limit to prevent any pre-
cipitate interfering with acoustic impedance measurements. Acoustic impedance was measured across a
range of concentrations from 0 mM to > 2M and high resolutions across select ranges. Five separate acqui-
sitions per concentration were obtained, each with the average return signal from >200 consecutive reflec-
tions. Standard deviation is reported across acquisitions as this is the larger of the intra- and inter-acqui-
sition variability and provides the more conservative measure of variability.
For the dynamic ionic content sensing for dissolution across temperatures, the following paragraphs
describe the methods used. Salt dissolution is sensed with the system depicted in Figures 1 and 5,and
described in the above sections. For dynamic signal change during dissolution, the sample rate should
be chosen by taking into consideration the expected rate of concentration change. The signal was sampled
using the oscilloscope ADC following the envelope detector and the sample-and-hold block. This method
reduced the frequency content of the dissolution rate from GHz carrier frequencies with a base-band pulse
repetition frequency of 1 MHz–100 kS/s sample rate, sufficient for the frequency content of salt dissolution
in water.
The temperature change was induced using an annular Silicone thermal element (TSA040020eR10), with a
temperature increase specified to vary approximately linearly with the voltage applied to the heater. The
temperature was recorded using a K-type digital thermocouple with a temperature resolution specified at 1
degree Celsius with an error of 1.5% of the reading. This method was used to set the temperature of the
solution at various values tested. To determine that water evaporation is not causing signal fluctuation,
a constant concentration of 1 M NaCl was tested with multiple temperature settings during the same
time course. The return signal in mV was determined to be consistent across the reading under the
same sample rate of 100 kS/s for 100 seconds, ensuring that the water was not evaporating and causing
a concentration spike, and only the dissolution is measured. Given that the dissolution is completed within
the first 10 seconds of acquisition, 20 seconds of data are depicted to visualize the dissolution dynamics
better. The reflection amplitude data was post-processed with a locally weighted linear regression
(LOWESS) filter across a moving window of 1% of the data points per sample.
QUANTIFICATION AND STATISTICAL ANALYSIS
All ionic content measurements for constant concentrations are reported as mean and standard deviation.
The graphical depictions with error bars depict standard deviation in all cases, for both the transducer char-
acterization and ionic content sensing applications. The mean is taken across >200 consecutive return echo
amplitudes for each measurement, and 5 measurements for each reported concentration. Dynamic
ll
OPEN ACCESS
18 iScience 26, 106907, June 16, 2023
iScienc
e
Article
measurements are filtered using locally weighted filtering techniques, with details for each data acquisition
reported in the method details section. Statistical testing and filtering is performed on Matlab 2021b.
ANOVA with Post-Hoc Tukey HSD testing is performed for the 0 to 5 mM concentration range, and results
are significant at a significance level of a= 0.05, the post-hoc testing suggests significance across almost
all pairs of concentrations with at least p<0.05 and many pairs significant with a p<0.01, except between
datapoints 1 mM and 2 mM which did not have a significant difference.
ll
OPEN ACCESS
iScience 26, 106907, June 16, 2023 19
iScienc
e
Article
