Gigahertz Ultrasonic Multi-Imaging of Soil Temperature, Morphology, Moisture, and Nematodes

Abstract

This work reports a single chip GHz ultrasonic micro-imager for imaging soil temperature, morphology, moisture, and pests such as nematodes. A 128×128 pixel array of 50×50 µm piezoelectric Aluminum Nitride (AlN) transducers is integrated onto 130nm CMOS substrates. The imager-surface is segmented into three sensing regions: soil temperature, moisture, and direct soil imaging of morphology and pest, such as nematodes. The imager's compact size and potentially low price can significantly reduce the barrier to the success of digital agriculture, which requires data collections over millions of acres in a cost-effective way with high sampling resolution.

Full text

GIGAHERTZ ULTRASONIC MULTI-IMAGING OF SOIL TEMPERATURE,
MORPHOLOGY, MOISTURE, AND NEMATODES
Anuj Baskota1, Justin Kuo1, and Amit Lal1
1Geegah Inc., Ithaca, NY, USA
ABSTRACT
This work reports a single chip GHz ultrasonic micro-
imager for imaging soil temperature, morphology,
moisture, and pests such as nematodes. A 128x128 pixel
array of 50x50 µm piezoelectric Aluminum Nitride (AlN)
transducers is integrated onto 130nm CMOS substrates.
The imager-surface is segmented into three sensing
regions: soil temperature, moisture, and direct soil imaging
of morphology and pest, such as nematodes. The imager’s
compact size and potentially low price can significantly
reduce the barrier to the success of digital agriculture,
which requires data collections over millions of acres in a
cost-effective way with high sampling resolution.
KEYWORDS
Gigahertz ultrasound, MEMS, ultrasonic imaging,
CMOS, digital agriculture
INTRODUCTION
Digital agriculture is transforming conventional
farming habits by incorporating technology and leading to
more data-driven decisions. Sensors used in agriculture
help quantify several biotic and abiotic factors impacting
crops. For example, soil moisture and temperature are
essential to soil properties that affect crops' growth and
quality. Water is necessary for photosynthesis, whereas the
soil temperature directly influences plants’ water and
nutrient intake. Similarly, pests such as nematodes
parasitize plants and reduce crop yields, causing high
economic damage[1]. The ability to monitor soil
temperature, moisture, and pests would provide vital
information to the farmers that can enhance agricultural
yield, quality and improve irrigation, fertilizer, and
pesticide use.
Commercially available moisture sensors, which are
primarily based on resistance and capacitance
measurement, have limitations such as electrolytic
corrosion and the readings being impacted by temperature
[2]. Current detection techniques of nematodes include
molecular methods such as real-time PCR or SCAR marker
[3] or imaging techniques like confocal microscopy [4].
These processes, including soil extraction/filtering, can be
labor intensive and expensive. In addition, the majority of
soil sensors focus on single-point/single-modality
measurements and therefore integrating different sensing
modalities with individual packaging and electronic
interfaces leads to bulky and expensive systems. Other
limitations also include battery life and low-resolution data
[5].
A compact, multi-modal sensor can be achieved by
addressing the limitations mentioned above, which can
potentially be deployed to sample fields with high
sampling resolutions owing to low cost and a small form
factor. We propose utilizing a GHz ultrasonic imager - as a
tool holding multi-sensing abilities in the agricultural-tech
arena. The GHz ultrasonic imager consists of an array of
ultrasonic transmit/receive pixels, with each pixel
comprised of a thin film Aluminum Nitride (AlN)
transducer and underlying CMOS electronics. Each pixel
sends a GHz frequency ultrasonic pulse into the silicon
substrate. The wave-packet travels through the silicon, hits
the material on the other side of the film, and is reflected to
the silicon where the AlN transducers receive it. The
magnitude of the received ultrasound from each pixel is
measured, thus allowing images to be formed by plotting
the values obtained by the pixel array.
The imager used in this work consists of a 128 x 128-
pixel array, with 50 µm pixel pitch as shown in Fig. 1A
and 1B. With several thousand pixels, the grouping enabled
the ability to analyze the return signal differently, allowing
specific information to be extracted from each of these
groups (Fig. 1C). The GHz imager used in this work uses
the same device layer stack as reported by Kuo et al [6].
The GHz ultrasonic imager has a few key features that
make it a powerful and compact tool to advance mass data
acquisition in agriculture. Using ultrasound instead of light
allows this GHz ultrasonic imager to be used under the soil
in a thin form factor. The GHz frequency also yields high-
resolution images due to the small ultrasonic wavelength –
around 4.5 µm at 1.85 GHz in silicon. This scaling is
particularly important to image nematodes accurately
considering their small size and observe any minor spatial
variation of moisture, temperature, and soil composition.
Similarly, the division of the pixels by function using a
Figure
1: A) 128 x 128 Pixel Imager Chip. B)
50 µm pixel with
integrated CMOS T/R circuits. C) Schematic showing GHz
ultrasonic imaging concept under soil.
978-1-6654-0911-7/22/$31.00 ©2022 IEEE 519 IEEE MEMS 2022, Tokyo, JAPAN
9 - 13 January 2022

simple 3D printed structure allows gathering important soil
data all at once. This integration of multiple sensing modes
in one device reduces the power, space, size, and the cost
of the digital-ag IoT device. In addition, the CMOS-MEMS
integration allows for features such as GPS and RF
communication to be integrated. This level of integration
would allow for wireless communication of these compact
imagers and the potential for mass data collections over
millions of acres of farmlands that could revolutionize
digital agriculture.
In this work, we successfully demonstrate the ability
of this imager to measure soil moisture, temperature, and
image nematodes as well as soil morphology maps over
several experiments. Similarly, we also show data
collection over several days, recording events of
nematodes coming near the imager and identifying soil
properties to optimize plant growth.
EXPERIMENTS
The GHz ultrasonic imager was used to perform
several experiments to gather appropriate soil moisture,
temperature, nematodes, and morphology data. The
ultrasound carrier frequency used for all the experiments
reported in this paper is 1.853 GHz. Similarly, the air
reading was taken as a baseline for every reading and later
subtracted from the actual experiment readings.
Initially, the imager was calibrated to measure soil
moisture and temperature. For moisture measurement,
loam soil was ground using a mortar and pestle to break
down large lumps. A fixed volume of this soil (6.4 cm3)
was mixed with varying water volume to obtain soil
samples of different Volumetric Water Content (VWC%),
ranging from 0% to 75%. These samples were placed on
the imager surface and were gently pressed against the
imager surface to ensure proper contact between the pixels
and the soil. For temperature calibration, the imager was
placed inside an oven (Tenney Jr. 9818-65) and the air was
imaged for different temperatures (-10 to 40 °C, interval of
5 °C). The sampling frequency for the calibration
experiments was 6.5 fps. Nematodes were visualized by
directly placing them in contact with the imager’s surface
and placing them in the soil near the imager. The
nematodes used in the experiments are Steinernema
carpocapsae (BioBest Sustainable Crop Management),
considered as beneficial nematodes. Beneficial nematodes
were used as they are easily available and disposable.
These commercially purchased nematodes are packaged
along with their food paste. Therefore, water was added to
the mixture to separate nematodes from the residues. Using
an optical microscope (Hayear HY-2307), their average
length and width was found to be 600 µm and 30 µm
respectively. This size and shape observed from the optical
images were also used to verify the nematode presence in
the ultrasonic images. The soil morphology was also
observed during the entire experiment using ultrasonic
images. The 3D printed structure for pixel separation was
printed using a 3D printer (Original Prusa i3 MK3), using
polylactic acid (PLA) filament. It was then attached to the
imager’s silicon surface using ethyl cyanoacrylate (Krazy
Glue KG98848R) which was chosen because this adhesive
could be removed using acetone.
Three sections (shown in Fig. 2A) – cavity for soil
moisture, closed region for temperature, and open region
for soil morphology and nematodes – in the 3D printed
structure is also prepared to allow for the best sensing of
the soil parameters. The cavity region is filled with 48 µm
hydrophilic stainless-steel spheres and covered with a 25
µm nylon mesh filter to allow only water to flow through
the cavity to the imager surface. In the “closed section”, the
pixels are completely covered by the 3D printed PLA from
exposure to soil or water. Therefore, the only changes in
the pixel signals in this region are from changes in
temperature. The open region is directly in contact with the
soil to visualize nematodes touching the imager. Similarly,
this region also allows visualizing the soil morphology,
which includes the composition of slit, clay, rocks, and
sand.
Multi-day experiments were carried out using the
imager and the 3D printed structure to monitor the change
of soil temperature and moisture over time and capture
instances where nematodes appear on the imager. The
imager was dipped in a large pot containing soil (Fig. 2B)
with tiny wood chips and rocks. Approximately 13 mL of
water was sprayed on the surface of the soil while watering.
Soil temperature was changed using an infrared lamp
Figure 2: A) The 3D printed structure on the imager
surface with the compartments marked (1: cavity
covered with mesh, 2: closed region, 3: open region). B)
Experiment setup with the imager under soil near a
plant. C) The overall experimental setup showing the
imager dipped in the soil, and the infrared lamp placed
approximately 20 cm away from the soil surface.
520

placed about 20 cm away from the soil facing the surface
of the imager (Fig. 2C). The temperature was increased by
turning on the lamp for 30 seconds. Nematodes were also
added a few times throughout each experiment. In each
instance, about 200 live nematodes mixed with 100 µL
water were pipetted on the surface of the soil. They were
gently spread out on the soil surface using a spatula. This
enabled the nematodes to travel through the soil to the
imaging surface below the soil. The sampling frequency
for multi-day experiments was approximately 13 fps. The
calibration curves and the soil/nematodes images from the
short-term experiments were used to quantify and analyze
long-term experimental data.
RESULTS AND DISCUSSION
The pixel outputs are the quadrature (I and Q)
components of the first acoustic echo from the silicon
backside interface. Two quadrature 128 x 128, 12-bit
matrices – [I_M] and [Q_M], which represent measured
signals for in-phase and out-of-phase data, are generated.
To improve the image contrast, the air-backed reading
matrices [I_B] and [Q_B]obtained at the beginning of each
experiment is subtracted from the image matrices:
[IIMAGE] = [I_M] – [I_B] (1)
[QIMAGE] = [Q_M] – [Q_B] (2)
For simplicity, values obtained from these matrices are
plotted to form images of the soil. To remove any DC
offsets, no-echo signals ([I_N] and [Q_N]) were also
obtained. These are the signals that were read after the echo
died out. For moisture calibration, the wet and dry pixel's
average return signal [QIM AGE] was measured to correlate
with the soil VWC%. The obtained calibration plot is
shown in Fig. 3A. Temperature can be measured through
the phase change of the return echoes, as the speed of sound
in silicon changes with temperature [7]. Therefore, the
phase was calculated using the following relation:
Phase (θ) = tan( __
_ _ ) (3)
The derived phase plotted for 5 pixels is shown in Fig.
3B. An experiment was also performed to compare the
imager’s temperature reading with a standard temperature
sensor (BME 280). Here, the BME 280 was placed next to
our imager’s surface and an infrared lamp located around
20 cm away (Fig. 2C) from the surface was turned on for
50 seconds. Fig. 3C shows the phase shift and the
interpolated temperature values with changing frames for
two random pixels (sampled at 6.5 fps) along with BME
280 temperature sensor data. The ultrasonically measured
temperature follows the BME 280 sensor data, validating
the ultrasonic temperature measurement.
One of the long-term experiment measurements of soil
VWC% and the average temperature is shown in Fig. 4A.
These results show a rapid rise in temperature when the
infrared lamp is turned on and a slight decrease in
temperature as the lamp is turned off. The measurement
showed moisture decreasing as the lamp was turned on and
consequently increasing as the lamp was turned off. Due to
the instantaneous high heat, water evaporates from the
surface, which decreases the VWC%. As the imager is
heated for brief period, there is a possibility of osmosis
from the wet soil in the surrounding regions to the
moisture-starved imager surface due to a large water
gradient [8] which increases the sensor value.
To visualize soil nematodes and soil morphology, [QIMAGE]
is plotted. The shape of nematodes appearing in these
ultrasonic images suggests the presence of nematodes in
the soil, as shown in Figure 4B. This identification can be
supported by checking the length and width of nematodes
in the acoustic images, which were usually found to be 10
– 15 pixels (500 – 750 µm) long and 1-2 pixels (50 – 100
µm) wide. This falls in the range of the average size for
S.carpocapsae [9]. Previous experiments have used optical
microscopes to confirm that these shapes correspond to the
presence of nematodes [6]. Throughout the multi-day
experiments, images of soil can also be seen changing with
water addition and soil drying. The overall imaging itself
maps out the morphology of the soil. Ultrasonic imaging
has a huge potential for researchers to study various abiotic
and biotic variations inside the soil environment. These
images could be further interpreted to find the potential
relationship of the ultrasonic signals with organic matter,
water retention, and soil porosity.
CONCLUSIONS AND FUTURE WORK
This work reports the first-ever single chip GHz
ultrasonic imager simultaneously measuring soil
temperature, moisture and imaging soil morphology and
nematodes. This monolithic CMOS-integrated GHz
ultrasonic imaging array successfully measured soil
parameters in multi-day experiments and shows great
Figure 3: A) Calibration curve of moisture: QIMAGE plotted against soil VWC%. B) Calibration curve of unwrapped
phase shift varying with temperature for 5 random pixels. C) Phase and the interpolated temperature changing over
time measured by the GHz imager and BME 280.
521

potential as a compact agricultural sensor for mass data
acquisition. One of the limitations for moisture
measurement was the calibration curve obtained using
loam soil. As water retention rate and density vary with
different soils (clay, sand, slits), the obtained calibration
curve might not derive accurate moisture measurements in
other soil types [10]. In future work, calibration curves for
different soil types will be addressed. In addition,
collecting data at high frames rates (>6 fps) for multi-day
experiments require huge memory spaces. This can be
worked out by implementing machine learning
to detect nematodes in the generated ultrasonic images in
real-time. The sampling frequency can be much lower at
first, and after nematode detection by the model, the
reading rate can be increased for a short period of time,
resulting in a large reduction in memory requirements.
Similarly, further experiments should be performed in
large fields and greenhouses to study the variation of soil
parameters in more realistic environments. Multiple
imagers placed in the same location may also increase SNR
by co-incidence of events being recorded.
ACKNOWLEDGEMENTS
This work was performed in part at the Cornell NanoScale
Facility, a member of the National Nanotechnology
Coordinated Infrastructure (NNCI), which is supported by
the National Science Foundation (Grant NNCI-2025233).
The ARPA-E also funded this work under the OPEN 2018
program (DE-AR0001049)
REFERENCES
[1] S. Koenning and K. Wrather, “Suppression of
Soybean Yield Potential in the Continental United
States by Plant Diseases from 2006 to 2009,” Plant
Health Progress, Nov. 2010.
[2] M. Saleh, I. H. Elhajj, D. Asmar, I. Bashour and S.
Kidess, "Experimental Evaluation of Low-Cost
Resistive Soil Moisture Sensors," 2016 IEEE
International Multidisciplinary Conference on
Engineering Technology (Imcet), 2016, Pp. 179-184
[3] C. Maleita, J. M. S. Cardoso, L. Rusinque, I. Esteves,
and I. Abrantes, “Species-Specific Molecular
Detection of the Root Knot Nematode Meloidogyne
luci,” Biology, vol. 10, no. 8, p. 775, Aug. 2021
[4] F. E. O’Callaghan, R. A. Braga, R. Neilson, S. A.
MacFarlane, and L. X. Dupuy, “New live screening of
plant-nematode interactions in the
rhizosphere,” Scientific Reports, vol. 8, no. 1, Jan.
2018
[5] Ray, P.P.: Internet of things for smart agriculture:
technologies, practices and future direction. J.
Ambient Intell. Smart Environ. 9(4), 395–420 (2017)
[6] J. Kuo, A. Baskota, S. Zimmerman, F. Hay, S.
Pethybridge and A. Lal. “Gigahertz Ultrasonic
Imaging of Nematodes in Liquids, Soil, and Air,”
IEEE International Ultrasonic Symposium, 2021.
[7] M. Abdelmejeed, “AlN Piezoelectric CMOS
Compatible Ultrasound Transducers: Applications on
Sensing and Computation,” Ph.D. Thesis, Cornell
Univ., Ithaca, NY, USA, 2020.
[8] C. J. Lynde, “Osmosis in Soils: Soils Act as Semi‐
Permeable Membranes,” Agrom.J, 7, 15-19, 1915
[9] K. B. Nguyen and D. J. Hunt, Entomopathogenic
nematodes : systematics, phylogeny and bacterial
symbionts. Leiden Brill Boston, 2007.
[10] R. D. Jong, C. A. Campbell, And W. Nicholaichuk,
“Water Retention Equations And Their Relationship
To Soil Organic Matter And Particle Size Distribution
For Disturbed Samples,” Canadian Journal Of Soil
Science, Vol. 63, No. 2, Pp. 291–302, May 1983, Doi:
10.4141/Cjss83-029.
CONTACT
*A. Lal, amit@geegah.com
Figure 4: A) Average of soil moisture (VWC%) and temperature changing over 35 hours span. The shaded region
represents night. The icons represent different events carried out such as watering the plants, heating the surface, or
adding nematodes to the soil. B) Ultrasonic image of soil with nematodes captured at one instance. The background
of the ultrasonic image also provides signal variation which is the soil morphology.
522