micromachines

Article
Detection of Particulate Matter of Size 2.5 µm with
a Surface-Acoustic-Wave Sensor Combined with
a Cyclone Separator
Fung-Yu Kuo 1 , Ying-Chen Lin 2 , Ling-Yi Ke 1 , Chuen-Jinn Tsai 2 and Da-Jeng Yao 1,3, * ID

1 Institute of NanoEngineering and MicroSystems, National Tsing Hua University, Hsinchu 30013, Taiwan;
fency726@gmail.com (F.-Y.K.); lingyi0412@gmail.com (L.-Y.K.)
2 Institute of Environmental Engineering, National Chiao Tung University, Hsinchu 30013, Taiwan;
eva770117.ev99g@nctu.edu.tw (Y.-C.L.); cjtsai@mail.nctu.edu.tw (C.-J.T.)
3 Department of Power Mechanical Engineering, National Tsing Hua University, Hsinchu 30013, Taiwan
* Correspondence: djyao@mx.nthu.edu.tw




Received: 26 June 2018; Accepted: 9 August 2018; Published: 12 August 2018


Abstract: A device to monitor particulate matter of size 2.5 µm (PM2.5) that has been designed and
developed includes a surface-acoustic-wave sensor operating in a shear horizontal mode (SH-SAW)
combined with a cyclone separator. In our tests, aerosols generated as incense smoke were first
separated and sampled inside a designed cyclone separator; the sampled PM2.5 was then introduced
into the sensing area of an SH-SAW sensor for detection. The use of microcentrifuge tubes as a cyclone
separator effectively decreases the size and power consumption of the device; the SAW sensor in
a well design and operating at 122 MHz was fabricated with MEMS techniques. After an explanation
of the design of the cyclone separator, a simulation of the efficiency and the SAW sensor detection are
discussed. A microcentrifuge tube (volume 0.2 mL, inlet and outlet diameters 0.5 mm) as a separator
has separation cutoff diameters 50% (d50 ) at 2.5 µm; the required rate of volumetric flow at the
inlet is 0.125 LPM, according to simulation with computational fluid dynamics (CFD) software;
the surface-acoustic-wave (SAW) sensor exhibits sensitivity approximately 9 Hz/ng; an experiment
for PM2.5 detection conducted with the combined device shows a strong positive linear correlation
with a commercial aerosol monitor. The limit of detection (LOD) is 11 µg/m3 with sample time 160 s
and total detection duration about 5 min.

Keywords: cyclone separator; surface-acoustic-wave sensor; environmental sensing; PM2.5

1. Introduction
The potential health effect caused by particulate matter in air, also called aerosol, is significant at
present as air pollution becomes severe. Anthropogenic and biological aerosols, such as those found in
haze, smoke and bacteria, fungi found in facilities for treatment of wastes and agricultural activity,
can invade the human respiratory system and cause asthma, lung, and cardiovascular diseases [1,2].
In 2004, the contribution of ambient and indoor air pollution to lung cancer was estimated by World
Health Organization’s Globe Burden of Disease to cause, worldwide, 62,000 and 16,000 deaths per
year; airborne PM is designated a carcinogen in Group 1 by WHO International Agency of Research
on Cancer [3]. In 2012, diseases contributed by air pollution caused 7 million deaths [4]; guidelines
and regulation of PM have hence been established by governments and organizations worldwide.
A common classification of PM is based on its size, such as PM2.5 for which the aerodynamic diameters
(AD) are smaller than 2.5 µm. WHO set guideline values of PM2.5 as annual mean concentration
10 µg/m3 and 24-h mean concentration 25 µg/m3 , but few countries can achieve this goal [5].

Micromachines 2018, 9, 398 ; doi:10.3390/mi9080398 www.mdpi.com/journal/micromachines

Micromachines 2018, 9, 398 2 of 11

The impact of the particles collected on the surface of the acoustic sensor induces a gravimetric
effect that modifies the conditions for acoustic-wave propagation [6]. The development of a device
that is convenient, cheap, and suitable for monitoring the personal exposure to aerosols in real time
thus becomes an important task [7].
PM2.5 monitoring is currently generally conducted using instruments based on, for example,
beta attenuation or a tapered-element oscillating microbalance, which are expensive, time-consuming,
and bulky, as shown in Table 1. Some miniature PM2.5 sensor modules for personal use are available
on the market; most are based on the principles of light scattering and piezoelectricity. For light
scattering, such a detection mechanism is in accordance with an assumption of the physical parameters
and size distribution parameters of the aerosol, which causes an uncertainty of measurement
because of the natural variability of PM2.5 and leads to a poor accuracy of a sensor [8,9]. Based on
the sensing mechanism of a mass-loading effect, a device with a surface-acoustic-wave sensor
operating in a shear horizontal mode (SH-SAW) can effectively decrease this inaccuracy and exhibit
a greater sensitivity than another piezoelectric sensor such as a quartz-crystal microbalance (QCM).
A surface-acoustic-wave (SAW) device can be fabricated in a simple process and operated without
a complicated setup or expensive instruments [10,11].

Table 1. Comparison of aerosol monitoring instruments.

Resolution
Principle Instrument Advantages Volume/Mass Price (NTD)
(m/m3 )
TSI DustTrak Real-time measurement
Optical 1 6 × 16 × 10cm3 /1 kg $300,000
8532 (1 s); Portable
Beta Met-One Continuous data;
0.1 31 × 43 × 40cm3 /24.5 kg $580,000
Attenuation BAM 1020 Error-handling software
Oscillating Thermo RP
0.1 Simple design principle 28 × 22 × 43cm3 /20 kg $550,000
Microbalance 1400
Spectrometer GRIMM 180 0.1 31 size channels 27 × 36 × 48cm3 /18 kg $850,000

In additional to the sensitivity and reliability of the sensor itself, without an effective aerosol
sampler most monitors might have a large disturbance signal from aerosols of kinds other than PM2.5.
In this work, a microcentrifuge tube was developed into a replaceable cyclone separator for PM2.5
sampling. For a comparison with the required flow rate of commercial separators, for a commercial
separator to achieve a cutoff diameter at 2.5 µm the flow rate is about 2–4 LPM [12]; our separator
with a much smaller required flow rate is compatible with a micro pump that can decrease the size of
a complete device.

2. Materials and Methods

2.1. Fabrication of SH-SAW Chips

The SH-SAW chip was designed with central frequency 122.4 MHz. A lithium-tantalate (LiTaO3 )
piezoelectric substrate with a large electromechanical coupling coefficient (K2 ) was chosen [13];
the interdigital transducer (IDT) made from gold was patterned with photolithography and an e-beam
evaporator. The IDT consist of 50 pairs with finger width 8.5 µm, and deposited thickness 100 nm;
an overall die size is 13.4 mm × 7.4 mm. The process included three main steps: (a) photolithography,
(b) e-beam evaporation and (c) lift-off, as shown in Figure 1.
 Micromachines 2018, 9, 398 3 of 11
Micromachines 2018, 9, x FOR PEER REVIEW 3 of 11

Figure 1. Fabrication of SH-SAW chips.

Figure 1. Fabrication of SH-SAW chips.

(a) Photolithography
(a) Photolithography
TheThe
LiTaO 3 wafer
LiTaO was
3 wafer wasfirst
firstcleaned
cleaned according tothe
according to thefollowing
following standards.
standards. TheThe
waferwafer was given
was given
a positive photoresist (AZ5214, MicroChemicals GmbH, Ulm, Germany) at 3000 rpm for 30 s. The
a positive photoresist (AZ5214, MicroChemicals GmbH, Ulm, Germany) at 3000 rpm for 30 s. The wafer
wafer
waswas
thenthen baked
baked at precisely 100 ◦ 100
at precisely C for°C for 1and
1 min minexposed
and exposed to ultraviolet
to ultraviolet (UV)
(UV) light. Thelight. Thewas
pattern pattern
was developed and inspected with a microscope to verify the completeness of the IDT structure.
developed and inspected with a microscope to verify the completeness of the IDT structure.
(b) E-beam evaporation
(b) E-beam evaporation
After the photolithographic process, Cr (thickness 20 nm) was deposited first on the substrate
Afteradhesive
as an the photolithographic process,
layer; Au (thickness Crnm)
100 (thickness 20 nm) was
was deposited deposited
second first structure
as a major on the substrate
with as
an adhesive layer; Au (thickness 100 nm) was deposited second as a major structure with e-beam
e-beam evaporation.
evaporation.
(c) Lift-off
After the evaporation, the wafer was immersed and sonicated in acetone to remove unnecessary
(c) Lift-off
photoresist, and the pattern was defined. The finished wafer was diced into chips with laser cutting.
After the evaporation, the wafer was immersed and sonicated in acetone to remove unnecessary
2.2. Cyclone Separator Design and Simulation
photoresist, and the pattern was defined. The finished wafer was diced into chips with laser cutting.
Lindsley and coworkers were the first to use a microcentrifuge tube as a replaceable cyclone
2.2. separator
Cyclone Separator Designsampling
for bioaerosol and Simulation
[14]. Based on this research, we developed a microcentrifuge
tube (volume 0.2 mL) into a smaller PM2.5 cyclone separator [15,16]. In our work, aerosols are first
Lindsley into
introduced andthecoworkers
tube from were the first
the slanted inletto use
and a microcentrifuge
separated tube asofaaverage
via a vortex; particles replaceable cyclone
diameter
separator for bioaerosol
(AD) greater than 2.5 µmsampling [14]. Based
are collected on thisofresearch,
in the bottom the tube,we developed
whereas a microcentrifuge
particles with AD smaller tube
(volume 0.2µm
than 2.5 mL) into a smaller
are transported to thePM2.5 cyclone
upper outlet and separator [15,16].
can be detected with In
theour work,
sensor. aerosols are
The dimensions of first
introduced into the tube from the slanted inlet and separated via a vortex; particles
the tube are fixed; two parameters of the inlet and outlet sizes are designed such that the ratio betweenof average
diameter (AD)and
tube, inlet greater
outletthan 2.5 D:
is about μmdinare collected
/dout in the
= (A) 10:1 andbottom ofshown
(B) 5:1, as the tube, whereas
in Figure 2. particles with AD
smaller than 2.5 μm are transported to the upper outlet and can be detected with the sensor. The
dimensions of the tube are fixed; two parameters of the inlet and outlet sizes are designed such that
the ratio between tube, inlet and outlet is about D: din/dout = (A) 10:1 and (B) 5:1, as shown in Figure 2.
Micromachines 2018, 9,2018,
Micromachines x FOR PEER REVIEW
9, 398 4 of 11 4 of 11

Figure 2. Dimensions and design parameters of microcentrifuge tube (0.2 mL) as a cyclone separator.
Figure 2. Dimensions and design parameters of microcentrifuge tube (0.2 mL) as a cyclone separator.
The optimum operating flow velocity of a cyclone separator is between 10 and 20 m/s;
Theforoptimum
operation operating
at a small flowflowrate,
velocity
the inletof and
a cyclone separator
outlet diameters areisdecreased,
between although
10 and the 20 m/s; for
operation at a small
separation flow might
efficiency rate, the
alsoinlet and
decline. Theoutlet diameters
designed cyclone are was decreased,
hence simulated although the separation
with software
efficiency might also fluid
for computational decline. The(CFD)
dynamics designed
to find thecyclone was between
best balance hence efficiency
simulated and with software for
operational
computational fluid dynamics (CFD) to find the best balance between efficiency and operational
consumption. CFD software (Fluent, ANSYS Fluent Inc., Lebanon, NH, USA) with the finite-element
method was used. The simulation had three major parts. (1) The cyclone geometry was drawn using
consumption. CFD software (Fluent, ANSYS Fluent Inc., Lebanon, NH, USA) with the finite-element
3D CAD software (SolidWorks) and meshed in pre-processing; the number of mesh elements was
methodabout
was used. The
a million. (2)simulation had three
Numerous solution major were
parameters parts.set,(1) The cyclone
including geometry
phase models. The airwaswasdrawn
set using
3D CADat software (SolidWorks)
normal temperature and as
and pressure meshed in pre-processing;
a continuous phase using a Reynoldsthe number
stress modelof mesh
to solveelements
with was
about a inert
million. (2) Numerous
particles (0.1–6 µm) assolution
a discreteparameters
phase using awere model set,
of including phase models.
discrete particulate matter (DPM)The to air was set
at normal temperature
solve; an uncoupledand pressure
calculation wasas a continuous
chosen phasetheusing
for them because volume a fraction
Reynolds stress in
of particles model
air is to solve
small, ≤12%, in which case the continuous phase is not impacted by the presence of the discrete phase.
with inert particles (0.1–6 μm) as a discrete phase using a model of discrete particulate matter (DPM)
For the phase characteristics, the discrete phase density is the average bulk density of the PM2.5 [17,18].
to solve;Previous
an uncoupled calculation
PM2.5 studies reportedwas chosen
particle for them
densities mainlybecause
between the volume
1 g/cm 3 and 3 fraction of particles in
g/cm3 [19,20].
air is small,
For the≤12%, in which
boundary case the
conditions, continuous
particles phase iswhen
became trapped not they
impacted
reachedby thethe presence
surface of the of the discrete
outlet.
phase. For the phase characteristics, the discrete phase density is the average bulk density of the
(3) After the calculation, an analysis of the flow field and particle trajectories analysis was undertaken
PM2.5 [17,18]. Previousas PM2.5
in post-processing, shown instudies
Figure 3. reported
The particlesparticle densities
were separated mainly
with the vortex between
as a function 1 g/cm
of 3 and
time. The particles of average diameter (AD) greater than 2.5 µm were collected at the bottom of the
3 g/cm3 [19,20]. For the boundary conditions, particles became trapped when they reached the surface
tube, whereas particles with AD smaller than 2.5 µm were transported to the upper outlet. The number
of the outlet. (3) After
of suspended the calculation,
particles was hence equal an toanalysis
the number of the flow field
of meshes of theand
face; particle
the particlestrajectories
that were analysis
was undertaken
calculated toinreach
post-processing,
the boundary could as beshown
trapped inorFigure 3. The
could escape. Theparticles
position ofwere separated
the particles in the with the
vortex asseparator
a function ofwith
varied time. The
time particlesonofthe
depending average
size. Thediameter (AD) greater
particles entered thanat2.5
the separator theμm were collected
maximum
at the bottom of the tube, whereas particles with AD smaller than 2.5 μm were transported
intake velocity; the reverse flow generated when reached the bottom brought together the particles of to the
small diameters; the particles of large diameter were deposited below. The particle cutoff diameters
upper outlet. The number of suspended particles was hence equal to the number of meshes of the
(d50 ) were near the experimentally determined values. The number of particles separated in the outlet
face; theisparticles
plotted as that were calculated
a cumulative-volume to reachofthe
distribution boundary
particle could
size so that the be
d50 trapped or could[21].
could be compared escape. The
position of the particles in the separator varied with time depending on the size. The particles entered
the separator at the maximum intake velocity; the reverse flow generated when reached the bottom
brought together the particles of small diameters; the particles of large diameter were deposited
below. The particle cutoff diameters (d50) were near the experimentally determined values. The
number of particles separated in the outlet is plotted as a cumulative-volume distribution of particle
size so that the d50 could be compared [21].
Micromachines 2018, 9, 398 5 of 11
Micromachines 2018, 9, x FOR PEER REVIEW 5 of 11

Figure 3. Particle traces colored according to the diameter at varied flow times. The simulation photos
have shown the different size particle in the cyclone.
Figure 3. Particle traces colored according to the diameter at varied flow times. The simulation photos
have shown
2.3. Sensing the different size particle in the cyclone.
System
The SH-SAW
2.3. Sensing System sensor was first tested to identify its sensitivity to PM2.5. Two-port sensors were
used in the experiment, one of which is considered a reference to monitor and to compensate for the
Theambient
variable SH-SAWcondition,
sensor was andfirst
thetested
othertois identify
for PM2.5 its detection,
sensitivityas toshown
PM2.5.in Two-port
Figure 4a.sensors were
The setup
used in the experiment, one of which is considered a reference to monitor
of the detection experiment is shown in Figure 5; an aerosol was generated on burning incense and to compensate for the
variable
sticks, ambient
which condition,
renders and the other
a distribution is for size
of particle PM2.5 detection,
from 0.06 to as
2.5shown
µm, with in Figure
average 4a.bulk
The setup
density of
the detection
3 experiment is shown in Figure 5; an aerosol was generated
1.1 g/cm [22]. Incense smoke was exhausted with a fan into a glass chamber (20 × 15 × 15 cm ), on burning incense sticks,
3
whichdry
with renders a distribution
air supplied to adjustofitsparticle size from
concentration from0.06 to 200
10 to 2.5 µg/m
μm, with3 . The average
chamber bulk density
connected
1.1 g/cm
into
3
two pipe [22]. channels,
Incense smoke was one
of which exhausted with atofan
led directly theinto a glasssensor
SH-SAW chamber with(20a ×PM2.5
15 × 15 cm3before
filter ), with
the sensor and an acrylic cap covering the chip to ensure that the particles can strike and adheretwo
dry air supplied to adjust its concentration from 10 to 200 μg/m . The chamber connected into
3
to
pipe channels, of which one led directly to the SH-SAW sensor with a
the sensing zone before leaving to the outlet, as shown in Figure 4b; the flow was controlled withPM2.5 filter before the sensor
aand an acrylic
mass-flow cap covering
controller (MFC)the andchip to ensure
a vacuum pump;thatthe
theother
particles can was
channel strike and adhere
connected intoto the sensing
a commercial
zone before leaving to the outlet, as shown in Figure 4b; the flow
monitor (DustTrak Aerosol Monitor 8530, TSI Inc., Shoreview, MN, USA) to provide a reference was controlled with a mass-flow of
controller (MFC) and a vacuum pump; the other channel was connected
real-time PM2.5 mass concentrations at every second. The sensors were connected to an oscillator into a commercial monitor
(DustTrak
circuit, a powerAerosol Monitor
supply, and a8530, TSI Inc.,
frequency Shoreview,
counter; MN, USA)
a computer recordedto provide a reference
the frequency signalof every
real-time2 s.
PM2.5 mass concentrations at every second. The sensors were connected
After identifying the sensitivity of the SH-SAW sensor, we tested the combined device to an oscillator circuit,
of thea
power supply,
SH-SAW sensorand and a frequency counter; (0.2
cyclone separator a computer
mL), as recorded
shown inthe frequency
Figure 4c,d; signal
the upper every 2 s. of the
outlet
After identifying the sensitivity of the SH-SAW sensor, we tested the combined device of the
SH-SAW sensor and cyclone separator (0.2 mL), as shown in Figure 4c,d; the upper outlet of the
Micromachines 2018, 9, 398 6 of 11

Micromachines 2018, 9, x FOR PEER REVIEW 6 of 11

Micromachines 2018, 9, x FOR PEER REVIEW 6 of 11
separator was connected to the inlet of a chip cap, for separated particles to flow to the sensor. A cyclone
separator was
separator was connected to to the
the inlet
inlet of
of aa chip
chip cap,
cap, for
for separated
separated particles
particles to
to flow
flow to
to the
the sensor.
sensor. AA
separator (0.2
cyclone mL)connected
was
separator used
(0.2 mL) as
was a used
substitute
as a for a PM2.5
substitute for a filter;filter;
PM2.5 the experiments
the were
experiments wereconducted
conducted with
cyclone separator (0.2 mL) was used as a substitute for a PM2.5 filter; the experiments were conducted
and without
anda without
with and cyclone aseparator
without to verifyto
cyclone separator
separator itsverify
efficiency. The experimental
its efficiency.
efficiency. parameters
The experimental
experimental of theofseparator
parameters the
with a cyclone to verify its The parameters of the
separator
depended depended
on the resultson of the
theresults of the CFD simulation.
CFD simulation.
separator depended on the results of the CFD simulation.

Figure 4. (a)
4. (a) Two-port SH-SAW sensors.
Two-port sensors. (b) Sensing chip cap. (c) Assembled cyclone separator using
Figure
Figure 4. (a) Two-portSH-SAW
SH-SAW sensors. (b)(b) Sensing
Sensing chipchip cap.
cap. (c) (c) Assembled
Assembled cyclone using
cyclone separator separator
usingaaamicrocentrifuge
microcentrifuge
microcentrifuge tube (0.2(0.2
mL) andand
channels made of acrylic. (d) PM2.5 detection devicedevice
tube (0.2 mL) and channels made of acrylic. (d) PM2.5 detection device with a
tube mL) channels made of acrylic. (d) PM2.5 detection with awith
SH-SAW sensor combined with a cyclone separator.
SH-SAW
a SH-SAW sensor
sensor combinedwith
combined with aa cyclone
cyclone separator.
separator.

Figure 5. Experimental setup.

Figure5.5.Experimental
Figure Experimental setup.
setup.
Micromachines 2018,
Micromachines 9, 398
2018, 9, x FOR PEER REVIEW 7 of7 11
of 11

3. Results
3. Results
3.1.
3.1. Simulation
Simulation of of
thethe Efficiency
Efficiency of the
of the Cyclone
Cyclone Separation
Separation
TwoTwo parametersofofinlet
parameters inletand
and outlet
outlet sizes
sizes were
weredesigned
designedsuch suchthat
thatthethe
ratio of the
ratio tubes
of the waswas
tubes about
(A) 10:1 and (B) 5:1 and were simulated under operating flow rates 10.6
about (A) 10:1 and (B) 5:1 and were simulated under operating flow rates 10.6 m/s and 21.2 m/s m/s and 21.2 m/s separately.
Figure 6 shows
separately. Figurea 6comparison of the separation
shows a comparison of the efficiency
separationofefficiency
the microcentrifuge tube (0.2 mL).
of the microcentrifuge With
tube
inlet
(0.2 mL).flow
Withvelocities
inlet flow21.2 and 10.6
velocities 21.2m/s,
and the
10.6tube
m/s,with design
the tube with(A) had (A)
design d50 at
had1.8d50μm and
at 1.8 µm2.5and
μm,
respectively; the tube with design (B) had d 50 at 1.1 μm and 1.5 μm, respectively. From the simulation
2.5 µm, respectively; the tube with design (B) had d50 at 1.1 µm and 1.5 µm, respectively. From the
results weresults
simulation concluded that tubes that
we concluded withtubes
design withB have a separation
design B have a efficiency
separationbetter than that
efficiency betterof than
design
A, of
that but the tube
design with
A, but thedesign A still
tube with has dA50 still
design at PM2.5
has d50atatinlet
PM2.5flowat velocity
inlet flow 10.6 m/s; the
velocity 10.6required
m/s;
volumetric flow rate is 0.125 LPM. It is hence impractical to decrease
the required volumetric flow rate is 0.125 LPM. It is hence impractical to decrease the size the size of the inlet and
of outlet
the
inlet and outlet ports without limit to decrease the flow demand, but the 0.2-mL centrifuge tube was to
ports without limit to decrease the flow demand, but the 0.2-mL centrifuge tube was designed (A)
achieve(A)
designed d50toequal
achievetod50PM2.5
equal to at PM2.5
inlet at flow
inletrate
flow 10.6 m/s;m/s;
rate 10.6 the the
required
requiredflow
flow rate wasonly
rate was only
0.125
0.125 LPM.LPM.

Figure
Figure 6. Comparison
6. Comparison of of separation
separation efficiency
efficiency of of
thethe tube
tube (0.2
(0.2 mL).
mL). TheThe results
results indicate
indicate that
that design
design (B)(B)
generally had a separation efficiency better than that of design
generally had a separation efficiency better than that of design (A). (A).

3.2. SH-SAW Sensor Aerosol Experiment

3.2. SH-SAW Sensor Aerosol Experiment
3.2.1.
3.2.1. Test
Test of of Sensor
Sensor Sensitivity
Sensitivity
AtAtthethe initial
initial ventilation,
ventilation, thethe frequency
frequency increased
increased greatly,
greatly, because
because of of
thethe mass-loading
mass-loading effect
effect
caused by the air flow, but, when the air flow stopped, the frequency slowly
caused by the air flow, but, when the air flow stopped, the frequency slowly returned to the original returned to the original
value
value because
because of of
thethe non-ventilation
non-ventilation condition.
condition. AtAtthethe same
same time,
time, thethe suspended
suspended particles
particles became
became
slowly deposited on the surface of the SAW, which on returning to the original value continued to
slowly deposited on the surface of the SAW, which on returning to the original value continued
to decrease.
decrease.TheTheSH-SAW
SH-SAWsensor sensorcan candetect
detecta amass-loading
mass-loading ofof
particles
particlesthat deposit
that depositon onthethe
sensing
sensingzone
andand
zone transmit
transmit a frequency
a frequencysignal;
signal;thethe frequency
frequency data are processed
data are processedincluding
includingcompensating
compensating thethe
reference value, and zero as the initial frequency value. As shown in Figure 7, the moment that thethe
reference value, and zero as the initial frequency value. As shown in Figure 7, the moment that
aerosol
aerosol and
and airair flow
flow into
into thethe sensor
sensor causes
causes a large
a large frequency
frequency change,
change, butbut
thethe sensor
sensor returns
returns to to a stable
a stable
state
state when
when thethe
airair stops;
stops; thethe deposited
deposited particles
particles leave
leave a frequency
a frequency shift,
shift, with
with a response
a response interval
interval about
about
1–2 min. On comparison of the frequency shift with the mass concentration
1–2 min. On comparison of the frequency shift with the mass concentration measured with the aerosol measured with the aerosol
monitor,
monitor, thethe sensitivity
sensitivity (S)(S)
of of
thethe SH-SAW
SH-SAW sensor
sensor was was calculated
calculated with
with Equation
Equation (1),(1), in which
in which ∆f Δf is the
is the
frequency shift (Hz), mc is mass concentration (μg/m3), Q is the input flow rate (m3/s), and t is the
sample interval (s).
Micromachines 2018, 9, x FOR PEER REVIEW 8 of 11
Micromachines
Micromachines 2018,
2018, 9,
9, x398
FOR PEER REVIEW 88of
of11
11

∆𝑓
∆𝑓
frequency shift (Hz), mc is mass concentration (µg/m3 ), Q is the input flow rate (m3 /s), and t is the
𝑆 = (1)
sample interval (s). 𝑆 = 𝑚𝑐 × 𝑄 × 𝑡 (1)
S=
∆f 𝑚𝑐 × 𝑄 × 𝑡 (1)
mc × Q × t

Figure 7. Frequency response curve of the SH-SAW sensor to incense smoke at varied concentration
Figure
Figure7.7.Frequency
Frequencyresponse
responsecurve
curveof
ofthe
theSH-SAW
SH-SAWsensor
sensorto
toincense
incense smoke
smoke at
at varied
varied concentration
concentration.

Figure 8 shows the frequency shift responses of an SH-SAW sensor to incense smoke versus
Figure 88 shows the
the frequency shift responses of
of an
an SH-SAW sensor to
to incense smoke versus
massFigure showsmeasured
concentrations frequency
with ashift responses
commercial detector SH-SAW sensorShoreview,
(TSI DustTrak, incenseMN, smoke versus
USA); the
mass
mass concentrations
concentrations measured
measured with
witha commercial
a commercial detector
detector(TSI DustTrak,
(TSI DustTrak,Shoreview,
Shoreview, MN, USA);
MN, the
USA);
tested mass concentration is from 11 to 170 μg/m3 ; the limit of detection is 11 μg/m3 at frequency shift
3 3
tested massmass
the tested concentration is fromfrom
concentration 11 to11170
to μg/m ; the3 limit of detection is 11 is
μg/m at frequency
3 shift
response 49 Hz. We focusedis on 170 µg/m
an experiment ; thealimit
with of detection
small concentration 11 µg/m
(10-50 at frequency
μg/m 3) that
response 49
shift response Hz. We focused on an experiment with a small concentration (10-50 μg/m )
33
) that
corresponds to 49 Hz. We
a PM2.5 focused on in
concentration ananexperiment with a small
actual environment whenconcentration (10-50 µg/m
the air is in effective that
circulation
corresponds
corresponds to a PM2.5
to a PM2.5 concentration
concentration in an actual
in an actual environment
environment when
when the air is in effective circulation
with no obvious PM source such as cigarette or incense smoke. The the air is in of
coefficient effective circulation
determination is
with
with no
no obvious
obvious PM
PM source
source such
such as
as cigarette
cigarette or
or incense
incense smoke.
smoke. The
The coefficient
coefficient of
of determination
determination is
is
calculated to be 0.9835; converting the frequency to the mass sensitivity of the sensor is approximately
calculated
calculated to
to be
be0.9835;
0.9835; converting
converting the
the frequency
frequency to
to the
the mass
mass sensitivity
sensitivity of
ofthe
the sensor
sensor isisapproximately
approximately
9 Hz/ng with flow rate 3 LPM and sampling interval 10 s.
99 Hz/ng
Hz/ngwith
withflow
flowrate
rate33LPM
LPMandandsampling
samplinginterval
interval10 10s.s.

Figure 8. Frequency response of SH-SAW sensor to varied PM2.5 concentration.

Figure 8. Frequency response of SH-SAW sensor to varied PM2.5 concentration.
Figure 8. Frequency response of SH-SAW sensor to varied PM2.5 concentration.
3.2.2. Test
3.2.2. Test of
of an
an SH-SAW
SH-SAW Sensor
Sensor Combined
Combined with
with aa Cyclone
Cyclone Separator
Separator
3.2.2. Test of an SH-SAW Sensor Combined with a Cyclone Separator
After identifying
After identifying the the PM2.5
PM2.5 mass
mass sensitivity
sensitivity ofof the
the SH-SAW,
SH-SAW, we we determined
determined the
the mass
mass
After identifying
concentration of the the PM2.5
detected PM2.5mass
on sensitivitythe
measuring offrequency
the SH-SAW,
shift; we mass
the determined the mass
concentration was
concentration of the detected PM2.5 on measuring the frequency shift; the mass concentration was
concentration
calculated with of the detected
with Equation
Equation (1).(1). PM2.5 on measuring the frequency shift; the mass concentration was
calculated
calculated with Equation (1).
Micromachines 2018, 9, x FOR PEER REVIEW 9 of 11
Micromachines 2018, 9, 398 9 of 11
The sensitivity was 9 Hz/ng; other experimental parameters are shown in Table 2. The cyclone
separator (0.2 mL, with inlet and outlet diameters 0.5 mm), at operating flow rate 0.125 LPM with d50
at 2.5μm
The was chosenwas
sensitivity to replace the other
9 Hz/ng; PM2.5experimental
filter for the test. Becauseare
parameters theshown
operating flow rate
in Table was
2. The small,
cyclone
the sample interval was increased to 160 s for sufficient particles to deposit. The test without a cyclone
separator (0.2 mL, with inlet and outlet diameters 0.5 mm), at operating flow rate 0.125 LPM with d50
and
at otherwas
2.5µm PM2.5 separator
chosen wasthe
to replace conducted withfor
PM2.5 filter operating
the test. flow
Becauseratethe
3 LPM and sampling
operating flow rate interval 10
was small,
s, to verify the necessity and efficiency of the cyclone separator (0.2 mL).
the sample interval was increased to 160 s for sufficient particles to deposit. The test without a cyclone
and other PM2.5 separator was conducted with operating flow rate 3 LPM and sampling interval 10 s,
Table 2. Parameters
to verify the necessity and efficiency for the
of the cyclone cyclone separator
separator (0.2 mL).test.

Cyclone Size din/dTable

out /mm Operating
2. Parameters Flow
for the Rateseparator
cyclone /LPM test.
d50 /µm Sample Time /s

Without Cyclone
Cyclone Size din /d
-out /mm 3 Rate/LPM
Operating Flow -
d50 /µm Sample10Time/s
Without Cyclone - 3 - 10
0.2
0.2mL
mL 0.5/0.5
0.5/0.5 0.125
0.125 2.5
2.5 160
160

Figure 9a
Figure 9a shows
shows thethe mass
mass concentration
concentration detected
detected with
with the
the SH-SAW
SH-SAW sensor
sensor versus
versus the
the mass
mass
concentration measured
concentration measured with with the
the reference
reference (DustTrak); the error
(DustTrak); the error of
of the
the SH-SAW
SH-SAW sensor
sensor measurement
measurement
was much
was much greater
greater than
than that
that of
of the
the aerosol
aerosol monitor
monitor when
when thethe sensor
sensor was
was tested
tested without
without aa PM2.5
PM2.5
separator, because
separator, because particles
particles larger
larger than
than 2.5
2.5 µm
μm also
also deposited
deposited in in the
the sensing
sensing zone
zone and
and influenced
influenced thethe
frequency change.
frequency change. An An error of this
error of this kind
kind is
is difficult
difficult to
to calibrate
calibrate with
with aa calculation
calculation oror other
other method
method
because ofofthe
because the variability
variability of aerosols
of aerosols depending
depending on the an
on the source, source, an environment
environment with varied with varied
composition
composition
and and its size distributions;
its size distributions; for this reason, foran
this reason, anPM2.5
appropriate appropriate PM2.5
separator separator is necessary.
is necessary.
Figure 9b
Figure 9bshows
showsthethemassmass concentration
concentration detected
detected with anwith an SH-SAW
SH-SAW sensor
sensor and and aseparator
a cyclone cyclone
separator (0.2 mL) versus mass concentration measured with the reference (DustTrak);
(0.2 mL) versus mass concentration measured with the reference (DustTrak); the operating flow rate the operating
flow0.125
was rate LPM,
was 0.125 LPM, theinterval
the sampling sampling wasinterval wascoefficient
160 s. The 160 s. Theofcoefficient of determination
determination was calculated was
to
calculated to be 0.9642. A strong positive linear correlation between the SAW sensor
be 0.9642. A strong positive linear correlation between the SAW sensor and the reference (DustTrak) and the reference
(DustTrak)
shows shows
that the that the
cyclone cyclone(0.2
separator separator (0.2 mL) excluded
mL) effectively effectivelytheexcluded
influencetheofinfluence
particles of particles
other than
other than
PM2.5. ThePM2.5.
requiredThe required
flow flowdesign
rate of this rate ofwas
thisdecreased
design was butdecreased but the total
the total detection detection
period period
was increased
was
to increased
4–5 min; theto 4–5 min;
ability the ability
of real-time of real-time
sensing sensing
was hence was hence decreased.
decreased.

Figure 9. (a) Aerosol test of an SH-SAW sensor with no PM2.5 separator. (b) Aerosol test of SH-SAW
Figure 9. (a) Aerosol test of an SH-SAW sensor with no PM2.5 separator. (b) Aerosol test of SH-SAW
sensor with a cyclone separator (0.2 mL).
sensor with a cyclone separator (0.2 mL).

4. Conclusions
4. Conclusion
In this work, we developed a PM2.5 monitoring device. The shear horizontal- mode
In this work, we developed a PM2.5 monitoring device. The shear horizontal- mode surface-
surface-acoustic-wave sensor fabricated with a MEMS technique exhibited sensitivity 9 Hz/ng to
acoustic-wave sensor fabricated with a MEMS technique exhibited sensitivity 9 Hz/ng to PM2.5; the
PM2.5; the limit of detection was3 11 µg/m3 . A cyclone separator with a microcentrifuge tube (volume
limit of detection was 11 μg/m . A cyclone separator with a microcentrifuge tube (volume 0.2 mL,
0.2 mL, diameter 0.5 mm of inlet and outlet) had a separation cutoff diameter at 2.5 µm; the required
diameter 0.5 mm of inlet and outlet) had a separation cutoff diameter at 2.5 μm; the required flow
Micromachines 2018, 9, 398 10 of 11

flow rate was 0.125 LPM. The power consumption and size (volume) were much decreased relative
to an existing PM2.5 separator. The aerosol experimental data measured with the combined device
show a strong positive linear correlation with a commercial aerosol monitor, with sample interval
160 s, total detection period 5 min. The results show a great potential for a real-time PM2.5-monitoring
device for the environment.
For a further pratical application, the design and fabrication of the SH-SAW sensor will be
improved on decreasing the IDT width to achieve a desired wave penetration depth (~2.5 µm), so that
the PM2.5 sensitivity and ability of the sensor will improve for real-time sensing [23].

Author Contributions: Conceptualization—F.-Y.K.; Formal analysis—F.-Y.K. and Y.C.L.; Methodology—Y.C.L.;

Project administration and revised the manuscript—L.-Y.K, and D.-J.Y.; Supervision—C.-J.T. and D.-J.Y.
Funding: The funding was supported by Minister of Science and Technology (MOST), Taiwan. (MOST 105-2221-E-007-062-MY3).
Acknowledgments: The authors also thank the Center for Nanotech, Material Science and Microsystems at
National Tsing Hua University for research and facility support.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Kulkarni, P.; Baron, P.A.; Willeke, K. Biological particle sampling: Principles, techniques, and applications.
In Aerosol Measurement, 3rd ed.; Wiley: Hoboken, NJ, USA, 2011; p. 22.
2. Douwes, J.; Thorne, P.; Pearce, N.; Heederik, D. Bioaerosol health effects and exposure assessment:
Progress and prospects. Ann. Occup. Hyg. 2003, 47, 187–200. [PubMed]
3. Air Pollution and Cancer. Available online: https://www.iarc.fr/en/publications/books/sp161/
AirPollutionandCancer161.pdf (accessed on 9 August 2018).
4. World Health Organization. Burden of Disease from the Joint Effects of Household and Ambient Air Pollution
for 2012. Available online: http://www.who.int/airpollution/data/AP_jointeffect_BoD_results_Nov2016.pdf
(accessed on 9 August 2018).
5. WHO Air Quality Guidelines for Particulate Matter, Ozone, Nitrogen Dioxide and Sulfur Dioxide. Available
online: http://apps.who.int/iris/bitstream/handle/10665/69477/WHO_SDE_PHE_OEH_06.02_eng.pdf?
sequence=1 (accessed on 9 August 2018).
6. Djoumi, L.; Vanotti, M.; Blondeau-Patissier, V. Real time cascade impactor based on surface acoustic wave
delay lines for PM10 and PM2. 5 mass concentration measurement. Sensors 2018, 18, 255. [CrossRef]
[PubMed]
7. Djoumi, L.; Blondeau-Patissier, V.; Vanotti, M.; Appert-Collin, J.C.; Thomas, D.; Fertier, L. Surface Acoustic
Wave Sensors for PM2. 5 and PM10 Concentration. Procedia Eng. 2016, 168, 696–699. [CrossRef]
8. Molenar, J.V. Theoretical Analysis of PM2.5 Mass Measurements by Nephelometry. Available online: http://vista.cira.
colostate.edu/improve/publications/graylit/014_AerosolByNeph/AerosolbyNeph.pdf (accessed on 9 August 2018).
9. Black, J.P.; Elium, A.; White, R.M.; Apte, M.G.; Gundel, L.A.; Cambie, R. MEMS-enabled miniaturized
particulate matter monitor employing 1.6 GHz aluminum nitride thin-film bulk acoustic wave resonator
(FBAR) and thermophoretic precipitator. IEEE Ultrason. Symp. 2007, 4, 476–479.
10. Thomas, S.; Cole, M.; Villa-López, F.H.; Gardner, J.W. High frequency surface acoustic wave resonator-based
sensor for particulate matter detection. Sens. Actuators A Phys. 2016, 244, 138–145. [CrossRef]
11. Rayleigh, L. On waves propagated along the plane surface of an elastic solid. Proc. Lond. Math. Soc. 1885, 1, 4–11.
[CrossRef]
12. Chen, B.T.; Feather, G.A.; Maynard, A.; Rao, C.Y. Development of a personal sampler for collecting fungal
spores. Aerosol Sci.Technol. 2004, 38, 926–937. [CrossRef]
13. Campbell, C. Surface Acoustic Wave Devices and Their Signal Processing Applications; Elsevier: New York, NY, USA, 2012.
14. Lindsley, W.G.; Schmechel, D.; Chen, B.T. A two-stage cyclone using microcentrifuge tubes for personal
bioaerosol sampling. J. Environ. Monit. 2006, 8, 1136–1142. [CrossRef] [PubMed]
15. Hoffman, A.C.; Stein, L.E.; Hoffmann, A.C. A closer look at centrifugal gas cleaning devices. In Gas Cyclones
and Swirl Tubes: Principles, Design, and Operation, 2nd ed.; Springer-Verlag: Heidelberg, Berlin, 2002.
16. Hoffman, A.C.; Stein, L.E.; Hoffmann, A.C. Particle motion. In Gas Cyclones and Swirl Tubes: Principles, Design,
and Operation, 2nd ed.; Springer-Verlag: Heidelberg, Berlin, 2002.
Micromachines 2018, 9, 398 11 of 11

17. Hand, J.L.; Kreidenweis, S.M. A new method for retrieving particle refractive index and effective density
from aerosol size distribution data. Aerosol Sci. Technol. 2002, 36, 1012–1026. [CrossRef]
18. Shi, Y.; Ji, Y.; Sun, H.; Hui, F.; Hu, J.; Wu, Y.; Fang, J.; Lin, H.; Wang, J.; Duan, H.; et al. Nanoscale
characterization of PM 2.5 airborne pollutants reveals high adhesiveness and aggregation capability of soot
particles. Sci. Rep. 2015, 5, 11232. [CrossRef] [PubMed]
19. Pitz, M.; Cyrys, J.; Karg, E.; Wiedensohler, A.; Wichmann, H.E.; Heinrich, J. Variability of apparent particle
density of an urban aerosol. Environ. Sci. Technol. 2003, 37, 4336–4342. [CrossRef] [PubMed]
20. Morawska, L.; Johnson, G.R.; Ristovski, Z.; Agranovski, V. Relation between particle mass and number for
submicrometer airborne particles. Atmos. Environ. 1999, 33, 1983–1990. [CrossRef]
21. Kulkarni, P.; Baron, P.A.; Willeke, K. Aerosol Measurement: Principles, Techniques, and Applications; John Wiley & Sons:
Hoboken, NJ, USA, 2011.
22. Jetter, J.J.; Guo, Z.; McBrian, J.A.; Flynn, M.R. Characterization of emissions from burning incense.
Sci. Total Environ. 2002, 295, 51–67. [CrossRef]
23. Kondoh, J.; Matsui, Y.; Shiokawa, S. New biosensor using shear horizontal surface acoustic wave device.
Jpn. J. Appl. Phys. 1993, 32, 2376. [CrossRef]

© 2018 by the authors. Licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC BY) license (http://creativecommons.org/licenses/by/4.0/).