L ABORATORY E XPERIENCE FOR S TUDYING A P IEZO MUMPS

D EVICE AS A M ULTIFUNCTIONAL S ENSOR E XPLOITING THE

P ROPERTIES OF THE A LUMINIUM N ITRIDE L AYER

Katia Samperi, Claudio Spitale 15/07/2021

katiasamperi@gmail.com
claudiospitale@gmail.com

1 Introduction
This report describes in detail the laboratory experiences we did. The objective of these ex-
perience is to verify if the Mems described in the first chapter is able to detect other physical
quantities such as DUV rays, UVA rays and temperature variations as well as vibrations. If
this is true, we are in presence of a multifunctional sensor, and it would be used to detect these
quantities or redesigned in an optimized way to detect these physical quantities. Therefore, a
MEMS, under the profile of Multifunctional Sensor, is a device capable of detecting multiple
physical quantities producing a variation on an output parameter such as voltage, current, fre-
quency, phase, etc. Our device could be seen as a transducer of physical quantities into voltage.
After that, a calibration is needed to go back getting from the output voltage a measure of that
physical quantity.
Therefore our question is: Is the MEMS realized in PIEZOMUMP’s technology a multifunc-
tional sensor? We will try to answer the question and in order to do that we will exploit some
properties of the AlN layer.
In this work, we made some experiences to determine the resonant frequency and to analyse
the response of our device to UVA, DUV rays and temperature variations.
The report is divided into three chapters. In the first chapter we will talk about the evaluation
of the MEMS resonant frequency, in the second chapter we will talk about the device response
to UVA and DUV light and in the last chapter we will talk about the device response to tem-
perature variations. Moreover, each chapter of this paper is generally divided into three parts:
state of the art, setup and experimental result. In the section ’state of the art’ is reported all the
theorical parts, therefore what we expect from that experience from the future acquired data.
In the ’set up’ section are listed briefly all the tools we needed and every step followed to carry
out the experiment. Lastly, in the section ‘experimental result’ are shown the graphs and data
obtained with a concluding discussion.

Page 1
 (a) Final result of PiezoMUMP’s technology steps (b) MEMS in PiezoMUMP’s technol-
and list of employed materials. ogy used in this experience.

Figure 1

2 MEMS in PiezoMUMPS technology

The MEMS used in our experience, shown in figure 1(b) is a cantilever realized in PiezoMUMPS
technology whose peculiarity is the presence of a piezoelectric material, Aluminium Nitride,
that is used to produce an electric output.
It is important to recall some process steps and characteristics that are strictly related to
the device properties. The PiezoMUMPS technology consist of several process steps which
employ the materials shown in figure 1. One of the deposited layer is the piezoelectric material,
Aluminium Nitride (AlN), which has a piezoelectric strain coefficient, d33, on the order of
[3.4 − 6.5] pC/N and a thickness of 0.5 µm. Since the AlN layer is responsible of many effects
such as pyroelectric effect and piezoelectric effect, in this experience, it will be the layer on
which we will focus our attention.
The physical structure, as it is highlighetd in Fig. is that of a simple Cantilever and will be
discussed in detail in the following subsection. it is important to note that for our purpose the
device structure is not so relevant, while the characteristics of the AlN layer are relevant.

2.1 Piezoelectric effect

"The piezoelectric effect is the ability of certain materials to generate an electric charge in re-
sponse to applied mechanical stress and it is reversable process" [2]. In other terms, when a
mechanical stress is applied to a piezoelectric material, the charges inside the material start
moving, with the consequent generation of an external electrical field (called direct piezo-
electric effect). On the contrary, an outer electrical field applied to a piezoelectric material
introduces stretches or compresses in it (called inverse piezoelectric effect).

Page 2
 Figure 2: Beam cantilever structure.

2.2 Beam cantilever in PiezoMUMPs technology

A beam cantilever (shown in Fig. 2) is an elastic body anchored at one extreme and free to
move in all the other points. It has a length much greater than its width and thickness, so that
the mechanical study can be simplified into a smaller number of dimensions.
A cantilever responds with a deformation when a force is applied on it.

2.2.1 U-Shaped beam cantilever in PiezoMUMPs technology used as current sensor

One particular structure that exploit its properties is the U-shaped beam cantilever. This cate-
gory of MEMS devices is typically used as current sensor, magnetic field sensor or accelerom-
eter. The following treatment is a summary of the key points described in [1]-[2].
When an electromagnetic field is applied, the Lorentz force will exert on the cantilever and
its expression is

F~L = q · ~E + q · ~v x~B (1)

where q is the charge of the particle, ~E is the electric field, ~v is the velocity of the particle and
~B is the magnetic induction. In practice the Lorentz force is induced by interaction between an
unknown electrical current, I, that produces the magnetic induction, ~B and a known current
Icant driven through the silicon cantilever beam. Consequently, the modulus of Lorentz Force
may be given by
µ0
FL = · Icant · I · L · sinθ (2)
2πr
where r is the distance between the cantilever tip and the wire with the unknown current and
L is the cantilever length (l4 in Fig. 2), θ is the angle between the magnetic field vector and the
beam length vector.
The Lorentz force in (1) deforms the microelectromechanical structure, and the resulting
displacement of the cantilever tip will be a function of the current to be measured. It is impor-
tant to define the transduction chain (Fig. 3) when it is used as the magnetometer as described

Page 3
 (a) Working principle. (b) Focus on the geometrical parameters.

Figure 3: 3D view of the U-shaped beam cantilever.

above: the electrical current to be evaluated is converted in a force, FL , then into a displace-
ment, z, of the cantilever beam free end, finally into a voltage, V, using a piezoelectric material
composing the MEMS device (Aluminium Nitride). The U-shaped beam cantilever exhibits
a dynamic behaviour which can be modelled using the well-known equivalent mass-spring
system equation which is a second-order differential equation

mz̈ + dż + ΛV (t) + kz = FL (3)

where m is the inertial mass of the microelectromechanical structure, d is the mechanical damp-
ing, Λ is the piezoelectric damping, k is the mechanical stiffness and V (t) is the voltage gener-
ated by the piezoelectric material as a response to the displacement, whose expression may be
given by
Z  
V (t)
V̄ (t) = Πż − +α
RC (4)
| {z }
V (t)

where Π is the coupling constant, RC is the load of the piezoelectric stack and α is an offset
voltage that has been added to the simulated output voltage, in order to take into consideration
the noise level.
The dynamic equation (3) is valid also for a simple cantilever.

Page 4
 Figure 4: Transduction chain for the U-shaped cantilever used as a current sensor.

2.2.2 Beam cantilever with different width Layers resonance frequency

The resonance frequency of a piezoelectric cantilever with different width layers has been eval-
uated in [7]. The starting point of this study is the motion equation of bending piezoelectric
cantilever which is given by Newton’s law:

∂2 ω ∂2 M p
m = (5)
∂t2 ∂x2
m= ∑ ρi ti Wi (6)
i
where m is the mass of the unit length of the piezoelectric cantilever, ρi , bi and ti are the mass
density, the width and the thickness of the i-th layer film respectively. Utilizing the moment
equilibrium, the total moment is equal to zero

∂2 ω
M=− K − Mp = 0 (7)
∂x2
K= ∑ Ki (8)
i
where Ki is the flexural rigidity of the i-th layer film and M p is the piezoelectric moment. Finally
combining () and () we get

∂4 ω ∂2 ω
K +m 2 = 0 (9)
∂x 4 ∂t
Applying the separated variable method and combining the boundary conditions of cantilever,
we can obtain the resonance frequency

β2i
r
K (10)
fi =
2πL2 m
where betai is a coefficient of modal function, whose value for the foundational frequency is
1.8746. It is important also to recall the expression of the flexural rigidity for each layer
Yi bi
Ki = ((zi − z N A )3 − (zi−1 − z N A )3 ) (11)
3
where Yi is the Young”s modul of the i-th layer, z N A is the position of the neutral plane and zi
is the position of the i-th layer.

Page 5
2.2.3 Beam cantilever in PiezoMUMPs technology used as temperature and UVC sensor

The objective of our experience is that of exploiting the properties of the aluminium nitride to
implement with the same device both a UVC and a temperature sensor in addition to a current
sensor, which is the function for which it has been designed. Some works whose reference will
be reported later have demonstrated that the Aluminium nitride is sensible to both temperature
variation and UVC rays. For this reason, we want to test our device to discover this relation
are valid in our case. The hypothetical transduction chains for UVC and temperature sensors
are shown in Fig. and respectively. The temperature sensor could be realized exploiting the
pyroelectricity of AlN layer. Pyroelectricity can be described as the ability of certain materials
to generate a temporary voltage when they are heated or cooled. It is important to note that
despite the case of the current sensor, every quantity causes a displacement in the cantilever is
here an interference quantity, hence it is strongly undesired. That’s the main challenge of our
experience.

(a) Transduction chain for the UVC sensor. (b) Transduction chain for the temperature sen-
sor.

Figure 5: Transduction chains for both the UVC and temperature sensors.

Page 6
3 MEMS resonant frequency evaluation

3.1 PLL
In Fig. 5 is represented the PLL block diagram. It consists of different components listed as
follows:

• Analog to Digital converter. The conversion of the input signal into a digital one is im-
portant for the cancellation of high frequency noise.

• Preamplifier Stage. It is needed to amplify the weak input signal. Indeed, the order of
magnitude of the input signal is about 10−3 V. It’s gain is K.

• two multipliers. It is a device that performs modulation. One multiplier executes the
modulation of the input amplified signal with the sinusoidal carrier signal at frequency
ωre f . The other multiplier executes the modulation of the input amplified signal with the
cosine carrier signal at frequency ωre f

• Low-pass filter that eliminates the harmonic components generated by the multiplier.

• an oscillator which produces the sinusoidal and cosine input signals of the multipliers at
frequency ωre f .

Figure 6: PLL block diagram

3.1.1 Working principle

Assumig, for simplicity, that the input amplified signal, x (t), is a sinusoidal signal at frequency
ωi

x (t) = Ksin(ωi t) (12)

Page 7
That signal follows two different paths. Along the upper path the signal is modulated by the
multiplier as follows

K K
x (t) · sin(ωre f t) = cos((ωi − ωre f )t) + cos((ωi + ωre f )t) (13)
2 2

Along the path below the signal is modulated by the multiplier as follows

K K
x (t) · cos(ωre f t) = sin((ωi − ωre f )t) + sin((ωi + ωre f )t) (14)
2 2

Choosing appropriately the cut off frequency of the low pass filter, the output signals, X and
Y, will be
K
X= cos((ωi − ωre f )t)
2 (15)
K
Y = sin((ωi − ωre f )t)
2
Therefore, we can evaluate module and phase of Z, that is defined as a complex number

Z = X + iY (16)

The module will be

p
|Z| = X2 + Y2 (17)

while the phase will be

Y π π
ϕ = tan−1 = tan−1 (tan((ωi − ωre f )t)) = (ωi − ωre f )t if − < (ωi − ωre f )t <
X 2 2
(18)

The most important signal is the phase, φ, since allows as to detect the resonant frequency. We
can observe from the expression (10) it is strictly linked to the frequencies of the input signal.
In particular, the phase is a linear function with respect to the time and the quantity (ωi − ωre f )
represents the slope of that linear function or straight line. This means, from a practical point
of view, that we can distinguish three main different cases as follows:

• when ωi > ωre f the slope of the straight line is positive;

• when ωi = ωre f the phase is constant and equal to zero. This means we have found the
resonant frequency;

• when ωi < ωre f the slope of the straight line is negative;

In the following chapter we will report the experimental results that confirm the above theory.

Page 8
3.2 Measurement experience on MEMS using the PLL
The objective of that experience is to find out the MEMS resonant frequency. We will talk about
the instruments configuration we have adopted for the PLL, oscilloscope, piezoelectric discs
and function generators. The strategy used to detect the resonant frequency is equivalent to the
step response study, but in our case the study of the step response would have been impossible
because the small voltage amplitude would have been indistinguishable from noise. Therefore,
we urged the device with a step function and its response has been evaluated through the PLL
to have an accurate result.

3.2.1 Setup

For this experience, as it is shown in Fig. 6, we used the following instrumentation:

• The PLL.

• two Function Generators.

• The cables: 2 BNC-BNC, 1 Bnc-Probe, 1 BNC-Crocodile.

• The Oscilloscope.

• Four piezomoves

Figure 7: Instrumentation used for this experience. In this photo we can see all the devices
listed in the setup section.

We connected one function generator to the PLL to set its reference frequency that will be
changed during the experience to find the resonant frequency, while we connected the other

Page 9
function generator to the piezoelectric discs sending a square wave with a frequency of 10
Hz and an amplitude of 10 V. In this way, the four piezoelectric discs began to vibrate and
the MEMS gives an higher output voltage. Afterwards, we connected mems output to the
PLL using crocodiles-BNC cables. The PLL allowed us through the output "display" to get the
graphs on the oscilloscope display. The objective is to urge the device with all the frequency
harmonics (a step function contains all the harmonics) to capture its resonant frequency or
the frequency at which the response has an higher amplitude. We have chosen the following
appropriate parameters values for the PLL:

• the low pass filter cut off frequency, τ = 1 and slope, roll − o f f = 24dB.

• the reference frequency, f re f , as been changed manually through the function generator
in order to find the resonant frequency or the frequency at which the phase is null.

3.2.2 Experimental results

The results obtained are shown in Fig. 7 and 8 with a detailed description.

(a) Phase signal when ωre f = 119 · 2π rad/s. Accord- (b) Phase signal when ωre f = 121 · 2π rad/s. Accord-
ing to the theory, since the slope is positive, we expect ing to the theory, since the slope is negative, we expect
that ωi > ωre f . that ωi < ωre f .

Figure 8: PLL Phase signal. The behaviour of that signal is consistent with what we expected.

Page 10
 (a) Phase signal when ωre f = 119 · 2π rad/s. According to the theory, since the slope is positive, we expect
that ωi > ωre f .

(b) Phase signal when ωre f = 120 · 2π rad/s. We can observe the signal is almost constant and approxi-
mately null. Therefore, we have found the resonant frequency and according to the theory we are in the
case ωi ≈ ωre f = 120 · 2π rad/s.

(c) Phase signal when ωre f = 121 · 2π rad/s. According to the theory, since the slope is negative, we expect
that ωi < ωre f .

Figure 9: PLL Phase signal. The behaviour of that signal is consistent with what we expected.
Page 11
 (a) Device layer structure (b) I-V characteristic of an AlN MSM photodetector.

Figure 10: Device layer structure and its dark current versus applied bias voltage studied in
[3].

4 UV response

4.1 Aluminium nitride UV response: state of the art

The AlN possesses some fundamental properties which can be exploited for different applica-
tions such as high breakdown voltage, high DUV (deep ultraviolet) to visible rejection ratio and
high responsivity. AlN is the subject of several studies for the development of deep ultraviolet
(DUV), vacuum UV (VUV), and extreme UV (EUV) detectors thanks to its particular proper-
ties. Indeed, it possesses the widest direct energy bandgap (6.1 eV) among all semiconductors
and a very high thermal conductivity.
In [3] the operation of 200 nm DUV photodetectors through AlN metal-semiconductor-
metal (MSM) photodetectors with very high DUV to visible rejection ratio has been demon-
strated. The device layer structure employed in this study [3] is schematically shown in Fig. 9
(a), which utilizes an undoped AlN epilayer as an active layer. Undoped AlN of about 1.5 um
thick were grown by MOCVD on sapphire substrates.
The typical I-V characteristics of AlN epilayer based MSM detectors are shown in Fig. 9
(b). The device exhibits a very low dark current1 and is able of substaining a very high voltage
value of about 200 V.
The spectral responses have been measured in [3] at different bias voltages, and the result is
shown in Fig. 10. "AlN MSM detectors exhibit a peak responsivity of 200 nm, an extremely
sharp cutoff wavelength around 207 nm, and more than four orders of magnitude of DUV
to UV/visible rejection ratio" [3].
The data of Fig. 10 suggest that MSM detector has a selective and high gain around 200
1 "It consists of the charges generated in the detector when no outside radiation is entering the detector. Physically,
dark current is due to the random generation of electrons and holes within the depletion region of the device."

Page 12
Figure 11: Spectral response of an AlN MSM detector at 30 V. The inset shows the detector
responsivity as a function of the applied bias.

nm of wavelenght, while it exhibits a strong attenuation as the wavelength increases. This

behavior is desirable since we want to capture DUV rays only, shielding the rays with a higher
wavelength. That may be attributed to the presence of dislocations or deep level defects in the
epilayers and its mechanism remains to be investigated. It is important to note that this is so
far the shortest cut-off wavelength achieved for semiconductor detectors.

4.2 Measurement experience with DUV lamp and UVA lamp

The objective of this experience is to demonstrate the responsivity of AlN layer inside the
MEMS. According to the above article, although we have a thinner layer with respect that
of the study, we expect that the device will provide a voltage as response to DUV light. The
thick of AlN layer in MUMPS technology is 0.5 µm. Surely, the device is not thought for this
purpose, thus probably the result will differ from the one in the paper. We will also show the
device response to UVA beam. We expect that there won’t be a voltage variation when the de-
vice is exposed to UVA beam since its responsivity should be very low for such a wavelength
according to the article [4].

4.2.1 Setup

For this experience, as we can see in Fig. 11, we used the following instrumentation:

• UVC and UVA lamps.

• one Data Acquisition device.

• one computer.

Page 13
 • a mechanical support to keep the device at a certain distance from the lamp

The MEMS was connected to the Data acquisition device which sent the processed data to the
PC through an USB cable. A LabVIEW program was prepared to recover the output device
response for a certain desired time interval.

4.2.2 Experimental results

In this study the device has been positioned in parallel with respect the beam direction. In this
way the beam has a direct access to the AlN layer inside the device, while if we positioned the
device in a perpendicular direction with respect the beam, the beam would be shielded by the
metal layer. The results are shown in figure 10.
The graph in Fig. 10 (a) shows the device response when the UVC lamp is on. Since the lamp is
a source of electromagnetic noise that could be seen as an interfering input, we must isolate the
UVC beam from that noise. For this purpose, we covered the UV lamp with a tick cardboard,
since the UVC beam can’t pass through this obstacle. Successively, we recovered the UVC re-
sponse in the following way: for approximately twenty seconds the lamp was covered with
cardboard (in figure we refer to this time interval with the word ’UVC OFF’) and for approx-
imately others twenty seconds the lamp wasn’t covered with cardboard (in figure we refer to
this time interval with the word ’UVC ON’). The same procedure was repeated for about two
minutes. As we can see the result is coherent with what we expected. The information seems
to be contained into the slope of the curve. Indeed, when the UVC beam is off the slope is neg-
ative, while, when the UVC beam is on the slope is positive. Moreover, we could attribute the
big oscillation in both cases (UVC on and UVC off) to the noise generated by the lamp. Thus,
we can say the aluminium nitride into the device reacts to that input producing a voltage.
The graph in figure 10 (b) shows the device response when the UVA lamp is on. Since the lamp
is a source of electromagnetic noise that could be seen as an interfering input, we must isolate
the UAC beam from that noise. For this purpose, we covered the UV lamp with a tick card-
board, since the UVC beam can’t pass through this obstacle. We recovered the UVA response
in the following way: for approximately twenty seconds the lamp was covered with cardboard
(in figure we refer to this time interval with the word ’UVA OFF’) and, for approximately oth-
ers twenty seconds, the lamp wasn’t covered with cardboard (in figure we refer to this time
interval with the word ’UVA ON’). The same procedure was repeated for about two minutes.
As we can see the result is coherent with what we expected. Indeed, in both cases the response
is characterized only by oscillations due to the noise generated by the lamp and the two cases
are indistinguishable. Thus, we can say the aluminium nitride into the device doesn’t react to
that input.

Page 14
 (a) UVC lamp, device and DAQ.

(b) UVA lamp, device and DAQ. (c) Device under UVA beam.

Figure 12: Instrumentation used for this experience. In this photo we can see all the devices
listed in the setup section.

Page 15
 (a) The graph shows the device response either when the UVC lamp is on or off.

(b) The graph shows the moving average of the device response either when the UVC lamp is on or off. It takes the
average of every 1000 consecutive samples of the waveform with the help of a Matlab f unction2 .

(c) The graph shows the device response when the UVA lamp is either on or off.

Figure 13 Page 16
Figure 14: The picture shows the results of the previous fig. (5) together. We observe the
oscillation amplitude and the mean value of the blue curve are higher than the ones of the
orange curve, probably because the DUV lamp generates a bigger noise with respect the UVA
lamp.

Page 17
Figure 15: Pyroelectric voltage signal vs temperature modulation measured for Au/Ti/Al-
N/Si/Ti/Au structure in [5].

5 Temperature variation response

5.1 Aluminium nitride temperature variation response: state of the art

The wide- band gap materials such as GaN and AlN have a high thermal stability. Indeed,
there are some studies on their pyroelectric effects. One of these was managed by the authors
of the article [5] who studied the pyroelectric properties of high-quality (0001) AlN films grown
on (111) Si. They measured the pyroelectric coefficient using the dynamic method. The pyro-
electric coefficient was calculated as

I
ρ= = 6 − 8µC/(m2 K ). (19)
2π f A∆T

where I is the pyroelectric current, f is the modulation frequency, A is the capacitor surface
area, and ∆T is the change in sample temperature. "The sample temperature was varied 15–35
°C using a two-stage thermoelectric cooler/heater configuration. One stage controlled the aver-
age sample temperature while the second stage imposed a small sinusoidal temperature mod-
ulation. The modulation frequency was 100 mHz and had a peak-to-peak value in the range of
0.20–0.30 °C." [5] In their experience was observed that the voltage response follows linearly
the temperature variation without any significant delay (fig. 12).

5.1.1 Equivalent circuit model of the pyroelectric crystal

The equivalent circuit model of the pyroelectric crystal, studied in [6], is shown in Fig 14. It
consists of a temperature-dependent voltage source in series with the capacitance of the elec-
troded crystal, shunted by the leakage resistance Rl as shown in Fig. 2. The voltage source

Page 18
vd (t) may be given by

Aλ
vd (t) = T (t) (20)
Cc d

where λ represents the pyroelectric coefficient of the crystal material and d is defined as the
rate of change with respect to temperature of the spontaneous polarization (at constant stress),
Td (t) is the crystal time-dependent temperature, A is the area of the large face Cc is the electrical
capacitance. for the crystal element and is defined as

e r e0 A
Cc = (21)
t

where er is the relative permittivity, or dielectric constant of the material in the direction of the
spontaneous polarization axis, and e0 to is the permittivity of free space.
The leakage resistance Rl is defined as

t
Rl = (22)
σA

where σ is the DC leakage conductance plus the ac loss conductance for the material. The
Thevenin equivalent series circuit with voltage source vd (t) can be converted to a Norton
equivalent. with parallel connected current source id (t) of value.

dTd (t)
id (t) = Aλ (23)
dt

"Clearly the pyroelectric current varies directly with the time rate of change of crystal tem-
perature, while the pyroelectric voltage varies linearly with its temperature"[6]. This ob-
servation is important since from our experience we will expect that the output voltage will
change linearly with temperature.

Figure 16: Pyroelectric equivalent model of a simple crystal structure. P is the spontaneous
polarization vector, A is the area of the large face and t is the thickness.

Page 19
Figure 17: Instrumentation used for this experience. This photo shows the complete set of
devices listed in the setup section.

5.2 Measurement experience with thermal oven

The objective of this experience is that of finding the response of the device to temperature
variations. As we said in the previous section, we expect the output voltage will follow the
temperature variation with a linear relation.

5.2.1 Setup

For this experience, as we can see in figure, the following instrumentation has been used:

• A thermal oven.

• one Data Acquisition device.

• one computer.

5.2.2 Experimental results

The oven temperature profile and the voltage response to that temperature prfile are shown in
Fig. 17 and 18 (a) respectively.
The oven temperature profile has been set as follows: it varies linearly from 50 °C to 100 °C
in about three minutes and then it remains constant for about one minute. This choice gives
rise to a curve slope inside the time interval [0 s,150 s] of 0.33 °C/s.

Page 20
 Temperature variation profile

100
Temperature (°C)

87.5

75

62.5

50
0 50 100 150 200 250
time (s)

Figure 18: The graph shows the temperature variation profile it varies linearly from 50 °C to
100 °C in about three minutes and then it remains constant for about one minute. Therefore,
the curve slope inside the time interval (0 s,150 s) is 0.33 °C/s.

The measured device voltage response shows a linear behaviour with respect to tempera-
ture variation. From figure 18 (a) the curve slope can be derived and, in the interval of interest,
[0 s,150 s], it results to be 2.67 mV/s. After this time interval the voltage remains constant.
From Fig. 18 (a) it can be also noted that a large amplitude noise is overlapped to the useful
signal. For this reason, with the help of Matlab2 the moving average of the voltage response in
fig 18 (a) has been performed and the resulting curve is shown in fig. 15 (b).
Finally, in Fig. 18 is shown the moving average of voltage response versus temperature.
From this graph we can link directly the temperature value to the corresponding voltage value,
but it should be clear that this correspondence depends on the temperature profile slope. The
main important information in contained into the voltage curve slope since it is proportional
to the temperature profile slope with a certain coefficient. If we change the temperature profile
slope, the voltage response will also change its slope. The obtained experimental results are
coherent with the pyroelectric equivalent model and theoretical analysis.
2 Matlab functions for implementing the moving average over a certain number of samples (var. sample) of the
desired voltage values array (var. array). x is the time array whose dimension is equal to that of the voltage array.
x=0.001:0.001:146
sample= 1000;
coeff24hMA = ones(1, sample)/sample;
avg24hTempC = filter(coeff24hMA, 1, array);
plot(x,[VarName2 avg24hTempC])

Page 21
 Voltage response to Temperature variation
1

0.5
Amplitude (V)

-0.5

0 20 40 60 80 100 120 140 160 180 200 220

time (s)

(a) The graph shows the device voltage response. We observe the response follows linearly
the temperature variation. The curve slope in the interval of interest, (0 s,150 s), is 2.67
mV/s. After this time interval the voltage remains constant.
Moving Average of Voltage Response to Temperature Variation Versus Time
0.5

0.4

0.3
Voltage (V)

0.2

0.1

-0.1

-0.2
0 20 40 60 80 100 120 140 160 180 200 220
Time (s)

(b) The graph shows the moving average of the voltage response in fig 15 (a). The moving
average is taken over every 1000 consecutive samples of the waveform with the help of a
Matlab function2 .

Figure 19: Measured device voltage response to temperature variation (whose profile is shown
in fig. 14) and its moving average versus time.

Page 22
 Moving Average of Voltage Response versus Temperature
0.5

0.4

0.3
Voltage (V)

0.2

0.1

-0.1

-0.2
0 12.5 25 37.5 50
Temperature (°C)

Figure 20: Moving average of voltage response of fig. 15 (a) versus temperature assuming
that the temperature profile variation has a slope of 0.33 °C/s into the time interval under
observation.

5.2.3 Pyroelectric coefficient calculation

Using the expression (9) and (10) we can estimate the pyroelectric coefficient

vd (t) v ( t ) e r e0 A v ( t ) e r e0
λ= Cc = d = d (24)
Td (t) A Td (t) A t Td (t) t

where t, the AlN thickness, is given by the technology t = 0.5 µm, er e0 is the dielectric constant
of AlN and is equal to er e0 = 7.7 · 10−11 . The ratio vd (t)/Td (t) was estimated from this expe-
rience assuming the oven temperature is equal to the device temperature and was found to be
vd (t)/Td (t) ≈ 0.01 mV/°C. Substituting these numbers into the previous expression we get

v d ( t ) e r e0
λ= ≈ 1.5µC/(m2 K ). (25)
Td (t) t

This result differs from the one of the article [2], but this is probably due to the different thick-
ness of the AlN layer. The thickness of the AlN layer used in [2] is t = 0.3 µm.

Page 23
Conclusion
An integrated microsensor for measuring UVC and temperature variation has been studied
and experimentlly valideted during this experience.
The most interesting advantage is represented by the possibility, offered by the piezoelec-
tric material embedded in the PiezoMUMPs technology, to convert the measurand in an electric
signal straight available in the output of the sensor. The experimental results obtained in this
experience shows a very good agreement with that theoretically predicted. We have demon-
strated qualitatively the possibility of using this MEMS for detecting more than one quantity.
In particular, the device is able of detecting UVC rays and temperature variation in addition
to displacements, thanks to the properties of the Aluminium Nitride layer. Indeed, the ex-
perimental results shows that it generates a voltage variation as response to UVC beam and
temperature variation. Moreover, during the experiences, when possible, the interfering in-
puts have been shielded.
Future works are necessary to optimize the device structure for detecting the three inputs
and realizing a more efficient multifunctional sensor. The results obtained encourage possi-
ble future works to optimize the device design in order to improve the performances. Further
research efforts should exploits the big advantages of piezoelectric and pyroelectric transduc-
tion mechanisms for measuring dc current and temperature variation respectively, but also the
ability of Aluminium nitride of etecting UVC rays.

Page 24
References
[1] C. Trigona, V. Sinatra, G. Crea, B. Andò and S. Baglio, "Characterization of a PiezoMUMPs
Microsensor for Contactless Measurements of DC Electrical Current," in IEEE Transactions on
Instrumentation and Measurement, vol. 69, no. 4, pp. 1387-1396, April 2020.

[2] V. M. Sinatra, "Development of innovative magnetic field sensors with tuning features for
a wide operative range", PhD Thesis, Doctor of Philosophy in System, Energy, Computer and
Telecommunication Engineering – XXXII Cycle, March 2020.

[3] V. Fuflyigin, E. Salley, A. Osinsky, and P. Norris , "Pyroelectric properties of AlN", Applied
Physics Letters 77, 3075-3077 (2000)

[4] J. Li and Z. Y. Fan III-N Technology, Inc., Manhattan, Kansas 66502 R. Dahal, M. L Nakarmi,
J. Y. Lin, and H. X. Jiang Department of Physics, Kansas State University, Manhattan, Kansas
66506-2601 , "200nm deep ultraviolet photodetectors based on AlN", Applied Physics Letters
89, 213510 (2006)

[5] Kansho Yamamoto, Fabian Goericke, Andre Guedes, Gerardo Jaramillo, Takuo Hada, Al-
bert P. Pisano, and David Horsley , "Pyroelectric aluminum nitride micro electromechanical
systems infrared sensor with wavelength-selective infrared absorber", Applied Physics Letters
104, 111111 (2014).

[6] W. P. Wheless, L. T. Wurtz and J. A. Well, "An equivalent-circuit radiation sensor model,"
Proceedings of SOUTHEASTCON ’94, 1994, pp. 7-11, doi: 10.1109/SECON.1994.324253.

[7] Liu, Z.; Chen, J.; Zou, X. Modeling the Piezoelectric Cantilever Resonator with Different
Width Layers. Sensors 2021, 21, 87.

Page 25