micromachines

Communication
A High-Performance MEMS Accelerometer with an Improved
TGV Process of Low Cost
Yingchun Fu 1,2 , Guowei Han 3 , Jiebin Gu 4 , Yongmei Zhao 3,5 , Jin Ning 3,5 , Zhenyu Wei 3,5 , Fuhua Yang 3,5
and Chaowei Si 3, *

1 Beijing Smart-Chip Microelectronics Technology Co., Ltd., Beijing 100192, China; [email protected]
2 Zhongguancun Xinhaizeyou Technology Co., Ltd., Beijing 100192, China
3 Institute of Semiconductors, Chinese Academy of Sciences, Beijing 100083, China;
[email protected] (G.H.); [email protected] (Y.Z.); [email protected] (J.N.);
[email protected] (Z.W.); [email protected] (F.Y.)
4 State Key Laboratory of Transducer Technology, Shanghai Institute of Microsystem and Information
Technology, Chinese Academy of Sciences, Shanghai 200050, China; [email protected]
5 Center of Materials Science and Optoelectronics Engineering, University of Chinese Academy of Sciences,
Beijing 100049, China
* Correspondence: [email protected]; Tel.: +86-10-8230-5147

Abstract: High-performance MEMS accelerometers usually use a pendulum structure with a larger
mass. Although the performance of the device is guaranteed, the manufacturing cost is high. This
paper proposes a method of fabricating high-performance MEMS accelerometers with a TGV process,
which can reduce the manufacturing cost and ensure the low-noise characteristics of the device. The
TGV processing relies on laser drilling, the metal filling in the hole is based on the casting mold and
CMP, and the packaging adopts the three-layer anodic bonding process. Moreover, for the first time,
the casting mold process is introduced to the preparation of MEMS devices. In terms of structural
design, the stopper uses distributed comb electrodes for overload displacement suppression, and
the gas released by the packaging method provides excellent mechanical damping characteristics.
Citation: Fu, Y.; Han, G.; Gu, J.; Zhao,
The prepared accelerometer has an anti-overload capability of 10,000 g, the noise density is less than
Y.; Ning, J.; Wei, Z.; Yang, F.; Si, C. A √
0.001◦ / Hz, and it has ultra-high performance in tilt measurement.
High-Performance MEMS
Accelerometer with an Improved
Keywords: MEMS accelerometer; TGV; distributed stoppers; comb structure
TGV Process of Low Cost.
Micromachines 2022, 13, 1071.
https://doi.org/10.3390/mi13071071

Academic Editor: Kun Li 1. Introduction

Received: 2 June 2022 MEMS devices have the advantage of being low cost and small in size, and they
Accepted: 2 July 2022 can be fabricated in batches. They are also successful sensors for various environmental
Published: 5 July 2022 monitoring uses. MEMS inclination sensors are widely used in buildings, bridges, power
Publisher’s Note: MDPI stays neutral
transmission networks, automobiles, and other engineering vehicles [1,2]. Accelerometers
with regard to jurisdictional claims in
with high performance and high reliability have great market value.
published maps and institutional affil-
At present, the accelerometer with the best performance is the sandwich structure
iations. from Murata and Safran. The devices of Colibris are based on three-layer silicon bonding
and have excellent temperature stability [3], but the preparation cost is high, and it is
often used in high-performance fields such as aerospace engineering [4]. The devices from
Murata use a silicon glass composite cover plate as the upper and lower electrodes, and the
Copyright: © 2022 by the authors. electrodes are grown on the sidewalls through a hard mask, which achieves a good balance
Licensee MDPI, Basel, Switzerland. between performance and cost, and has a wide range of applications in the automotive and
This article is an open access article industrial fields [5]. However, since Murata’s structure requires electrodes to be grown on
distributed under the terms and the sidewalls, it is not possible to screen the yield of devices on the wafer [6], and the cost
conditions of the Creative Commons is still higher than that of consumer-grade accelerometers.
Attribution (CC BY) license (https://
In this paper, an accelerometer with comb-tooth structures is proposed. It is imple-
creativecommons.org/licenses/by/
mented based on SOG process, and the structure layer adopts 100 µm silicon to ensure the
4.0/).

Micromachines 2022, 13, 1071. https://doi.org/10.3390/mi13071071 https://www.mdpi.com/journal/micromachines

Micromachines 2022, 13, 1071 2 of 9

size of the mass block, thereby reducing the mechanical noise of the device. The packaged
capacitor electrodes are led out to the surface of the device through vertical leads, which
can meet wafer-level tests and reduce production costs. Relying on the distributed stopper
set between the comb teeth, the sensitive mass displacement is constrained, depending
on the distributed stopper structure set between the comb teeth. Metal filled in through
vias on BF33 glasses serves as the vertical lead cover, the anodic bonding process achieves
hermetic packaging, and the released gas provides additional damping, making the device
excellent in vibration resistance.
The proposed process solution has the capability of low-cost mass production and can
guarantee the performance of the device, which is extremely innovative.

2. Structure Design
From the perspective of detection accuracy, the sensitivity of capacitance change per
unit acceleration needs to be considered when designing the tilt angle structure, which
is determined by the resonant frequency of the structure, the distance between the ca-
pacitor plates, and the total capacitance. Under the action of inertial force, the response
displacement of the accelerometer can be obtained from Newton’s second law, as follows:

x = a/ω02 (1)

where x is the displacement under the acceleration a, and ω 0 is the resonant frequency of
the accelerometer. If C0 is the initial capacitor, then d0 is the initial distance of capacitor
plates, and the capacitance change ∆C is

∆C = xC0 /do (2)

The accelerometer resolution is determined by its mechanical noise, and the noise 〈an 〉
is related to the resonant frequency and the size of the mass m [7]. Where k is the Boltzmann
constant, T is the ambient temperature, m is the mass of the movable structure, and Q is the
Quality factor. p
h an i = 4kTω0 /mQ (3)
To ensure sensitivity, the typical resonant frequency of an inclination accelerometer is
several hundred hertz. Considering the anti-overload capability of the accelerometer, the
shock response displacement should not be too large, which is designed to be 1 KHz here.
To ensure that the accelerometer has a detection accuracy of 0.001◦ , its output electrical
√
noise should be less than 3 µg/ Hz, and the corresponding mechanical noise should be at
the µg level.
Considering the over-damping situation, the quality factor of the accelerometer is
less than 1, and the mass of the accelerometer should be greater than 0.12 mg. At present,
the detection capability of the accelerometer detection circuit for differential capacitance
is at the zF level. Considering the tolerance, it is hoped that 0.001◦ can correspond to a
capacitance change rate of 10 zF, then the capacitance sensitivity of an ideal inclination
accelerometer is expected to reach 0.57 pF/g.
For a 100 µm-thick silicon structure MEMS device, considering the ability of deep
silicon etching, the capacitor plate spacing should be greater than 3 µm. Considering the
processing accuracy, the capacitor plate spacing is set to 4 µm, the length of the plate is
150 µm, and the width is 6 µm. The overlap length is 140 µm. One anchor point can support
80 pairs of comb capacitor plates with a capacitance of 1 pF. The capacitor plates are shown
in Figure 1.
Micromachines 2022, 13, x FOR PEER REVIEW 3 of 9

Micromachines 2022,
Micromachines 2022, 13,
13, 1071
x FOR PEER REVIEW 33of
of 99

Figure 1. The comb capacitor.

1. The comb
Figure 1.
Figure comb capacitor.
The The
designedcapacitor.
accelerometer has four and a half pairs of comb structures as detection
capacitors on one side.
The designed
designed The total has
accelerometer capacitance andisaa half
four and 4.5 pF,
half pairstheofsize
combof structures
the deviceas is detection
2.4 × 3.3
mm 2,The
the size of the accelerometer
mass block is hasmg,
0.6 four the resonant pairs of comb
frequency of structures
the device as
is detection
1.14 kHz,
capacitors on is 2.4 × 2
capacitors
and the onone
oneside.
theoretical side. The
Thetotal
mechanical totalcapacitance
capacitance
noise is 1.43
is 4.5 pF, pF,
is 4.5
μg/√Hz.
the size
The the of the
size ofdevice
displacement the device
of the is 3.3
mass
mm
× 3.3,
2.4block
the size
mm 2, theofsize
the of
mass
theblock
mass is 0.6 mg,
block themg,
resonant
isacceleration
0.6 frequency of the device
of theisdevice
1.14 kHz, and the
caused by the unit gravitational √the resonant
is the frequency
capacitance sensitivity of is 1.14
0.19 kHz,
μm/g,
theoretical
and mechanical
the theoretical
theoretical noise is 1.43
mechanical noise µg/ Hz.μg/√Hz.
is 1.43 The displacement of the mass
Thealldisplacement of the block caused
and
by thethe unit gravitationalcapacitance sensitivity
acceleration is theiscapacitance
0.71 pF/g, of whichofmeet
sensitivity the mass
0.19 µm/g, design block
and there-
caused
quirements.by the unit gravitational acceleration is the capacitance sensitivity of 0.19 μm/g,
theoretical capacitance sensitivity is 0.71 pF/g, all of which meet the design requirements.
and Thethe theoretical capacitanceinsensitivity is 0.71 pF/g, of all of whichwas meet the design re-
The anti-overload
anti-overloaddesign design inthe thedetection
detectiondirection
direction ofstructure
structure wasrealized
realizedby bydis-
dis-
quirements.
tributed
tributed stopper
stopper structure.
structure. According
According to to Yoon’s
Yoon’s conclusions
conclusions [8],[8], when
when thethe stopper
stopper collides
collides
with Thestructure,
anti-overload design in the detection direction of structure was realized by dis-
with thethe structure, if if aa flexible
flexible collision
collision cancan be be achieved,
achieved, the the stress
stress caused
caused by by the
the collision
collision
tributed
will stopper structure. According to Yoon’s conclusions [8], when the stopper collides
will be
be greatly
greatly reduced,
reduced, and and the
the overload
overload resistance
resistance of of the
the structure
structure will
will be
be improved.
improved. In In
with the to
addition structure,
the low if a flexible
resonant collisionof
frequency canthe bestructure
achieved, the stress
itself, the causedfrequency
resonant by the collision
of an
addition to the low resonant frequency of the structure itself, the resonant frequency of
will be greatly
electrode plate reduced,
is about andkHz,
100 the overload
which is aresistance
good of the
choice as astructure
stopper willTherefore,
[9]. be improved. in thisIn
an electrode plate is about 100 kHz, which is a good choice as a stopper [9]. Therefore, in
addition
paper, to the low
a hypotenuse resonant
was was frequency
designed of
between the structure
the mass itself,
block the
and andresonant
the electrodefrequency
plate of an
to re-
this paper, a hypotenuse designed between the mass block the electrode plate to
electrode
duce thetheplate is about
distance between 100them
kHz,andwhich theisthea good choice
collision distanceas was
a stopper [9]. Therefore,
reduced, as shown inFig-
in this
reduce distance between them and collision distance was reduced, as shown in
paper,
ure 2. a2.hypotenuse was designed between the mass block and the electrode plate to re-
Figure
duce the distance between them and the collision distance was reduced, as shown in Fig-
ure 2.

Figure
Figure 2.
2. Stoppers
Stoppers located
located on
on electrodes.
electrodes.
Figure
The2. Stoppers located
stopper has twoon electrodes.
other advantages. First, the collision between the capacitor plate
The stopper has two other advantages. First, the collision between the capacitor plate
and
and the
thehypotenuse
hypotenuseisisananedge contact,
edge which
contact, does
which doesnotnot
easily cause
easily adhesion;
cause second,
adhesion; alt-
second,
hough The
thestopper hasdistance
collision two other advantages.
between the First, the collision
hypotenuse and the between
electrodethe capacitor
plate is plate
reduced,
although the collision distance between the hypotenuse and the electrode plate is reduced,
and
the the hypotenuse
the stopper
stopper is locatedisin
is located anan
in edge
an contact,
open
open which
area and does nottime
the etching easily cause
will adhesion;
not be second, alt-
increased.
hough the collision distance between the hypotenuse and the electrode plate is reduced,
the stopper is located
3. Fabrication
3. Fabrication Processin an open area and the etching time will not be increased.
Process
The designed
The designed inclination
inclination accelerometer
accelerometer was was prepared
prepared by by the
the TGV
TGV process.
process. TheThe metal
metal
3. Fabrication
filling in the Process
TGV was realized by the electroforming method, the structure layer was
filling in the TGV was realized by the electroforming method, the structure layer was pre-
The
prepared designed
by 110 inclination
µm-thick accelerometer
(100) crystal was
silicon, and prepared
the by the
resistivity
pared by 110 μm-thick (100) crystal silicon, and the resistivity is less than 0.005 Ω cm. TGV
is less process.
than The
0.005 Ω metal
cm.
filling in the
First,
First, theTGV
the was photoresist
AZ6130
AZ6130 realized
photoresistby was
the
waselectroforming
patternedon
patterned on110method,
110 µm-thick
μm-thick the monocrystalline
structure layer was
monocrystalline pre-
silicon,
silicon,
pared by
10 μm-deep
10 110 μm-thick
silicon was
µm-deep silicon (100)
was etched, crystal
etched, and silicon,
and the and
the bonding the
bonding area resistivity
area was
was left. is less than 0.005
left. Magnetron-sputtered
Magnetron-sputtered ITO Ω cm. ITO
served First,
as the
the AZ6130
sacrificial photoresist
layer, and was patterned
stripping and on 110
cleaning μm-thick
were monocrystalline
carried
served as the sacrificial layer, and stripping and cleaning were carried out. Then, the sili- out. Then, the silicon,
silicon
10 μm-deep
wafer
con waswas
wafer silicon
bonded to
bonded was
a to etched,
300 and the
a µm-thick
300 μm-thickBF33bonding
glass.
BF33 area was left. Magnetron-sputtered ITO
glass.
served ITO as the
also sacrificial
served as layer,
the maskandforstripping
structural
ITO also served as the mask for structural etching and cleaning
etching [10]. were
When
[10]. When carried
ITO ITO out.
is used Then,
is used athe sili-
as aassacrificial
sacri-
con
layer,wafer
the was bonded
material is to a
transmits300 μm-thick
light, and BF33
the glass.
etching of the bottom
ficial layer, the material is transmits light, and the etching of the bottom of the silicon can of the silicon can be
ITO
observed also
from served
the as
glass the mask
layer, for
avoiding structural
the etching
occurrence [10].
of When
over-etching.
be observed from the glass layer, avoiding the occurrence of over-etching. In addition, ITO isInused as
addition, a sacri-
ITO
materials
ficial layer,with good conductivity
the material is transmitscan absorb
light, and thetheetching
charged of particles
the bottom ofofthetheetching
silicon gas, can
reducing
be observed the from
footprint
the phenomenon caused by
glass layer, avoiding thethe reflection of
occurrence of over-etching.
F+ particles. In addition,
Micromachines 2022, 13, x 2022,
Micromachines FOR PEER REVIEW
13, x FOR PEER REVIEW 4 of 9 4 of 9

ITO materials with good

ITO materials conductivity
with can absorb
good conductivity canthe charged
absorb particles particles
the charged of the etching
of thegas,
etching4 gas,
Micromachines 2022, 13, 1071 of 9
reducingreducing
the footprint phenomenon caused by the reflection of F+ particles.
the footprint phenomenon caused by the reflection of F+ particles.

The release
Theofrelease
the structure was carried
of the structure wasout in sulfuric
carried out inacid–hydrogen peroxide.peroxide.
sulfuric acid–hydrogen After After
rinsing with The
purerelease
water, ofitthe
wasstructure
boiled was
in carried outalcohol
MOS-grade in sulfuric
for acid–hydrogen
30 min to peroxide.
displace the After
rinsing with pure water, it was boiled in MOS-grade alcohol for 30 min to displace the
rinsing
water in water with
the structure,pure water,
and dried it was boiled
ondried
a hot on in
plate.MOS-grade
The alcohol
preparation for 30 min to
process isprocessdisplace
Fig-the in
shownisinshown water
in the structure, and a hot plate. The preparation Fig-
ure 3. in the structure, and dried on a hot plate. The preparation process is shown in Figure 3.
ure 3.

(a) (a) (b) (b) (c) (c)

Figure 3.Figure
Preparation
Figure of the accelerometer
3.3. Preparation
Preparation of the
of structure:structure:
theaccelerometer
accelerometer (a) preparation
structure: of structures;
(a) preparation
(a) preparation (b) structures
of structures;
of structures; (b)structures
(b) structures
etching; (c) structures
etching;(c)
etching; release.
(c)structures
structuresrelease.
release.

The advantage
The of ITO as
The advantage
advantage ofan
of ITO
ITOetching
as
as an mask
an etching
etching and
masksacrificial
mask and layer material
and sacrificial
sacrificial layer is that the
layer material
material isis that
that the
the
materialmaterial
has different
materialhas corrosion
hasdifferent resistance
differentcorrosion properties
corrosionresistance to
resistanceproperties silicon
propertiesto and
tosiliconglass.
siliconand When
andglass.
glass.etched
Whenin
When etched
etchedin in
an acid solution,
an
anacid it can beititeffectively
acidsolution,
solution, can
canbe removedremoved
beeffectively
effectively withoutwithout
removed affecting
withoutaffectingthe structure.
affecting the Compared
thestructure.
structure. Compared
Compared
with thewith
traditional
with the silicon oxide
the traditional mask,
silicon
silicon oxidethemask,
oxide removal
mask, the
the method
removal
removal is more
method
method economical
is more
is more and con-and and
economical
economical con-
venient.convenient.
Using a photoresist
Using aas the mask,
photoresist it
asis difficult
the mask, toitachieve
is high
difficult toaspect
achieve
venient. Using a photoresist as the mask, it is difficult to achieve high aspect ratio pattern ratio pattern
high aspect ratio
etching pattern
on the bonding
etching etching wafer
on
on the bonding with
the bonding poor
wafer heat
wafer
with dissipation,
with
poor anddissipation,
poordissipation,
heat heat duringandetching, it is also
and during
during etching, etching, it
it is also
difficultisdifficult
toalso
ensure the
difficult etching
to ensure accuracy
the of
etching the structure
accuracy of due
the to the
structure shrinkage
due
to ensure the etching accuracy of the structure due to the shrinkage of the edge to of
the the edge
shrinkage of the
of the photoresist.
edge
of theofphotoresist.
the photoresist.
The fabricated
The structures
Thefabricated
fabricated are shown
structures
structures are
are inshown
shownFigure
in in4.Figure
The4.designed
Figure The
4. The plate plate
designed
designed distances are 4are 4are
distances
plate distances µm4
μm andand 8 μm,8 the
µm, etched
the sizes
etched are
sizes 3.959
are 3.959μm µm and
and 8.125
8.125 μm µm onon the
the top
μm and 8 μm, the etched sizes are 3.959 μm and 8.125 μm on the top side of the silicon,top side
side of
of the
the silicon,
silicon, and the
and the sizes
sizes are
are 4.306
4.306 μm
µm and
and 8.472
8.472µmμm on the
on backside.
the backside. These
These values
and the sizes are 4.306 μm and 8.472 μm on the backside. These values show excellent values show excellent
show excellentagreement
agreementbetween
between
agreement design andand
design
between actual
design values,
actual
and andand
values,
actual the the
footprint
values, footprint
and the is effectively
is effectively
footprint issuppressed.
suppressed.
effectively suppressed.

Figure 4. The etched

Figure structure.
Figure 4.4.The
The etchedstructure.
etched structure.

The TGV Thecover

The TGV
TGV plate needed
cover
cover plate to be prepared.
plateneeded
needed tobe
to First, laser
beprepared.
prepared. drilling
First,
First, was performed
laserdrilling
laser drilling on
wasperformed
was performed on
on
300 μm-thick glass
300µm-thick
300 [11],
μm-thickglassthe via
glass[11], was
[11],the designed
thevia
viawas to the
wasdesigned shape
designed to to the of a
theshape square,
shape of the length
of aasquare,
square,theof the
thelength
length of
of the
the
inlet sideinlet
wasside
inlet 120was
side μm, and
was120 the and
120µm,
μm, length
andthe oflength
the the outlet
length ofthe
of side
the wasside
outlet
outlet about
side 90about
was
was μm. The
about metal
90µm.
90 μm. filling
The
The metalfilling
metal filling
in the hole
in was
inthe realized
thehole
holewas by metal
wasrealized
realizedby casting
bymetal mold
metalcasting [12],
castingmold and the
mold [12], excess
[12],and
andthe metal
theexcess was removed
excessmetal
metalwas
wasremoved
removed
through the CMP process. After CMP, CrAu was used as a mask for glass etching, and the
cover plate was prepared. The process is shown in Figure 5.
Micromachines 2022, 13, 2022,
Micromachines
Micromachines x 2022,
FOR13,PEER
x FOR
13, REVIEW
x FORPEER
PEERREVIEW
REVIEW 5 of 9 5 5ofof9 9

Micromachines 2022, 13, 1071 5 of 9

through the CMP
through
throughtheprocess.
CMP
the After CMP,
CMPprocess. After
process. CrAu
CMP,
After was
CrAu
CMP, CrAuused asused
was
was aused
mask
asasafor glass
amask
mask etching,
for glass
for and theand
glassetching,
etching, andthe
the
cover plate
coverwas prepared.
coverplate
platewas The process
wasprepared.
prepared.The is shown
Theprocess in Figure
processisisshown 5.
shownininFigure
Figure5.5.

(a) (a)
(a) (b) (b)
(b) (c) (c)(c)
Figure 5. Preparation
Figure 5.5.
Figure
Figure of the TGV
5.Preparation
Preparationofof
Preparation ofcover:
the TGV
the
theTGV(a)
TGV laser
cover:
cover: drilling;
(a)
cover:(a)laser
laser
(a) (b) metal
drilling;
drilling;
laser filling
(b) metal
(b)
drilling; and
metal
(b) CMP;
filling
filling
metal and (c)CMP;
and
filling active
CMP;
and CMP;area
(c)(c)
active
activearea
area
(c) active
corrosion.
corrosion.
corrosion.
area corrosion.

AnodicAnodic
bonding
Anodic
Anodicbondingwas used
bonding
bonding was
was
wastoused
realize
used
usedtoto the packaging
realize
torealize
realize the
the of the tilt
packaging
thepackaging
packaging ofof accelerometer.
the
ofthe tilt
thetilt When When
accelerometer.
tiltaccelerometer.
accelerometer. When
When
bonding, a three-layer
bonding,
bonding, bonding
bonding,aaathree-layer
three-layer method
bonding
three-layerbonding was
method
bondingmethod used.
methodwas wasA positive
used.
used.AA
wasused. voltage
positive
Apositive of 280
voltage
positivevoltage V
voltageofof was280applied
280VVVwas
of280 was applied
wasapplied
applied
to the silicon
toto the
tothe structure,
thesilicon and the
siliconstructure,
structure, and
structure, glass
and the
and the substrate
glass
theglass
glass and and
substrate
substrate
substratetheand
TGV
the
and TGV
the cover
the cover
TGVTGV were
were
cover
cover grounded,
grounded,
were
weregrounded,asas shown
grounded, asas
shown shown
in
in Figure
Figure
shown 6.6.An
ininFigureAnexcessively
Figure excessively
6.6.An high
highbonding
Anexcessively
excessively bonding
high voltage
voltage
highbonding
bonding was
was
voltagelikely
voltage likely
was
was tolikely
cause
to cause
likely the
toto structure
the
cause structure
cause the to be
thestructure
structure
to be adsorbed
toadsorbed upward
upward
tobebeadsorbed
adsorbed orordownward
upward downward
upward on
orordownwardonthe
downward glass
theon
glass
onthe cover.
the glassTo
cover.
glass ensure
Tocover.
ensure
cover. bond
bond
ToToensure
ensure quality
quality
bondat
bond atquality
lowvoltages,
low
quality atatlow
low
voltages, good
good
voltages, plasma
plasma
voltages,good cleaning
cleaning
goodplasma ofof the
the
plasmacleaning structure
structure
cleaningofofthe and
and cover
cover
thestructure
structureand was
was required.
required.
andcover
coverwas Finally,
Finally,
wasrequired. thethe inter-
interconnection
required.Finally,
Finally,the theinter-
inter-
connection between
between
connection
connection different
different
between
between electrode
electrode
different
different groups
groups
electrode
electrode was
wasgroups achieved
achieved
groups waswas through
through
achieved
achieved metal metal
traces.
through
through traces.
metal
metaltraces.
traces.

Figure 6. Anodic
Figure bonding
6.6.6.Anodic
Figure was used
Anodicbonding
bonding forused
was
was wafer-level
usedfor package.package.
forwafer-level
wafer-level package.
Figure Anodic bonding was used for wafer-level package.

The preparation
The
The methodmethod
preparation
Thepreparation
preparation used silicon
method used
used
used and
siliconglass
silicon
siliconand
and as
and theasmain
glass
glass
glassasthe
asthe materials,
main
themain
main and and
materials,
materials, theand
materials, cost
the cost
andthe was
thecost
cost
was low.
wasIn
low. addition,
In
low. In the
addition, anodic
the
addition, the bonding
anodic bonding
anodic method
bondingmethod was
method used
was to
used
was realize
to
used the
realize
to wafer-level
the
realize
was low. In addition, the anodic bonding method was used to realize the wafer-level pack- wafer-level
the pack-
packaging,
wafer-level pack-
aging, which
aging,also
which
aging, also
which has
whichhasgood
good
also
alsohas air
hasairtightness.
good tightness.
good air The
Theoxygen
airtightness.
tightness.oxygen
The released
released
Theoxygen
oxygen ininthe
released theanodic
released anodic
ininthe bonding
bonding
theanodic in-
increased
anodicbonding
bonding in-
in-
creasedcreased
the
the mechanical
mechanical
creased the damping.
damping.
themechanical
mechanical The
damping.
damping. anti-vibration
The anti-vibration
The capability
Theanti-vibration
anti-vibration of
capabilitycapability
of the
capability structure
the structure thewas
ofofthe was im-
improved.
structure
structure was The
wasim-
im-
proved.proved.
The processed
processed
proved. The
The device
device isdevice
is shown
processed
processed shown
device in is inshown
Figure Figure
isshown 7.Figure7.7.
7. ininFigure

Figure 7. The processed

Figure 7.7.7.The
Figure accelerometers.
Theprocessed
processedaccelerometers.
accelerometers.
Figure The processed accelerometers.

4.4.
4. Experiment Experiment Results
ResultsResults
4.Experiment
Experiment Results
The dynamic The
The dynamic
Thedynamic
dynamic signal
signal analyzer
signal analyzer
HP35665A
signalanalyzer
analyzer HP35665A
was used
HP35665A
HP35665A was
was
was used
toused
test to
used to
the test the
frequency
totest
testthe frequency
response
thefrequency
frequency response
response
response
characteristics
characteristics
characteristics
characteristics of the
ofofthe accelerometer.
of the accelerometer.
theaccelerometer.
accelerometer. It
It was It was
found
Itwas found
wasthat
found
foundthat the
thethat position
position
that the of the frequency
of the frequency
theposition
position ofofthe re-response
thefrequency
frequency re-
re-
curve
sponse sponse
curve fluctuated
fluctuated
sponsecurve with
with
curvefluctuated the
the
fluctuatedwith bias
bias
withthe voltage,
voltage,
thebias but
but
biasvoltage, no
no
voltage,but obvious
obvious harmonic
harmonic
butnonoobvious
obviousharmonic response
response characteristics
charac-
harmonicresponse
responsecharac-
charac-
were
teristicsteristics
were measured.
measured.
teristics were So,
So,the
weremeasured.
measured.thereleased
released
So,
So,the gas
gasininthe
thereleased
released the
gaspackaging
packaging
gas ininthe process
processprovided
thepackaging
packaging provided
process sufficient
sufficient
processprovided
provided damping.
sufficient
sufficient
damping. damping. A CV test was carried out on the device using Keysight’s B1500, and the test results
damping.
are
A CV test shownwas incarried
Figurecarried
8. The structure ofusing
the device on the test surface thehad a results
good response
AACV CVtest wasout
testwas on out
carried the
outdevice
ononthe
thedevice
device Keysight’s
using B1500,
usingKeysight’s
Keysight’s and
B1500,
B1500, test
and
andthe thetest
testresults
results
are shown to the electrostatic
inshown
Figure 8.Figure force,
The structure the etching and
of the device release
ondevice process
the test of
surface the
had device
a good was good, and the
are
areshown ininFigure 8.8.The
Thestructure
structure ofofthethedevice ononthethetest
testsurface
surface hadaresponse
had agood
goodresponse
response
device was force,
to the electrostatic intact the
andetching
free of adhesion.
and release The capacitances
process of between
the device thegood,
was mass andand the positive
theand
totothe
theelectrostatic
electrostatic force,
force,the
theetching
etching andandrelease
releaseprocess
process ofofthe thedevice
device was
was good,
good, andthe
the
and negative electrodes were about 5.52 pF and 5.32 pF, respectively. Considering the
parasitic effect, the design value was consistent with the measured value. The differential
capacitance deviation caused by the process was only 3.7%.
Micromachines 2022, 13, x FOR PEER REVIEW 6 of 9

device waswas
device intact andand
intact free free
of adhesion. The The
of adhesion. capacitances between
capacitances the mass
between andand
the mass the posi-
the posi-
tive tive
andand
negative electrodes
negative werewere
electrodes about 5.525.52
about pF and 5.325.32
pF and pF, respectively. Considering
pF, respectively. the the
Considering
Micromachines 2022, 13, 1071 6 of 9
parasitic effect,
parasitic the design
effect, value
the design waswas
value consistent withwith
consistent the measured
the measuredvalue. The The
value. differential
differential
capacitance deviation
capacitance caused
deviation by the
caused by process waswas
the process onlyonly
3.7%.3.7%.

(a) (a) (b) (b)

Figure 8. CV
Figure
Figure 8.characteristics
8. CV of the
CV characteristics
characteristics ofaccelerometer:
of the (a) positive
the accelerometer:
accelerometer: (a) direction;
(a) positive
positive (b) negative
direction;
direction; (b) direction.
(b) negative
negative direction.
direction.

The The
capacitance detection
capacitance
capacitance waswas
detection
detection realized
was through
realized
realized a dedicated
through
through a adedicated
dedicatedASIC with
ASIC
ASIC digital
with
with out-
digital
digital out-
output.
put.put.
Thesensitivity
The sensitivity
The ofofthe
sensitivity the accelerometer
ofaccelerometer
the was
was
accelerometer tested
tested
was on on aoncentrifuge.
a centrifuge.
tested TheThe
a centrifuge. test
testThe
resultsresults
testshowed
results
showedthat that
showed the prepared
that the prepared
the prepared accelerometer had had
accelerometer
accelerometer had good good linearity
good
linearity inrange
linearity
in the the
in range
the of plus
of range
plus or minus
ofminus
or plus or2 minus
g. The
2 g. 2sensitivity
The sensitivity
g. The of
sensitivitythe
of accelerometer
the was
accelerometer 35,459.4
was bits/g,
35,459.4 and
bits/g, the
and corresponding
the corresponding
of the accelerometer was 35,459.4 bits/g, and the corresponding sensitivity sensi-
sensi-
was
tivity waswas
about
tivity about
28.2 28.228.2
µg/bit,
about μg/bit,
which which haddetection
hadwhich
μg/bit, high highhigh
had detection accuracy,
accuracy,
detection as shown
accuracy,as shown
in in Figure
as Figure
shown 9. 9.
9.in Figure

Figure 9. The
Figure
Figure 9. test
Theresult
9. The test of the
test result
result ofscale
of the factor.
the scale
scale factor.
factor.

The The
accelerometer
The output
accelerometer
accelerometer waswas
output
output continuously
was continuously
continuously sampledsampled
sampled at 120 Hz
120 for
at 120
at Hz more
Hz for more
for thanthan
more twotwo
than two
hours when
hours the input
when waswas
the input zero,zero,
and and
the result is shown
the result is shown in Figure 10. The
in Figure corresponding
10. The corresponding
RMSRMS
noise
RMS waswas
noise about
was 163.8
about μg, μg,
163.8 the the
µg, corresponding tilt noise waswas
0.0094°,
0.0094 ◦ , and
and the the
noise
the noise
noise
noise about 163.8 corresponding
√ tilt noise 0.0094°, and
density nearnear
density 10 Hz10 was about
Hz was 1.2 μg/√Hz,
about µg/ the
1.2 μg/√Hz, test
Hz, results
thetest
the are shown
testresults
results in Figure
areshown
are shown 10. 10.
ininFigure
Figure 10.

Figure 10. The

Figure
Figure zero-acceleration
10. The output.
The zero-acceleration
zero-acceleration output.

The Allan deviation of the accelerometer was calculated, and its zero-bias instability
was 9.4 µg, which showed that the device has excellent performance as an inclination
detection accelerometer, as shown Figure 11.
Micromachines 2022, 13, x FOR PEER REVIEW 7 of 9

The Allan deviation of the accelerometer was calculated, and its zero-bias instability
Micromachines 2022, 13, 1071 was 9.4 μg, which showed that the device has excellent performance as an inclination7 of
de-9

tection accelerometer, as shown Figure 11.

Figure 11.
Figure 11. The
The Allan
Allan deviation.
deviation.

The device was

The device wasbonded
bondedtoto thethe ceramic
ceramic tubetube
shellshell
withwith DOWSIL™
DOWSIL™ ME-1030
ME-1030 die-
die-bond-
bonding adhesive.
ing adhesive. The impact
The impact test test
waswas carried
carried out out 6 consecutive
6 consecutive times
times at 10,000
at 10,000 g @g 0.5
@ 0.5
msmsin
in the detection direction. The structures of the tested 10 devices were not damaged
the detection direction. The structures of the tested 10 devices were not damaged after the after
the impact,
impact, andand
the the device
device function
function waswasgoodgood
afterafter the power-on
the power-on test,test, indicating
indicating that that the
the pre-
prepared inclination
pared inclination accelerometer
accelerometer had had
an an anti-overload
anti-overload capability
capability of 10,000
of 10,000 g. g.

5.
5. Discussion
Discussion
The
The accelerometer designed in
accelerometer designed in this
thispaper
paperhas
haslow
lowmechanical
mechanicalnoise noisedueduetoto
thethe use
use of
of 100 µm-thick silicon for the structural layer. The method of using ITO
100 μm-thick silicon for the structural layer. The method of using ITO as the sacrificial as the sacrificial
layer
layer reduces
reduces the
the foot
foot effect
effect and
and ensures
ensures the the consistency
consistency ofof the
the device
device structure
structure and and the
the
design value. Glasses with metal-filled through holes are used as the package cover, and
design value. Glasses with metal-filled through holes are used as the package cover, and
the oxygen released by anodic bonding provides sufficient gas damping. Therefore, the
the oxygen released by anodic bonding provides sufficient gas damping. Therefore, the
accelerometer prepared by this process has extremely low mechanical noise and is suitable
accelerometer prepared by this process has extremely low mechanical noise and is suitable
for high-precision inclination detection.
for high-precision inclination detection.
Compared with the devices that have been successfully commercialized on the market,
Compared with the devices that have been successfully commercialized on the mar-
the performance of the accelerometer prepared here also shows certain advantages, as
ket, the performance of the accelerometer prepared here also shows certain advantages,
shown in Table 1. Colibris’s and Murata’s pendulum accelerometers have larger masses
as shown in Table 1. Colibris’s and Murata’s pendulum accelerometers have larger masses
and thus have lower mechanical noise. The advantage of Colibris‘s device is that the device
and thus have lower mechanical noise. The advantage of Colibris‘s device is that the de-
is made of all-silicon material, including the upper and lower cover plates and the sensitive
vice is made of all-silicon material, including the upper and lower cover plates and the
structure in the middle. When the ambient temperature changes, the deformation caused by
sensitive structure in the middle. When the ambient temperature changes, the defor-
thermal expansion is consistent, so it has excellent temperature stability and repeatability,
mationiscaused
which by thermal
incomparable for expansion is consistent,
other accelerometers, sothe
but it has excellent
cost temperature
is too high stability
to be suitable for
and repeatability, which is incomparable for other accelerometers, but
mass adoption in the industry. Murata’s upper and lower cover plates are made of siliconthe cost is too high
to be composite
glass suitable forstructure,
mass adoption
and theingas-filled
the industry.
methodMurata’s
is usedupper and lower
to provide cover plates
over-damping for
sensitive structures during bonding, which has excellent reliability and good to
are made of silicon glass composite structure, and the gas-filled method is used provide
resistance
over-damping
to vibration and forshock.
sensitive structures
However, the during bonding,
electrodes whichonhas
are grown theexcellent
sidewalls reliability
and cannotand
good resistance to vibration and shock. However, the electrodes are grown
be screened at the wafer level, and the price is much higher than that of consumer-grade on the side-
walls and
MEMS cannot be screened at the wafer level, and the price is much higher than that of
accelerometers.
consumer-grade MEMS accelerometers.
Table 1. Comparison of parameters about MEMS accelerometers for inclination detection.
Table 1. Comparison of parameters about MEMS accelerometers for inclination detection.
Package Noise
√ (µg Bandwidth
Manufacturer Model Structure
Manufac- Package
Form Noise / (μg
Hz) Bandwidth
(Hz)
Model Structure
turer Form
Silicon– /√Hz) (Hz)
Colibris [13] RS9000 Pendulous 30 30–80
Silicon–Sili-
Silicon
Colibris [13]
Murata [14] RS9000
SCA820 Pendulous
Pendulous Anodic 30220 30–80
18
con
ST [15] IIS2ICLX comb Al-Si 15 12.5
Murata [14] SCA820 Pendulous Anodic 220 18
Dual-layer
BOSCH [16]
ST [15] BMA456
IIS2ICLX comb Al-Ge
Al-Si 15120 1.5–1.6
12.5 k
comb
SDI [17] Model 1521 Teeter–totter 7 0–250
Gatech [18] Hinge-shaped 72
This Work comb Anodic 1.2 10
Micromachines 2022, 13, 1071 8 of 9

Bosch has designed multi-layer materials as structural layers to achieve the purpose of
increasing the size of sensitive structural mass blocks and reducing the impact of packaging
stress on device performance, and is widely used in automobiles. The structure layer
should be prepared by the epitaxial polysilicon method, and the thickness of the structure
layer is about 30 µm. Compared with the sensitive structure prepared with 100 µm-thick
single crystal silicon proposed in this paper, the noise control is slightly insufficient.
In Silicon Designs Inc., the low-noise accelerometer is realized by the double-seesaw
process, which has lower noise, but the device is not packaged at the wafer level, which is
very unfavorable for wafer-level testing and cost control.
The hinge-shaped accelerometer designed by Galtech has excellent sensitivity, and if
the mass size is appropriately increased, then it can also achieve lower mechanical noise;
however, this does not solve the packaging problem.
Although the sensitive structure of the high-performance MEMS accelerometer pro-
posed in this paper adopts the traditional comb structure, the mechanical noise of the device
is greatly reduced by increasing the thickness of the sensitive structure. Wafer-level packag-
ing in glass with metal-filled through holes also has low-cost mass production capabilities
and is a competitive implementation of high-performance MEMS accelerometers.

Author Contributions: Conceived the experiment and wrote the original draft, Y.F.; performed the
experiments, G.H.; contributed the metal casting mold process, J.G.; contributed reagents, materials,
and key process parameters, Y.Z. and J.N.; performed test and data analysis, Z.W.; technical super-
vision, F.Y.; writing—review and editing, C.S. All authors have read and agreed to the published
version of the manuscript.
Funding: This work is supported by The Laboratory Open Fund of Beijing Smart-chip Microelec-
tronics Technology Co., Ltd., and the Chinese National Science Foundation (Contract No. 52075519
and 61974136).
Institutional Review Board Statement: The study was conducted in accordance with the Declaration
of Helsinki, and approved by the Ethics Committee of the Institute of Semiconductors, Chinese
Academy of Sciences (1 July 2022).
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Yu, Y.; Ou, J.; Zhang, J.; Zhang, C.; Li, L. Development of Wireless MEMS Inclination Sensor System for Swing Monitoring of
Large-Scale Hook Structures. IEEE Trans. Ind. Electron. 2009, 56, 1072–1078. [CrossRef]
2. Łuczak, S. Erratum to: Guidelines for tilt measurements realized by MEMS accelerometers. Int. J. Precis. Eng. Manuf. 2014, 15,
2011–2019. [CrossRef]
3. Mejias, L.; Correa, J.F.; Mondragon, I.; Campoy, P. COLIBRI: A vision-Guided UAV for Surveillance and Visual Inspection. In
Proceedings of the 2007 IEEE International Conference on Robotics and Automation, Rome, Italy, 10–14 April 2007; pp. 2760–2761.
4. Gonseth, S.; Brisson, R.; Balmain, D.; Di-Gisi, M. Tactical grade MEMS accelerometer. In Proceedings of the 2017 IEEE DGON
Inertial Sensors and Systems (ISS), Karlsruhe, Germany, 19–20 September 2017; pp. 1–11. [CrossRef]
5. Vagner, M.; Benes, P. Start-up response improvement for a MEMS inclinometer. In Proceedings of the 2015 IEEE International
Instrumentation and Measurement Technology Conference (I2MTC), Pisa, Italy, 11–14 May 2015.
6. Kuisma, H. Glass isolated TSVs for MEMS. In Proceedings of the 5th Electronics System-Integration Technology Conference
(ESTC), Helsinki, Finland, 16–18 September 2014; pp. 1–5.
7. Thong, Y.K.; Woolfson, M.S.; Crowe, J.; Hayes-Gill, B.R.; E Challis, R. Dependence of inertial measurements of distance on
accelerometer noise. Meas. Sci. Technol. 2002, 13, 1163. [CrossRef]
8. Yoon, S.W.; Lee, S.; Perkins, N.C.; Najafi, K. Shock-Protection Improvement Using Integrated Novel Shock-Protection Technologies.
J. Microelectromech. Syst. 2011, 20, 1016–1031. [CrossRef]
9. Si, C.; Han, G.; Ning, J. Shock Resistance Design of a High-Performance MEMS Tuning-Fork Gyroscope. Micronanoelectron.
Technol. 2014, 51, 302–307.
10. Cai, M.; Si, C.; Han, G.; Ning, J.; Yang, F. Deep Silicon Etching of Bonded Wafer Based on ITO Mask. Micronanoelectron. Technol.
2020, 57, 6.
11. Yang, F.; Han, G.; Yang, J.; Zhang, M.; Ning, J.; Yang, F.; Si, C. Research on Wafer-Level MEMS Packaging with Through-Glass
vias. Micromachines 2018, 10, 15. [CrossRef] [PubMed]
Micromachines 2022, 13, 1071 9 of 9

12. Gu, J.; Xia, X.; Zhang, W.; Li, X. A Modified MEMS-Casting Based TSV Filling Method with Universal Nozzle Piece That Uses
Surface Trenches as Nozzles. In Proceedings of the 2018 19th International Conference on Electronic Packaging Technology
(ICEPT), Shanghai, China, 8–11 August 2018; pp. 536–539. [CrossRef]
13. Stauffer, J.M.; Dietrich, O.; Dutoit, B. RS9000, a novel MEMS accelerometer family for Mil/Aerospace and safety critical
applications. In Proceedings of the IEEE/ION Position, Location and Navigation Symposium, Indian Wells, CA, USA, 4–6 May
2010; pp. 1–5. [CrossRef]
14. SCA820-D04. Available online: https://www.murata.com/zh-cn/products/productdetail?partno=SCA820-D04 (accessed on
7 June 2022).
15. High-Accuracy, High-Resolution, Low-Power, 2-Axis Digital Inclinometer with Embedded Machine Learning Core. Available
online: https://www.st.com/en/mems-and-sensors/iis2iclx.html (accessed on 9 June 2022).
16. High Performance. Thin Package. Low TCO. Acceleration Sensors BMA456. Available online: https://www.bosch-sensortec.
com/products/motion-sensors/accelerometers/bma456/ (accessed on 11 June 2022).
17. SDI’S TECHNOLOGY. Available online: https://www.silicondesigns.com/tech (accessed on 13 June 2022).
18. Jeong, Y.; Daruwalla, A.; Wen, H.; Ayazi, F. An out-of-plane hinge-shaped nano-gap accelerometer with high sensitivity and
wide bandwidth. In Proceedings of the 2017 19th International Conference on Solid-State Sensors, Actuators and Microsystems
(TRANSDUCERS), Kaohsiung, Taiwan, 18–22 June 2017; pp. 2131–2134. [CrossRef]