a flexible liquid crystal polymer mems …seagrant.mit.edu/publications/mitsg_12-18.pdf · a...

Sea Grant College Program Massachusetts Institute of Technology Cambridge, Massachusetts 02139 NOAA Grant No. NA10OAR4170086 Project No. 2011-R/RCM-30 Reprinted from Smart Materials and Structures, Vol. 21 (11), doi: 10.1088/0964-1726/21/11/115030, October 2012.

A FLEXIBLE LIQUID CRYSTAL POLYMER MEMS PRESSURE SENSOR ARRAY FOR FISH-LIKE

UNDERWATER SENSING

A. G. P. Kottapalli, M. Asadnia, J.M. Miao, G. Barbastathis and M. S. Triantafyllou

MITSG 12-18

lOP Pum.rslliNG

Smart Mater. Struct. 11 (2012) 115030 (13pp) doi: 10.108810964-1726/21111/115030

A flexible liquid crystal polymer MEMS pressure sensor array for fish-like underwater sensing

A G P Kottapalli1, M Asadnia1, J M Miao1, G Barbastathis2 and M S Triantafyllou2

1 School of Mechanical and Aerospace Engineering, Nanyang Technological University, 50 Nanyang Alvenue,639798,Singapore 2 Department of Mechanical Engineering, Massachusetts Institute of Technology, 77 Massachusetts Alvenue, Cambridge, MAl 02139, USA

E-mail: mjmmiao @ntu.edu.sg

Received 19 June 2012, in final form 29 August 2012 Published 23 October 2012 Online at stacks.iop.org/SMS/21/115030

Abstract In order to perform underwater surveillance, autonomous underwater vehicles (AUVs) require flexible, light-weight, reliable and robust sensing systems that are capable of flow sensing and detecting underwater objects. Underwater animals like fish perform a similar task using an efficient and ubiquitous sensory system called a lateral-line constituting of an array of pressure-gradient sensors. We demonstrate here the development of arrays of polymer microelectromechanical systems (MEMS) pressure sensors which are flexible and can be readily mounted on curved surfaces of AUV bodies. An array of ten sensors with a footprint of 60 (L) mm x 25 (W) mm x 0.4 (H) mm is fabricated using liquid crystal polymer (LCP) as the sensing membrane material. The flow sensing and object detection capabilities of the array are illustrated with proof-of-concept experiments conducted in a water tunneL The sensors demonstrate a pressure sensitivity of 14.3 1-L V Pa-1. A high resolution of 25 mm s-1 is achieved in water flow sensing. The sensors can passively sense underwater objects by transducing the pressure variations generated underwater by the movement of objects. The experimental results demonstrate the array's ability to detect the velocity of underwater objects towed past by with high accuracy, and an average error of only 2.5%.

(Some figures may appear in colour only in the online journal)

1. IntroducUon

Autonomous underwater vehicles (AUVs) have been used more frequently for underwater surveillance in recent decades. In order to navigate and explore underwater information, an AUV requires a broad suite of sensors that can withstand harsh, cluttered and murky seawater environments, which pose a great challenge to most sensing systems for imaging objects. Amongst the plethora of sensors on the body of an AUV, one of the most important and ineluctable sensing systems detects objects around the vehicle imparting vision to the AUV [1, 2]. In the past, sonar and optical methods have been used to meet this purpose, but they

0964-1726/121115030+ 13$33.00

suffer from serious disadvantages. Sonars are highly impeded by their poor resolution of detection and they reveal the location of the source due to active sensing [3]. In addition, there have been reports in the past about the devastating deaths of marine animals due to the intense sound generated by sonar sensing [4]. Optical methods, on the other hand, fail to work in dirty, clouded waters. Sensors based on Doppler frequency shift consist of an acoustic launcher and receiver and are usually quite bulky [5]. Additionally, all these methods perform active sensing and must emit energy in the form of acoustic waves or light in order to work, and are not energy efficient. Although acoustic sensing has shown great promise in this regard, its operation needs a 40% upward

© 2012 lOP Publishing Ltd Printed in the UK & the USA

Smart Mater. Struct. 21 (2012) 115030

increase of a small AUV's total power [6]. Therefore, there is a need to develop alternative low-powered, low-cost and miniaturized sensors that are robust enough to withstand harsh seawater environments and capable of generating the 'feel' of surrounding flows imparting vision to AUVs.

Fish that navigate in similar complex environments have an uncanny ability to adeptly swim at high speeds, while avoiding any collision with underwater objects even in regions that are completely devoid of vision [7- 9]. They are capable of performing this impressive feat by relying on arrays of sensors on their bodies called lateral-lines [7- 12]. The blind cave fish is one such unique fish species that relies completely on its lateral-line of sensors for its navigation. Due to its inhabitation of dark caves for long periods of time, this fish developed an atrophied eye over evolution [10]. The lateral-lines are arrays of sensors present on and beneath the skin of the fish running down their sides and around their heads. Individual sensors of the lateral-line called neuromasts (standing cylindrical structures of soft polymer-like material) extend into the surrounding flow [7]. Surrounding flow variations cause the neuromasts to bend which triggers an electric impulse that is sent to the fish's brain [7- 9]. The lateral-line consists of two sensory sub-modalities: a system of velocity sensitive superficial neuromasts that responds to flow variations and a canal neuromast system located under the skin that responds to pressure variations [7]. The superficial neuromasts are present on the surface of the skin, while the canal neuromasts are submerged in fluid-filled canals and communicate with the surrounding water through a series of pores [11, 12]. A pressure difference between successive pores actuates an individual canal neuromast located between the pores. It has been hypothesized in the past that the lateral-line potentially contributes in generating a hydrodynamic image of surroundings of the fish [12]. The fish performs a 'silent' or passive sensing in the water without generating any electric field, optical or ultrasonic signal. Pressure sensor arrays inspired by an efficient and ubiquitous sensory system like the lateral-line of the fish could fill the gap left by sonars and optical systems and provide an energy efficient alternative sensing system to enable pressure sensing around the bodies of AUVs. Passive surveillance does not reveal the source or interfere with other vehicles or life forms. Another aspect of inspiration from the fish that could greatly benefit AUVs is its ability of rapid maneuvering and energy efficient propulsion by active flow manipulation. Arrays of pressure sensors on the body of the vehicle could envision pressure generated by flows around the vehicle so that it can utilize rather than fight those flows, saving energy and improving maneuverability [13]. This work through proof-of-concept experiments demonstrates the LCP sensor array's abilities in performing underwater sensing by measuring pressure variations similarly to the pressure-gradient sensing performed by the canal neuromasts.

Microelectromechanical systems (MEMS) devices fabricated using micromachining technology offer high resolution and high sensitivity, while the ability to shrink device size is of benefit in fabricating dense arrays of pressure sensors in a small area. MEMS make it possible to develop pressure

2

A G P Kottapalli et al

sensor arrays that consume less power, at low fabrication cost (due to batch production). In light of these issues, a considerable interest has been devoted in the past to the development of MEMS flow sensor arrays that mimic the biological lateral-line on the fish. Liu et al [14, 15] were among the first to develop MEMS artificial hair cell sensors bio-mimicking the individual neuromast sensors of the fish's lateral-line. Their device design consisted of artificial hair cells represented by high-aspect ratio SU-8 pillars standing on 2 J,tm thick silicon cantilever beams doped with piezoresistive sensing elements at the anchor. The silicon-based MEMS flow sensors that they developed demonstrated ultrahigh sensitivity to flow velocity. Venturelli et al employed artificial lateral-line pressure sensor arrays for fluid dynamic studies in demonstrating the array's ability to discriminate between Karman vortex sheets from uniform flows [16, 17]. Klein and Bleckmann demonstrated the use of optical flow sensors to sense bulk flow velocity and also to track the size of a cylindrical object underwater [18]. Qualtieri et al developed biomimetic parylene coated stress-driven out-of-plane bent multilayered cantilever beams for air and water flow sensing [19]. Salumae et al employed lateral-line like pressure sensing systems on board an underwater robot [20]. Silicon piezoelectric diaphragm based pressure sensors have also been developed for underwater sensing in the past [21, 22]. In spite of the possibility of fabricating, densely packed arrays of sensors on a silicon substrate, silicon does not offer the flexibility that is needed for the array to be mounted on underwater vehicles because silicon is rather too brittle a material with low fracture strength. Moreover, silicon does not exhibit the chemical robustness and hermeticity needed for an underwater setting. This work, based on [23], addresses these issues by constructing a pressure sensor array almost entirely using liquid crystal polymer (LCP) material, allowing the array to be flexible, and robust against chemical attack. and moisture absorption in a harsh seawater setting. LCP is used as a membrane material and also as a backing substrate to package the array of sensors on a flexible platform that can be readily mounted on streamlined vehicle surfaces.

Previous works [3, 14, 15, 19] on fabricating MEMS artificial lateral-lines have taken a biomimetic approach, closely mimicking the structural features of the individual biological sensor on the fish by recreating sophisticated hair cells using micromachining technology. The current work, instead, in view of application, takes a bio-inspired approach, focusing on developing a robust and flexible array of sensors that can be readily deployed on underwater vehicles. The work focuses on the functional implications of the lateral-line and bears significant engineering benefits such as ease of fabrication and the ability to develop low-powered, low-cost, flexible arrays with a robust sensor design. In previous work, LCP has been mainly used for electronic packaging applications [24, 25]. Despite the unique and beneficial features of LCP with respect to the underwater atmosphere and its fabrication compatibility with conventional microfabrication technologies it has not been deployed as a material to fabricate MEMS devices for underwater applications in past works.


Almost all artificial MEMS biomimetic sensors developed previously have used a vibrating dipole to quantify the sensor array as an artificial lateral-line [3, 14, 15]. Although the hydrodynamic stimuli generated by the oscillating tail of a small prey can be approximated to a vibrating dipole, the dipole fails to model the stimuli generated by larger fish, which could be better approximated by blunt moving objects [2]. In fact the dipole model is rather inapplicable to both passive and active object detection, as discussed in [13]. This paper investigates the ability of a flexible LCP based pressure sensor array to accurately detect the velocity and distance of a non-oscillatory stimulus. An array of ten sensors is packaged on a flexible LCP backing, with individual sensors fabricated using an LCP membrane with metal piezoresistors deposited on it. The performance of the sensors in sensing steady-state laminar water flow will be discussed in this paper. The individual sensors developed demonstrate a resolution of 25 mm s-1 in sensing steady -state laminar water flow velocity. The array's capability in passive underwater object detection is explained through proof-of-concept experiments conducted by towing objects at various velocities and distances past the array. The purpose of these experiments is to investigate the LCP MEMS sensor array's capability in detecting the velocity and distance of a non-oscillatory stimulus. The array could predict the velocity of an object towed past the sensors with a high accuracy of 2.5% average error. It could sense objects that passed by at a distance as large as about 300 mm away from the array. The results demonstrate the promise towards a lateral-line like array, capable of withstanding harsh environments for AUV applications.

2. The LCP MEMS flexible sensor array design

The current design of the array consists of ten sensors, arranged into two rows, with five sensors in each row, with the entire array having a small footprint of 60 mm x 25 mm. Individual sensors have a thickness of 330 ~-tm and occupy a footprint of 6 mm x 3.5 mm. The sensors are arranged with a center-to-center spacing of 10 mm, providing a gap of 4 mm between edges of adjacent sensors to prevent the possibility of crosstalk. Figure l (a) shows all the physical dimensions of the array. All the sensors are firmly mounted using a non-conductive epoxy on a 100 11-m thick flexible LCP backing. The LCP backing consists of sputter coated and patterned gold interconnects that link to the contact pads of the sensors. The sensors are fabricated and packaged into very thin arrays of thickness approximately 450 IJ.m, which in addition to increasing flexibility, also ensures that the physical height of the array does not influence the flow sensed by the array when mounted and used on a streamlined underwater body.

The strain concentrating LCP diaphragm transduces the pressure difference between the environment and the sensor's internal cavity (sealed between the LCP sensing membrane and the flexible LCP backing layer) into a resistance change in metal piezoresistors. Individual sensors are connected to external Wheatstone bridge circuits mounted on a PCB board

3


to convert the resistance change into voltage change. The sensor is operated as one arm of the Wheatstone bridge circuit. The other resistances of the circuit are chosen to be the same as the resistance of the metal piezoresistor in order to maintain a zero output voltage when no external pressure acts on the membrane.

2.1. LCP as a sensor membrane material

In comparison to silicon, LCP is an excellent material for MEMS devices for underwater applications. LCP has a lower moisture absorption coefficient (""0.02%) and permeability, and higher fracture strength than silicon. Chemically, LCP is more resistant to a wide range of chemical attacks (present in seawater) as compared to silicon and offers notable advantages as a viable material for harsh environment MEMS devices [23]. LCP material is compatible with conventional silicon fabrication processes. Our group has developed various fabrication processes for LCP, including wafer bonding (LCP to LCP and also LCP to silicon), deep reactive ion etching (DRIE), spin coating and lithography, deposition and lift-off.

The device fabricated in the current work consists of an LCP membrane with gold piezoresistors deposited on it. In spite of the low gauge factor of gold (KAu = 2), as compared to piezoresistive silicon (Ksi = 140), a combination of gold strain gauges on LCP offers higher sensitivity in comparison to piezoresistive silicon membranes (of the same membrane dimensions and thickness) due to the lower Young's modulus of LCP as compared to silicon. The reason for this is that the factor that affects the sensitivity is the ratio of KfE as described in equation (1) [23], where K is the gauge factor and E is the Young's modulus of the material. The sensitivity is defined as the change in resistance of the strain gauge for unit stress generated.

(1)

where u stands for the stress. The Young's moduli of silicon and LCP are Esi = 185 GPa and ELCP = 2.16 GPa respectively and R is the strain gauge's resistance. Comparing two strain gauge pressure sensors of the same design of strain gauge dimensions and membrane size and thickness, but one made of the silicon membrane and the other made of the LCP membrane, the LCP membrane sensor turns out to be 19% more sensitive, as can be calculated from equation (1).

Therefore, in comparison to LCP membranes, very thin silicon membranes (the most commonly used are 2, 5 or 10 ~-tm) are required in order to preserve sensitivity, whereas in the current sensor design an LCP diaphragm as thick as 25 11-m offers high sensitivity to flow. LCP membranes can withstand a higher pressure impact generated by pressure waves in water than silicon due to LCP's higher fracture strength and thicker membranes. The choice of LCP material in this work offers high robustness to the sensors as compared to silicon-based hair cell like standing structures or thin cantilever beams for harsh seawater applications [23, 24].

Smart Mater. Struct. ll (2012) 115030 A G P Kottapalli et al

4mm

(b)

Conductive silver paste

LCP membrane sensor

Au Contact pads

Au routing Flexible

LCP backing

Figure 1. A schematic (not drawn to scale) showing the array. (a) Design, dimensions and materials used in the array fabrication and packaging. (b) 3D view of a flexible array of 10 sensors.

2.2. Strain gauge and LCP membrane design

The main design parameters that influence the pressure sensitivity of the device are the membrane dimensions and the strain gauge design. Considering the diaphragm to be a thick film diaphragm. i.e. small center deflection (I)() as compared to diaphragm thickness t, the deflection w(r) at any point on the diaphragm under uniform pressure can be approximated by

Ptiowll' ( (r)2)2 w(r) = 64D 1 - ~ (2)

where Pfiow is the pressure generated by flow variations on the diaphragm, a is the radius of the diaphragm, r is the position along the radial direction (0 < r < a) and D is the flexural rigidity of the membrane, given by

Er3 D = 12(1- v2)

(3)

4

where E is the Young's modulus and v is the Poisson's ratio. Using equations (2) and (3), the membrane dimensions must be chosen to keep the deflection much smaller than the membrane thickness for pressure values up to the maximum pressure of concern. The maximum pressures generated in steady-state direct currents in water flow and underwater waveforms generated by moving objects will never exceed 1000 Pa [13]. This assumption is made from results presented in [13] by Fernandez et al. They demonstrate that the pressure generated by an object towed past an array of pressure sensors is well below 1000 Pa in very similar experimental conditions and ranges of distance and velocity of the object to those used in this work. LCP is available in thin films of thickness 25 JJ.m and therefore the membrane thickness parameter is fixed. Solving using equations (2) and (3), the radius of the membrane must not be more than 1.38 mm in order to ensure that it is operating in the thick film approximation for the


maximum pressures of interest. The radius of the membrane was chosen to be 1 mm in order to be able to design an optimal resistance of the metal strain gauge on it and also to preserve the response of the sensor. The metal piezoresistor design consists of gold resistors with a resistance value of 600 Sl placed in the periphery of the membrane in a serpentine pattern. A zig-zag pattern of resistors consisting of long and thin radial elements and short and thick tangential elements was designed to maximize the resistance change collected. The wires linking the neighboring radial elements were made short and thick because they are strained perpendicular to the electric current and reduce the overall signal [26].

2.3. Flow sensor design

An external flow past the sensor membrane sets a pressure difference between the atmosphere and the membrane cavity, therefore the membrane bends. The change in resistance of the metal piezoresistors that is read-out as voltage change could be calibrated to flow velocity. The piezoresistors collect the maximum strain due to membrane displacement when placed at the periphery of the membrane. The bending strains in the radial and tangential directions are given, respectively, by [27]

(4)

and

(5)

where Br and Bt are the strains along the radial and tangential directions, respectively.

At r = a, the maximum radial strain occurs, Br = Emax, and Et = 0 (where the radial strain gauge is placed). In reality, although the radial strain gauge is placed at the circumference, it extends 400 1-1-m towards the center from the circumference of the membrane forming a meander and the stress gets averaged over this length [28]. However, for simplicity of calculations, we consider the resistor to be placed at r = a where maximum stress occurs. Substituting for wo in equation (4), we get the maximum strain Ermax as

(6)

It can be shown that the relative change in resistance for a resistor segment deformed by being bonded to the top of a plate is [27]

AR 1 2v -1 - ~ --Er + --Et R l-v 1-v

(7)

where Br and Et are as defined in equations (4) and (5). Using equations (6) and (7) and substituting for the values E = 2.1 GPa (LCP), a= 1000 1-1-m, v = 0.3 and t = 25 1-1-m. Using equations (6) and (7), the relative change in resistance depends on the pressure as follows:

AR - = (7.22 x 10-7 Pa-1)P R

where P is the pressure difference across the diaphragm.

(8)

5


2.4. LCP flexible array design

The individual sensors developed in this work consist of LCP bonded to silicon and the sensors are mounted in an array on a 100 1-1-m thick LCP backing layer to incorporate flexibility of the array as illustrated in figure 1 (b). In cases where high flexibility of the array is desired, it is possible to eliminate silicon completely in the array design and achieve an all LCP MEMS sensor for underwater sensing. On one hand, this could improve the flexibility by a great extent, especially in the case of densely packed arrays where the presence of silicon could pose a constraint, while on the other hand, the thickness of the array could be greatly reduced to ~ 130 1-1-m. Through holes can be drilled by DRIE into an added layer of LCP that replaces silicon in the current design. LCP layers can be easily bonded to each other forming a stack of three layers with a 25 J.t.m thick diaphragm membrane, a 100 J.t.m thick layer with DRIE holes and a 100 J.t.m thick flexible backing layer. When the array is mounted on a curved surface, an extra parameter needs to be considered in the design. The strain induced on the membranes due to the curvature of the surface on which the array is mounted, Ecurvature. should not exceed the maximum strain experienced by the diaphragms, Bmax. In the small deflection regime Bcurvatun:: should be kept much smaller than Bmax so that the curvature does not significantly influence the deflection of the diaphragm by a signal [27].

where the maximum pressure across the diaphragm, Pmax, is 1000 Pa. Bcurvatun:: can be estimated as follows:

(10)

where p is the radius of curvature and z is the distance from the neutral axis to the plane of the array of membranes. The thickness of the entire array is 430 1-1-m. therefore z is approximately 215 J.t.m. For p = 1 m, we get Bcurvature = 0.42emax, therefore the sensor's diaphragm will still operate linearly in spite of the curvature.

3. LCP MEMS device fabrication and packaging

LCP 3908 is commercially supplied in the form of thin films of thickness 25 J.t.m with 18 J.t.m thick copper cladding on both sides. The copper cladding on both the sides is etched away to expose fresh surfaces of LCP. The device fabrication is illustrated through schematics in figure 2. The fabrication process commences with bonding a 25 J.t.m thick LCP 3908 film to a 300 1-1-m thick silicon wafer by using an intermediate adhesion layer of 2 J.t.m thick SU-8 2002 spin-coated on both the silicon wafer and the LCP film. In order to be able to spin-coat SU-8 on a flexible film like LCP, the LCP film is temporarily bonded to another silicon support wafer as shown in figure 2(a). A film of 5 1-1-m thick photoresist is coated on the temporary support wafer and the LCP is bonded onto it. The SU-8 coated faces of both the silicon and LCP layers are then exposed to oxygen plasma in a reactive ion etching (RIE)

Smart Mater. Struct. ll (2012) 115030

(a)r------------,.

S11icon 3001Jm

-----~-.-----ii 211fll SU-o--..... ,_ ______ ll. ___ lllllt 211m

1----___..:L:.:C~P------a 251Jm

Silicon 300~-~M L---------------------~

(c) 100nmAu

LCP

Silicon


(b) 111111 5kPa Uniform pres5Ure

Silicon

LCP SU-8

Silicon

(d) 100 nm Au

LCP

I

Figure l. A schematic of the fabrication processes: (a) LCP and silicon both spin-<:<>ated with 211-m thick SU-8 2002layers, (b) LCP-silicon bonding, (c) photoresist patterning and Cr (20 run)/ Au (100 nm) sputter-deposition, (d) the lift-off process to form gold piezoresistors.

chamber for 60 s to improve the bonding quality. Immediately after that, the SU-8 faces of both the LCP and the silicon wafers are brought into contact. The wafer pair is subjected to a pre-optimized thermal cycle to harden the SU-8 intermediate adhesive layer and enhance the bond strength between the silicon and the LCP. The thermal cycle involves increasing the temperature in steps, starting at 45 oc for 20 min, then 90 oc for 10 min, followed by 150 oc for 50 min. The temperature is then slowly ramped down to ambient temperature. A uniform pressure of 5 kPa is applied throughout the bonding cycle on the wafer pair as shown in figure 2(b). The intermediate SU-8 layer bonding process offered excellent bonding between the LCP and silicon surfaces. An optical micrograph of the cross-sectional view of the bonded pair is shown in figure 3(a) which demonstrates a good bonding between the LCP and the silicon with no delamination or trapped air bubbles. A thin conformal layer of SU-8 binds the two layers together. The silicon temporary support wafer is removed by dipping the wafer pair in acetone which rapidly etches away photoresist but does not attack SU-8.

We also developed a wafer bonding recipe for direct bonding between the LCP and the silicon. In direct bonding, the wafer pair is heated to a temperature slightly above the glass transition temperature (Tg) of LCP 3908 at 280 °C. The bonding process only ends after the substrates have cooled down sufficiently so that the reflowed LCP can solidify to form a strong adhesive bond. Figures 3(b) and (c) show the result of direct bonding between the LCP and the silicon. However, during direct bonding, stress generated at the interface due to reflow of LCP leads to wafer cracking during the dicing process. The stress at the interface is as a result of mismatch of the coefficient of thermal expansion (CTE) between silicon and LCP. During the dicing process, some broken silicon pieces can be observed to be still bonded to the LCP, demonstrating the strong bonding strength between

6

silicon and LCP. In addition, reflow at the interface between silicon and LCP could change some of the material properties of LCP which would influence the mechanical performance of the diaphragm. We found that the intermediate SU-8 layer that is used in this work for bonding LCP and silicon not only eliminates all the problems of direct bonding but also enhances the bonding strength. This process is followed by spin coating a 5 ,.,_m photoresist (PR) on the LCP and patterning it for the lift-off process. Cr (20 nm)/Au (100 nm) are sputter-deposited, and the lift-off process is performed to form gold piezoresistors. A thin intermediate layer of Cr enhances the adhesion of Au on the LCP. A DRIE through-hole, aligned with the resistors on the LCP on the top side, is etched on the silicon from the back side while releasing the LCP membrane as shown in figure 2(d).

Figure 4 shows optical microscope images of the front and back sides of the fabricated devices. These devices are then diced into individual sensors which are then packaged as flexible arrays on the LCP. The interconnects required for wiring are sputtered and patterned on the flexible 100 ,.,_m thick LCP backing layer. All interconnects and contacts are waterproofed using a thin layer ofPDMS. Since the fabricated sensors have only a single metal piezoresistor and form only a quarter-bridge of the Wheatstone circuit, external resistor circuits are needed for device operation. We also developed LCP sensors which incorporated all the four metal piezoresistors on the sensor diaphragm eliminating the need for any external resistor circuits for operation. Moreover, this design increases the sensitivity of the device fourfold. An optical microscopic image of the full-bridge LCP sensor incorporating two spiral and two radial strain gauges is shown in figure 5.

In order to package the devices on a flexible substrate, a Cr (20 nm)/ Au (700 nm) thick gold layer is sputtered and


(b) (c)

LCP

' Reflowof LCP

l0f.1m 10 f,lffi

Figure 3. Summary of the bonding experiments. Cross-sectional view of (a) the dicing plane of the silicon-LCP wafer pair bonded using a 4 (J.m thick SU-8 intermediate adhesion layer demonstrating a good conformal bonding, (b) LCP and silicon direct bonding before reflow heat treatment, (c) after reflow with a pedect bonding.

(a) (b)

Figure 4. Optical microscope images of fabricated LCP sensors: (a) the front side of Au piezoresistors in serpentine form, (b) the back side DRIE through holes in silicon that stopped at LCP membranes.

patterned to form electrical connections on 100 ~-tm thick LCP for the flexible backing. The devices are bonded at ambient conditions on the flexible LCP layer using a non-conductive epoxy and the contact pads are connected using silver paste

7

for electrical contacts on the LCP backing layer. The sensor cavity is formed between the LCP sensing membrane and the LCP backing layer. Figure 6 shows a photograph of the fabricated flexible array with ten sensors. Figure 7 shows

FigureS. Optical microscopic image of a fabricated full-bridge LCP sensor with two radial and two spiral gold piezoresistors.

Figure 6. Photograph of an array of ten sensors arranged into two rows on a 100 ~m thick flexible LCP backing patterned with gold interconnections.

a photograph of the sensor array mounted on a cylinder to demonstrate the high flexibility of the array.

4. Experimental results and discussion

This section elucidates proof-cf-concept experiments that demonstrate the abilities of the fabricated pressure sensor array in sensing water flow, and detecting the velocity and distance of underwater objects with good accuracy. The sensors are connected to external Wheatstone bridge circuits mounted on a PCB that is placed in close proximity to the sensor array in order to minimize noise possibly introduced due to the wire cables. The Wheatstone bridge circuit is biased with 5 V, and the signal from the sensor is acquired in LAB VIEW at a sampling rate of 1000 Hz using a National Instruments data acquisition (NI-DAQ) system USB-6289 M-series model. The same array of sensors is used in all the experiments documented in this paper, and the data reported are unfiltered data. In all the underwater experiments, the hydrostatic pressure, found as the pressure at the membrane when no water flow is generated or before

8

Figure 7. A photograph illustrating the flexibility of the packaged array.

any stimulus approaches the sensor, is removed by subtracting the corresponding value of the voltage from the data of each sensor measurement in each run. All voltage quantities in the figures and in the text are therefore the voltages corresponding to pressure differences with respect to the hydrostatic pressure.

The output voltage of the sensors is calibrated to pressure by using a vacuum pump test described in this section. In order to validate the ability of the array to detect underwater objects, two experiments are conducted by towing a cylinder past the array of sensors; one at fixed distance with velocity varied and the other with a fixed velocity at various distances. The results depict that the array has a good accuracy in detecting object velocity with a low average error of 2.5% calculated from repeated experiments. The results are consistent with repeated trails of experiments conducted so far.

4.1. Pressure calibration of the LCP sensor

A vacuum pump test is conducted to calibrate the sensor's output to the pressure exerted on the sensor membrane. The schematic of the experimental setup used to perform this test is illustrated in figure 8(a). A single sensor is mounted on an aluminum fixture fabricated with two holes drilled in perpendicular planes that can transfer pressure from the vacuum pump to the membrane as shown in figure 8(a). The sensor is mounted at the center of an aluminum fixture with its membrane aligned to the circular hole of 2 mm diameter drilled into the fixture. It is ensured that there is no loss of pressure by providing an air-tight sealing at all connections. The connection between the vacuum pump and the sensor has a pressure regulator and a pressure valve to regulate the pressure exerted on the membrane and to read the value of the pressure respectively. The sensor output is recorded as the pressure is varied from 0 to 1500 Pa (the range of


(a)

(b)

30

24 I

18 > E -; 12 C) Ill .. ][

0 6 >

0

0 300 600 900 1200 1500 Pressure (Pa)

Figure 8. Voltage-pressure calibration. (a) Schematic describing the experimental setup used for pressure calibration. (b) Pressure calibration plot. The error bars represent the standard deviation obtained over a number of repeats of the experiment.

interest for underwater applications). Figure 8(b) shows the voltage output of the sensor as a function of the pressure on the membrane. The error bars are plotted from the standard deviation values acquired over a number of experiments repeated at similar experimental conditions to demonstrate the repeatability of the calibration data. The sensor demonstrates good repeatability and a linear behavior in this range. The value of sensitivity as derived from the slope of the linear calibration plot shown in figure 8(b) is 14.3 ttY Pa-1.

4.2. Water flow sensing

Experiments for water flow sensing are performed in a Long win LW-3457 model closed circuit type water tunnel. The water tunnel has a large test section of dimensions 0.3 (W) x 0.4 (H) x 1 (L) m3 and is fitted with a turbulence reducing steel screen and two honeycomb layers. The large dimensions of the tank reduce the chances of the sensors finding the pressure signals generated due to reflections of waves off the walls.

9

The sensor array is mounted in the center of the test section of the water tunnel immersed 100 mm deep in water. Initially, the water flow velocity is kept at a maximum of 600 mm s-1

and then the flow velocity is reduced to 0 mm s-1 in steps of 25 mm s-1 while the data are continuously acquired using a DAQ. It is preferred to conduct the experiment with the flow rate changing downstream since the flow turbulence observed in the water tunnel is lower in this case as compared to varying the flow rate upstream. The sensors show clear steps of decreasing voltage for the corresponding decrease in the flow pressure. Figure 9 shows a profile of the voltage signal acquired versus the water flow rate.

The sensitivity of the sensors for water flow sensing as calculated from the curve fitting in figure 9 is 90.5 m V /IDB-1.

The sensors demonstrate a good resolution of 25 mm s-1

for steady-state laminar water flow sensing. The values of sensitivity presented here are average values found from a repeated set of experiments. The water flow in the tunnel is also at the same time monitored by a commercial GE Druck LPM 9381 model Pitot tube sensor as a reference.


40

35

-30 > .§.. 25 '5 .e- 20 :I 0 0 15 1/)

~ 10 (/)

5

0

o< :7

9!' /> _/.

(0

~·

y -~· ~

0 100 200 300 400 500 600

Water flow velocity (mms"1)

Figure 9. Experimental testing results in the water tunnel. The solid line is a second order polynomial fit passing through the experimental points. The flow rate was reduced from 600 to 0 mm s-1 in steps of 25 rom s-1•

4.3. Underwater object sensing experiments

Separate tests are conducted to evaluate the array's ability to determine the velocity and distance of an object when it is towed past the array. Although it is not necessary to generate water flow for the object detection tests conducted, it was still

(a)

Gantry for towing

Sensor output to ---.-:.__---"-=;;.._-11~ ./., DAQ and LABVIEW

100 mm deep ~------1~

Sensor array --+------1~

Stimulus lnhio>rt\ -+------f--1

Commercial reference oono,n.·-------~


chosen to conduct the tests in the water tunnel since the gantry of the water tunnel is used for the purpose of towing an object past the sensor array. The gantry is a translational stage that can move the object attached to it at a specific velocity along the length of the test section of the water tunnel. The position of the object in the test section and velocity of its pass can be adjusted by controlling the gantry through a LABVIEW code. All object detection tests presented in this work consist of a stationary array and a moving stimulus (resembling a passive fish-like sensing case), which is a cylinder of radius 5 mm. In all these experiments, the array is mounted fiat onto the wall of the test section of the water tunnel and submerged such that the membrane is 100 mm below the water surface. The object is attached firmly to the gantry of the water tunnel. The gantry is adjusted in such a way that the cylindrical object (stimulus) moves in a straight line parallel to the array when towed past in front of the array of sensors as shown in figure 10. The experimental setup for the object detection experiment conducted in the water tunnel is described in figure 10.

4.3.1. Underwater object velocity sensing. Tests for determining the velocity of the stimulus are conducted by passing the stimulus at a fixed distance 50 mm away from the array but at different velocities passing the array. Three runs of the experiment are conducted involving towing at three different velocities of 50, 100 and 200 mm s-1, all at a fixed distance. While towing the object, the output signals

(b)

Figure 10. Photograph of the experimental setup used for underwater object sensing. (a) The test section of the water tunnel with the mounted sensor array, the program controlled water tunnel gantry system for towing the object, the cylindrical stimulus and a reference sensor, (b) a close-up view of the sensor array and the object.

10


(a)

0 2 4 0

'iU 'iU ~ ~ Q) ... 0 ~ -100 II) - 200 ~ -300 0.. 0 2 3 4

Time (s)

(c)

0 2 3 4 5 6 7 8

70

60

E 50 g c .~ 40 .. ·;;;

&. 30 0

'iU II)

~ 20 0.. Cll

10

2 3 4 Time(s)

5 6 7 8


(b)

2 3 4 5

2 . 3 T1me(s)

4 5

(d)

I I I I I I I I I I

0 Aetu.al 'clocity SO mrm;· l E1pcrir•t'lllall~ dett rruia(d SJJ rur~afl

0 Arlusl H'lotitylll() mms"I I;'JK'rimu tallydtttrmint'd 99.!3 m111f l

0 AC'lUJI ,·t i<K'ity 200 mmfl l:\l)f'rim(nt.:~tlytkttrminfd 198.9-1 mm_,-1

I{ IJ ,_(J IT

I I .l.

I j I -2.0 2.2 2.4 2.6 2.8 3.0 3.2 3.4 3.6 3.8 4.0

Time (s)

Figure 11. Experimental results of object velocity sensing. Data collected from all five sensors simultaneously when an object is towed past sensors Sl-S5 at (a) 50 mm s-1, (b) 100 mm s-1 and (c) 200 mm s-1; (d) the slopes of the peak dips obtained in (a)-( c) that estimate the actual towing velocities.

from the array of all ten sensors are simultaneously collected through ten different inputs of the NI-DAQ and acquired in LAB VIEW.

The results presented in this section are the output signals obtained from five sensors Sl-S5. Due to the symmetry, a very similar pressure signal is generated at the sensor rows Sl-S5 and S6-S10 during the act of towing the cylinder. Figures 11(a)-{c) describe the signal output of sensors Sl-S5 acquired simultaneously, when the stimulus is towed at 50, 100 and 200 nun s-1 respectively. As can be observed from the plots, the sensors demonstrate a very good correlation in the output signal. The sensors are able to measure the flow velocity with high accuracy with an average percentage error of 2.5% calculated from all the towing experiments. A peak increase in the pressure is generated at the sensor when the

1l

object is right in front of the sensor and is transduced as a peak drop in voltage signal. Due to the time lag of approach of the object towards consecutive sensors in the array, there will be a time-shift between the peak voltage dips obtained at each sensor. From this time lag between peak voltage dips obtained from the experiment, and the known distance between sensors, the velocity of the object passing by can be easily calculated using the formula d = Vt, where dis the interval of the spatial sampling, which is the same as the center-to-center distance between sensors, V is the velocity of the stimulus and t is the time lag. This is an accurate approximation whenever the flow is steady from the frame of moving the stimulus. It does not hold well in the wake region of the cylinders, where, in addition to being turbulent, the presence and location of vortices is highly time dependent. Figure 11 illustrates a series


of plots that describe the time delay of occurrence of peak voltage dips among successive sensors of the array that are generated due to the gaps between sensors and a finite velocity of approach of the object. From the plots in figures ll(a)-{c), by calculating the slope, the velocity of the object can be determined. The slopes of the plots calculated in figure ll(d) estimate the actual velocities of stimulus of 50, 100 and 200 mm s-1 as 53.3, 99.23 and 198.9 mm s-1 respectively, which closely match the actual towing velocities.

A cylinder moving at higher velocity generates a larger amplitude of pressure response. This is evident in the sensor responses in figures ll(a}-(c) and follows the potential flow model. A potential flow model of a circular cylinder in an infinite steady flow shows that the pressure behaves with a velocity-squared relation

p = U2C(x, r, p) (11)

where U is the velocity of the cylinder, x is the position from the center of the cylinder, r is the radius of the cylinder and p is the density of water. The results presented in figure 11 might not completely follow this trend because the potential flow model of a circular cylinder is not an ideal comparison in this case, but it gives an estimate. The potential flow model presented in equation (11) is based on a round cross-section and does not take into account the wake region; it can estimate the expected trend of pressure signals transduced by the sensor.

4.3.2. Underwater object distance sensing. The goal of the experiment demonstrated in this section is to determine whether or not it is possible to distinguish between various distances of object passes by employing the sensors. The stimulus is towed at a constant velocity of 100 mm s-1 at different distances (d =50, 100, 150 and 200 mm) from the plane of the array of sensors. The results obtained from a single sensor for towing the object at different distances are illustrated in figure 12. The sensors can detect objects as far as 200 mm away from them which is quite on a par with the biological sensors of the blind cave fish. Using information from its lateral-line the fish can detect objects about 2 body lengths (the body length of an adult Astyanax mexicanus fasciatus is approximately 100-130 mm) away from itself [29]. As the distance between the object and the sensor array (d) increases, the peak dip of the sensor output signal is reduced approximately proportional to the inverse square of the distance d. The observed trend of peak pressure decrease with distance of the object as observed in figure 12 matches with the theory. Considering the case of a potential flow around a circular cylinder, it can be derived that the pressure (P) along the plane of the sensor array depends in an inverse square manner on the distance d as described by the following equation:

(12)

where R is the radius of the cylinder, p is the density of water, U is the velocity of flow past the cylinder and din this

12

1

0

ti -11 c..

oft ~ :l -22 1/) 1/)

~ c.. -33

-44 0

~

~

0 2


I.<! ~ • .O. L

IJ ' 'TA ... ""

~ rr

¥11 ~somm

~100mm

\} ~150mm I-

~200mm

I 4 6 8 10

Time(s)

~gure 12. Experimental results of sensor outputs when the object 1s towed at a constant velocity of 100 mm s-1 at different distances from the sensor.

case is the distance of the sensor array from the center of the cylinder. The pressure dependence on d can be approximated to inverse square neglecting the fact that the second term that is negligible compared to the first term.

Therefore, using the experimental results presented in this paper and the techniques presented in [13] the array can be employed to interrogate the velocity and distance of the object, although that analysis is not performed here since it falls outside the scope of this paper.

5. Conclusion

Nearly all UAVs use active sensing systems like sonar and optical methods for object detection in water. However, sonar is impeded by problems of high power consumption, large size and poor resolution of sensing in sonar blind zones. Blind cave fish elegantly perform passive object detection in water using arrays of pressure sensors on their bodies. This paper develops a flexible array of LCP based MEMS pressure sensors that could be readily mounted on the streamlined surfaces of underwater vehicles. This work, through proof-of-concept experiments, demonstrates that an array of MEMS pressure sensors could be used to detect the distance and velocity of objects passing by the array. This work also investigates LCP as the sensing material choice to develop MEMS sensors for a harsh seawater environment. The ideal properties of LCP like chemical robustness, high hermeticity and high fracture strength qualify it to be better than silicon for harsh environment applications. As an MEMS structural layer material, LCP also offers better sensitivity as compared to silicon. Most work in the literature has taken a biomimetic approach, mimicking the hair cell like structural features of fish. With the focus of application, and to fabricate a reliable array for real-time applications, we took a bio-inspired approach focusing on the functional implications of the lateral-line for AUVs. A simple and inexpensive fabrication process is developed. The fabrication process developed does not employ thin and fragile hot wires or delicate cantilever beams which are not feasible in real AUV


applications. The sensors have been tested for water flow sensing and perform well with good sensitivity and resolution. Object detection experiments were conducted by towing a cylindrical object (stimulus) at different velocities and distances past the array. The array is capable of identifying the velocity of stimulus with a high accuracy and an average percentage error of only 2.5%. Finally, the work presents the development of a flexible LCP MEMS sensor array for AUV applications with promising proof-of-concept experimental results demonstrating the flow sensing and object detection abilities of the array.

Acknowledgments

The authors would like to thank Dr Chee Wee Tan of the Singapore-MIT Alliance for Research and Technology (SMART) and Dr Nay Lin and Ms Meghali for their inputs, ideas and assistance with the experiments. This research was funded by the Singapore National Research Foundation (NRF) through the Singapore--MIT Alliance for Research and Technology (SMART) Centre, Center for Environmental Sensing and Modeling (CENSAM) IRG.

References

[ 1] Griffiths G 2002 Technology and AppUcations of Autonomous Underwater Vehicles (London: Taylor and Francis)

[2] Fernandez V I, Hou S M, Hover F S, Lang J H and Triantafyllou M S 2007 Lateral-line inspire MEMS-array pressure sensing for passive underwater navigation parts I and ll Proc. 15th Int. Symp. on U1111101111ed Untethered Submersible Technology (Durham, NH)

[3] Yang Y C, Chen J, Engel J, Pandya S, Chen N, Tucker C, Coombs S, Jones D L and Liu C 2006 Distant touch hydrodynamic imaging with an artificial lateral line Proc. NatlAcad. Sci. USA 103 18891-5

[4] Schrope M 2002 Whale deaths caused by US navys sonar Nature 415 106

[5] Richter M, Wackerle M, Woias P and Hillerich B 1999 A novel flow sensor with high time resolution based on differential pressure principle Proc. 12th Int. Conf on Micro Electro Mechanical Systems (Orlando, FL) pp llS--23

[6] Akyildiz IF, Pompili D and Melodia T 2004 Challenges for efficient communication in underwater acoustic sensor networks ACM SJGBED Rev. 1 3--8 (Special issue on embedded sensor networs and wireless computing)

[7] Montgomery J C, Coombs S and Baker C F 2001 The mechanosensory lateral line system of the hypogean form of Astyanax fasciatus EvoL BioL Fish. 62 87-96

[8] Von Cam.penhausen C, Riess I and Weissert R 1981 Detection of stationary objects by the blind cave fish Anoptichthys jordani (characidae) J. Comp. PhysioL 143 369-74

[9] Montgomery J, Coombs S and Halstead M 1995 Biology of the mechanosensory lateral line in fishes Rev. Fish. BioL Fisher. S 399--416

[10] Montgomery J C, Baker C F and Carton A G 1997 The lateral line can mediate rheotaxis in fish Nature 389 960--3

[11] Burt de Perera T and Braithwaite 2005 Laterality in a non-visual sensory modality-the lateral line of fish Curr. BioL 15 R241-2

[12] Wmdsor S P, Tan D and Montgomery J C 2008 Swimming kinematics and hydrodynamic imaging in the blind Mexican cave fish (Astyanax fasciatus) J. Exp. BioL lll 2950--9

13


[13] Fernandez VI, Maertens A, Yaul F M, Dahl J, LangJ Hand Triantafyllou M S 2011 Lateral-line inspire sensor arrays for navigation and object identification J. Marine Tech. Soc. 45 130-46

[14] Peleshanko Set al2007 Hydrogel-encapsulated microfabricated haircells mimicking fish cupula neuromast Adv. Mater. 19 2903--9

[15] Chen N, Tucker C, Engel J, Yang Y, Pandya Sand Liu C 2007 Design and characterization of artificial haircell sensor for flow sensing with ultrahigh velocity and angular sensitivity J. Microelectromech. Syst. 16 999-1014

[16] Venturelli R, Akanyeti 0, Visentin F, Jezov J, Chambers L D, Toming G, Brown J, Kruusmaa M, Megill W M and Fiorini P 2012 Hydrodynamic pressure sensing with an artificial lateral-line in steady and unsteady flows Bioinspir. Biomim. 7 036004

[17] Akanyeti 0, Venturelli R, Visentin F, Chambers L D, Megill W M and Fiorini P 2011 What information do Karman streets offer to flow sensing? Bioinspir. Biomim. 6 036001

[18] Klein A and Bleckmann H 2011 Determination of object position, vortex shedding frequency, and flow velocily using artificial lateral-line canals Beilstein J. NanotechnoL 2276--83

[19] Qualtieri A, Rizzi F, Epifani G, Emits A, Kruusmaa M and Vittorio M D 2012 Parylene coated bioinspired artificial haircell for liquid flow sensing Microelectron. Eng. 98 516--9

[20] Salumae T, Rano I, Akanyeti 0 and Kruusmaa M 2012 Against the flow: a Britenberg controller for a fish robot (ICRA '12 ): IEEE Int. Conf Robotics and Automation (St. Paul, MN) pp 4210--5

[21] Olfatnia M, Xu T, Ong L S, Jing X M and Norford L 2010 Piezoelectric circular microdiaphragm based pressure sensors Sensors Actuators A 163 32--6

[22] Tan C W, Kottapalli A G P, Wang Z. Xu J, Miao J M, Barbastathis G and Triantafyllou M 2011 Damping characteristics of a micromachined piezoelectric diaphragm-based pressure sensor for underwater applications (Transducers'll): 16th Int. Conf Solid-state Sensors, Actuators and Microsystems (Beijing) pp 72-5

[23] Kottapalli A G P, Tan C W, Olfatnia M, Miao J M, Barbastathis G and Triantafyllou M 2011 A liquid crystal polymer membrane MEMS sensor for flow rate and flow direction sensing applications J. Micromech. Microeng. ll 085006

[24] Xuefeng W, Jonathan E and Liu C 2003 Liquid crystal polymer (LCP) for MEMS: processes and applications J. Micromech. Microeng. 13 62S--33

[25] Culbertson E C 1995 A new laminate material for high performance PCBs: liquid crystal polymer copper clad films Proc. 45th Electronic Components and Technology Conf pp 520--3

[26] Schomburg W K, Rummier z. Shao P, Wuff K and Xie L 2004 The design of metal strain gauges on diaphragms J. Micromech. Microeng. 14 1101-8

[27] Senturia S D 2001 Microsystems Design (Boston, MA: Kluwer Academic Publishers)

[28] Clark S K and Wille K D 1979 Pressure sensitivity in anisotropically etched thin-diaphragm pressure sensors IEEE Trans. Electron. Dev. 26 1887-96

[29] Braun C B and Coombs S 2000 The overlapping roles of the inner ear and lateral line: the active space of dipole source detection Phil. Trans. R. Soc. B 355 1115-9

a flexible liquid crystal polymer mems …seagrant.mit.edu/publications/mitsg_12-18.pdf · a...

Documents