Implantable MEMS sensors are an enabling technology for diagnostic analysis
and therapy in medicine. The encapsulation of such miniaturized implants
remains a largely unsolved problem. Medically approved encapsulation
materials include titanium or ceramics; however, these result in bulky and
thick-walled encapsulations which are not suitable for MEMS sensors. In
particular, for MEMS pressure sensors the chip surface comprising the
pressure membranes must be free of rigid encapsulation material and in direct
contact with tissue or body fluids. This work describes a new kind of
encapsulation approach for a capacitive pressure sensor module consisting of
two integrated circuits. The micromechanical membrane of the pressure sensor
may be covered only by very thin layers, to ensure high pressure sensitivity.
A suitable passivation method for the high topography of the pressure sensor
is atomic layer deposition (ALD) of aluminium oxide (Al
The development of small MEMS (micro-electro-mechanical systems) sensors with
functional surfaces enables the ability to perform diagnostic and therapeutic
tasks inside the human body in the form of a long-term implant. Application
examples include continuous pressure monitoring (Mokwa, 2007; Gembaczka et
al., 2013), nerve and muscle stimulation (Eick et al., 2009) and drug
delivery systems (Hang Tng et al., 2012). Usually medically
approved micro-electronic implants such as pacemakers (Park and Lakes, 2007),
cochlear implants (Loeb, 1990) or brain stimulators (Rezai et al., 2002) are
encapsulated by non-flexible materials such as titanium or ceramics with
considerable sidewall thickness in the range of 100
The measurement of pressure in different areas of the human body, such as blood pressure (in large arteries or veins), intraocular pressure or intracranial pressure is important for medical diagnostics and therapy (Mokwa, 2007). The continuous monitoring of the corresponding pressure values with a permanent implant can be beneficial for the treatment of medical conditions such as intracranial and arterial hypertension, heart insufficiency and glaucoma.
The development of a miniaturized encapsulation technology for MEMS pressure sensors is particularly challenging as the pressure outside of the housing must be transferred to a pressure-sensitive membrane. A number of approaches for the encapsulation of implantable pressure sensors have been suggested previously (Mokwa, 2007). Silicone is often used as encapsulation material because it is soft and can transmit the pressure well (Cleven et al., 2012; Bradford et al., 2010; Stangel et al., 2001). However, silicones absorb water and the material properties change over time leading to a drift of the pressure value measured by the sensor. Preconditioning of the material and offset compensation of the sensor can alleviate the problem (Gräfe et al., 2009). Nevertheless, water diffusion remains problematic especially with electrical components inside the human body. Materials like Parylene-C were applied to further reduce the water uptake (Schlierf et al., 2005) or the membrane of the pressure sensor was directly made of parylene (Ha et al., 2012; Chen et al., 2008). However, a hermetically sealed encapsulation is not achievable based on polymer alone, therefore novel solutions must be developed.
The pressure sensor employed in this work is fabricated in CMOS technology
allowing the integration of a capacitive pressure sensor with the read-out
electronics on one chip. For some applications a second chip is required for
signal processing. A hard and thick passivation would change the
characteristics of the sensitive membranes of the pressure sensor chip so
strongly that no sufficient pressure transmission would be possible. In
previous preliminary work, an encapsulation concept for the pressure sensor
surface based on atomic layer deposition (ALD) was successfully tested (Betz,
2011). A passivation layer consisting of 50 nm Al
Photo of the pressure sensor chip after 1 h in PBS at
150
In a novel encapsulation concept, therefore, the electrical contacts of the
chip, the bond wires and the dicing-edges have to be protected before ALD
passivation. ALD deposition is performed at a temperature of 275
Particularly problematic is that polyimide–epoxy significantly shrinks during the curing phase. A process had to be developed to circumvent the shrinkage. For the first time, this work reports on a biocompatible encapsulation method combining ALD and polymers.
The paper is organized as follows: in Sect. 2 an assembly of a two-chip pressure sensor module (pressure sensor chip including read-out electronics and signal processing chip) is introduced to facilitate the development of novel encapsulation and assembly technique. Section 3 describes the high-temperature resistant polyimide–epoxy composite and its deployment. In Sect. 4, the ALD layer is discussed and test results regarding the hermetic seal of the sensor are presented. Section 5 presents pressure sensor module calibration and measurement results at various temperatures, and Sect. 6 accelerated life time testing results. Finally, in Sect. 7 some conclusions are drawn.
The pressure sensor module mainly consists of two application-specific integrated circuits (ASICs). The first is the capacitive pressure sensor chip (PS) including the read-out electronics, the second a signal post-processor chip (SPP). Both are mounted on a silicon chip carrier with bond pads and signal tracks. Additionally, some surface mount components (SMD) are required. The cross-section of the pressure sensor module is depicted in Fig. 2. In the first step, the chips were glued on the chip carrier using a polyimide–epoxy composite and electrically connected to the carrier with bond wires. The bond wires and the entire module, except for the pressure sensor membranes, were covered with the same polyimide–epoxy composite (as described in Sect. 3) used for die attachment to minimize thermal expansion stress.
We choose here the approach to encapsulate the chips by polyimide–epoxy first, and then by an ALD layer to achieve a hermetically sealed passivation. Considering the opposite approach, application of a polymer potting layer on top of the ALD layer is not a viable solution, as the bond wires mechanically deform die to shrinkage of the polyimide–epoxy in the curing process, which, in turn, would lead to miniature fractures of the ALD layer. Another disadvantage is the water uptake of the polyimide–epoxy which leads to changes of the material properties of the polymer over time. The water uptake of polymer is very low but the pressure sensor is highly sensitive to any kind of mechanical stress and, in general, any change of the material properties leads to drift of the measured pressure value. Here, it was therefore chosen to apply first a layer of polyimide–epoxy followed by ALD deposition to create a water impermeable, hermetic encapsulation.
A ceramic disc with an opening for the chip carrier was placed next to the
ASICs and fixed with the same polymer. The disc serves as a separation
between the ASICs that are exposed to the medium on one side and the discrete
circuit components on the other side; this is depicted in Fig. 2b. The
overall assembly of the separation disc and the discrete components is not
the subject of this paper and therefore shall not be described in further
detail. The final step (described in more detail in Sect. 4) is the
passivation of the entire assembly in an ALD tool with aluminium oxide
(Al
Only the hermetically sealed area located left of the ceramic disc is later exposed to the liquid medium.
The carrier for the entire sensor module is made of a silicon chip with aluminium tracks (for the discrete components), pads and space for the pressure sensor chip and the SPP, as shown in Fig. 3. The chip carriers are processed at wafer level and diced into separate chips.
Photo of the silicon chip carrier.
Silicon as a carrier material was selected because it has the same thermal expansion coefficient as the ASICs. The thermal expansion of different materials is critical in this passivation method, as thermal stress can lead to fractures.
The pressure sensor chip has been fabricated in a CMOS technology allowing
the integration of capacitive pressure sensor membranes with the readout
electronics in a surface micromachining process. The chip size is
1.8
Photograph of the capacitive pressure sensor chip before encapsulation (Betz, 2011).
The bottom electrode of the pressure sensor is made of
Section through a pressure sensor membrane (Trieu, 2011).
In order to achieve a sufficiently large change of the capacitance, 20
identical pressure sensor membranes are combined into an array. The pressure
sensor chip is not only sensitive to the applied pressure, but due to its
micro-mechanical structure, also to mechanical stress induced, for example,
by the different thermal expansion coefficients of the chip and the adhesive
materials. Consequently, a soft silicone is typically used as adhesive to
mechanically decouple the chip. However, due to the ALD deposition
temperatures of 275
To get a functional pressure sensor module a second ASIC is required to
digitize the analogue pressure sensor signals. The SPP chip shown in Fig. 7
was fabricated in a 0.35
The width of the ASIC is identical to the width of the pressure sensor chip. The ASIC has a temperature sensor integrated based on a bipolar transistor with the necessary read-out circuits. The temperature measurement is used to compensate the temperature dependence of the pressure sensor. In addition to an analogue-to-digital converter the SPP includes a voltage regulator, a local oscillator and an EEPROM for permanent storage of calibration and identification data (Gembaczka et al., 2013). With a two-wire connection cable, the digitized sensor data is transmitted to a data reader. The connection is used both for data transmission and power supply.
Cross-section SEM image of an area of the pressure sensor illustrating the difficult topography for encapsulation.
For the first encapsulation layer of the pressure sensor module a polymer had to be found satisfying all requirements. The same material should be used for the die-attach and the potting of the pressure sensor module including the bond wires. This has the advantage that thermal expansion coefficients are identical avoiding tension and thus cracks in the assembly.
Ideally, the material should satisfy the following requirements:
long-term temperature stability up to 275 high glass transition temperature low thermal expansion thermally conductive electrically insulating good adhesive properties smooth surface without pores dispensable thixotropic CMOS compatible.
The desired properties, in particular the temperature stability, drastically
reduce the list of candidates of possible polymers. With regard to the
processability and adhesion, an epoxy is a good choice for potting. Epoxy
resins with different modifiers are available to improve material properties,
for example rubber additives for flexibility (Boyle et al., 2001). One
disadvantage is the decomposition and deformation of polymers (Beyler and
Hirschler, 2001; Cheng et al., 2009). Some electrically insulating high-temperature epoxy composites already exist that allow operating temperatures
up to about 260
Photograph of the SPP-chip.
Alternatively, polyimides are materials satisfying nearly all requirements
and are well established for applications in microsystem technology
(Walewyns et al., 2013). However, processing of polyimide is challenging due
to the evaporation of the solvent (for example N,N
In particular, polyimide–epoxy composite was identified as a suitable material; here a small amount of epoxy (typically 10–15 %) is added to the polyimide partially replacing the solvent (Gaw and Kakimoto, 1999). This increases adhesive strength, shape stability and reduces the formation of pores.
For the encapsulation of the pressure sensor module we used
The potting of the chips and the bond wires with a polyimide–epoxy composite is cumbersome due to the lack of dimensional stability during the application process. The polyimide–epoxy composite flows over the chip surface depending on the material and topography making it difficult to control and predict area coverage. The pressure sensor membranes need to remain free of the polyimide–epoxy composite. Pre-heating of the chip carrier allows application of the polyimide–epoxy composite in certain areas only. The polyimide–epoxy composite was applied manually using a dispenser. For this reason the edge of the polyimide–epoxy composite around the area of the membranes is irregular.
A further critical point is the shrinkage of the material due to the
evaporation of the solvent during curing. Preliminary tests showed that the
25
Casting of the entire sensor module before the ALD passivation.
To avoid this problem, a high degree of imidization was achieved by heating
the pressure sensor module up to 290
The requirements for the ALD passivation material are as follows:
good adhesion to all materials (silicon, silicon nitride, polyimide–epoxy composite) must not alter the mechanical characteristic and properties of the pressure membranes conformal deposition of high surface topography biocompatible electrically insulating impermeable to water vapour corrosion resistant deposition at moderate temperatures to preserve the underlying potting material and CMOS circuitry.
Photograph of a test chip (left) and SEM image of the transition
region at a 45
Based on a previous study (Betz, 2011) aluminium oxide (Al
It should be noted that the pressure sensor membrane deflection with 100 nm
Ta
Table 1 summarizes the process parameters for the deposition of the ALD layers.
Schematic representation of the test setup for the amperometric defect investigation.
Intact probe after an amperometric defect test (left) and a sample having a defect in the ALD passivation (right). The sacrificial AlSi layer is completely decomposed.
The interface between the chip surface and the polyimide is a critical area which had to be examined in more detail since fractures due to stress caused by the cooling process may damage the ALD passivation layers. Another failure mechanism is particle contamination. In order to locate defects in the passivation a destructive amperometric measurement test was carried out. For this purpose a silicon test wafer was fabricated with the following layers: 10 nm Ti, 40 nm TiN, 900 nm AlSi deposited by sputtering.
The wafer was diced with a wafer-saw into individual chips with a size of
25
For the amperometric defect investigation, the test chips were prepared by gluing a plastic tube with silicone in the middle of the encapsulated region thus creating a container for PBS. The AlSi layer directly underneath the passivation test layers was used as the anode. The cathode and the reference electrodes were placed in the PBS. Figure 10 shows the measurement setup.
Current flow at a defective passivation at 3 V.
For the measurements an impedance analyser (
With the test samples the following amperometric measurements were carried out at room temperature: increasing the voltage from 1 to 5 V in 1 V steps for 5 minutes each and continuous measurement for 25 h at 5 V. A voltage of 5 V was selected since many medical implants operate at this voltage.
The graph in Fig. 12 shows the current versus time for a sample with a defect. For the measurement using an increasing voltage the sample showed a first measurable current flow at a voltage of 3 V which continuously increased with time.
SEM image of boundary between polyimide–epoxy composite and chip surface with an ALD layer (not visible) over the entire area.
Process parameters for the deposition of the passivation layers.
A sample without defect showed no measurable current flow, even after 25 h at 5 V.
The preliminary tests showed that the adhesion of the polyimide–epoxy composites was sufficient despite the high-temperature stress it was exposed to. Dust particles were identified as the main source for defects; this is attributed to the fact that the test chips were not coated at wafer level. We tested four samples which were cleaned only with a stream of nitrogen before ALD. All of these samples showed defects. Four further samples were cleaned first with acetone, isopropanol and finally with DI-water; these showed no defects. The experiments indicated that the cleaning of the surface and full processing in clean-room conditions are important for a defect-free coating.
The amperometric defect investigation showed that the transition region
between silicon and polyimide–epoxy composite exhibited sufficient adhesion
and that ALD passivation of polyimide–epoxy composite without fractures is in
principle possible. Therefore, the pressure sensor modules were coated in the
ALD tool. After the coating, a sample was analysed in the SEM. Figure 13
shows a close-up of the transition region between the chip surface and the
polyimide–epoxy composite at an angle of 45
Figure 14 shows the encapsulation of the entire pressure sensor chip as a SEM picture. The brightness is due to charging of the tantalum pentoxide layer in the SEM, which has the benefit that defects can be detected easily.
Additionally, wafer pieces of pressure sensor chips were coated in the same ALD process step to control the layer thickness and conformity. Figure 15 shows an area of the pressure sensor after the ALD coating. The bright tantalum pentoxide layer on the surface is clearly visible and indicates excellent conformal coverage.
SEM image of a defect-free passivation.
Cross-section SEM image of the ALD passivation layers.
Test measurements to evaluate and characterize the pressure sensor module
after encapsulation were carried out. Prior to the measurement the sensor
modules were calibrated allowing the temperature dependency of the pressure
sensor to be compensated. The calibration took place in a pressure range from
800 to 1400 hPa and a temperature range of 24 to 40
To apply a defined pressure in a test fluid the pressure sensor module needed a suitable housing. The housing for the sensor module was made of polyether ether ketone (PEEK) which offers excellent chemical resistance and good temperature stability. The housing is designed in a modular way to facilitate different types of measurements. Figure 16 shows the calibration setup consisting of the PEEK housing and a transparent tube screwed onto the housing. Test liquid could then be dispensed into the tube creating a well-defined pressure.
Photo of the housing with the transparent tube for holding the test fluid and creating a well-defined pressure.
The resulting liquid column provided a sufficient fluid reservoir. It also allowed connecting an air hose to apply a defined external pressure required for the calibration. The entire setup is placed in a climatic chamber to set the desired temperature.
With the presented measuring system it is possible to measure not only
pressure but also temperature. The measuring system consisted of the housed
pressure sensor module placed in a climate chamber and connected with a hose
to the pressure calibrator. The pressure sensor module transmits the measured
values through a two-wire cable to a data reader device outside of the
climatic chamber. The data reader contains the interface and energy
controller (IEC) and a microcontroller. The IEC represents the link between
the pressure sensor module and the microcontroller to control the
communication and power supply. The data reader transmits the data via USB to
a PC running a purpose programmed
The digital output of the pressure sensor module is in a form of values
between 0 and 8191. The calibration of the pressure sensor module associates
the digital values with the real pressure measured by a reference pressure
sensor (
Overview of the pressure measurement system.
The pressure curves used for calibration were recorded at various
temperatures in 0.9 % saline solution. Figure 18 shows an example of a
calibration measurement against the reference pressure sensor. The pressure
curves were recorded in a range of 800 to 1400 hPa in a temperature range
between 24 and 40
Pressure value of an encapsulated sensor module at various temperatures in saline solution.
Pressure–calibration error (including 1-sigma noise) at different temperatures.
With the measurement output values for the pressure (
For low temperature coefficients of the pressure sensor module (as for the measurement result shown in Fig. 18) calibration with the polynomial is possible. In Fig. 19 the calibration error of the pressure sensor module is shown, which is below 2 hPa (including 1-sigma noise). Based on the measurement range this corresponds to a total error of about 0.34 % full scale span.
Figure 20 shows the temperature sensor values at different pressures between
24 and 40
Temperature value of an encapsulated sensor module at various temperatures in saline solution.
Temperature–calibration error (including 1-sigma noise) at different temperatures.
Pressure curves of different calibrations at 36
To estimate the long-term stability accelerated life tests of the pressure
sensor modules were carried out. Several calibrated pressure sensor modules
were placed in a sealed housing filled with 0.9 % saline solution. The
housed pressure sensor modules were stored in an oven at a temperature of
60
Error of the pressure curves after a different accelerated life
test time at 60
An upward shift of the pressure curves can be observed with increasing aging time. The calibration polynomial of the first measurement before the aging period was used to calculate the pressure value error. The measurement total calibration error in hPa is shown in Fig. 23. It should be noted that the housed pressure sensor module had to be removed from the oven and the housing opened before every control measurement to allow mounting the transparent tube for the pressure calibrator. This process is likely to cause mechanical stress on the housing and thus influence the sensor output. The error after 524 h and after 744 h of aging was about 17 hPa. This corresponds to a total error of about 2.8 % full scale. Between the measurement curves at 744 and 1482 h the complete pressure sensor module housing was opened for a visual investigation. The last measurement curve after 1482 h indicates that the system changed after the inspection as the error increased up to 30 hPa.
The sensor modules failed in continuous operation after a total of 1500 h at
60
The interpretation of the measurement data with respect to a prediction of
long-term stability is not straightforward. Further accelerated aging tests
at 100 % humidity and a temperature of 120
An encapsulation process has been developed to use a high-temperature
resistant polyimide–epoxy composite as a die-attach material and sealing
compound for the bond wire and parts of the chip surface excluding the MEMS
pressure membranes. The process prepares the pressure-sensor module for ALD
passivation. It could be shown that a conformal ALD of Al
Photo of the sensor module with the wider silicon chip carrier.
Moreover, it has been shown that a complete pressure sensor assembly can be encapsulated and hermetically sealed. The passivated pressure sensor was calibrated and tested in a 0.9 % saline solution showing excellent results.
An accelerated life test at 60
Further optimization and automation of the fabrication process may allow a
new kind of hermetic encapsulation for human implants based on MEMS sensors.
As a next step, it is intended to test the pressure sensor module in blood
and in an animal model in the near future. However, similar results are
expected at least for the Ta