Scalable hybrid microelectronic-microfluidic integration of highly sensitive biosensors

Amongst all disciplines of packaging of microelectronics, sensor packaging sometimes seems to be the most challenging. When packaging a standard microelectronic component, the package provides protection against environmental load and ensures the reliability that is needed, it does reroute the electrical interconnects from the die so that the package can be mounted easily. Packaging a sensor often brings in some more challenges: provide access of selected media to the sensor surface while keeping others away, protect temperature sensitive or mechanically delicate sensor surfaces from thermal and mechanical load, combine the sensor with an adjacent ASIC for optimum signal integrity, i.e. leading to heterogeneous integration and do all that in an economically reasonable way. Some of these challenges are true for the sensor packaging tasks encountered in the POC-ID project [4], where especially the timing for the functionalization of the bio-sensor ICs during packaging was of ample importance. The basics for bio-sensor functionalization, the basic sensor principles and the technology flows are described in the following sections.

For the specific detection of the biomarkers, biomimetic recognition molecules – aptamers – are employed. Aptamers are made of nucleic acids and generated by a process called in vitro selection to bind highly specifically to a given target molecule with high affinity. They are chemically synthesized and can be modified by functional groups, e.g., for immobilization on sensor surfaces without compromising their functionality. Moreover, aptamers are fairly sturdy to heat up to around 100 °C, can be dehumidified to dryness and re-dissolved again to be fully active. They are stable for years in the dry state even at room temperature but also when dissolved in H2O and stored at 4 °C. [5]

For sensing purposes, these aptamers have to be immobilized on the sensor surfaces. Since they are considerably smaller and more lightweight than antibodies, their use can be advantageous. This is generally the case for field-effect and mass-based sensors. As the materials of sensor surfaces and the detection principles differ, different functionalization procedures can be employed.

In graphene FET based sensors, graphene is used as both: the channel material of an amplifier and the sensor material. The devices are able to detect changes in the conductivity of the channel due to electric fields generated by charged molecules or ions. [6] GrFET biosensors can intrinsically possess a very high sensitivity up to the fM level [7]. However, this sensitivity is difficult to achieve in combination with high selectivity and specificity of the biomolecular binding events. Covalent functionalization of GrFETs results in impairing the electronic structure of graphene, which significantly reduces the device mobility. The functionalization of graphene via physisorption is typically less stable, causes larger distances between the analyte and graphene plane, or reduces the density of probe molecules. These effects reduce sensitivity and stability of GrFET biosensors. In the POC-ID project a sensor is developed, which uses a carbon nanomembrane (CNM) [8, 9]. CNMs are molecular nanosheets with an about 1 nm thickness and tailored physical or chemical properties. [8]. A CNM/graphene hybrid structure has the potential to master the challenges of graphene [9]. The highest sensitivity is expected as the conducting channel is closest to the sensor surface (distance of only 1 nm). Furthermore, no additional charge carrier scattering is generated in graphene when an amino-terminated CNM is placed on top of forming an all-carbon van der Waals heterostructure [9]. Aptamers can be covalently immobilized on the CNM-surface for detection of the specific biomarkers. A schematic of such a device can be seen in Figure 1.

The detection of the analytes by the sensor is done by measuring the electron transport characteristic of the FET where the drain-source current is monitored as a function of the gate voltage. An array of 15 devices fabricated on silicon/silicon oxide substrate in the frame of the PoC-ID project is shown in Figure 2.

The ultra-high mobility of charge carriers in graphene is important for the high sensitivity of the FET. Charged impurities lead to charge carrier scattering and thus reduce mobility and sensitivity of the sensor. Also, high temperature treatment needs to be avoided as it may lead to more intimate contact of graphene with the underlying substrate, which also creates additional charge carrier scattering in graphene. Furthermore, bombardment with high energetic particles can create defects in the atomic structure of graphene, which in turn can form covalent bindings with the substrate or molecules in solution. These effects can dramatically decrease the sensitivity, create hysteresis, or long response times of the sensors. In addition, van-der-Waals heterostructures are missing covalent bindings to the substrate and avoiding mechanical damage to the FET via delamination is a particular challenge for the packaging process.

The BioMEMS chip has biosensors based on a Flexural Plate Wave (FPW) element with the feature of a very thin membrane. The thinning is done using anisotropic etching of the Si-membrane employing the Bosch process. This leads to an increased mass sensitivity of the device in comparison to Surface Acoustic Wave (SAW) elements.

The very thin membrane is a challenge for the assembly of the BioMEMS chips because of their sensitivity to mechanical and thermal stress.

The fabrication of a FPW-sensor involves different technological steps. One of the main challenges in the course of the fabrication is the patterning of the piezoelectric PZT-layer, a key layer constituting the sensor. As a result, a multiple measurement array of 8 different FPW can be placed on top of one setup, called devices. These measures allow also for side effect compensation, since typical disturbances, such as stress and temperature, are comparable for all sensor elements. The layout of the sensor is depicted in Figure 3.

Additionally, a new protocol was developed, which allows for direct functionalization with aptamers of the native membrane material, silicon dioxide. In this way the typical used additional gold layer on the sensor membrane becomes obsolete and the setup is more compact with a reproducible performance. The coupling method is based on Click Chemistry approach, which is proved to give reliable results. [10] Soluted aptamers are dispensed into the cavity and incubated on the thin membrane.

Figure 4 shows the spectrum of a FPW sensor device after integration of a BioMEMS chip into the microelectronic package. It was assembled after the process presented further down. The insertion loss at resonance reads −18 dB and is comparable with literature values and the reference measurement [11]. The signal to noise ratio reads +40 dB and thus the integration is considered successful.

The hybrid microelectronic-microfluidic package must enable the electrical readout and also the exposure of the active sensor surface to the sample fluid. Therefor designed, a 3D-printed so called luer connector (Figure 5, left) enables the connection of three syringes, one syringe for the probe, one for the buffer and one for the waste. The luer connector is attached to the microfluidic side of the package, enabling the fluid to flow over the active sensor surface.

Pads on the electrical side enable the electrical connection to a readout electronic (Figure 5, right). Placed into the readout device they can electrically connect to diverse connectors as e.g. pogo pins.

The main challenge of combining microfluidics with a silicon sensor is the electrical connections being on the same side as the active sensor area which must be brought into contact with the fluid. Various solutions are conceivable:

1. A narrow microfluidic channel can be passed between the electrical contacts.

2. Through silicon vias can lead the electrical contacts to the other side, separating the electrical and the fluidic side of the package. This enables a planar surface of the microfluidic side.

3. Through mold / PCB vias can lead the contacts to the other side, separating the electrical and the fluidic side of the package. This enables a planar surface of the microfluidic side.

It was decided to use the third variant because of the following advantages:

The planar surface of the microfluidic side enables an easy attachment of the microfluidic components. Using the polymer surface for the microfluidic channels saves expensive silicon space. Using through mold / PCB vias and not using silicon vias takes also less space of the expensive silicon.

Being insensitive against mechanical pressure the selected packaging strategy for the BioGrFET chips is the Fan-Out Wafer Level Packaging (FOWLP) process including the compression molding with high pressure treatment [2]. It enables a planar package surface for the silicon-saving and easy attachment of the microfluidic. Avoiding impurities on the sensor surfaces and the bombardment with high energetic particles during processing the Redistribution Layer (RDL), protection layers for the sensor surfaces are needed.

FOWLP of the BioGrFET chips compromises five steps (Figure 6). The first step is the creation of the reconfigured polymer wafer. It consists of assembling the silicon sensor chips on a carrier covered with a thermal release film and creating a reconfigured mold wafer using compression molding technology. The compression molding is done at temperatures of 150 °C for 1 h. Thermal release is done at 175 °C for 3 min. Electrical characterization of functional BioGrFET chips before and after the compression molding showed no significant degradation of sensing structures.

The second step is the creation of the redistribution layer (RDL). Holes are laser-drilled into the mold for electrical routing the non-active back side of the wafer. Several steps, comprising vacuum resist lamination and photolithographic structuring of the layers, sputtering of Ti:W and copper and electro-plating and etching copper, follow.

The third step is the creation of the microfluidic. A planar surface of the mold wafer enables the vacuum lamination of the photoresist. Channels and openings for the sensor devices are created photolithographically. Afterwards, the chips are laser-cut into single packages (Figure 7).

The first three steps involve methods that could attack the graphene or the CNM with amino groups of the sensor surface.

A PMMA layer turned out to be a nice protection layer. It protects the sensor surface against contamination from the thermal release film used for compression molding and avoids contaminations from different solvents during the RDL process and the application of the microfluidic. It is processed by spin coating before the compression molding process and is removed with acetone right before the multiplexing. Nevertheless there is still the bombardment with the high energetic particles from sputtering. Tests with aptamers immobilized on glass showed that the thin PMMA layer (around 500 nm thick) did not protect against the high energetic particles from Ti:W- and Cu-sputtering. Fluorescence markers usually docked to the immobilized aptamers could not be captured on the surface after sputtering. An additional poly vinyl alcohol [PVA] layer leads to an adequate protection. After sputtering on aptamers immobilized on glass with the additional PVA protection layer, capture aptamers are docked to the surface and were detected by fluorescence microscope. PVA can easily be removed with DI-water after the RDL processing.

A PVA global protection layer cured on the surface of a BioGrFET chip led to a delamination of the graphene due to shrinkage of the PVA layer. A protection with local PVA layers on each device showed no delamination after removing the PVA (see Figure 8 and Figure 9).

The fourth step is the application of different aptamers on the devices called multiplexing. It is done by spotting an aptamer solution on the devices with a nanoplotter. The aptamers can be incubated overnight in wet conditions. The late addition of the biological capture molecule in step 4 is due to their low temperature stability (< 100 °C). Thereafter, there is no thermal load on the disposables.

Afterwards the open channels are covered (5.). The luer connector is attached to the package using a laser-cut double-sided adhesive tape, placed with a standard pick and place tool.

The charge carrier mobility of graphene in BioGrFET devices of 2000 - 3000 cm²/Vs retains after the packaging process, which can clearly be deduced from the unchanged slope of the I_ds(V_g) transfer curves (Figure 10). Moreover, the transfer data demonstrate a lower p-doping level in graphene after the packaging, as the minima of I_ds(V_g) are shifted to lower gate voltages in comparison to the unpackaged BioGrFET devices. Such a change can be even favorable for the sensing applications.

As the BioMEMS chips with their thin membranes are considered to be mechanically too sensitive for the pressures occurring during compression molding, the FOWLP process is not the first option for packaging the BioMEMS. So the process chosen is the chip integration into a PCB with an adapted cavity. Using the so called Chip in Board Packaging (CIBP) process, even a planar surface for the silicon-saving and easy attachment of the microfluidic is possible. CIBP of the BioMEMS chips comprises of 4 steps (Figure 11). The first step is the integration of the BioMEMS chip inside a PCB with openings for the BioMEMS chips. The PCB is attached to a thermal release film on a flat carrier. The BioMEMS chips are placed into the openings of the PCB with a standard pick and place tool. The vacuum tip of the pick and place tool was developed in such a way, that it does not touch the membranes of the chip when handling it.

The gap between the PCB and a BioMEMS chip is filled with the adhesive 2035SC of Loctite Ablestik with an industrial dispensing system. The adhesive has a low tension modulus chosen for compensating the mechanical stress caused by different coefficients of thermal extension of the chips and the PCB (Figure 12). Attaching the BioMEMS chips and the PCB on the flat surface of a carrier before filling the gap with an adhesive makes sure that there is no unevenness of the surface and that the microfluidic can be attached later on (Figure 13). After curing the adhesive at 100 °C for 15 min, the carrier is released by heating up the PCB.

The adhesion and the release of the PCB and the chips were tested on different thermal release films. Best adhesion and a residue free release was found with the thermal release tape 3195V of Nitto. The release needs a temperature of 175 °C for 3 min.

The second step is the installation of the electrical connections between the BioMEMS chips and the PCB creating gold wire bonds using an industrial wire bonder. The PCB has an ENEPIG surface optimized for gold wire bonding. A microelectronic package with successfully created wire bonds is depicted in Figure 14. Figure 15 depicts the wire bonds in detail. Afterwards (3.) the chip is multiplexed by spotting and incubating different aptamers on the sensor surfaces.

The last step (4.) is the forming of the microfluidic channel with a laser-cut double-sided adhesive. The luer connector (microfluidic interface) is attached to the microelectronic package with the laser-cut double-sided adhesive. A 3D-printed protection cap for the wire bonds is bonded to the electrical side considering the feasibility of electrical readout (Figure 16).

Mechanical stress cannot be completely avoided assembling the hybrid microelectronic-microfluidic package. Curing the adhesive, thermal release and forming wire bonds lead to a thermal load on the disposable and so on the thin membranes. So, investigating the membranes on their intactness and measuring the electrical characteristic after the assembly is very important.

Two possible paths of packaging highly sensitive microelectronic biosensor chips for hybrid microelectronic-microfluidic disposables, considering the different properties of the sensors, were successfully demonstrated.

Chips reasonably insensitive to mechanical stress like the BioGrFET chips can be packaged with the FOWLP technology. Avoiding serial process steps and therefor being presumable cheap the FOWLP technology is preferable.

Chips with mechanical sensitive membranes like the BioMEMS can be packaged with the CIBP technology, avoiding the mechanically stressing molding process. Packaging using adhesive with a low elasticity modulus ensures the compensation of different coefficients of thermal expansion.

Both processes ensure their biocompatibility. Deposition of protective layers on the BioGrFET surface ensures clean surfaces and prevents surface damage by high-energetic particles. Applying the aptamers for multiplexing just before covering the microfluidic channels makes sure that the temperature sensitive aptamers are not destroyed.

Both, using through mold / PCB vias and not wasting silicon space, and, having a planar surface on the microfluidic side of the microelectronic package which enables distributing the microfluidic channels on the polymer material, saves expensive silicon. Both processes are adaptable even for much smaller chips without much space around the sensor devices. This allows a higher chip density for every production step. By using microelectronic production equipment, cost for such devices can be very low at high volumes. For the FOWLP technology, all process steps are scalable to typical wafer sizes (Ø200 mm / 300 mm) and even full PCB panel size (610 × 460 mm²) potentially reducing manufacturing cost due to scaling effects as described in [3].

The PoC-ID project has received funding from the EU Horizon2020 Program under Grant Agreement number 634415. The authors would like to thank Damian Freimund for supporting this study in die assembly, Duy Nguyen for wafer molding, Marcus Voitel for data evaluation and Tang Zian for the schematic representation of BioGrFET.