Polymeric Packaging for Fully Implantable Wireless Neural Microsensors

A. Cortical Sensor: The “Front-End”

As previously mentioned, a crucial feature for the success of these technologies has been the detection of high quality neural signals with rich temporal and spatial information. These types of signals are sensed in our devices using microfabricated silicon MEAs (Blackrock Microsystems). These arrays are integrated with our custom designed Application Specific Integrated Circuit (ASIC) amplifier microchips to create the sensor unit “front-end”. These high performance low-power and low-noise microchips require only five interfacing signals to the back-end (embedded within the common Kapton substrate): two power supply connections (VDD and GND), one sample clock, one synchronization and acknowledge signal, and one multiplexed analog output.

Each of the preamplifiers in the ASIC use a capacitive-feedback, folded cascode operational transconductance amplifier (OTA) configuration with a source follower output buffer. The details of the preamplifier circuit are discussed in [5]; it is important to underscore that our systems allow broadband data recording and telemetry on all 100 channels of low frequency field potentials (FPs) as well as single cell spiking activity.

B. Power and Telemetry: The Epicranial “Back-End”

The epicranial subunit consists of five major blocks: the LC tank for RF inductive power, a diode rectifier, two SAR-ADCs, a digital control ASIC, and a Vertical Cavity Surface Emitting Laser (VCSEL). The system is inductively powered at 13.56 MHz RF power carrier frequency. The digital control ASIC produces the clock and chip select signals for the ADCs from the power carrier, and generates the sampling clock for the preamplifier ASIC in the “front-end” sensor unit. The digitized neural data from the two ADCs are packaged and combined into a single serial bit stream to modulate a current source that drives the infrared laser (VCSEL). The VCSEL transmits the optical signal through the skin to an external infrared receiver, where the original neural data can be recovered [6].

A. Fabrication of a 100-Channel Small Footprint Implantable Neural Interface

The 10×10 (100 channel) intracortical microelectrode arrays are purchased from Blackrock Microsystems with electrode heights of 1.5 mm and Pt/Ir tip metallization. These arrays are flip-chip bonded to the previously described ASIC amplifier using our developed process described in [4]. The passive and active device components are populated on the “back-end” and “front-end” of the substrate using silver epoxy. The controller and amplifier ASICs are wirebonded to the substrate using an ultrasonic wedge bonder, and finally a 5mil thick Pt/Ir wire is attached to a pad on the “back-end” to serve as common voltage reference input to all the preamplifiers.

The active neurosensors are bench top tested after fabrication and before their final encapsulation. The 100 channel system has a 45.6 dB gain, 1 Hz – 7.8 kHz bandwidth, 8.6 μV_rms rms input referred noise, and consumes only 52 μW of power. The power dissipation of each block is shown in Table I. Finally, a 100 Hz, 750 μV_pp artificial spike function (Figure 2) is used to test the device before silicone encapsulation.

B. Electronic Packaging Methods

Packaging the system to ensure good isolation from the saline rich body environment (and vice versa) is the last step in completing the full fabrication of the device. For our systems, we developed a hybrid encapsulation method using a Hysol epoxy (M-31 CL) followed by complete system encapsulation with polydimethylsiloxane (PDMS, Nusil R-2188). The Hysol epoxy mechanically secures all components and wirebonds, while the PDMS provides a soft, biocompatible, and flexible protective barrier. Although we have successfully packaged devices for long term in-vivo applications, we emphasize that this process is highly sensitive to deposition conditions, adhesion, and surface treatment before application of the insulator, e.g. cleaning procedures. We have summarized in Table II the results from various device implantations with this encapsulation scheme into swine (Yorkshire; practical time limits due to animal growth apply) and non-human primates (rhesus macaques).

We consider this initial encapsulation approach with PDMS a useful starting point with viable applications for short-term neural recording demonstration purposes (at least ~months). However, to create long-term, fully implantable devices more reliable packaging schemes and materials are necessary. To achieve significant statistical certainty on the quality of materials and packaging processes we have created a system capable of testing the encapsulation reliability of polymeric insulators. This system, shown in Figure 3, is capable of accelerated characterization of 100 test units simultaneously during periods of extended soaking in hot saline solution at variable temperatures.

The test units for this system are uniquely designed to simulate the topographical, thermal, mechanical and electrical stresses, as well as the fabrication procedures, materials and components that form the previously described fully implantable active neurosensors. With the test units we are capable of monitoring continuously the resistance between interdigitated conductors on the substrate surface as well as leakage current through the encapsulation material. A picture of these test units is shown in Figure 4.

These units are currently being employed to test multiple types of polymeric encapsulation materials, processes and configurations. It is our objective to build a trustworthy database of material properties to share with the scientific community. Work with PDMS (but also Polyimide, Parylene, LCP, and SU8 among others) is underway, as shown in Figure 5 for three of the many PDMS encapsulated test units (here showing development of ionic leakage over period of months, and unacceptable cases). However, we are also interested in exploring other types of materials with augmenting potential for truly chronic implantable biomedical devices. Our approach is to include material strategies based on the combinations of organic and inorganic thin film multilayers. Here, we outline the adaptation to our implants of a recently developed packaging material based on a nanocomposite of PDMS and SiO₂ [7]. The thin conformal films are produced with a Plasma Enhanced Chemical Vapor Deposition (PECVD) system. This system deposits layers of material from a HMDSO and O₂ plasma mixture. The deposited materials have been reported to provide a reliable hermetic permeation barrier for highly water sensitive Organic Light Emitting Diodes (OLEDs) [7].

The films deposited by this process are a heterogeneous micro/nanoscale mixture composed of SiO₂ and silicone, where networked SiO₂ is viewed to play a role for successfully blocking diffusion paths within the silicone. It has been reported through infrared absorption techniques, wetting angle measurements, and indentation hardness experiments that these films are uniform on the macroscale. Hence, this supports the theory of a SiO₂ filled silicone hybrid film. In addition to the permeation barrier properties, the films are also highly flexible. They have been reported to remain intact with over 58,600 cycles of bending [3].

In our lab, we have successfully developed our own custom in-house PECVD system for device encapsulation. Preliminary infrared absorption photospectroscopy tests support the reported results of the hybrid SiO₂/silicon film. For example, monitoring the IR absorption by the O-Si-O stretch resonance with increase in O₂ flow with respect to HMDSO is one of the methods to define optimal process parameters as shown in Figure 6. Future work will investigate these thin (~100nm) films using AFM, hardness, and wet etching, as well as results from applying these coatings on implantable active neural sensors and test structures for accelerated lifetime verification and material analysis.