High Density Neural Implants

The exponential advances in microfabrication coupled with our increased understanding of neural diseases enables the design of highly sophisticated novel devices that benefit patients suffering from neurodegenerative diseases. Most of these devices have mainly concentrated on stimulating the remaining healthy layers of the appropriate neural network. This approach works well with the central nervous system, however when dealing with the peripheral nervous system this approach does not work as well and an alternative approach must be taken. For diseases associated with the peripheral nervous system it is not enough to stimulate but also essential to record from the brain to effectively complete the feedback-loop that is absent due to disease or trauma. Alternatively, one could envision that by recording signals from the brain one could enable patients to control their environment and thereby enhancing their quality of life.

This naturally leads to an equally important aspect of neural recording: the possibility to mapping the neural code. The ability to close the feedback loop in a prosthesis that uses processed signals from the brain is essential for its success. Fortuitously from a design point of view the electrical domain is ideal for recording unlike the stimulation domain where additional challenges of preventing faradic reactions occurring at the electrode must be prevented. However, recording poses its own unique challenges and arguably the main difference is one of scale. The state of the art stimulation device, which impressively restores partial hearing, is the cochlear implant, which possesses only 21 electrodes. Neural recording devices, albeit in-vitro or tethered, in the lab on the other hand routinely record from hundreds of neurons. Furthermore, to truly be able to perform effective mapping of neural signals it would be ideal to record from a large population of cells.

This brings unique challenges to the design of the implant, and necessitates the use of microfabrication techniques. To effectively record from the brain the following requirements must be fulfilled:
1. Electrodes must be penetrating
2. Electrode material must cause minimal tissue damage upon insertion and must allow for surgical handling
3. Implant must be electrically active and incorporate an electronics platform for wireless power and data telemetry to enable un-tethered operation
4. Implant must be biocompatible to prevent adverse tissue reactions and inflammation

Each of these tasks is a formidable one and though our team possesses great expertise in VLSI circuit design, especially in implantable electronics, we believe that microfabrication and biocompatibility design of high-density neural recording array are crucial.

Bio-ionic Neural Interface

Designing a synthetic neural interface that is inherently biocompatible, allows for high-density neural integration and consumes very little power has been a goal for many decades. Realizing this vision would have tremendous impact from restoring lost function and building more efficient neural prosthetic devices, to understanding neural systems and enables the design of better machines. Tremendous advances in lithography, materials and computation enabled the silicon revolution. Unfortunately, this level of success has not been enjoyed by neural interfaces. Since, the discovery of the electrical excitable cells by Galvani in the late 1800s, the primary means of stimulating neural tissue is by placing an electrode near it and passing current between two electrodes. This technique has the advantage that it uses the electronic variable that we can manipulate easily. However, it suffers from the fact that electron transport though electrolytes are accompanied by terminal redox reactions. Furthermore, the conductive nature of the electrolyte shunts most of the current away from the target neuron. These factors often results in issues with long-term stability, neural toxicity and leads to vastly lower efficiency.

We have recently established that a a small increase in extracellular potassium ion concentration can effectively stimulate neural tissue. More quantitatively the background concentration of potassium is 5mM and an increase to just 10mM can effectively stimulate neural tissue. This increase is less than 1 log unit increase in concentration that translates to a Nernst potential of 18mV and an energy dissipation of approximately 600pW assuming a 1pL volume dispensed in 30 msecs. This is 5 orders of magnitude more efficient than an electrical interface, assuming 1µA from a 10V supply. The other key advantage of using an ionic interface is that the potassium ions can be effectively sequestered from the extracellular fluid unlike other chemical interfaces that have been proposed.

DNA Nanopore Sequencing

Genes, proteins, and other biomolecules provide a wealth of information that is essential for the diagnosis and treatment of diseases. One of the key bottlenecks to the widespread use of personalized medicine is the lack of a low-cost, high-throughput, accurate, and easy-to-use biomolecule detection platform. The realization of such a detection platform will also aid in the early detection of diseases, which in cases such as cancer will greatly improve the survival rate.

Solid-state nanopores have emerged as a new tool for electrically sensing individual Deoxyribonucleic acid (DNA) molecules. A nanopore is a nano-scale hole in a thin insulating membrane isolating two reservoirs filled with ionic solution. Since DNA molecule is negatively charged in solution, it will be pulled through the pore at high speed along with ionic solution when a voltage gradient is applied across the reservoirs. The phenomenon is known as translocation. Because the diameter of the nanopore (~10nm) is comparable to that of DNA (~2nm), a measurable fraction of the current will be blocked (current blockade) by the insertion of a single molecule until it completely traverses the nanopore. The information of DNA strand length and relative physical sizes therefore can be extracted from the duration and magnitude of their current blockade signals.

The focus of our research in nanopore sensor involves the design of a novel solid-state platform that allows rapid characterization of DNA. The main contribution is to develop on-chip electronics that can be co-integrated with the solid-state nanopore. Co-integration allows us to reduce the parasitics, develop low cost, high bandwidth and low noise biosensing interface. Furthermore, it enables the parallel processing of DNA molecules. This technology provides a faster and cheaper solution for DNA sequencing. Additionally this engineering approach will transform the subject of highly specialized research into a practical diagnostic technology, improving DNA-based medical resource.

Photonic Switching for Data Center Applications

Switching fabrics in data centers that rely on traditional electrical switches face scaling issues in terms of power consumption. Fast optical switches based on a silicon photonics platform can enable the high port speed and high interconnection density needed while still maintaining a small footprint and low power consumption.

The interconnect fabric is taking an ever-more dominant portion of the power budget and optical interconnects are already used today within data centers for rack to rack interconnection to overcome the limits of electrical signaling. However, the switches interconnecting these racks have inadequate capacity to continue to scale with complementary-metal-oxide semiconductor (CMOS)-based technology due to fundamental limitation of on-chip interconnects and power consumption and pin count of scheduler application-specific integrated circuits. Further, not only does the switch fabric require significant power, but the conversion from the optical signal to electrical and back again (OEO) adds a significant amount of power and space. We present here a unique switching technology that allows for high radix and high bandwidth, which is not achievable in conventional electrical interconnects, while dissipating very low power. We achieve these diametrically opposing goals by utilizing the large-bandwidth enabled by optics fabricated in a cost-efficient CMOS platform. Furthermore, we integrate conventional CMOS electronics within the optical platform using a chips last integration process. (Picture provided by Martijn Heck)

Heterogeneous Electronic/Photonic Circuits

There is no question that photonic systems require a diverse set of materials. Examination of the components in a typical telecom system (the most advanced example) will reveal a diversity of materials (semiconductors, glass, crystals, and polymers) even in subsystems that have components in close enough proximity that they could be integrated. Even in purely semiconductor Photonic Integrated Circuits (PICs), a number of different materials are required, with the most advanced InP PICs containing 4-5 different optical materials on the same substrate . The fundamental limits of integration with this approach are the limited wafer sizes of InP, the need for lattice matching, and the lack of a pervasive, low cost infrastructure for making these PICs.

The scaling laws that have driven the electronics industry forward do not offer the same return on investment for photonic devices. However, the close integration of photonics with electronics can lead to a new class of systems that offer superior performance to conventional macroscale integration at the board level. This disparity in scaling benefits calls for a different approach to integration than the conventional front-end of the line (FEOL) integration pursued currently by the majority of silicon electronic/photonic integration approaches. In our approach we exploit the diversity of silicon processes for the best electronics and utilize a systems in package interconnect approach developed at UCSB to enable close integration of electronics and photonics. (Picture provided by Martijn Heck)

The basic approach is as follows: a wafer containing the photonic integrated circuit (PIC) is designed to allow for the creation of a slot that will house the CMOS chip. A chip specific hole is placed in the PIC wafer via a self-aligned masking method that we have pioneered and the chip is subsequently bonded to the PIC using a spin-on-glass and planarized. The planarized die can now be electrically connected via conventional lithography and metallization allowing for high-density integration with parasitics similar to VLSI interconnects. We have successfully shown that connections > 100 wires with pitch of 50µm can be easily made with this method. This method will be extended to smaller dimensions, limited by the minimum passivation opening of the bond pads allowed by the foundry, if the same pad is to be used for electrical testing. Alternatively, custom etching can allow for opening of 5µm on a side and much higher density of wires. This process can also be extended to multichip and multilayer metallization for greater flexibility.

Tailored Triblock Copolymers for Biomedical Applications

Self-assembly on the nanoscale is a key property that nature relies on to generate biological membranes. These membranes provide a structural framework via a microenvironment and a functional framework by incorporation of channels, receptors, and pumps. By exploiting hydrophobic–hydrophilic interactions, these membranes assemble into bilayers and other structures. If we hope to mimic the principles of natural nanoarchitectures, the selfassembling membrane is a critical component. Recently, a plethora of polymeric systems has successfully exploited self-assembly for the purposes of drug delivery, biomedical coatings, virus-assisted gene delivery, and nanoreactors. These are amphiphilic diblock or triblock copolymers that self-assemble into micelles, worm-like micelles, tubular structures, membranes, or vesicles in a suitable solvent.

Recently, click chemistry has found considerable use in polymer–polymer conjugation via the copper-catalyzed azide–alkyne cycloaddtion (CuAAC) reaction. Click-based conjugation allows for the modular synthesis of block copolymer architectures with well-defined block lengths and endgroups. This allows for systematic investigation of the effect of a single parameter variation on the self-assembling properties of the macromolecular system. For example, by using individually well-characterized hydrophobic and hydrophilic blocks, the effect of the hydrophilic block length on the self assembling properties could be studied. This exquisite level of control allows for sophisticated scientific exploration and will enable the understanding of the underlying structure–function relationships that influence the physics of self assembly in these macromolecules.