A System‐Level Feasibility Study of a Lead‐Free Ultrasonically Powered Light Delivery Implant for Optogenetics

Body implants play a crucial role in clinical applications, encompassing data acquisition, diagnosis, and disease treatment. However, challenges in size, power consumption, and biocompatibility, particularly in brain applications requiring small, battery‐free devices for deep areas, hinder their development. Despite potential advances through simplified, single‐purpose devices, such as recording or stimulation, overcoming the power and biocompatibility issues remains a hurdle. Addressing this, the article introduces an ultrasonically powered light delivery implant (LDI) utilizing lead‐free piezoelectric material (Li0.08K0.46Na0.46) NbO3 to harvest energy from external ultrasonic waves. The prototype includes a piezoelectric cube, a chip fabricated in 180 nm CMOS technology, and a microscale light‐emitting diode (μ‐LED) for optogenetics. Achieving an end‐to‐end efficiency of 0.75%, the LDI holds promise for various optogenetic studies, particularly in animal studies targeting specific brain areas for treating Parkinson's disease. The delivered optical power on the μ‐LED surface, at 14.1 mW mm−2, presents applicability to diverse studies involving specific opsins.


Introduction
Nowadays, neurodegenerative movement disorders (NMDs) stand as a significant health concern.As an example, to understand the seriousness of neurodegenerative diseases, Parkinson's disease (PD) is a prevalent and rapidly progressing neurodegenerative movement disorder. [1,2]The global burden of PD has doubled in the past decade and is projected to reach 12 million cases by 2040. [3]The primary approach for treating PD involves medication; however, as the disease advances, the efficacy of these treatments may diminish or result in adverse side effects.[6] DBS stimulation is restricted to a population of neurons, making precise spatiotemporal level stimulation challenging. [7]0] Optogenetics, introduced by Deisseroth et al. in 2006, [11][12][13] elegantly combines genetic engineering and optics to accurately modulate the biological functions of the cell(s) with remarkable temporal and spatial precision. [8,14,15]This is achieved by genetically controlling a population of light-responsive neurons through photosensitive proteins, [16] allowing for precise temporal control on a millisecond scale. [17]However, optogenetics does come with its own challenges, such as genetic modification, invasiveness, and the need to identify target cells with suitable gene expression control.The core of an optogenetic system is a light source for optical stimulation of neurons.Traditional optogenetics employs optical fibers to direct light to the target brain regions [18][19][20][21][22][23][24] New methods such as using microscale injectable needles with integrated microscale inorganic light-emitting diodes (μ-LEDs) have been proposed in literature. [25,26]Nevertheless, using tethered technologies for animal studies hinder the natural mobility of freely moving animals, for example, mice, and may cause micrometer-scale movements between probes and soft tissues during movements, which potentially can damage brain tissues, create artefacts, and degrade long-term signal quality, especially in neural interfaces with multiple channels. [27]Additionally, these devices require a well-designed power management unit/chip.The power supply for the μ-LED, often relies on bulky, short-lived batteries, that call for surgery to replace the battery, which can cause leakage risks. [7,28]herefore, implants require reliable power sources with significantly longer lifetime equipped with wireless energy and data transmission. [29][61][62] Far-field technologies utilize radiation by capturing RF radiation with antennas and converting it into usable power. [48,63]ar-field technologies rely on radiation, capturing RF radiation with antennas and converting it into usable power. [48,63]iniaturization necessitates higher frequencies in the GHz range, but these frequencies are absorbed by moisture and tissues, leading to attenuation and tissue heating.The alignment between transmitter and receiver antennas affects far-field transmission efficiency.However, the limited capture capability of small receiver antennas restricts their use to devices near or beneath the skin, limiting applications like DBS. [44] Near-field wireless power transmission, a non-radiative method that does not emit electromagnetic waves indefinitely in free space, [64][65][66][67] relies on inductive coupling between coils.It operates at lower frequencies, typically below 20 MHz, [68,69] with living tissue being largely transparent to this range.While near-field is a well-established technology used in many electronic devices in recent years, [70][71][72] it does present challenges related to coil alignments and a limited transmission range for biomedical implantable devices. [48,59,66]ptical power transfer, a less-explored possibility, exhibits lower efficiency and is vulnerable to power loss due to misalignment between the receiver and the light source.Despite its drawbacks, tissue safety, it provides a compact solution with the potential for smaller CMOS-based power-receiving devices compared to other wireless power transfer systems. [58,73]ltrasonic Power Transfer (UPT) is a wireless technology that uses ultrasonic waves for power transmission, without suffering from electromagnetic interference.It employs a piezoelectric transducer to convert mechanical energy from ultrasonic waves into electrical power.It offers advantages over RF and inductive technologies, especially in delivering power to greater depths (>5 cm) and in smaller volumes.The millimeter-sized wavelength of ultrasonic waves (%1.5 mm at 1 MHz) [74] allows for higher power efficiency, effective beam forming, and focusing compared to RF and inductive technologies.Additionally, the slow speed of ultrasonic waves enables more directional transducers and receivers at a given frequency.Ultrasonic power transmission features low propagation loss (%1 dB cm À1 MHz À1 ) and high FDA-approved intensity (I SPTA = 7.2 mW mm À2 ), ensuring safe and reliable power delivery through deep tissue (>5 cm). [50,62,75]The accompanying electronics in UPT, operating at lower frequencies than RF, are simpler to implement, [76,77] enabling streamlined packaging and operation in a high-impedance environment.This results in high power recovery efficiency even at negligible power levels, addressing a challenge faced by RF power harvesting systems. [76]arious piezoelectric materials are used for ultrasonic transducers.PZT is extensively studied but contains lead, thus, limiting its use in implantable devices.Consequently, lead-free materials have gained attention.Ferroelectric perovskite structures such as BaTiO 3 (BT), KNbO 3 (KN), and (K 0.5 Na 0.5 ) NbO 3 (KNN) have drawn researchers' interest.Among these, BT exhibits relatively favorable electromechanical properties but has a low Curie temperature (%120 °C), [78,79] restricting its use in implants requiring sterilization at temperatures above 120 °C.(However, alternative low-temperature sterilization techniques, such as plasma sterilization, are documented in the literature.Plasma sterilization, while effective, has limitations, notably the low penetration power of plasma species.This limits its effectiveness when dealing with organic residues from packaging materials (e.g., PDMS and Parylene-C, common packaging materials), complex geometries, or multiple devices to be sterilized in a single chamber. [80]The other low-temperature sterilization method is ethylene oxide sterilization-which has several drawbacks including lengthy cycle time, cost, and potential hazards to humans [81] ).In contrast, KN-based materials, especially KNN, possess a high Curie temperature (420 °C) and promising electromechanical properties (e.g., remnant polarization of 33 μC cm À2 ), making them suitable for high-temperature applications. [82,83]n this article, we introduce an LDI comprising a biocompatible piezoelectric cube using (Li 0.08 K 0.46 Na 0.46 )NbO 3 (LKNN) for ultrasonic energy harvesting, a 180 nm CMOS-based rectifier chip, [84] and a miniaturized blue μ-LED controlled by the chip through commands from an external device called the Ultrasonic Power Source (UPS) (Figure 1).The article's structure is as follows: Section 2 details the energy harvesting approach using LKNN and presents results.The CMOS chip design and characterization results are briefly discussed in the same section.μ-LED fabrication is covered in Section 3. Section 4 outlines the integration approach employed.Results are analyzed in Section 5, and the article concludes in Section 6.

Energy Harvesting Subsystem
The energy harvesting subsystem includes the rectifier circuit and a piezoelectric material used to harvest energy from ultrasonic waves.For the energy harvesting material, we have applied a green chemistry approach using hydrothermal synthesis to prepare the biocompatible piezoelectric material, [85] serving as the ultrasonic energy harvester.The alkali niobates KNbO 3 (KN), and NaNbO 3 (NN) were synthesized separately, using a modified version of the approach used by Goh et al. [86] (details given as follows) and subsequently thoroughly mixed by ball milling with commercially sourced LiNbO 3 (LN).The target composition, Li 0.08 K 0.46 Na 0.46 NbO 3 (LKNN), was chosen based on the previous work by Wang et al. [87] suggesting enhanced piezoelectric properties at around 8% Li substitution.The sintering and compaction of the material were carried out using a modified version of the induction hot pressing procedure used by Souza et al. [88] (details given as follows).The as-synthesized material was then annealed to repopulate the oxygen vacancies formed in the hot pressing procedure.For the CMOS circuit, we need to have a low-power energy-efficient design to convert the AC to DC for driving the load, i.e., μ-LED, which is used for optogenetic purposes.

Synthesis and Compaction of LKNN
KN and NN precursors were prepared using hydrothermal synthesis.Nb 2 O 5 (Sigma-Aldrich, 99.99%, CAS: 1313-96-8) was mixed with an excess of 9 M KOH (for KN, Sigma Aldrich, ≥85%, CAS: 1310-58-3) and 9 M NaOH (for NN, Merck, ≥98%, CAS: 1310-73-2) in a Teflon-lined steel autoclave and reacted at 200 °C for 24 h.The product was washed using mixtures of Ethanol (VWR Chemicals, 96%, 34-17-5) and demineralized water, starting with pure demineralized water and ending with pure ethanol, once the pH was neutral and dried for 24 h at 60 °C.Powder X-ray diffraction confirmed the identity and purity of the product (Figure S1, Supporting Information).The KN and NN precursors were then mixed with commercial LN (Sigma-Aldrich, 99.9%, CAS: 12031-63-9) in a molar ratio corresponding to K 0.46 Na 0.46 Li 0.08 NbO 3 in a porcelain vial with alumina balls, and ball milled for 1 h at low energy (Fritsch Pulverisette).The resulting LKNN powder mix was reacted and compacted in a 0.5 in graphite die using a home-built induction furnace at 40 MPa and 950 °C for 2 h with a heating rate of 97.5 °C min À1 .The density of the compacted samples was determined geometrically and found to be above 97% of the theoretical density as determined from the refined X-ray diffraction data.The compacted pellet was annealed in the air for 72 h at 800 °C.

Postprocessing of LKNN
Prior to poling, the sample surfaces were covered with Ag-paste (Ted Pella Inc., Pelco Colloidal Silver 16031) for electrical contact and cured in the oven for 10 min at 125 °C.The sample was poled in the air using a DC electric field of 1.2 kV mm À1 (Heinzinger PNC 6000-300 pos) at 130 °C, with the field maintained through slow cooling (%0.5 °C min À1 ) to room temperature.

Characterization of the LKNN
The sample was characterized using Synchrotron Powder X-ray Diffraction (SPXRD).Room temperature data was collected from powder obtained from the poled pellet of LKNN at the BL44B2 beamline at the SPring-8, Japan, and modelled using the Rietveld refinement procedure (details in the Supporting Information Section 2).SPXRD confirms the formation of the LKNN phase with a slight impurity of unreacted LN of 2 wt% (Figure 2).In the initial studies related to this work, several types of minor impurity phases were observed as resulting from the described synthesis and processing.Slight variations in relative powder diffraction peak intensities have also been observed, owing to the strain generated in the material.The synthesis delicately depends on purity of precursors and synthesis vessels, pressure and temperature and so upscaling to industrial quantities will require further studies.It should be noted that the superior data quality of the synchrotron measurements allows very sensitive detection of impurities that would not be identified using normal laboratory XRD measurements.The relative amounts of the alkaline elements in the sample were determined using Induction Coupled Plasma Optical Emission Spectroscopy (ICP-OES).This corresponded to a stoichiometry of K 0.44 Na 0.42 Li 0.07 NbO 3 (Table S2, Supporting Information).ICP further revealed trace elements (%0.45 mol%) of Al.

Power Management Circuit
The power management circuit sits between the crystal that provides AC power, and the μ-LED load that receives the rectified DC-power output.Historically AC-to-DC conversion has been done with diode-based rectifier bridges; however, they suffer from extensive power loss in the diodes.Thus, a different approach where a rectifier based on integrated CMOS circuitry replaces the diodes is used here instead.The rectifier utilizes passive diode-connected transistors in parallel with active diodes, with adaptable high and low power modes, in order to ensure a safe and fast startup before the internal DC supply rail reaches a threshold level for normal operation.Figure 3 shows the chip micrograph of the rectifier fabricated in 180 nm CMOS technology occupying 300 μm Â 300 μm, with the core representing the circuit implementation area, and the rest is dedicated to the output capacitor (95 pF).The proposed chip operates at frequencies up to 5 MHz and power levels in the 0-10 mW range.It takes 1.8 μs from the first impact of the acoustic wave until the active rectifier starts normal operation at 2.9 MHz and delivering 14.4 mW mm À2 (Figure 4) and achieving 98.1% efficiency @ 2 mW (Figure 5).

Optical Subsystem
Gallium nitride (GaN)-based LEDs are an ideal source for implantable optogenetic applications. [89]Unlike lasers, the devices have no threshold for light emission and the emission wavelength can be selected by choice of active region during wafer growth.It is coincidental that the optimum wavelength for light-gated excitation of the cation-selective membrane channel Channelrhodopsin-2 (ChR) is around 480 nm, [90] where GaN materials have their highest efficiency.The operating voltage for LEDs is determined by the emission wavelength; for example, it will be %2.6 V for blue devices.LEDs maintain many of their performance properties as the size of the device is decreased to 100 microns and belowthe devices are termed μ-LEDs.It is essential to engineer the devices for the application, and here, the design of the LED involves several key parameters: 1) Being capable of generating sufficient power to activate the target opsin(s) (Table S3, Supporting Information); 2) Having an operating voltage below 3.3 V, the supply limit of the chip; 3) Highly efficient while operating at a low forward bias (%1 mA) and finally; and 4) The physical dimension is matched with the electronic die so that it can be integrated directly on the electronic chip.Here, the electronic die has a footprint of 300 Â 300 μm 2 with two 50 Â 100 μm 2 Al-contact pads for the μ-LED.The layout of the electronic chip is presented in Figure 6A.Substrate-emitting circular μ-LEDs with emitting diameter of 100 μm were designed with a chip footprint of 130 Â 300 μm 2 having p and n contacts of size 100 Â 72.5 μm 2 (Figure 6B).Smaller sized devices resulted in higher resistivity reducing the wall-plug efficiency (WPE).Larger sized devices are restricted by the size of the electronic chip.

Fabrication of the μ-LED for Integration with the LDI
An InGaN multiple quantum well (MQW)-based wafer structure emitting at 480 nm, grown on patterned sapphire substrate (PSS) was used for the device fabrication to enhance out-scattering of light.The schematic fabrication process flow is shown in Figure 7. Pd/Ni/Au metal was deposited as the p-contact followed by mesa definition where the photoresist was re-flown at an elevated temperature (150 °C for 3 min) to form a sloped profile which was transferred into the material using Cl 2 assisted etch.Ti/Al/Ti/Au was deposited as n-contact.300 nm of SiO 2 was deposited to passivate the mesas and small sections were opened by C 4 F 6 assisted dry etch.Ti/Au was deposited as bond pads.The wafer was thinned down to 110 μm and an anti-reflection coating was deposited on the back surface of the PSS to enhance the light output and finally, the chips were diced.

Efficiency of the μ-LED
The efficiency of the μ-LEDs plays a key role in the performance of the system.The efficiencies of the μ-LEDs were measured in terms of external quantum efficiency (EQE) and WPE, where EQE refers .LDI prototype start-up time with an acoustic intensity of 14.4 mW mm À2 at a frequency of 2.9 MHz (cf.also ref. [100]).
Figure 5.The chip power conversion efficiency and power consumption versus frequency (cf.also ref. [100]).
to the ratio of the number of photons coming out of the μ-LED chip with the number of injected electrons and WPE refers to the optical power out for input electrical power.
Figure 8 represents the current-voltage (IV) characteristics and measured power of the μ-LEDs operating in DC mode.A turn-on voltage of 2.3 V was measured while operating at a current of less than 20 μA.At 1 mA, an optical power of 0.11 mW was measured at a numerical aperture (NA) of 0.5 (collection of light at a solid angle of 30°), whereas 0.71 mW (before anti-reflection coating (ARC)) when measured at an NA of 1 (collection of all the light only from the bottom surface of the chip).Theoretical calculations showed that a power of 7 times more can be collected in an NA of 1 than in NA of 0.5 which is in good agreement with experimental data.The measurements show that about 10% power improvement (0.77 mW) has been achieved by the ARC coating under the same operating conditions.Using an integrating sphere, the total optical power of 1 mW was measured at 1 mA, 2.5 V.This gives a power density of 16 mW mm À2 from the surface of the LED into air, which is increased into tissue and meets the requirement for activation of most of the opsins (Table S3, Supporting Information).The highest EQE was measured at 0.4 mA where a very high EQE of 31% was obtained into a NA of 1% and 39% when measured with the integrating sphere.An extremely high WPE of 41% was achieved in this case.As the current increases, different non-radiative recombination mechanisms start to dominate, causing an efficiency drop of about 2% at 1 mA.Further improvement on the efficiency of about 8% can be expected by encapsulating the μ-LED in an index matching material.

A B
Figure 6.A) The micrograph of the electronic chip (cf.also ref. [84]) and B) the layout of the compatible μ-LED (cf.also ref. [104]).
A B
A 3 mm Â 1.5 mm sub-mount was designed and fabricated on a glass wafer by vapor deposition of gold pads to connect the individual subsystems.The μ-LED pads are 500 μm Â 500 μm and the LKNN crystal pad is 1 mm 2 .An open area was dedicated to the rectifier chip.Two gold studs were put down on the μ-LED pads and 50 μm solder bumps were placed on the studs.The bumps were then coined.Then the μ-LED was bonded to the studs of the μ-LED pads using flip-chip bonding technology and thermal compression at 270 °C for 20 s.To hold the μ-LED firmly on the glass substrate, non-conductive epoxy was applied to the μ-LED.The crystal was attached with conductive silver epoxy.To prevent the crystal from being damaged by heat, a room-temperature curable epoxy (curing time: 8 h) was used.The rectifier chip was then attached to the glass submount with a non-conductive epoxy.Since the rectifier chip does not have electrostatic discharge (ESD) protection, precautions were taken while taking care of the rectifier chip.Finally, the connections were realized by gold wire bonding.The device was encapsulated with a PDMS layer to protect it from environmental influences and increase its mechanical robustness (Figure 10).Furthermore, this layer serves as an impedance matching layer to reduce the reflection of the ultrasonic waves at the interface between the piezo and the surrounding fluid, i.e., the water.First, a PDMS layer was spun onto a cleaned silicon wafer at 500 rpm for 30 s, followed by baking in an oven at 80 °C for 15 min.Next, the second layer was deposited on top of the first layer at 2000 rpm for 30 s.The integrated system using a tweezer, and then the sample was placed in the oven at 80 °C for 15 min.Finally, the last layer was spun coated at 500 rpm for 30 s and baked at 80 °C for 15 min.The finished device was obtained by carefully peeling the system off the substrate.

Results and Discussion
To determine the viability of the LKNN crystal as an energy harvester in a wireless energy transfer system, an experiment is performed to gauge the amount of power one could expect from this depending on the incident acoustic intensity.The experiment is performed in a tank filled with demineralized water (Figure 11a).An industrial transducer (V303-SU) from Olympus driven by a signal generator (Agilent Technologies 81160A Pulse Function Arbitrary Generator) through a 50 dB gain power amplifier (Electronics & Innovation RF Power Amplifier 50 dB) acts as the ultrasonic power source.
An HGL-0400 Hydrophone with an AH-2010 Pre-Amplifier from Onda Corporation connected to an oscilloscope (Rhode & Schwarz RTO 1044 Oscilloscope) is used to calibrate the ultrasonic intensity to a desired level before experimentation.The LKNN crystal is then subjected to the calibrated ultrasonic power, and electrical energy is then harvested and rectified to a DC with a diode bridge with a 1 nF storage capacitor to reduce ripple noise.The electrical output of the bridge (V þ with respect to V À ) is then connected to a Keithley 2450 Source Meter, which first determines the open circuit voltage (V OC ), and then sweeps the voltage from 0 V up to V OC in 50 mV steps while measuring the current along the way.A functional block diagram of the experimental setup can be seen in Figure 11b.The LKNN crystal dimensions are approximately (0.8 Â 0.8 Â 0.8) mm 3 , and the optimum frequency was empirically found to be approximately 2.73 MHz for this crystal.From the measured data, the harvested power is calculated versus the voltage which can be seen in Figure 12.
Given the efficiency of different stages in the proposed system, we calculate the total efficiency of the whole LDI.For example, assuming the η UPS and η Chip close to one (98.2%),and η US!Electric from Figure 12, and a WPE of 37%, [84] with the total area of 0.8 mm 2 Â 0.8 mm 2 , we achieve a total efficiency of 0.754% (Equation (1)). .Current-voltage (I-V ) characteristics and optical power measured from the fabricated μ-LED as a function of collection numerical aperture (NA) (cf.also ref. [84]).
Considering an area of 0.01 mm 2 for the blue μ-LED, we can achieve a total optical output power of 14.1 mW mm À2 at the surface of the μ-LED.This is with a duty cycle of 20% to be within the allowed FDA-approved range of 7.2 mW mm À2 .Nevertheless, employing extremely low duty cycles, such as 1%, [91] gives the possibility of increasing the input ultrasonic power, thereby, increasing the output optical power on the μ-LED surface by a factor of at least 20 (in comparison to the duty cycle of 20%), reaching approximately 320 mW mm À2 .This level of intensity would be sufficient for the majority of optogenetic applications (Table S3, Supporting Information).As depicted in Figure 12, the output power of the energy harvester depends on the input ultrasonic power.By adjusting the input ultrasonic power, which is provided by the ultrasonic transducer, it is possible to modify the electric power delivered to the μ-LED, thereby meeting the required light intensity at different depths within the body as long as the FDA limit is followed.
The results achieved in this article show the feasibility of making a more complex system where other blocks such as recording interfaces, processing, and storage, can be added with a small power overhead.Literature shows implementation of recording front-ends with a power consumption of <10 μW/channel proving the feasibility of the development of more complex but biocompatible systems. [92]One can even consider biocompatible closed-loop systems for recording, processing, and stimulation, which can be a breakthrough within the brain and other neural implants.
It is worth noting that the measurements were performed before encapsulation with PDMS.If the device were encapsulated with PDMS, it would restrict access to the contact pads, making it sensible to measure the electrical output before encapsulation.Applying PDMS as an encapsulating layer can have advantages due to its acoustic impedance being between that Figure 9. Schematics for device integration steps.Adapted with permission. [84]Copyright 2020, IEEE.
of water and LKNN.Therefore, in addition to its role as an encapsulating layer, PDMS can also serve as a matching layer, enhancing the electromechanical coupling and reducing reflected power at the LKNN/water interface. [93]This can lead to improved performance of the device.Furthermore, one may have concerns regarding the intensity decay after PDMS encapsulation.
However, it is important to note that the absorption coefficient of PDMS for blue light is generally very low (for a membrane of 1 cm thickness, it is roughly 1% [94] )making it almost transparent. [95]As a result, the optical power generated by the μ-LED is unlikely to be significantly absorbed by the PDMS encapsulation layer.This characteristic allows the generated light to pass through the PDMS and reach the desired target without substantial losses.
The current longevity of medical devices encapsulated by PDMS is insufficient for their use in long-term implant applications.This limitation stems from moisture penetration through the packaging layer.However, some reports have introduced a promising approach to enhance the lifetime of PDMS as a encapsulating material by up to 5.5 years at physiological temperatures. [96]Nevertheless, it is important to note that this proof-of-concept device cannot be used for human studies as a long-lasting encapsulation is necessary.However, in animal studies, the PDMS approach can work as the study is short.As an example, Park et al. developed a stretchable, fully implantable miniaturized optoelectronic system for wireless optogenetics, which was encapsulated with PDMS and tested on a mouse. [97]nother significant concern about this system is the potential heat generation.The main power consuming part that is expected to generate heat in the tissue is the light (associated with μ-LED) absorption of the tissue.According to the paper published by Park et al. [97] targeting the nerves with blue light (470 nm) at an intensity of 10 mW mm À2 , similar to our experiments, infrared imaging of an anaesthetized mouse during device operation revealed that an optical power density of 10 mW mm À2 (40% duty cycle, 20 Hz period, 20 ms pulse width) did not cause detectable temperature changes.The electronics does not consume too much power to generate any heat on (A-F) Adapted with permission. [84]Copyright 2020, IEEE.

A B
Figure 11.A) The water tank test setup representation.Reproduced with permission. [84]Copyright 2020, IEEE.B) Block diagram of the LKNN wireless power transfer experiment.The diode bridge consists of four Schottky diodes (1PS70SB14).
the chip (the chip efficency is 98.2%).Even if it increases, the surrounding is a great heat sink to cool down the chip to 37 °C.Therefore, there is no concern as the electronics consume very low energy.Note that, in comparison to other energy harvesting technologies, such as Thermoelectric Generators (TEGs), which their open circuit voltage and output power heavily depend on the temperature gradient between the cold and hot sides, [98] and theoretically achieve an output power of 180 μW cm À2 with an 8-K temperature difference, [99] the proposed system in this article outperforms them.More specifically, considering the human body as a heat source, the absence of a significant temperature gradient makes this comparison irrelevant.Table 1.Comparison with state-of-the-art similar implants in the literature.Charthad et al. [105] Kim et al. [106] Montgomery et al. [49] Jia et al. [107] Khan et al. [108] This

Conclusion
Within this article, for the first time, we developed an LDI with an efficiency of 0.754% based on the lead-free LKNN material used to harvest energy from ultrasonic waves.As a proof-of-concept, a device was built which included an integrated chip used to rectify the received ultrasonic waves from an external transducer to drive an μ-LED used for optogenetics.Compared to our recently published work utilizing PZT as a piezoelectric energy harvester, which achieved an end-to-end efficiency of 11%, [100] LKNN exhibits a comparatively lower efficiency at 0.754%.The proposed LDI in this article is compared with the existing counterparts found in the literature in Table 1.
However, the lead-free LKNN crystal is a significantly crucial factor for brain or body implants powered by piezoelectric components.A future path for this work is to optimize the LKNN to enhance its efficiency, which can be potentially achieved through chemical modifications or post-processing of the material.
All in all, the device shows the feasibility of being used in implants for freely moving animal studies while in vivo studies certainly bring up new challenges and complexities which need to be tackled as well.Given the amount of power available on the device, there is a possibility, also, to add more complex functions such as recording, processing, and communication of the data to an external receiver to build a closed-loop system.
The next possible step from the perspective of miniaturization, would be to have a more compact integration stacking different components on top of each other to reduce the total volume of the device while minimizing the tissue damage.To achieve device miniaturization and realize a full system including other blocks such as recording interfaces, communication, and storage, the size of the integrated circuit (IC) is not the primary bottleneck.The proposed chip in this article was designed and implemented using the 0.18 μm technology node, and further size reduction is possible by utilizing smaller technology nodes, for example, 28 nm CMOS.
Although the measurements were performed under ideal conditions in a water tank, it is important to consider that tissues are inherently inhomogeneous with multiple transitions between the transmitter and receiver.While water and living tissue have roughly similar acoustic impedance, it is crucial to conduct further in vivo experiments to investigate the non-ideal conditions and understand how they might affect the effectiveness of the light delivery system for the specific tissue of interest.
Another potential concern is that the brain would surround the implanted LDI with connective tissue and astrocytes as part of a foreign body response (scarring).This tissue could have several adverse effects, such as reducing the transparency for 480 nm light, increasing the distance between the μ-LED and the target region, introducing light scattering that would decrease the efficiency of stimulation, and finally altering the device orientation relative to the target region and lowering the efficiency of ultrasonic powering or communication.
However numerous papers have indicated the biocompatibility of PDMS and its suitability for body implants, [101,102] the biocompatibility of the encapsulating layer, i.e., PDMS, is a crucial concern that needs to be addressed through supplementary cytotoxicity testing which is beyond the scope of the current work.

Figure 2 .Figure 3 .
Figure 2. Synchrotron powder X-ray diffractogram (XRD) of powder from the biocompatible piezoelectric material confirming the formation of the LKNN phase.The data is modelled using Rietveld refinement, revealing the coexistence of 2 wt% unreacted LN.

Figure 4
Figure 4. LDI prototype start-up time with an acoustic intensity of 14.4 mW mm À2 at a frequency of 2.9 MHz (cf.also ref.[100]).

Figure 8
Figure 8. Current-voltage (I-V ) characteristics and optical power measured from the fabricated μ-LED as a function of collection numerical aperture (NA) (cf.also ref.[84]).

Figure 10 .
Figure 10.Encapsulation process of the LDI and the final device size visualization.A) Spin coating and baking the first PDMS layer.B) Spin coating the 2nd PDMS layer.C) Placing the integrated LDI and baking in the oven.D) Spin coating and baking the last PDMS layer on top of the whole sample.E) Peeling off the encapsulated integrated LDI from the Si wafer.F) Microscopic image of the encapsulated LDI.G) A size comparison by placing the LDI on the fingertip and H) beside a sesame seed.(A-F) Adapted with permission.[84]Copyright 2020, IEEE.

Figure 12 .
Figure 12.Harvested power versus voltage at five different acoustic intensities.