Fabrication and Characterization of a Thermoelectric Generator with High Aspect Ratio Thermolegs for Electrically Active Implants

Implantable thermoelectric generators hold significant promise as an alternative or supplementary energy source for implantable medical devices, but their performance is constrained by spatial limitations within implants and relies heavily on design optimization. In this study, an optimized thermoelectric generator featuring high aspect ratio thermolegs (0.5 × 0.5 × 4 mm3) is presented, which achieves a remarkable power output of 110 µW with a simulated temperature difference (3 K) in an implantation scenario. This achievement is realized through the use of a removable assembly rig during the soldering process, streamlining the fabrication process. Furthermore, the integration of a boost converter within the thermoelectric generator results in a voltage output of 2.5 V, addressing the voltage requirements of active implants. These findings highlight the potential that implantable thermoelectric generators can be used as a reliable, quiet and self‐sufficient power source for implantable medical devices.


Introduction
Implantable medical devices (IMDs) have made enormous achievements in saving lives and curing diseases since the DOI: 10.1002/admt.202301157introduction of the first cardiac pacemaker in 1958.Today, IMDs are more widely used in the treatment of disease, medical diagnosis, and prognosis.Their demands will be unceasingly stimulated to overcome the challenges caused by the growing aging population.Despite advances in battery technology, batteries within IMDs have significant limitations in terms of energy capacity, size, and lifetime.Only half of the lithium-powered pacemakers can survive for more than eight years. [1]As a result, implantable thermoelectric generators (TEGs) have been proposed as an alternative or supplemental source of power, which can provide electrical power by harnessing the human body's temperature gradients. [2]The use of TEGs in medical applications offers several advantages over traditional power sources, including maintenance-free, long lifespan, and cost-effectiveness.This eliminates the need for surgical intervention to replace batteries or rewire the device, reducing the risk of infection and improving patient outcomes.
In light of the burgeoning Internet of Things (IoT), there has been a surging drive to explore wearable TEGs with the primary objective of energizing intelligent wearable electronic devices.The vast potential for the application of wearable TEGs has been substantiated in various domains, including healthcare monitoring, [3] motion recognition, [4] and fire warning systems. [5]The field has witnessed remarkable advancements, propelled by material innovations, [4][5][6] structural enhancements, [7][8][9] and thermal optimization strategies. [3,10]onetheless, it is imperative to underscore that implantable TEGs, which hold significant promise for applications within the human body, have not garnered a commensurate level of attention when compared to their wearable counterparts.Persistent doubts linger regarding the ability of implantable TEGs to generate the requisite levels of power and voltage output for IMDs.
The performance of thermoelectric generators (TEGs) relies heavily on the figure-of-merit (ZT) of the thermoelectric (TE) materials used.This metric, expressed as ZT =  2  −1  −1 T, where , , , and T represent the Seebeck coefficient, electrical resistivity, thermal conductivity, and absolute temperature, respectively, plays a pivotal role in determining TEG performance.[14] Unfortunately, conventional TEGs using BiTe material struggle to generate the required power due to thermal mismatch issues with the implantation environment.In a relevant study, [15] an in vivo experiment involving the implantation of a commercially available TEG in a rabbit failed to provide sufficient power and voltage for IMDs.This research underscores the need for optimized TEGs with a significant number of high aspect ratio thermolegs to meet the power and voltage specifications of IMDs. [16]While densely packing micro TEGs has the potential to achieve high packing density and voltage output, [17,18] their limited height, typically within a few hundred micrometers, results in inadequate temperature gradients across the thermolegs when implanted in the human body, leading to low power output.Overcoming this challenge through the use of multiple stacked micro TEGs has proven complex due to manufacturing difficulties and increased heat losses at the interfaces.A previous study [19] attempted to print thin-film thermolegs on a flexible polyimide substrate and wound them to create high aspect ratio devices.However, the introduction of flexible material between the thermolegs increased thermal shunt significantly, reducing the device's performance.
Obtaining free-standing high-aspect-ratio thermolegs through bulk material presents a more feasible approach, rendering them a more desirable option for implantation purposes.In previous works such as those by Leonov et al. [20] and Vasilevskiy et al., [21] TEGs with 0.1 × 0.1 × 1 mm 3 BiTe thermolegs were presented, but the fabrication process was not described.Wesolowski et al. [22] utilized sinker electro-discharge machining to cut semiconductor wafers, which were then held by multiple alumina carrier plates.The TEG with high-aspect-ratio thermolegs (0.15 × 0.3 × 7 mm 3 ) was fabricated by stacking and cementing the alumina carrier plates.This approach eliminates the need for a pick-and-place process for the thermolegs but is still complex due to the numerous cuts required in the alumina plates.Additionally, it is important to note that the TEGs mentioned above were not specifically designed and optimized for electrically active implants.The performance of TEGs with high aspect ratio thermolegs for implantation use is still unclear and requires further investigation.
This study introduces a fabrication approach for implantable TEGs that simplifies the manufacturing process and enhances production efficiency.The method utilizes dual-level removable assembly fixtures to secure the high-aspect-ratio thermolegs during a single soldering process.This enables the fabrication of a TEG comprising a 16 × 16 array of thermolegs (0.5 × 0.5 × 4 mm 3 ) with 508 simultaneous solder connections.To adapt the voltage output of the implanted TEG to meet the requirements of IMDs, we fabricated a compact TEG with 62 thermolegs positioned on a circular ceramic substrate (diameter 30 mm) with an integrated voltage boost converter.This method is well-suited for accommodating the space constraints typically associated with IMDs.The fabricated TEGs were characterized by using the temperature difference across the TEGs in a simulated implantation scenario.

Device Design
The geometry of TEGs, designed for electrically active implants, was obtained based on simulation results using a simplified tissue model that incorporates bio-heat effects. [23]Housing made of low thermally conductive polyetheretherketone (PEEK) was introduced to encapsulate the TEG, with metal plates serving as thermal contacts between the TEG and the tissue.The size of the housing is a critical parameter in terms of performance and patient comfort.A traditional pacemaker typically measures approximately 50 mm by 40 mm, with a thickness of about 6 or 7 mm. [24]As a result, we have determined that a housing with dimensions measuring 65 mm in diameter and 8 mm in height is deemed acceptable within the scope of this study.This size is anticipated to offer around 100 μW when the packaged TEG is implanted within the subcutaneous fat layer.
In our previous study, [16] it was found that the power output depends on the geometry ratio between the height and crosssectional area of the entire thermopile.This ratio determines the thermal resistance of the TEG, and an optimum geometry ratio exists to maximize power output.The number of thermolegs does not affect the power output when the geometry ratio is maintained.However, having more thermolegs results in higher voltage output, while also increasing the aspect ratio (the ratio between height and width) of the thermolegs.Hence, the height of the thermolegs in this study was established at 4 mm.This decision takes into account both the requirement for adequate voltage output and the heightened difficulties associated with manufacturing and assembling thermolegs with higher aspect ratios.
Subsequently, a TEG designated as TEG-1 was selected due to its near-optimum fill factor of 15.5%.The fill factor represents the ratio between the cross-sectional area occupied by the total thermolegs and the cross-sectional area of the TEG itself.TEG-2 and TEG-3 were designed to have the same thermal conductance as TEG-1, but their configurations were modified to achieve different voltage outputs.This was achieved by either changing the number of thermolegs or incorporating a boost converter into the design.

Fabrication Approach
TEG-1 consists of a total of 62 p-type BiSbTe and n-type BiTeSeS thermolegs (1 × 1 × 4 mm 3 ) and Al 2 O 3 ceramic substrates (20 × 20 × 0.38 mm 3 ) with direct bonded copper of 0.1 mm thickness.Solder pasting was conducted through screen printing using a squeegee blade to scrape the solder paste onto copper interconnects of the substrates, as shown in Figure 1a.An assembly rig was developed, consisting of a holder, two spacers, and four alignment fixtures (Figure 1b).This rig was designed to securely hold the thermolegs in place during the soldering process, preventing any unintended movements or misalignments that could compromise the quality of the solder joints.The thermolegs were dispensed and placed into the assembly rig using a vacuum pumpactivated nozzle (Figure 1c).Subsequently, the top substrate with pasted solder was positioned onto the thermolegs, and the assembly rig was subjected to reflow soldering in a reflow oven.The alignment fixtures were removed from the TEG after soldering (Figure 1d).The fabrication process is detailed in the Experimental Section.
To explore the potential of the fabrication approach and aim for higher voltage output, the TEG-2 (Figure 1e) was manufactured with a configuration comprising a 16 × 16 array of thermolegs (0.5 × 0.5 × 4 mm 3 ).This configuration retains the same fill factor as TEG-1 while featuring a fourfold increase in the number of thermolegs.Consequently, TEG-2 was expected to yield the same power output as TEG-1 but with a fourfold increase in voltage output.To ensure the proper positioning and securement of thermolegs, the alignment fixtures were modified accordingly.However, fabricating thermolegs with such dimensions via wirecutting resulted in a low yield.To address this, a grinding procedure was implemented to reduce the cross-section of the BiTe legs from 1 × 1 mm 2 to 0.5 × 0.5 mm 2 .Additional details and information regarding this process can be found in the Experimental Section.
Despite TEG-2 having thermolegs with an aspect ratio of 1:8, the voltage output it achieved was still insufficient for IMDs.Increasing the aspect ratio of the thermolegs further would require a more time-consuming and error-prone assembly process.For instance, to maintain the power output and obtain a 2 V voltage output, it would be necessary to incorporate around 6000 thermolegs with dimensions of 0.1 × 0.1 × 4 mm 3 .However, achieving such a configuration using the semi-manual fabrication approach is not feasible due to the challenges associated with assembly at that scale.To address this limitation, a com-mercially available boost converter was introduced for voltage upconversion.This approach can meet the voltage requirements of IMDs without the need for excessively high aspect ratios or an impractical number of thermolegs.
To accommodate the restricted space requirements of IMDs, the boost converter's circuit layout was designed and routed into a single layer.This layout was then fabricated on the bottom substrate, as depicted in Figure 2a.In this configuration, the bottom substrate serves as the connection point between the thermolegs and the components of the boost converter.On the other hand, the top ceramic substrate is solely responsible for connecting the thermolegs.Importantly, the electronics in the boost converter have a maximum height of 3.7 mm, which is less than the 4 mm height of the thermolegs.Thus, a small air gap is intentionally maintained between the electronic components of the boost converter and the top substrate.The presence of the air gap serves a specific purpose, namely to minimize heat transfer from the boost converter to the surrounding substrate.Excessive heat transfer could negatively impact the temperature gradient across the TEG, potentially reducing its performance.The assembly rig was adapted to accommodate round substrates, providing a suitable platform for the placement and alignment of the thermolegs.Additionally, appropriate assembly fixtures were utilized to securely hold the thermolegs in place during the soldering process (Figure 2b).This ensured that the boost converter could be seamlessly integrated into the TEG assembly, enhancing the overall functionality and assembly efficiency.Using the same fabrication approach, a compact TEG, referred to as TEG-3, was Table 1.Material properties of p-type thermolegs (BiSbTe) and n-type thermolegs (BiTeSeS).

Characterization and Analysis
The material characteristics of the BiTe thermolegs employed in our TEGs are provided in Table 1.Specifically, we conducted measurements for the Seebeck coefficient (), electrical resistivity () and power factor (PF), while the thermal conductivity values () were sourced from ref. [25].
To comprehensively characterize the thermal and thermoelectric properties of the TEGs, we employed a customized experimental setup.This setup comprises a measurement environment, as illustrated in Figure 3a, accompanied by a Peltier controller and a display unit.Within the experimental arrangement, it's important to note that the sensing area of the heat flux sensor is slightly larger than the TEG being measured.Consequently, the measured thermal resistance of the TEG encompasses not only the thermal resistance intrinsic to the TEG itself but also contributions from other associated components.These components encompass the alumina platform on which the TEG is affixed, as well as the air filling the space within and surrounding the TEG, as depicted in Figure 3b.For an accurate determination of the thermal resistance attributed solely to the TEG, it is imperative to undertake appropriate calculations while considering the thermal properties and dimensions of each component.This approach enables us to effectively isolate and quantify the specific thermal resistance associated exclusively with the TEG.To verify the accuracy and reliability of the thermal conductance values obtained, the Mini-PEM system was used for an additional round of measurement.The Mini-PEM system is specifically designed for characterizing devices with cross-sectional areas less than 10 × 10 mm 2 .To adapt our TEGs to the measurement system, two adapters were affixed to the heating and cooling blocks.As depicted in Figure 3c, the heat flow through the TEG was determined by comparing the inlet and outlet temperatures of the liquid stream with a constant velocity.This measurement system operates within a vacuum environment.
In order to evaluate the efficacy and suitability of our fabrication approach, a series of TEGs with thermoleg configurations of 6 × 6, 8 × 8, and 10 × 10 were produced.All of these TEGs were manufactured using thermolegs with identical dimensions, measuring 1 × 1 × 4 mm 3 .This allowed for a direct comparison of the TEGs' performance.The characterization results for the electrical resistance, open-circuit voltage per Kelvin and thermal conductance of the fabricated TEGs are depicted in Figure 4ac.The TEGs were measured under operating conditions with a cold temperature of 25 °C and a hot temperature of 35 °C.The estimated values were directly calculated from the geometry of the thermolegs and material properties in Table 1.The effects of interconnects, solder material, and contact resistance in the TEG were not considered.As a result, the estimated values exhibit a linear increase proportional to the number of thermolegs.To assess the impact of solder joints and copper interconnects on the electrical resistance of the TEGs, a set of devices were systematically fabricated for evaluation.These devices were designed with copper legs instead of BiTe thermolegs and had dimensions of 1 × 1 × 4 mm 3 , arranged in an 8 × 8 array.The TEG-shaped devices exhibit an average electrical resistance of only 0.07 Ω, which is merely 2% of the electrical resistance observed in the measurement of the 8 × 8 TEG.This suggests that the contacts, solder joints, and copper interconnects have a negligible impact on the overall electrical resistance.The notable discrepancy observed in Figure 4a can be attributed to the deflection of current at the interfaces between the thermolegs and interconnects.This deflection causes a portion of the current to deviate from its intended vertical path, resulting in deviations in the measured data.To mitigate this error, employing simulated data from comprehensive finite element models may be more accurate than relying solely on calculations based on geometric parameters.The measured voltage output results exhibit a high level of agreement with the estimated values.The discrepancy observed in Figure 4c may arise from the inherent errors associated with the two measurement systems.The thermal conductance measured using the custom experimental setup primarily accounts for the thermal conductance of the TEG itself and the surrounding air.This is why the measured thermal conductance is higher than the estimated values, which might not consider these factors accurately.However, through calculations that isolated the thermal conductance of the TEG, the measured and estimated results were brought into alignment.On the other hand, the introduction of adapters in the Mini-PEM system resulted in errors  during the measurements.The presence of the adapters in the Mini-PEM system may have caused non-homogeneity in the heat flow, resulting in an uneven distribution of heat with a greater amount passing through the center of the TEG, which was originally the designated sensing area of the Mini-PEM system, as previously illustrated.Consequently, the measured values obtained from the Mini-PEM system are higher than the estimated values.
TEG-3 was characterized in the temperature range of 31 to 35 °C, as shown in Figure 4d.The boost converter necessitates a TEG output voltage of no less than 20 mV and a minimum temperature difference of 1.5 K.In response, the boost converter delivers an open-circuit voltage ranging from 3.3 to 4.2 V, corresponding to temperature differences between 2 and 4 K. Higher temperature differences result in higher input voltages for the boost converter and require smaller matched loads for optimal performance.TEG-1, TEG-2 and TEG-3 were designed to show identical thermal conductance, which is matched to the surrounding tissue environment.Consequently, the simulation results [16] reveal a consistent temperature difference of 3 K across all TEGs, maintaining the cold side at 33 °C and the hot side at 36 °C, while considering an ambient temperature of 21 °C.The properties of these TEGs measured under simulated thermal conditions are presented in Table 2. Due to the increased number of thermolegs, TEG-2 has four times higher output voltage compared to TEG-1.Since the fill factor is not changed, the power delivery to a matched resistive load remains almost constant at 113 μW.The integrated boost converter of TEG-3 provides a voltage of 2.5 V to the matched load.The conversion efficiency of the boost converter decreases the available power output to 53 μW.While the incorporation of the boost converter does entail a reduction in available power, it is a trade-off that proves advantageous.This is because the up-converted voltage output effectively aligns with the requisites of typical IMDs, including pacemakers and neurostimulators.In the foreseeable future, surmounting the fabrication hurdles associated with the production of micro TEGs featuring high aspect ratio thermolegs, thereby increasing the count of thermolegs while maintaining the optimum fill factor, holds the potential to obviate the neces-    sity for the boost converter.This, in turn, could lead to further reductions in the device's overall dimensions.A thermal cycling experiment was conducted on TEG-1 to evaluate its robustness and reliability, where the applied temperature difference was periodically varied between 0 and 10 K, corresponding to a temperature range fluctuating from 25 to 35 °C.The temperature difference was adjusted every five minutes.Once the set point temperature was reached, a dwell time period of 200 s was implemented to stabilize the temperature.The ramp rate for temperature change was set at 0.1 Ks -1 .During the thermal cycling experiment, a matched resistive load was connected to TEG-1.The load voltage, as depicted in Figure 5, was observed to remain relatively constant at 64 mV.This stable load voltage was maintained after 5000 cycles, equivalent to approximately 35 days.

Packaged TEG
The general requirements for active IMDs have been delineated in directives and standards, including 90/385/EEC, [26] EN 45502-1, [27] and ISO-14708. [28]The standards cover various aspects including biocompatibility, long-term stability, safety, and overall performance, highlighting the criticality of packaging TEGs for implantation.In Section 2.1, a disk-shaped housing for the TEG was introduced, as depicted in Figure 6a.The housing with an outer diameter of 65 mm comprises a lid, a side wall, and two metal plates.The total height of the housing measures 8 mm.In order to optimize performance, the lid and side wall were fabricated from biocompatible PEEK, which offers relatively high thermal resistivity compared to conventional packaging materials such as ceramics and metals.Both the lid and side wall are 2 mm thick.The two metal plates, made of gold-coated brass, enable an efficient thermal interface to the surrounding tissue.These metal plates have a diameter of 55 mm and a thickness of 1.5 mm.The interior of the housing is filled with air.
The fabricated and assembled module, depicted in Figure 6b, was constructed according to the aforementioned design.The simulated thermal conductance of the packaged TEG was determined to be 54.6 mWK -1 , while the experimental measurement yielded a slightly lower value of 48.5 mWK -1 .The observed discrepancy can be attributed to mechanical errors in the connections of the housing and the use of epoxy glue, which resulted in the device not achieving complete flatness.Despite attempts to polish the device's surface, noticeable contact resistance is still introduced as a consequence of these factors.
Long-term stability is of paramount importance to ensure the safety of patients with IMDs.A packaged and implanted TEG, which will be exposed to body fluids, shall withstand ambient conditions equivalent to 100% relative humidity.The extracellular space surrounding the TEG is warm and contains electrolytes and proteins, significantly increasing the likelihood of electronic failures. [29]However, the polymer PEEK is not hermetic, which leads to water diffusion.One common solution reducing water diffusion is coating by Parylene C, which is a polymer material widely employed in medical microdevices. [30]It features good biocompatibility, stable chemical properties and excellent permeation resistance.We performed a helium leak test for the housing to evaluate the improvement of Parylene C coating.An unprocessed PEEK housing and a second housing coated with 25 μm of Parylene C were tested.The helium leak rate of the PEEK housing is approximately 120 times higher compared to the Parylene C coated device, suggesting that the Parylene C coating significantly increases the hermeticity, as shown in Figure 6c.
Implanted devices with polymeric coatings face water ingress as their greatest challenge.Utilizing the practically equivalent quasi-steady-state (QSS) model, [31] Dahan et al. [32] introduced a refined analogical model to assess and forecast the internal relative humidity change of housings crafted from various materials and uniform thickness when implanted into the human body or submerged in water.The alteration in relative humidity of a submerged, sealed housing can be calculated as where RH t signifies the relative humidity at time t, RH i denotes the relative humidity at t = 0 (initial relative humidity), and RH a represents the ambient relative humidity (RH a is 100% when the package is submerged in water). is the time constant.
In order to gauge the rate of moisture ingress in our housing, an aging test was performed on an uncoated device, a de-vice with a Parylene C coating and a reference device fabricated from brass.The housing with integrated humidity sensor was submerged in the water with a temperature of 53 °C.The test results are depicted in Figure 6d.It becomes evident that the QSS model does not align well with the experimental data for the Parylene C coated housing.This discrepancy might be attributed to potential epoxy degradation or minor leakages during the test, leading to accelerated moisture penetration in the latter half of the experiment.Consequently, the actual time constant for the Parylene C coated housing should be larger than the one derived from our experimental data.By applying the ten-degree rule, an approximate conversion of the Arrhenius reaction rate function, as described in ref. [33], allows us to convert the measured time constant of the devices at 53 °C to the equivalent time constant at 37 °C, as presented in Table 3.
The results indicate that the Parylene C coating increases the time constant by 71.9%, translating to a significant enhancement in the device's longevity.Nevertheless, this improvement falls short of the desired standard.In order to ensure the longevity of our equipment for a period exceeding ten years, with relative humidity not surpassing 85% over that duration, starting from an initial relative humidity of 10%, it is imperative that the time constant exceeds 48000 h, as calculated using Equation (1).It is worth noting that even with the housing constructed from brass, this target remains considerably distant.The notable and rapid escalation of relative humidity, even within the brass housing, strongly suggests that moisture ingress predominantly occurs through the epoxy connections.The incorporation of a PEEK housing and metal plates indeed leads to a substantial improvement in TEG performance; however, it also introduces a significant challenge concerning moisture ingress.Notably, the thermal conductivity of air with varying humidity levels exhibits only negligible deviations around human body temperature. [34]This implies that moisture ingress is unlikely to exert a significant impact on the power generated by the TEG.To mitigate potential electronic failures stemming from moisture infiltration, we propose the application of Parylene C coating on electronic components.Experimental findings have substantiated the efficacy of Parylene in safeguarding silicon chips, with a demonstrated longevity of Parylene-coated metal at 37 °C exceeding 60 years, as reported in ref. [35].However, it is crucial to remain vigilant regarding safety concerns arising from interactions between the BiTe material and the implantation environment.Addressing this concern can be divided into two directions.First, efforts should focus on developing non-toxic thermoelectric materials that offer comparable performance to BiTe around human body temperature.Second, future research endeavors should prioritize enhancing the PEEK-to-metal seal and polymer encapsulation mechanisms to further fortify device safety.Promisingly, more effective hermetic coating methods, such as atomic layer deposition (ALD), show significant potential. [36]

Conclusion
In summary, our study showcases implantable thermoelectric generators (TEGs) purposefully designed to power implantable medical devices (IMDs).Employing a disk-shaped housing made of polyetheretherketone (PEEK) with a diameter of 65 mm and a height of 8 mm, we achieved dual functionality: biocompatible encapsulation and heightened TEG performance due to PEEK's high thermal resistivity.The TEGs exhibit thermolegs with high aspect ratios (height to width), ensuring adequate voltage output, and their thermal resistance is matched to the surrounding implantation environment to optimize power generation.Our fabrication approach employs dual-layer detachable assembly grids, enabling the assembly and soldering of TEGs with 0.5 × 0.5 × 4 mm 3 bismuth telluride thermolegs in a 16 × 16 array.This configuration yields a power output of 113 μW and a voltage of 79 mV under matched load conditions at ΔT = 3 K in the implantation scenario.To further boost voltage output to meet IMD requirements, we seamlessly integrated a boost converter within the TEG by routing circuitry on the ceramic substrate.This integrated system delivers 53 μW of power at a voltage of 2.5 V, and it adheres to the specific requirements of pacemakers and neurostimulators.Helium leak testing and aging tests on the packaged TEG were conducted, demonstrating that coating technology has the potential to address the moisture-proof limitations associated with polymer-based packaging.The advancement in this work is poised to amplify interest in TEGs for implantation use and serve as a source of inspiration for the design of both implantable TEGs and their packaging.

Experimental Section
Fabrication of TEGs: The solder pasting process involved screen printing equipment (S20, MECHATRONIC ENGINEERING) with stencils (RO-STOCK LEITEPLATTEN) of 100 μm thickness.A lead-free solder material Sn 42 Bi 57 Ag 1 with a melting temperature of 138 °C was applied onto Al 2 O 3 substrates with direct bonded copper, 0.1 mm thick (TIANJIN CEN-TURY ELECTRONICS).The p-type BiSbTe and n-type BiTeSeS thermolegs (Powercool Technology) with nickel and gold coatings on the top and bottom surfaces had dimensions of 1 × 1 × 4 mm 3 .Alignment fixtures were created through laser-cutting (LPKF Protolaser S4) copper sheets, 100 μm thick.Other components of the assembly rig, constructed from X8CrNiS18-9, were manufactured via computer numerical control (CNC) machining.The assembly process comprised placing the bottom substrate with solder paste into a holder, followed by the placement of four alignment fixtures in different directions.After the placement of every two alignment fixtures, one spacer was inserted.Thermolegs were picked and placed using a manipulator (LM901, FRITSCH).Soldering was conducted in a reflow oven (SMT 400C), with a weight of 50 g applied to the top substrate to ensure secure solder connections.Due to the thermal mass within the assembly rig, it was observed that the temperature of the TEG inside the oven was slightly lower than the desired set point temperature, potentially leading to incomplete soldering.To address this issue, a temperature sensor was integrated into the assembly rig to enable accurate temperature regulation.The thermal profile of the reflow soldering process is available in the Figure S1 (Supporting Information).After one round of soldering, the entire assembly rig was removed.In contrast to TEG-1, TEG-3 integrated a boost converter (ECT 310, EnOcean) on the bottom substrate.
Grinding of BiTe Thermolegs: Segmenting the BiTe ingots by wirecutting was limited due to its brittleness.Therefore, the thermolegs utilized in TEG-2 were obtained through a grinding process.The legs, with dimensions of 1 × 1 × 4 mm 3 , were affixed to a grinding jig using a water-soluble adhesive (CrystalbondTM 555, EMS) (Figure S2a, Supporting Information).The grinding jig featured 0.5 mm deep notches designed for mechanical clamping of the legs and served as a stopper during grinding.Following the initial grinding, the cross-section of the legs was reduced to 0.5 × 1 mm 2 (Figure S2b, Supporting Information).Subsequently, the grinding jig was immersed in water to dissolve the adhesive.Every two extracted legs were combined and re-attached to the grinding jig (Figure S2c, Supporting Information).After the second grinding round, the cross-section of the legs was further reduced to 0.5 × 0.5 mm 2 (Figure S2d, Supporting Information).The cross-section of the BiTe legs at each stage of the grinding process is illustrated in Figure S3 (Supporting Information).
Characterization of TEGs: In the measurement environment of the custom experimental setup (Figure 3a), two Peltier modules (TEC1-12706, MIKROELEKTRONIKA) were affixed to aluminum blocks, which were equipped with heat sinks and fans to effectively dissipate the generated heat.A heat flux sensor (PHFS-01, FluxTeq) was securely attached to the bottom aluminum block.On top of the heat flux sensor, an alumina platform served as a substrate for the sample objects.Additionally, two temperature sensors were attached to the platform and the top aluminum block, respectively.To ensure a uniform and controlled clamping force applied to the sample, a set of clamps was specifically designed for this purpose, incorporating three spring-loaded buttons and socket bars.The Peltier controller (TEC-1122-SV, Meerstetter Engineering) allowed for independent regulation of the upper and lower side temperatures, enabling precise temperature control.
The Seebeck coefficient and electrical resistivity measurements (LSR-3, Linseis Messgeraete) were performed at the Leibniz Institute for Solid State and Materials Research (IFW) in Dresden, Germany.Additionally, the thermal conductance of the TEGs (Mini-PEM, ULVAC) was measured at IFW as part of the verification process.
Encapsulation of TEGs: For electrical wiring, a hole with a diameter of 2 mm was drilled in the side wall of the housing, which was made of PEEK (KTK Kunststofftechnik).Thermally conductive glue (Arctic Alumina) was employed to attach the TEG and metal plates, while the remaining parts

Figure 2 .
Figure 2. Fabrication process of compact TEG with 62 thermolegs (1 × 1 × 4 mm 3 ) and an integrated boost converter: a) boost converter circuit laid on the bottom ceramic substrate, b) one of the adapted alignment fixtures facilitating the assembly process, c) assembled TEG-3 with a diameter of 30 mm and a height of 5 mm.

Figure 3 .
Figure 3. a) Custom experimental setup for TEG characterization.b) Method of thermal characterization in the custom experimental setup.c) Method of thermal characterization in the Mini-PEM.

Figure 4 .
Figure 4. a) Electrical resistance; b) open-circuit voltage; c) thermal conductance of TEGs dependent on the number of thermolegs (1 × 1 × 4 mm 3 ).Data 1 was measured by the custom experimental setup, then it was revised by computationally isolating the TEG from the air; Data 2 was measured by the Mini-PEM.d) Power output of TEG-3 dependent on the load resistance with a temperature difference from 2-4 K.

Figure 5 .
Figure 5. Voltage output of TEG-1 under 0 K / 10 K temperature difference cycle with a matched load.

Table 2 .thermolegs 1 × 1 × 1 × 1 × 4
Characterization of TEGs with a cold side temperature of 33 °C and a hot side temperature of 36 °C.The thermal conductance was measured with an open circuit, where data 1 was obtained by using the custom experimental setup, and data 2 was obtained by using the Mini-PEM.The voltage and power output were measured with a matched resistive load.4 mm 3 0.5 × 0.5 × 4 mm3

Figure 6 .
Figure 6.a) Geometry of packaged TEG for electrically active implants.The TEG is embedded between two metal plates and surrounded by a side wall and a lid.The device is filled with air.b) Packaged TEG with PEEK housing and metal plates for efficient thermal interfaces.c) Helium leak rate of the Parylene C coated housing and the unprocessed PEEK housing.d) Evolution of RH with time for brass, PEEK, and Parylene C coated Housing.Dotted curves represent the QSS model with the time constant extracted from the corresponding experimental data.

Table 3 .
The time constant of PEEK, Parylene C coated PEEK, and brass housing at 53 °C and equivalent time constants at 37 °C.