Final Technical Report (FTR) Cover Page a. Federal Agency Department of Energy b. Award Number DE-EE0009638 (non-lab) c. Project Title Front Side Copper Metallization d. Recipient Organization Bert Thin Films Inc e. Project Period Start: 01/01/22 End: 03/31/26 Name: Ruvini Dharmadasa Title: CTO Email address: ruvini@bertthinfilms.com f. Principal Investigator (PI) Phone number: 502 738 7001 ext. 2 Name: Thad Druffel Title: CEO Email address: thad@bertthinfilms.com g. Business Contact (BC) Phone number: 502 738 7001 ext. 1 Name Title h. Certifying Official (if different Email address from the PI or BC) Phone number 9th June 2026 Signature of Certifying Official Date By signing this report, I certify to the best of my knowledge and belief that the report is true, complete, and accurate. I am aware that any false, fictitious, or fraudulent information, misrepresentations, half-truths, or the omission of any material fact, may subject me to criminal, civil or administrative penalties for fraud, false statements, false claims or otherwise. (U.S. Code Title 18, Section 1001, Section 287 and Title 31, Sections 3729-3730). I further understand and agree that the information contained in this report are material to Federal agency’s funding decisions and I have any ongoing responsibility to promptly update the report within the time frames stated in the terms and conditions of the above referenced Award, to ensure that my responses remain accurate and complete. Page 1 of 50 Acknowledgement: "This material is based upon work supported by the U.S. Department of Energy's Office of Energy Efficiency and Renewable Energy (EERE) Solar Energy Technologies Office (SETO) under the Systems Integration and Hardware Incubator Funding Program [DE-FOA-0002437-1594], Award Number: DE-EE0009638." Disclaimer: “This report was prepared as an account of work sponsored by an agency of the United States Government. Neither the United States Government nor any agency thereof, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product, or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process, or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof.” Frequently used Abbreviations: Cu = Copper Ag = Silver Al = Aluminum Si = Silicon PV = Photovoltaic η = Efficiency Voc = Open Circuit Voltage Jsc = Short Circuit Current Density FF = Fill Factor pFF = Pseudo Fill Factor RSH = Shunt Resistance RS = Series Resistance Rf = Lateral Finger Resistance RC = Contact Resistance ρC = Contact Resistivity IR = Infra-Red EL = Electroluminescence BSF = Back Surface Field LCO = Localized Contact Opening PERC = Passivated Emitter Rear Contact HE = Homogeneous Emitter SE = Selective Emitter TOPCon = Tunnel Oxide Passivated Contact BTF = Bert Thin Films Inc. UNCC = University of North Carolina Charlotte GIT = Georgia Institute of Technology NREL = National Renewable Energy Laboratory EOP = End of Project Goals GNG = Go / No-Go Goals Page 2 of 50 Executive Summary: The global drive to lower carbon emissions has resulted in a boom in renewable energy technologies, with the International Energy Agency (IEA) reporting that Solar Photovoltaics (PV) contributed to over 75% of new renewable capacity in 2024; with installed PV capacity exceeding 2 TW, globally. i Crystalline-silicon (c-Si) solar cells account for over 95% of this market with most of these cells utilizing metal contacts produced by screen printing silver (Ag) metallization pastes. The application and processing of these contacts is a critical step in producing high efficiency (η) devices and is also the highest non-Si material cost in cell manufacturing. The pastes typically contain over 85% Ag and are thus subject to the volatility of this highly priced metal. In the first quarter of 2021 the price of Ag was ~ $850/kg, but on the 14th October 2025 the metal reached historical highs of $1785/kg and is now currently hovering ~$1560/kg. These prices cut into the cell manufacturer’s profit margins with the costs eventually passed on to consumers; potentially impacting the Solar Energy Technology Office’s (SETO) 2030 target of 3¢/kWh. The PV industry is also competing against other consumer goods for supply, so Ag is projected to become an expensive bottleneck on the route to reach multi- terawatt production; thus, alternative metallization routes are being actively sought. Bert Thin Films (BTF) Inc.’s approach was to develop a screen printable copper (Cu) based paste (CuBertTM) that can be fired in atmospheric IR belt furnaces without the use of additional steps. In other words, producing a direct drop-in replacement for the Ag pastes. The primary goals of this project were to demonstrate that CuBertTM could be delivered to market at a 90% cost reduction of the established Ag pastes, as well as demonstrating the performance and durability of the Cu contacts on high η Si solar cells. Cu contacts made with BTF’s air fired CuBertTM metallization pastes achieved a maximum η of 21.4% when used as the front contacts on PERC cells, and an η of 24.4% when used as the rear contact on TOPCon cells. Thus, resulting in a relative η of 94% for PERC and 96% for TOPCon, when compared to the same structures commercially metallized with Ag. These results were achieved by improvements to the paste composition and processing conditions. The reliability of the Cu contacts was studied by thermally stressing the bare cells at 200°C, as well as damp heat (85°C/85%RH) testing of mini- modules containing these cells. This work was also supported by micro-structural and elemental analysis of the interface. Improved device performance was assisted by a reduction in the dimensions of the printed contacts; reducing shading and charge-carrier recombination at the interface. This also reduced paste usage to below 0.39 mg/cm2. Cost models demonstrated that even with this amount of paste, the product was still economically viable, with modeled costs showing that when the pastes are manufactured at scale, the costs are calculated to be below 10% of the cost of analogous Ag paste. Domestic and non-US based suppliers of the main components of the paste were sourced to produce a more robust supply chain for the product. Both the cost model and supply chain were verified by third parties. The findings from this project allowed BTF to reduce the risk of switching from Ag to Cu pastes; thus, creating an opportunity to produce metallization pastes in the US. Access to lower cost, domestically produced materials can also potentially spur on domestic manufacturing of cells and modules. BTF continues to actively work with industrial partners to improve and demonstrate the product at scale. Page 3 of 50 Table of Contents Acknowledgement: ............................................................................................................... 2 Disclaimer: ............................................................................................................................ 2 Frequently used Abbreviations: ............................................................................................ 2 Executive Summary: ............................................................................................................. 3 Background: ......................................................................................................................... 5 Project Objectives: ................................................................................................................ 6 Project Results and Discussion: ......................................................................................... 10 Task 1: Paste Formulation and Optimization................................................................... 10 Subtask 1.1: Paste Formulation ................................................................................... 10 Subtask 1.2: Screen Printing........................................................................................ 12 Subtask 1.3: Firing ....................................................................................................... 15 Subtask 1.4: Metallization Design ................................................................................ 19 Task 2: The Copper – Silicon Interface Optimization ...................................................... 21 Subtask 2.1: Preliminary Cell Durability ....................................................................... 22 Subtask 2.2: Copper Diffusion at the Interface ............................................................. 23 Subtask 2.3: Materials Analysis at the Interface........................................................... 23 Task 3: Cell Efficiency Measurement .............................................................................. 26 Task 4: Stakeholder Engagement, Cost Model Development, and Supply Chain Analysis ........................................................................................................................................ 30 Subtask 4.1: Ordering of Supplies ............................................................................... 31 Subtask 4.2: Stakeholder Engagement ........................................................................ 32 Subtask 4.3: Cost Model Development ........................................................................ 33 Task 5: Mini Modules and/or Module Fabrication and Lifetime Measurement ................. 34 Subtask 5.1: Mini Modules and/or Module Fabrication and Lifetime Measurement ...... 35 Subtask 5.2: Module Architecture Optimization............................................................ 36 Subtask 5.3: Increasing Cell Durability......................................................................... 37 Subtask 5.4: Module Construction and Testing ............................................................ 38 Task 6: Co-Firing Optimization ........................................................................................ 40 Task 7: Materials Sourcing, Cost Model Refinement and Manufacturing ........................ 44 Subtask 7.1: Supply Chain Evaluation ......................................................................... 45 Subtask 7.2: Updated Cost Model ............................................................................... 45 Subtask 7.3: Stakeholder Engagement in Pilot Stage Manufacturing .......................... 46 Significant Accomplishments and Conclusions: .................................................................. 47 Path Forward: ..................................................................................................................... 47 Products: ............................................................................................................................ 48 Project Team and Roles: .................................................................................................... 49 References: ........................................................................................................................ 50 Page 4 of 50 Background: The fabrication of metal contacts on c-Si solar cells is a critical step in the manufacturing process; impacting both efficiency (η) and durability of the PV module. For the last 50+ years, these contacts have predominantly used screen printed silver (Ag) paste. PV manufacturing capacity was ~ 135 GW in 2020, and the industry used a total of 2,860 metric tons of Ag representing approximately 10% of the global supply. Capacity has grown rapidly in the last few years with reports indicating a manufacturing capacity of ~700 GW. ii As a result, the PV sector used ~30% of global industrial Ag supply in 2024; roughly 16% of total worldwide Ag demand. iii,iv The International Technology Roadmap for Photovoltaics (ITRPV) expects that as manufacturing capacity grows, the amount of Ag used in solar cell production will also grow significantly in the next decade. Innovations in cell design, screen technology and paste formulation have allowed the industry to reduce Ag usage, but there is a limiting floor. Current average Ag usage is ~9, ~14 and, ~17 mg/W for passivated emitter rear contact (PERC), tunnel oxide passivated contact (TOPCon) and silicon heterojunction (SHJ) architectures, respectively. v SHJ cells require manufacturing temperatures below 250°C due to the inclusion of amorphous Si passivation layers. Consequently, Ag usage has been significantly higher than the other technologies. Fine line printing in the form of knotless screens with openings < 25 µm’s are starting to be employed with Ag pastes, as well as copper (Cu) and hybrid AgCu approaches; η's above 23% have been achieved. vi Reactive Ag inks printed using a syringe-based dispersion system are also under development for SHJ solar cells. The inks, theoretically result in denser contacts using less Ag, than conventionally cured or fired pastes. vii Other printing technologies that produce high aspect ratio prints with lower paste usage are also being explored for SHJ, but screen printing for all cell types is forecast to be the dominant method for the foreseeable future.v Advancements within the higher temperature Ag paste technology have been focused on TOPCon cells, that now command >60% of the market share. Major changes have been made to the frit technology within the paste, to lower the temperature so that they don’t damage the poly-Si passivation layers, as well as optimizing them for Laser Enhanced Contact Optimization (LECO) tools, that have been introduced after the firing step to increase both yield and η. For the front p-type emitter this has meant aluminum (Al) has been significantly reduced or removed entirely from the AgAl pastes; leading to a reduction in recombination. viii Ag free approaches include modification of Al pastes on the rear n-type contact of TOPCon. However, an extra processing step is required in this case, where the silicon nitride (SiNx) anti-reflective coating is first laser ablated to create a local contact opening, before the Al paste is screen printed and fired. ix Cu electroplating has been around since the 1980’s but can also require the use of laser ablation when applied to TOPCon. A nickel (Ni) barrier layer is also deposited to the Si, to prevent Cu diffusion and aid adhesion. However, these contacts often employ an Ag capping layer to reduce oxidation and improve solderability. x Other methods to reduce Ag usage have taken a hybrid approach. These include Ni pastes doped with ~4% Ag nanoparticles. xi However Ag seed layers, deposited by first screen printing a thin Ag layer are the most commonly used approach. In the case of Cu electroplating, they can be used to etch through the SiNx, thus avoiding the need for laser ablation. xii The industry has been trying to adopt Cu plating for years, but the added process complexity has so far made it a cost-prohibitive step for mainstream Page 5 of 50 manufacturing. Leaner Ag seed layer approaches include printing the Ag paste in the form of dashes. The pastes are fired, to make electrical contact with the Si. Fingers and bus bars are then screen printed on top using either a high temperature Al or low temperature Cu paste. The Al and Cu pastes are formulated to not etch through the SiNx coating during their respective firing and curving steps. xiii Many of these methods require more processing steps than conventional Ag screen printing technology, which cuts into the cost savings made by using a lower cost metal. A Cu paste that uses the exact same manufacturing processes from paste roll milling through screen printing to furnace firing would be the ideal method to deliver the most cost savings to the industry. Project Objectives: The industry has been trying to adopt Cu for years but concerns around low-cost processing, performance and durability have limited this. Cu is of course much less expensive than Ag but is also ~800 times more abundant and has nearly the same conductivity. However, it’s small atomic radius means it can rapidly diffuse into Si, damaging the performance of the device. In addition, it oxidizes more rapidly. Bert Thin Films (BTF) Inc’s approach has been to develop a Cu based screen printable paste that can fire through the SiNx coating. The ability to air-fire this paste in a typical fast fire IR belt furnace without oxidation or Cu migration into the underlying Si is the basis of the BTF technology. The chemistry of the paste creates a Cu diffusion barrier upon firing and initial tests have demonstrated durability. The primary aim of this project was to show that BTF’s Cu-based screen printable paste can be delivered to the market at a 90% cost reduction of the established Ag pastes. This project was designed to demonstrate the performance and durability of the Cu contacts on high η c-Si solar cells. At the start of the project, these devices utilized less than 110 mg of Ag paste to form the front-side electrical contacts, but as a highly priced commodity, Ag continues to be a significant material cost. This project aims to prove the performance and economic viability of a direct drop-in replacement for the Ag pastes. As PV cell manufacturing shifted to the Asia Pacific region, so did the manufacturing of the Ag pastes. Earlier this year, one of the remaining European Ag paste suppliers for solar, sold its PV division, while a US supplier divested of its PV paste business back in 2021; resulting in over >90% of Ag paste manufacturing for solar now occurring in China. Therefore, for US cell manufacturing to grow, it must also have a low cost and domestic metallization source. The following project tasks were identified to de-risk the Cu paste and create the opportunity to produce the paste domestically. Task 1: Paste Formulation and Optimization Task Summary: The Cu pastes are primarily composed of Cu powders of varying size ranges. The paste also includes frits, binders, solvent and viscosity modifiers. This task included optimizing the material ratios in the paste, adjusting the composition of the glass frits in the paste, optimizing the processing conditions for a given finger volume, and metallization design. The aim of this task was to reduce the amount of Cu paste required to produce cells with the target efficiency and ensure that devices’ lifetime/durability. Subtask 1.1: Paste Formulation Subtask Summary: This subtask involved modifying and optimizing the formulation, including and not limited to changes to the Cu particle size, the ratios of the different components in the paste and frit composition. Page 6 of 50 Subtask 1.2: Screen Printing Subtask Summary: The volume of the fingers is defined by the screen-printing screen. In addition to the screen opening (finger width), the mesh count, wire thickness and emulsion thickness can impact the paste deposit. This subtask included and was not limited to designing and procuring the screens required to meet the target paste deposit. Subtask 1.3: Firing Subtask Summary: As the Cu metallized cells travel through the IR belt furnace, the paste is exposed to an oxidizing environment. Lowering the air flow in the furnace, would reduce the rate of oxidation. However, the airflow in the furnace determines lamp power usage. A high flow rate uses a large amount of power, while too low a flow rate allows the lamps to turn off for periods, creating unstable firing conditions. In addition, the temperatures in the multiple heated zones, as well as belt speeds can be adjusted to create different firing profiles. This subtask included and was not limited to studies on the impact of the firing profile on the lateral finger resistance (Rf). Subtask 1.4: Metallization Design Subtask Summary: The move towards high efficiency devices also impacted the front metallization design. This subtask monitored Rf as formulation and processing changes were made so that the metallization design (including the minimum number of bus bars, and finger dimensions) required to produce the target η’s could be defined. Task 2: The Copper – Silicon Interface Optimization Task Summary: This task included and was not limited to material analysis of the materials at the interface, as well as accelerated stress tests of the cells. The aim was to show that the Cu does not diffuse into the device and to lower the contact resistance (RC) to achieve the target cell η. Subtask 2.1: Preliminary Cell Durability Subtask Summary: Changes to the frit and firing conditions directly impacts the pastes’ ability to prevent Cu from diffusing into the device. The quality of the inherent diffusion barrier was studied following accelerated aging protocols developed to investigate the performance of diffusion barriers in Cu electroplated cells. The cells were placed in an inert atmosphere to minimize oxidation and then thermally stressed at elevated temperatures (up to 200°C) for up to 1000 hrs. The open circuit voltage (Voc) and pseudo fill factor (pFF) of the device were then periodically measured. Subtask 2.2: Copper Diffusion at the Interface Subtask Summary: The aim of this task was to select cells which had been thermally stressed to evaluate their interface. Analysis of the interface included and was not limited to secondary ion mass spectroscopy (SIMS). This technique was used to study the depth profile of the elements in the Si, directly beneath the contact region, and was a further method of evaluating the diffusion of Cu into the device. Subtask 2.3: Materials Analysis at the Interface Subtask Summary: Changes to the frit composition and additives impacts the rate the paste etches the SiNx anti-reflective coating during the firing process. The remaining materials at the interface will impact both the diffusion barrier properties of the paste and the RC. This subtask aimed to use analytical techniques including but not limited to SIMS, Page 7 of 50 Raman spectroscopy, scanning electron microscopy (SEM), energy dispersive x-ray spectroscopy (EDX) and RC measurements to inform the decisions made on the paste modifications and correlate the results to the initial minority lifetime and η of the cell. Task 3: Cell Efficiency Measurement Task Summary: This task compared the performance of the Cu paste on high efficiency silicon solar cells to the same cells metallized with Ag paste. The cells had a PERC) and/or TOPCon structure. The Ag metallized cells will be sourced directly from the manufacturer, consequently the paste formulation and processing conditions had already been refined for these pastes. As a result, the aim was to produce Cu metallized cells which have a target η relative to the Ag metallized cells. Task 4: Stakeholder Engagement, Cost Model Development, and Supply Chain Analysis Task Summary: This task was aimed at identifying, engaging and collecting feedback from relevant stakeholders who would be willing to take this technology to the next stage in manufacturing or that would be needed to gather feedback on technical metrics. A cost model / tecno-economic analysis would also be developed for this technology. Subtask 4.1: Ordering of Supplies Subtask Summary: The purchase orders/invoices for the materials directly related to the paste were used to develop the draft cost model. Supplies including but not limited to the following items were purchased: (1) Cu powders, (2) Si precursor cells, (3) Si solar cells metallized with Ag and (4) screen printing screens. Subtask 4.2: Stakeholder Engagement The aim was to identify, engage, and get feedback on the technical metrics from relevant stakeholders including but not limited to cell, PV module, paste manufacturers. Stakeholder engagement was also needed to independently verify the cost model / tecno- economic analysis for this technology. Subtask 4.3: Cost Model Development Subtask Summary: This subtask was used to build a cost model that was informed by vendor pricing up to production scale. The model included materials, processing (labor and depreciation), overheads and margin. The baseline cost of Ag pastes was derived from quotes for batches of Ag paste from different suppliers. The model aimed to adjust the cost of the Ag paste to account for differences in operational expenditures (OpEx) as well as any possible impact on capital expenditures (CapEx). This task included collecting material cost data from suppliers and vendors. Task 5: Mini Modules and/or Module Fabrication and Lifetime Measurement Task Summary: The goal was to show that the Cu paste would not adversely affect the performance of modules that are warrantied for up to 30 years. This task utilized industry accepted protocols for modules as well as accelerated testing for the solar cells. The task included also producing and testing different sized modules. Subtask 5.1: Mini Modules Fabrication Subtask Summary: To study durability of the Cu contacts containing, mini modules using two full sized cells would be constructed and tested. These tests included and were not limited to the damp heat test, where the mini-modules or modules were exposed to Page 8 of 50 85°C/85%RH environment for up to 1000 hours and measurements are taken periodically. Measurements included and were not limited to current-voltage (I-V) and electroluminescence (EL) measurements. The test evaluates the ability of the Cu metallization to withstand damage (e.g. corrosion) that would be experienced in the field. Subtask 5.2: Module Architecture Optimization Subtask Summary: Based on the accelerated aging tests on the mini-modules, the module architecture (e.g. Glass/glass, glass/backsheet, encapsulant material) for the certification test were determined. Subtask 5.3: Increasing Cell Durability Subtask Summary: Any changes made to the frit and firing conditions directly impacts the pastes’ ability to prevent Cu from diffusing into the device. The aim of this subtask was to ensure that cell durability was improved further, so that the cells used in the modules testing would not fail due to Cu diffusion into the cell. The quality of the diffusion barrier was studied using accelerated aging protocols developed to investigate the performance of diffusion barriers in Cu electroplated cells. Subtask 5.4: Module Construction and Testing Subtask Summary: Larger mini-module structures will be fabricated to undergo industry accepted testing procedures used in the IEC 61215 certification. Task 6: Co-Firing Optimization Task Summary: This task continues to use an IR belt furnace to develop and optimize the firing conditions for the Cu paste. This task aimed to determine the amount of paste required to achieve the target η relative to η of the same cell structure metallized with Ag pastes when the Cu paste is co-fired with the rear contact. Task 7: Materials Sourcing, Cost Model Refinement and Manufacturing Task Summary: This task was focused on evaluating the economic and logistical challenges of scaling the technology up to production level. Subtask 7.1: Supply Chain Evaluation Subtask Summary: Supplies including but not limited to the following items were purchased and the quotes analyzed to understand risks and challenges associated to this technology supply chain. This included (1) Cu powders, (2) Si precursor cells, (3) Si solar cells metallized with Ag, and (4) screen printing screens. The robustness of the supply chain to be able to scale the paste up to production levels will be assessed. Subtask 7.2: Updated Cost Model Subtask Summary: This subtask refined the cost model for the Cu paste. The updates included and were not limited to feedback from the independent reviewer, formulation changes and updated cost data projections. Subtask 7.3: Stakeholder Engagement in Pilot Stage Manufacturing Subtask Summary: The aim was to engage in discussions with relevant stakeholders including and not limited to cell, PV module, paste manufacturers to ensure that the technology meets all the specifications and requirements needed by the private sector to move it to the next stage in the manufacturing process. Page 9 of 50 Project Results and Discussion: Task 1: Paste Formulation and Optimization Milestone Performance Assessment Tool / Method of # Metric Measuring Success Value 1.3 Paste ≥ 8 Paste Formulations Viscometer and grind gauge. Formulation Viscosity less than 1,500 Pa.s at 0.5 rpm. Grind gauge lower than 25/10. 1.4 Finger ≥ 72 samples Measurement of Resistance Down-selection of formulation Finger resistance. meeting finger resistance <8 Ω/cm at 0.50 mg/cm2 ± 10% paste on a 4 × 4 cm2 sample or larger. 1.7 Number of Bus Number of Bus Bars required Lateral finger resistance data with Bars to produce a cell with an respect to firing conditions in the IR efficiency ≥ 80% relative to belt furnace. the Ag metallized cells using ≤ 0.50 mg/cm2 ± 10% of paste on a cell ≥ 4 × 4 cm2 in size. 1.10 Metallization 0.42 mg/cm2 ± 10% paste on Values for the following parameters to Design and a cell ≥ 4 × 4 cm2 in size. produce 3 cells using the Cu paste Processing Cell efficiency ≥ 85% relative with an efficiency defined in GNG-1A: to the Ag metallized cells. - Number of Bus Bars Standard deviation ± 1% - Printing Parameters (absolute efficiency). - Firing Conditions - Mass of Paste - Minority Carrier Lifetime GNG-1C Mass of Cu ≤ 100 mg ± 10 mg (0.39 Average, standard deviation of the 50 paste deposited mg/cm2 ± 0.04 mg/cm2) on a cells measured as part of GNG-1A. cell ≥ M4 (161.7 mm) in size. Subtask 1.1: Paste Formulation Milestone 1.3. Paste Formulation The CuBertTM Cu pastes are primarily composed of Cu powders of varying size ranges coated by BTF. The pastes also include frits, binders, solvent and viscosity modifiers. Initially nine 1.5 kg sized batches (P1 to P9) of paste were produced, with a variation in the particle size distribution (PSD) of the Cu powders and the molecular weight (MW) of the binder (Table 1). The PSD of the Cu powders in pastes P1 to P5 were varied by changing the type and blend of the powders. The smaller the particle size, the higher the surface area and the more susceptible the Cu powders were to oxidation. However, increasing the particle size, limited how small the fingers could be printed. P1 used powders with the highest surface area (0.90 m2/g), while pastes P2 to P8 had surface areas ranging from 0.59 – 0.65 m2/g. Both the PSD of the Cu and the MW of the binder affect the rheology of the pastes, thus impacting printability. Figure 1(a) shows the initial viscosity data for the pastes. The impact of these formulation changes on Rf was studied as part of Milestone 1.4. Page 10 of 50 Over the course of the project, over 50 different paste formulations were made. The composition of the glass frits, solvent and viscosity modifiers were all changed to improve the pastes rheology, printability and electrical performance. By the end of the project, the screen openings had been reduced to 25 µm. This required a significant reduction to the paste’s viscosity, to allow the paste to be printed in fine lines using industrial print speeds (≥ 200 mm/s). Figure 1(b) shows the viscosity of the paste (F53) used to print the fingers on PERC cells used in the End of Project (EOP-B) goals related to cell η. Table 1. Paste compositions highlighted within this report. PSD = Particle Size Distribution. MW = Molecular Weight. FOG = Finess of Grind for Grind. Paste Copper Binder MW Binder in Total Frit Ratio of Frit FOG ID Powder PSD Paste (wt%) (wt%) 1 to Frit 2 (µm/µm) P1 Powder 1 Medium MW 3.8 11.1 0.29 5/2 P2 Powder 2 Medium MW 3.8 11.2 0.30 8/5 P3 Blend 1 Medium MW 3.8 11.3 0.30 8/5 P4 Blend 2 Medium MW 3.8 11.2 0.30 8/4 P5 Blend 3 Medium MW 3.8 11.3 0.30 9/5 P6 Blend 1 High MW 3.8 11.0 0.30 8/5 P7 Blend 1 Medium MW 6.1 10.8 0.30 7/4 P8 Blend 1 Blend 1: Medium & High MW 6.1 10.8 0.30 7/3 P9 Blend 1 Blend 2: Medium & High MW 6.1 10.6 0.30 7/4 P2-1 Powder 2 Medium MW 4.0 11.7 0.35 - P2-2 Powder 2 Medium MW 4.0 10.8 0.24 - P2-3 Powder 2 Medium MW 6.8 11.3 0.26 - P2-4 Powder 2 Medium MW 6.9 10.5 0.23 - P2-5 Powder 2 Medium MW 5.4 11.5 0.35 - P2-6 Powder 2 Medium MW 5.6 7.6 0.72 - F10 Powder 2 Medium MW 5.6 7.6 0.73 8/5 F11 Powder 2 Medium MW 5.7 6.7 0.49 8/5 F12 Powder 2 Medium MW 5.7 7.2 0.60 8/5 F6 Powder 2 Medium MW 3.9 10.1 0.12 8/3 F13 Powder 2 Medium MW 4.9 7.3 0.59 10/3 F18 Powder 2 Medium MW 4.3 10.4 0.40 13/3 F100 Powder 1&2 Medium MW 5.4 6.6 0.54 10/6 (a) (b) Figure 1. Examples of the viscosity vs. spindle speed for (a) Pastes P1, P3, P5, P6, and P9 used in the finger resistance studies for Milestone 1.4. (b) Viscosity for Paste F53 used to print the fingers on the cells used to fulfill EOP-B. The viscosities were measured using a Brookfield Ametek HBDV1M cone and plate viscometer with CPA-52Z spindle. Page 11 of 50 Subtask 1.2: Screen Printing During the project’s first quarter, BTF received its upgraded screen printer (ASM E by DEK). The printer allowed cells up to M12 (441 cm2) in size to be printed and has visual alignment capabilities, making it easier to align the prints with the selective emitter regions. Moving from the old printer (HMI MSP885) to the upgraded printer, required time to develop new “standard” settings. The printer settings had to be adjusted for each paste (P1 to P9), in order to deposit similar amounts of paste. As the project progressed, the Cu powder and MW of binder used in Paste P2 were selected for further paste optimization. Changes to the paste included increases in the proportion of binder, frit and rheology modifier used in the paste. The proportion of binder in the pastes was varied with P2-1 using 4.0 wt% and P2-3 using 6.8 wt%. The printing conditions were kept constant for all screens. A significant variation in height was observed for both pastes and examples of the roughness for pastes printed with 30 and 65 μm screen openings are shown in Figure 2. The resistivity of the lines was calculated using the smallest cross-sectional area. The lowest ρ values were observed for the 65 μm prints, and had values in the 16 – 30 μΩcm range. (a) (b) (c) (d) Figure 2. Examples of the optical profiles of fingers printed with (a, b) P2-1 and (c, d) P2-3, using screens with (a, c) 30 μm and (b, d) 65 μm openings. Fingers were printed with a print gap of 3.0 mm, squeegee pressure of 0.33 kg/cm and print speed of 50 mm/s. BTF’s screen manufacturer recommended a finer mesh for screen openings < 50 μm. Thus, reducing the open area of the screens from 56% for a 400 mesh to 39% for a 640 mesh. Print studies demonstrated that even with the finer mesh the viscosity of the paste had to be lowered, in order to reduce the roughness (peaks and valleys) along the fingers. The rheology was adjusted by adding extra solvent to the paste. However, this also resulted in significant growth in the line width, a reduction in height and divergence from a rectangular shaped cross-section. The organic vehicle was further investigated, with changes made to the rheology modifiers and solvent system to increase the solids Page 12 of 50 loading. The largest impact on the paste rheology came from a change in solvent. Figure 3(a) shows the impact of switching the solvent from Solvent 1 (used in pastes P2-1 and P2-3) to Solvent 3 for a concentrated solution of the binder. A significant reduction in viscosity was observed due to Solvent 3 having (i) a relatively larger sized molecule compared to the binder; (ii) containing only one isomer with no branches, compared to two in Solvent 1; and (iii) having fewer functional groups per molecule (which can create intermolecular forces with the binder). When the binder and rheology modifier were dissolved in the solvents, the viscosity of the organic vehicle (Medium B) using Solvent 3 was significantly reduced at the higher shear rates (Figure 3). Once the inorganic components of the paste were added, the viscosity of the final paste at the high shear rates were similar. However, paste F4 made using Solvent 3 ended up requiring less solvent than paste F2 prepared with Solvent 1 (Figure 3(c)). When printed with screen gap of 1 mm, paste F4 could be printed with a squeegee pressure of 0.33 kg/cm and print speed of 200 mm/s when using a screen with 400 mesh and 65 µm opening. However, when printed through with a finer mesh (#640), the squeegee pressure had to be increased to 0.40 kg/cm and the print speed had to be reduced to 35 mm/s to print through 50, 40 and 30 µm openings. Even at these conditions, the paste struggled to pass through the 30 μm openings; with fewer samples being printed with this line width. Figure 4 shows examples of the fingers printed with paste F4 through a 65, 50, 40 and 30 μm screen opening. The fingers for the 50 μm prints were the most rectangular in shape. As expected, both the Voc and Jsc of the cells were observed to increase slightly, as the finger width decreased and shading was reduced (not shown). The performance of cells printed using a screen opening of 30 μm, was limited by the finger resistance, as the fingers were less uniform and more susceptible to oxidation at slow belt speeds. (a) (b) (c) Figure 3. Viscosity of (a) Binder solutions prepared with 55 wt% Binder 1 and either 45 wt% Solvent 1 or 45 wt% Solvent 3, (b) Organic vehicle prepared with the two solutions shown in (a); containing 42 wt% Binder 1, 5 wt% Rheology modifier and 53 wt% Solvent 1 or 3. (c) Shows the viscosity of the finished pastes F2 and F4 prepared with Mediums A (Solvent 1) and B (Solvent 2), respectively. Apart from the solvent, F2 and F4 had the same paste composition. F4 required less solvent to achieve a similar viscosity to F2 at the higher shear rates (i.e. spindle speed). As the project progressed, further formulation adjustments were made to the paste to improve rheological and electrical properties. An alternative screen manufacturer was also sourced, who supplied screens for industrial PV production lines. They were able to produce knotless screens for fine line printing. Figure 5 shows single prints of the same paste (standard screen) through a screen with the mesh angle of 22.5° and a knotless screen with a 0° mesh angle. Both screens had a 25 µm opening. A clear improvement to the uniformity of the fingers was observed with a reduction in the roughness of the Page 13 of 50 prints. After being printed as an M6 sized cell, the cells were scribed into 16 cm2 sections and fired in BTF’s 3-zone customized furnace. The cells were measured using a Sinton FCT-650 flash tester. The group of 24 sections achieved an average η of ~92% relative to commercially produced Ag cells, and the champion cell had a ~95% relative η compared to Ag sections taken from a commercially produced cell with the same structure (Figure 5). (a) (c) (b) (d) Figure 4. Optical profiles of fingers printed with paste F4 through a (a) 65 μm (Screen SA), (b) 50 μm (Screen SE), (c) 40 μm (Screen SE) and (d) 30 μm (Screen SE) screen opening (Table 3). 22.5° Profile of finger using 22.5° biased screen. 0° Profile of finger using 0° biased screen. Efficiency Relative to Silver, Ag Reference Champion 16 cm2 96% cell printed with a zero biased screen 94% and fired in a 3-zone furnace. 92% 90% 88% 86% Number of Cells Printed with 0 Biased Screen = 24 Figure 5. Examples of the profiles of fingers printed with a screen opening of 25 µm using a 22.5° and 0° angled mesh screen. Relative efficiency results for twenty-four 16 cm2 cells printed with the 0° biased screen and fired in BTF’s 3- zone furnace. The current-voltage (I-V) and Suns-Voc curves, as well as the electroluminescence image for the champion cell from this set is also shown. Page 14 of 50 Subtask 1.3: Firing The firing process is a critical step in the metallization process and ended up being one of the more challenging parts of the project. The project was originally conceived with the firing and measurement of the full-sized (wafers with sizes ≥ 261 cm2) cells occurring at the University of North Carolina Charlotte’s (UNCC) using their 6-zone IR belt furnace. Initially only one profile measured at the start was taken as the correct temperature, and zone 6’s setpoint was adjusted accordingly, but this resulted in unclear trends in the data. Consequently, the number of profiles collected before and after firing the cells was increased. On the advice of BTF’s advisors the firing procedure was also modified so that ~100 “dummy” cells were passed through the furnace prior the actual cells, to try to stabilize the furnace temperature. These changes improved results, but temperature variations >20°C were still observed at the same settings. Some of this variation was attributed to the thermocouple shifting position as it moved through the furnace. Figure 6. Images of the belt and a cell on the belt in the six zone IR belt furnace at UNCC. The planned upgrade to the belt did not meet the project’s timeline. (a) (b) Figure 7. Shows the average (from 5 cells) aluminum sheet resistance with respective to the distance from (a) the left edge of the cell and (b) the top edge of the cell. The left and right edges are in contact with the belt, while the top and bottom edges are not. The cells were fired at: (i) a belt speed of 300 in/min, zone 6 temperature of 680°C, and air flow of 150 SCFH through all zones, (ii) a belt speed of 300 in/min, zone 6 temperature of 700°C, and air flow of 150 SCFH through all zones, (iii) a belt speed of 300 in/min, zone 6 temperature of 720°C, and air flow of 150 SCFH through all zones and (iv) the aluminum contact firing conditions using a belt speed of 210 in/min, zone 6 temperature of 785°C, and air flow of 200 and 300 SCFH through zones 1 to 4 and 5 to 6, respectively. The second issue with the furnace was that the cells were only supported at the edges of the cells (Figure 6). As a significantly older furnace, the belt was originally designed for much smaller cells, as a result the more modern wafers which are wider and thinner bow Page 15 of 50 significantly as they are transported. During the co-firing of cells with a full aluminum paste coverage on the rear, the impact of the non-uniform temperature variation was clearly observed, through the measurement of the aluminum sheet resistance (Figure 7). Figure 7 shows that ~two thirds of the cell was at a significantly lower temperature than the center. The center of the cell was unsupported, resulting in this area getting hotter and a lower Al sheet resistance in this region. This variation in temperature also impacted the formation of the localized back surface field (L-BSF) on the rear Al contact but additionally the front Cu metallization; resulting in regional differences in the RC and Rf values of the Cu fingers across the entire cell. The impact of this variation is likely to be a significant factor in the wide range of Rf and η values observed for the cells. To meet the target η while continuing to use this furnace, attempts were made to reformulate the paste so that it could be fired at higher temperatures and slower belt speeds, while alternative furnaces were sought. Initially a six zone BTU furnace in Singapore was located. The furnace had a belt made from a nichrome wire mesh with raised wires; so that the cell would not sit directly on the mesh. However, the speed range for this furnace was limited to 20 – 220 in/min. Cells were prepared for this furnace by printing paste P2-5 through a 65 μm opening. Paste P2-5 was selected at the time as it had the largest firing window of the pastes being studied. Around ~200 M6 sized PERC cells were printed with 0.41 ± 0.4 mg/cm2 of paste being used on the fingers. Although the cells were ready to be shipped, a communication module in the furnace failed. Alternative firing sites were looked at and eventually an agreement was reached with Georgia Institute of Technology (GIT) who were in the process of bringing an old furnace back online to fire Cu. Like UNCC, the furnace at GIT also had 6 zones, however its belt was similar to the BTU furnace (Figure 8(a)). GIT tried matching the firing profiles to those obtained at UNCC (Figure 8(b)). A faster belt speed in addition to different temperature and air flow settings was required, due to the longer length of the heated zones. GIT’s furnace was also observed to have a slower cooling rate. At the time GIT only had current-voltage (I-V) system that could probe 5 of the 9 bus bars. This resulted in a maximum η of 15.3%, at a peak temperature of ~630°C. Consequently, the cells were sent to the UNCC for measurement using their Sinton FCT- 650 system. This system could contact all 9 bus bars and an efficiency of ~20.2% was measured; achieving a relative η of approximately 86% of a commercially prepared Ag metallized cell. However, only 1 – 2 cells were fired at this condition. When firing 25 cells in a row, a maximum relative η of 78.5% was achieved at the same settings. By tracking the cell η and EL images vs. firing order, it was clear that the thermal mass and throughput rate were causing the temperature to drop. Subsequently, the number of “dummy” cells prior to firing the actual cells was increased and the power input to the lamps was monitored. Although GIT’s furnace produced a more uniform firing, its furnace was designed over 17 years ago, so lacked supports for larger cells, resulting in the center of the cell being cooler than the edges. Although most firings at GIT resulted in a very large spread in results, one or two cells per firing set could be obtained with high η. In parallel, paste development work continued at BTF using a custom made 3-zone firing furnace for 16 cm2 cells (Figure 9(a)). The stage was motorized to produce a similar temperature/firing profile to those obtained in the IR belt furnaces (Figure 9(b)). Due to the issues with UNCC’s and GIT’s furnaces, the Go/No-Go goal’s GNG-1A, B and C were modified to use 16 cm2 PERC cells fired at BTF. Page 16 of 50 Set 1 Wafer Temperature (°C) (b) (a) Time (s) Figure 8. (a) Belt used in Georgia Tech’s furnace. (b) Firing profile for Set 1 at Georgia Tech (UCEP TPS) compared to a profile taken at UNCC (Bert Film). (a) (b) Figure 9. (a) Schematic image of the 3-zone firing furnace at BTF. (b) Firing profile in the furnace with respect to time. The thermocouple is attached to the wafer using a high temperature cement, as a result, a time lag results when measuring the firing profiles, which leads to slightly lower temperatures being measured. During the latter half of the project, BTF raised further capital (EOP-F). The first part of the raise funded the buildout of BTF’s facilities as well as the purchase of an IR belt furnace (Figure 10(a)). The most up to date firing profiles from GIT were used as the starting point for the firing profile in BTF’s 8-zone furnace. This ended up requiring two software upgrades. The first upgrade was to increase the belt speed to account for the longer heating and cooling sections. The second involved allowing the setpoint temperature of zones 6 – 8 to be increased. The original belt had sloping supports, designed to accommodate cells with widths ranging from M6 (166 mm) to M12 (210 mm). Due to the design of the furnace drums, no cell supports could be placed in the center of the belt. Figure 10(a) shows an EL image of an M6 sized PERC cell fired using only the belt supports. A large portion of the center of the cell clearly touched the belt, where it acted as a heat sink. The result was a higher series resistance, Rs in this region, and hence lower performance. Therefore, temporary tray supports were designed and manufactured to prevent the cell contacting the woven region of the belt; resulting in a more uniform firing (Figure 10(b)). BTF worked with the manufacturer of the furnace to design a belt with ~50% lower thermal mass that wouldn’t need additional supports for the cells. However, this newer belt had a lead time of 2-3 months and was not available in time to fire the cells for the end of project goals (EOP-A and B). Page 17 of 50 (a) (b) (c) Figure 10. (a) Eight zone IR belt furnace installed at BTF in October 2024. Electroluminescence (EL) images of a Cu metallized PERC cell from print group B1 fired (b) using only the belt supports and (c) with additional tray supports. The cells were fired with the same firing conditions. Funds raised during the project were also used to purchase a LECO tool. In recent years the LECO process has been implemented into high throughput manufacturing lines, to increase the yield and narrow the cell η distribution. The process uses a relatively lower firing temperature profile to create the initial electrical contact. The cell is placed under reverse bias and scanned with a focused laser beam, creating high current densities. Since metallization pastes typically do not create a uniform contact area, the high current is localized. It is believed that heat generates a “current induced firing” at the interface. The smaller, and more resistive the contact points, the higher the localized temperature.xiv In the case of Ag pastes, significant changes have been made to the glass frit in the paste to lower the etching rate and create an initial resistive contact, that improves with LECO. The LECO tool was used on cells for the EOP-A and EOP-B; resulting in a further reduction to the amount of Cu paste needed to produce higher η devices. Milestone 1.4. Finger Resistance Pastes P1 to P9 (Table 1) were printed in a half cell pattern with nine bus bars on M6 sized PERC cells. The pastes were printed through 400 mesh screens with 40 and 50 µm openings. To improve printability, the viscosities were adjusted with the addition of extra solvent (Figure 1). The cells were fired using UNCC’s 6-zone IR belt furnace. The belt speed was changed from 250 to 325 in/min. For each set, 3 firing profiles were measured before and after the cells were fired (Figure 11(a)). The Rf was measured between bus bar 1 and 9 on each half cell, with a distance of 14.4 cm. Each half cell had 64 fingers and two resistance measurements were taken on each side. Some of the cells showed an order of magnitude difference in Rf for the half-cell patterns on either side of the cells; indicating either print or firing uniformity issues across the cell. All the pastes printed using the 40 µm and 50 µm screen openings, had weights well below the target mass deposition of 0.38 mg/cm2. Consequently, at the slower belt speeds, Rf was extremely high. As expected, the smaller the PSD, the higher the total Cu surface area, and the more susceptible the particles were to oxidation. As a result, the finger resistance for P1 did not drop below 100 Ω/cm. Pastes P2, P4 and P5 were selected for further investigation with screen openings of 40, 50, 55 and 65 µm’s. Figure 11(b,c) show that there was a lot of variation in Rf values for Pastes P2 and P4. This was later attributed to issues related to a uniform firing. However, trends in the Rf values with respect to the printed weight showed that pastes P2 and P4 achieved target resistances of < 8 Ω/cm with < 0.4 mg/cm2 of paste. Thus, hitting milestone 1.4’s target. Page 18 of 50 Z6 = 715°C (a) (b) (c) Figure 11. (a) Average peak wafer temperature with respect to belt speed and the corresponding trendline. The error bars are the standard deviation. (b, c) Average resistance of a single finger with respect to the mass of paste. The M6 sized cells were fired at UNCC with a belt speed of 325 in/min and a zone 6 set point temperature of 715°C. The mass of paste was varied by changing the screen (40, 50, 55 and 65 µm) openings. Fingers were printed with pastes (b) P2, (c) P4. The cells used a half cell design; thus, two finger resistance values were measured for each cell. Subtask 1.4: Metallization Design Milestone 1.7. Number of Bus Bars Unreliable firing and print issues created a lot of noise and made paste development challenging. The majority of the work carried out in this project utilized M6 (276 cm2) sized PERC cells, with the relative η of the Cu metallized cells compared to the same structure metallized with Ag on commercial production manufacturing lines. Therefore, the same screen design (i.e. finger and bus bar pitch) was used to print the Cu fingers and bus bars. Consequently, the majority of the M6 sized PERC cells printed for this project had 127 fingers (~1.30 mm pitch) and 9 bus bars (~18 mm pitch). Thus, the focus was not on trying to change the metallization design, but to try to lower the bulk resistivity and Rf values to achieve the η targets. To achieve the η target of ≥ 90% relative η compared to commercially produced Ag cells for GNG-1A, improvements to the paste rheology, composition and firing conditions were made. The PERC cells were screen printed as M6 sized wafers using ≤ 0.42 mg/cm2 of Cu paste for the fingers and then scribed into 16 cm2 sections. These sections were then fired in BTF’s 3-zone IR furnace. These sections had 3 bus bars with a pitch of 18 mm, and an average Rf of ~2.0 Ω/cm; resulting in a relative η ~88% (Figure 12). Milestone 1.10. Metallization Design and Processing To achieve relative cell η’s ≥ 85% relative to the Ag metallized cells using only 110 mg (~0.42 mg/cm2) of paste, M6 PERC cells with a selective Emitter (Set B, Table 3) were used. Fingers were printed with 40 µm screen openings using multiple prints to obtain the required cross-sectional area of ~600 µm2 using ~106 mg (0.39 mg/cm2) of paste. The cells were printed with nine bus bars (18 mm pitch). To achieve η’s >85%, these cells also had to be scribed into the 16 cm2 sections and fired in BTF’s 3-zone furnace using a belt speed of 250 in/min, with zone 1, 2 and 3 setpoint temperatures of 450, 450 and 905°C, respectively. Carrier lifetime at Vmpp averaged ~60 µs. No edge passivation was used on the sections. The larger edge to area ratio of the scribed sections were thought to be the major contributor for increased recombination. Therefore, this milestone was achieved in terms of performance, if not cell size. Page 19 of 50 Figure 12. Box plot showing the average resistance and profile of a single finger from samples prepared for GNG-1C. The cells had a relative efficiency of ~88% relative to commercially produced Ag cells. Go/No-Go Milestone: GNG-1C. Mass of Cu Paste Deposited At the start of the project, approximately 0.60 mg/cm2 of paste was being used to print the fingers on G1 (250 cm2) sized wafers. Over the course of the project particle size distribution of the powders was lowered, to shrink the screen openings, and thus, the mass of paste used to print the fingers. Figure 13 shows the mass of paste used to print the cells for GNG-1A & 1C. 70% of the cells had a mass ≤ 0.39 m2/g, and 100% of the cells had a mass ≤ 0.43 mg/cm2 on an M6 sized cell; thus, Go/No-Go goal GNG-1C was deemed to have been met. By EOP-B the mass of paste used to produce cells with η’s > 90% on M6 sized PERC cells was ~0.31 mg/cm2 (Figure 14). Further reductions in paste are expected with improvements to the print quality. Figure 13. Mass of Copper Paste (Dried) Used to Print 119 M6 Sized Cells used in GNG-1C. 200 100% Number of M6 Sized PERC Cells with 180 90% 160 80% 140 70% Selective Emitter 120 60% 100 50% 80 40% 60 30% 40 20% 20 10% 0 0% 0.26 0.27 0.28 0.29 0.30 0.31 0.32 0.33 0.34 0.35 0.36 0.37 0.38 0.39 0.40 0.41 0.42 More Mass of Copper Paste (Dried) Used to Print the Fingers (mg/cm2) Frequency Cumulative % Figure 14. Mass of Copper Paste (Dried) Used to Print 588 M6 Sized Cells used in EOP-B. Page 20 of 50 Task 2: The Copper – Silicon Interface Optimization Milestone Performance Assessment Tool / Method of # Metric Measuring Success Value 1.6 Start of Copper ≥ 8 cells in test First set of samples into oven, so that (Cu) Diffusion Selected from ≥ 2 paste samples can be selected for milestone Studies formulations and 1.8. ≥ 2 different firing conditions 1.8 Cu Diffusion Pseudo Fill Factor (pFF) and ≥ 10 (4 × 4 cm2) samples will reach the open circuit voltage (Voc) target. after 1000 hrs at 200°C is ≥ 80 % of its initial value. The current-voltage (I-V) performance From cells with an initial of the cells at the start and at the end efficiency ≥ 80% relative to of the 1000 hrs will be measured. the Ag metallized cells. Standard deviation ± 1% The data will be correlated to the initial (absolute efficiency). minority lifetime. 1.9 Contact <1 Ω ± 10% Analysis will include and not be limited Resistance to: From cells with efficiency ≥ • Raman spectroscopy 85% relative to the Ag • Scanning electron microscopy metallized cells. Standard (SEM) deviation ± 1% (absolute • Energy dispersive x-ray efficiency). spectroscopy (EDX) • Contact resistance measurements • SIMS spectrometry The data will be correlated to the initial minority lifetime and efficiency of the cell. 1.11 Cell durability ≥ 8 cells being tested Last date to add samples into oven for start date for Cells are metallized from the cell durability test for samples if doing samples in same batch of paste(s). the test in two batches. GNG-1B GNG-1B Cell Durability Voc and Pseudo fill factor Data tracking pFF and Voc for the cells. (pFF) remains within 10% of cells initial value after 1000 The performance of the cells is hrs at 200°C. measured at 0, 1, 10, 30, 70, 110, 150, 200, 300, 400, 500, 600, 700, 850, and Total 10-20 cells. 1000 hour time points. The number of cells is dependent on the capacity of Initial values for the cells the oven. tested with the Cu metallization is ± 3% of the Current-voltage (I-V) measurements of Voc and pFF values for the the cells at the start (0 hrs.) and end cells metallized with Ag. (1000 hrs.) of the durability tests will be performed. Minimum number of cells to be measured: 5 – 10 cells metallized Cu paste 5 – 10 cells metallized Ag paste Paired t-test for Suns-Voc vs. actual I-V measurement per IEC 60904-1 for each of the 2 groups. Page 21 of 50 Subtask 2.1: Preliminary Cell Durability Milestone 1.6 & 1.8. Start of Copper (Cu) Diffusion Studies and Cu Diffusion The Cu pastes developed by BTF are being designed to avoid the use of extra diffusion barriers; thus, avoiding the additional processing steps. The quality of the “inherent” diffusion barrier is studied internally at BTF following accelerated aging protocols developed to investigate the performance of Cu electroplated cells. xv,xvi The cells are thermally stressed at 200°C, and the Voc and pFF are monitored over time. The loss in pFF is assumed to represent a certain depth that the Cu has diffused into the Si and the time taken (t95) for the pFF to drop by 5% of its initial value is monitored. The longer the time, the more durable the cell, and the greater the probability that the module’s constructed with these cells will last. The initial round of cells was printed with six different pastes with varying amounts of frit in the paste (P2-1 to P2-6 Table 1). After printing, the cells were scribed into 16 cm2 sections and fired in BTF’s 3-zone IR furnace. The cells were thermally stressed at 200°C under an inert atmosphere. The cells were periodically cooled down and the Voc and pFF were measured using a Sinton Suns-Voc tester. Due to time constraints, after the 70 time point, I-V measurements were halted until the final measurement at 1000 hrs. By the end of the test, 28 of the 53 Cu metallized cells had a Voc within 10% of their initial values, while only 12 of these cells had both the Voc and pFF within the targeted range. All the cells printed with paste P2-5 failed early on and were removed after 70 hrs of testing. The I-V curves clearly show that failure was due to the Cu’s diffusion into the junction, resulting in a rapid and significant drop in pFF and Voc after only 1 hr of testing (Figure 15). Sections printed with P2-4 and P2-6 also showed a decrease in pFF over the 1000 hr test period; however, it occurred at much slower rate. The final I-V curves and EL images indicated that rather than the deterioration being purely due to Cu diffusion, increased resistance of the contacts appears to be the root cause; with the decrease in pFF, coinciding with an increase in series resistance (RS), but not with a large drop in Voc (Figure 15). Although only ~50% of samples met the target performance of a Voc and pFF ≥80% of their initial values, the results were used to adjust the paste formulation (F10 to F12, Table 1) for the second round of accelerated aging tests runs for GNG-1B. P2-4 P2-6 P2-5 Normalized pFF after 1000 hrs at 200°C Normalized pFF after 1000 hrs at 200°C Normalized pFF after 70 hrs at 200°C = 0.911 = 0.847 = 0.669 Normalized Voc (Suns-Voc) after 1000 hrs Normalized Voc (Suns-Voc) after 1000 hrs Normalized Voc (Suns-Voc) after 70 hrs at at 200°C = 0.995 at 200°C = 1.012 200°C = 0.538 Figure 15. Examples of the I-V curves of 16 cm2 cells printed with P2-4, P2-6 and P2-5, thermally stressed at 200°C. The normalized Voc and pFF values at the end of the 200°C tests cells are shown. Page 22 of 50 Subtask 2.2: Copper Diffusion at the Interface Due to issues producing uniformly fired cells at UNCC, SIMS of the interface was initially delayed and primary investigation into Cu diffusion was carried out through electrical characterization. UNCC used the ION-TOF Time of Flight (TOF) SIMS at North Carolina State University to evaluate if the Cu had diffused into the bulk of the wafer (Figure 16). The cells were PERC cells with fill factor’s (FF) < 70%, due to underfired region at the edges. Consequently, samples were taken from the center of the cell, where EL imaging showed good contact had been achieved. The Cu contacts were etched off using hydrofluoric (HF) acid. Measurements were taken by sputtering from the front and back of the interface. Cu was detected in the glass region, but not in the bulk. From the frontside measurement Cu showed a long tail into the Si substrate due to the residue from etching. From the back measurement, Cu was observed in the SiNx anti-reflective coating region, but not in the bulk. Residual HF acid was also observed (not shown), which can also potentially increase the Cu ion yield. Although it was not possible with this project, the next steps would be to repeat the measurement without etching off the Cu, as well as measuring thermally stressed cells. (a) (b) Figure 16. (a) TOF-SIMS analysis from the front side of cell underneath the Cu gridline. (b) TOF-SIMS analysis from the back side of Cu finger underneath the Cu gridline after gridline etched in HF. The bulk of the Si substrate was first removed using a FIB microscope. SIMS conditions: Sputtering rate =1 nm/s, Beam = 10 KeV Cs, 20 nA, Sputtering area = 120 µm × 120 µm. Subtask 2.3: Materials Analysis at the Interface Milestone 1.9. Contact Resistance In addition to SIMS analysis, UNCC also carried out Scanning Electron Microscopy (SEM) and Energy Dispersive X-ray Analysis of the interface. The results showed incomplete etching of the SiNx coating, which could account for the relatively higher RS of the cells (results not shown). However, due to the small regions studied no definitive correlation was able to be made to the measured RC values. GIT, in addition to firing cells, were able to measure the RC of 16 cm2 Cu metallized cells prepared for GNG-1A. Cells with FF’s less than and greater than 0.70 were analyzed using the Transmission Line Method (TLM). TLM is a destructive technique as it cuts a small strip (~1 cm wide) from the cell containing only fingers. The resistance between the fingers was measured, with increasing spacing between the two fingers. Resistance vs. distance, was plotted and the slope was used to calculate the RC and contact resistivity (ρc). These values were compared to commercially produced Ag metallized cells. In the Page 23 of 50 low FF group, FF was predominated by a high RC due to non-uniform firing, while in the high FF group, the grid resistance dominated. RC’s below 1 Ω were achieved (Examples shown in Table 2 and Figure 17). Table 2. PV performance of a 16 cm2 Cu and Ag metallized PERC with SE cells from GNG-1A. Cells were measured at GIT. Voc = Open Circuit Voltage, Jsc = Short Circuit Current Density, FF = Fill Factor, RS = Series Resistance, RSH = Shunt Resistance, Rf = Lateral Resistance of a Single Finger, RC = Contact Resistance, ρC = Contact Resistivity. Group (Cell ID) Voc Jsc FF η RS RSH Rf Rc ρc (V) (mAcm-2) (%) (%) (Ωcm2) (Ωcm2) (Ω/cm) (Ω) (mΩcm2) Cu Low FF (Cu1-9) 0.657 38.2 67.4 15.4 3.1 3012 2.93 2.06 22.21 Cu High FF (Cu 47-1) 0.657 38.9 75.1 19.2 0.68 3024 2.91 0.52 1.50 Ag Ref (Ag-9) 0.672 40.4 78.5 21.3 0.64 14399 1.98 0.70 5.04 Ag-9 Cu 1-9 Cu 47-11 Figure 17. Electroluminescence (EL) images of the 16 cm2 PERC cells with SE commercially metallized with silver (Ag- 9) and metallized with copper (Cu 1-9 & Cu 47-11) at BTF. PV performance is shown in Table 5. In addition to building a TLM system at BTF to study the impact of formulation and processing conditions on the interface. Measurements were also taken at other external facilities on commercially available RC tools to verify the values. In addition, to the technique being a destructive measurement, it also requires hardware changes as the fingers become smaller and the pitch/spacing narrows. Go/No-Go Milestone: GNG-1B. Cell Durability A second round of dry-heat testing included one hundred and fifty-four 16 cm2 sections cut from M6 sized PERC cells with a selective emitter. The test focused on 3 pastes (F10, F11, F12 Table 1), 2 screen openings and 2 belt speeds. Screens with 40 μm and 65 μm openings were used to print fingers with cross sectional areas ranging from 190 – 320 μm2 and 435 – 570 μm2, respectively. Twelve sections per set were thermally stressed at 200°C under an inert atmosphere. The test included commercially produced Ag reference cells cut into 16 cm2 sections as a control group. The cells were periodically cooled down, and the Suns-Voc was measured at the 0, 1, 10, 30, 70, 110, 150, 200, 250, 300, 400, 500, 600, 700, 750, 850 and 1000 hr time points. IV curves under dark and illuminated conditions were measured at the 0, 250, 500, 700 and 1000 hr time points, while EL images were taken after 0, 500 and 1000 hrs. At around the 600 hr time point, the oven suffered a power failure, and the samples had to be transferred to an older oven, where a loss on vacuum was observed for ~2.5 hrs. Unlike the previous round of testing, no catastrophic loss in Voc or pFF was observed by any of the samples during the first few hours of testing (Figure 18 and Figure 19). Etching depth is impacted by the paste volume, proportion of frit (Frit 1) and firing temperature. The proportion of frit in the paste had the largest impact on durability, followed by the finger dimension. With increasing amounts of Frit 1, a decrease to the pFF and Voc was observed (Paste F10). In addition to the increasing recombination, the Page 24 of 50 RS was observed to increase over time and EL images showed the images getting darker as the tests progressed (not shown). While some of the RS increases appear to be localized to individual fingers, regions which had a high RS to begin with (due to non- uniform firing), appeared to increase in RS more rapidly. Eighty-seven of the Cu metallized sections ended with both Voc and pFF remaining within 10% of the cell’s initial value after 1000 hrs at 200°C. Most of these sections were printed using pastes F11 and F12. (a) (b) (c) (e) (f) (g) Figure 18. The average effect of accelerated aging 16 cm2 PERC cells at 200°C on the open circuit voltage, Voc [Suns- Voc]. The cells were printed with (a, e) F11, (b, f) F12 and (c, g) F10 using screens with either (a, b, c) 65 μm or (e, f, g) 40 μm openings. The Voc were normalized to their initial values. The error bars reflect the standard deviation. (a) (b) (c) (e) (f) (g) Figure 19. The average effect of accelerated aging 16 cm2 PERC cells at 200°C on the pseudo fill factor, pFF [Suns- Voc]. The cells were printed with (a, e) F11, (b, f) F12 and (c, g) F10 using screens with either (a, b, c) 65 μm or (e, f, g) 40 μm openings. The pFF were normalized to their initial values. The error bars reflect the standard deviation. Page 25 of 50 Task 3: Cell Efficiency Measurement Milestone Performance Assessment Tool / Method of # Metric Measuring Success Value GNG-1A Relative Cell Efficiency relative to the Grouped T-test. Average, one efficiency silver (Ag) metallized cells. standard deviation of the 50 cells Target efficiency ≥ 90% of producing efficiencies ≥ 90% of the the efficiency of the same cell efficiency of the same cell structure structure metallized with metallized with Ag. Standard deviation silver. Standard deviation ± ± 1% (absolute efficiency). 1% (absolute efficiency). (Target is dependent on the Baseline for the Ag metallized cells of efficiency of the Ag the same structure ≥ 20%. metallized cell coming directly from the manufacturer). EOP-B Efficiency Distribution Curve Distribution plots (histograms) for the Deviation Efficiency relative to the efficiency of 500 cells printed and fired silver (Ag) metallized cells. with the Cu paste. Target efficiency ≥ 95% of A subset of ≥ 50 cells for Ag and Cu, the efficiency of the same cell respectively, will be characterized via structure metallized with actual I-V measurements per IEC silver. Standard deviation ± 60904-1. 1% for Suns-Voc and ± 0.5% for actual I-V measurements. Paired t-test for Suns-Voc vs. actual I-V (absolute efficiency). measurement for each of the 2 groups. Go/No-Go Milestone: GNG-1A. Relative Cell Efficiency The biggest challenge in achieving the targeted cell η has been the firing of full-sized (≥ M4 (≥ 261 cm2)) cells. Two furnaces at external facilities were tested. The primary furnace was at the subrecipient’s facility at UNCC, and the second was at GIT. Both furnaces were manufactured by the same manufacturer but used different types of conveyor belts. The temperature variation within both furnaces was unknown, consequently many of the cells processed in these furnaces struggled to produce repeatable results, and often resulting in a large spread in performance, with only a few cells per firing set obtaining a high η (Figure 20).  Mass of Paste ~0.58 mg/cm2  Mass of Paste ~0.41 mg/cm2  η = 19.4% (FF~77%)  η = 20.2 % (FF~73%)  M6 PERC with SE  M6 PERC with SE  Rear Al contact fired by Manufacturer  Rear Al contact fired by UNCC  Front Cu contact fired by UNCC  Front Cu contact fired by GIT  IV at UNCC = 9 Bus Bars  IV at UNCC = 9 Bus Bars  EL at BTF  EL at UNCC March 2022 January 2023 Figure 20. Examples of the highest efficiency PERC cells with sizes ≥ 261 cm2 achieved over the first budget period. PERC cells were printed for GNG-1A (Figure 13) and split into two groups. One group was sent to GIT where they were fired at two peak wafer temperatures (610 & 630°C). The remaining group were scribed and fired at BTF. GIT measured the cells by contacting 5 of the 9 bus bars on the cell. Two distinct η distributions correlating to the two firing temperatures were observed (not shown). The cells were then sent to UNCC where they Page 26 of 50 were measured using their 9-bus bar IV system. It was expected that the Rs would be lower, as the impact of grid resistance was reduced. However, a loss of the two distinct η distributions occurred. Further analysis of the results and additional measurements carried out at BTF by cutting the cells into 16 cm2 sections, implied that there was a measurement issue (most likely between the probes and the bus bars) that was impacting the results (not shown). None of these cells met the relative cell η target of 90%. The second group of cells printed for GNG-1A were processed in parallel at BTF. The remaining cells were laser scribed into one hundred and thirteen 16 cm2 sections and fired in BTF’s three zone furnace (Figure 9). The stage was motorized to produce a similar temperature profile to those obtained in the IR belt furnaces. The cell η was compared to the average η of twelve 16 cm2 sections cut from commercially produced Ag metallized cells. Figure 21 shows the relative η of the Cu metallized cells, with the average relative η being ~88%. To understand the η distribution, the cells with the lowest and highest performance were compared to the Ag samples. EL images showed that cells with lower performances had a darker region to one side of the cell, causing increased Rs in this area (Figure 17). This regional variation appears to be a result of the cell moving on the stage when fired, and the thicker wire creating a heat sink. Removing data for cells whose EL images showed that they moved during firing, resulted in the average relative η shifting to ~90%; thus meeting GNG-1C’s target. Figure 21. Histogram showing the relative efficiency of one hundred and thirteen 16 cm2 sections cut from PERC with SE cells metallized with Cu. The efficiencies were compared to the average efficiency of twelve 16 cm2 sections cut from PERC with SE cells metallized with silver by the manufacturer of the blue cells. End of Project Milestone: EOP-B. Efficiency Deviation Issues with older furnaces at R&D sites, as well as concerns about potential Cu contamination of processing equipment at potential customer sites, prevented this milestone being carried out at a third party. Therefore, this milestone was aided by BTF’s 2024 investment round which allowed BTF to purchase a Sinton FCT-650 flash tester to measure the cells, an 8-zone IR belt furnace for firing and a LECO tool to aid in improving the yield and performance. Over five hundred M6 sized PERC cells with a selective emitter and rear aluminum contact pre-fired were printed with BTF’s Cu paste using a knotless 0°, 480 mesh screen with 25 µm opening. This resulted in approximately 0.31 mg/cm2 of paste being used to printed the fingers (Figure 14) The cells were printed with the same number of fingers and bus bars as the Ag metallized cells. Figure 23 show the Page 27 of 50 histogram plots of the PV performance for 525 cells fired in this study. The results were compared to 50 commercially produced Ag metallized PERC cells. The Voc for all the cells had a relative performance >95%, while the pFF for all 525 cells had a relative performance ≥ 96%. Since the Sinton flash tester simultaneously measured both IV and Suns-Voc curves for all cells, no paired t-test needed to be carried out. Both Jsc and FF histograms for the group skewed right, with a long “tail” (not shown). This observation was mirrored by the RS which skewed left; with 198 cells having a Rs > 1.2 Ωcm2. This increase in RS appears to be the largest reason for only 282 of the 525 cells having a relative η ≥ 90% compared to the Ag metallized cells. None of the cells achieved the target relative η ≥ 95%. FF loss analysis was performed on the results and compared to 50 commercially produced silver (Ag) metallized PERC cells. xvii The FFJ01,limit (FF excluding RS, RSH and J02), for the Cu metallized cells was very close (~99.7%) to the Ag metallized cells. The analysis showed that at FF’s < 76%abs. both RS and J02 losses were a contributing factor to the lower FF; indicating that the contact resistance had increased significantly. However, 368 cells had a FF ≥ 76%abs., and although the ΔFFJ02, remained higher than the commercially produced cells, the losses due to RS of the Cu metallized cells remained the largest reason for the relative gap in performance. The, EL images also highlighted that the firing step was a contributing factor to the lower performance (Figure 22) due to the cells experiencing an uneven exposure to heat; with the central region remaining cooler. BTF purchased a new belt with a more open weave using thinner wires and elevated supports; thus, reducing the thermal mass by ~50%. The new design results in the cell reaching a higher temperature, with slower heating and cooling rates. Therefore, additional firing studies need to be carried out prior to the next η distribution study. However, this falls outside the project’s timeline. η = 21.4%, η = 20.5% η = 19.0% η = 17.1% η = 15.1% Figure 22. Electroluminescence, EL images of selected cells taken from the 525 cell firing study, with respect to their efficiency, η. Page 28 of 50 (a) (b) (c) (d) Figure 23. Histograms showing the distribution in PV performance (a) Efficiency, η (b) Open Circuit Voltage, Voc, (c) Pseudo Fill Factor, pFF and (d) Series Resistance, RS for 525 M6 sized PERC cells metallized with copper paste, fired in BTF’s furnace and LECO processed. Page 29 of 50 Task 4: Stakeholder Engagement, Cost Model Development, and Supply Chain Analysis Milestone Performance Assessment Tool / Method of # Metric Measuring Success Value 1.1 Ordering Ordering of critical supplies Items to be ordered to meet the Supplies milestones in the first 3 months include Including but not limited to: and are not limited to: 1) ≥ 20 kg Cu powder • Copper (Cu) powders 2) ≥ 2 screens • Screen-printing screens 3) ≥ 2000 silicon cell • Silicon solar cells precursors. precursors from per • Silver metallized cells (with an technology/ type. Cell efficiency ≥ 20%) size ≥ M4 (161.7 mm). 4) ≥ 200 silver metallized cells from the same lot as the cell precursors. 1.2 Draft Cost Model 1 Draft Model in Excel Cost model contains: • Defined sensitivity range (contains worst case scenario for material and fixed costs) – Minimum of 3 scenarios. • All materials in paste. • Scaled up to a minimum of 1000 kg. 1.5 Contact ≥ 2 stakeholders identified Begin identifying and contacting Stakeholders and contacted stakeholders defined in GNG-1E. GNG-1D Preliminary Cost < 10% the cost of Ag paste Excel sheet containing the Model comprehensive cost model. Includes sensitivity analysis with ranges agreed upon in milestone 1.2 GNG-1E Letter of Intent ≥ 4 Letters or Count. A minimum of four letters or communication(s) communications from different stakeholders, including but not limited to: (1) Cell manufacturer stating the minimum performance metrics/requirements of the Cu paste to test the product on their R&D line. (2) PV module manufacturer stating the requirements the paste must meet to be adopted into a pilot line test. (3) Paste manufacturer stating interest in joint development or financial partner (investor, bank) interested in providing follow on funding with or without milestones (4) End user (utility, developer; installer; Independent Engineer, Engineering, Procurement, and Construction firm) stating that the cost reduction would improve sales. Page 30 of 50 Subtask 4.1: Ordering of Supplies Milestone 1.1. Ordering Supplies Cu powders, chemicals, screens and Si solar cells were periodically ordered during the project. Where possible, materials from US suppliers were sourced first. The Cu powders account for > 80 wt% of the product and were sourced from US suppliers. To develop the pastes, Si solar cells with and without Ag metallization needed to be purchased. The majority of PV cell manufacturing is in Asia, consequently most cells used in this project were sourced from a company in Taiwan. At the start of the project, PERC cells had the largest market share and were easier to source. Although, cell sizes over the last few years have increased from G1 (252 cm2) up to M12 (441 cm2), BTF’s supplier was able to provide a steady supply of M6 (276 cm2) sized PERC cells for the project. However, the cell manufacturer did transition away from fully printing the rear side of the cells with Al paste, to printing the rear Al contact as fingers and bus bars, making the cells bi-facial. Both homogeneous (HE) and selective emitter (SE) cells were tested over the course of the project (Table 3). The same cell structures metallized with Ag were also purchased directly from the manufacturer for relative comparison of the performance. Screen printing screens were ordered from a US based company, but to get to finer printed fingers, the supplier had to be switched to a supplier in Taiwan that is a supplier of screens for industrial PV manufacturing lines. Table 3. Cell structures discussed. Set B cells were used in GNG-1A, B & C. Set E cells were used in EOP- A & B. Set C & D cells were used in Task 6.0. Mfr.4 (TW) - B Mfr.4 (TW) - C Mfr.4 (TW) - D Mfr.4 (TW) - E Cell Type PERC with SE(B) PERC with HE(C) PERC with HE(D) PERC with SE(E) Wafer Size M6 (276 cm2) M6 (276 cm2) M6 (276 cm2) M6 (276 cm2) Emitter n-type n-type n-type n-type Bulk Wafer p-type p-type p-type p-type Sheet Resistance of 150 - 160 100 - 110 100 - 110 150 - 160 Emitter (Ω/□) Sheet Resistance of Selective Emitter 110 - 120 None None 110 - 120 Regions (Ω/□) Thickness of the anti- 70 – 80 nm 70 – 80 nm 70 – 80 nm 70 – 80 nm reflective Coating Wafer Thickness (μm) 160 (+20,-10) 160 (+20,-10) 160 (+20,-10) 160 (+20,-10) Resistivity (Ωcm) 0.4-1.1 0.4-1.1 0.4-1.1 0.4-1.1 Rear of the cells Full Aluminum Full Aluminum Bi-facial Pattern Bi-facial Pattern Coverage with Coverage with Localized Laser Localized Laser Localized Laser Localized Laser Openings (LCO’s) Openings (LCO’s) Openings (LCO’s) Openings (LCO’s) Number of Front Side 9 9 9 9 Bus Bars Page 31 of 50 Subtask 4.2: Stakeholder Engagement Milestone 1.5. Contact Stakeholders Cu powder suppliers, cell manufacturers and academic institutions were contacted over the course of the project. In some cases, these stakeholders provided additional cell structures, materials and cell characterization. Go/No-Go Milestone: GNG-1E. Letter of Intent Although no formal letters were issued by stakeholders. BTF actively engaged with the following research institutions and manufacturers. Paste and cells were exchanged to gain feedback from the relevant stakeholders (Table 4, Table 5). In lieu of a letter, this information was shared with the DOE. Table 4. Research and Development Facilities contacted during Budget Period 1. R&D Location Details UNCC USA Firing, measurement, analytical analysis. (Subrecipient). GIT USA Firing, measurement, analytical analysis. Started a project using BTF’s Cu paste on TOPCon cells. (Subrecipient). NREL USA Measurements. Collaborated on a DuraMat project in April 2023. Fraunhofer DE Sent 1 kg of paste to print on TOPCon. Concerns about Cu contamination, so set up a separate printer. Cells sent to GIT for firing. R1 DE PL tests to evaluate “floating” fingers to be used in combination with a Ag paste. Discussed the funding of additional tasks and whether the Cu pastes can be used in their facilities (i.e. contamination concerns). R2 CH Planned to send paste at the end of July for TOPCon printing. – Were willing to print and fire at their facilities. R3 SG Printing and firing discussions. Table 5. Module and Cell Manufacturers contacted during budget period 1. Company Location Details ID A CA/USA MWT – 1000 cells – Interested in Cu’s interaction with encapsulant. The cells had issues with initial pre-firing of the rear Al metallization. Cells had to be fired again prior to Cu printing. Cells had a large spread in performance. 10 modules were built with the cells. Modules were placed into damp heat to evaluate encapsulation interaction. Change of personnel at company A. B CN PERC – 1000 cells – Rejected by customs due lack of documentation for WRO. – Recently reengaged at IEEE PVSC. Company is planning a US footprint. C CN PERC – 500 cells – Want to use a floating Ag bus bar with Cu fingers and co-fire with Al paste. Concerns that frits in the bus bars would interact with the Cu paste. Interested in TOPCon, but could not supply TOPCon cells due to IP concerns. D Asia/US PERC/shingled – 100 cells were originally sent for evaluation. 1 kg of paste was sent to the company, but at last minute they were unwilling to use it in their facilities. Sent additional 1200 cells to BTF. E Asia/US TOPCon – 150 cells – Cu printed and fired on rear of cells. Cells were sent back for feedback and further processing of the front contact. F CN TOPCon – 600 cells – Evaluation of the Cu on the back and front. On hold due to issues with firing M10 (331 cm2) cells. Page 32 of 50 Subtask 4.3: Cost Model Development Milestone 1.2. Draft Cost Model Production of BTF’s CuBert™ paste was broken into materials, labor, and equipment. For the analysis, the material was evaluated at varying scales up to 10,000 metric tons/year, but the labor and equipment remained as a per kilogram cost. This assumed that operation (OPEX) and capital (CAPEX) expenses were significant. The model assumed: (1) the paste was predominately Cu, (2) a power law could be used to estimate the materials cost based on the cost at low volumes, (3) labor rates were based on Bureau of Labor Statistics for Louisville, Kentucky and (4) a 10-year equipment depreciation rate with 4% maintenance cost per year. The labor and equipment costs were built around the production of a 600 kg batch of paste. Monte Carlo sensitivity analysis on the materials and labor was carried out; yielding a maximum price, minimum price, and the standard deviation. To estimate the costs of Cu paste it was first necessary to place a bound around the amount of paste needed. This was done using the PV CAGR and net zero policies to determine the amount of paste required by industry. The penetration of the Cu paste into the industry was modeled using a fast (10 years to reach 80%) and slow (20 years to reach 40%) s-curve adoption approximations. Three of the scenarios easily approach 1,000 tons of paste per year. Based on this analysis the materials costs of the Cu paste was below 10% the cost of just the Ag in the current metallization pastes at quantities greater than 100 tons. The cost of the Ag paste is a very closely guarded secret within the industry. At the time of developing the model, the cost of Ag pastes ranged between $1100 and $1400 per kg, with a median of $1250 per kg. xviii Using a full factory of 5 teams, it is expected that 150 tons of Cu paste can be manufactured per year. An Excel sheet and pdf document describing the model were submitted. Go/No-Go Milestone: GNG-1D. Preliminary Cost Model A Senior Energy Technologies, Economics and Policy Analyst at NREL, reviewed the documents. The reviewer shared NREL’s “…industry best practices for detailed cost modeling including solar photovoltaic (PV) materials manufacturing. The NREL methods and approaches are the result of more than 12 years of work with solar industry materials and equipment suppliers and integrated manufacturers.” Training in methods shown in Figure 24 were used to produce the preliminary cost model for GNG-1D. With the updated model, the cost of Cu paste remained below 10% of the cost of Ag paste. The preliminary cost model was submitted to both parties for verification. Figure 24. Methods and Calculation approaches provided by NREL and utilized by Bert Thin Films. Page 33 of 50 Task 5: Mini Modules and/or Module Fabrication and Lifetime Measurement Milestone Performance Assessment Tool / Method of # Metric Measuring Success Value 2.1 Damp Heat Test ≥10 cells shipped to partner First set of tests on mini-modules site. started using cells from GNG-1A. ≥ 3 mini-modules fabricated and placed in test chamber 2.3 Improvement in Voc and Pseudo fill factor Data tracking pFF and Voc for the cells. Cell Durability (pFF) remains within 5% of The performance of the cells is cells initial value after 1000 measured at 0, 1, 10, 30, 70, 110, 150, hrs at 200°C. 200, 300, 400, 500, 600, 700, 850, and 1000 hour time points. The number of 10-20 cells cells is dependent on the capacity of the oven. Initial values for the cells Current-voltage (I-V) measurements of tested with the Cu the cells at the start (0 hrs.) and end metallization is ± 2% of the (1000 hrs.) of the durability tests will be Voc and pFF values for the performed. cells metallized with Ag. Minimum number of cells to be measured: 5 – 10 cells metallized Cu paste 5 – 10 cells metallized Ag paste. 2.4 Mini Module 1 CAD drawing CAD drawing shows the proposed Architecture 1 Bill of Materials module layout. The bill of materials includes the specifications for all the materials used to construct the mini modules (e.g. glass type and thickness, backsheet, etc.). 2.5 Damp Heat Test < 5% reduction in Pmax Degradation of Pmax < 5% of the initial on ≥5 2-cell Mini- Pmax value at the start of the test per Modules IEC 61215 protocol (i.e. incl. pre- and post-stabilization) The mini modules contain two full size cells connected in series. 2.6 Partial IEC 8 mini modules containing 16 Modules fabricated and testing 61215 cells fabricated and sent out started. certification for testing at ISO 17025 - 4 modules for thermal cycling. accredited CTL - 4 modules for damp heat test. Cells ≥ M4 EOP-A Module Lifetime Partial Certification from an Degradation of Pmax after each test < ISO 17025 CTL. Any failure 5% of the initial Pmax value at the start should not be attributed to of each test, and not more than 8% the metallization. across each testing sequence. - 4 mini modules for thermal cycling. The results from 8 mini - 4 mini modules for damp heat test. modules containing 16 cells each. Page 34 of 50 Subtask 5.1: Mini Modules and/or Module Fabrication and Lifetime Measurement Milestone 2.1. Damp Heat Test Five prototypes were constructed to see if the module architecture and process settings were viable. A supplier of edge sealants for PV modules allowed BTF access to their laminator. The first round of mini-modules used mono-facial PERC cells with a SE and pre-fired rear Al contact. The cells were fired as full M6 sized cells with the bus bars parallel to the furnace entrance and cut in half after the front metallization had been fired. Table 6 shows the materials used to construct these modules. Due to limited access to equipment, the 5 mini-modules were tested continuously for 1000 hrs. Since BTF had not yet received its Sinton FCT-650 flash tester, the 0 hr and 1000 hr measurements were taken at two different locations. Consequently, measurements were taken with slightly different setups for the system. With no other modules available the system was calibrated to Module E at both timepoints. The results demonstrated that the module bill of materials (BOM) was viable, but further improvements to the cells processing were required (not shown). Milestone 2.5. Damp Heat Tests on ≥5 2-cell Mini Modules The same external facility was used to produce an additional set of fifteen mini- modules. With limited access to the laminator, the procedure was also adjusted from module to module to try to reduce the number of voids present in the encapsulant. Lower laminating temperatures created more voids, while higher temperatures reduced the number of voids, but at some settings delamination of the encapsulant from the glass was observed. Silicone frames were also used around the modules to prevent glass “pinching” at the edges and improve the wetting of the edge sealant on the glass surface. The second-round of modules were constructed with both the pre-fired and co-fired cells. All fifteen mini-modules contained two half-M6 sized PERC cells connected in series. Although the η targets had yet to be realized on full sized (≥ M4) cells, cells were selected for the modules, to identify the PV parameters that would indicate whether the Cu contacts would fail or succeed during the accelerated aging tests. As a result, a range of paste formulations and firing groups were tested. The selected cells were measured at the University of North Carolina, UNCC and had pseudo fill factors, pFF >78% (i.e. pFF’s > 94% relative to Ag). Modules containing commercially produced Ag metallized cells were included for all the cell structures tested. The damp heat tests were carried out at BTF. Due to the number of variables, modules were studied first by cell group and then on a case-by-case basis. Modules I, J and K used pre-fired cells, while modules L, M, N, O, P, Q, R and S used co-fired cells. The performance was monitored for ~1000 hrs, with measurement intervals gradually increasing until the modules were continually exposed to damp heat (85°C/85%RH) for the final 500 hrs. Figure 25 shows the IV and Suns-Voc curves for four of these modules (N, Q, R, S) showing different types of failures. None of the modules displayed a drop in Voc below 90% of its starting value. However, modules Q and N were the only modules that resulted in a relative Voc drop of more than 95%. In the case of module Q, EL imaging showed that one of the cells was failing faster than the other, while for module N, both cells appeared to decrease at a similar rate. After 1000 hrs modules I, J and K (using the pre-fired cells) had relative Voc’s of 99, 99, and 98%, respectively, while the remaining modules containing co-fired cells, L, M, O, P, R and S had relative Voc’s of 100, 100, 97, 98, 97, 98 and 97%, respectively. For the majority of Page 35 of 50 the modules, decreases to the Pmax, appeared to be driven by an increasing RS, which occurred during the first 100 hrs of testing. EL images showed that many of the regions displaying high RS corresponded to areas of the cell that were observed to have experienced a higher temperature during firing. A third set of modules were produced by a third-party company that specializes in building and testing prototype modules. The previous two sets of mini-modules had previously been encapsulated and interconnected using a variation of the Smart Wire Connection Technology (SWCT) for the front contacts and manual soldering for the rear contacts (Table 6). The third set of mini modules were produced using cells fired at GIT. This time both the front and rear contacts were manually soldered for both the Cu and Ag modules. The interconnected cells were shipped along with the glass, encapsulant and edge sealant to the company. The company recommended switching the older thermoplastic polyolefin from 2016 for a newer cross-linking polyolefin encapsulant (Table 7). The resulting modules did not contain any bubbles. Due to the need to accelerate product development, BTF purchased a secondhand laminator. With the addition of perforated tape to aid air evacuation, bubble free minimodules were produced at BTF. This resulted in modules for EOP-A being constructed in-house before being shipped to NREL for testing. 0 hrs 0 hrs Module N Module Q 500 hrs 500 hrs 1007 hrs 1007 hrs 0 hrs 0 hrs Module S Module M 500 hrs 500 hrs 1007 hrs 1007 hrs Figure 25. Examples of the different types of failures observed during the damp heat tests. Modules N, Q, S and M contained co-fired cells. Cells were fired at GIT as a pre-cut half-cell and used ~0.48 – 0.52 mg/cm2 of paste. Subtask 5.2: Module Architecture Optimization Milestone 2.4. Mini Module Architecture A CAD file detailing the module structure and materials for the first prototypes were submitted (Table 6). Changes were made to the BOM for the modules containing 16 half- M6 sized cells used in the partial IEC testing (Table 7). Page 36 of 50 Table 6. Bill of materials for glass-glass mini-modules. Material Name Description Size Qty per Module Manufacturer 1 Glass Low-iron, tempered glass with 240 mm × 2 Solite textured surface 206 mm, 3.2 mm thick 2 Sealant Tape Desiccated polyisobutylene, PIB 8 mm × 0.74 1.72 m length Quanex sealant tape mm 3 Encapsulant Thermoplastic polyolefin encapsulant 224 mm × 2 sheets DNP Solar sheet 190 mm, ~0.4 mm thick 4 Cells PERC cell laser-scribed in half. 166 mm × 83 2 TW: PERC mm Cells (see Table 2) 5 Tabbing PV ribbon for secondary connections 0.5 mm × 0.2 ~528 mm length Vikocell Ribbons mm 6 Rear Contact Cu ribbons manually soldered (0.50 ± ~2.275 m length Ulbrich of Ribbons Rm = max 270 N/mm²; Rp = max 90 (Nine per cell) Austria GmbH 0.050) mm × N/mm²; A = min 25 % Solder: 15 - 25 µm Peak Sn60Pb40 (0.200 ± 0.005) mm 7 Front Cu wires soldered during lamination (0.250 rd ± ~4.6 m length Ulbrich of Contact process (Smart Wire Connection (Seventeen Austria GmbH 0.010) mm Wires Technology, SWCT). spaced ~0.9 cm Rm = 220 - 290 N/mm²; Rp = 100 - 160 intervals) N/mm²; A = 10 - 40 % Solder: 5 - 7 µm Peak Sn42Bi57Ag1 Table 7. Bill of materials for final 16 half-cell glass-glass modules. Material Name Description Size Qty per Module Manufacturer 1 Glass Low-iron, tempered glass with textured 735 mm × 435 2 Solite surface mm, 3.2 mm thick 2 Sealant Tape SolarGain Edge Sealant - desiccated 8 mm × 0.74 mm 4.68 m length Quanex polyisobutylene, PIB sealant tape 3 Encapsulant SOLAR ASCE™ Cross linking Polyolefin 715 mm × 415 2 sheets Mitsui encapsulant mm, ~0.45 mm Chemicals thick Tohcello, Inc. 4 Cells PERC cell laser-scribed in half. 166 mm × 83 16 TW: PERC mm Cells (see Table 1) 5 Bus Ribbons PV tinned copper ribbon, manually 5 mm × 0.2 mm ~3.41 m length Vikocell soldered 6 Tabbing Cu ribbons manually soldered (0.300 rd ± ~25.38 m Ulbrich of Wires 7746-RD0, 0.005) mm length Austria GmbH Rm = max 290 N/mm²; Rp = max 100 N/mm²; A = min 25 % Solder: 5 - 15 µm Peak Sn60Pb40 Subtask 5.3: Increasing Cell Durability Milestone 2.3. Improvement in Cell Durability Dry-heat accelerated aging tests continued on the bare cells in parallel to the damp heat testing. Cells (6 print groups and 7 firing conditions) fired in GIT’s furnace were tested. Over two thirds of the cells met the minimum relative pFF and Voc of ≥ 95% of its initial value after 1000 hrs (not shown). For these cells, increased RS was a contributing factor for η decreases. However, the η of this group of cells had yet to achieve the target η of EOP-B. Previous dry heat tests showed that when Cu diffusion into the bulk is the Page 37 of 50 predominant reason for reduced cell lifetime, it occurs rapidly, therefore shorter tests can quickly eliminate formulation and processing changes that make this failure pathway to occur. Consequently, 1000 hour dry-heat testing was paused in favor of shorter tests. Figure 26 shows examples of the IV curves of M6 sized PERC cells fired in BTF’s 8-zone IR belt furnace before and after being thermally stressed at 200°C for 50 hrs. A similar degradation pathway was observed, regardless of whether the contacts had been produced by firing or through a combination of the firing and LECO process; with a lower pFF indicating diffusion into the bulk and increased recombination currents. However, by utilizing the LECO process it was possible to produce shallower contacts, resulting in little to no change to the pFF after thermal stressing. Further optimization is required to make reproducible contacts. Fired Contacts Fired & LECO Processed Contacts Fired & LECO Processed Contacts Relative pFF after 50 hrs at 200 °C Relative pFF after 50 hrs at 200 °C Relative pFF after 50 hrs at 200 °C = 95.2% = 93.1% = 99.1% Relative Voc after 50 hrs at 200 °C Relative Voc after 50 hrs at 200 °C Relative Voc after 50 hrs at 200 °C = 99.7% = 99.6% = 100.0% (a) (b) (c) Figure 26. Examples of the IV and Suns-Voc curves for M6 sized PERC cells with a pre-fired Al contact thermally stressed at 200°C for 50 hrs. The graphs show examples of (a) fired and (b, c) fired and LECO processed Cu contacts with differing degrees of durability. Subtask 5.4: Module Construction and Testing Milestone 2.6. Partial IEC 61215 Certification After firing and testing, PERC cells from EOP-B were laser scribed in half. The front and rear contacts of the cells were then manually soldered to create four strings containing four half cells connected in series. The four strings were connected in parallel. No by-pass diodes or junction modules were used. The modules were constructed at BTF using the BOM shown in Table 7. Eight modules containing 16 half cells metallized with Cu were produced. In addition, a single module containing commercially made Ag metallized cells was produced as a control. Due to the size of the modules BTF was unable to make any measurements except for EL imaging prior to shipping (not shown). The images did show some damage due to cell cracking, as well as areas of high RS where the metal contacts had been damaged. These defects had not been present after firing and were attributed to the soldering and lamination process. End of Project Milestone: EOP-A. Module Lifetime NREL’s certification lab was selected to run the partial IEC 61215 tests on the 16 half- cell modules as it is an ISO accredited laboratory. At the request of the lab, a module containing a single cell was also produced for Quantum Efficiency measurements (not shown). The PERC cells use p-type Si wafer, which is known to be susceptible to Light Induced Degradation (LID). This degradation mechanism can be minimized by light soaking the modules, and since the IEC 61215 protocol uses an initial stabilization step prior to running the tests, the modules were light soaked outside for 15 kWh/m2. Figure 27 shows examples of the initial IV curves for the Ag metallized module and two of the Page 38 of 50 Cu metallized modules showing the highest and lowest performance. The Ag module had a lower FF and maximum power (Pmax) than expected, indicating further improvements to the interconnection method are needed to lower RS. The Cu metallized modules had a relative performance of between 91 – 88% when compared to the Ag module. Due to budget and time constraints only the tests in the IEC61215 standard that would directly impact the metallization were carried out. This included thermal cycling of four of the modules and damp heat testing of the remaining four. The Ag metallized module was only used to calibrate the flash tester before and after testing. The modules are constructed with materials that have different thermal expansion coefficients. By exposing the modules to extreme temperature swings, the interfaces, in particular the metallization and solder points are mechanically stressed, leading to cracking and increased RS. xix The first four modules were thermally cycled between -40°C and +85°C, for 200 cycles while biased under a voltage of 10 V and 9.2 A. The modules were held at extreme temperatures for 10 mins and each cycle took ~ 4.8 hrs. Figure 28(a-d) shows the impact of the thermal cycling on the IV curves of the modules. Only a small increase in RS was observed, with the modules retaining an average Pmax of 97% of their initial values. The final four modules were exposed to damp heat testing; with the modules being continuously exposed to a temperature and relative humidity of 85°C and 85%RH, for 1000 hrs. The damp heat test typically identifies failures such as corrosion of the metal contacts, as well as interactions between the encapsulant and metal, that can occur as a result of moisture ingress into the module. Figure 28(e-h) shows the IV curves before and after damp heat tests. The results show a significant reduction in performance with an average Pmax of 74% of the modules initial values. Based on dry-heat tests of single cells and damp heat measurements of the mini-modules the degradation appears to be due to increased recombination potentially as a result of Cu diffusion. The average Voc of the modules had decreased to ~97% of their initial values; roughly a 21 mV reduction per half-cell. EL images of the tested modules, showed that some cells on the strings remained relatively bright, indicating that the performance of these cells had not decreased significantly (not shown). Although the modules passed the thermal cycling tests, they did not pass the damp heat tests. However, a route for producing durable PERC modules still exists with further optimization of the paste and processes. (a) (b) (c) Figure 27. Certified IV measurements for the 16 half-M6 sized PERC cell modules constructed for partial IEC 61215 testing at NREL. (a) Module constructed using commercially produced Ag metallized PERC cells as a control group. Examples of the (b) highest (M4) and (c) lowest (M8) performing modules made using copper metallized PERC cells. Page 39 of 50 Thermal Cycling Thermal Cycling Thermal Cycling Thermal Cycling (a) (b) (c) (d) Damp Heat Damp Heat Damp Heat Damp Heat (e) (f) (g) (h) Figure 28. IV curves for 16 half-M6 sized PERC modules: (a-d) M1, M2, M3 and M4 were thermally cycled between - 40 to 80°C for 0 and 200 cycles. The modules were placed under a bias voltage of 10 V and 9.2 A. Modules (e-h) M5, M6, M7 and M8 were exposed to damp heat (85°C/85%RH) for 0 and 1000 hrs. Task 6: Co-Firing Optimization Milestone Performance Assessment Tool / Method of # Metric Measuring Success Value 2.2 Co-Firing Mass of Paste required to Average, standard deviation of 50 achieve ≥ 90% of the cells. efficiency of the same cell structure metallized with silver. Standard deviation ± 1% (absolute efficiency). Maximum amount of paste used < 175 mg ± 10 % on a cell ≥ M4 (161.7 mm) in size. Milestone 2.2. Co-Firing BTF’s Cu paste has a lower firing temperature than Ag metallization pastes, therefore the Al pastes used to print the rear Al contact have been designed to co-fire at the higher temperatures utilized by the Ag pastes. Consequently, the Cu paste was initially developed to work on cells with the rear Al contact pre-fired. Paste F13 in Table 1 shows an example of a paste developed for pre-fired PERC cells. The amount of frit and nanoparticles present in F13 would produce a working device when printed on a cell with a pre-fired Al contact. When the paste was co-fired and the belt speed reduced (i.e. increasing the firing time and peak wafer temperature), the paste significantly over etched; resulting in increased recombination currents and lower Voc’s (Figure 29(a)). By lowering the amount of Frit 1 (the more aggressive) etchant in the paste (F6), etching could be reduced (Figure 29(b)). However, at higher temperatures the contacts were more prone to oxidization resulting in a high RS. The current hypothesis is that the hydrogen released from the SiNx coating during co-firing aids in creating a reducing environment close to the interface, which accelerates the etching rate of the etchants in the Cu paste. In addition, the SiNx becomes denser during firing, requiring more aggressive/higher quantities of etchants when the paste is printed on a pre-fired cell. Page 40 of 50 (a) (b) Figure 29. Examples of the current-voltage (IV) curves for a paste with (a) an average (F13) and (b) relatively low (F6) proportion of Frit 1 (Table 3) with respect to IV curves for a Ag cell. The pastes were co-fired on PERC cells with an HE (100 – 110 Ω/□) in BTF’s 3-zone furnace (Z1 = 550°C, Z2 = 450°C, Z3 = 925°C, B.S. = 225 – 275 in/min). Initial co-firing studies were conducted in BTF’s 3-zone furnace using PERC with a HE emitter and full rear Al paste coverage. Cells were fired at numerous furnace settings with both zone and temperature setpoints being varied. The sheet resistance of these samples was ~2 – 4 times higher than the Al contact (~0.0025 Ω/□) from commercially produced Ag metallized cells. However, it is not only the conductivity of the Al contact that will impact the device’s performance, but the back surface field (BSF) that forms through the narrow local contact openings (LCO) in the rear passivation layer. At the LCO regions, dark lines appear in the contact where the Si diffuses into the Al. The concentration of Si and hence area of the dark regions is dependent on temperature. Bulk Al melts ~660°C, but Si will start to dissolve in Al ~300°C, and solubility will reach a maximum around the Al-Si eutectic point at ~577°C. The narrower the region, the shallower the BSF. A reduction to the BSF of the cell can increase recombination at the rear surface, resulting in a lower Voc. Figure 30 shows the box plots for LCO area’s measured using optical microscope images taken from the center of cells fired by the manufacturer, UNCC, GIT and BTF. The smaller LCO areas indicate that BTF’s furnace, and the settings used to fire the Cu contacts at GIT were significantly lower than the respective Ag co-firing settings used by the manufacturer or the settings used by UNCC to pre-fire the cells. M6 sized PERC cells were also co-fired in GIT’s 6-zone furnace. Figure 31 shows the impact of the firing conditions on the Al sheet resistance. In all cases the left and right edges experienced a higher temperature, while the center was cooler. However, the cooler region was not uniformly distributed, with one side (most likely the edge leading into the furnace) appearing to experience a higher temperature. This variation in conductivity was also observed for the front contact, on both pre- and co-fired cells (not shown). Increasing zone 6’s temperatures aid in improving the conductivity of both the Al and Cu contact but does not completely alleviate the non-uniformity. Not only is conductivity impacted, but the localized BSF is as well. The size and thickness of the wafer as well as the location of the wafer supports on the furnace belt is believed to be a contributing factor to the lack of homogeneity, due to increased bending and warping. M6 cells were also cut in half before being put through the furnace to try to minimize this effect. Page 41 of 50 600,000 500,000 Area of the LCO (μm2) 400,000 (a) Fired by Manufacturer (b) Fired at GIT at 290 300,000 at 236 in/min in/min 200,000 100,000 0 UNCC at 210 in/min BTF at 225 in/min BTF at 275 in/min (c) Fired at BTF at 225 (d) Fired at BTF at 275 Manufacturer at GIT at 290 in/min BTF at 250 in/min in/min in/min 236 in/min Location of Furnace and Belt Speed Figure 30. Area of the Localized Contact Openings (LCO) on the rear of the cells fired in IR belt furnaces by the Manufacturer using Ag co-firing settings, and UNCC using their Al pre-firing settings and GIT using settings for Cu contact. The areas were compared to the samples fired in BTF’s 3 zone furnace (Z1 = 550°C, Z2 = 450°C, Z3 = 925°C) using belt speeds for 225, 250 and 275 in/min. The images show a Ag solder pad and the LCO regions of samples fired by (a) the manufacturer, (b) at GIT, (c) at BTF at 225 in/min and (d) at BTF at 275 in/min. Between 11 – 16 LCO’s towards the center of each sample were measured. F1) VC: Z6 = 760°C, B.P. = 25% F2) VC: Z6 = 780°C, B.P. = 25% F3) VC: Z6 = 800°C, B.P. = 25% F5) VC: Z6 = 820°C, B.P. = 25% F6) VC: Z6 = 840°C, B.P. = 25% F4) VC: Z6 = 800°C, B.P. = 50% F7) VC: Z6 = 790°C, F9) VC & BPTT: Z6 = 745°C, F10) VC & BPTT: Z6 = 805°C, F11) VC & BPTT: Z6 = 790°C, F12) VC & BPTT: Z6 = 840°C, F13) VC & BPTT: Z6 = 790°C, B.P. = 25% B.P. = 25% B.P. = 50% B.P. = 100% B.P. = 100% B.P. = 150% BASELINES: Ag Ref: Al Co-fired by Al Pre-Fired at UNCC Al Pre-Fired at GIT Manufacturer 11 Figure 31. Heat maps showing the sheet resistance of the rear aluminum (Al) contact on a M6 (16.6 cm × 16.6 cm) sized PERC cell with HE(C) (Table 3) with respect to the location on the cell. The cells were co-fired at GIT with changing Zone 6’s (Z6) setpoint temperature and the power (B.P.) going to the bottom lamps. Al sheet resistance was compared to cells pre-fired by UNCC and GIT, as well as co-fired by the cell manufacturer. At the start of this task ~0.80 mg of paste/cm2 of paste was used to print the fingers on the co-fired cells. The larger fingers had the advantage of allowing for a broader firing window. However, the highest η achieved by a co-fired cell was 17.6%. FF loss analysis indicated that RS remained the largest contributor to the lower FF values. A reduction in paste usage was made through formulation and screen changes. However, early attempts at firing the smaller fingers resulted in the cells having η’s below 8% when 0.47 Page 42 of 50 mg2/cm2 of paste was used. To avoid the non-uniformity shown in Figure 31, cells were cut in half prior to firing. This helped prevent one side of the cell from being more fired than the other, but the cut edge was not passivated, which led to a relatively lower RSH. By reducing the temperatures in the burnout zone, and increasing the speed, the time the contacts were exposed to elevated temperatures decreased; thus, minimizing the potential for the smaller fingers to oxidize and producing an η of 17.8% using 0.52 mg/cm2 of paste (Table 8). However, the lower temperatures impacted the formation of the L-BSF and conductivity of the rear Al contact. As the project progressed, the manufacturer of the PERC cells made changes to the rear Al contact by printing Al fingers and bus bars, so that the cells were effectively bi-facial. With the assumption that the wafers were roughly the same thickness and thus weight, the bi-facial cells used approximately two thirds less Al paste than the mono-facial cells (full Al contact). Consequently, changes to the firing settings were necessary since the bifacial cells heated up differently due to the difference in thermal mass and conductivity. However, the green-strength of the Al on the bi-facial cells was significantly lower, leading to the aluminum being easily damaged prior to firing. Further adjustments to the firing process resulted in a maximum η of 18.4% using 0.43 mg/cm2 of paste (Table 8). Co-fired cells from this group were also selected for damp heat testing. However, the mechanical integrity of the Ag solder pads near the center of the cells failed with the pads detaching. This region appeared to be underfired relative to the rest of the cell. To date, the highest co-fired cell η achieved was 19.7% (~89% relative η to Ag on the same structure) using 0.32 mg/cm2 of paste (F18 - Table 1) on an M6 sized PERC cell fired and LECO processed at BTF (Table 8). Performance remains limited by the mismatch in firing temperatures between the front and rear contact. Progress on this task was first delayed by work on the EOP targets and then by development of the pastes for TOPCon. Table 8. PV Performance of highest efficiency cells with respect to mass of paste used to print the fingers. Paste Mass Jsc Voc FF RS RSH pFF Cell ID PERC Firing η (%) (mg/cm2) (mA/cm2) (mV) (%) (Ωcm2) (Ωcm2) (%) C116 M6 at Set C 0.80 36.7 655 73.5 17.6 1.6 5.6×104 81.6 (F26/F11) GIT C257A ½ M6 at Set C 0.52 37.2 653 73.2 17.8 1.4 3.2×103 80.3 (N/F26) GIT D37A Set D ½ M6 at 0.52 37.8 654 71.3 17.6 2.0 2.4×103 81.1 (V/F26) Bifacial GIT D147 Set D M6 at 0.43 37.6 656 74.7 18.4 1.2 4.0×103 80.5 (AN/F36) Bifacial GIT M6 at CP108 Set C BTF & 0.32 38.3 664 77.5 19.7 0.7 6.3×103 80.9 (Set 52) LECO M6 at DL141 Set D BTF & 0.16 37.5 660 76.6 19.0 0.8 2.6×104 80.4 (CPF6) Bifacial LECO M6 at DL58 Set D BTF & 0.13 37.5 661 75.8 18.8 1.0 3.1×104 81.0 (CPF2) Bifacial LECO Page 43 of 50 Paste development was primarily carried out on PERC cells, but over the latter part of the project TOPCon structures started to become more readily available, as the industry switched to higher η technologies. Initially testing was on M10 (331 cm2) cells with the front p-type emitter printed and fired with Ag contacts by the manufacturer. The rear n- type side was printed with the Cu pastes developed for PERC cells. A limited number of cells were fired at GIT and measured at NREL due to limitations with GIT not having the ability to measure an 11 bus bar cells at the time. The cells achieved a maximum η of 19.1%. Photoluminescence imaging showed that the η was limited by the edges of the cells being overfired, due to the lack of supports for this size wafer (not shown). A third- party R&D facility also tested a version of the Cu pastes on TOPCon. They printed the paste on M2 (246 cm2) sized cells and BTF fired the cells at GIT. The cells were then shipped back for LECO processing and measurement. A maximum η of 21.6% was achieved. The pastes developed for co-fired PERC cells were observed to work the best on TOPCon structures, due to the lower aggressiveness of the etchants and the relatively higher firing temperatures. Further development of the pastes and metallization designs on the rear side resulted in η’s >24% being achieved at BTF and third-party sites. Table 9. PV Performance of the highest efficiency TOPCon cell using paste F100 (Table 1) to print the rear fingers. Paste Mass Jsc Voc FF RS RSH pFF Cell ID Type Firing η (%) (mg/cm2) (mA/cm2) (mV) (%) (Ωcm2) (Ωcm2) (%) TOPCon M10 at QB65 Ag front BTF & 0.20 41.8 734 79.4 24.4 1.1 41.8×104 85.2 (TNA52) Cu Rear LECO Task 7: Materials Sourcing, Cost Model Refinement and Manufacturing Assessment Tool / Milestone Performance Metric Success Value Method of Measuring # EOP-C Material Supply Chain Supply Chain Map and Completed map and risk Supplier Risk Assessment assessment for all the materials in the paste. EOP-D Refined Cost Model < 10% the cost of Ag paste Excel sheet containing the Sensitivity range will be comprehensive cost model. updated depending on feedback from the independent reviewer in GNG-1D. Baseline cost of Ag paste defined as an average of commercially available Ag pastes, verified by third party. EOP-E Dissemination of Results >1 Scientific article in a peer review journal or industry recognized technical publication. EOP-F Joint Development ≥1 Count. Agreement or Investment or Binding letter of agreement with customer to test the Cu paste in their manufacturing line. Page 44 of 50 Subtask 7.1: Supply Chain Evaluation End of Project Milestone: EOP-C. Material Supply Chain A supply chain map and risk assessment for the materials used in the paste was completed. Each material was assigned a score of one or zero depending on whether it had (1) a US Source, (2) Additional non-China Source, (3) a China Source, (4) had an Internal Qualification Assessment Process, (5) at least one Source of Material would pass an Internal Qualification Process, (6) each Vendor Source could scale to a 1000 MT and (7) on the Vendor ISO status. The higher the score, the lower the risk. No material scored lower than 6. These documents were reviewed by a Solar Techno-Economic Analyst at NREL. The reviewer provided the following assessment: “The Supply Chain Risk Assessment shows that all constituent materials be sourced from domestic producers and from an addition[al] non-China source. All but one material can be sourced from the US up [to] volumes needed to produce 1000 MT of copper paste.” The full third-party review, Supply Chain Map and Risk Assessment, was submitted to the DOE for verification. Subtask 7.2: Updated Cost Model End of Project Milestone: EOP-D. Refined Cost Model The changes recommended by the third-party reviewer were built upon and updated with the latest Cu paste formulation. This model and summary description were reviewed by a Solar Techno-Economic Analyst at NREL. Based on the analyst’s feedback, the model was further refined. The updated model looked at scaling CuBertTM production up to 1000 MT/year, and compared the resulting price to the target Cu paste cost of less than 10% of Ag paste price. Ag paste pricing was obtained from the Shanghai Metals Market ($1001/kg and $1,131/kg with and without VAT, respectively on the 14th March 2025). xx Tariff and duty to the US are estimated at 25% tariff and 3% duty. xxi This gave a target Cu Paste cost of ~$120/kg, based on Ag front-side paste pricing. For modeling purposes, Ag paste price (for US customers) was estimated to be Ag market price/0.85. As background, the cost of Ag has fluctuated between $400 and $1500 per kg over the past decade and in March 2025 was ~$1000/ kg, while the cost of Cu has fluctuated between $3 and $10/ kg and is currently ~$10/kg. xxii [N.B. On the 14th October 2025 the price of Ag and Cu had increased to $1785/kg and $11/kg, respectively]. Bulk Cu has 94.6% of the conductivity of Ag. Although not included in this model, BTF commercial calculations assume an additional ~5 – 10 wt% Cu paste will be required to meet η parity with Ag paste applications. BTF has modeled that near parity in η between Ag and Cu is required to commercialize Cu pastes. The model also used learning rates to extrapolate the cost of production at much higher quantities of production than what BTF currently produces. BTF has obtained budgetary quotes at the 10, 100, and 1000 MT scale for the Cu powders that make up the majority of the paste at the 1 and 100 kg batch sizes. Based on these empirical data, learning rates were determined for quantities between 1 and 10,000 and 10,000 and 1,000,000 kg. To scale capital equipment, labor, QA, facility space, utilities, and other requirements, the model also utilized economies of scale and η gains with increased production. The model indicated at the current formulation that Cu paste manufacturing cost would be under 10% of an analogous Ag paste cost. The model did not consider various tariff scenarios, and assumes Ag continues to trade near the current market value. The reviewer’s assessment included “ the Bert Thin Films Inc. Cost Model Summary compares production of the CuBert paste to commercial Ag pastes in domestic production Page 45 of 50 of solar cells in the US. The model accurately compares Cu paste production costs and Ag paste market prices to the degree that assumptions can be validated. Ag paste prices, Cu costs, electricity rates, labor rates, and other defined input assumptions are in line with publicly available information for US production. The individual costs of materials other than Cu in the paste are not disclosed and can therefore not be verified. Similarly, the process equipment is not defined, but cost is listed and falls within expectation for manufacturing volumes under 100 MT/yr.” The full third-party review, model and summary were supplied to the DOE for verification. Subtask 7.3: Stakeholder Engagement in Pilot Stage Manufacturing End of Project Milestone: EOP-E. Dissemination of Results This task was completed by the publication of a peer reviewed article co-authored with UNCC in the Solar Energy Materials and Solar Cells Journal (Figure 32). Figure 32. Abstract for the published article. End of Project Milestone: EOP-F. Joint Development Agreement or Investment or Binding letter of agreement with customer to test the Cu paste in their manufacturing line. Cu is a known contaminant in c-Si solar cells that can have a detrimental effect on device performance if it reaches the junction. As a result, it was a major barrier in testing the performance of the Cu paste at a third-party customer site, without providing them with initial performance and reliability data showing that the paste works on their cells. Due to barriers related to accessing equipment capable of processing industrial sized cells, a third investment round was entered into during the project. Based on the results of cells processed and tested using equipment purchased with the funds, two Joint Development Agreements (JDA) with potential customers were signed during the project. Seed Round III closed at the end of 2024. Total amount = $2.5M. This round involved previous Angel investors and matching funds from a Kentucky seed stage venture capital investment group. Funds were used for product scaling and to facilitate equipment for customer testing at BTF’s facilities in Louisville, Kentucky. JDA with a module manufacturer with manufacturing facilities in North America: The manufacturer will supply cells, which will be printed and processed at BTF’s facilities. Data and processed cells will be shared across both groups. IEC testing will be carried out by the manufacturer. The JDA also included related business terms. JDA with a solar metallization paste manufacturer: Processed cells and data will be shared with the manufacturer. The manufacturer will work with BTF on improving the input materials to increase the performance of the pastes. Once the performance target has been reached, further discussions on the next steps will be taken. Page 46 of 50 Significant Accomplishments and Conclusions: Cu contacts made with BTF’s air fired CuBertTM metallization pastes achieved a maximum η of 21.4% when used as the front contacts on M6 (276 cm2) PERC cells, and an η of 24.4% when used as the rear contact on M10 (331 cm2) sized TOPCon cells. Thus, resulting in a relative η of 94% for PERC and 96% for TOPCon, when compared to the same structures metallized with Ag. The amount of Cu paste was reduced to below 0.39 mg/cm2, making the product more economically viable. Cost models on the production of the paste established that paste production would be under 10% of the cost of analogous Ag paste. Although the first attempts at partial IEC-61215 testing of 16 half- cell modules contact Cu metallized PERC cells were not successful, smaller scale tests on single and half-cell modules indicate that performance can be improved through further optimization of the pastes and processing parameters. The modules, however, were able to pass the thermal cycling component of the tests, indicating that the manually soldered interconnection was viable. Early stage testing of single cell TOPCon modules indicate that the Cu contacts are much more resilient to Cu diffusion, due to their structure. In the last 5 years major changes have been occurring in the PV industry, with cell sizes increasing from M2 (246 cm2) up to M12 (441 cm2), increased number of bus bars, shallower emitters and a major shift away from PERC to TOPCon. This has meant that many of the processing and measurement equipment available at R&D facilities were unable to handle these larger devices; resulting in sub-optimal performance. Consequently, a large proportion of the paste development work was carried out on 16 cm2 cells. However, this did force BTF to expand its capabilities with an IR belt furnace, Sinton flash tester, LECO unit and Laminator. Thus, enabling an acceleration in product development. Path Forward: BTF has demonstrated the CuBertTM Cu metallization pastes on both PERC and TOPCon technologies and is actively working on closing the performance gap to Ag metallized cells, as well as improving durability. The company is also focused on providing a frontside solution for TOPCon. This work is being carried out with both cell and module manufacturers. In addition to collaborating with Cu powder suppliers to make a robust supply chain, BTF is also discussing further scaling manufacturing with paste production facilities and has already demonstrated the production of a 25 kg batch of Cu paste. C-Si solar cells are typically produced on turnkey production lines, resulting in less manual handling of the product and wafers. Therefore, to understand how the product will perform under these conditions, BTF is looking to test the product on pilot or production lines. However, the industry is reluctant to fire Cu pastes in their production furnaces, and some are even reluctant to bring the product on to their site without proof that the material is viable. Many manufacturers site the potential of Cu to indirectly poison cells via the contamination of manufacturing equipment. BTF has mitigated some of their concerns by building an applications laboratory in Louisville, KY that allows testing on customers cells, prior to the Cu pastes being potentially transferred to their development and production lines. Alternative metallization strategies to Ag paste have been discussed for years, but between the imminent terra-watt scale PV manufacturing and recent historical rises in the price of silver ($1785/kg on the 14th October 2025), the drive to develop and ultimately adopt Cu metallization has been accelerating. Page 47 of 50 Products: Products developed or submitted under this award. Accepted Manuscripts of Journal Article OSTI ID Article Author(s) DOI Journal Volume Publication Status Acknowledgement Legal Title Name and Date of Federal Disclaimer Page No. Support Language 2282892 Screen Abasifreke Ebong, 10.1016/j.solma Solar Volume 1/15/2024 published Yes Yes printable Donald Intal, Sandra t.2023.112633 Energy 265, copper Huneycutt, Thad Materials 112633 pastes for Druffel, Ruvini and Solar silicon Dharmadasa, Kevin Cells solar cells Elmer, Apolo Nambo Conference Proceedings, Papers, Presentations OSTI ID Presentation, Author/Speaker DOI Conference Conference Status Acknowledgement Legal Paper Title Name Date of Federal Support Disclaimer Language 1992500 Presentation: Thad Druffel, 10.2172/19 11th 8th – 9th published Yes No Screen Ruvini Dharmadasa, 92500 Metallization May 2023 Printable Kevin Elmer, and Copper Pastes Apolo Nambo, Interconnection for Silicon Solar Aba Ebong, Workshop, Cells Sandra Huneycutt Session IV: Donald Intal Copper-based cell metallization 2507326 Poster: Thad Druffel, Dustin 10.21272/25 12th 23rd – 24th accepted Yes Yes Reliability of c- Williams, Kevin 07326 Metallization October Silicon cells Elmer, Erin Yenney, and 2024 using Screen- Apolo Nambo, Ruvini Interconnection Printed Copper Dharmadasa, Workshop Contacts Fired Vijaykumar in Air Upadhyaya, Ajeet Rohatgi, Ajay Upadhyaya, Bill Nemeth, Steve Johnston, Harvey Guthrey, Peter Hacke, Paul Stradins Page 48 of 50 Project Team and Roles: List of all the individuals that participated in the project. Name / Job Title/ Organization Role in the Project Directly funded Ruvini Dharmadasa • Principal investigator on the project. She Yes Chief Technology Officer was also involved in materials, paste and Bert Thin Films Inc. device development. Thad Druffel • Business and technical support within the No Chief Executive Officer organization. He assisted with contacting Bert Thin Films Inc. key stakeholders and technical advice. Matt Healy • Business lead. In the project he assisted No Chief Strategy Officer with commercializing the product, as well Bert Thin Films Inc. as the supply chain and cost model Krishnamraju Ankireddy • He was involved in paste formulation, Yes Photovoltaic Engineer and optimization of large area PERC Bert Thin Films Inc. devices. Kevin Elmer • He was involved in establishing paste Yes Quality Assurance Chemist SOP’s and was involved in paste and Bert Thin Films Inc. firing optimization, and durability testing. Hugo Apolo Nambo Salgado • He was involved in materials, synthesis Yes Materials Engineer and was also involved in interconnection Bert Thin Films Inc. and durability testing. Surabhi Sharma • She was involved in the synthesis and Yes Chemist paste formulation within the organization. Bert Thin Films Inc. Dustin Williams • He was involved in device measurement, Yes Research Engineer interconnection development and Bert Thin Films Inc. durability testing within the organization. Erin Yenney • She was involved with adjusting paste Yes Research Engineer rheology, developing the screen Bert Thin Films Inc. parameters, analyzing the contacts. James Ian Ammons • He was involved in producing the No Technician materials used in the paste. Bert Thin Films Inc. Stephanie Schaupp • She was involved in developing the No SMILE Intern refined cost model. Abasifreke Ebong • Principial Investigator. He was involved Yes Professor with directing the subrecipients’ work on University of North Carolina Charlotte the project; including cell processing, (Subrecipient) measurement and characterization. Sandra Huneycutt • She was involved in firing cells, device Yes Student / Research Assistant measurement and materials/device University of North Carolina Charlotte interface characterization. Donald Intal • He was involved in firing cells, device Yes Student / Research Assistant measurement and materials/device University of North Carolina Charlotte interface characterization. Ajeet Rohatgi • He was involved with directing and Yes Professor and Director of UCEP reporting the work on the project. Georgia Institute of Technology (Collaborating Organization) Ajay Upadhyaya • He was involved in the projects Yes Associate Director – Research Admin administration. Georgia Institute of Technology Vijay Upadhyaya • He was involved in firing cells, and Yes Research Engineer device measurement. Georgia Institute of Technology Page 49 of 50 References: i IEA Photovoltaic Power Systems Technology Collaboration Programme (2025), Trends in Photovoltaic Applications 2025, Report IEA PVPS T1-48:2025, ii D. Feldman, J. Zuboy, K. Dummit, L. Reed, D. Stright, M. Heine, H. Mirletz, R. Margolis, Winter 2024 Solar Industry Update. NREL, 25th January 2024. iii Metals Focus, Despite Emerging Challenges PV Silver Demand Set to Hit a New High in 2024. 11th September 2024, https://www.moneymetals.com/news/ iv The Silver Institute, Silver Supply and Demand,13th May 2024, https://silverinstitute.org/silver-supply- demand/ v International Technology Roadmap for Photovoltaic (ITRPV). Results 2023. Fifteenth Edition, May 2024. vi M.I. Yacouba, A. Lambertz, Y. Liu, H. Gattermann, V. Lauterbach, K. Bittkau, U. Rau, K. Ding. Achieving High Efficiencies for Silicon Heterojunction Solar Cells Using Silver‐Free Metallization. Progress in Photovoltaics: Research and Applications, 33(11), pp.1223-1235. 2025. vii M.W. Martinez-Szewczyk, S.J. DiGregorio, O. Hildreth, M.I. Bertoni. Reactive silver inks: a path to solar cells with 82% less silver. Energy & Environmental Science, 17(9), pp.3218-3227. 2024. viii X. Wu, X. Wang, W. Yang, J. Nie, J. Yuan, M.U. Khan, A. Ciesla, C. Sen, Z. Qiao, B. Hoex. Enhancing the reliability of TOPCon technology by laser-enhanced contact firing. Solar Energy Materials and Solar Cells, 271, p.112846. 2024. ix Y. Cheng, Y. Zhang, Y. Xu, A. Stokes, M. Dhamrin, S. Deng, L. Sun, K. Tsuji, J. Seidel, D. Chen, Y. Chen. Integration of aluminum contacts in TOPCon solar cells: A pathway to reduce silver usage. Solar Energy Materials and Solar Cells, 285, p.113559. 2025. x B. Grübel, G. Cimiotti, C. Schmiga, S. Schellinger, B. Steinhauser, A.A. Brand, M. Kamp, M. Sieber, D. Brunner, S. Fox, S. Kluska. Progress of plated metallization for industrial bifacial TOPCon silicon solar cells. Progress in Photovoltaics: Research and Applications, 30(6), pp.615-621. 2022. xi V. Unsur, M.K. Arslan, E. Alloji, B. Akgayev, H. Tokgoz, R. Turan, S. Akbayrak. Screen printed Ag-doped nickel metallization for industrial n-TOPCon silicon solar cells. Solar Energy Materials and Solar Cells, 287, p.113602. 2025. xii Y.C. Chang, Y. Zhang, L. Wang, S. Wang, H. Wang, C.Y. Huang, R. Chen, C. Chan, B. Hallam. Silver‐ lean metallization and hybrid contacts via plating on screen‐printed metal for silicon solar cells manufacturing. Progress in Photovoltaics: Research and Applications, 33(1), pp.158-169. 2025. xiii Y. Zhang, S. Wang, L. Wang, Z. Sun, Y.C. Chang, R. Chen, C. Chan, Y. K. Okamoto, D. Wang, M. Dhamrin. Ultra‐Lean Silver Screen‐Printing for Sustainable Terawatt‐Scale Photovoltaic. Solar RRL, 8(17), p.2400478. 2024. xiv H. Höffler, F. Simon, E. Krassowski, J. Greulich. Understanding current paths and temperature distributions during ‘Laser Enhanced Contact Optimization’ (LECO). In AIP Conference Proceedings (Vol. 2826, No. 1). 2023. xv A. Kraft, Plated Copper Front Side Metallization on Printed Seed-Layers for Silicon Solar Cells. Fraunhofer ISE, Fraunhofer Verlag, (2015). xvi J. Bartsch, A. Mondon, K. Bayer, C. Schetter, M. Hörteis, and S. Glunz, Quick determination of copper- metallization long-term impact on silicon solar cells. Journal of the electrochemical Society, 157(10), p.H942. 2010. xvii K. Ankit, T. Mueller, R.A. Stangl, B. Hoex, P.K. Basu, A.G. Aberle. A fill factor loss analysis method for silicon wafer solar cells. IEEE Journal of Photovoltaics 3, no. 4 (2013): 1170-1177. xviii A. Lachowicz, N. Badel, J. Gay, G. Andreatta, N. Blondiaux, A. Faes, P. Wyss, J. Geissbühler, J. Diaz Leon, J. Zhao, A. Descoeudres, B. Paviet-Salomon, C. Ballif. Reinforcement of Screen-Printed Copper Paste by Electrodeposited Copper for Metallization of Silicon Heterojunction Solar Cells presented at the 10th Metallization and Interconnection Workshop, Genk, Belgium, November 2021. xix R.K. Satpathy, V, Pamuru, Solar PV power: design, manufacturing and applications from sand to systems. Academic Press. Ch. 5. (2020). xx Shanghai Metals Market, 14th March 2025, https://www.metal.com/Solar/202112230005 xxiHarmonized Tariff Schedule, Silver Paste = 7115.90.4000, 15th March 2025, https://hts.usitc.gov/search?query=711590 xxii Markets Insider, 15th March 2025, https://markets.businessinsider.com/commodities/silver-price Page 50 of 50