Investigation of Coffee Qualities through Electrostatic and Electrochemical Methods by Robin Bumbaugh A dissertation accepted and approved in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Chemistry Dissertation Committee: James Prell, Chair Christopher H. Hendon, Advisor Carl Brozek, Core Member Benjamı́n Alemán, Institutional Representative University of Oregon Summer 2024 © 2024 Robin Bumbaugh This work is openly licensed via CC BY-NC-ND 4.0. 2 https://creativecommons.org/licenses/by-nd/4.0/ DISSERTATION ABSTRACT Robin Bumbaugh Doctor of Philosophy in Chemistry Title: Investigation of Coffee Qualities through Electrostatic and Electrochemical Methods This thesis presents groundbreaking research on using electrochemical methods in conjunction with % Total Dissolved Solids (%TDS) as a marker for brewed coffee qualities. The study emphasizes the necessity of brewing reproducible coffee for accurate measurements. It highlights the impact of adding water to coffee beans before grinding, which reduces electrostatic charge and results in a more uniform particle size distribution, enhancing consistent extraction during brewing. The research further examines the relationship between roast degree, measured by Agtron value, and %TDS, utilizing cyclic voltammetry (CV) as a novel technique for analyzing brewed coffee. A strong correlation is found between the integrated area of the observed reduction wave and %TDS, linked to hydrogen underpotential deposition (HUPD) on a platinum electrode. It is also found that coffee matrix molecules adsorb to the electrode surface, block reaction sites, and suppress the HUPD signal with multiple CV cycles. The study further explores the effects of varying brew parameters (grind size, water temperature, water amount, bean amount, and brew time) on CV characteristics, demonstrating linear correlations between %TDS, HUPD reduction wave area, and peak height, with shifts in brew parameters impacting these metrics. Additional CV characteristics, peak center and peak full-width-half-maximum both of which are known to relate to solution composition, are observed to shift with brew parameters but do not correlate to 3 %TDS. The innovative use of CV for assessing coffee quality opens new avenues for electrochemistry techniques in food science, with potential applications in other acidic liquids such as wine and tea. Future research could leverage multifactor analysis for standard protocols in coffee scaling and flavor targeting, possibly incorporating electrochemical devices in brewing processes to allow consumers to adjust for individual flavor preferences. This dissertation includes previously published and unpublished co-authored material. 4 CURRICULUM VITAE NAME OF AUTHOR: Robin Bumbaugh GRADUATE AND UNDERGRADUATE SCHOOLS ATTENDED: University of Oregon, Eugene, OR, USA California State University, Chico, Chico, CA, USA DEGREES AWARDED: Doctor of Philosophy, Chemistry, 2024, University of Oregon Master of Science, Chemistry, 2020, University of Oregon Bachelor of Science, Chemistry and Biochemistry, 2019, California State University, Chico AREAS OF SPECIAL INTEREST: Coffee Analysis Analytical Chemistry Electrochemistry PROFESSIONAL EXPERIENCE: Graduate Researcher, Chemistry, 2021-2024 Scientist I, Oregon’s Wild Harvest, 2020-2021 Graduate Teaching Assistant, General Chemistry, 2019-2020 Professional Intern, NASA Kennedy Space Center, 2019 GRANTS, AWARDS AND HONORS: Graduate Student Community Builder Award, University of Oregon, 2023 Outstanding Graduating Senior in Chemistry, California State University, Chico, 2019 5 PUBLICATIONS: Bumbaugh, R. E., Pennington, D., Sartori, C., Barrales, J., Rheingold, E. J., Wehn, L. C., Tomasino, E., Hendon, C. H., Bulk electrolysis of coffee directly affects sweetness and acidity, In Preparation, 2024 Bumbaugh, R. E., Huang, J., Wehn, L. C., Rheingold E. J., McDonald C. S., Alemán B., Boettcher S. W., Brozek, C. K., Hendon, C. H., An Electrochemical Descriptor for Coffee Quality, In Preparation, 2024 Bumbaugh, R. E., Aléman B., Hendon C. H., A Universal Voltametric Model for Designer Coffee, In Preparation, 2024 Méndez Harper, J., McDonald, C. S., Rheingold, E. J., Wehn, L. C., Bumbaugh, R. E., et al., Moisture controlled triboelectrification during coffee grinding, Matter, 2023, DOI: 10.1016/j.matt.2023.11.005 Bumbaugh, R. E.; Ott, L. S. Preparing and Testing Novel Deep Eutectic Solvents from Biodiesel Co-Product Glycerol for Use as Green Solvents in Organic Chemistry Teaching Laboratories. Environmental Research Literacy: Classroom, Laboratory, and Beyond ACS Symposium Series 2020, 1351, 113–130 6 ACKNOWLEDGEMENTS First and foremost, I would like to thank Professor Christopher H. Hendon for his invaluable guidance as my principal investigator. I have greatly enjoyed working on such a unique project and am honored to be the first of his students to receive a degree focusing on coffee chemistry. My heartfelt thanks also go to my committee members: Professor James Prell, Professor Carl Brozek, and Professor Benjamı́n Alemán, for their scientific insights and moral support. Professor Alemán, in particular, has been a personal support, for which I am deeply grateful. I would also like to thank Dr. Joshua Méndez Harper for his camaraderie and expertise in electrostatics. Additionally, I would like to thank Professors Shannon Boettcher and Paul Kempler for their invaluable knowledge and insight into electrochemistry. I want to extend my warmest thanks to my undergraduate mentor, Professor Lisa Ott, for being a scientist and a person I strive to emulate. My deepest gratitude also goes to my middle school science teacher, Mrs. Jana Morris, for sparking my love for science. I would not be where I am today without the influence of these two remarkable women. A special thank you goes to Dr. Joshua Williams at Drexel University for his assistance with chromatography instrumentation. It is always a joy to see researchers at different institutions collaborating to advance science. I am also grateful to Dr. Lawrence Scatena for his help with Raman spectroscopy, and to Dr. Jiawei Huang, Tingting Zhai, and Sarah Beaudoin for their assistance with QCM instruments. 7 I also want to thank the many friends I have made at the University of Oregon. Notable mentions include Dr. Khoa Le, Dr. Michael LeRoy, Dr. Samantha Shepherd, and my Hendon Labmates, especially Dr. Eoghan Gormley and Doran Pennington. Finally, I would be remiss not to thank the undergraduate students I have had the joy of working with, particularly Lena Wehn, Eli Rheingold, Connor McDonald, and Julio Barrales. My deepest gratitude goes to my outstanding family. I could not have achieved this without the unwavering support and encouragement of my mother, Rhonda Bumbaugh, my sister, Rachel Aschbacher, and my dear friend, Natalie Schlosser. Finally, my heartfelt thanks to my fiancée, Brandon Brewington, whose constant encouragement and love remind me to enjoy life and whose presence makes every achievement more meaningful. 8 To my loud, proud, gregarious father Raymond Allen Bumbaugh. I wish you were here to celebrate with me. 9 TABLE OF CONTENTS Chapter Page I. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . 28 1.1. Making Measurements on Coffee . . . . . . . . . . . . . . . . 28 1.1.1. Separation Techniques Applied to Coffee Analysis . . . . . . 31 1.1.2. Electrochemistry Techniques for Coffee Analysis . . . . . . 35 1.2. Co-Authors and Publication Details . . . . . . . . . . . . . . . 42 1.2.1. Chapter 2, Electrostatics of Coffee Grinding . . . . . . . . 42 1.2.2. Chapter 3, Determining Roast Degree . . . . . . . . . . . 42 1.2.3. Chapter 4, Impact of Brewing Parameters . . . . . . . . . 43 II. ELECTROSTATICS OF COFFEE GRINDING . . . . . . . . . . . 44 2.1. Moisture-controlled triboelectrification during coffee grinding . . . . 44 2.1.1. Author Contributions . . . . . . . . . . . . . . . . . . 44 2.1.2. Experimental procedures . . . . . . . . . . . . . . . . . 44 2.1.3. Progress and Potential . . . . . . . . . . . . . . . . . . 46 2.1.4. Introduction . . . . . . . . . . . . . . . . . . . . . . 47 2.1.5. Electrifying Coffee . . . . . . . . . . . . . . . . . . . 48 2.1.6. General trends in electrification of grinding commercially roasted coffee . . . . . . . . . . . . . . . . 51 2.1.7. Isolating the impact of roast profile . . . . . . . . . . . . 55 2.1.8. Grind setting-dependent charging . . . . . . . . . . . . . 58 2.1.9. How granular mechanics influence electrification . . . . . . 61 2.1.10. Grinding with supplemental external water and its impact on brewing . . . . . . . . . . . . . . . . . . 63 2.1.11. A theoretical explanation for the role of water in charge passivisation . . . . . . . . . . . . . . . . . . . 69 10 Chapter Page 2.1.12. Outlook and conclusion . . . . . . . . . . . . . . . . . 71 2.1.13. Supplemental Information . . . . . . . . . . . . . . . . 72 2.2. Strategies to mitigate electrostatic charging during coffee grinding . . 72 2.2.1. Author Contributions . . . . . . . . . . . . . . . . . . 72 2.2.2. Method Details . . . . . . . . . . . . . . . . . . . . . 72 2.2.3. Introduction . . . . . . . . . . . . . . . . . . . . . . 74 2.2.4. Canvasing charge passivation strategies . . . . . . . . . . 75 2.2.4.1. De-electrification through external water inclusion . . . . . . . . . . . . . . . . . 76 2.2.4.2. Static reduction using ion beams . . . . . . . . . 77 2.2.5. Aggregate formation . . . . . . . . . . . . . . . . . . . 81 2.2.6. The effect of charge mitigation on espresso quality . . . . . 82 2.2.7. Conclusion . . . . . . . . . . . . . . . . . . . . . . . 84 2.2.8. Limitations of the study . . . . . . . . . . . . . . . . . 85 2.2.9. Supplemental Files . . . . . . . . . . . . . . . . . . . 85 III. DETERMINING ROAST DEGREE . . . . . . . . . . . . . . . . 93 3.1. An electrochemical descriptor for coffee quality . . . . . . . . . . 93 3.1.1. Author Contributions . . . . . . . . . . . . . . . . . . 93 3.1.2. Experimental Methods . . . . . . . . . . . . . . . . . . 93 3.1.3. Introduction . . . . . . . . . . . . . . . . . . . . . . 99 3.1.4. Results and Discussion . . . . . . . . . . . . . . . . . 100 3.1.5. Conclusion . . . . . . . . . . . . . . . . . . . . . . . 105 IV. IMPACT OF BREWING PARAMETERS . . . . . . . . . . . . . 108 4.1. A Universal Voltametric Model for Designer Coffee . . . . . . . . 108 4.1.1. Author Contributions . . . . . . . . . . . . . . . . . . 108 4.1.2. Methods . . . . . . . . . . . . . . . . . . . . . . . . 108 11 Chapter Page 4.1.3. Introduction . . . . . . . . . . . . . . . . . . . . . . 110 4.1.4. Results and Discussion . . . . . . . . . . . . . . . . . 111 4.1.5. Conclusion . . . . . . . . . . . . . . . . . . . . . . . 123 V. CONCLUSION . . . . . . . . . . . . . . . . . . . . . . . . . 125 APPENDICES A. CHAPTER 2 MOISTURE-CONTROLLED TRIBOELECTRIFICATION DURING COFFEE GRINDING SUPPLEMENTAL INFORMATION . . . . . . . . . . . 128 B. CHAPTER 2 STRATEGIES TO MITIGATE ELECTROSTATIC CHARGING DURING COFFEE GRINDING SUPPLEMENTAL INFORMATION . . . . . . . . . . . 135 C. CHAPTER 3 AN ELECTROCHEMICAL DESCRIPTOR FOR COFFEE QUALITY SUPPLEMENTAL INFORMATION . . . . . . . . . . . . . . . . . . . . . . . . . 139 REFERENCES CITED . . . . . . . . . . . . . . . . . . . . . . . . 149 12 LIST OF FIGURES Figure Page 1. Coffee Brewing Control Chart originally designed by Lockhart in 1957. Reprinted with permission from Sci Rep 10, 16450 (2020), doi.org/10.1038/s41598-020-73341-4. . . . . . . 29 2. Flow chart of various specialty coffee processing methods. Reprinted with permission from LWT 172 (2022) 114245, doi.org/10.1016/j.lwt.2022.114245. . . . . . . . . . . . . . . . . . 32 3. Headspace GCMS of volatile compounds from 70.0g of whole beans heated in 60°C water bath for 60 minutes. Comparison of two samples that were characterized as having or not having the potato-taste defect, common in Rwandan coffee, via cupping after GCMS data was collected. Reprinted with permission from J. Agric. Food Chem. 2014, 62, 42, 10222-10229. Copyright 2014 American Chemical Society. . . . . . . . . . . . . . . . . . . . . . . . . 34 4. HPLC chromatogram performed on two samples of commercially available undiluted coffee brewed via coffee pods at a 1:6 ratio. (coffee pods 5g in, 30 mL out aka 1:6 ratio). Reprinted with permission from J. Chem. Educ. 2023, 100, 4, 1564–1570. Copyright 2023 American Chemical Society. . . . . . . . . . . . . . . . . . . . . . . . . 36 5. Visual description of applied potential versus time and associated current versus potential outputs for cyclic voltammetry (top, A), differential pulse voltammetry (middle, B), and square wave voltammetry (bottom, C). Reprinted with permission from ECS Sensors Plus, 2024 3 027001, doi.org/10.1149/ 2754-2726/ad3c4f, and Creative Commons https://creativecommons.org/licenses/by-nc-nd/4.0/ . . . . . 39 13 Figure Page 6. Square Wave Voltammetry of unroasted (aka green beans) coffee extracted at a 1:100 ratio and diluted 50-fold (green trace). Traces in blue, red, and black are each spiked with the same amount of a different chlorogenic acid isomer. Reprinted with permission from Int. J. Electrochem. Sci., 9 (2014) 6134 – 6154, doi.org/10.1016/S1452-3981(23)10876-5, and Creative Commons https://creativecommons.org/licenses/by-nc-nd/4.0/. . . . . 40 7. Differential Pulse Voltammetry of diluted green bean extract 1:100 ratio (opposed to 1:16.5 we use) (black trace, a). Traces b through d were each spiked with different chlorogenic acid isomers. Reprinted with permission from Int. J. Electrochem. Sci., 11 (2016) 2854 - 2876, doi.org/10.1016/S1452-3981(23)16146-3, and Creative Commons https://creativecommons.org/licenses/by-nc-nd/4.0/. . . . . 41 8. Electrification of coffee beans and particles. a) Whole coffee beans accumulate charge when rolled down a vibrating ramp coated in a variety of materials. b) These surfaces materials can be arranged according to their capacity to charge whole beans. Here, Starbucks Blonde Espresso Roast weakly charges against steel, while glass and nylon result in positive charging, and plastics like PVC and Mylar lead to negative charging. c) During fracture, coffee particles accumulate charges from the burr-coffee interface and coffee-coffee rubbing (tribocharging), as well as fracture points (fractocharging). d) After grinding, the charge is quantified by alloying the particles to fall between two sub-parallel electrodes with a potential difference of 8.2 kV. The electric field separates particles by charge polarity, and particles are collected in negative, neutral, and positive bins at the base of the separator. . . . . . . . . . . . . . . . . . . 49 14 Figure Page 9. Charging regimes of commercially sourced coffees a) Three Mexican coffees – Tacámbaro (N), Mané (B), and Temascaltepec (W), show three possible charging regimes. b) Charge-to-mass ratio as a function Agtron color for a variety of coffees ground at setting 2.0. There is no strong relationship between charge magnitude and polarity, and coffee processing method. Confidence bands are plotted at 90%). A polarity switch is observed for darker coffees (Agtron = ≤ 70). The magnitude of negative charge continues to increase with darkness. c) Internal moisture - a property proportional to roast color - results in more positive charging. Moisture content is a slightly better predictor of charge-to-mass ratio than color. d) Examination of the particles collected for two representative coffees, Kolla (W, a positive charging, coffee) and Mané (a negative charging coffee), reveals that positively charged particles are generally smaller than the boulders, as denoted by the average arrows in blue. . . . . . . . . . . . . . . . . . . . 52 10. The effect of roast profile on charging a) Two sets of roast profiles were explored: Short Morse (blue) and Long Morse (purple). b) Charge-to-mass ratios as a function of roast time for a representative Ethiopian coffee (YirgZ) shows that charging increases with darker roasts. Prolonged Morse transitions to the negative charging regime at a slower rate than the shorter Morse. An additional Mexican coffee (Yogondoy) was roasted using the short Morse profile, yielding a similar electrification behavior to the Ethiopian coffee. c) Charge-to-mass ratios for Ethiopian and Mexican coffees as functions of Agtron color. d) Charge-to-mass ratios for Ethiopian and Mexican coffees as functions of residual water content. . . . . . . . . . . . . . . . . . . . . . . 56 15 Figure Page 11. Charging as a function of grind setting and roast a) Charge-to-mass ratios of two representative coffees ground at settings spanning espresso (fine) to French press (coarse). In general, grinding finer increases charging, and at settings around 1.0 (the espresso region), positively charging coffees a large spread due to the formation of b) aggregates as the coffee exists the grind chamber. For negatively charging coffees, grinding at 1.0 yields a slightly more positive average charge, with relatively small spread. We attribute this to the higher charging resulting in the very rapid formation of aggregates. c) At a constant grind setting, dark coffees have approximately 100 µm finer mean particle size than light coffees. . . . . . . . . . . . . . . . . . . . 59 12. Twice-grinding coffee to assess the role of fracture in electrification. The first grinding of whole bean coffee at setting 2.0 results in charging from both triboelectric and fractoelectric processes, with a corresponding particle size distribution presented inset in grey. Regrinding the same particles at setting 6.0 results in a reduction in surface charge, with essentially no change in volumetric particle size distribution. First grinding is presented in grey, and second grinding in black. . . . . . . . . . . . . . . . . . . . . . . . . . 62 13. External moisture controls surface charging and causes particle deaggregation a) Charge-to-mass ratios for several coffees that span positive, neutral, and negative charging, with increasing amounts of water introduced to the whole beans prior to grinding. The red-to-blue coloring is indicative of the roast color, where blue are darker roasts. The upper x-axis provides an estimate for the effective change in total moisture content contained within, and coated on the surface, of the whole beans. b) The inclusion of minerals in the water solution has no effect on the magnitude of charge suppression achieved by the water itself. c) The inclusion of water during grinding causes deaggregation of fines from boulders. d) The redistribution of particle polarity upon the addition of 20 µL g−1 water added to the whole beans. . . . . . . . . . . . . . . . . . . . . . 64 16 Figure Page 14. Shot time and flow rate dependence with and without water added to whole bean coffee a) Without changing any other parameter, coffee prepared using the addition of water to whole beans during grinding produces consistently slower shots with increased beverage strength. b) The change in flow rate can be fit using a logistic function such that the permeability of the bed approaches a constant within conventional espresso brewing times. The time in which the sigmoid takes to reach its inflection is nearly half that of the shots prepared with water indicating that the bed is more permeable, and progresses towards equilibrium permeability more rapidly. . . . . . . . . . . . . . . . 68 15. Electrification of coffee during grinding. a) Schematic of the setup used to assess the electrification of coffee during grinding. During fracture, coffee particles accumulate charge from the burr-coffee and coffee-coffee interfaces (tribocharging), as well as fracture points (fractocharging). Charge-to-mass ratios can be measured with a Faraday cup and scale. b)Particle size distributions for our in-house roasted coffees ground at setting 2.0 on our Mahlkönig EK43. c) Example charging curves (raw data from Faraday cup) for lighter/wetter and darker/dryer coffees. d) Photograph of a spark discharge spanning the gap between a metal cup containing freshly ground coffee and the lead author’s finger. Assuming a breakdown field of 3 MV m-1 (air at 101 kPa), the potential difference between the two surfaces is ∼7.5 kV. . . . . . . . . . . . . . . . . . . . . . . . 88 16. Time-resolved and water-mediated charge reduction strategies. a) Schematic of setup used to measure the charge decay in freshly ground coffee. A non-contact voltmeter was placed 5 mm above 10 g of ground coffee and its potential was measured every 0.5 sec. b) Charging dissipates with time as charge-carriers recombine with each other through surface and bulk conduction. Some charge may also be lost directly to the atmosphere. Charge in lighter roasts decays faster than in dark coffee. c) During the grinding process, charge accumulation is hindered by the addition of small amounts of water (0 - 30 µL g-1) to the whole beans prior to grinding. . . . . . . . . . . . . . . . . . . . 89 17 Figure Page 17. Ion beam charge reduction strategies. Charging may be counteracted by ionizing the air around the coffee grounds as these exit the grinder. Free negative and/or positive ions adhere to solid particle surfaces, tuning their charge. a) Ionization may accomplished via high-voltage gaseous breakdown. b) While potentially effective, the number of positive and negative ions generated must be adjusted to balance the charging characteristics of a coffee. The nominal charging behavior of coffee with no de-electrification is presented in white. Dark coffees generally natively charge negatively, while light coffees charge positively. The distance between the ionizer and coffee is presented in mm. c) Negative and positive ions may also be generated via a bipolar high-voltage source. d) Exposing negatively charging coffee to a balanced ionizer reduces its charge by at least 50 %. . . . . . . . . . . . . . . . . . 90 18 Figure Page 18. Particle aggregation and grinder retention a) For dry-ground coffee (Yogondoy [dark]), expelled grounds follow the particle size distribution is presented in brown. Grounds retained within the grinding cavity concentrate fines (dotted, brown curve). Fines have higher electrostatic- to-gravitational ratios, meaning they are more likely to adhere to surfaces when charged. Adding even a small amount of water (10 µl g-1) can significantly reduce electrostatic aggregation, reducing retention and shifting expelled ground particle sizes toward smaller diameters (dashed, blue curve). b) Water contents in the range of 0-50 µL g-1 continue to shift particle sizes toward smaller mean diameters. Water contents above 50 µL g-1 again increases the mean particle size, indicating the activation of wet (capillary) aggregation processes. c) A linear shift in mean particle size and ion density is not observed for coffee treated with a unipolar corona ionizer at different chute- ionizer distances. These data suggest that fine particles within the grinder are not included in the measurement sample (that is, they remain electrostatically adhered to the inner surfaces of the grinder), and the aggregates are formed before deionization, which is to be expected since the corona ionizer is placed after the chute. d) Because the water addition technique (RDT) hinders electrification throughout the grinder, the wet method (using 10 µL g-1) has the ability to greatly reduce retention. Ionization (7.8 × 106 cm-3), addressing static only at the grinder chute, involves retention masses similar to those of grinding with no static mitigation treatment. Grind data in panels a-c were collected in triplicate and the averages are presented. . . . . . . . 91 19 Figure Page 19. Espresso shot time and flow rate dependence with and without charge mitigation for a dark roast (Temascaltepec) a) Without changing any brewing parameters, coffee prepared using the addition of water to whole beans during grinding produces consistently longer shots (left panel) with reduced flow rates. The shot flow rate can be fit using a generalized logistic function such that the permeability of the bed approaches a constant (right panel). Note that the time it takes an espresso prepared with water to reach this plateau is significantly longer than that of an espresso brewed conventionally. b) Using a positive high-voltage ionizer, we do not observe an appreciable increase in shot time or a reduced flow rate. We do observe a modest increase in % TDS. The departure from the behavior observed with the Ross droplet technique highlights the fact that ionization methods at the grinder chute do not address electrostatic effects within the grinder cavity and highlights the importance of de-electrification during grinding. Shots were run in triplicate, and compared to the untreated samples. For a p-value of 0.05, the ionization treatment is not significantly different from the native charging sample (t-value of -1.625). However, the water addition treatment was found to be significantly different (t-value of 6.059) at the same p-value. . . . . . . 92 20. Electrodes used in experiments. a) Platinum wire counter electrodes. Left, as purchased from CH Instruments. Right, coiled to maximize surface area. b) Silver/silver chloride reference electrodes. Left, used and right, new. c) Platinum working electrodes. Left, platinum mesh electrode. Middle, platinum dot electrode. Right, platinum quartz crystal microbalance electrode. . . . . . . . . . . . . . . . . . . . 94 20 Figure Page 21. Assessing the voltametric features present in coffee extracts. a) Cyclic voltammetry performed at 200 mV/s using boron-doped diamond, glassy carbon, and platinum working electrodes. Previously studied potential regions are highlighted in grey and red. The hydrogen underpotential deposition (HUPD) region is highlighted in blue. b) HUPD is suppressed with subsequent CV cycling due to deposition of coffee material on the surface of the electrode. c) The HUPD region depends on %TDS of the brewed coffee. d) Reductive scanning results and mass accumulation on the working electrode, leading to electrode fouling. e) Atomistic representation of proton-assisted deposition of coffee material onto the platinum electrode surface. . . . . . . . . . . 102 22. Mass accumulation and determination of Pt-surface adsorbed species. a) CV cycles with shortened scan window (upper) and the associated QCM mass increasing (lower). Only four cycles have been shown for clarity; experimental conditions involved 50-100 cycles. b) HPLC- MS chromatogram of coffee material collected from QCM electrode showing caffeine present with associated calibration curve of caffeine. . . . . . . . . . . . . . . . . . . . . 104 23. Relating roast profile to hydrogen underpotential deposition suppression. a) The temperature profiles used to generate six systematically darker coffees. b) The electrochemical response for identically brewed coffee diluted to ∼1 %TDS. c) Subtracting the background hydrogen evolution, the total charge passed is linear with %TDS, but depends on roast color, with darker roasts more effectively suppressing HUPD. d) Plotting the total charge and %TDS against the whole bean Agtron value (roast color) to create a “brew control plane” for our in-house roasted coffee. . . . . . . . . . . . . . . . . . . . . . . . . . . 106 24. a) A typical electrochemical response of filter strength coffee exhibits features like bulk, contaminated acidic water, Pt—Pt—Ag/AgCl. b) Beginning at 0 V and cycling reductively to -1.1 V, the hydrogen underpotential deposition feature is probed. c) The corresponding background-subtracted reductive feature. Key electrochemical features are labeled. . . . . . . . . . . . . . . 113 21 Figure Page 25. a) Clustered multivariate correlation matrix for CV feature vs. %TDS. The heat map represents the statistical correlation r with positive correlation in red and negative in blue, which have been clustered by correlation. b) Scatterplots of Ip1, A1, δE, and Ep1 vs. %TDS for the entire factorial data set. Linear fits are shown as solid lines with the shaded region giving the 95% confidence interval. The R2 values for the linear fit are also provided for each plot. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 114 26. a) Full-effects reduced models for %TDS, Ip1, A1, δE, and Ep1. Each plot shows the − log(P-value) for sources of the full-factorial experiment, including factors mb, mw, g, T , and τ . The plots were filtered to retain sources that had a significance error of α = 1% (i.e., − log(P-value) = 1) for at least one response variable (e.g., %TDS); hatched filling indicates sources above α = 5% (emphasized with transparent boxes). The full reduced model parameter analysis is provided in Appendix D. The %TDS model was transformed to stabilize the variance using log(%TDS). b) Scatter plots of the measured response versus the predicted response for log(%TDS), Ip1, A1, δE, and Ep1 using the respective reduced models. . . . . . . . . . . . . . . . . . . . . . 116 27. Effects plots for the measured variables log(TDS), Ip1, and Ep1 with respect to the main experimental factors mb, mw, g, T , and τ . The Ip1 data shows the magnitude of the current; the raw values have a negative sign. The levels for each factor are indicated above the factor label (e.g., mb levels are 8 g and 20 g). Centers correspond to the mean and the error bars indicate the standard error, which for log(TDS), Ip1, and Ep1 plots are approximately 0.01, 0.5 µA, and 0.0004 V, respectively. All effects are significant below P = 0.01, except those with means connected by a dashed line, whose P -value is given below the effect. . . . . . . . . . 119 22 Figure Page 28. The vertical lines are t-test cutoffs with N = 32 for (α = 5%, tα = 2.04), (α = 0.01%, tα = 4.48). Tukey-Kramer HSD test with 95% confidence rings using family error; circles overlapping with circle centers are not statistically significant below 5%; connecting letter reports: same letter is the same group. Kernel density plots for cycle-to-cycle fractional differences with smoothness factor set to zero; color coding indicates cycle differences as follows: i” - ”j/i, where i and j are cycle numbers. So 1” - ”2/1 for Ep data is the distribution for fractional difference (Ep1 − Ep2)/Ep1. . . . . . . . 122 A.1. Roast profiles of the 12 in-house roasts. The temperature versus time plots for each sample roasted on the Ikawa Pro100. The roasts resulted in the numerical values presented in Table A2. Short and long refer to Morse times on the Ikawa. . . . . . . . . . . . . . . . . . . . . . . . . 128 A.2. Empirically correcting for grind size drift due to roast color. Four arbitrary coffees – ranging from dark (1) to light (4) – were first ground at 2.0 and their particle size and charge was measured. To account for the difference in particle size distribution due to roast, the grind setting was then changed so that the particle size distributions of each coffee matched one another. In effect, the was achieved by grinding the darkest coffee 0.3 more coarse. A minor reduction in charging was observed, but most of the charging persists, highlighting that a drift in particle size due to roast is not the determining reason for increased charging in dark coffees. . . . . . . . . . . . . . . . . . . . . . 129 B.1. Temperature profiles used to generate light and dark roasts, related to Figure 16. We roasted the coffees listed in Table 1 using an Ikawa Pro roaster. We roasted coffee in batches of 50 grams. The profiles may be downloaded from 10.6084/m9.figshare.23277320.v1. . . . . . . . . . . 135 23 Figure Page B.2. Changes in humidity within the grinder as function of added water, related to Figure 18. A concern associated with the water droplet technique is that moisture could lead to corrosion within the grinder. Thus, we investigated the manner in which small quantities of water changed the environmental conditions inside the grinding cavity. For these experiments, we inserted a small relative humidity (RH) senso (Honeywell HIH-4030) into the grinder chute and secured it against the stator (non- spinning burr). Relative humidity was then recorded during grinding. Baseline humidity during the experiment was ˜40%. The curves rendered above show the variation in RH over time for added water contents in the range of 0–50 µl g−1. Experiments were conducted within minutes of each other in the order indicated by the numbers in parenthesis. For low water contents (˜10 µl g−1) we find that humidity in the chamber increases by a few percent and then decays back to background in a few seconds. For higher water contents, RH can increase by up to 30%, however we did not register condensation. While the most humid conditions were short-lived, we did find that several minutes were required for the internal RH to return to background for high water contents (>20 µl g−1). This longer-lived excess humidity, however, could be removed effectively by grinding a small amount of dry coffee (red curve, trial 6). Thus, if corrosion is a concern with the water droplet technique, we suggest that a sacrificial ”purge” grind be performed when coffee preparation is concluded. . . . . . . . . . . . . . . . . . . 136 B.3. Espresso shot time and flow rate dependence with and without charge mitigation for a light roast (Temascaltepec), related to Figure 19. a). Shot dynamics with (blue curves) and without (brown curves) the addition of extrinsic water. b) Shot dynamics with (red curves) and without (brown curves) treatment with high-voltage ionization. . . . . . . . . . . . . . . . . . . . . . . 137 24 Figure Page B.4. Although potentially harmful to humans, we also examined de-electrification using helium nuclei, related to Figure 17. a) Negative and positive ions may also be generated via alpha decay. b) Exposing negatively charging coffee, typically dark roasted, reduces charge comparable to the balanced ionizing method presented in the main text. These experiments were conducted for advancement of basic science and, besides being an inferior de-electrification technique compared to conventional ion beams and/or water addition, the risk associated with the use of radioactive elements makes this approach untenable. . . . . . . 138 C.1. Particle size distributions for EK43 Mahlkonig grinder settings 2 and 12. Setting 2 was used for ground bean Agtron value assessment and setting 12 was used for coffee cupping brews. . . . . . . . . . . . . . . . . . . . . . . . . . . 139 C.2. Example of electrochemical cleaning of platinum working electrode in 50 mM H2SO4. Cycles were repeated until the features visibly overlaid with one another. Scan rate 300 mV/s. . . . . 140 C.3. Background CV of water purified using the Pentair Everpure Conserv 75E Reverse Osmosis system. Scan rate of 200 mV/s. . . . . . . . . . . . . . . . . . . . . . . . . . . 140 C.4. Overlay of Roast 1a at 1.56 %TDS with a background scan of the reverse osmosis H2O used for coffee brewing and 50 mM solution of H2SO4 used for electrochemical polishing of electrodes. Scan rates for H2O and coffee were 200 mV/s and 300 mV/s for H2SO4. . . . . . . . . . . . . . . . . . . . . . 142 C.5. Scan rate dependence of HUPD peaks in undiluted brewed coffee. Reduction and oxidation peak centers do not correlate linearly with log base 10 of the scan rate, indicating non-Nernstian behavior. . . . . . . . . . . . . . . . . . 142 C.6. Example of background subtraction as performed in OriginPro9 to integrate reduction peak area. The top panel shows the first CV curve of roast 1a at 1.56 %TDS. The bottom panel shows the reduction sweep of the CV after the baseline correction was subtracted. . . . . . . . . . . . . . . . . . 143 C.7. Overlay of the first CV cycle of Roast 1a via cupping brew method at multiple %TDS values. Scan rate 200 mV/s. . . . . . . . . 144 25 Figure Page C.8. Overlay of the first CV cycle of Roast 1b via cupping brew method at multiple %TDS values. Scan rate 200 mV/s. . . . . . . . . 144 C.9. Overlay of the first CV cycle of Roast 1c via cupping brew method at multiple %TDS values. Scan rate 200 mV/s. . . . . . . . . 145 C.10.Overlay of the first CV cycle of Roast 1d via cupping brew method at multiple %TDS values. Scan rate 200 mV/s. . . . . . . . . 145 C.11.Overlay of the first CV cycle of Roast 1e via cupping brew method at multiple %TDS values. Scan rate 200 mV/s. . . . . . . . 146 C.12.Overlay of the first CV cycle of Roast 1f via cupping brew method at multiple %TDS values. Scan rate 200 mV/s. . . . . . . . 146 C.13.Total charge passed vs TDS performed in triplicate using Roast 1e with associated error bars. . . . . . . . . . . . . . . . . 147 C.14.Overlay of the first CV cycle of brewed coffee with and without sparging. Dark blue trace is the brewed coffee prior to any sparging; light blue trace is following 20 minutes of sparging with N2, green trace is following an additional 20 minutes of sparging with O2. Scan rate 200 mV/s. . . . . . . . . . . 147 C.15.CV and associated QCM mass data in coffee at 1.36 %TDS. Began at -1.2 V vs. Ag/AgCl and scanned oxidatively, as opposed to starting at 0 V vs. Ag/AgCl and scanning reductively to start. QCM shows no mass gain during the initial oxidative sweep (orange trace). Mass is only accumulated following the onset of HUPD during the first reductive sweep (blue trace) since a build-up of adhered H+ on the electrode surface is necessary for side reactions that lead to a detectable mass gain. . . . . . . . . . . . . . . . . . . . 148 26 LIST OF TABLES Table Page 1. Characteristics of in-house roasted coffee used in this work. . . . 87 A.1.Details of the commercially sources coffees. * entries were not measured. . . . . . . . . . . . . . . . . . . . . . . . 130 A.2.Roast profile details of the in-house roasted coffees. Target curves are available for download from DOI:10.6084/m9.figshare.23277320. . . . . . . . . . . . . . . . 134 C.1. Roasting conditions and their associated Agtron numbers. Grind setting 2 was used on an EK-43 to obtain the ground Agtron values. . . . . . . . . . . . . . . . . . . . . . . . . . . 139 C.2. %TDS, resistance, and pH values for coffee samples used for cyclic voltammetry. . . . . . . . . . . . . . . . . . . . . . . . . 141 27 CHAPTER I INTRODUCTION 1.1 Making Measurements on Coffee The coffee industry involves a complex supply chain that includes millions of smallholder farmers, international traders, roasters, and retailers. Global revenue for hot coffee drinks in 2023 alone totaled 445.5 billion USD.1 The keystone of the coffee market is the consumers’ enjoyment of the end product – brewed coffee. Intense care is taken throughout the coffee bean’s life cycle to ensure the tastiest flavors reach the end brew. Many factors along the way influence the chemical composition which give rise to both desirable and undesirable flavors. To fully understand how these different chemical compositions affect the brew, we must have a way to measure the different chemicals present. The coffee industry relies on a measurement of total dissolved solids (%TDS), popularized as a quality metric by the Coffee Brewing Institute in the early 1960’s, to quickly determine if the brew is in the range that is considered “ideal” (see Figure 12,3). Despite the consumer enjoyment depending on the chemical composition of the brew, %TDS fails to capture these nuances that lead to different flavor profiles. Most coffee studies employ separation techniques to quantify the hundreds of compounds present with high fidelity. Others focus on only a few molecules that can be targeted by diluting the coffee far past what a consumer would classify as coffee. The work presented herein has been conducted on coffee under normal consumption conditions and is the first to relate compositional differences to the industry standard of %TDS. There are many steps that impact flavor and composition of brewed coffee, starting with the bean species. The two main species of coffee that are cultivated and sold commercially are Coffea arabica, known as Arabica, and Coffea canephora, 28 Figure 1. Coffee Brewing Control Chart originally designed by Lockhart in 1957. Reprinted with permission from Sci Rep 10, 16450 (2020), doi.org/10.1038/s41598- 020-73341-4. 29 known as Robusta4. Brazil has been the top coffee producing country since 1852 and specializes in Arabica varieties while most of the world’s supply of Robusta coffee comes from Vietnam.5 The World Coffee Research website provides detailed profiles of 55 Arabica varieties and 47 Robusta varieties.6,7 Many of the varieties can be distinguished from one another based on physical characteristics such as plant size, bean size, optimal growing altitude, and susceptibility to diseases such as coffee leaf rust, coffee berry disease, and nematodes. Robusta varieties are generally more robust in terms of disease resistance and tolerance to differing climate conditions.8,9 However, Robusta coffee also has higher caffeine content in both green and roasted beans which contributes to the increased bitter flavor.10 In an attempt to combine the disease resistance of Robusta with the cup quality of Arabica, some producers graft Arabica plants into Robusta roots, though this information rarely reaches the consumer. Due to its preferred flavor profile, the specialty coffee industry focuses almost exclusively on Arabica coffees. When the coffee plant cherries are ripe, they are harvested and processed (see Figure 211). This processing step is crucial in determining the end flavors that will reach the cup. Coffee beans considered “naturally” processed are laid out with the cherry mucilage still intact and allowed to ferment while drying, leading to distinct fruity flavors in the brewed coffee. Those that are “washed” have the cherry mucilage removed and are purposefully fermented without their natural fruit before drying. Washed coffees tend to taste strongly of chocolate, caramel, and tea. These are the flavors many American consumers expect to find in their morning brew. In the middle of washed and natural lies the “honey” process in which the cherries are pulped to remove the outer fruit skin, but some mucilage is allowed to remain coated on the beans during 30 fermentation. As one might expect, these coffees have more fruity flavors than washed coffees but are less fruit-forward than natural coffees. The processing step is in part controlled by the economy of the producing region. Washed coffees are more expensive in terms of energetic input and water usage.12 As the specialty coffee industry has continued to expand, so too has innovation within the fermentation techniques which are applied to coffee. It is now common to find commercially available coffees that have undergone anaerobic fermentation, carbonic maceration, or co-fermentation in which a flavor addition, such as cinnamon sticks or coconut, has been incorporated into the fermentation step. All of these processing decisions impact the end flavor. 1.1.1 Separation Techniques Applied to Coffee Analysis. A common chemistry technique for separating, identifying, and quantifying compounds from a complex mixture is chromatography. Chromatography is based on different components partitioning between a stationary phase and a mobile phase. The stationary phase is typically a column packed with fine particles, while the mobile phase is a liquid or gas that flows through the column.13 Chromatography alone is a separation technique and must be adequately coupled with a detector to be used as an identification technique. Common detectors used with liquid chromatography are mass spectrometry, UV-Vis, fluorescent, refractive index, or diode array detectors. Each of these detectors works by recognizing known properties of the compounds, such as their individual refractive indices or UV-Vis absorption wavelengths. In these cases, some prior knowledge about the compounds is useful so that the detector can be chosen with appropriate settings. Gas chromatography is often coupled with flame ionization or mass spectrometry detectors which ionize the compounds in predicted ways and are connected to large 31 Figure 2. Flow chart of various specialty coffee processing methods. Reprinted with permission from LWT 172 (2022) 114245, doi.org/10.1016/j.lwt.2022.114245. 32 libraries of known compounds for identification.14 Gas chromatography is more often used for volatile compounds while liquid chromatography is best with soluble compounds. Brewed coffee has its share of both soluble and volatile compounds, so it is unsurprising that both liquid and gas chromatography techniques have been used for its analysis. Traditional gas chromatography uses a liquid sample that is volatilized upon entry into the instrument. However, headspace gas chromatography coupled with mass spectrometry (headspace GCMS) is a variation that involves using gaseous sample. In Figure 315, Jackels et al. used headspace GCMS to compare volatiles obtained from whole roasted bean samples. The whole beans were heated in a sealed container via a 60°C water bath for 60 minutes. After this equilibration time, gas was extracted from the sealed container (hence the name “headspace”) and underwent chromatographic separation. The two samples pictured show differences in the overall number of peaks present and especially in peak height of several of the labeled peaks. These two coffee samples, among others, were then brewed via cupping technique and tasted by professional coffee tasters. The top panel of Figure 3 shows a sample which was determined not to have a potato-taste defect while the bottom panel shows a sample that was found positive for potato-taste defect.15 Additional work has used GCMS for purposes such as determining metabolite composition of Indonesian coffees,16 aroma component comparison among coffee blends,17 and determination of polycyclic aromatic hydrocarbons present in coffee brew.18 Liquid chromatography is a natural choice for coffee analysis since most work is performed on brewed coffee. In fact, high-performance liquid chromatography (HPLC) systems have been used to analyze coffee samples for 33 Figure 3. Headspace GCMS of volatile compounds from 70.0g of whole beans heated in 60°C water bath for 60 minutes. Comparison of two samples that were characterized as having or not having the potato-taste defect, common in Rwandan coffee, via cupping after GCMS data was collected. Reprinted with permission from J. Agric. Food Chem. 2014, 62, 42, 10222-10229. Copyright 2014 American Chemical Society. 34 such purposes as determining tocopherol and triglyceride content,19 cholorgenic acid content,20 caffeine content,21 organic acid content,22 and bioactive compound content.23 HPLC analysis of coffee samples is effective enough that Stathakis et al. proposed using it for an educational lab. The lab is reliable enough that HPLC alone is not the focus of the experience, but instead the authors encourage its use for teaching post-column derivatization and statistical analysis.24 Figure is 424 from Stathakis et al. and shows the HPLC chromatogram of two different commercially available coffee pods. The two samples are very similar in overall compounds present, as indicated by the largely overlapping peaks, but in different concentrations, as shown by the differing peak heights. HPLC, therefore, proves itself to be a highly sensitive technique which can inform on the minute differences from one brew to another. However, there is evidence that compounds within the complex matrix of brewed coffee interact with one another, such as the aggregate that forms between caffeine and chlorogenic acid,25,26 which are inherently missed in analysis performed involving separation techniques. 1.1.2 Electrochemistry Techniques for Coffee Analysis. Electrochemistry is another approach used in the analysis of liquid brewed coffee but, unlike chromatography techniques, it does not involve compound separation. The field of electrochemistry studies chemical processes which involve the movement of electrons, particularly those transferring between the surface of an electrode (typically a metal) and a molecule present in the electrolyte solution.27 These methods measure the electrical properties associated with oxidation, the loss of electrons, and reduction, the gain of electrons. For an electrochemical measurement to be viable, the system must allow for both the movement of electrons through the potentiostat (used to control the potential applied to the 35 Figure 4. HPLC chromatogram performed on two samples of commercially available undiluted coffee brewed via coffee pods at a 1:6 ratio. (coffee pods 5g in, 30 mL out aka 1:6 ratio). Reprinted with permission from J. Chem. Educ. 2023, 100, 4, 1564–1570. Copyright 2023 American Chemical Society. 36 electrode surface) as well as the movement of charged species in the liquid sample. It is imperative that the liquid contains a supporting electrolyte to ensure that it is conductive and able to move charged species appropriately.28 Resistance values observed in the coffee samples examined in the studies performed in chapters 3 and 4 of this work were between 500–2000 Ω, which was found to be sufficiently conductive for electrochemical measurements without an additional electrolyte. By analyzing the current response produced by the movement of electrons as the applied potential changes, one can determine the concentration of chemical species, gain insight into reaction mechanisms, and assess the composition of liquid samples. Electrochemical techniques like square wave voltammetry (SWV) and differential pulse voltammetry (DPV) can provide highly sensitive and specific measurements of trace amounts of substances. SWV applies a series of potential pulses, via the potentiostat, in a staircase manner, with each pulse consisting of a forward and reverse potential step (see Figure 529). The current is measured at the end of each forward and reverse pulse, and the difference in current is plotted against the potential. Similarly, DPV superimposes a series of regular potential pulses on a linear potential sweep. The current is measured just before the pulse and at the end of the pulse, and the difference in current is plotted against the potential. It is through the application of these potential pulses that SWV and DPV techniques minimize background noise and achieve high sensitivity.29 Figures 630 and 731 show the application of SWV and DPV to green coffee extract by Sˇeruga and Tomac. In both publications, the green coffee extract was spiked with 10−5M amounts of known isomers of chlorogenic acids and the resulting peak heights increased as expected. While these works highlight the use of both electrochemical techniques on specific compounds present in brewed coffee, no 37 proof that chlorogenic acid is the only compound contributing to these peaks is provided. SWV has also been used for the determination of caffeine32 and the presence of adulterants in coffee samples.33 Along with determination of caffeine,34 DPV has also been used to measure the total antioxidant capacity in coffee, juices, and wine.35 SWV is commonly used in biological analysis for reversible or quasi- reversible reactions while DPV is better suited for irreversible reactions.29,36–38 A common technique for examining the electrochemical reversibility of a redox active reaction is cyclic voltammetry (CV). CV involves sweeping the potential of the working electrode linearly back and forth between two set values at a given scan rate. The current is continuously measured and plotted against the potential, producing a cyclic voltammogram (see Figure 5). While less sensitive than SWV and DPV, CV provides detailed information on redox processes and reaction mechanisms, revealing both kinetic and thermodynamic information through careful interpretation of the voltammograms. Despite the broad applications of CV, it has been an underutilized technique for studying brewed coffee. Literature examples of its use for coffee include estimating phenolic content,39 evidence of acrylamide contamination,40 and preliminary electrode surface characterization followed by SWV for caffeine quantification.41,42 These chromatographic and electrochemical techniques inform researchers about the compounds present in coffee samples, but they lack a connection to the consumers’ experience of brewed coffee. A typical cup of pour-over coffee is brewed with a ratio between 1:15 and 1:18, coffee to water. In the study performed by Kilmartin and Hsu, the coffee was extracted at a ratio of 1:100 by weight and then further diluted 50-fold into a phosphate buffer as the supporting electrolyte before CV analysis of phenolic content. Samples used for SWV and DPV by Sˇeruga and 38 Figure 5. Visual description of applied potential versus time and associated current versus potential outputs for cyclic voltammetry (top, A), differential pulse voltammetry (middle, B), and square wave voltammetry (bottom, C). Reprinted with permission from ECS Sensors Plus, 2024 3 027001, doi.org/10.1149/ 2754- 2726/ad3c4f, and Creative Commons https://creativecommons.org/licenses/by-nc- nd/4.0/ 39 Figure 6. Square Wave Voltammetry of unroasted (aka green beans) coffee extracted at a 1:100 ratio and diluted 50-fold (green trace). Traces in blue, red, and black are each spiked with the same amount of a different chlorogenic acid isomer. Reprinted with permission from Int. J. Electrochem. Sci., 9 (2014) 6134 – 6154, doi.org/10.1016/S1452-3981(23)10876-5, and Creative Commons https://creativecommons.org/licenses/by-nc-nd/4.0/. Tomac in Figures 6 and 7 also used a 1:100 brew ratio with a 50-fold dilution. Stathakis et al. samples had a brew volume more suitable for espresso coffee (30 mL) but used a 1:6 brew ratio as opposed to a more typical espresso ratio of 1:2.5. Many researchers may choose to use more dilute brew ratios in part to ensure that data remains in the dynamic linear range for their instrumentation. While a comparison of brew ratios gives some insight into the apparent concentration or strength of the brew, the coffee industry standard is to use a refractometer to determine the %TDS. This measurement is notably lacking in the literature 40 Figure 7. Differential Pulse Voltammetry of diluted green bean extract 1:100 ratio (opposed to 1:16.5 we use) (black trace, a). Traces b through d were each spiked with different chlorogenic acid isomers. Reprinted with permission from Int. J. Electrochem. Sci., 11 (2016) 2854 - 2876, doi.org/10.1016/S1452-3981(23)16146-3, and Creative Commons https://creativecommons.org/licenses/by-nc-nd/4.0/. on coffee studies. Even the plethora of coffee studies employing human sensory panels43–51 fail to include %TDS measurements. To successfully analyze minute differences in brewed coffee composition, one must first ensure that reproducible samples can be prepared. This work begins with an examination of the electrostatics of coffee grinding and its impact on brewing. The following chapters detail work that is the first to report %TDS in conjunction with other chemical analysis, namely CV, on brewed coffee samples. It is also the first to use electrochemical analysis on as-brewed strength coffee, with no addition of a supporting electrolyte or dilution of the sample. Moreover, this 41 work demonstrates the use of multi-factor design of experiments to gain insight into brew factor interactions that cannot be captured through single factor experiments. These factor interactions are then leveraged for economical optimization of brew parameters and used to understand why a simple brewing scale-up does not result in the same composition and flavor profile. 1.2 Co-Authors and Publication Details 1.2.1 Chapter 2, Electrostatics of Coffee Grinding. The work presented in Chapter 2 of this dissertation has previously been published52,53 and is co-authored with Joshua Méndez Harper, Connor S. McDonald, Elias J. Rheingold, Lena C. Wehn, Elana J. Cope, Leif E. Lindberg, Justin Pham, Yong-Hyun Kim, Josef Dufek, and Christopher H. Hendon. JMH acknowledges support from start-up funds at Portland State University. CHH acknowledges the support from the National Science Foundation under Grant No. 2237345 and support from the Camille and Henry Dreyfus Foundation. This work was supported by the Coffee Science Foundation, underwritten by Nuova Simonelli. The work was enabled by donations of green coffee from Farmers Union and Finca La Ilusión. We are grateful to Ikawa for providing the roaster, Pentair for providing the reverse osmosis water filtration system, and Tailored Coffee, Eugene, OR for the use of the EK43. 1.2.2 Chapter 3, Determining Roast Degree. The work presented in Chapter 3 of this dissertation will be published and is co-authored with Doran Pennington, Zachary S. Walbrun, Lena C. Wehn, Elias J. Rheingold, Joshua Williams, and Christopher H. Hendon. CHH acknowledges the support from the National Science Foundation under Grant No. 2237345 and support from the Camille and Henry Dreyfus Foundation. 42 This work was supported by the Coffee Science Foundation, underwritten by Nuova Simonelli. We are grateful to Pentair for providing the reverse osmosis water filtration system, and Tailored Coffee, Eugene, OR for the use of the Ikawa Pro50 and for providing green coffee. 1.2.3 Chapter 4, Impact of Brewing Parameters. The work presented in Chapter 4 of this dissertation will be published and is co-authored with Benjamı́n Alemán and Christopher H. Hendon. CHH acknowledges the support from the National Science Foundation under Grant No. 2237345 and support from the Camille and Henry Dreyfus Foundation. This work was supported by the Coffee Science Foundation, underwritten by Nuova Simonelli. We are grateful to Pentair for providing the reverse osmosis water filtration system. 43 CHAPTER II ELECTROSTATICS OF COFFEE GRINDING 2.1 Moisture-controlled triboelectrification during coffee grinding 2.1.1 Author Contributions. The study was conceived by JMH, JP, REB, JD, and CHH. Coffee was roasted by LCW and JMH. Laser diffraction measurements were performed by EJR, CSM, REB, LEL, and JMH. Triboelectric measurements were performed by JMH, LEL, and REB. Espresso measurements were performed and interpreted by EJC and CHH. JMH and CHH wrote the manuscript and all authors contributed to the final version. 2.1.2 Experimental procedures. Coffee roasting was performed using an Ikawa Pro100. The roast profiles are available for download.54 The coffee color/roast degree and internal water content of both commercially sourced and in- house roasted coffee were measured using The Dipper KN-201 and the Roastrite RM-800, respectively. The Roastrite compensates for variations in ambient temperature and humidity. Coffee grinding was performed on a Mahlkönig EK 43 flat burr grinder using stock coffee burrs. The burrs were aligned using the Mahlkönig burr alignment tool. A grind setting of 0.0 was set to when the burrs were brought together to create a chirping sound. The axial, radial, and angular alignment was measured by applying a marker to the outermost edges of the burrs and then bringing the burrs to grind setting 0.0. The burrs were sufficiently aligned where all radial marker had been rubbed away, and there was homogeneity of the grind distribution from laser diffraction particle size analysis. Laser measurements were performed on a Malvern Mastersizer 2000 with the solid-particle feed system, Scirocco 2000. Vibration feed rate and air pressure were set to 60 % and 2 bar, respectively. The SOP parameters 44 were three measurements per aliquot, 2-second delay between measurements, an estimated refractive index and absorption of 1.59 and 0.1, respectively (similar to chocolate), measurement time of 10 seconds, 10,000 measurement snaps, background time of 5 seconds, and 5000 background snaps. Raw data for plots presented in Figures 5 and 6 are available for download.54 Twice ground coffee in Figure 12 was obtained by first grinding at 2.0, then at 6.0. These settings were empirically determined by isolating when the particle size distribution was unchanged during the second grind (see Supplementary Information, Figure S4 for further details). Grind settings less than 6.0 showed a minor increase in fines from random fracturing events of large particles. The raw triplicate data can be found in Ref.54. Photographic images were collected using a Keyence VHZ-100UR microscope. Charge-to-mass quantification was performed using a custom build Faraday cup, machined to fit the chute of the EK 43 grinder. The charge on particles entering the cup was then measured by a Keithley 614 electrometer operating in Coulomb mode with a range setting of ± 2 nC (see Figure 8c). The coffee collected in the Faraday cup was then weighed using a generic laboratory scale to obtain its mass. The mass anc charge were then used to compute charge-to- mass ratio. The same Faraday cup setup was employed to measure the triboelectric response of whole coffee beans rolling down a coated vibrating ramp presented in Figure 8a. Particle charge polarity measurements were performed using the electrostatic separator following a configuration presented in a previous study.55 The system shown in Figure 8c consists of two sub-parallel 1 meter long electrodes with a potential difference of 16.4 kV. The electric field separates particles by charge 45 polarity, and particles are collected in negative, neutral, and positive bins at the base of the separator. Water solutions used for fractoelectric charge reduction were constructed from reverse osmosis water from a Pentair Conserv 75E, and mineralized using sodium chloride obtained from Sigma Aldrich. A pipette was used to introduce water onto whole bean coffee, and the coffee was shaken in a sealed container to ensure homogeneous distribution. Samples were prepared in triplicate. Whole bean coffee was stored in H2O impermeable vacuum bags, and kept at -20oC. The coffee was allowed to equilibrate to room temperature in the vacuum bag before grinding. Humidity measurements were performed in a glove box, and the grinder was allowed to reach equilibrium with the atmospheric water. Coffee was not allowed to equilibrate, to isolate the role of internal versus external water. Otherwise, all experiments were conducted in air at 25±2oC, 20-45%RH, and 101±1 kPa. Espresso was prepared using a Nuova Simonelli Black Eagle prepared at 7 bar of static water pressure, with 94oC water, and a fixed brew ratio of 18 g of coffee to produce 45.0 g of espresso. The grounds were tamped using PUQpress Q1, set to apply 196 N, and a normcore 58.5 mm diffusing screen was added to the top of the compacted bed. Flow rate data was computed from the gravimetric change measured at the scale. 2.1.3 Progress and Potential. Coffee grinding produces large quantities of static charge due to both fracturing and rubbing. Charge causes particle aggregation and discharge, a familiar problem in industrial coffee production. This study demonstrates that the magnitude of charge depends on the roast profile, and more importantly, the internal moisture content of whole 46 bean coffee. In an effort to control the charge, we demonstrate that the addition of external water mitigates its accumulation during grinding and promotes particle declumping. Notable differences in brew parameters are achieved. Implementation of our findings directly addresses a key issue of static accumulation and particle clumping, and highlights the challenges of making physical property predictions based on bean color. 2.1.4 Introduction. Triboelectrification is the physical process where materials acquire surface charge from frictional interactions at their interfaces.56 The magnitude of charge depends on the interfacial material composition57 and can be harnessed in emergent technologies for energy generation.58–62 The mechanism of electrostatic accumulation is complex, and is further obscured in granular materials where collisions are sufficiently energetic to cause fracturing. In this ”fractoelectric” regime, crack initiation and propagation is thought to charge particles through transfer of electrons and/or ions at the hot crack interface.63,64 Whether a material’s charging is dominated by tribo- or fractoelectrification, fracture-generated granular flows often comprise of particles whose surface charge density may exceed the theoretical maximum value of 27 µC m-2 65–68, or charge- to-mass ratios in the range of 0.1 to 100 nC g−1.69,70 There remains fundamental interest in studying the mechanism and magnitude of charging, and methods to control the process, in particular to mitigate spurious effects such as electrostatic discharges and agglomeration within industrial settings.71–75 The electrification of chemically complex materials (e.g. foodstuffs, wood)76 present unique and complex problems in materials science. While most food is not subject to fracturing resulting in appreciable electrical charge formation, coffee is a paragon of material complexity, as all coffee is ground, and the chemical 47 composition of whole-bean coffee depends on numerous factors (e.g. roast, origin).77–79 The effects are particularly emphasized in espresso preparation, where the coffee must be ground fine, imparting large amounts of static charge. Here, we use coffee to provide fundamental insights into the electrification processes in organic materials composed of a variety of molecules. We demonstrate that i) conventional coffee parameters – roast, internal water content, grind setting – dictate the charging of roasted coffee and offer an explanation why; ii) both triboelectric and fractoelectric processes occur during grinding, with the majority of charging coming from fracturing events, and iii) charging depends on the coffee bean’s internal and external water content, with higher water content suppressing charge accumulation. For industrial-scale operations, uncontrolled coffee charging can cause clumping, leading to product heterogeneities and clogged conduits. At the brewing level, aggregation may also affect liquid-solid accessibility,80 leading to inhomogeneous extraction, and unpredictably unpleasant espresso.81 In the context of both understanding the fundamentals of triboelectrification, and bolstering our efforts towards brewing more reproducible and sustainable coffee, this paper offers strategies to control the charging of coffee particles and posits opportunities therefrom. 2.1.5 Electrifying Coffee. We first sourced numerous commercially roasted coffees, canvasing many major producing countries and coffee processing paradigms. These coffees are further categorized by their processing method – natural (N), washed (W), decaffeinated (D) – and are all single origin unless designated as a blend (B). Full details of the commercial coffees are presented in Table S1. The color of roasted coffee can be quantified using a spectrophotometric method that places coffee on the ”Agtron Gourment Scale”.82 The scale ranges 48 4.1 kV c Electrification of coffee by grinding + +++ _ _ _ _ __ + fractocharging crack propagation grounds exit stainless steel electrodes negative grains neutral grains positive grains -4.1 kV Fd Fe Fg Faraday cup d Charge separation in electric field to electrometer grounds exit + +++ + ++ + + + ++ + _ ______ _ metal burr tribocharging a 0.50 0.25 0.00 0.25 0.50 Coffee charge-to-mass ratio [nC g–1] PVC Mylar Steel Paper Aluminum Nylon Glass Heterointerfacial triboelectric polarization of whole-bean coffee - + Coffee beans rolling down vibrating ramp Faraday cup charge amplifier coated ramp b Figure 8. Electrification of coffee beans and particles. a) Whole coffee beans accumulate charge when rolled down a vibrating ramp coated in a variety of materials. b) These surfaces materials can be arranged according to their capacity to charge whole beans. Here, Starbucks Blonde Espresso Roast weakly charges against steel, while glass and nylon result in positive charging, and plastics like PVC and Mylar lead to negative charging. c) During fracture, coffee particles accumulate charges from the burr-coffee interface and coffee-coffee rubbing (tribocharging), as well as fracture points (fractocharging). d) After grinding, the charge is quantified by alloying the particles to fall between two sub-parallel electrodes with a potential difference of 8.2 kV. The electric field separates particles by charge polarity, and particles are collected in negative, neutral, and positive bins at the base of the separator. 49 from 0 (black/carbonized) to 150 (green/unroasted), with most specialty coffees falling within the range of 40-90 as measured on our spectrophotometer. Examples are presented in Figure 9. To assess surface charging of whole beans, we first selected Starbucks Blonde Espresso Roast (a dark roast coffee: Agtron 65.2, water content 1.3%) and rolled it down a vibrating ramp coated in various materials. At the end of the ramp, beans were collected in a Faraday cup where the charge and weight are recorded, Figure 8a. Coffee generally charges poorly against metallic surfaces but may acquire relatively large charge when contacting dielectrics. To assess the charging of coffee against other materials, we created a series of heterointerfaces between whole been coffee and other common materials found in coffee grinding environments. From Figure 8b, it is found that coffee charges positively against plastics such as polyvinyl chloride (PVC) and biaxially-oriented polyethylene terephthalate (Mylar, a material widely used in coffee grinder technologies), but acquires negative charge when rubbed against glass and nylon, Figure 8b. Coffee gains almost no charge against office paper. These data indicate that coffee is similar to wood, cellulose, and grain in the triboelectric series.57 Coffee beans must be ground — a process that results in significant electrification. The process typically produces particles with sizes ranging from 100 nm to 2 mm. The distribution is controlled by grind setting and bean temperature.83 In flat-burr grinder architectures such as the configuration housed within the Mahlkönig EK 43 (a grinder with steel 98 mm burrs, Figure 8c), the grind setting is determined by the separation between rotating metal plates. Finer grind settings result in more fracturing events, longer coffee-burr contact time, the production of more fines (sub 100 µm particulates), and smaller boulders (post 100 50 µm particulates). While grinding can lead to minute spark discharges, especially if the grinder is not grounded, the primary consequence of electrification is the formation of particle aggregates held together by electrostatic forces. The net charge acquired by a coffee sample is measured by placing a Faraday cup under the grinder chute, Figure 8c. An electrometer then reports a voltage proportional to the particle charge, with a sensitivity of 10 nC V-1. Although we used 1 g (5-10 beans) of coffee across all experiments, the amount of ground coffee that enters the Faraday cup varies between experiments (some material is retained in the space between the burrs). To account for this variation, we normalized the measured charge by the mass collected in the cup, allowing us to calculate a cumulative charge-to-mass (Q/m) ratio (i.e. the total charge of the coffee in the cup). Furthermore, we performed a different experiment to separate the positive, negative, and neutral particles. By replacing the Faraday cup with an electrostatic separator consisting of two sub-parallel plates held at a potential difference of ∼8.2 kV and grinding 10 g of coffee, the negatively-charged grains drift towards the positive plate, the positive grains towards the negative plate, and net neutral particles fall straight down, Figure 8d. Imaging and laser diffraction particle size analysis can then be used to determine whether polarity impacts size distributions, in addition to determining a Q/m for each bin. 2.1.6 General trends in electrification of grinding commercially roasted coffee. Using this experimental setup, we first examined three Mexican coffees, Figure 9a. Those samples showed net positive, net negative, and both positive and negative charging. But those coffees were roasted by different roasters, to different colors and the data suggests, perhaps unsurprisingly, that origin alone does not dictate the polarity of charge. Figure 9b summarizes the charging 51 40 60 80 100 –100 –50 0 50 2 0 2 0.4 0.2 0.0 2 0 2 time [s] 12.5 10.0 7.5 5.0 2.5 0.0 2 0 2 0 2 4 C ha rg e [n C ] a Tacámbaro (N) C ha rg e [n C ] C ha rg e [n C ] Charging regimes of three Mexican coffees Mané (B) Temascaltepec (W) Charging in commercially sourced coffees Charging as a function of moisture content 102 103 Size [μm] 0.00 0.05 0.10 0.15 R el at iv e fre qu en cy positive negative 0.00 0.10 0.20 102 103 Size [μm] 0 5 Time [s] 0 2 C ha rg e [n C ] Kolla (W) npos = 672 nneg = 660 Mané (B) npos = 167 nneg = 1098 Particle size distributions of oppositely-charged grains collected by the electrostatic separatord averages averages Agtron color C ha rg e- to -m as s ra tio [n C g –1 ] decaf (D) blend (B) washed (W) natural (N) 1.0 1.5 2.0 2.5 3.0 –100 –50 0 50 Water content [%] error in water % ~ ± 0.1%error in Agtron color ~ ± 1 washed, roasted in-house R2 = 0.350 R2 = 0.41 Agtron 70.5 Agtron 57.7 Agtron 74.2 b c Figure 9. Charging regimes of commercially sourced coffees a) Three Mexican coffees – Tacámbaro (N), Mané (B), and Temascaltepec (W), show three possible charging regimes. b) Charge-to-mass ratio as a function Agtron color for a variety of coffees ground at setting 2.0. There is no strong relationship between charge magnitude and polarity, and coffee processing method. Confidence bands are plotted at 90%). A polarity switch is observed for darker coffees (Agtron = ≤ 70). The magnitude of negative charge continues to increase with darkness. c) Internal moisture - a property proportional to roast color - results in more positive charging. Moisture content is a slightly better predictor of charge-to-mass ratio than color. d) Examination of the particles collected for two representative coffees, Kolla (W, a positive charging, coffee) and Mané (a negative charging coffee), reveals that positively charged particles are generally smaller than the boulders, as denoted by the average arrows in blue. 52 behavior of ∼30 coffees as a function of Agtron color. Although we observed both positive and negative charging, the charge-to-mass ratio magnitudes of positive samples are generally smaller (<50 nC g−1) than that of negative coffees (up to 120 nC g−1). We observed a weak relationship between roast color and charging, with positive charging occurring only at Agtron values exceeding 70. Post-roast internal moisture content showed a slightly better relation to the sign and magnitude of charge, Figure 9c. Here, the transition from negative to positive charging occurs when the water content increases above ∼2 % by mass. The substantial scatter in Figure 9b and c likely reflect the fact that there are infinite ways to roast to an arbitrary Agtron color and resultant chemical composition. For example, one could obtain a dark coffee by roasting at low temperature for prolonged time, or roasting hot and fast. Additionally, the internal water content of pre-roast green coffee varies with time, and depends on storage environment.84 Both variables have significant impact on roast chemical composition,85 and are also known to affect the resultant beverage properties.86 However, the relationship between water content in roasted coffee and charging transition from negative to positive is somewhat surprising, given other reports of decreasing charge-to-mass ratios with increasing moisture.87 One possibility is that the polarity flip reflects degrees of strain at fracture.88 In that work, lower degrees of strain were associated with negative charging. Since dark coffee is more brittle, they may support less strain before failure than their more ductile, light roast counterparts.89 Another possibility is that the water is affecting the physical properties of the coffee, which we will discuss later in the paper. Beyond performing net charge-to-mass measurements, we also separated particles by polarity and then sized them photographically. An example particle 53 size distribution of positive and negative coffees is presented in Figure 9d. These data suggest that the boulders carry negative charge independent of roast, process, and origin. High-speed videography reveals that these boulders exit the grinder first, explaining why we occasionally observe negative charge entering the Faraday cup at the onset of grinding, even if the net charge is ultimately positive (see inset in Figure 9d and the third panel in Figure 9a). Fine particles (<100 µm) tend to be biased slightly toward negative charging, or have comparable positive and negative abundances. The distribution of charge on mid-size particles is more complicated. We observed a peak in the abundance of positively-charged particles at diameters of 100-300 µm, regardless of net charge polarity. For light roasts, the abundance of particles in this size range exceeds that of negative grains. For coffees with lower Agtron values, the positive maximum is still present, but the quantity of negatively-charged particles outnumbers positive particles at all sizes. In other words, the polarity of the particles in this intermediate range appears to dictate the overall polarity of a coffee’s net charge-to-mass ratio. This size range also corresponds to the sizes of particles typically produced for espresso format brewing, adding to the ever-increasing challenge of brewing reproducible espresso.81 Regardless of relative abundances, positive particles generally have smaller mean diameters than negative ones across all coffees (noted in Figure 9d by the average arrows). This size-dependent bipolar-charging provides insight into the basic electrification mechanisms operating during grinding. The observation that the larger particles charge negatively is consistent with the charge separation described by James and coworkers65 in the context of volcanic pumice fracture. Although an equal number of positive and negative surfaces are likely created during any given fracture event, those authors hypothesize that subsequent ion 54 scavenging processes lead to particles of different sizes concentrating opposite polarity charges. This bias – larger, negative particles and smaller, positive particles – is opposite to that often reported in purely triboelectric systems (i.e. processes with little to no fracture).55 In those contexts, charge segregation has been attributed to the exchange of trapped electrons, polarization,90 and hydrated ions.91 For now, we can summarize that dark roast coffees appear to charge more negatively than light roasts, large particles carry negative charge, and that commercially roasted coffees charge in an a seemingly unpredictable way. 2.1.7 Isolating the impact of roast profile. The data presented in Figure 9 points to a general challenge in the coffee industry: the words ”light,” ”medium,” and ”dark” describe the end color and to some extent provide a touch point for flavor profile.92 But roast color does not yield sufficient information about the chemical composition and resultant tribocharging. The large amount of scatter in the data likely reflects the compounded effects of origin and processing93 in addition to the temperature profile used to take it from green to brown.94,95 Many commercial coffee roasters treat their roast profiles as proprietary, and it is impossible to back-calculate the precise profile from examination of only the roasted whole beans. There is academic value in standardizing roast profiles across the industry, thereby allowing for direct comparisons between coffees. But we are not advocating for this on the industrial scale, as that would sterilize an artisanal aspect of the industry. Instead, we developed our own profiles with the aim of isolating roast through the development of systematically “darker” coffees. Noting that pre-roast internal moisture content is known to dictate roast induced swelling and other properties,96 we sourced coffees with moisture content representative of conventional specialty coffees. We obtained a green Ethiopian 55 1.0 1.5 2.0 2.5 3.0 Water content [%] -100 -75 -50 -25 0 25 60 70 80 90 Agtron color -100 -50 0 50 RMSE = 19.37 RMSE = 7.93 300 400 500 600 700 Roast time [sec] -100 -75 -50 -25 0 25 a Representative roast profiles b Q/m ratios as a function of roast time c Q/m ratios as a function of Agtron color d Q/m ratios as a function of water content Q/m = aA + b Q/m = a (1 – e–bW) + c 0 200 400 600 Time [s] 50 100 150 200 Short Morse Long Morse Short YirgZ (W) Short Yogondoy (W) Long YirgZ (W) best fit to YirgZ databest fit to YirgZ data Run 1 Run 6Run 1 Run 6 Positive region R oa st er h ea ds pa ce a ir te m p. [º C ] C ha rg e- to -m as s ra tio [n C g –1 ] Morse time C ha rg e- to -m as s ra tio [n C g –1 ] C ha rg e- to -m as s ra tio [n C g –1 ] Figure 10. The effect of roast profile on charging a) Two sets of roast profiles were explored: Short Morse (blue) and Long Morse (purple). b) Charge-to-mass ratios as a function of roast time for a representative Ethiopian coffee (YirgZ) shows that charging increases with darker roasts. Prolonged Morse transitions to the negative charging regime at a slower rate than the shorter Morse. An additional Mexican coffee (Yogondoy) was roasted using the short Morse profile, yielding a similar electrification behavior to the Ethiopian coffee. c) Charge-to-mass ratios for Ethiopian and Mexican coffees as functions of Agtron color. d) Charge-to-mass ratios for Ethiopian and Mexican coffees as functions of residual water content. 56 coffee from the Yirgacheffe region, ”YirgZ”, featuring 12% internal moisture at the time of roasting. This coffee was roasted using an Ikawa Pro100 to achieve five different roasts by systematically increasing the terminal headspace air temperature and time by 2 ◦C and 60 s, respectively (Figure 10a in blue). A sixth coffee was generated by adding 8 ◦C and 180 seconds to the fifth profile. Additionally, a second roast profile was employed, differing by a parameter we call the ”Morse time” (the empirical time taken for the headspace thermocouple to read a temperature equal to the initial temperature in an Ikawa roaster). The long Morse profiles were constructed in increments of 3 ◦C and 60 s, Figure 10a, purple. Figure 10a shows the profiles for the shortest (solid) and longest (dotted) roasts. In summary, 12 dissimilar roasts were achieved; their details are presented in Table S2. After degassing for 24 h, the coffees were ground at a setting of 2.0. Figure 10b shows the charge-to-mass ratio of the YirgZ as a function of roast time for the two profiles. In both cases, the resultant charging is positive for short roasts (i.e., lighter coffees) with a transition to negative charging as the roast time increases. However, we observe an earlier transition to negative charging for the short Morse roast. To test whether these behaviors are specific to the YirgZ, we repeated the short roast experiment with a washed Mexican coffee, Yogondoy (9% internal moisture at time of roasting). The result of this ancillary experiment is presented in Figure 10b (grey squares). Within error, the charge-to-mass ratios of the Ethiopian and Mexican coffees roasted using the same roast profile are comparable. Such congruence hints that the product of the roast, more than the characteristics of the green coffee, ultimately determines the charging behavior of coffee when ground. 57 We can further test the hypothesis that roast is a primary control on charging by casting our data in terms of roast color and residual water content, Figure 10c and 10d, respectively. In agreement with the data presented in Figure 9, we observed weak linear relationship between color and electrification, with a transition from positive to negative charging at Agtron colors in the vicinity of 70-80. Lightly roasted coffee also retained more internal moisture than the darker roasted analogues, Figure 10d. In line with our findings in Figure 9c, we observe an abrupt transition to negative charging at water contents <2%. Notably, the internal moisture content appears to be an very good predictor for charging (RMSE = 7.93 for charge versus moisture content, compared to RMSE = 19.37 for charge versus Agtron). Also, dehydration follows an exponential relationship between moisture and charge, in line with the dehydration profile of bananas,97 seeds,98 and other foodstuffs. Together, these data suggest that internal water content is a primary factor in the electrification behavior of roasted coffee. 2.1.8 Grind setting-dependent charging. Finer grind settings necessitate more fractures for the coffee to exit the grind chamber. Additionally, kinetic theory predicts that finer particle flows have higher granular temperatures (provided that particles have significant inertia to overcome fluid modification of granular temperature99), with individual grains undergoing large numbers of collisions.100 Thus, grinding finer should generate more charging through both fracto- and triboelectrification, regardless of polarity. To probe this hypothesis, we performed the same experiments presented in Figure 9, but at varied the grind setting. Two examples are displayed in Figure 11a, revealing that coarser settings yield lower charge, independent of whether the coffee is positive charging (Amatepec, W) or negative charging (Kicking Horse, D). 58 5 10 15 20 EK 43 grind setting -90 -80 -70 -60 -50 -40 -30 C ha rg e- to -m as ss ra tio [n C g –1 ] 10 15 20 0 20 40 60 Boulder-fine aggregateb a Q/m ratio as a function of grind setting Amatepec (W) EK 43 grind setting Kicking Horse (D) 1000 µm 40 60 80 Agtron color 300 325 350 375 400 425 M ea n pa rti cl e di am et er [µ m ] net charge _ + Particle size as a function of roast color c 5 Espresso size region 1 1 Figure 11. Charging as a function of grind setting and roast a) Charge- to-mass ratios of two representative coffees ground at settings spanning espresso (fine) to French press (coarse). In general, grinding finer increases charging, and at settings around 1.0 (the espresso region), positively charging coffees a large spread due to the formation of b) aggregates as the coffee exists the grind chamber. For negatively charging coffees, grinding at 1.0 yields a slightly more positive average charge, with relatively small spread. We attribute this to the higher charging resulting in the very rapid formation of aggregates. c) At a constant grind setting, dark coffees have approximately 100 µm finer mean particle size than light coffees. 59 The general trend is that the Q/m magnitude increases as coffee is ground finer. Yet some unexpected behavior is observed at the finest grind settings. For instance, Amatepec shows a large spread in the charge-to-mass ratio values and microscopic examination reveals that the variance is attributed to the formation of aggregates, i.e., particles sticking to each other and other grinder surfaces, Figure 11b. The effect is exemplified in the Kicking Horse samples, where the finer particles have even higher surface charge and result in a minor reduction in observed electrification at the finest grind setting. We attribute this reduction to expeditious formation of the aggregates, analogous to those shown in Figure 11b. Additionally, our measurements reveal that dark roast coffees do produce much finer particles when ground at the same setting, Figure 11c. The darkest coffee in our dataset exhibits a -100µm shift in particle size relative to the lightest coffee. These data build upon a previous study that showed that four light roast coffees produced similar particle sizes83 because they were similar in roast color. Figure 11c offers one explanation as to why darker coffees may yield slower espresso shots for the same brew parameters. It may not only be the increased volatile content94, but also the reduced bed permeability. Since finer grinding generates more charge, a direct relationship between roast and charge must include an empirical correction for the difference in particle sizes shown in Figure 11c. To do this, we can manually change the grind setting and monitor the particle size data until the dark and light roast coffees produce the same size distribution. In practice, this meant grinding dark roasts at a slightly coarser setting, 2.3, to achieve the same particle size distribution as our lightest coffee ground at 2.0. Figure S2 reveals that variations in particle size alone cannot account for the trends we observed in Figure 9. That is, dark roast 60 coffees accumulate more negative surface charge during grinding than their lighter relatives, independent of the size differences. 2.1.9 How granular mechanics influence electrification. With the impact of roast and grind isolated (where dark roasts and fine grinding yields the most charge), we next sought to investigate the impact of granular mechanics on electrification during coffee grinding. Much effort has been devoted to ascertain the roles of fragmentation and triboelectric charging in granular flows,55,101,102 and some authors invoke a common physio-chemical origin for charging,103,104. Although this matter remains unsettled, it seems as though flows that do include material fracture behave differently from those that do not. For example, Lim and colleagues showed that pre-ground coffee particulates accumulated approximately -2 to -5 nC g−1 by simply rubbing upon a stainless steel mixing auger and against themselves.105 In contrast, coffees ground in our experiments gained absolute Q/m ratios exceeding 100 nC g−1. This lays the foundation for the hypothesis that fragmentation processes are, to a large degree, responsible for electrostatic charging in coffee. To isolate the impact of fracturing, we allowed ground coffee to pass through the grinder a second time at a coarser setting, preventing additional comminution. Without additional fracturing events, most charge should arise from coffee-coffee and coffee-burr interactions. The inset in Figure 12 shows the particle size distribution before and after the coffee traversed the grinder a second time, first at setting 2.0 and second at a much larger burr aperture (setting 6.0). The latter setting was selected because it was empirically shown to not alter the particle size distribution upon re-grinding (see Figure S4 for further details). The minimal differences between the particle distributions indicates that pre-ground coffee 61 Pike 's Plac e Swee t L ife Unic yc le Red B ric k Đà L ạt Clou dh op pe r Plan ad as –100 –80 –40 –20 0 20 40 C ha rg e- to -m as s ra tio [n C g –1 ] –60 Charge-to-mass ratio of once- and twice-ground coffeea First grind (2.0) Regrind (6.0) 100 1000 Size [μm] 0.00 0.05 0.10 Fr eq ue nc y [v ol um e % ] 2.0 6.0 Figure 12. Twice-grinding coffee to assess the role of fracture in electrification. The first grinding of whole bean coffee at setting 2.0 results in charging from both triboelectric and fractoelectric processes, with a corresponding particle size distribution presented inset in grey. Regrinding the same particles at setting 6.0 results in a reduction in surface charge, with essentially no change in volumetric particle size distribution. First grinding is presented in grey, and second grinding in black. 62 particles are sufficiently small to exit the grind chamber without further fracture, but still accumulating some charge. To assess the generality of this observation, we performed these experiments on seven coffees, Figure 12. Twice-ground coffees (black) acquired significantly less charge than their primary ground counterparts (grey). Indeed, we observe a reduction of charge by up to 90%, with most pre- ground coffees acquiring Q/m ratios in the range of 5-10 nC g−1. Such values are comparable to the findings made by Lim and others105 during the fluidization of powdered coffee. While we cannot isolate coffee-coffee and coffee-burr rubbing interactions, this experiment points to a general observation that reduced fracturing greatly reduces charging. 2.1.10 Grinding with supplemental external water and its impact on brewing. Recalling the progression towards zero/positive charging with increasing internal moisture, Figure 10d, we next sought to understand the impact of adding external moisture. This process - known in the coffee industry as the ”Ross Droplet Technique” - was posited to have been originally presented in an online message board.106–108 Anecdotally, baristas have observed that the incorporation of small amounts of liquid water onto the whole-bean coffee prior to grinding results in seemingly reduced charging. In our hands it also resulted in near-zero grounds being retained by the grinder, an observation that has implications for reducing waste and increasing quality of beverages. Perhaps we will revisit this in a future study, but for now we are more interested in whether the addition of water neutralizes the effects of fracto- and triboelectrification or modulates particle aggregation via capillary forces. Indeed, while abundant water does seem to preclude charge buildup109, recent work suggests that small amounts of free water during granular electrification can produce unexpected behaviors. For 63 0 5 10 15 20 Water content [µL g–1] 60 40 20 0 20 40 Unicycle.com (B) Red Brick (B) Amatepec (W) Huila (N) Kicking Horse (D) Bekele (N) Temascaltepec (W) Mané (B) a Charge suppression by water inclusion pre-grinding b Charge altering using saline solution Huila (N) Negative Neutral Positive 0 20 40 60 80 100 C ha rg e- to -m as s ra tio [n C g –1 ] Bin polarity N or m al iz ed w ei gh t [ % ] c Charge binning of Huila coffee dry, field 20 µL g–1, field dry, no field 0 5 10 15 20 Water content [µL g–1] 80 60 40 20 0 20 40 C ha rg e- to -m as s ra tio [n C g –1 ] 1 M NaCl 0.5 M NaCl 0 M NaCl 0 µL 5 µL 10 µL 15 µL 20 µL 0 2 4 6 8 10 12 0 500 1000 1500 2000 clumping region Particle Diameter [μm] V ol um e % dParticle size distributions with increasing water content 0.0 0.5 1.0 1.5 2.0 Added water [dry coffee mass %] Figure 13. External moisture controls surface charging and causes particle deaggregation a) Charge-to-mass ratios for several coffees that span positive, neutral, and negative charging, with increasing amounts of water introduced to the whole beans prior to grinding. The red-to-blue coloring is indicative of the roast color, where blue are darker roasts. The upper x-axis provides an estimate for the effective change in total moisture content contained within, and coated on the surface, of the whole beans. b) The inclusion of minerals in the water solution has no effect on the magnitude of charge suppression achieved by the water itself. c) The inclusion of water during grinding causes deaggregation of fines from boulders. d) The redistribution of particle polarity upon the addition of 20 µL g−1 water added to the whole beans. 64 instance, Grosjean and Waitukaitis110 showed that water can change the charging behavior randomly and irreversibly. Hu and colleagues found that the open-circuit voltage of a triboelectric nanogenerator increases with relative humidity up until values of 50%.111 We also performed humidity experiments, Figure S4, and found that humidity only affected charging above approximately 60 %RH, in line with Ref.111. To assess the the impact of water addition, we systematically introduced water to whole bean coffee and ground at setting 2.0. Charge-to-mass ratios as a function of added water are presented in Figure 13a. All commercially sourced coffees of varying darkness, moisture content, and origin/processing methods (see Supplementary Table S1) show a systematic reduction in charging with increasing external water content. As water content approaches 20µL g−1, the charging approaches towards 0 nC g−1. We performed further experiments to isolate the impact of ambient humidity, and found while humidity can begin to reduce charging above 60 %RH (see Figure S4), all of our measurements were performed between 25-45 %RH, and were largely unaffected by atmospheric water. This result is somewhat surprising, given the majority of charging comes from fractoelectrifying events, and the water is only introduced to the surface of the whole beans near instantaneously prior to grinding (preventing uptake and homogeneous wetting of the insides of the coffee). Here, water may be acting to reduce interfacial temperatures during fracturing (the coffee does exit the chamber cooler in the presence of water, see Figure S5), or perhaps it is facilitating some other physical process, e.g. enabling rapid solvated ion transfer. Since both tribo- and fractocharging may originate from electronic ionization, nuclear ion transfer, or a combination of the two,112 we further 65 developed an experiment to suppress ion transfer through the inclusion of ions directly into the wetting solution. If ion transfer is an operative mechanism, the inclusion of salt water should produce markedly reduced charging, dissimilar to that of pure water. Figure 13b reveals that the inclusion of NaCl, at either 0.5 M or 1.0 M, shows the same fractoelectric charge reduction as that of pure water. These data lead us to conclude that ionic electrification is likely not the mechanism of charging in coffee, but rather electron transfer. To deduce whether the coffee particles were forming neutral aggregates we examined both the laser diffraction particle size distributions as well as the electrostatic binning, Figure 13c and d. From the particle size distribution it is clear that the inclusion of even small quantities of water (as low as 5 µL g−1) results in an immediate reduction in electrostatic aggregates of boulders and fines (the clumping region in Figure 13c).From Figure 13d, the data reveals that a positive charging coffee transitions from to ne