Effects of Temperature and Relative Humidity on Filter Loading by Simulated Atmospheric Aerosols & COVID-19 Related Mask and Respirator Filtration Study

Home , Air filter, Clean Air Delivery Rate

A DISSERTATION SUBMITTED TO THE FACULTY OF THE UNIVERSITY OF MINNESOTA BY

Chenxing Pei

IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY

Dr. David Y.H. Pui, Advisor

August 2020

Acknowledgments

First and foremost, I am deeply indebted to my advisor, Professor David Y. H. Pui, for his continuous guidance, mentoring, and inspiration throughout my undergraduate and graduate studies. He provided me with the encouragement and freedom to work on any filtration related studies that interested me. His support, patience, and trust have made my years spent in the Particle Technology Laboratory an enjoyable, enriching, and memorable experience in my life.

I would like to express my deepest appreciation to my committee members: Prof.

Thomas Kuehn, Prof. David Kittelson, and Prof. Kevin Janni for reviewing my dissertation, offering insightful comments and suggestions, and serving on my Ph. D. exam committees.

I am extremely grateful to my committee members not only for their time and extreme patience, but for their intellectual contributions to my development as a scientist.

I would also like to thank Dr. Qisheng Ou, whose help cannot be overestimated. Dr.

Ou provided me with the tools that I needed to choose the right direction and successfully complete my dissertation. Special thanks to Dr. Young H. Chung, who led to many interesting and motivating discussions. His help and advice were invaluable for my research and life. Thanks also to Weiqi Chen, who collaborated with me on the smart sensor project. I am also grateful to Kai Xiao for enlightening me the first glance of research.

I would also like to extend my deepest gratitude to my current and former colleagues in the Particle Technology Laboratory: Dr. Young H. Chung, Dr. Lin Li, Dr.

Sheng-Chieh Chen, Dr. Qisheng Ou, Dr. Shigeru Kimoto, Dr. Min Tang, Dr. Yan Ye, Dr.

i Seong Chan Kim, Dr. Qingfeng Cao, Dr. Handol Lee, Dr. Ziyi Li, Dr. Tsz Yan Ling, Dr.

Zhili Zuo, Dr. Chang Hyuk Kim, Dr. Drew Thompson, Dr. Nanying Cao, Dr. Seungkoo

Kang, Dr. Lipeng Su, Dr. Xinjiao Tian, Dr. Qiang Lyu, Kai Xiao, Gustaf Lindquist, Swathi

Satish, Weiqi Chen, Dongbin Kwak, Tao Song, Jihe Chen, Ren Jeik Ong, Qi Heng Lee,

Jonathon Anderson, Kaushik Kanna, Dr. Mamoru Yamada, Dr. Dongbin Kim, Dr. Yun

Liang, Dr. Deqiang Chang, Dr. Cheng Chang, and Dr. Ningning Zhang. Their company, discussion, and assistance have been invaluable.

I would like to express my gratefulness to ME Connect Seminar Series committee members and our mentor Prof. Suhasa Kodandaramaiah. It was rewarding and interesting to work with all of them to organize seminars every week and learn others’ research in the

Department of Mechanical Engineering.

I would like to acknowledge members of the Center for Filtration Research for their financial support, comments and interests in my studies: 3M Company, Applied Materials,

Inc., BASF Corporation, Boeing Company, Corning Incorporated, Cummins Filtration Inc.,

Donaldson Company, Inc., Entegris, Inc., Ford Motor Company, LG Electronics Inc.,

Parker Hanniﬁn, Samsung Electronics Co., Ltd., Shengda Filtration Technology Co., Ltd.,

Shigematsu Works Co., Ltd., TSI Inc., W. L. Gore & Associates, Inc., WatYuan Filtration

System Co., Ltd, Yancheng Environmental Protection Science and Technology City, and the aﬃliate member National Institute for Occupational Safety and Health (NIOSH). Parts of this work were carried out in the Characterization Facility, University of Minnesota, which receives partial support from NSF through the MRSEC program.

Last but not least, my deep and sincere gratitude to my wife, Kunlin Wang.

Understanding me best as a Ph.D. herself, Kunlin has been my best friend and terrific

ii companion, who gives me continuous, unparalleled, unconditional love and support. I am also extremely grateful to my parents and parents-in-law, who give me opportunity to explore my new possibilities and support me emotionally and financially. It is their unreserved love, unbounded trust, endless patience, and timely encouragement make me who I am today. Thanks also to my other family members and friends along my way, who bring me love and support. In addition, My thanks go to my 5-year-old twin cats, Laobing and Bantang, who have grown up as the best ‘coworkers’ during my Ph.D. study.

iii

To my parents, my love, and my future children

iv Abstract

This dissertation focuses on the effects of temperature and relative humidity on filter loading by simulated atmospheric aerosols and COVID-19 related mask and respirator filtration study. There are two objectives of this dissertation: the first one is to fill the gap between lab filter testing and actual filter operation, and the second one is to help curb the COVID-19 spreading from filtration perspective.

Filter life is an important criterion to evaluate air filters, and longer life is preferred before reaching the replacing pressure. Filter life could be evaluated in the laboratory by loading lab-generated contaminants on the test filter sample, typically at high concentrations to accelerate the testing process. However, the current filter loading testing protocols and standards use the non-hygroscopic micron dust or less-hygroscopic lab salt particles (NaCl, KCl) as challenging particles. Meanwhile, the testing environment, especially relative humidity, have not been strictly controlled. With increasingly more stringent pollution control requirement in place in past decades, sub-micron particles are becoming more important, such as inorganic salts ((NH4)2SO4, NH4NO3), soot, organic compounds, which are more hygroscopic than lab salts. Therefore, there have been discrepancies between lab filter testing and actual filter operation, including challenging particle species, testing relative humidity and temperature. It is vital to understand how air filters perform in the actual environment so that both filter recommendations and optimization could be done confidently and wisely.

This dissertation thrives on the effects of temperature and relative humidity on filter loading by simulated atmospheric aerosols. A Lab-simulated Ambient Environment Air

v Filter Test Rig was firstly built with controlled temperature and relative humidity testing environment, simulated atmospheric aerosols generation, and advanced filter testing capability. The testing temperature and relative humidity are well controlled to simulate different filter operating conditions. The techniques to generate atmospheric aerosols were developed, including inorganic salt generation and dry/wet state control at various testing relative humidities, fresh soot and aged soot by organic compounds coating. The versatile atmospheric aerosols generation techniques enable filter loading tests with “real” atmospheric aerosols rather than lab salts.

The temperature effect on filter loading was studied, and it is found that temperature is a minor factor affecting filter life, compared to relative humidity. The relative humidity effect on filter loading was then investigated with conventional cellulose filter and nanofiber coated cellulose filter. The testing relative humidity covered from below salts’ efflorescence relative humidity to above their deliquescence relative humidity, and both dry and wet particles at different relative humidities were loaded on test filters. It is found that, in general, the higher the loading relative humidity, the longer the filter life. But the filter life also depends on the filter structure and challenging particle state. Filtration efficiency evolution was also studied.

To better simulate atmospheric aerosols, test filters were loaded with (NH4)2SO4 and NH4NO3 mixture particles and soot aged by organic compounds coating separately at various relative humidities. It is found that relative humidity is a very strong factor affecting filter life, particularly for salt mixtures. There is a discrepancy between fresh soot and aged soot, and aged soot should be used in filter loading since it can represent atmospheric soot more realistically. These studies reveal that actual filter operation with

vi atmospheric aerosols is complex, and the filter recommendation should be carefully made, and the actual operating conditions should be considered as much as possible.

The above research was conducted at constant loading relative humidity, which is not possible in actual filter operation. Hence the effect of relative humidity change on hygroscopic salt loaded filters was studied. This study broadens the understanding of the loaded filter behavior in varying relative humidity which could benefit potential techniques to prolong filter life or to clean filters by reverse pulsing.

The ultimate goal of the first objective is to recommend filters based on operating locations. Whereas it is still challenging to achieve filter selection and optimization based on all research mentioned above. We developed and established a Global Air

Pollutant/Meteorological Condition Database and a Smart Sensor for Filter Performance

Monitoring System. For a city of interest, this Database can report historical pollutants speciation and concentration, historical monthly average temperature and RH, and pollutant concentration trend in past years. The Smart Sensor could monitor filtration efficiency, the differential pressure across the filter, and operating temperature, RH in real- time. With millions of the Smart sensor installations, massive filter operation data could be received for machine learning and big data analysis. With the aid of the Database and Smart

Sensor, the above experiment conclusions could be utilized to better test and recommend filters based on operating locations.

COVID-19 becomes a global challenge in 2020. Two mask and respirator studies were carried out to help curb the spread of COVID-19 and relieve the severe shortage of masks and respirators. The first one compared common materials that have the potential to be used as alternative masks for the public in daily protection. The second one aimed to

vii evaluate the effect of decontamination of the commercial and alternative mask and respirator materials from the filtration perspective.

These seven studies below constitute the main body of this dissertation and have been published, currently under review, or in preparation.

Chapter 5: Pei, C., Ou, Q. and Pui, D. Y. H. (2020). Effects of Temperature and Relative Humidity on Laboratory Air Filter Loading Test by Hygroscopic Salts Separation and Purification Technology 255: 117679. doi:10.1016/j.seppur.2020.117679

Chapter 6: Pei, C., Ou, Q., and Pui, D. Y. H. (2019). Effect of relative humidity on loading characteristics of cellulose filter media by submicrometer potassium chloride, ammonium sulfate, and ammonium nitrate particles Separation and Purification Technology 212: 75-83. doi:10.1016/j.seppur.2018.11.009 Pei, C., Ou, Q., Yu, T. and Pui, D. Y. H. (2019). Loading characteristics of nanofiber coated air intake filter media by potassium chloride, ammonium sulfate, and ammonium nitrate fine particles and the comparison with conventional cellulose filter media Separation and Purification Technology 228: 115734. doi:10.1016/j.seppur.2019.115734

Chapter 7: Pei, C., Ou, Q., Wang, K., Sun, C.C., and Pui, D. Y. H., Effect of different ammonium salt mixing ratios on the loading performances of two nanofiber coated cellulose filters In preparation

Chapter 8: Pei, C., Ou, Q., and Pui, D. Y. H., Effect of relative humidity change on pressure drop of loaded cellulose filter media with hygroscopic deposits In preparation

Chapter 9: Pei, C.1, Ou, Q.1, Kim, S. C., Chen, S-C., and Pui, D. Y. H. (2020). Alternative Face Masks Made of Common Materials for General Public: Fractional Filtration Efficiency and Breathability Perspective Aerosol and Air Quality Research 20 doi:10.4209/aaqr.2020.07.0423

Chapter 10: Ou, Q.1, Pei, C.1, Kim, S. C., Abell, E. and Pui, D. Y. H. (2020). Evaluation of decontamination methods for commercial and alternative respirator and mask materials – view from filtration aspect Journal of Aerosol Science 150: 105609. doi:10.1016/j.jaerosci.2020.105609

viii Table of Contents Acknowledgments ...... i Abstract ...... v Table of Contents ...... ix List of Tables ...... xiii List of Figures ...... xiv

Part 1: Methodology Development ...... 1 Chapter 1: Lab-simulated Ambient Environment Air Filter Test Rig ...... 3 1.1 Introduction ...... 3 1.2 Temperature Control ...... 8 1.3 Air Flow Control and RH Control ...... 10 1.4 Salt Dry/Wet State Control ...... 13 1.5 Electric Control System and LabVIEW Program ...... 17 1.6 Filter Sample Preparation ...... 19 Chapter 2: Soot Generation and Organic Compounds Soot Coating ...... 22 2.1 Soot Generation ...... 22 2.2 Organic Compounds Coating Method ...... 25 Chapter 3: Smart Sensor for Filter Performance Monitoring System ...... 27 3.1 Background ...... 27 3.2 System Description ...... 30 3.3 Case study 1: Indoor Air Purifier ...... 33 3.4 Case study 2: Delhi SALSCS Operating Monitoring ...... 38 3.5 Conclusion ...... 41 Chapter 4: Global Air Pollutant/Meteorological Condition Database ...... 43

Part 2: Temperature and RH Effect on Filter Performance ...... 48 Chapter 5: Effects of Temperature on Filter Loading by Hygroscopic Salts at Different RH ...... 50 5.1 Introduction ...... 50 5.2 Experimental Method and Material ...... 52 5.3 Results ...... 55 5.3.1 Dry KCl ...... 55

5.3.2 Dry (NH4)2SO4 ...... 57

ix 5.3.3 Wet NH4NO3 ...... 60 5.4 Discussion ...... 63 5.4.1 DRH Temperature Dependence ...... 63 5.4.2 Capillary Condensation ...... 65 5.4.3 Suggestions to Atmospheric Particulate Pollutants Control Technology . 67 5.5 Conclusion ...... 68 Chapter 6: RH Effect on Filter Loading by Single Species ...... 70 6.1 Conventional Cellulose Filter Media ...... 70 6.1.1 Introduction ...... 70 6.1.2 Experimental Method and Material ...... 72 6.1.3 Results and Discussion ...... 75 6.1.3.1 RH Effect of Deposited Hygroscopic Particles ...... 75 6.1.3.2 Loading Curves ...... 77 6.1.3.3 Efficiency Curves ...... 81 6.1.3.4 Volume Loadings ...... 83 6.1.3.5 Microscopic Observation ...... 85 6.1.4 Conclusions ...... 88 6.2 Nano-Fiber Coated Cellulose Filter Media ...... 90 6.2.1 Introduction ...... 90 6.2.2 Experimental Method and Material ...... 94 6.2.3 Results and Discussion ...... 98 6.2.3.1 Particle State Effect on Filter Deposition ...... 98 6.2.3.2 RH Effect on Filter Deposition ...... 100 6.2.3.3 Loading Curves ...... 101 6.2.3.4 Volume Loading ...... 105 6.2.3.4.1 Dry Particles Volume Loading ...... 105 6.2.3.4.2 Wet Particles Volume Loading ...... 106 6.2.3.5 Comparison with Conventional Cellulose Filter Media ...... 109 6.2.3.6 Microscopic Observation ...... 111 6.2.4 Conclusions ...... 114 Chapter 7: RH Effect on Filter Loading by Multiple Species ...... 118

7.1 (NH4)2SO4 and NH4NO3 Mixture ...... 118 7.1.1 Introduction ...... 118 7.1.2 Experimental Method and Material ...... 120 x 7.1.3 Results and Discussion ...... 123 7.1.3.1 Dry Mixture Particles Loading ...... 123 7.1.3.2 Wet Mixture Particles Loading ...... 129 7.1.4 Conclusions ...... 131 7.2 Soot with Organic Compounds Coating ...... 132 7.2.1 Introduction ...... 132 7.2.2 Experimental Method and Material ...... 134 7.2.3 Results and Discussion ...... 135 7.2.4 Conclusions ...... 139 Chapter 8: Effect of RH Change on Pressure Drop of Hygroscopic Particles Loaded Cellulose Filter Media ...... 141 8.1 Introduction ...... 141 8.2 Experimental Method and Material ...... 143 8.3 Results and Discussion ...... 147 8.3.1 KCl Loaded Filter Sample ...... 147

8.3.2 (NH4)2SO4 Loaded Filter Sample ...... 150

8.3.3 NH4NO3 Loaded Filter Sample ...... 153 8.4 Conclusions ...... 155

Part 3: COVID-19 Related Study: Mask and Respirator Filtration Study ...... 157 Chapter 9: Alternative Face Masks Made of Common Materials for General Public: Fractional Filtration Efficiency and Breathability Perspective ...... 159 9.1 Introduction ...... 159 9.2 Experimental Method and Material ...... 162 9.3 Results And Discussion ...... 167 9.3.1 Commercial Respirators and Masks ...... 167 9.3.2 Furnace Filters ...... 170 9.3.3 Vacuum Cleaner Filters ...... 171 9.3.4 Common Household Materials ...... 172 9.3.5 Discussion of The Figure of Merit ...... 175 9.3.6 Some Discussions ...... 177 9.4 Conclusions ...... 180 Chapter 10: Evaluation of Decontamination Methods for Commercial and Alternative Respirator and Mask Materials – View from Filtration Aspect ...... 181 10.1 Introduction ...... 181 xi 10.2 Experimental Method and Material ...... 185 10.2.1 Decontamination Methods ...... 185 10.2.2 Filtration Performance Testing ...... 185 10.3 Results and Discussion ...... 187 10.3.1 Base Performance Before Decontamination ...... 188 10.3.2 Decontamination by Isopropanol Alcohol ...... 190 10.3.3 Decontamination by UVGI ...... 193 10.3.4 Decontamination by Thermal Treatments ...... 195 10.3.5 Autoclave Treatment on Masks Made of Sterilization Wrap ...... 202 10.4 Conclusions ...... 204

Chapter 11: Summary and Future Work ...... 207 11.1 Effects of Temperature and RH on Filter Loading by Simulated Atmospheric Aerosols ...... 207 11.2 COVID-19 Related Mask and Respirator Filtration Study ...... 212 11.3 Future Work ...... 213

Bibliography ...... 216 Appendix ...... 236

xii List of Tables Table 1–1. Testing conditions of several filter testing standards ...... 4 Table 1–2. Deliquescence and efflorescence relative humidity of potassium chloride, ammonium sulfate, and ammonium nitrate from literature ...... 15 Table 3–1. Comparison of current commercially available products ...... 29 Table 3–2. Bill of material of the current prototype ...... 33 Table 3–3. Sample readings of the filter performance monitoring system at medium fan speed ...... 35 Table 5–1. Analysis of Variance Table of KCl ...... 57 Table 5–2. Analysis of Variance Table (Type II test) of dry (NH4)2SO4 ...... 60 Table 5–3. Analysis of Variance Table (Type II test) of wet NH4NO3 ...... 62 Table 6–1. Test matrix of this study ...... 97 Table 7–1. Chemical and Physical Properties of glutaric acid and oleic acid ...... 134 Table 7–2. Mass fractal dimensions of pure soot and glutaric acid coated soot at different RHs ...... 137 Table 8–1. All possible process from loading RH to Conditioning RH ...... 146 Table 10–1. Quantitative fit testing results of the new N95 respirator and after oven and steam heat treatment cycles...... 199

xiii List of Figures Figure 1–1. Schematic diagram of the Lab-simulated Ambient Environment Air Filter Test Rig ...... 7 Figure 1–2. Photo of the Lab-simulated Ambient Environment Air Filter Test Rig ...... 7 Figure 1–3. Photos of the temperature-controlled chamber and details inside ...... 9 Figure 1–4. Detailed schematic diagram of the Lab-simulated Ambient Environment Air Filter Test Rig (only one filter holder is depicted here for simplicity) ...... 10 Figure 1–5. Demonstration of the hydration and dehydration process of a salt particle .. 14 Figure 1–6. Schematic diagram of the LAEAF Test Rig control system ...... 18 Figure 1–7. Filter sample preparing process ...... 21 Figure 2–1. Cross section drawing of the homemade soot generator ...... 23 Figure 2–2. Photo of the homemade soot generator ...... 24 Figure 2–3. Sample of soot particles generated by the homemade soot generator ...... 24 Figure 2–4. Soot organic compound coating apparatus ...... 26 Figure 3–1. Schematic of the filter performance monitoring system ...... 30 Figure 3–2. Photo of the filter performance monitoring prototype ...... 32 Figure 3–3. The filter performance monitoring system on an indoor air purifier (HEPA filter is not in place to show the downstream module) ...... 34 Figure 3–4. Differential pressures of the HEPA filter at different fan speeds ...... 36 Figure 3–5. Filtration efficiencies of the HEPA filter at different fan speeds ...... 37 Figure 3–6. Schematic diagram of the Delhi SALSCS and filter performance monitoring system (modified from Dr. Sheng-Chieh Chen’s diagram) ...... 39 Figure 3–7. Locations of filter performance monitoring system sensors and other measurement equipment in the Delhi SALSCS ...... 40 Figure 3–8. Sensing and control system diagram of the Delhi SALSCS ...... 41 Figure 4–1. Monthly average PM2.5 concentration of Beijing, Shanghai, and Guangzhou from 2014...... 44 Figure 4–2. Yearly average PM2.5 concentration of Beijing, Shanghai, and Guangzhou from 2014 ...... 45 Figure 4–3. Sample data extracted from the global air pollutant meteorological condition database ...... 47 Figure 5–1. (NH4)2SO4, NH4NO3 testing points and corresponding AHs. Testing temperatures are carefully picked so that the AH of low temperature high RH is close to the high temperature low RH (two dotted lines indicate two similar AHs)...... 54 Figure 5–2. SEM image of a clean filter sample. Thin fibers are synthetic nanofibers (Nylon), and thick fibers beneath the nanofibers are cellulose fibers ...... 55 Figure 5–3. Volume loading of dry KCl at different loading AHs ...... 56 Figure 5–4. Volume loading of dry (NH4)2SO4 at different loading temperatures and RHs ...... 58 Figure 5–5. Volume loading of dry (NH4)2SO4 at different loading AHs ...... 59 Figure 5–6. Volume loading of wet NH4NO3 at different loading temperatures and RH 61 Figure 5–7. Volume loading of wet NH4NO3 at different loading AHs ...... 62 Figure 5–8. The plot of Mean Curvature of the Meniscus as a function of RH and temperature ...... 67

xiv Figure 6–1. SEM image of a test filter media sample. It is a conventional cellulose filter media with an efficiency rating of F8 according to EN779 ...... 73 Figure 6–2. Salt particle size distribution ...... 74 Figure 6–3. Dendrites restructuring with capillary condensed water ...... 76 Figure 6–4. Selected loading curves under different RHs...... 80 Figure 6–5. Efficiency curve as a function of volume loading...... 82 Figure 6–6. Volume loadings of KCl, (NH4)2SO4, and NH4NO3 at different RH ...... 84 Figure 6–7. SEM images of the KCl and (NH4)2SO4 under 30% and 75% RHs, NH4NO3 under 15% and 30% RHs...... 86 Figure 6–8. Higher magnification (10000X) of the SEM images of the KCl and (NH4)2SO4 under 30% and 75% RHs, lower magnification (5000X) of the NH4NO3 under 15% and 30% RHs...... 88 Figure 6–9. Number-based and mass-based size distributions of three salts. This is the averaged size distribution of all samples in this study for each salt...... 96 Figure 6–10. KCl, (NH4)2SO4, and NH4NO3 sample loading curves under different RH and different particle states ...... 104 Figure 6–11. Volume loadings of KCl, (NH4)2SO4, and NH4NO3 at different RH on nanofiber coated cellulose filter (the present study) ...... 107 Figure 6–12. SEM image of (NH4)2SO4 at 90% RH ...... 109 Figure 6–13. Volume loadings of KCl, (NH4)2SO4, and NH4NO3 at different RH on conventional cellulose filter (Adopted from Pei et al. (2019a)) ...... 110 Figure 6–14. SEM images of test filter media loaded with dry KCl and (NH4)2SO4 at different RH...... 112 Figure 6–15. SEM images of test filter media loaded with dry (NH4)2SO4 at 30% and 75% RH ...... 113 Figure 6–16. SEM images of test filter media loaded with wet KCl, (NH4)2SO4 and NH4NO3 at different RH...... 114 Figure 7–1. SEM image of a clean nanofiber coated cellulose filter with thicker and denser nanofibers...... 122 Figure 7–2. Mass gains of dry (NH4)2SO4 and NH4NO3 mixture particles on filter samples at different loading RHs ...... 124 Figure 7–3. Mass gains and sample loading curves close to the MDRH of salt mixture when m((NH4)2SO4):m(NH4NO3)=80%:20% ...... 125 Figure 7–4. SEM images and particle deposition illustrations at different loading RH . 126 Figure 7–5. Small cavities on m((NH4)2SO4):m(NH4NO3)=50%:50% mixture dry particles at 30% loading RH...... 127 Figure 7–6. Mass gains and loading curves close to the MDRH of salt mixture when m((NH4)2SO4):m(NH4NO3)=50%:50% ...... 128 Figure 7–7. Mass gains of wet (NH4)2SO4 and NH4NO3 mixture particles on filter samples at different loading RHs ...... 129 Figure 7–8. SEM images of filter samples loaded with pure NH4NO3 wet particles (left) and wet m((NH4)2SO4):m(NH4NO3)=20%:80% mixture particles (right) at 60% RH... 130 Figure 7–9. Effective densities of pure soot and glutaric acid coated soot particles with different mobility diameters ...... 136

xv Figure 7–10. SEM images of a filter sample loaded with pure soot in the bottom and glutaric coated soot on the top ...... 138 Figure 7–11. Mass Gains of pure soot and glutaric coated soot at different loading RH 139 Figure 8–1. Loading and conditioning processes of the test filter sample ...... 144 Figure 8–2. Illustration of possible states of loading and conditioning processes (KCl and (NH4)2SO4) ...... 146 Figure 8–3. KCl loaded filter samples pressure stabilization during the RH conditioning process...... 148 Figure 8–4. KCl terminal pressures of different loading and conditioning RHs ...... 149 Figure 8–5. SEM images of KCl loaded filter sample that was loaded at 30% RH and conditioned at 90% RH. Left: 1000X; Right: 5000X...... 150 Figure 8–6. (NH4)2SO4 terminal pressures of different loading and conditioning RHs . 151 Figure 8–7. SEM images of (NH4)2SO4 loaded filter sample that was loaded at 90% RH and conditioning at 30% RH ...... 152 Figure 8–8. SEM images of (NH4)2SO4 loaded filter sample that was loaded at 45% RH and conditioning at 100% RH. Left: dry particle loading; Right: wet particle loading .. 153 Figure 8–9. NH4NO3 terminal pressures of different loading and conditioning RH ...... 154 Figure 8–10. SEM images of NH4NO3 loaded filter sample that was loading at 60% and conditioning at 30% RH...... 154 Figure 9–1. Schematic of the fractional efficiency measurement setup ...... 167 Figure 9–2. Fractional filtration efficiency and breathing resistance of commercial respirators and masks ...... 169 Figure 9–3. Fractional filtration efficiency and breathing resistance of furnace filters .. 171 Figure 9–4. Fractional filtration efficiency and breathing resistance of vacuum cleaner filters ...... 172 Figure 9–5. Fractional filtration efficiency and breathing resistance of common fabrics ...... 175 Figure 9–6. The figure of merit of common materials ...... 177 Figure 10–1. Fractional efficiency (a), breathing resistance, and quality factor (b) of 3M 8210 respirator, Halyard 48207 surgical mask, 3M 1820 procedure mask, Halyard H600 sterilization wrap, and Cummins EX101 material, before any decontamination treatments ...... 190 Figure 10–2. Fractional efficiency and breathing resistance of the 5 filter materials before and after IPA treatments (soaking or spraying) ...... 193 Figure 10–3. Fractional efficiency (a-c) and breathing resistance (d) of N95 respirator, surgical mask, and procedure mask after multiple UVGI treatment cycles ...... 195 Figure 10–4. Fractional efficiency and breathing resistance of all five filter materials after dry (a) and steam (b) heat treatments...... 201 Figure 10–5. Filtration performance of unworn and worn H600 sterilization wrap masks after one or multiple cycle(s) of autoclave treatment. IPA soaked same material is included for reference purpose ...... 204 Figure 11–1. Structure of the first objective of this dissertation ...... 208

xvi

Part 1: Methodology Development

In Part 1, the methodologies used in this dissertation are introduced. Chapter 1 introduces the Lab-simulated Ambient Environment Filter Test Rig, which can test filters under specified temperature, relative humidity, particle compositions and states (dry/wet), etc. Chapter 2 presents the soot generation and organic compound soot coating method.

Chapter 3 introduces a novel filter performance monitoring system, which can monitor air filter efficiency and pressure drop in real-time. Chapter 4 covers the Global Air Pollutant

Meteorological Condition Database, which has the potential to recommend filters with the aid of the filter performance monitoring system.

2 Chapter 1: Lab-simulated Ambient Environment Air Filter Test Rig

1.1 Introduction

Air filters are designed to remove particulate air pollutants to provide cleaner air.

It is a particularly important protection for the equipment or people behind the filter. The best air filter should have low pressure drop, high efficiency, and long filter life. It is essential to understand how the pressure drop and efficiency of an air filter evolution during the air filter operation, and how long it takes before the air filter reaches a target pressure drop. The characterization of an air filter could be achieved with a filter test rig, which can control the flowrate through the filter media, measure the pressure drop across the filter media, and measure the filter media efficiency.

To simulate the aging process of an air filter, filter loading could be performed on a filter test rig. The loading process is to load the air filter with high concentration of challenging particles. Some filter test standards, such as ASHRAE 52.2, ISO 5011, ISO

16890, and EN 779, use the synthetic loading dust as the ﬁlter loading particles, which consists of ﬁne dust, powdered carbon, and cotton linter (ASHRAE 2017, European

Committee for Standardization 2012, ISO 2014, 2016). Some respirator loading testing protocols, such as 42 CFR Part 84, use sodium chloride (NaCl) or dioctyl phthalate (DOP) as the filter loading particles (42 C.F.R. § 84, Office of the Federal Register 1995).

However, there are several shortcomings of the current testing methods. The first one is that the synthetic loading dust and NaCl particle could not well represent the real particles 3 during the operation. The second one is that the relative humidity (RH) and temperature are also not carefully controlled and cannot represent the operating conditions. The testing particles, RH and temperatures of several filter testing standards are shown in Table 1–1.

Table 1–1. Testing conditions of several filter testing standards Relative Standard Testing Particles Temperature Humidity ASHRAE 52.2-2017 KCl / Synthetic Loading Dust 45 ± 10% 10 ⁓ 38℃ ISO 5011:2014 ISO A2 & A4 Dust 55 ± 15% 23 ± 5℃ ISO 16890:2016 Synthetic Loading Dust 45 ± 10% 23 ± 5℃ Room or EN 779:2012 DEHS / Synthetic Loading Dust < 75% outdoor air 42 CFR Part 84 NaCl(N-series)/DOP(R,P-series) 30 ± 10% 25±5℃

For the first shortcoming, in laboratory studies, sodium chloride (NaCl) and potassium chloride (KCl) are used as the testing particles for the ﬁlter loading research, because they are inexpensive, widely available, and non-toxic. The most common method of generating salt particles is by atomizing a salt solution and dry atomized droplets to yield dry salt particles. According to the ASHRAE standard 52.2-2017, to dry droplets, the RH of the particle-laden air must be controlled less than 55% for NaCl, and 70% for KCl, which is feasible to achieve in the laboratory by diﬀusion drying and mixing with dry air

(ASHRAE 2017).

On the other hand, the global population-weighted PM2.5 composition study by

Philip et al. (2014) shows that mineral dust is about 30% in PM2.5 composition. In another study, a Surface Particulate Matter Network (SPARTAN) measures particulate matter concentrations, speciation of 21 cities around the world, and the average crustal material percentage in PM2.5 is 13.4 % ± 9.9 % (Snider et al. 2015, Snider et al. 2016). With more stringent pollution control in past decades, sub-micron particles are concerned, such as

4 inorganic salts, soot, organic compounds. Many filed measurements have depicted that inorganic salts, ammonium sulfate ((NH4)2SO4) and ammonium nitrate (NH4NO3), accounts for at most 50% of the PM2.5 in some highly urbanized regions, and soot and organic compounds are also major pollutants (Cheng et al. 2016, He et al. 2001, Philip et al. 2014, Snider et al. 2015, Snider et al. 2016, Sun et al. 2004, Yang et al. 2011, Yao et al.

2002, Zhang et al. 2007).

For the second shortcoming, due to the use of the synthetic loading dust, the RH and temperature are not strictly controlled in those loading process. This is because moisture has the limited effect on the dust loading. Also, synthetic loading dust is not hygroscopic, and the particle size is in the micron range so that the capillaries between particles are diﬃcult for water vapor to condense. However, as mentioned above, the synthetic loading dust does not include salt particles, which is a critical component of ambient pollutants. The use of KCl and NaCl particles avoids the delicate RH control. The salt particles would be solid, provided that the testing RH is below the deliquescence relative humidity (DRH) of the testing salt. In addition, when the RH is considered, the temperature control is also critical as temperature variation would affect the RH stabilization.

Therefore, to simulate the atmospheric inorganic salt pollutants that air filters may encounter, (NH4)2SO4 and NH4NO3 are better challenging particles in filter tests. However,

(NH4)2SO4 and NH4NO3 are more hygroscopic than KCl and NaCl, which means that they are more active with water vapor at a given RH, and diﬀerent ambient air RHs may aﬀect the state of (NH4)2SO4 and NH4NO3. For soot and organic compounds, their hygroscopicities vary from hydrophobic to hydrophilic. Therefore, their hygroscopicities

5 will potentially affect soot deposition on air filters and further affecting the air filters performance.

The existing filter test apparatus are not capable to achieve a better control of the testing conditions mentioned above. To understand the effect of different particulate pollutants and operating conditions on filter performance, a Lab-simulated Ambient

Environment Air Filter Test Rig (LAEAF Test Rig) was built in the Particle Technology

Laboratory. The LAEAF Test Rig could test air filter samples under controlled RH and temperature, with various simulated pollutants ((NH4)2SO4, NH4NO3, soot, and organic compounds). It can perform both filter loading test and filter efficiency test.

The schematic diagram of the LAEAF Test Rig is shown in Figure 1–1, and the photo of the LAEAF Test Rig is shown in Figure 1–2. The LAEAF Test rig consists of pollutants generation, temperature-controlled environment, moisture generation, and filter monitoring. With different generation methods, salt particles, soot, and organic compounds can be generated and introduced to the LAEAF Test Rig, which will be described in detail later. The temperature-controlled environment encloses the test filter holders, electric heating elements, a circulating cooling coil, and other necessary parts, which can achieve filter testing at different temperatures. Details will be described in Section 1.2 Temperature

Control. To study the RH effect on the filter performance with different challenging particles, the testing RH can also be actively controlled. The details about the moisture generation and RH control together with the air flow control will be described in Section

1.3 Air Flow Control and RH Control. The electric control and testing automation will be covered in Section 1.5 Electric Control System and LabVIEW Program.

Figure 1–1. Schematic diagram of the Lab-simulated Ambient Environment Air Filter Test Rig

Figure 1–2. Photo of the Lab-simulated Ambient Environment Air Filter Test Rig

7 1.2 Temperature Control

To study the temperature effect on the filter performance, essential components of the LAEAF Test Rig are enclosed in a temperature-controlled chamber. The chamber is built with 80/20® T-slot construction system (Columbia City, IN), and covered with 1-inch thick polystyrene foam insulation boards, including the bottom. To increase the overall insulation performance, all exposed aluminum frames are covered with insulation boards as the aluminum is the thermal bridge in the system (see Figure 1–2). The front door is covered with two acrylic sheets, which are mounted on the outside and inside of the frame, separately. This double-panel design ensures the insulation and visibility of the front door.

An extra 1-inch thick polystyrene foam insulation board can be mounted when handling temperature around 0 ℃ (not shown in Figure 1–2).

Electric heating elements and a circulating cooling coil connected to a chiller are installed in the enclosure to control the testing temperature. Currently, the testing temperature ranges from 0 ℃ to 50 ℃. To efficiently cool the chamber near 0 ℃, ethylene glycol and water 1:1 liquid is used so that the cooling liquid would not freeze in the circulating tubes. When testing temperature is below room temperature, the temperature control is achieved by both cooling and heating since the electric heating elements could be controlled with a quicker response so that a more stable temperature could be achieved.

Consequently, chiller temperature is set to several degrees below the desired testing temperature, and the temperature inside the chamber is maintained to the desired testing temperature by PID controlled electric heating elements. When the testing temperature is above room temperature, only electric heating elements are employed to reach the desired

8 temperature. Two pairs of fans are installed below the cooling coil and electric heating elements, and the third pair of circulating fans are installed on the top portion of the chamber. The air inside the chamber circulates to reach homogenous desired temperature.

However, the particle-laden air is isolated from the temperature-controlled environment. Two sections of the stainless tubes are used to enhance the heat transfer and achieve the desired testing temperature in the particle-laden air system. Figure 1–3 shows the details inside the temperature-controlled chamber. There are several temperature sensors in the particle-laden air system and temperature-controlled chamber to monitor the process temperature and offer temperature controller feedback signal. When the temperature is stabilized and equilibrated, the temperature in the particle-laden air system reaches the desired testing temperature.

Figure 1–3. Photos of the temperature-controlled chamber and details inside

9 Only essential components are installed in the temperature-controlled chamber, so that unnecessary heat loads could be reduced, such as the hotplate for moisture generation and soot generator. Meanwhile, the components that can be affected by the temperature are also placed outside of the chamber, such as mass flow controllers (MFC) and differential pressure sensors (See Figure 1–4).

Figure 1–4. Detailed schematic diagram of the Lab-simulated Ambient Environment Air Filter Test Rig (only one filter holder is depicted here for simplicity)

1.3 Air Flow Control and RH Control

The air flow in the LAEAF Test Rig is labeled with color arrow in Figure 1–4. As shown on the left, salt particles are generated by a homemade Collision-type pneumatic atomizer, and soot particles are generated by a homemade diffusion flame soot generator, which will be described in detail in Section 2.1 Soot Generation. Salt and soot particles can be generated separately or simultaneously according to the needs. Dry filtered compressed air could be added to desiccate salt or soot particles based on needs. A chilled mirror 10 hygrometer (System 1100DP, GE General Eastern Instruments, Wilmington, MA) is installed to sample the particle-laden air in order to measure the RH. The reason why a RH sensor is installed at this location will be discussed in Section 1.4 Salt Dry/Wet State

Control. A capacitive RH and temperature sensor (AM2302, ASAIR, Guangzhou,

Guangdong, China) is also installed in series to measure the RH and temperature as a reference. Moisture generation is on the right side, and the moisture is generated by sparging the ﬁltered air into warm water bath at a controlled temperature. There are two proportional valves in the moisture generation part. One controls the flow rate of dry compressed air through the distilled water (wet air) and the other one controls the bypass flow rate (dry air). The total ﬂow rate through two proportional valves is about a constant, and the amount of the moisture introduced into the LAEAF Test Rig depends on opening percentages of two proportional valves. The opening percentage is PID controlled by a

LabVIEW program, which will be covered in Section 1.5 Electric Control System and

LabVIEW Program. The purpose of moisture generation is to increase the RH but only water vapor is preferred. To avoid introducing water droplets, the dry and wet combined air is connected vertically to a reservoir cup so that no water droplets could reach the main

ﬂow that would interfere the humidity measurement and filter testing.

When the particle-laden air meets the generated moisture, they would mix in the stainless tube. The mixed air is drawn to the test filter holder. The RH of test filter upstream air is sampled to the second chilled mirror hygrometer (DewMaster, Edgetech Instruments

Inc. Hudson, MA) and the second AM2302 RH/Temp sensor, where the testing RH is measured. Before the test particles reach the test ﬁlter, a Po-210 Staticmaster® Alpha

Ionizer (NRD, LLC. Grand Island, NY) neutralizer brought the charge level of challenging

11 particles to Boltzmann equilibrium charge distribution. The test filter is placed on a 47mm filter holder (Gelman Science 2220, Pall Corporation, Port Washington, NY), which has the pressure ports on both upstream and downstream. The third AM2302 sensor is installed at the downstream of the filter holder. The pressure drop across the test ﬁlter media was measured by a diﬀerential pressure sensor (Setra M267, Setra Systems, Inc. Boxborough,

MA). To measure the challenging particle size distribution and filter efficiency during the testing, a scanning mobility particle sizer (SMPS, TSI 3938L56(3082, 3081A, 3756)

Shoreview, MN) is connected to the upstream and downstream of the test ﬁlter by a 3-way ball valve. Since different RHs are used in testing, and the hygroscopic particle size is sensitive to RH, a silica gel dryer column is added in the differential mobility analyzer

(DMA) sheath flow to maintain a low RH in the DMA. As shown in Figure 1–3, there are two sample holders installed in parallel, which can test two test filters in parallel to achieve the efficient testing. Part of the mixed air is exhausted to satisfy the flow equilibrium, and the typical flow rate is about 30 liter per minute (LPM). The pressure in the LAEAF Test

Rig is slightly below the ambient pressure, the pressure is balanced by the pressure balance

HEPA filter near the moisture generation.

The sample flow rate of two chilled mirror hygrometers are controlled by critical orifices. When the downstream of the critical orifice is connected to the house vacuum, the pressure ratio of the upstream and downstream of the critical orifice is greater than the minimum pressure ratio for the chocked flow and the flowrate is limited to a constant value for a given orifice size. The flow rates of two test filters are controlled by two separate

MFCs, which could maintain a constant face velocity of test filters and accommodate the filter pressure increasing. Since the specific heat of the moist air varies slightly at different

12 RH (Boukhriss et al. 2013), a desiccant column is used before MFCs to remove water vapor in the stream. To accommodate the different air flow configurations, the main exhaust flow rate is also modulated by an MFC. Similarly, a desiccant column is connected to remove moisture. To protect critical orifices and MFCs from challenging particles, HEPA filters are installed before them.

There are six RH sensors in the LAEAF Test Rig: two chilled mirror hygrometers and four AM2302 sensors. In Figure 1–4 there are three RH sensor locations. Chilled mirror hygrometers and AM2302 sensors are installed in series after the particle generation and at the upstream of the test filter. The other two AM2302 sensors are installed at the downstream of each test filter. The chilled mirror RH sensors measure RH accurately and the one at the upstream of the test filter is used as the feedback signal of the PID RH control.

Other AM2302 sensors offer temperature measurements and reference RH measurements.

1.4 Salt Dry/Wet State Control

One of the important properties of hygroscopic salts (e.g., NaCl, KCl, (NH4)2SO4,

NH4NO3) is the hysteresis of the hydration and dehydration behavior. Salt particles are dry at low environment RH. If the environment RH increases then, salt particles does not absorb water until the environment RH reaches its deliquescence RH (DRH). At the DRH, salt particle abruptly absorbs water and undergoes a phase change from solid to its aqueous solution, i.e. a droplet. If the environment RH keeps increasing above the DRH, the salt droplet size grows by absorbing more water. If the environment RH decreases from a point above the DRH, salt droplet loses water with decreasing RH. The salt droplet does not lose

13 all the water it carries at the DRH, whereas loses all the water at its efflorescence RH (ERH) instead. Salt particle stays dry when the environment RH is below the ERH. The hydration and dehydration processes are demonstrated in Figure 1–5. The growth factor is the ratio of the particle size at different RH to its dry state size, indicating the salt particle size change. The hydration path and dehydration path are labeled in the figure with arrows. The salt particle dry and wet states are also labeled with wide black and red lines.

Figure 1–5. Demonstration of the hydration and dehydration process of a salt particle

The ERH of certain salt, when it exists, is always lower than its DRH. When the environment RH is between the ERH and DRH of a certain salt, the salt particle can be in either solid or droplet state, depending on its hydration-dehydration history. When the environment RH is above the DRH of a certain salt, it is in wet state; when the environment

RH is below the its ERH, it is in dry state (3 regions in Figure 1–5). Table 1–2 lists the

DRHs and ERHs of the three salts used. It is worth to note that the ERH of NH4NO3 is hard to determine and could be considered to not exist. Therefore, NH4NO3 particles remain droplet state while smaller in size than droplet generated from atomizer. 14 Table 1–2. Deliquescence and efflorescence relative humidity of potassium chloride, ammonium sulfate, and ammonium nitrate from literature Salt DRH ERH Sources 84.60.3% 58.70.7% (Ahn et al. 2010) Potassium 84.70.2% 57.70.5% (Li et al. 2014) chloride KCl 851% 561% (Freney et al. 2009) 83-86% 59% (Cohen et al. 1987a, c) 79.5% ~36% (Tang 1980) 80.6-81.3% 48% (Cohen et al. 1987a, c) (Tang and Munkelwitz 80% 37~40% 1994b) 791 332 (Cziczo et al. 1997) Ammonium sulfate 80% 42% (Dougle et al. 1998) (NH4)2SO4 75% 33% (Liu et al. 2008) 79.90.3% 37.90.5% (Ahn et al. 2010) (Laskina et al. 2015) 80.40.6% 37.74.1% HTDMA (Laskina et al. 2015) 82.32.5% 43.52.1% m-Raman (ten Brink and 60% <10% Veefkind 1995) Ammonium 62% 25~32% (Tang 1996) nitrate 60% <8% (Dougle et al. 1998) NH4NO3 61.80.3% N/A (Lightstone et al. 2000) 63% 32% (Lee and Hsu 2000)

When salt particles are dry, they are solid particles; while they are in paste or liquid states when salt particles absorb some water. Consequently, the salt particle dry and wet states can dramatically affect the filter performance. To better understand the filter performance under different RHs, dry and wet salt particle should be researched separately.

Two distinct techniques were performed to generate dry and wet salt particles before humidifying to the desired testing RH.

15 As aforementioned, salt particles are generated by a homemade Collision-type pneumatic atomizer from salt aqueous solution. To generate dry salt particles, salt droplets generated from the atomizer must be dried below the salt’s ERH before conditioning to the desired testing RH. Salt droplets were firstly dried by a diffusion dryer filled with silica gel desiccant, and dry clean air was supplied around the aerosol stream to desiccate particles and achieve the RH below its ERH. Conversely, to generate wet salt particles, the environment RH must be maintained above its ERH before conditioning to the desired testing RH. Hence, silica gel desiccant in diffusion dryer was removed, and dry clean air was reduced to maintain the RH above its ERH. Then, Salt particles were conditioned to desired testing RH by the introduced moisture mentioned in Section 1.3 Air Flow Control and RH Control.

The particle state control technique mentioned above relies on the accurate RH measurement and control before salt particles meet the generated moisture. The System

1100DP chilled mirror hygrometer (System 1100DP, GE General Eastern Instruments,

Wilmington, MA) can offer such precise RH measurement. According to Gao (2006), who reported that the ERH of (NH4)2SO4 depends on particle size, and ERH of the particles with diameter less than 100 nm could be as low as 29.7% RH. To generate dry salt particles, this effect should be considered. The RH before conditioning was controlled below 25% to prompt (NH4)2SO4 particles drying. The ERH of KCl is about 58%, and the RH before humidifying was controlled below 35%. Assuming KCl has the similar size eﬀect on the

ERH decreasing, 35% is suﬃcient to dry KCl particles.

For NH4NO3, it is impossible to generate dry NH4NO3 particles from its aqueous solution since it doesn’t have ERH. However, to compare the eﬀect of the RH history on

16 the filter performance of NH4NO3, two methods are applied, that is, keeping the RH as low as possible after particle generation and then conditioning to the desired RH (dry particle generation method); or directly conditioning to the desired loading RH (wet particle generation method). It is worth noting that the dry particle generation method cannot generate dry NH4NO3 particle.

1.5 Electric Control System and LabVIEW Program

The LAEAF Test Rig is automatic controlled by a customized LabVIEW program.

The customized LabVIEW program is the brain of the LAEAF Test Rig. It analyzes all input signals and transmits the output signals. The schematic diagram of the control system is demonstrated in Figure 1–6, the NI Data Acquisition chassis (cDAQ-9172, NI, Austin,

TX) connects most of the sensors and actuators, which is the bridge between the customized LabVIEW program and the computer. Two pressure sensors measure two test filter holders separately, and both transmit voltage signal to the NI-DAQ. Similarly, two chilled mirror hygrometers also transmit voltage signals to the NI-DAQ. Precise flow control is critical on the LAEAF Test Rig, and 3 MFCs adjust air flows of two test filters and the main exhaust accurately. The soot generator, which will be introduced detailly in

Section 2.1 Soot Generation, has 6 MFCs controlling propane, air, and nitrogen flows, which are all controlled by the NI-DAQ. The testing RH in the LAEAF Test Rig is PID controlled, and, as mentioned above, two proportional valves are voltage controlled by the

NI-DAQ. A solenoid valve that can shut-off the compressed air to the atomizer is also connected to the NI-DAQ. So that salt particles generation can halt when the testing filter

17 reaches preset pressure drop, which is useful in the study of Chapter 8. Since the AM2302 communicates in digital signal, an Arduino microcontroller is employed to modulate

AM2302 signals. The Arduino microcontroller communicates with the customized

LabVIEW program by serial communication.

Figure 1–6. Schematic diagram of the LAEAF Test Rig control system

To measure the challenging particle size distribution and filter efficiency during the testing, a SMPS is also connected to the computer. The SMPS communication is accomplished by the Aerosol Instrument Manager® from TSI. To measure the soot particle mass and mass fractal dimension, an Aerosol Particle Mass Analyzer (APM, Kanomax Inc.,

18 Andover, NJ, USA) is also employed, which is used in combination with the modified

SMPS system.

The electric heating elements are controlled with two temperature controllers in series. The first temperature controller is set to heating on-off control mode with the preset of 50 ℃, and it will cut off the power once the temperature inside the temperature- controlled chamber is above 50 ℃. The second temperature controller is set to heating PID mode so that the temperature inside the chamber could be precisely controlled.

1.6 Filter Sample Preparation

In this study, test filter media sheets are cut with a 47 mm circular cutter to fit in the test filter holders. The filter life could be indicated by the weight of deposited particles on the test filter when the test filter is loaded to a certain pressure drop. By weighing the test filter media sample before and after the loading, mass of salt loaded on test filter media can be calculated. However, it was observed that cellulose filter media could absorb moisture during storing or loading, and the amount of water absorbed could be even comparable to the weight of the salt loaded on the filter. The deposited hygroscopic salt particles can also absorb different amounts of moisture depending on surrounding RH levels. Therefore, to accurately measure the weight of the deposited salt on the test filter, it is critical to maintain the RH stable during the filter storing and weighing. According to

EPA PM 2.5 mass weighing SOP (EPA 1998), mean RH conditioning ﬁlters should be between 30% and 40%. Whereas, it is not applicable to this study because of the salts’ hygroscopicity may lead to inaccurate mass measurements. For example, when the

19 conditioning RH is between 30% and 40%, it is possible that (NH4)2SO4 loaded on test

ﬁlter media is not completely dried. This is because the ERH of (NH4)2SO4 is about 37%, which is in between the 30% and 40%, not to mention the ERH lowering for nanoparticles.

To solve this problem, a dry storing and weighing chamber was built, in which the RH is close to 0%. The low RH inside the chamber is achieved by charging the chamber with dried compressed air. All pre-cut test ﬁlters are stored in the chamber for at least 48 hours before weighing and loading. The post-loading weighing is conducted 24 hours after loading, which is suﬃcient to remove the water absorbed during the loading process. The filter sample preparation process is shown in Figure 1–7. A Cahn C-31 micro-balance

(Thermo Fisher Scientific Inc. Waltham, MA) is used to weigh test filters. One of the reasons that EPA regulates conditioning RH between 30% and 40% is that static electricity affects the weighing accuracy. In order to avoid the effect of the static electricity, 2 Po-210

Staticmaster® Nuclespot™ Alpha Ionizers (NRD, LLC. Grand Island, NY) were placed inside the balance chamber to eliminate the static electricity. There is a concern that

NH4NO3 would evaporate faster at the very low RH (Lightstone et al. 2000) An NH4NO3 loaded test filter was weighed one week after loading; no significant mass loss was observed compared to the mass measured 24 hours after loading. This results consistent with the conclusion of Lee and Hsu (2000), no significant evaporation loss of NH4NO3was found.

20 Condition new filter sample @ 0.1% RH

Weigh new filter sample

Load filter sample at testing condition to a certain pressure drop

Condition loaded filter sample @ 0.1% RH

Weigh loaded filter sample

Figure 1–7. Filter sample preparing process

21 Chapter 2: Soot Generation and Organic Compounds Soot Coating

2.1 Soot Generation

As mentioned in Section 1.3 Air Flow Control and RH Control, a homemade diffusion flame soot generator was design and constructed. The design of this soot generator was modified from a commercially available soot generator called Standard

Combustion Aerosol generator (Jing 1999). It has a closed combustion chamber with controlled propane, air, and nitrogen supply. The cross section drawing of the soot generator is shown in Figure 2–1. Propane is supplied to the center tube of the combustion chamber; air is delivered from the outside annular gap in the combustion chamber after a laminar flow screen. Since the combustion chamber is closed, a pair of the spark igniter is installed to ignite the flame.

One of the advantages is the fuel/air ratio could be changed to control the concentration of the generated soot. When it is the lean burn, the soot concentration is low, which could be used to warm-up the soot generator before filter tests, while the rich burn can be used to generate high concentration of soot particles for filter tests. Another advantage is that, comparing to the open flame design mentioned in Kim et al. (2009b), the soot generator can deliver the generated soot to the LAEAF Test Rig with an air-tight connection, with more flexibility in flow rates settings. Besides the closed combustion chamber, another advanced design is the transverse quench flow. A horizontal nitrogen flow is introduced to quench the freshly generated soot particles. The transverse quench 22 flow can cool down the stream temperature and bring soot particles away from the high- temperature zone above the flame. Since the pure nitrogen is used, it can prevent further oxidation and stabilize the morphology of soot particles. The extra dilution air can be added to reduce the soot concentration (top of Figure 2–1).

Figure 2–1. Cross section drawing of the homemade soot generator

The gas flows of the soot generator are all controlled by MFCs (see Figure 2–2), and the flow rates can be precisely controlled. As mentioned above, the soot concentration could be adjusted by changing propane/air ratio, as well as the dilution air flow rate. The

23 soot particle peak size depends on the flow rate of the nitrogen quench flow. The higher the quench flow, the smaller soot particle peak size due to reduced agglomeration time.

Samples of the soot particles collected is shown in Figure 2–3.

Figure 2–2. Photo of the homemade soot generator

Figure 2–3. Sample of soot particles generated by the homemade soot generator

24 2.2 Organic Compounds Coating Method

Soot generated from this soot generator could be regarded as the fresh soot when the oxidation gas flowrate is higher than a certain point (Barthazy et al. 2006). In the atmosphere, soot could be coated by various species, and the secondary organic aerosol is a very representative one. The aging of soot in laboratory could be achieved by coating fresh soot with organic compounds. A controlled amount of filtered compressed air will be used to bubble selected organic liquid and bring a certain amount of organic vapor to coat the freshly generated soot.

Similar to the soot coating process apparatus in Xue et al. (2009), Dalirian et al.

(2018), and Wang et al. (2018), a vaporization-condensation apparatus is design as Figure

2–4. The selected organic compound (e.g. glutaric acid) is placed in the mineral oil bath at

150 ℃. Filtered air flows into the organic compound bottle with constant flow rate to carry organic vapor to the outlet of the soot generator. A cooling fan is installed on the upper portion of the stainless tube to cool down the organic compound vapor and soot stream so that the organic compound vapor will condense on the soot particles, which is similar to the working principle of condensation particle counter. The coated soot particles will then be introduced to the LAEAF Test Rig.

Figure 2–4. Soot organic compound coating apparatus

26 Chapter 3: Smart Sensor for Filter Performance Monitoring System

3.1 Background

Currently, air filters are installed and operated with limited performance monitoring system, such as pressure drop. Meanwhile, criteria of filter replacements are mainly based on the installed time or mileage, for example, 3 months for residential HVAC filters,

~12000 miles for vehicle cabin air filters. However, the actual filter operating time and pollutant concentration are not considered. It is quite common that air filters are used with prolonged time or cleaned with compressed air and then reused. However, the performance of air filters is not closely monitored, especially the efficiency.

HVAC Air filters are designed to protect people staying indoor, and occupants’ health is invaluable. It has been a rising concern that the HVAC system may transmit

COVID-19, and ASHRAE recommends that MERV-13 or above level filtration should be used when it is achievable (ASHRAE 2020). The purpose of using high-level filters is to capture virus aerosol in room air, whereas some HVAC engineers worry the improper installation of air filters or the leaks around filter edges may jeopardize the overall filtration efficacy. Improper installation or defective filters can give people false sense of safety, which is not desired. Therefore, it is critical to ensure the HVAC filtration system is operating as designed.

Gas turbines, internal combustion engines, and dust collection systems are all using air filters. Comparing to the equipment behind air filters, the price of air filters is minimal, 27 not to mention the repair cost if the equipment is damaged due to defective air filters or improper installations.

Without sufficient air filter performance information, it is unclear whether air filters are functioning as designed. When switching to a new filter brand, it is unrealistic for end- users to judge the quality of the new filters. The uncertainties of air filters performance make people or equipment behind the air filters more vulnerable. On the other hand, the lack of information could cause early or prolonged filter replacements. When air filters are not replaced confidently, it could be either a waste of filters or loss of protections or waste of energy to overcome extra filter resistance.

Concluded from examples above, filter performance monitoring system is highly demanded. Currently, there are several commercially available filter monitoring sensors.

Table 3–1 lists four sensors that can monitor air filters. In fact, all four sensors are based on the pressure difference measurements, and none of them could measure filter efficiencies, temperature and RH. 3M smart filter and Clean Alert filter monitor are designed for residential HVAC applications. 3M smart filter uses Bluetooth to connect with customer’s cell phone and it will alert when it is time to replace the filter. However, the smart sensor is glued on the HVAC filter and cannot be reused, therefore its cost is higher than conventional HVAC filters. The Clean Alert filter monitor can only send a message to customers when it is time to replace the HVAC filters. The other two industrial filter monitoring sensors have the online monitoring functions, with paid subscription. The

Donaldson iCue is designed for dust collectors, which gives the differential pressure measurement to monitor and predict filter’s life. Mann+Hummel Senzit is designed for the truck intake air filter monitoring. It is installed at the downstream of the intake air filter.

28 The data is transmitted through the cellular network and the location of the truck is also reported. It is designed for fleet managers to arrange the maintenance in fleet base to reduce the fleet operation cost.

Table 3–1. Comparison of current commercially available products Filter Pressure Online RH, T Product Application Efficiency transducer Monitoring Monitoring Monitoring No Residential 3M smart filter Single Bluetooth No No HVAC only Clean Alert Residential Single or No No No filter monitor HVAC Differential Industrial Yes Paid Donaldson iCue dust Differential No No Subscription collector Mann+Hummel Industrial Yes Paid Single No No Senzit Fleet Subscription

The current available commercially available sensors are not capable in RH and temperature measurements. In addition, filter efficiency is also not reported. The actual protections that filters can offer is unclear. With the current monitoring methods, It is highly unlikely to detect the improper filter installation and defective filters. Therefore, it is vital to design a filter monitoring system with filter efficiency measurement capability.

With that, filter operation will not be a “black box”, and the protections that filters can offer can be reported. At the same time, the comprehensive filter operation condition could be recorded, which could be used to predict the filter life more accurate. For example, the temperature and RH effect on the filter life will be discussed in Part 2 later.

29 3.2 System Description

To measure the filter efficiency, the particle concentrations at the upstream and downstream of the filter are needed. Similarly, the differential pressure can be measured by a pair of pressure sensors. Figure 3–1 shows the schematic of the filter performance monitoring system. Upstream and downstream measurement signals are processed by an

Arduino microcontroller, and an I2C multiplexer is employed to switching the I2C addresses of each sensor. Currently, the filter operating parameters are uploaded to the thingspeak.com via Wi-Fi, which is an open platform for IoT devices. Thingspeak.com can be accessed from a computer, a laptop, a tablet, or a cellphone, therefore, the filter operating parameters are available wherever there is internet access.

Figure 3–1. Schematic of the filter performance monitoring system

The microcontroller used in this system is Arduino Uno Wi-Fi R2, the most common microcontroller for electronic prototyping. The Arduino Uno Wi-Fi R2 has the

30 built-in Wi-Fi module so it can connect to the internet. The communications between the microcontroller and sensors are accomplished by I2C protocol. Each of the sensor has a fixed I2C address. Since particle sensors and environmental sensors are used in pairs, the

I2C addresses will be duplicated. A I2C multiplexer is employed to switch the communications between sensors, so that the data transmission is achieved sequentially.

The particle sensor used in this system is Sensirion SPS30, which is a commercially available optical particulate matter sensor. According to the manufacture, the innovative contaminant-resistant technology can keep the detection chamber clean to report accurate concentration throughout its life for at least 8 years continuous operation. The mass concentration limit is up to 1000 µg/m3, that is high enough for the ambient pollutant application. Another advantage of the SPS30 is that it can offer 4-channel particle size concentrations, PM1.0, PM2.5, PM4.0, PM10. The particle size information is particularly important to characterize the particulate pollutants and monitor the filter performance.

The environmental sensor used in this system is Bosch BME680, which is a multifunction sensor. It can measure temperature, RH, pressure, and breathable VOC. The

RH detection range is 0 ~ 100 %; The temperature detection range is -25 °C ~ 85 °C; The pressure range is 300 hPa ~ 1100 hPa. All three parameter’s operating ranges are suitable for the ambient air measurement. At the downstream of the filter, the pressure is lower than the ambient pressure, so this sensor is suitable for the absolute pressure measurements.

This sensor has a small footprint of 3.0mm × 3.0mm and it is a surface-mount device that can be mounted on a customized printed circuit board in later design. In the current design a sensor module with larger footprint is used for prototyping purpose.

31 A photo of the filter performance monitoring system is shown in Figure 3–2. An additional display is installed to show the operating status. The one shown in the photo is made to compare the sensor readings parallelly. All readings are within 2.5% error and hence sensors are interchangeable as claimed.

Figure 3–2. Photo of the filter performance monitoring prototype

The cost of the filter monitoring system is another aspect worth considering. The bill of material of the current prototype is listed in Table 3–2, and the total cost is about

$190. Unlike some sensors that are attached to air filters, this filter performance monitoring system doesn’t need extra replacement except battery. The total cost can be regarded as the initial investment, which can be averaged for the life of the system. It should be noticed that the Arduino is used here since it is a common platform for prototyping conveniently, and the BME680 used here is also mounted on a prototype circuit board. The actual cost of the filter performance monitoring could be reduced if Arduino is replaced with ESP32

32 and a customized printed circuit board is design for ESP32 and BME680. In that way the total cost could be below $120.

Table 3–2. Bill of material of the current prototype Item Price/unit Quantity Notes Arduino Uno WiFi R2 $44.95 1 Microcontroller Sensirion SPS30 $46.95 2 Particle sensor Bosch BME680 $19.95 2 Environmental sensor TCA9548A I2C Multiplexer $6.95 1 Communication component Other (wires, breadboard) $5 Total Cost $190.7

Two case studies will be covered here to project potential applications of this smart sensor for filter performance monitoring system.

3.3 Case study 1: Indoor Air Purifier

Indoor air purifiers are common residential air cleaning devices. This prototype is installed on an indoor air purifier (Oreck Air Instinct 200, Cookeville, TN) to monitor the

HEPA filter performance. Figure 3–3 shows how the filter performance monitoring system is installed on the air purifier. The Arduino and other circuit boards are enclosed in an acrylic box, which is mounted on the front panel of the indoor air purifier. The upstream module is mounted on the inside of the front panel to measure operating parameters of the

HEPA filter, which is measuring the indoor air quality, absolute pressure, temperature, and

RH. Behind the HEPA filter, the downstream module is installed, to detect the air quality, temperature, and RH after HEPA filtration. In Figure 3–3, the HEPA filter is not in place so that the downstream module can be seen. It was found that the HEPA filter paper frame

33 doesn’t seal tightly, and thus extra foam weather-strip is applied around the HEPA paper frame to improve the sealing between the HEPA filter and the air purifier.

Figure 3–3. The filter performance monitoring system on an indoor air purifier (HEPA filter is not in place to show the downstream module)

The sample readings of this system are listed in Table 3–3, where the air purifier is set at medium fan speed. Both mass and number concentrations of PM1.0, PM2.5, PM4.0,

PM10 are reported. It can be found that for the sample readings showing below that overwhelming particles are PM1.0, with limited PM2.5 particles, as the readings of PM2.5,

PM4.0, PM10 are the same. Since PM10 concentrations representing all detected particles, both PM10 mass and number efficiencies are calculated at the bottom of Table 3–3. The mass and number efficiencies are slightly less than the nominal HEPA efficiency of

99.97%. which might be due to leaks around the HEPA filter, though weather-strip is applied. This is a good example that even with careful sealing, the overall filtration efficacy is less than the designed filtration efficiency. The non-sealed filtration efficiency will be

34 discussed later in this section. In addition, the SPS30 sensor can also report the average particle size, which is 0.47 µm and 0.39 µm at upstream and downstream, respectively.

Thus, a slight particle size distribution shift occurred during filtration. The indoor air PM2.5

3 mass concentration is 1.94 µg/m , which is far below the EPA annual PM2.5 standard, 12

3 µg/m . Whereas the PM2.5 mass concentration at the downstream of the air purifier is only

0.04 µg/m3. This indicates the effectiveness of an air purifier in removing indoor particulate matters. At the meantime, the absolute pressures, temperatures, and RHs are also reported.

The HEPA filter differential pressure could be calculated based on the pressure data, which is 44 Pa.

Table 3–3. Sample readings of the filter performance monitoring system at medium fan speed Upstream Number Concentration / #/cm3 Average 0.47 µm PM1.0 PM2.5 PM4.0 PM10 Particle Size 14.51 14.61 14.61 14.61 Pressure 97878 Pa Mass Concentration / µg/m3 Temperature 25.63 ℃ PM1.0 PM2.5 PM4.0 PM10 Relative 64.09 % 1.83 1.94 1.94 1.94 Humidity Downstream Number Concentration / #/cm3 Average 0.39 µm PM1.0 PM2.5 PM4.0 PM10 Particle Size 0.28 0.28 0.28 0.28 Pressure 97834 Pa Mass Concentration / µg/m3 Temperature 25.9 ℃ PM1.0 PM2.5 PM4.0 PM10 Relative 63.2 % 0.03 0.04 0.04 0.04 Humidity

PM10 Number Efficiency PM10 Mass Efficiency Differential Pressure 98.08% 97.94% 44 Pa

The air purifier used here has four fan speeds, low (100 CFM), medium (180 CFM), high (240 CFM), and turbo (350 CFM). With the same filtration area, the higher the fan

35 speed, the higher the differential pressure. A comparison of differential pressures at different fan speeds is shown in Figure 3–4. The average differential pressures for each fan speed are calculated from 75 sample points, when the fan speed is stable. It could be found a very clear differential pressure raise with increasing fan speed. As mentioned before, the

HEPA filter does not fit the air purifier tightly, non-sealed differential pressures at various fan speeds are also shown in Figure 3–4. It is reasonable that differential pressures of non- sealed are lower than the sealed conditions due to leaking points. However, differential pressure can only be used to verify the installation when the leak is significant enough. It is impossible to tell whether the HEPA filter is sealed and whether the HEPA filter is performing as expected only by comparing the differential pressures information.

Sealed 58.5 ) Non-Sealed

a 60

P 51

(

e r 42.6

s 37.2 s 40 34

e r 29.6

l 25.7

a 23.4

r 20

0 Low Medium High Turbo Fan Speed

Figure 3–4. Differential pressures of the HEPA filter at different fan speeds

To verify that the HEPA filter is performing as expected, the filtration efficiency is the only criterion. Figure 3–5 shows the comparison of mass and number efficiencies of

PM10 at various fan speeds. Significant efficiency discrepancies show that the non-sealed

HEPA filter in the air purifier can only achieve about 70% ~ 80% overall filtration efficacy. 36 The clean air delivery rate is overestimated when the actual filtration efficacy is below the designed. This condition may also happen to the residential HVAC furnace filters, whose frames are also made of cardboard without sealing foam around them. Hence the protection

HVAC furnace filters can offer is also overestimated due to the imperfect installation.

Since indoor air purifiers and residential HVAC furnace filters are circulating indoor air in a room which can be regarded as a closed space, the deficient clean air delivery capability can be mitigated by extra circulation and longer operating time. However, when air filters are handling air with the “single pass” application, such as engine intake air filter, the deficient filtration capability can result in severe damage to the engine.

Sealed, PM10, Number-based 150% Sealed, PM10, Mass-based

) Non-Sealed, PM10, Number-based

(

Non-Sealed, PM10, Mass-based

c 95% 95% n 98% 98% 97% 97% 98% 98%

i 100%

c 80% 81% i 74% 75% 71% 73% f 72% 73%

a 50%

0% Low Medium High Turbo Fan Speed

Figure 3–5. Filtration efficiencies of the HEPA filter at different fan speeds

From the case study of the indoor air purifier, it can be found that the differential pressure measurement cannot offer sufficient information to verify that filters are performing as designed and there are potential hazards to occupants or equipment behind air filters. A pair of particle sensors across the air filter can support a confident operation

37 with filtration efficiency reported. Or at least a single particle sensor at the downstream of air filters can offer clean air delivery capability when the absolute cleanness is required.

3.4 Case study 2: Delhi SALSCS Operating Monitoring

A solar-assisted large-scale cleaning system (SALSCS) was designed to handle the regional air pollutions (Cao et al. 2015, Cao et al. 2018a, Cao et al. 2018b). The first generation of SALSCS works based on the stack effect, that the heated ground air under the transparent solar collector can flow to the chimney by passing the air filter bank. The ambient air can be cleaned in this way to mitigate regional air pollutions. In the second generation SALSCS, the air flow is reversed, a fan is installed in the chimney to push the ambient air to the air filter bank and supply the clean air to the ground level. A demonstration system of SALSCS has been built in Xi’an, China.

The India air quality is a concern for the public health. According to the data compiled by IQAir, among the 30 cities with worst air pollution globally, 21 cities are from

India (IQAir 2020). Many urban area in India are experiencing air pollutions far higher than EPA standards (Guttikunda and Calori 2013). To reduce the air pollution burden, India government plans to reduce the PM2.5 by 30% in 5 years. The SALSCS demonstration system in Xi’an proves that the SALSCS is a useful tool to improve regional air quality, hence Delhi is planning to build multiple second generation SALSCS to be compliant with the 5-year air quality target. The filter performance monitoring system can be installed on the SALSCS to offer filter and system operating monitoring. The schematic diagram of the

Delhi SALSCS and filter performance monitoring system is shown in Figure 3–6.

Figure 3–6. Schematic diagram of the Delhi SALSCS and filter performance monitoring system (modified from Dr. Sheng-Chieh Chen’s diagram)

The filter monitoring system is mainly shown on the right side of Figure 3–6, and a detailed sensor location diagram is shown in Figure 3–7. 16 yellow circles representing the sensor module (particle sensor SPS30+ environmental sensor BME680) are installed at the upstream and downstream of the air filter bank (green zigzag). There are two pairs of sensor modules on each side of the filter bank to monitor the filter performance (efficiency, differential pressure). Because the filter wall has a large area, two sensor modules are used to sensing more points. Besides the filter performance monitoring system, 5 TSI DustTrak

(8543, Shoreview, MN) are also installed around the SALSCS. There is one DustTrak inside the chimney to sample the inlet ambient air, and the other four are placed on each downstream side to sample the filtered air. The cost advantage of the described filter performance monitoring system can be found here. 16 sensor modules can be deployed while only 5 DustTrak is planned due to the cost consideration. The more the sample points, the more accurate the system operating information can be collected. In addition to the filter efficiency monitoring, the fan speed control is also integrated here. 9 TSI 39 micromanometers (Alnor EBT730, Shoreview, MN) are proposed to measure the face velocity so that the fan speed can adjust accordingly.

Figure 3–7. Locations of filter performance monitoring system sensors and other measurement equipment in the Delhi SALSCS

The sensing and control system is shown in Figure 3–8. With aforementioned sensors, the filter resistance, filtration efficiency, and clean air delivery rate (CADR) can be calculated while the fan speed can be adjusted based on the CADR requirement. With the filter performance monitoring system reporting the filter resistance and filtration efficiency, the SALSCS system can operate more efficiently and effectively and the filter replacement can be more confidently.

Figure 3–8. Sensing and control system diagram of the Delhi SALSCS

3.5 Conclusion

With case studies described above, it could be concluded that the filter performance monitoring system can offer the filtration efficiency of air cleaning devices together with the differential pressure, temperature, and RH. This improvement of the filter monitoring enables direct reading of the filtration efficiency, which is the critical parameter to the air cleaning devices. The filter will not be a black box with the aid of this system, and the protections an air cleaning device offers can be quantified. Residential customers can rest assured with their air purifiers or HVAC furnace filter; industrial customers can monitor the filtered air cleanness that will enter their internal combustion engines or gas turbines; 41 filters or air cleaning devices manufactures can utilize the collected operating parameters to better design and improve their products or to recommend proper products to their customers.

42 Chapter 4: Global Air Pollutant/Meteorological Condition Database

With the understanding of the effect of the different particulate pollutants and operating conditions on the filter performance, which will be discussed in Part 2, the filter selection can be decided based on the location to gain a longer filter life. In addition, the atmosphere pollution trend is also important information to estimate the filter life. Some international organizations (World Health Organization, World Bank) and air pollution protection agencies (United States Environmental Protection Agency (EPA), Ministry of

Ecology and Environment of China (MEE), and European Environment Agency (EEA)) have air pollution database. Their data has a very wide coverage, whereas some data is not up to date. For some locations, only annual average data is available. Meanwhile, meteorological data is not included in the database.

Considering the insufficient information current databases can offer, a global air pollutant/meteorological condition database is built on the Center for Filtration Research website. It is designed to offer monthly average temperature, RH, criteria air pollutant concentration of the city of interest for past 5 years. It can also show the current weather conditions and air quality index (AQI) of the city of interest. The current weather is retrieved from the OpenWeather, a platform that provides weather data through application programming interface (API). An inquiry will be sent to the OpenWeather when a city of interest is input, and the requested weather data will be returned within a second. The current AQI data is retrieved in the same method as the weather data from aqicn.org, who has a World Air Quality Index project, by its API. 43 Currently, historical air pollution data cover the US cities and Chinese cities. The data source of the US cities is the US EPA, and EPA has the API available to access its Air

Quality System (AQS). The AQS can offer ambient air sample data across the country, and it also includes the meteorological data and PM2.5 speciation data. Similarly, MEE of China announces its ambient air monitoring data of more than three hundred Chinese cities hourly, and a website (www.aqistudy.cn) recorded the published data. The China data used in the database here are retrieved from the aqistudy.cn. Both data collected are raw data and they are processed to the monthly average to show the pollution trend in a year. Figure 4–1 is an example of monthly average PM2.5 of Beijing, Shanghai, and Guangzhou from 2014. It could be found that the PM2.5 concentration shows a seasonal variation. The standard deviation of Beijing’s data is very high, which is due to PM2.5 declining in recent years.

140 Beijing Shanghai 120 Guangzhou

100

)

m / 80

(

2 60

P 40

0 Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec

Figure 4–1. Monthly average PM2.5 concentration of Beijing, Shanghai, and Guangzhou from 2014.

44 Figure 4–2 is the yearly average PM2.5 concentrations of Beijing, Shanghai, and

Guangzhou from 2014. It could be found that there is a significant decay of Beijing’s PM2.5 concentration. The decay partially explains the high standard deviation of Beijing in Figure

4–1. The similar plots could be generated on the Center for Filtration Research website for any city of interest.

120 Beijing Shanghai 100 Guangzhou

)

(

M P 40

0 2014 2015 2016 2017 2018 2019 2020

Figure 4–2. Yearly average PM2.5 concentration of Beijing, Shanghai, and Guangzhou from 2014

Meteorological data are also included so that the historical ambient temperature and

RH could be used as references to select air filters. There are several data sources for the historical meteorological data, including EPA, weather-and-climate.com, and weather- atlas.com. Their data are all included in the database as references.

All processed data are visualized on the CFR website for quick review. Figure 4–3 is the sample data replotted from the processed data, and Phoenix and Minneapolis are 45 example cities. It can be found that the temperature band of Phoenix is above 0 ℃ all year around, while the temperature band of Minneapolis is shifted to below 0 ℃ in winter.

Ambient air filter may encounter different RHs in Phoenix and Minneapolis. The RH in

Minneapolis is relatively stable, fluctuating around 70%, whereas the RH in Phoenix can be as low as 20% in June and the highest RH is below 60%. The RH is an important factor that may affect filter life, which will be covered in Part 2.

The fine particle and coarse particle ratio may also affect the filter life according to the previous study (Ou et al. 2014). More importantly, the PM2.5 composition can also affect the filter life, such as the percentage of the hygroscopic particles. It has been found that non-hygroscopic particles, such as aluminum oxide and fine dust, are not sensitive to the RH during filter loading (Gupta et al. 1993, Ou et al. 2015); while hygroscopic salt particles filter loading can be affected at different RHs (Gupta et al. 1993, Joubert et al.

2010, Montgomery et al. 2015b, Pei et al. 2019a, Pei et al. 2019b, Ribeyre et al. 2017). The filter selection could be done with the supportive information from the database. The sulfate and nitrate in Phoenix PM2.5 are only 24%, while they account for more than a half in the Minneapolis PM2.5. The PM2.5/PM10 ratio can be used to infer the fine and coarse particle ratio. The PM2.5/PM10 ratio of Minneapolis is about 13% higher than that of

Phoenix. Both the particulate pollutant composition and fine/coarse particle ratio can be used to recommend appropriate air filters so that longer filter life could be achieved.

Figure 4–3. Sample data extracted from the global air pollutant meteorological condition database

Part 2: Temperature and RH Effect on Filter Performance

In Part 2, temperature and RH effect on filter performance will be discussed. Since

RH is temperature dependent, the temperature effect on the filter performance will be discussed in Chapter 5 in the first place. It was found that temperature effect on the filter loading is minor, comparing to the RH effect. So that the rest of studies were conducted at room temperature. There are two conditions to study RH effect during filter operating: constant RH and varying RH. Chapter 6 and 7 studied the constant RH filter loading with single species and multiple species, accordingly. Chapter 6 covers the hygroscopic salt filter loading on conventional cellulose filter media, nano-fiber coated filter, and the efficiency evolution for different salt states. Chapter 7 covers the (NH4)2SO4 and NH4NO3 mixture filter loading study, and organic compounds coated soot filter loading study. Lastly, the varying RH condition was studied on the single species salt loaded air filters in Chapter

49 Chapter 5: Effects of Temperature on Filter Loading by Hygroscopic Salts at Different RH

This section has been published: Pei, C., Ou, Q. and Pui, D. Y. H. (2020). Effects of Temperature and Relative Humidity on Laboratory Air Filter Loading Test by Hygroscopic Salts Separation and Purification Technology 255: 117679. doi:10.1016/j.seppur.2020.117679

5.1 Introduction

Air filters are designed to remove particulate pollutants from ambient air. Once they are installed, they could operate in various climates, including tropical wet, arid,

Mediterranean, humid continental, tundra, highland climate, etc. However, as mentioned in Section 1.1, the filter testing standards only specify a testing temperature range close to room temperature or the testing temperature is not mentioned (see Table 1–1); the RH range is broad in an uncontrolled fashion as well. One of the important composition of the ambient air particulate pollutants is the hygroscopic portions, including sulfate, nitrate, ammonium, etc. (Cheng et al. 2016, He et al. 2001, Philip et al. 2014, Snider et al. 2015,

Snider et al. 2016, Sun et al. 2004, Yang et al. 2011, Yao et al. 2002, Zhang et al. 2007).

The hygroscopic pollutant is sensitive to the surrounding temperature and RH, which are very important factors affecting the particulate pollutants properties, such as morphology.

Therefore, the performance of air filters removing the ambient particulate pollutants could be also affected by operating temperature and RH due to existence of the hygroscopic pollutant.

The air filter performance under the effect of operating temperature has been studied little. Most of the studies of air filter performance were conducted at room temperature. Several temperature effects on the coalescence filter have been discussed by 50 Hajra et al. (2003), and they found that the higher the temperature, the higher the filter performance. It is worth noting that RH depends on temperature. Therefore, it is important to study that whether the RH effect is valid only at room temperature or it is valid in a broader temperature range. To describe the RH at a certain temperature, absolute humidity

(AH) could be used, which is independent to the temperature. Hence, the question of whether the temperature affects filter performance at a given RH could be understood as the effect of AH on filter performance. For a given AH, different temperatures will lead to different RHs, such as AH of 75% RH at 5℃ and 30% RH at 19℃ are very close. By combining the temperature and RH to AH, not only the filter performance of the same RH at different temperatures could be compared, but that of the same AH at different temperatures could also be compared (see Figure 5–1).

The focus of this study is the temperature effect on filter performance, and the RH effect will be covered in detail in Chapter 6, which has been studied a little previously. In general, the operating RH is positively correlated with the filter holding capacity, that is, the higher the operating RH, the greater the holding capacity at a given terminal pressure, which indicates a longer filter life (Gupta et al. 1993, Joubert et al. 2010, Joubert et al.

2011, Miguel 2003, Montgomery et al. 2015a, Montgomery et al. 2015b, Pei et al. 2019a,

Pei et al. 2019b). Greater holding capacity at a given terminal pressure reveals a slower pressure increasing rate at higher operating RH. Among all the mentioned studies, different filter structures and different hygroscopic salt particles have been studied. It is agreed that the filter structure is a minor factor regarding the slower pressure increasing rate. The particle to particle interaction plays an important role in reducing the pressure increasing rate. Deposited hygroscopic particles stick to each other to form the dendrite on the filter

51 with small capillaries between each other. The capillary condensation introduces surface water into small gaps between hygroscopic particles even when the surrounding RH is below salt’s DRH. Hygroscopic particles can coalesce or pack themselves denser, which reduces the specific surface area, therefore less resistance to the incoming airflow. The higher the surrounding RH, the more water can condense, consequently the greater the mass loading.

The filter life is a very important factor affecting filter operation, hence it is worthwhile to pay attention to the temperature and RH effect on the filter holding capacity.

To better simulate ambient particulate pollutants, (NH4)2SO4 and NH4NO3 were selected, as ammonium, nitrate, and sulfate are the most abundant inorganic ionic components in ambient aerosols and PM2.5(Cheng et al. 2016, He et al. 2001, Philip et al. 2014, Snider et al. 2015, Snider et al. 2016, Sun et al. 2004, Yang et al. 2011, Yao et al. 2002, Zhang et al.

2007). As mentioned in Section 1.4, NH4NO3 cannot be dried and remain in wet state, so it is used to represent wet ambient salt particles; (NH4)2SO4 represents dry hygroscopic ambient salt particles. In addition, a common lab salt, KCl was also studied. In this study, three salts were studied individually under different operating conditions (temperature, RH and dry/wet particle state). The holding capacities of each operating condition were compared to verify the primary factor affecting filter performance.

5.2 Experimental Method and Material

This study was conducted with the LAEAF Test Rig described in Chapter 1. The testing temperature and RH were well controlled.

52 As aforementioned, (NH4)2SO4, NH4NO3 and KCl were selected to represent dry and wet ambient hygroscopic salt particles and common lab filter test salt particles, respectively. The salt particle dry/wet state control technique has been discussed in Section

1.4. For each salt, three temperatures were selected. (NH4)2SO4, NH4NO3, and KCl were tested at two, three and one RHs at each temperature, respectively. 1 % volume concentration salt solutions were atomized by the homemade pneumatic atomizer and droplets are dried by a diffusional dryer (see Figure 1–4). The count median diameter is about 100 nm and the geometric standard deviation is about 1.8.

For simplicity, the testing matrix can be several testing RH at different testing temperatures. However, as mentioned in Section 5.1, AH can represent the combination of

RH and temperature. To better compare the effect of AH on filter holding capacity, testing temperatures are carefully picked so that the AH of low temperature high RH is comparable to that of high temperature low RH. For example in Figure 5–1, the AH of 5℃ 75% RH is

5.09 g/m3, and the AH of 19℃ 30% RH is 4.89 g/m3; similarly, the AH of 19℃ 75% RH is 12.23 g/m3, and the AH of 35℃ 30% RH is 11.88 g/m3. In Figure 5–1, testing temperatures and RHs of NH4NO3 were selected in the same fashion. Since the 75% RH

NH4NO3 filter loading is unstable due to its wet particle state. 45% RH was selected to compare the filter mass loading. In addition to 45% RH, 60% RH loading was also performed for 9℃ and 16℃ to compare the AH and RH effect. KCl was only tested at 30%

RH for 10℃, 30℃, and 50℃. At each of the testing condition mentioned above, at least three samples were repeated.

53 10 29.69 9.26 30 30% RH 30% RH

) ) 8.18

3 75% RH 3 45% RH 8

/ / 60% RH

(

6.17

y y 6.13

i 20 i 6

i i 5.29

u u 3.97 4.09

12.23 11.88 4

l 10 l 2.65

b 5.09 4.89 b 2

A 2.04 (NH4)2SO4 NH4NO3 0 0 0 10 20 30 40 0 5 10 15 20 25 Testing Temperature (°C) Testing Temperature (°C)

Figure 5–1. (NH4)2SO4, NH4NO3 testing points and corresponding AHs. Testing temperatures are carefully picked so that the AH of low temperature high RH is close to the high temperature low RH (two dotted lines indicate two similar AHs).

The test filter media used in this study is a composite of a synthetic nanofiber layer and conventional cellulose fiber filter media. It is used in the application of air intake particle removal. Figure 5–2 is the scanning electron microscope (SEM) image of a clean test filter sample used in this study. Thin fibers are synthetic nanofibers (nylon). Larger cellulose fiber is beneath the synthetic nanofiber layer, which is much thicker than the synthetic nanofibers. Though filters were tested under different conditions, all filters were prepared and weighed according to Section 1.6. Since the densities of three salts used in the study are different, we used the volume loading to represent the amount of the material deposited on filters. The volume loading could be calculated by dividing mass loadings with corresponding salt densities.

Figure 5–2. SEM image of a clean filter sample. Thin fibers are synthetic nanofibers (Nylon), and thick fibers beneath the nanofibers are cellulose fibers

5.3 Results 5.3.1 Dry KCl

KCl was only tested at 30% RH for 10℃, 30℃, and 50℃, since KCl effloresces around 59% RH (Ahn et al. 2010, Cohen et al. 1987a, c, Freney et al. 2009, Li et al. 2014),

KCl particles are at solid state at 30% RH. The KCl volume loading at 4 inches of water is plotted as a function of loading AH in Figure 5–3. It could be found that at the same RH

30%, the AH of 50℃ is about eight times of that of 10℃, and the AH of 30℃ is about three times of that of 10℃. However, the KCl volume loadings are essentially 0.3 mm3/cm2 for all three temperatures.

55 ) 2 0.4

m ( 0.3

4 0.2

g 10°C 30°C 50°C

a 0.1

e RH

u 30%

o 0 V 0 5 10 15 20 25 30 Loading Absolute Humidity (g/m3)

Figure 5–3. Volume loading of dry KCl at different loading AHs

The comparison between three different loading temperatures could be achieved by the analysis of variance (ANOVA) test to check whether any significant difference between them. In the KCl loading test, the temperature is the only independent variable, which affects filter loading results. One-way ANOVA was used to analyze data sets.

In this ANOVA test, the null hypothesis is that all volume loadings of three temperatures are the same, and the alternative hypothesis is that at least one volume loading is different from that of other temperatures. The ANOVA test was processed with R and

RStudio (R Core Team 2019, RStudio Team 2019). The ANOVA test result is shown in

Table 5–1. The Sum Sq column is a measurement of the variation, which is the sum of squares of the difference between samples and corresponding means. The Df column is degrees of freedom. The Mean Sq is Sum Sq/Df, which is similar to standard deviation. The

Temperature row lists the above results between different temperatures; the Residuals row lists the above results of each sample inside its testing temperature. The F value is the ratio of Mean Sq of Temperature and Residuals. The Pr(>F) is the probability that is larger than 56 the F value given the null hypothesis is true. Generally, if the Pr(>F) is less than 0.05, one could reject the null hypothesis and accept the alternative hypothesis. In the ANOVA test for KCl, the Pr(>F) is 0.8254 which is much greater than 0.05, indicating that it is highly likely that the null hypothesis is true. Hence, the temperature effect of the volume loading is insignificant, in another word, the AH doesn’t affect the volume loading substantially.

Table 5–1. Analysis of Variance Table of KCl Sum sq Df Mean sq F value Pr(>f) Temperature 0.001022 2 0.0005108 0.1972 0.8254 Residuals 0.01813 7 0.002590

5.3.2 Dry (NH4)2SO4

(NH4)2SO4 was also tested in their dry state, while under two RHs (30% and 75%) and three temperatures (5℃, 19℃ and 35℃). The reason for RH and temperature selection has been stated in Section 5.2. The volume loading of (NH4)2SO4 was depicted in Figure

5–4. The volume loadings of each loading RH are grouped. The average volume loading of 30% loading RH is less than that of 75% loading RH, while the volume loading varies among the loading temperatures.

57 0.7 Loading Temperature

)

2 5°C

m 0.6 c 19°C

/ 3 35°C

( 0.5

i 0.4

@ 0.3

o 0.2

l 0.1

V 0 30% 75% Loading Relative Humidity (%)

Figure 5–4. Volume loading of dry (NH4)2SO4 at different loading temperatures and RHs

When volume loadings are plotted as a function of loading AH (Figure 5–5), it is much clear that RH is a more significant factor compared to AH, in another word, temperature. Two pairs of the volume loadings of AH 5 g/m3 and 12 g/m3 are nice examples.

The volume loadings are comparable for the same loading RH, and no significant change for the loading AH increases. However, it is a substantial difference between different loading RHs while loading AH remains the same. It is important to note that for 75% RH cases, the volume loadings are comparable when the loading AHs are about 5 g/m3 and 30 g/m3, which is 6 times increase.

Figure 5–5. Volume loading of dry (NH4)2SO4 at different loading AHs

Similarly, the results of (NH4)2SO4 are analyzed by ANOVA. For KCl, only a single factor, loading temperature, is considered. Hence one-way ANOVA was used for

KCl. For (NH4)2SO4, two-way ANOVA is used, since both loading RH and temperature are considered. When there are two factors, the interactions between two factors may also affect outputs. In Table 5–2, three rows of temperature, RH, and their interactions

Temp:RH are listed. In the Pr(>F) column, both Pr(>F) of the Temperature and RH are less than 0.05, and Temp:RH is greater than 0.05. From the ANOVA table, both

Temperature and RH are significant factors affecting the volume loadings by 0.05 criteria.

However, the Pr(>F) of RH is much less than that of Temperature (1.977e-11 vs.

0.000672). These results could be interpreted as that both temperature and RH affect the volume loading, whereas the loading RH effect is much more significant compared to the temperature. This is consistent with the KCl results, that the loading RH plays a dominant role in volume loadings.

59 Table 5–2. Analysis of Variance Table (Type II test) of dry (NH4)2SO4 Sum Sq Df F value Pr(>F) Temperature 0.06338 2 9.010 0.000672 RH 0.3222 1 91.61 1.977e-11 Temp:RH 0.01701 2 2.419 0.1034 Residuals 0.1266 36

5.3.3 Wet NH4NO3

Unlike KCl and (NH4)2SO4, NH4NO3 stays in its wet state when it is generated from its aqueous solution since its efflorescence RH doesn’t exist (Dougle et al. 1998, Lightstone et al. 2000). Therefore, NH4NO3 was tested in its wet state at 30%, 45%, and 60% RH. It represents the wet particles in the atmosphere, which is not limited to NH4NO3, and other salts could also be in the wet state. Two 60% RH test conditions at 9℃ and 16℃ are added besides the 30% and 45% RHs, whose AH is indeed less than that of the 45% RH at 16℃ or 23℃, respectively (see Figure 5–1).

The volume loadings are presented in Figure 5–6, which are categorized by loading

RH. Despite the standard variation is relatively large, the volume loadings increase as the loading RH increases.

60 Loading Temperature

) 2 0.4 9°C

c 16°C

/ 3 23°C

(

O 0.3

0.2

e 0.1

V 0 30% 45% 60% Loading Relative Humidity (%)

Figure 5–6. Volume loading of wet NH4NO3 at different loading temperatures and RH

In Figure 5–7, volume loadings are plotted as the function of loading AH. Similar to (NH4)2SO4 results, volume loadings strongly depend on the loading RH rather than the loading AH. For example, the AH of 30% and 45% RH at 23℃ are about two times of those at 9℃ (9.26 g/m3 to 3.97 g/m3 and 6.17 g/m3 to 2.65 g/m3), whereas the volume loadings are comparable. When the loading AH are the same, volume loadings depend on the loading RHs.

Figure 5–7. Volume loading of wet NH4NO3 at different loading AHs

The two-way ANOVA test is also used to analyze the NH4NO3 results. In Table 5–

3, the Pr(>F) of the Temperature is 0.9403, which is far more than 0.05, indicating that the temperature effect on the volume loadings is insignificant. In contrast, the Pr(>F) of the

RH is 0.0002172, which is less than 0.05. Therefore, the RH effect on the volume loadings is significant. Meanwhile, the Pr(>F) of interactions between the Temperature and RH is about 0.6652, showing a small influence on the volume loadings.

Table 5–3. Analysis of Variance Table (Type II test) of wet NH4NO3 Sum Sq Df F value Pr(>F) Temperature 0.000419 2 0.0616 0.9403 RH 0.07061 2 10.38 0.0002172 Temp:RH 0.005391 3 0.5284 0.6652 Residuals 0.1428 42

62 5.4 Discussion

The root cause of the filter volume loading variation at different RHs and temperatures is the capillary condensation of moisture below the DRH of salts. The dendrite restructuring occurs when water condenses between hygroscopic particles, which reduce the overall cake resistance and lead to a greater holding capacity for given terminal pressure. The details about the RH effect on the dendrite structure and filter holding capacity will be discussed in Chapter 6. In this study, the temperature dependence of both the DRH and capillary condensation should be considered.

5.4.1 DRH Temperature Dependence

For a given loading RH, if the DRH of salts is dependent on the temperature, the relative extent from loading RH to the DRH will also change. The closer the loading RH to the DRH, the more water could be condensed due to some larger capillary can satisfy the Kelvin equation. Therefore, the DRH temperature dependence can affect filter holding capacities.

The temperature dependence of salt DRH has been extensively studied by many scientists and researchers, including Wexler and Hasegawa (1954), Rockland (1960),

Greenspan (1977), Wexler and Seinfeld Wexler and Seinfeld (1991), Tang and Munkelwitz

(1993), Tang and Munkelwitz (1994a), Xu et al. (1998), Onasch et al. (1999), Gysel et al.

(2002), and Ebert et al. (2002). The experimental techniques employed include saturated solution saturation RH measurement, single-particle levitation observation, FTIR particle measurement, and modeling, and studied temperatures cover from -15℃ to 50℃.

63 According to studies mentioned, the temperature dependence of salt DRH is not strong in this temperature range. Although the salt DRH is lightly negative proportional to the temperature, some researchers claimed this could be neglected.

In Wexler and Hasegawa’s study (1954), they used saturated salt solutions at a controlled temperature to measure the steady-state saturation RH. The results covered three older literature results(Adams and Merz 1929, Edgar and Swan 1922, O'Brien 1948) and their measurements were from 0℃ to 50℃ with an increment of 5℃. Their results show that the DRH of (NH4)2SO4 at 0℃ is 83.7% and 50℃ is 79.1%. Rockland (1960) used a similar method and found that the DRH of (NH4)2SO4 at 5℃ is 81 % and that at 40℃ is

79%, which is consistent with the Wexler and Hasegawa’s study (1954). The single- particle levitation method was carried out by Tang and Munkelwitz (1993, 1994a), Xu et al. (1998). The experiment temperature covers from -9.4℃ to 50℃. Their shared conclusion is that the DRH of (NH4)2SO4 decreases as the temperature increases, while Xu et al. (1998) pointed out that the DRH of (NH4)2SO4 variation is only 2.2 percentage points from -9.4℃ to 25℃. Onasch et al. (1999) utilized FTIR to measure the DRH variation from -10℃ to 20℃, and their results agree with some of the studies mentioned above.

Although a slight decrease of the DRH of (NH4)2SO4 with temperature increase was observed, the authors stated: “both the deliquescence and efflorescence relative humidities of (NH4)2SO4 are not strongly dependent on the ambient temperature”(Onasch et al. 1999).

Gysel et al. (2002) utilized the HTDMA developed by Weingartner et al. (2002) to measure the RH and temperature dependence of the (NH4)2SO4 particles. Gysel et al. (2002) claimed that the temperature dependence could be neglected in the temperature range -10℃ < T <

25℃. Similar to (NH4)2SO4 results, Ebert et al. (2002) used environmental scanning

64 electron microscopy (ESEM) to determine the temperature dependence of NH4NO3 DRH in the temperature range 5℃ < T < 15℃; Tang and Munkelwitz (1993) measured the DRH of KCl in the temperature range 0℃ < T < 50℃. Both DRHs of NH4NO3 and KCl decrease with temperature increases, whereas the temperature dependence is not strong enough in the testing temperature range. In addition, NH4NO3 was tested in its wet state, the DRH is not observable, and the DRH doesn’t affect the amount of water absorbed by a wet NH4NO3 particle. This will be discussed in Section 6.2 in detail. From the thermodynamic property modeling perspective, Wexler and Seinfeld’s model (1991), Tang and Munkelwitz’s model

(1993), and Greenspan’s review (1977) agree with the results of the experiments.

In conclusion, the DRH temperature dependence is not strong and can be regarded as unchanged in the temperature range of 5 ℃ ~ 50 ℃. Therefore, the temperature change will not affect the DRH of salts. So that filter holding capacities will not be affected from the DRH perspective.

5.4.2 Capillary Condensation

The capillary condensation is governed by the Kelvin equation below, which correlates the equilibrium vapor pressure, saturation vapor pressure with the mean curvature of the meniscus, liquid/vapor surface tension, and temperature.

푃 2퐻훾푉 푙푛 푣 = − 푙 푃푠푎푡 푅푇

Where the ratio of equilibrium vapor pressure and saturation vapor pressure on the left-hand side of the equation is the RH. Therefore, the mean curvature of the meniscus H could be expressed as a function of the RH, temperature T, surface tension γ and molar 65 volume Vl within the temperature range (273.15 K ~ 323.15 K). Since the surface tension

γ and molar volume Vl are dependent on the temperature. The mean curvature of the meniscus H, which is an indication of how much water is condensed in the capillaries between particles, could be plotted as a function of the RH and temperature T as Figure 5–

8. It should be noted that the mean curvature of the meniscus H is reverse proportional to the capillary dimensions, which depend on the particle diameters and particle gaps. The smaller the H, the larger the capillary that water can condense. For a given capillary site at a certain temperature and RH, if the H is above the surface, there will be capillary condensations.

In the surface plot below, the mean curvature of the meniscus H strongly depends on the RH. However, the temperature effect on the mean curvature of the meniscus H is not strong. It is clear that the RH is the major factor affecting the mean curvature of the meniscus H, and the temperature in the plotted range is a minor factor. This would reflect on the macro filter loading behavior at different RH and temperature. Filter volume loading is mainly loading RH dependent and the loading temperature plays a weaker role in volume loading.

Figure 5–8. The plot of Mean Curvature of the Meniscus as a function of RH and temperature

5.4.3 Suggestions to Atmospheric Particulate Pollutants Control Technology

As an effective atmospheric particulate pollutants mitigation technology, air filters are widely used in various scenarios. For example, air filters could be installed on gas turbines, internal combustion engines, and any equipment that requires clean air. Air filters could also be installed on indoor air purifiers. The operating conditions are based on locations of the equipment. Therefore, it is important to understand the effect of operating conditions on the filter’s performance, particularly the filter life. Both the experiment and theory studies imply that the RH is a major factor affecting the filter loading while the temperature is only a minor factor when the hygroscopic particles are involved. In fact, not only the inorganic salt particles are hygroscopic, but some other pollutants in the

67 atmosphere are also hygroscopic, such as aged soot, hydrophilic organics. The RH could change the morphology or the state, which may affect the filter life significantly, of the atmospheric particulate particles based on their hygroscopicities. Therefore, the RH is a critical parameter affecting the filter life when it is used to remove atmospheric particulate pollutants, and it should be well controlled during the filter loading test, which is ignored nowadays. This study suggests the current filter testing standards may be revised by considering the air filter operating RH and temperature. In addition, the particle state may be distinct from the RH variation due to the hysteresis of the hygroscopic particles. Hence, the filter design should be carefully considered based on the application scenarios and the possible particle states, rather than assuming all particles are dry. This study also offers an insight into the link between the atmospheric particles hygroscopicity and the ambient particulate matter mitigation technologies.

5.5 Conclusion

In order to understand temperature and RH effect on filter loading to better predict the filter life, this study was performed by loading air filter samples with KCl, (NH4)2SO4, and NH4NO3 at different temperatures and RHs. A LAEAF Test Rig was employed to achieve a temperature and RH controlled filter sample loading process. The testing temperature covers from 5 ℃ to 50 ℃ and the testing RH covers from 30% to 75%. KCl and (NH4)2SO4 particles were tested in dry state, and NH4NO3 is in its wet state, which represent the different particle states in the atmosphere. From experiment results, RH plays a significant role in affecting the filter life. In contrast, temperature is a minor factor. By analyzing the possible reasons that affect the filter life, including DRH temperature

68 dependence, capillary condensation, they also support the experiment results. This study suggests that RH should be considered for the filter testing methods and standards, whereas the temperature could be controlled in a relatively wide range. For the atmospheric particulate pollutants mitigation applications, the operating location, i.e. filter operating environment, local pollution particle hygroscopicity, should be considered as a factor when choosing air filters to optimize the filter life.

69 Chapter 6: RH Effect on Filter Loading by Single Species

6.1 Conventional Cellulose Filter Media

This section has been published: Pei, C., Ou, Q., and Pui, D. Y. H. (2019). Effect of relative humidity on loading characteristics of cellulose filter media by submicrometer potassium chloride, ammonium sulfate, and ammonium nitrate particles Separation and Purification Technology 212: 75-83. doi:10.1016/j.seppur.2018.11.009

6.1.1 Introduction

Cellulose filter media is commonly used to make air intake filters because of their low cost and easiness of production. Air intake filters are often used at air inlet of internal combustion engines or gas turbines to clean ambient air prior to entering combustion chamber. In the use of air intake filters, the holding capacity of filters is an important measure of how much contaminant the filter could hold at a given pressure drop. For mechanical filter media, the pressure drop across the media gradually increases as the deposited particles accumulate. The holding capacity of a filter media can be characterized as the amount (usually mass) of particles deposited when the media pressure drop reaches a specified value. A greater holding capacity media, which is desired, means it holds more particles than media with less holding capacities at the same terminal pressure. Air intake filters require replacement at a certain terminal pressure to maintain the desired intake airflow and power output level; thus, intake filters with greater mass loading capacity, which have longer service time between replacements, are desired. Although it may or may not represent the real application well enough, this filter loading process is often evaluated

70 in the laboratory, and the filter holding capacity is often determined by challenging filters or filter media with laboratory-generated contaminants, typically at high concentration to accelerate the testing process. Besides test dusts (e.g., Arizona Road Dust), often used in various industrial filter test standards (e.g., ISO 5011, ISO 16890, etc.), which represent the coarse fraction of ambient contaminants, NaCl and KCl are the most commonly used model test particles in the laboratory to simulate the finer fraction. As mentioned in Section

1.1, these two salts are non-toxic, inexpensive, and have relatively high ERH (~47% for

NaCl and ~57% for KCl), which ease the drying process of salt droplets after being generated from an atomizer. In the past, mass loading capacity tests in the lab often ignored the RH effect due to the experiment complexity.

Gupta et al. (1993) studied the sub-micron NaCl and Al2O3 loading characteristics on HEPA filter. It was found the specific cake resistance of NaCl decreases with increasing

RH when it is below the deliquescence relative humidity of NaCl. A nonlinear increasing pressure drop and low mass loading capacity were found for RH above the DRH of NaCl.

Joubert et al. (2010) did a similar research on flat and pleated filters using NaCl and Al2O3, in which similar results were found and the hygroscopic particles loading above its DRH was characterized as liquid aerosol filtration. In addition to the experiment, a semi- empirical model was established to predict the pressure drop across a HEPA filter in the cake filtration regime when the loading RH is below the DRH. Miguel (2003) also found that the performance of the filter was affected by particle hygroscopicity and RH. These results suggest that the RH plays an important role on filter loading by hygroscopic salt particles.

71 Nevertheless, as mentioned in Section 1.1, neither NaCl nor KCl is a major composition of ambient PM2.5 that one of major contaminates for air intake filters. Previous studies on PM2.5 chemical speciation show that the most abundant inorganic ionic components in PM2.5 are ammonium, nitrate, and sulfate. In terms of the salt compounds,

NH4NO3 and (NH4)2SO4 are the most common forms (Cheng et al. 2016, He et al. 2001,

Philip et al. 2014, Snider et al. 2015, Snider et al. 2016, Sun et al. 2004, Yang et al. 2011,

Yao et al. 2002, Zhang et al. 2007). Thus, (NH4)2SO4 and NH4NO3 could be better representative particles to mimic the hygroscopic behavior of ambient aerosols encountered in air intake filtration applications. (NH4)2SO4 and NH4NO3 are much more hygroscopic than NaCl and KCl, which means that they are more active with water vapor at higher RH. Also, (NH4)2SO4 and NH4NO3 represent realistic air pollutants.

6.1.2 Experimental Method and Material

This study was conducted with early version of the LAEAF Test Rig described in

Chapter 1, which didn’t have the temperature control function. Therefore, experiments were carried-out at room temperature. The testing RH control was the same as the LAEAF

Test Rig. The hygroscopic salt properties and the dry/wet state control technique have been explained thoroughly in Section 1.4. In this study, dry KCl, (NH4)2SO4 and wet NH4NO3 were used as challenging particles. Cellulose air intake filter media were loaded with sub- micron KCl, (NH4)2SO4, or NH4NO3 particles at different RHs below their DRHs. This study simulated the ambient salt particle loading on the cellulose filter media by using more realistic ambient salt particles rather than laboratory model salts. The effect of RH on the

72 salt particle loading behaviors was also considered and studied. The terminal pressure was

10 inches of water (≈2500Pa) when the filter face velocity is 5.3 cm/s.

The filter media tested in this study is designed for filters that removing ambient air pollution particles. The test filter media, which is made of cellulose, has an efficiency rating of F8 according to EN779. An example of the SEM image of the test filter media is shown in Figure 6–1. The amount of salts loaded on the filter media at different testing RH were compared. The test filter media sample was prepared according to Section 1.6. The mass gain is the dry salt mass deposited on the test filter media. Since the densities of three salts used in the study are different, again, we used the volume loading to represent the amount of the salt deposited on filters. The volume loading could be calculated by dividing mass loadings with corresponding salt densities.

Figure 6–1. SEM image of a test filter media sample. It is a conventional cellulose filter media with an efficiency rating of F8 according to EN779

The test matrix was determined according to the hygroscopic properties (ERH and

DRH) of three salts. For KCl and (NH4)2SO4, test filters were loaded at RH levels of 30%,

45%, 60%, and 75%. As mentioned above, all salt particles were assumed to release all water content before being conditioned to desired loading RH. While for NH4NO3, RH 73 levels of 15%, 30%, and 45% were selected. Considering ERH of NH4NO3 doesn’t exist,

NH4NO3 particles employed in this study were in wet state, which has been verified by Li et al. (2016b). For each test condition, at least two repeats were conducted. Some cases were repeated for three times.

Previous studies suggests filter holding capacity is very sensitive to the size of particles being collected (Cheng and Tsai 1998). Therefore, it is very important to keep a consistent particle size distribution among different test runs, especially when different salts were tested, in order to make the mass loadings comparisons meaningful among tests runs. 1% volume concentration solutions of three salts were prepared to generate salt particles with similar size distributions. In addition, the atomizer was operated at a constant pressure (40 psi  275.79 kPa) and other flow rates in the test system were kept unchanged to maintain repeatable particle size distributions and concentration. Figure 6–2 is the number-based and mass-based size distribution of three salts. The mean value and error bar in Figure 6–2 represent the mean and standard deviation of distributions from all valid runs.

6 5×10 KCl Number (NH4)2SO4 Number NH4NO3 Number 4

6 3×10 d )

4×10 KCl Mass W 3

/ m

(NH4)2SO4 Mass d c

l / NH4NO3 Mass o

# 6

( 3×10

4 p

2×10 p

( g

6 µ o

g l 2×10

/ d

m /

3 N 4

)

d 10 1×106

0 0 10 100 1000 Mobility Diameter (nm) Figure 6–2. Salt particle size distribution

74 6.1.3 Results and Discussion 6.1.3.1 RH Effect of Deposited Hygroscopic Particles

Filter loading at elevated RHs has been studied by many researchers as mentioned in the introduction. Some assumptions have been given to explain the reduction of filter flow resistance at elevated RHs. Gupta et al. (1993) proposed that inter-particle adhesive force is higher at elevated RHs, and particles would stick to deposited particles rather than fill the interstitial spaces during filter loading. This could construct more open cake structure, yielding lower cake resistance. Miguel (2003) advised that the stronger adhesion between particle at higher RH cause higher compact arrangement of deposited particles.

Joubert et al. (2010) emphasized the bonding forces between particles caused by liquid water could generate a cake with higher porosity. Butt and Kappl (2009) mentioned that the water vapor could condense into capillaries or pores of deposited particles when the temperature is far above the dew point of surrounding environment. Joubert et al. (2011) explained the pressure drop change of a hygroscopic salt loaded filter at elevated RHs as the water vapor in the air was absorbed by deposited particles causing a restructuring of the deposit. Montgomery et al. (2015a, 2015b) observed both aggregate and dendrite morphology change at elevated RHs, with the explanation of absorbing moisture on deposited hygroscopic particles impose the restructuring.

In this study, we propose that when the KCl and (NH4)2SO4 particles are loaded on the test filter media, the gaps formed among adjacent deposited particles are the ideal places for water vapor to condense at even when RH is less than their DRH (Thomson 1871). The water condensed in the gaps among particles could lead to necking which reduces the total surface area of the dendrite structure. With more water absorbed, particles could coalesce,

75 the dendrite structure could shrink and open the air passageway therefore further reduce the resistance to the air (Figure 6–3). Montgomery et al. (2015b) depicts the process of salt dendrites restructuring with capillary condensed water. Besides salt particles, Weingartner et al. (1997) studies soot morphology change with RH and finds capillary condensed water could cause soot aggregates collapse. According to their study, the restructuring process could happen as low as 30% RH, which is far below the DRH of corresponding salt. The filter resistance could be reduced if the total surface area of the loaded particles decreases and the air passageways are more open at elevated RH.

It should be noted that the morphology change of the salt dendrites on the filter fiber at elevated RH is independent of filter media. The similar phenomena have been observed on different filter media (Gupta et al. 1993, Joubert et al. 2011, Miguel 2003,

Montgomery et al. 2015b). However, this kind of morphology change may help increase holding capacity of filters, like cellulose filter in this study; it may have the negative impact on the holding capacity in certain circumstance, which will be discussed in Section 6.2.

Figure 6–3. Dendrites restructuring with capillary condensed water

76 6.1.3.2 Loading Curves

Loading curve represents the pressure drop across the test filter media as the particles loading on the test filter media. In this study, as aforementioned, the pressure drops as a function of the volume loading of three salts at different RHs will be presented and compared.

Figure 6–4 shows the selected loading curves of the test filter media as a function of the volume loading under different RHs for all three salts. The volume loadings were compared because the densities of three salts are different. Densities of KCl, (NH4)2SO4,

3 and NH4NO3 are 1.99, 1.77, and 1.72 g/cm , respectively (Haynes 2014, Lide 2005). The test filter media has a relatively low initial efficiency which gradually improves over time as loading proceeds, so that the pressure drop dependent on the loading time (which is raw data recorded by the LabVIEW program) cannot be directly converted as the function of the volume loading. Instead, these raw data were converted by considering the filter efficiency evolution during the loading. At initial stage of the loading, the actual mass loaded on the filter was factored by the efficiency at that time, until the efficiency reached nearly 100% at later stage of loading. The final volume loading at the end of each test was determined from the dry mass gain from sample weighing, with known density of each salt.

In Figure 6–4, the loading curves of KCl, (NH4)2SO4 at all RHs and NH4NO3 at 15%

RH have similar shapes, which share a typical pattern for cellulose media loaded with dry submicron particles. It starts from depth filtration at the beginning of the loading, during which particles were captured inside the filter and the slope of the pressure drop increasing was high. While it proceeds into the transition regime at later loading stage, the increase of pressure drop slowed down, asymptotic to pure surface filtration when the slope of pressure 77 drop remains constant. In this study, all loadings stopped at transitional stage (not yet reaching pure surface loading) when reaching 10 inches of water (2500Pa) terminal pressure. As presented in Figure 6–4, rates of the pressure drop increasing were different at each RH. Especially at the later filtration stage, when the particles accumulated near the surface of the filter, the pressure drop increased faster when the loading RH were lower.

The differences in pressure drop increasing rates resulted in the variations in the final volume loading on the filter at the same terminal pressure.

The NH4NO3 loading condition was different from KCl or (NH4)2SO4, considering its state is different from the other two, which can be the reason for different shapes of the loading curves in Figure 6–4. Interestingly, the NH4NO3 loading curve at 15% RH shared the same shape of the loading curve as the other two salts. Although NH4NO3 particles absorb moisture they remain in wet state and do not lose water under test environment in this study. The water carried by NH4NO3 droplet is limited at 15% RH, thus NH4NO3 droplets have high viscosity and therefore behave similarly to solid particle loading.

Whereas 15% RH NH4NO3 loading is slower compared to the other two salts. As for higher loading RHs, the pressure drop increasing much slower at 30% and 45% loading RH than

15% loading RH case at the beginning of loading. Even though the loaded droplets could coalesce to reduce the overall resistance, with more and more droplets loaded into the test filter, air passageways in the test filter are less and smaller. Eventually, air passageways are reduced both in numbers and size, so that the pressure drop increasing dramatically.

Similar trend was observed by Gupta et al. (1993) and Joubert et al. (2010), it was classified as liquid aerosol filtration. According to Contal et al. (2004), droplets fill the interstitial spaces at initial stage and the abrupt increasing of the pressure drop is caused by a liquid

78 film forming on top of the filter media with more droplets accumulated. In this study, liquid film may not be able to form on top of the test filter media as a cake layer but inside the test filter media instead, which will be discussed later in Section 6.1.3.5. It should be noticed that the pressure increase of the 30% and 45% loading RHs are abrupt at the end of the loading and if the same media was tested until a higher terminal pressure than 10 inches of water(2500Pa, although rare in practice), their volume loading could become close to or even lower than those loaded with KCl or (NH4)2SO4. Because the pressure curves of KCl and (NH4)2SO4 is a relatively steady increasing process, and more salt can be deposited if a higher terminal pressure is preferred. Whereas for NH4NO3, the pressure curve is nearly vertical, a higher terminal pressure will not allow too much salt to be deposited.

79 10 RH ) 30%

O 2 8 45%

i 60%

( 6 p 75%

e 4

r 2 P KCl 0 10 RH ) 30%

O 2 8 45%

i 60%

( 6 p 75%

e 4

r 2

P (NH4)2SO4 0 10 RH ) 15%

O 2 8 30%

i 45%

( 6

e 4

r 2

P NH4NO3 0 0 0.5 1.0 3 2 Volume Loading (mm /cm ) Figure 6–4. Selected loading curves under different RHs.

Note: The NH4NO3 at 30% and 45% RH loadings have the different shapes of loading curves than all other cases, they behave close to liquid aerosol filtration; all other cases share the similar shape of loading curves from the depth filtration to the transition to the surface filtration.

80 6.1.3.3 Efficiency Curves

As aforementioned, efficiency curves are not only important to factorize the loading curves, they are also beneficial in determining the loading process and state. Efficiency of the test filter media could be calculated by alternating particle size distribution measurements of upstream and downstream.

Figure 6–5 depicts the efficiency curves as a function of volume loading for different salts under various loading RHs. Similar to loading curves, except for NH4NO3 at 30% and 45% RH, all other cases reach 100% efficiency before reaching terminal pressure. However, for 75% RH of KCl and (NH4)2SO4, the volumes needed to reach 100% efficiency are greater than the other low RH cases. Because the efficiency increasing is mainly due to the growth of dendrite sticking out from fiber surface, high RH negatively affects the dendrite morphology and the speed of its growing. When curves in Figure 6–4 and Figure 6–5 are matched together, it could be found that at lower RH, both pressure drop increasing and the efficiency increasing are faster, which is a mutual effect. For

NH4NO3 at 30% and 45% RH, the growth of the dendrite is even slower because of their hygroscopic properties discussed in previous sections, so that the efficiency could not reach

100% even at the end of the loading. It was interesting that for 45% loading RH case, the efficiency was fluctuating over loading, suggesting an unstable state of the loaded particles on the test filter media.

81 1.0 RH 30% 0.9 45% 60%

y c 0.8 75%

f 0.7

E 0.6

0.5 KCl 1.0 RH 30% 0.9 45% 60%

c 0.8 75%

f 0.7

E 0.6

0.5 (NH4)2SO4 1.0

0.8

e i RH

f f 15% E 0.6 30% 45%

0.4 NH4NO3 0 0.5 1.0 3 2 Volume Loading (mm /cm ) Figure 6–5. Efficiency curve as a function of volume loading.

Note: All efficiency curves except NH4NO3 at 30% and 45% RH loadings achieve 100% efficiency during the loading process; NH4NO3 at 30% and 45% RH loadings’ efficiencies increase with fluctuation indicating an unstable state of the loaded particles on the filter media.

82 6.1.3.4 Volume Loadings

Figure 6–6 is the volume loadings of the KCl, (NH4)2SO4, and NH4NO3 at different loading RHs when the pressure drops of 10 inches of water (2500Pa) were reached. As shown in Figure 6–6, in general, the higher the loading RH, the greater the volume loading for same pressure drop. KCl and (NH4)2SO4 have the similar volume loading for all loading cases except 75% RH, at which RH (NH4)2SO4 has greater volume loading than KCl. These indicate that both salts particles clogged the filter with the similar mechanism given the comparable size distribution at the same loading RH. NH4NO3 has the much greater volume loading at each test RH than the other two salts due to it is in droplet state.

Besides the capillary condensation mentioned above, the size dependence of DRH plays a role in the volume loading deviation at 75% RH between KCl and (NH4)2SO4.

Reported DRHs of KCl and (NH4)2SO4 are about 84% and 80%, respectively. Thus, 75% is below their DRHs, and theoretically volume loading should be comparable as other lower RH conditions. The 75% RH volume loading of (NH4)2SO4 is about 60% more than that of the KCl. This indicates that the particle states of the KCl and (NH4)2SO4 are different compare to their states at lower RHs. According to Orr et al. (1958), the DRH is size dependent. For 100 nm diameter (NH4)2SO4 particle, the DRH is a little below 80%; while the DRH of 50 nm diameter particle and 20 nm diameter particle are about 75% and 63%, respectively. In contrast, the DRH of KCl particles for 106nm, 56nm, and 26nm diameter particles are about 84%, 82%, and 77%, respectively. KCl particles nearly don’t absorb water at 75% RH, while (NH4)2SO4 particles whose diameter were less than 50 nm absorb water when the loading RH was at 75%. Those particles would introduce liquid water on the dendrites and test filter media. Therefore, 75% is low enough for (NH4)2SO4 particles 83 to absorb more water than KCl particles, which affected the dendrite morphology and further volume loading results.

)

2 1.0

m KCl

3 (NH4)2SO4 m NH4NO3

m 0.8

(

i 0.6

@ 0.4

0.2

V 0 0 20 40 60 80 100 Relative Humidity (%)

Figure 6–6. Volume loadings of KCl, (NH4)2SO4, and NH4NO3 at different RH

Considering the water absorption behaviors of the particles at 75% RH, (NH4)2SO4 had more chance to absorb water from surrounding environment than KCl to change dendrite morphology. Thus, greater volume loading would be loaded when test filter media were stopped loading at same terminal pressure. While at the 30%, 45%, and 60% RHs, the dominant mechanism absorbing water both were capillary condensation. In this study, the particle size of the KCl and (NH4)2SO4 were comparable. Therefore, the amount of the water absorbed were close, which could lead to a similar volume loading.

In Figure 6–6, volume loadings of the NH4NO3 are much greater than the other two salts. As explained before, NH4NO3 droplets coalesce into larger particles much easier than the solid state during the loading process. Therefore, more volume of particles could be

84 hold at the same terminal pressure. The volume loading results verified the assumption.

The NH4NO3 volume loadings were much greater than the other two salts’ volume loadings at same loading RH.

6.1.3.5 Microscopic Observation

Figure 6–7 and Figure 6–8 are the SEM images of loaded test filter media. Figure

6–7 shows the images at the 200X magnification. Figure 6–8 intends to compare the details at higher magnifications, with KCl and (NH4)2SO4 samples at 10000X and NH4NO3 sample at 5000X due to significantly larger particle size on the sample. It should be noticed that all SEM samples were prepared after the samples were dried in the 0% condition, and the water content on the sample should be removed. During the sample preparation, transportation, and coating process, RH was controlled to less than the DRH of the salts.

Since the SEM chamber was vacuumed, during the observation, samples were all water- free as well.

In Figure 6–7, the top surface of test filter media was not covered by salt particles which indicates that filter loading did not reach surface filtration. It agrees with the filter type and the loading curves discussed above. For KCl and (NH4)2SO4, images of 30% and

75% loading RH samples were taken, representing the lowest and the highest loading RHs in this study. Both 30% and 75% RHs were below two salts’ DRHs, thus the overall appearances of the loaded particles are similar. For NH4NO3, images of 15% and 30% loading RHs samples were taken. The one at 30% could be compared with the other two salts at same RH of 30%. The 30% loading RH NH4NO3 particles appeared less on the surface of the filter, even with greater volume loading compared to the one at 15% RH. 85 However, the appearance of 15% one is comparable to the other two salts. This is consistent with Figure 6–4, that NH4NO3 loading curves are significantly different between 15% and

30% loading RHs. As mentioned in Section 6.1.3.2, NH4NO3 at 30% RH does not appear covered by a liquid film. Gupta et al. (1993) and Joubert et al. (2010) used HEPA filter, which is surface loading dominant. This study used cellulose filter, where particle accumulation starts from inside of the filter. Droplets would fill interior of the filter and clog the test filter media from inside. Hence the liquid film is not observed here.

30% KCl 30% (NH4)2SO4 15% NH4NO3

75% KCl 75% (NH4)2SO4 30% NH4NO3

Figure 6–7. SEM images of the KCl and (NH4)2SO4 under 30% and 75% RHs, NH4NO3 under 15% and 30% RHs. Note: The test filter media were not fully covered by salt particles indicating the surface filtration did not reach, which agrees with the filter type and the loading curves. Note that the NH4NO3 under 30% RH appearance is different from the other two salts at 30% RH, whereas the NH4NO3 under 15% RH is relatively close to that of the other two salts at 30% RH.

Figure 6–8 is a closer look of the same samples shown in Figure 6–7. It could be found that particles in the nanometer range are missing. There are two reasons explaining 86 the missing of small particles. The first one is the particle size measurement limitation.

Only particle size less than 1m could be measured by the SMPS system. In fact, the particles larger than 1m exist in the loading particles. The second reason is that very small particles could coalesce with large particles, which verifies the assumption in this study.

For KCl at 30% and 75% loading RHs, the edges of the particles did not exhibit much differences. However, the particle size increase is very significant from 30% to 75%.

As for (NH4)2SO4, as described above, the size dependence of DRH promoted more water to be absorbed by particles. Necking happened on the boundaries between loaded particles.

Necks between particles were larger than that of KCl. The comparison in this micro level image between KCl and (NH4)2SO4 particles at the same RH could explain that the volume loadings at 30% loading RH were close; while at 75% loading RH, different degrees of the necking caused the different volume loading results.

It should be noticed that NH4NO3 particles in Figure 6–8 has a different magnification level comparing to the other two salts. Although it is at lower magnification, the particle size still appears larger than the KCl and (NH4)2SO4 particles. At the same time, given the similar size distribution, smaller particles are deficient. It is because that NH4NO3 droplets were loaded on the test filter media during loading, and smaller droplets coalesced with larger droplets. The fact that overall surface area of the loaded NH4NO3 droplets was less than the other two salts could be verified by the microstructure of dendrite. The reduced overall surface area helped holding more volume loading. The sublimation of ammonium nitrate at low RH was another concern. As mentioned by Lightstone et al. (2000), the sublimation rate of the ammonium at low RH is faster than at high RH. However, Lee and

Hsu (2000) observed insignificant NH4NO3 mass change at around 15% RH for 10 hours. 87 In this study, samples were stored in the 0% RH environment continuously before being prepared for the SEM observation. The cracks on particles might be the marks of the quick evaporation of water content, due to the extreme RH change between loading and storage

(0% RH).

30% KCl 30% (NH4)2SO4 15% NH4NO3

75% KCl 75% (NH4)2SO4 30% NH4NO3

Figure 6–8. Higher magnification (10000X) of the SEM images of the KCl and (NH4)2SO4 under 30% and 75% RHs, lower magnification (5000X) of the NH4NO3 under 15% and 30% RHs. Note: In general, the higher the loading RH, the more necking happened on the loaded salt particles. Due to the size effect on the DRH, (NH4)2SO4 necking larger than the KCl by absorbing more water at 75% RH. The absence of the small NH4NO3 particles indirectly indicates that the small particle may be admitted by the larger particles due to the existence of the water absorbed by the salt particles.

6.1.4 Conclusions

In this study, we tested loading characteristics of conventional cellulose filter media that is used for intake air filtration by KCl, (NH4)2SO4, and NH4NO3 sub-micron particles under the RH that are below their DRH. Three salts with comparable size distributions were loaded separately onto the test filter media until the pressure drop of the test filter

88 media reaches 10 inches of water (2500Pa). During the loading, pressure drop was monitored and recorded; The mass gain on the test filter media was weighed. The loading curves of KCl, (NH4)2SO4 at all test RHs and that of NH4NO3 at 15% loading RH presented dry particle loading curves as expected, while the loading curves of NH4NO3 at 30% and

45% loading RHs exhibited droplet loading curves due to incomplete drying of NH4NO3 droplets. The efficiency curves of KCl, (NH4)2SO4 at all test RHs and that of NH4NO3 at

15% loading RH reached 100% before the end of the loadings, while efficiency curves of

NH4NO3 at 30% and 45% loading RHs never reached 100% even at the end of loadings.

For all three salts, the higher the loading RH, the greater the volume loading, when loading

RH were below their DRH. However, the volume loadings of NH4NO3 were much greater than the other two salts given the same loading RH. The volume loadings of KCl and

(NH4)2SO4 were comparable, and those of (NH4)2SO4 were slightly greater than KCl with the increasing loading RH. This study implies that the service life of filters designed for ambient particles will be longer when the RH during service was higher especially when high hygroscopic content exists. It also implies that the service life of an ambient filter will be longer when the percentage of NH4NO3 is dominant. This study proposed a new method to prolong the filter service life by increasing the RH when the hygroscopic particles are being filtered. However, this study only tested particle loading when the loading RH is below the DRH of the salts. Whereas the loading characteristics of the cellulose filter media when the salt particle in droplet state while below their DRHs or the loading RH above their DRH still unclear and more tests should be conducted to reveal the loading behaviors under those conditions.

89 6.2 Nano-Fiber Coated Cellulose Filter Media

This section has been published: Pei, C., Ou, Q., Yu, T. and Pui, D. Y. H. (2019). Loading characteristics of nanofiber coated air intake filter media by potassium chloride, ammonium sulfate, and ammonium nitrate fine particles and the comparison with conventional cellulose filter media Separation and Purification Technology 228: 115734. doi:10.1016/j.seppur.2019.115734

6.2.1 Introduction

Air intake filter media are commonly made of cellulose, because cellulose is cheap and easy to process. However, there are some disadvantages of cellulose filters, such as the low initial efficiency, relatively high pressure drops across the filter media, and concerns on the filter life. To solve these problems, filter media manufacturers found that a new media structure, which is the composite of a synthetic nanofiber layer and conventional cellulose filter media, can achieve high initial efficiency, low pressure drop, and longer filter service life. A longer filter service life indicates that the holding capacity, typically in mass, of the filter is greater, at the same terminal pressure across the filter. The holding capacity is an important measure of the filter media service life. The holding capacity can be measured in the lab by loading the sample filter media with lab-generated high concentration particles. It is an important research topic how to maintaining the low pressure drop across the filter while holding more mass, which could reduce costs of the filter itself and maintenance costs of the air cleaning equipment.

As aforementioned, some filter test standards, such as ASHRAE 52.2, ISO 5011,

ISO 16890, and EN 779, emphasize on the efficiency measurements after filter loading.

Synthetic loading dust, which consists of fine dust, powdered carbon, and cotton linter, is used as the filter loading particles. Relative humidity (RH) is not strictly controlled in those

90 loading process because there is limited moisture involved in the process. Also, synthetic loading dust is not hygroscopic, and the particle size is in the micron level so that the capillaries between particles are difficult for water vapor to condense. However, the synthetic loading dust does not include salt particles, which is an important component of ambient pollutants.

In laboratory studies, NaCl and KCl are used as the testing particles for the filter loading research, because they are cheap, widely available, and non-toxic. The most common method of generating salt particles is by atomizing the salt solution and dry the droplets to yield the dry salt particle. According to the ASHRAE standards 52.2:2017, to dry droplets of its solution, the RH of the particle-laden air must be controlled less than 55% for NaCl, and 70% for KCl, which is feasible to achieve in the laboratory by diffusion drying and mixing with dry air (ASHRAE 2017). Once the particles are dried, NaCl remains dry state when the loading RH is less than its DRH of 75~76% (Hu et al. 2010,

Langlet et al. 2011), and KCl could achieve that when the loading RH is below its DRH

83~86% (Ahn et al. 2010, Cohen et al. 1987a, c, Freney et al. 2009, Li et al. 2014). RH below those DRH is achievable in the laboratory without additional delicate RH monitoring and controlling design.

Gupta et al., Joubert et al., and Miguel studied NaCl loading characteristics of the

HEPA filter, at different loading RHs. They all found that RH plays an important role in filter loading by hygroscopic particles (Gupta et al. 1993, Joubert et al. 2010, Joubert et al.

2011, Miguel 2003). Montgomery et al. also studied hygroscopic deposit structure change with elevated RH and confirmed with fluorescent microscopy (Montgomery et al. 2015a,

Montgomery et al. 2015b). Besides, Ribeyre et al. studied the effect of RH on the non-

91 hygroscopic particles nanostructure deposit on a filter, which showed nanostructure might be altered when the deposit is non-hygroscopic (Ribeyre et al. 2017).Their researches are fundamental in the RH effect of filter loading.

However, air intake filters are not designed to remove laboratory generated NaCl or KCl particles but to remove particulate pollutants in ambient air to protect equipment, such as vehicles, gas turbines. Previous research found that the most abundant inorganic compositions of ambient air pollutants are ammonium, sulfate, and nitrate ions (Cheng et al. 2016, He et al. 2001, Philip et al. 2014, Snider et al. 2015, Snider et al. 2016, Sun et al.

2004, Yang et al. 2011, Yao et al. 2002, Zhang et al. 2007). Therefore, the possible salts air intake filters may meet are (NH4)2SO4 and NH4NO3. Whereas, (NH4)2SO4 and NH4NO3 are more hygroscopic than NaCl and KCl. It means different ambient air RHs may affect the state of (NH4)2SO4 and NH4NO3 and further affect the filter loading behavior of those particles; therefore, (NH4)2SO4 and NH4NO3 are more suitable in filter loading test than the NaCl and KCl to simulate ambient pollutants filter loading in laboratories.

Due to hygroscopicity differences, additional considerations should be taken into the RH monitoring and controlling. As introduced in Section 1.4, the ERH of (NH4)2SO4 is 36%~43% (Ahn et al. 2010, Cohen et al. 1987a, c, Cziczo et al. 1997, Dougle et al. 1998,

Laskina et al. 2015, Liu et al. 2008, Tang 1980, Tang and Munkelwitz 1994b), which is less than that of NaCl and KCl, 45%~48% and 56%~59%, respectively (Ahn et al. 2010,

Cohen et al. 1987a, c, Freney et al. 2009, Lee and Hsu 2000, Li et al. 2014, Liu et al. 2008,

Tang et al. 1997), and NH4NO3 doesn’t have an ERH (Dougle et al. 1998, Lightstone et al.

2000). Therefore, additional dryer or dry gas is required to generate solid state (NH4)2SO4 particles, and it is impossible to generate solid state NH4NO3 particles from its aqueous

92 solution. DRH of (NH4)2SO4 is 75%~82% RH (Ahn et al. 2010, Cziczo et al. 1997, Tang and Munkelwitz 1994b). If the loading RH is not carefully controlled, it may be over

75%~82%, which will change the state of (NH4)2SO4 particles, from dry to wet. Because

(NH4)2SO4 and NH4NO3 are more hygroscopic, which means they are more sensitive to loading RH than NaCl and KCl, loading RH is critical to them. In addition, the RH in the ambient varies seasonally, even diurnally. It is even possible to filter salt particles with absorbed water in some operation scenarios; therefore, if (NH4)2SO4 and NH4NO3 are selected to load air intake filter media, the loading RH and the state of salt particles must be carefully controlled. The control technique has been thoroughly introduced in Section

1.4.

In Section 6.1, KCl, (NH4)2SO4, and NH4NO3 sub-micron particles were used to load conventional cellulose air intake filter under different loading RH (Pei et al. 2019a).

The results show that when loading RH is below salts’ DRH, the higher the loading RH, the greater the holding capacity of the filter. In that study, we used dry KCl and (NH4)2SO4 particles, wet NH4NO3 particles due to its hygroscopicity. The holding capacity comparisons among salts show that the filter media holding capacities of KCl and

(NH4)2SO4 are comparable when the loading RH is much less than their DRH; When the loading RH is 75%, which is close to their DRH (KCl at 84%, (NH4)2SO4 at 80%), the holding capacity of (NH4)2SO4 is higher than that of KCl. The holding capacity of NH4NO3 is much greater than that of the other two salts. It is possible that the difference in the state of particles causes the holding capacity differences. However, only dry KCl and (NH4)2SO4 particles were used to load conventional cellulose filter media, but wet KCl and (NH4)2SO4 particles were not tested. It is not clear that if the volume loadings of the wet KCl and

93 (NH4)2SO4 will be greater than that of the dry KCl and (NH4)2SO4 while it is clear that the differences in the hygroscopicity do lead to different results in the holding capacities of the same filter media.

In this study, synthetic nanofiber coated cellulose filter media was loaded by KCl,

(NH4)2SO4, and NH4NO3 sub-micron particles at a certain constant loading RH. The loading RH cover from below their ERH, to above their DRH. When the loading RH was between the ERH and DRH, particles may have two possible states, dry (solid state) or wet

(possessing water). This study evaluated dry and wet particle filter loading separately.

6.2.2 Experimental Method and Material

Generally, filter loading tests in the laboratory simulate filter operation in real conditions. The filter loading test measures two aspects of the filter performance. The first one is the amount of the particles gained on the filter at a given pressure drop; the second one is the pressure drop curve as a function of the mass loaded on the filter. As mentioned above, this study chose three different salts and loaded test filter media at different RHs.

Those two aspects of filter loading performance would be measured and compared.

The test filter media used in this study is a composite of a synthetic nanofiber layer and conventional cellulose fiber filter media, which is the same at the media used in

Chapter 5. The SEM image of clean test filter media used in this study is shown in Figure

5–2. Similar to the previous chapters, the test filter samples were prepared and weighed according to Section 1.6 sample preparation to yield the dry mass gain under different testing conditions.

94 This study was conducted on the LAEAF Test Rig at room temperature. The RH control function was used to maintain a stable filter loading RH. Filters’ holding capacity at the same pressure drop would be compared in this study. The terminal pressure was 4 inches of water (≈1000Pa) when the operating face velocity is 5.3 cm/s.

The size distribution deviation affects the holding capacity given the same pressure drop (Cheng and Tsai 1998). Thus, it is important to keep the size distribution consistent for all loadings of all three salts. Firstly, the pressure drove the homemade pneumatic atomizer was controlled by a regulator and kept the same for all tests. Secondly, the volume concentration of the three salts solutions were all 1%. These two methods could assure the size distribution consistency. Considering the difference in the density of three salts

3 3 3 (KCl:1.99 g/cm ; (NH4)2SO4:1.77g/cm ; NH4NO3:1.72 g/cm (Haynes 2014, Lide 2005)), volume loading will be compared in this study, which is obtained by dividing mass loading with corresponding salt density.

In this study, challenging particles of the test filter media could be either dry or wet.

For dry particles, the particle size will not change when the loading RH is below the DRH of the salt. For wet particles, the particle size distribution shifts to a larger size compared to the dry particle size distribution due to the absorption of water. The extent of the size distribution shift depends on the loading RH. Therefore, it is difficult to ensure and compare the particle size distribution consistency. In this study, we installed a silica gel desiccant dryer column in the sheath flow of the DMA to keep the DMA sheath flow dry.

When wet particles enter the DMA, the low RH sheath flow will dry wet particles and revert them to their dry state (minimal water for NH4NO3). For all loading cases, we

95 maintain the dry particle size distribution consistency for convenience. Figure 6–9 is the averaged size distribution of all samples in this study for each salt.

6×106 KCl Number (NH4)2SO4 Number 4 6 3×10 5×10 NH4NO3 Number

) KCl Mass W

m 6 (NH4)2SO4 Mass d

c 4×10 l

/ NH4NO3 Mass o

# 4 g

(

D 2×10

6 p

D 3×10

(

6 m / 2×10 4

3 N 10

)

d 1×106

0 0 10 100 1000 Mobility Diameter (nm) Figure 6–9. Number-based and mass-based size distributions of three salts. This is the averaged size distribution of all samples in this study for each salt.

As aforementioned, the particle state is critical in filter loading test, the salt particle state was carefully controlled according to the method described in Section 1.4. In this study, dry and wet salt particle loadings were researched separately, two distinct techniques were performed to generate dry and wet salt particles before humidifying to the desired loading RH. Briefly, to generate dry salt particles, droplet generated from atomizer must be dried below the salt’s ERH before humidifying to the desired loading RH. Salt droplets were firstly dried by the diffusion dryer filled with silica gel desiccant, and dry clean air was supplied around the aerosol stream to dry particles and achieve the RH below its ERH.

Conversely, to generate wet salt particles, the environment RH must be above its ERH before humidifying to the desired loading RH. Silica gel desiccant in the diffusion dryer was removed, and dry clean air was reduced to maintain the RH above its ERH. Then, Salt particles were humidified to the desired loading RH directly.

96 Both dry and wet particles will be tested in this study. By referring to the ERH and

DRH of KCl, (NH4)2SO4, and NH4NO3 in Table 1–2, and salt states illustration in Figure

1–5, the test matrix of this study could be concluded as Table 6–1.

Table 6–1. Test matrix of this study State Salt Dry Wet KCl 30%, 45%, 60%, 75% 75%, 90%, 95% (NH4)2SO4 30%, 45%, 60%, 75% 45%, 60%, 75%, 85%, 90% NH4NO3 30%*, 45%*, 60%* 30%, 45%, 60% *NH4NO3 particles undergo the dry particle generation method (keep the RH as low as possible after particle generation and humidified to the desired loading RH).

To compare the volume loadings of the wet and dry salt particle, the same RH is selected. For example, the ERH of KCl is 56%~59%; the wet RH could start from 60% with an increment of 15%. Whereas due to the sensor accuracy, 60% RH was marginal, and only 75% RH condition was applied to compare volume loadings of dry and wet KCl salt particles under the same RH. 90% and 95% RH represent the loading RH above the

DRH and near the saturation. As for the (NH4)2SO4, it is compared at 45%, 60%, and 75%

RH between the dry and wet states. 85% and 90% RH represent the loading RH above the

DRH and near the saturation. NH4NO3 doesn’t have an ERH (Dougle et al. 1998,

Lightstone et al. 2000), and 30%, 45%, and 60% RH are tested. It is impossible to generate dry NH4NO3 particles from its aqueous solution due to the non-existence of its ERH.

However, to compare the effect of the RH history on the loading curves and volume loadings of NH4NO3, two methods were applied, that is, keep the RH as low as possible after particle generation and then humidified to the desired RH (dry particle generation method); or directly humidified to the desired loading RH (wet particle generation method).

97 The dry particle generation method cannot generate dry NH4NO3 particle, but it is listed in

Table 6–1 under the dry state.

6.2.3 Results and Discussion

Loading curves as a function of particle loading under various loading RHs and particle states are discussed here. These loading curves are useful in characterizing the loading process. Besides the loading curves, volume loadings of the different loading RHs and particle states would also be presented to compare their effects on volume loadings.

SEM images of selected samples would be shown at the end of this section. Before discussing the loading curves and the volume loadings, the particle state effect and RH effect on the filter deposition will be discussed first to understand the loading curves and the volume loadings better.

6.2.3.1 Particle State Effect on Filter Deposition

When dry KCl particles are loaded on the test filter media, the dendrites could grow in the direction of air flow that comes from, perpendicular to the filter media, to form a cake. The particles accumulate in a 3-dimensional space. It can be found in Figure 6–10 that dry particle loading curves increase gradually with greater volume loading. Whereas the wet particles could not build up as easy as dry particles due to their fluidity. Wet salt particles can hardly accumulate in the direction perpendicular to the filter media, which yields a low porosity and high resistance deposition layer on the top of the nanofiber layer.

In this scenario, particles accumulate nearly in a 2-dimensional surface. When the test filter

98 media is clean, the resistance growing is not significant. However, the available open area is finite on the test filter media. With the deposition of the wet particles of given concentration, the possible open area on the test filter media is reducing. The same concentration of the wet particles could only flow through the reducing open area, which accelerates the clogging of the air passageways. As in Figure 6–10, the slopes of the wet particles loading curves increase with the loading. Once all available open area is covered by the low porosity and high resistance layer, the pressure drop across the test filter media increases abruptly. Besides this, according to Ahn et al. (2010) when the surrounding RH is about 75%, the diameter of a wet KCl particle is about 1.7 times of its dry state. The increase in the particle size by absorbing water could also contribute to the resistance developing due to the total surface area increases.

Moreover, wet particles may fully block the space in between particles because they could coalesce with each other when loading RH is close to saturation, and particle growth is significant. Wet particles, or more accurately salt droplets, block possible air passageways on the test filter media as they are being loaded on it. Eventually, salt droplets form a continuous film on the top of the nanofiber layer of test filter media (see Figure 6–

16 NH4NO3 at 75% RH). Gupta et al. (1993) reported a similar phenomenon. When HEPA filter (similar surface loading mechanism as nanofiber coated cellulose filter media) was loaded at RH close to the DRH of NaCl (80% and 90% RH), NaCl droplet could percolate into the filter fibers and eventually form a continuous liquid film causing pressure drastically increasing. Liew and Conder (1985) proposed liquid aerosol deposited on the filter could either locate at the fiber intersections or form liquid bridges between fibers depending on the filter packing density. Both Walsh et al. (1996) and Contal et al. (2004)

99 found that there is a redistribution during the filter loading. Contal et al. (2004) found a formation of a liquid film on the top of the filter, causing an exponential increase in pressure drop.

6.2.3.2 RH Effect on Filter Deposition

For the constant loading RH, as researched by Gupta et al. (1993), Joubert et al.

(2010, 2011), and Miguel (2003), increasing of the loading RH could reduce the hygroscopic particle specific resistance, which is due to water vapor capillary condensation between particles. Montgomery et al. (2015a) found that when loading RH is below the

ERH of hygroscopic particles, an increase of the loading RH could result in deposit restructuring and pressure drop decrease; deposit restructuring is caused by water absorption. Montgomery et al. (2015b) verified that deposit restructuring during hygroscopic particles are in the dry state. In this study, loading RH is controlled as a constant. The higher the loading RH, the slower the pressure drop increases. A similar phenomenon has been explained by Pei et al. (2019a) with conventional cellulose filter media. Briefly, water could condense in the gaps formed between dry salt particles even the surrounding RH is below DRH of corresponding salt. The resistance of the deposited particles reducing at higher loading RH by changing of the dendrite morphology and expanding of the air passageways. Eventually, at the same terminal pressure, greater volume loading was yield at higher loading RH.

The wet particles loading curves presented distinct shape from the dry particles, which has been discussed in previous Section 6.2.3.1. Generally, for wet particle loading, the higher the loading RH, the greater the volume loading. Although wet particle size 100 increases with loading RH, which causes surface area of individual particle increasing, wet particles could coalesce with each other to reduce the overall surface area. The overall resistance from the deposited particle will decrease with the overall surface area reduction.

Therefore, volume loading is greater at higher loading RH at a given terminal pressure, although confined by the wet particle size growth effect. Besides this, Gupta et al. (1993) explained the loading RH effect on the surface tension of a droplet on the percolating of the test filter media. In this case, for a given wet particle, salt concentration in the droplet is higher at lower loading RH due to less water absorbed. Moreover, it is believed that the surface tension of a droplet is positively correlated with its concentration. It is more difficult for a droplet to percolate the test filter media when its surface tension is higher and easier to form the continuous liquid film. Subsequently, less volume loading at lower loading RH.

6.2.3.3 Loading Curves

Figure 6–10 shows the typical loading curves of the KCl, (NH4)2SO4 and NH4NO3 under different RHs and particle states. In each panel, loading curves of different states are labeled in different colors. Loading curves of the dry state KCl and (NH4)2SO4 are plotted in black, and that of the wet state KCl and (NH4)2SO4 are plotted in red. For NH4NO3, methods of generating dry and wet particles are labeled as method I and method II. Since the ERH of NH4NO3 does not exist, both methods should generate the same state of testing particles. Both methods were performed to verify the effect of the RH history on the

NH4NO3 loading curves and volume loadings. It should be noted that all loading curves

101 are originally recorded as a function of loading time, and they are all transferred as a function of volume loadings for convenient comparison.

It is obvious that different particle accumulation mechanism causes distinct shapes of dry and wet particles. For all dry KCl and (NH4)2SO4 particles loading, the more the particles deposited, the slower the pressure drop increase. All loading curves showed the transition from steep to gradual, which indicated particle deposition from the initial stage to cake formation on the test filter media. It enters the surface filtration stage when the particle accumulation reaches a certain level. The final deposition conditions could be verified by SEM images of loaded test filter media, which will be discussed in Section

6.2.3.6. During the entire loading process, it is clear that the slopes of the loading curves decrease with the loading RH increasing, as discussed in previous Section 6.2.3.2.

For wet KCl, (NH4)2SO4, and NH4NO3 particle loading, the abrupt pressure drop increasing could be found. As discussed in Section 6.2.3.1, the initial pressure drop change is not significant when the amount of the particles loaded is small. Whereas the pressure drop increases abruptly when all available open area is covered by the wet particle layer.

Moreover, under high RH loading conditions, the pressure drops did not show steady growths as the lower loading RH, such as KCl at 95% RH and (NH4)2SO4 at 90% RH. In contrast, the loading curve of KCl at 90% is relatively steadier than that of (NH4)2SO4 at

90%. The difference reflects the distinct hygroscopic properties, considering the DRH of

KCl is about 4% higher than that of the (NH4)2SO4. The reason for loading curves fluctuating will be discussed later with the aid of the SEM images of (NH4)2SO4 loaded test filter media at 90% RH.

102 As for NH4NO3, both method I and II generate the wet particles according to that

ERH of NH4NO3 does not exist. Loading curves of method I and II are comparable at the same loading RH, indicating both methods produced wet NH4NO3 particles and the RH history does not affect the filter loading. The method I and II converge to a single condition of the wet NH4NO3 particle. In Figure 6–10, all six selected loading curves of NH4NO3 exhibit a similar shape to those of wet KCl and (NH4)2SO4 particles and follow the trend that greater volume loading at higher loading RH.

103 4

)

H 3

i KCl

(

p Dry 30%

o r 2 Dry 45%

e Dry 60%

u Dry 75%

s s 1 Wet 75%

P Wet 90% Wet 95% 0 4

) (NH4)2SO4

O 2 Dry 30% H 3

i Dry 45%

(

p Dry 60%

o r 2 Dry 75%

e Wet 45%

u Wet 60%

s s 1 Wet 75%

P Wet 85% Wet 90% 0 4

)

H 3

(

p NH4NO3

o r 2 Method I 30%

e Method I 45%

u Method I 60%

s s 1 Method II 30%

P Method II 45% Method II 60% 0 0 0.5 1.0 1.5 3 2 Volume Loading (mm /cm )

Figure 6–10. KCl, (NH4)2SO4, and NH4NO3 sample loading curves under different RH and different particle states

Note: Black curves represent the dry KCl and (NH4)2SO4 particle loading, and red curves represent the wet KCl and (NH4)2SO4 particle loading; for NH4NO3, black curves represent the dry particle generation method (I), and red curves represent the wet particle generation method (II).

104 6.2.3.4 Volume Loading

Figure 6–11 shows volume loadings of test filter media when the pressure drop across the test filter media is 4 inches of water at different loading RHs and particle states.

Analog to the loading curves figure, volume loadings of dry particles are labeled in black, and that of wet particles are labeled in red. For each point, at least three repeated tests were done to receive reliable results. Error bars show how spread the volume loadings for a given condition. It is worth to point out that since both method I and II produced wet

NH4NO3 particles, the results of method I and II at the same loading RH are combined in the volume loading results.

6.2.3.4.1 Dry Particles Volume Loading

For dry KCl and (NH4)2SO4 particle loadings, the higher the loading RH, the greater the volume loading. The effect of the loading RH has been explained in Section 6.2.3.2.

From 30% RH to 75% RH, the volume loading increases for 2-3 times, which is very significant.

When it is compared salt-wise, volume loadings of (NH4)2SO4 are all greater than that of KCl at the same loading RH. In the previous study on conventional cellulose filter media (Section 6.1), volume loadings of (NH4)2SO4 were also greater than that of KCl but not as much as the present study. One possible explanation is that the structure of the filter media can affect the volume loading. When particles are deposited on conventional cellulose filter media at the early stage, particles will enter the filter media initially, known as depth filtration. Whereas for the nanofiber coated cellulose filter media, particles directly start with the surface filtration on top of the nanofiber layer. In the latter case,

105 particles are abundant on the top surface of the filter media to form salt particle cake, which contains more gaps between particles than that of dendrites in the depth filtration. When

(NH4)2SO4 and KCl form cake, (NH4)2SO4 could absorb more moisture and neck between each other due to its hygroscopicities than KCl. Therefore, the resistance of (NH4)2SO4 particles would reduce. Eventually, it could cause greater deposition than KCl. The cake filtration is more sensitive to the salt hygroscopicity than depth filtration due to more particle contacting could occur in particle cake, which increases the amount of the capillary condensed water. In addition to that, the depth filtration of cellulose could be affected by more factors, such as cellulose and binder hygroscopicity, pore size and fiber morphology, while surface filtration could be affected by fewer factors.

6.2.3.4.2 Wet Particles Volume Loading

As explained in Section 6.2.3.2, all wet particles follow the trend that volume loadings are increasing with the loading RH.

Volume loadings between dry and wet particles of KCl and (NH4)2SO4 at the same loading RH could be compared. For KCl at 75% RH, the average volume loading of dry particles is 0.8523 mm3/cm2, and that of wet particles is 0.2899 mm3/cm2. The average volume loading of dry particles is about three times of that of wet particles. For (NH4)2SO4, volume loadings between dry and wet particles also show large discrepancies. The average volume loading of dry particles is about 3 to 4 times of that of wet particles. As explained in Section 6.2.3.1, dry particles can accumulate in a 3-dimensional space, while wet particles only accumulate in a nearly 2-dimensional surface. The accumulation mechanism differences cause large discrepancies in volume loading. 106 ) 2 1.5

c Dry KCl

3 Dry (NH4)2SO4 m Wet KCl

m ( Wet (NH4)2SO4

O 1.02 Wet NH4NO3

n i 0.5

o 0 V 0 20 40 60 80 100 Relative Humidity (%)

Figure 6–11. Volume loadings of KCl, (NH4)2SO4, and NH4NO3 at different RH on nanofiber coated cellulose filter (the present study)

When the loading RH is higher than the salts’ DRH, the volume loading is greater than that of the lower loading RH. For example, volume loading of KCl at 90% RH is greater than its volume loading at 75% RH and volume loading of (NH4)2SO4 at 85% RH is greater than its volume loading of 75% RH loading as well. When wet particles are loaded, DRH is meaningless to wet particles. The increase of the volume loading is roughly linear as a function of the loading RH until the loading RH is above a certain level, for example, the volume loading of KCl at 95% RH, the volume loading of (NH4)2SO4 at 90%

RH and that of NH4NO3 at 75% RH are extremely great. In terms of wet KCl, volume loading at 95% RH is about 2 times of volume loading at 90% RH; In terms of wet

(NH4)2SO4, volume loading at 90% RH is about 3 times of volume loading at 85% RH; in

107 terms of wet NH4NO3, volume loading at 75% RH is about 2 times of volume loading at

60% RH.

The possible reason for the extremely high-volume loadings is nanofiber layer damage. As shown in Figure 6–12, there is a small hole and a large hole on the filter media.

The reason causing the nanofiber layer damage is that the filter media is not uniform, and there are some fewer resistance locations, where particles can flow through easier.

However, salt particles absorb water and become a droplet above its DRH. Unlike the solid particles, droplets could fully block the preferred locations on the top of the nanofiber layer.

Both the pressure difference and the gravitational force could cause damage to the nanofiber layer. Once the nanofiber layer is broken, the pressure drop will decrease. It is worth to point out that particles in Figure 6–12 are more likely the crystals of the (NH4)2SO4 rather than the originally generated particles, which proves the existence of the (NH4)2SO4 solutions on the test filter media. The nanofiber damage could expose the base cellulose filter media to the loading particles, locally alter the nanofiber coated cellulose filter media to conventional cellulose filter media. When loading particles are directly loaded on the base cellulose filter media, the filtration stage switches from surface filtration on the nanofiber layer to the depth filtration in the cellulose filter media, which is preferable for wet particles to achieve a greater volume loading (Pei et al. 2019a).

The damage to the nanofiber layer on the filter media occurs randomly, which causes an unstable loading process for a single loading test and increases the standard deviation of volume loadings sharply among test filter media samples. As shown in Figure

6–10, high RH loading curves are not as stable as that of lower loading RHs. The pressure drops are not monotonic increasing but fluctuating during the loading. In Figure 6–11, the

108 standard deviation of volume loading of KCl at 95% RH and that of (NH4)2SO4 at 90% RH are significantly greater than the standard deviations of other cases. Especially the test filter media sample used in this study is not large enough (=47mm) and the nanofiber coating conditions (stiffness, density, cellulose fiber distance, etc.) may vary among samples, the occurrence of nanofiber damage on each test filter media sample differs. On the macro scale, this difference will lead to variations in volume loadings.

Figure 6–12. SEM image of (NH4)2SO4 at 90% RH

6.2.3.5 Comparison with Conventional Cellulose Filter Media

It is worthwhile to compare the volume loading of dry and wet particles on different filter media. Section 6.1 used conventional cellulose filter media to perform filter loading with dry particles (similar size distribution). In that study, conventional cellulose filter was loaded to 10 inches of water (2500Pa), while the terminal pressure in the present study is

4 inches of water (1000Pa). To better compare the volume loadings with this study, Figure

6–13 shows the adopted volume loading at 4 inches of water on conventional cellulose 109 filter. Compared to conventional cellulose filter media, it is clear that the volume loading of dry particles on nanofiber coated cellulose filter media increased from around 0.1 mm3/cm2 to at least 0.3 mm3/cm2 and at most 1.1 mm3/cm2, which is a very significant increase. Nanofiber layer on conventional cellulose filter can considerably improve the filter holding capacity. However, it may reduce the holding capacity when the challenging particles are wet. For example, volume loading of wet NH4NO3 particle at 30% RH on conventional cellulose filter media in Figure 6–13 is about 0.5 mm3/cm2; the volume loading of wet NH4NO3 particle at same 30% RH on nanofiber coated cellulose filter media in Figure 6–11 is only 0.25 mm3/cm2, which is only a half of that of conventional cellulose filter media.

) 2 1.5

c Dry KCl

3 Dry (NH4)2SO4 m Wet NH4NO3

(

O 1.02

n i 0.5

o 0 V 0 20 40 60 80 100 Relative Humidity (%)

Figure 6–13. Volume loadings of KCl, (NH4)2SO4, and NH4NO3 at different RH on conventional cellulose filter (Adopted from Pei et al. (2019a))

110 Based on the comparison above, the selection of the filter media should be based on the particle state that is planned to remove. For hygroscopic salt, the state of the salt depends on environment RH and RH history. For a dry salt particle, if the surrounding RH was always below its DRH, the particle will stay in dry state, such as desert area or low precipitation locations. Alternatively, if the surrounding RH of that dry salt particle exceeded its DRH once and decreased below its ERH afterward, the salt particle reverts to dry state. In the above scenarios, because of the existence of the dry salt particles, nanofiber coated cellulose filter media should be used to improve the filter life and reduce the cost.

However, if the surrounding RH exceeded its DRH and maintained above its ERH, state of salt particles switches from dry to wet. When the overwhelming particles are wet, a conventional cellulose filter should be chosen to enhance the holding capacity of the filter.

6.2.3.6 Microscopic Observation

Figure 6–14 and Figure 6–16 are the SEM images of the loaded test filter media under different salts, loading RHs, and particle states. All images are in the same magnification, 1000X, to compare the particle size and state conveniently.

In Figure 6–14, dry KCl and (NH4)2SO4 loaded test filter media are shown, in which eight images correspond with the eight black points in Figure 6–11. Half of the images

(KCl 75% RH, (NH4)2SO4 45%, 60%, 75% RH) show that the filter media has been covered by salt particles. The remaining four images are taken on purpose to check the nanofiber and particle interaction. The cellulose fiber in the cellulose base layer is not evenly distributed so that air passageways are not distributed evenly. Airflow cannot flow through some locations on the top of the filter; therefore, the test filter media cannot be fully 111 covered by salt particles. For the four images below (KCl 30%, 45%, 60% RH, (NH4)2SO4

30% RH), salt particles cover other open air passageways on the test filter media.

30% 45% 60% 75%

KCl

(NH4)2SO4

Figure 6–14. SEM images of test filter media loaded with dry KCl and (NH4)2SO4 at different RH. Note: Image size is about 120 μm in width, 96 μm in height. Scale bars in images are 10 μm.

Figure 6–14 offers an overview of the loaded test filter media while it is not magnified enough to compare the morphology difference between loading RHs. Figure 6–

15 is an example of the magnified salt dendrites of (NH4)2SO4 at 30% and 75% RH. As mentioned in section 6.2.3.2, capillary condensation occurs between particles. It could be found that particle coalescence occurs at both RHs while it is significant at 75% RH. At

30% RH, salt particles connect to each other while it shows a clear boundary between particles. At 75% RH, salt particles coalesce with adjacent particles with an unidentified boundary. Such microscopic observation of the dendrites’ morphology change supports the macroscope observation of the volume loading increase with loading RH increase.

112 (NH4)2SO4 30% RH (NH4)2SO4 75% RH

Figure 6–15. SEM images of test filter media loaded with dry (NH4)2SO4 at 30% and 75% RH

Figure 6–16 is the SEM images of test filter media loaded with wet salt particles at different loading RHs, which correspond with the red points in Figure 6–11. Comparing to dry particle loading in Figure 6–14, deposited particles cover the filter media with a relatively thin layer. Volume loadings of wet salt particles are much less than that of the dry particles. As introduced in Section 6.2.3.1, wet particles can merely build upon a 2- dimensional surface. This is verified on their microscopic observation. KCl is the least hygroscopic salt among three salts in this study, and it could be found that wet particle at

75%, 90% RH images still look similar to the dry particle deposition. (NH4)2SO4 is the intermediate hygroscopic salt. 45%, 60%, and 75% RH conditions show an almost single layer of the wet particles deposited on the test filter media. Loaded test filter at 90% RH is also shown in Figure 6–12. It is clear that (NH4)2SO4 recrystallized on the test filter media, providing the size and the shape of the particles on the test filter media. NH4NO3 is the most hygroscopic salt, and all 4 SEM images show distinct appearances from the other two salts. NH4NO3 are all dried after loading while remaining in a much larger size and droplet

113 morphology. It is worth noting that there are two holes on the nanofiber layer in the image of NH4NO3 at 45% RH. The nanofiber layer damage most likely occurs when the challenging particles are wet. As explained aforementioned in section 6.2.3.4.2.

30% 45% 60% 75% 90%

KCl

(NH4)2SO4

NH4NO3

Figure 6–16. SEM images of test filter media loaded with wet KCl, (NH4)2SO4 and NH4NO3 at different RH. Note: Image size is about 120 μm in width, 96 μm in height. Scale bars in images are 10 μm.

6.2.4 Conclusions

In this study, the loading characterization of the nanofiber coated cellulose filter media was tested with KCl, (NH4)2SO4, and NH4NO3 separately at different loading RHs.

The loading RHs covered from below the ERH of salts to above the DRH of salts, which meant that challenging particle states covered both dry solid particles and wet droplet particles. Salt particles had similar dry particle size distribution. All test filter media were loaded to 4 inches of water (1000 Pa). The test filter media were weighed in a dry chamber

114 before and after the loading to eliminate the effect of the moisture absorbed in both test filter media and loaded salt.

For dry salt particles, the higher the loading RH, the slower the pressure drop increase and the greater the volume loading, which is consistent with conventional cellulose filter media results. However, conventional cellulose filter media starts with depth filtration followed by surface filtration, but nanofiber coated cellulose filter media starts with surface filtration directly. The differences in filtration mechanism cause a subtle loading curve shape deviation while do not affect the overall loading behavior. The surface filtration is more sensitive to salt hygroscopicity and loading RH than the depth filtration.

The volume loadings of nanofiber coated cellulose filter media are greater than that of conventional cellulose filter media for all RH conditions due to the coating of nanofiber.

The capillary condensed water between deposited particles contribute to the dendrite morphology change and possible air passageway opening, furthermore, reduce the overall resistance and improve the volume loading at elevated loading RH.

As for wet salt particles, it also follows the trend that the higher the loading RH, the slower the pressure drop increase and the greater the volume loading. The wet particles exhibit a distinct loading curve shape from that of dry particles. The pressure drop increased much slower than that of the dry particles initially, whereas the pressure increases abruptly at a certain time and reach the terminal pressure in a short time. When the loading

RH is close to the saturation, the pressure drop fluctuates during the loading process, which leads to an unstable loading curve. The loading curves’ shape of the wet particles loading on nanofiber coated cellulose filter media is comparable to that of the wet particles loading on conventional cellulose filter media. However, the volume loadings of the nanofiber

115 coated cellulose filter media reduced about 50% than that of conventional cellulose filter media. It is because that wet particles can only deposit in a nearly 2-dimensional surface on the nanofiber layer coating due to their wet state. The wet particle coalescence and the surface tension variation at different loading RH affect the volume loadings at different loading RHs.

This study mainly investigated the RH effect on the nanofiber coated cellulose filter media, which is mainly used in the application of the ambient air particulate pollutant removal. Filter manufactures claim that the nanofiber coating layer could effectively improve the service life of the filter. However, this study reveals that the improvement of the filter holding capacity depends on the operation RH and the state of the challenging particles. When the particles are dry, the higher the loading RH, the longer the service life.

The switch from conventional cellulose filter media to the nanofiber coated filter media could improve the filter service life for at least three times based on the media used in this study. However, when the particles are wet, the nanofiber layer reduces the volume loading to about 50% based on the media used in this study. In this scenario, volume loading still correlates with loading RH positively.

From the results of this study, we learned that the selection of the filter media should be cautious. Nanofiber coated cellulose filter media is not always an upgrade in terms of improving the filter life. When the operation RH is low enough to maintain most of the challenging particles in dry conditions, nanofiber coated cellulose filter will be a good option to switch to. Whereas conventional cellulose filter media should be chosen when the operating RH is relatively high, at which most of hygroscopic salt particles absorb water

116 and transform to wet state. For a specific filter, increase the operating RH to a moderate level could improve filter life when hygroscopic particles are being removed.

117 Chapter 7: RH Effect on Filter Loading by Multiple Species

7.1 (NH4)2SO4 and NH4NO3 Mixture

This section is in preparation: Pei, C., Ou, Q., Wang, K., Sun, C.C., and Pui, D. Y. H., Effect of different ammonium salt mixing ratios on the loading performances of two nanofiber coated cellulose filters

7.1.1 Introduction

As described in Chapter 1, many filed measurements have depicted that inorganic salt, (NH4)2SO4 and NH4NO3 accounts for at most 50% of the PM2.5 in some highly urbanized regions, and soot and organic compounds are also major pollutants (Cheng et al.

2016, He et al. 2001, Philip et al. 2014, Snider et al. 2015, Snider et al. 2016, Sun et al.

2004, Yang et al. 2011, Yao et al. 2002, Zhang et al. 2007). For example, sulfate and nitrate account for 25% and 30% mass of PM2.5 in Minneapolis, respectively (Figure 4–3). The hygroscopicity of pollutant particles depends on the inorganic particles components, even when they are mixture of organic and inorganic particles (Jing et al. 2016, Peckhaus et al.

2012). It was also found that internally mixed (NH4)2SO4 and NH4NO3 particles could be a vital surface provider in the severe haze episodes (Li et al. 2016a). Therefore, the RH effect of salt mixture particles on filter loading could provide insight of the ambient particle loading on filters.

The single species constant RH filter loading was studied in Chapter 6 on two types of filter media. It was found that both salt species and particle state can affect filter loading

118 performance at different loading RHs. However, the test conditions in Chapter 6 is not close to the ambient condition as it is impossible that the ambient pollutant consists of single salt species. To better understand how the hygroscopicity of salt mixture can affect filter loading, different mixing ratios of (NH4)2SO4 and NH4NO3 particles were tested at different loading RHs in this study. Both dry and wet particles were tested to simulate different ambient particle states.

Two-component inorganic salt mixture hygroscopicity studies depicted that stepwise phase change may occur during hydration and dehydration processes (Chang and

Lee 2002, Cohen et al. 1987b, Ge et al. 1996, 1998, Tang 1976, Tang and Munkelwitz

1993, Tang and Munkelwitz 1994a, Wexler and Seinfeld 1991). Similar to the single- component salt, two-component salt can deliquesce and effloresce at different RH with stepwise change. For a solid mixture particle under increasing environmental RH, the first step of water uptake occurs at its mutual deliquescence relative humidity (MDRH), at which both salts dissolved in the absorbed water with eutonic composition. With further

RH increasing, partially dissolved salt keeps absorbing water until reaching DRH of the mixture, which depends on the original mixture composition (Ge et al. 1998, Gupta et al.

2015, Wexler and Seinfeld 1991, Yang et al. 2006). The MDRH is normally below the

DRH of salt mixture. When the RH is lowered, one of the salts starts crystallization as the pure salt core in a salt solution droplet when environmental RH reaches its ERH. With further RH decreasing, eutonic composition salt solution effloresce around the pure salt core at its mutual efflorescence relative humidity (MERH) to form a heterogenous core- shell particle (Ge et al. 1996, Gupta et al. 2015, Wexler and Seinfeld 1991). In this study,

MDRH will be used as the reference RH, which is analog to the DRH of a single-

119 component salt. At or below MDRH, absorbed or capillary condensed water may alter the morphology of dendrites and further affect the filter loading performance.

Hygroscopicity of (NH4)2SO4 and NH4NO3 mixtures were studied extensively.

Dougle et al. (1998) and ten Brink and Veefkind (1995) studied hygroscopicities of both lab-generated (NH4)2SO4 and NH4NO3 and ambient particles of Netherlands. By comparing particles hydration and dehydration processes of particles generated in the lab and collected from atmosphere, they found that ambient (NH4)2SO4 and NH4NO3 mixtures are internally mixed with different mixing ratios. In this study, internal mixture of

(NH4)2SO4 and NH4NO3 with different mixing ratios will also be employed. Sun et al.

(2018) and Wu et al. (2019) conducted similar studies of (NH4)2SO4 and NH4NO3 mixture particles of different mixing ratios with individual particle hygroscopicity system and in- situ Raman microspectrometry, respectively. They found the MDRH of (NH4)2SO4 and

NH4NO3 mixture particles varies with composition, which is due to different initial crystal structures for various composition. The MDRH is in the range of 63% ~ 68% for

(NH4)2SO4 and NH4NO3 mixture.

7.1.2 Experimental Method and Material

This experiment was conducted at constant RH and at room temperature on the

LAEAF Test Rig introduced in Chapter 1. The RH control function was used to maintain a constant loading RH, while the temperature was left at room temperature. As described in the introduction, internally mixed (NH4)2SO4 and NH4NO3 particles were generated.

They were dissolved into a single solution and the solution was then atomized with the

120 homemade atomizer in the LAEAF Test Rig, so that internally mixed salt particles could be generated. Since both dry and wet mixture particles were planned to be loaded, two methods introduced in Section 6.2.2 were used. Briefly, to generate dry salt particles, atomizer generated droplets were first dry to below the ERH and then humidified to the loading RH; to generate wet particles, atomizer generated droplets were conditioned to the loading RH while maintaining RH above the ERH. According to Wu et al. (2019), ERH of the mixture particles is in the range of 20% ~ 40% depending on the composition. Sun et al. (2018) reported that ERH of the mixture particles is in the range of 38% ~ 45% when the molar mixing percentage of (NH4)2SO4 is more than 30%, and when NH4NO3 is dominant ( >70%) no crystallization was observed. The similar phenomenon was also reported by Dougle et al. (1998) when NH4NO3 is more than 68.3%. In this study, when dry particles were being generated, regenerated silica gel desiccant were used in every run to treat mixture particles as dry as possible.

From the PM2.5 speciation data of selected US cities in the database introduced in

Chapter 4, it was found that (NH4)2SO4 and NH4NO3 ratios vary from 81%:19% (Orlando) to 38%:62% (Los Angeles). To represent different scenarios, the following (NH4)2SO4 and

NH4NO3 mass ratios were used: 100%:0% (pure (NH4)2SO4), 80%:20%, 50%:50%,

20%:80%, 0%:100% (pure NH4NO3). Since the MDRH is around 60%, three testing RH were picked in this study, 30%, 45%, and 60%. The volume concentration was about 1% so that the salt mixture particle count median diameter is about 100 nm and the geometric standard deviation is about 1.8.

Another kind of nanofiber coated cellulose filter media were used in this study.

Comparing to the one used in Chapter 5 and Section 6.2, it is with thicker and denser

121 nanofiber layer (Figure 7–1). All filter samples were treated following the process in

Section 1.6. Briefly, all filter samples were conditioned in 0% RH environment before and after loading. The dry mass gain was yielded by comparing the filter dry mass before and after loading. Since densities of (NH4)2SO4 and NH4NO3 are close to each other (1.77 and

1.72 g/cm3, respectively (Haynes 2014, Lide 2005)), mass gain was used instead of volume loadings. All filter samples were loaded to 4 inches of water (≈1000 Pa). Multiple samples were tested at the same test conditions, and the averages and standard deviations were calculated to have the statistically meaningful results.

Figure 7–1. SEM image of a clean nanofiber coated cellulose filter with thicker and denser nanofibers.

122 7.1.3 Results and Discussion 7.1.3.1 Dry Mixture Particles Loading

Mass gains of dry mixture particles on filter samples are shown in Figure 7–2. It should be noticed that pure NH4NO3 particles could be considered as wet particles from the findings of Section 6.2. By checking the pure (NH4)2SO4 and pure NH4NO3 mass gain, it agrees with the conclusions drawn from Section 6.2 that mass loading of (NH4)2SO4 dry particles is greater than that of NH4NO3 wet particles. According to Ge et al. (1996) and

Wu et al. (2019), the eutonic composition of (NH4)2SO4 and NH4NO3 is about 5.6%:94.4%, which means all mixtures tested in this study is (NH4)2SO4-rich particles. Wu et al. (2019) reported ERH of mixtures with different compositions. The ERH is in the range of 25.8%

~ 19%, when m((NH4)2SO4):m(NH4NO3) = 20%:80%. This RH was reached during testing.

In Sun et al. (2018), the particles with this composition were reported behaving like pure

NH4NO3 particles, that is, particles remained wet even at 5% RH. The contradiction between two studies may be used to explain the mass gains of m((NH4)2SO4):m(NH4NO3)

= 20%:80%, which is close pure NH4NO3 in Figure 7–2 at 30% and 45%.

123 25 AS:AN=100%:0%

) AS:AN=80%:20% g AS:AN=50%:50%

( 20 AS:AN=20%:80%

2 AS:AN=0%:100%

i 15

n 10

s 5

0 30 45 60 Loading Relative Humidity (%)

Figure 7–2. Mass gains of dry (NH4)2SO4 and NH4NO3 mixture particles on filter samples at different loading RHs

For the other two compositions, more RH were tested when it is close to the MDRH.

Mass gains of m((NH4)2SO4):m(NH4NO3) = 80%:20% mixture particles were loaded at

55%, 58%, and 62% in addition to 30% and 45% loading RH. Figure 7–3 shows mass gains and sample loading curves of five loading RHs. From mass gain results, it could be found that the mass gain of 45% loading RH is about 50% greater than that of 30% loading RH.

Mixture particles were in dry state when loading RH were 30% and 45%, which can be found from the loading curves. When loading particles are dry, the increasing rate slows down after the initial stage (30% and 45% in Figure 7–3). Whereas, the loading curve of

55% RH is slightly different from that of 45% RH, and the mass loading of 55% RH is slightly less than that of 45% RH. By comparing the mass gains and loading curves of 58% and 62% RH, it is clear that 55% is a transition RH between dry and wet particles, though it is below the MDRH. The MDRH of this composition is about 68.7% RH, which was 124 measured from a single salt mixture measurement (Sun et al. 2018, Wu et al. 2019).

However, in the filter loading scenario, sub-micron size salt particles accumulated on the filter sample at elevated RH, which is close to the MDRH. As described in Section 6.1.3.1, capillary condensation could prompt water condensation on salt particles especially when the surrounding RH is very close to the MDRH.

25 4

)

( 20 )

O O

2 3

H H

n n

i 15 i

(

4 p

o r

@ 2

n 10

e r

u Dry 30%

G s

s Dry 45% e

s 1 5 r

P Dry 55%

M Dry 58% Dry 62% 0 0 30 45 55 58 62 0 5 10 15 20 Loading Relative Humidity (%) Mass Loading (mg) Figure 7–3. Mass gains and sample loading curves close to the MDRH of salt mixture when m((NH4)2SO4):m(NH4NO3)=80%:20%

SEM images (Figure 7–4) of filter samples at 30%, 55% and 62% demonstrate the deposition conditions on filter samples. In Figure 7–4, filter samples were artificially bended before preparing SEM samples, so that salt cake would break, and the cross-section can be seen. By comparing SEM images of 30% RH and 55% RH, salt particles coalesce more at 55% RH, and a similar phenomenon was observed in previous chapters. However, the particle deposition at 62% RH was distinct from the other two RHs, and a thin layer of wet mixture particles cover the filter media. Figure 7–4 is a good example of salt particle deposition arrangements at different loading RH.

125

Figure 7–4. SEM images and particle deposition illustrations at different loading RH

It should be recalled that in single species loading studies, particle state conversion below its DRH was not observed. Whereas it was found here, even with 20% NH4NO3 in mass. Ge et al. (1998) observed water absorption below MDRH for various (NH4)2SO4 and

NH4NO3 composition. The absorption lowering depends on the composition, the higher the 126 NH4NO3 percentage, the lower the RH water absorption starts. Authors explained that cracks formed on the surface of dry particles may cause this phenomenon. Small cavities were also observed in this study (Figure 7–5), which could have the similar effect as the cracks mentioned in Ge et al. (1998).

Figure 7–5. Small cavities on m((NH4)2SO4):m(NH4NO3)=50%:50% mixture dry particles at 30% loading RH.

Based on the discussion above, mass gains of m((NH4)2SO4):m(NH4NO3)=50%:50% could be explained that mass gain starts declining at 50% loading RH. With 50% NH4NO3 in the mixture, the particle state conversion started at 50% RH, which was 58% when

NH4NO3 is only 20%. This trend agrees with findings of Ge et al. (1998). Single particle analysis was studied in Ge et al. (1998), while capillary condensation plays a role in this study, which prompts more water absorption. Loading curves of 50%, 55%, and 60% RH in Figure 7–6 exhibit the shape of wet particle loading, which indicating the particle state conversion as well. 127 25 4

)

( 20 )

O O

2 3

H H

n n

i 15 i

(

4 p

o r

@ 2

n 10

e r

u Dry 30%

G s

s Dry 45% e

s 1 5 r

P Dry 50%

M Dry 55% Dry 60% 0 0 30 45 50 55 60 0 5 10 15 20 Loading Relative Humidity (%) Mass Loading (mg) Figure 7–6. Mass gains and loading curves close to the MDRH of salt mixture when m((NH4)2SO4):m(NH4NO3)=50%:50%

The unusual deliquescence process of (NH4)2SO4 and NH4NO3 was also discussed in earlier publications (Ge et al. 1998, Wu et al. 2019). As described in previous study, the low ERH of NH4NO3 may lead to some metastable droplets at very low RH, where particle should be completely dry per thermodynamic prediction (Ge et al. 1998). The ERH of the mixture particles were found lower than that of the pure (NH4)2SO4, which may cause incomplete efflorescence and maintain aqueous state over a wide RH range (Wu et al.

2019). Ge et al. (1998) also claimed that the (NH4)2SO4 and NH4NO3 results were difficult to explain, and it is not sufficient to use thermodynamics and surface morphology of mixture particles. Therefore, the filter loading behavior of the salt mixture particles will also be difficult to predict with other particles attached on salt mixture particles in the atmosphere. The (NH4)2SO4 and NH4NO3 mixture particle composition, loading RH, and particle dehydration history could affect the final loading results, which is a complex process needing more research.

128 7.1.3.2 Wet Mixture Particles Loading

As mentioned in the introduction, various composition wet mixture particles were also used for filter loading tests. Based on the previous studies mentioned in Section 7.1.2,

ERH of the mixture particle is generally below 45%, hence 45% and 60% RH were tested for wet mixture particle loading. Figure 7–7 shows mass loadings of various composition wet mixture particle loading. It could be found that most of mass gains are comparable except mass gains of pure NH4NO3 at 60% RH.

15 AS:AN=100%:0%

)

g AS:AN=80%:20%

( AS:AN=50%:50%

2 AS:AN=20%:80%

H 10 AS:AN=0%:100%

a 5

0 45 60 Loading Relative Humidity (%)

Figure 7–7. Mass gains of wet (NH4)2SO4 and NH4NO3 mixture particles on filter samples at different loading RHs

Wet salt particles remain wet until surrounding RH is below its ERH. The amount of water particles carrying depends on the surrounding RH. As described in Section 6.2.3.1, wet particles form a liquid film on the top of the nanofiber layer, which cause lower mass gains than dry particles. SEM images of all conditions in Figure 7–7 show the similar

129 deposition patterns as well. While the mass gain of pure NH4NO3 wet particles is about double of the other mass gains. SEM images of filter samples loaded with pure NH4NO3 wet particles and wet m((NH4)2SO4):m(NH4NO3)=20%:80% mixture particles at 60% RH are showing in Figure 7–8. Again, filter samples were artificially bended before preparing

SEM samples, so that salt cake would break. For pure NH4NO3 wet particles loaded filter sample, wet particles percolate through the nanofiber layer without damaging it. For salt mixture wet particles loaded filter sample, wet particle film formed on the nanofiber layer without percolating it. From hydration and dehydration curves in Wu et al. (2019) at 60%

RH, the pure NH4NO3 wet particle growth factor is about 2.8 times in volume than its dry state; while it is no more than 1.8 times for other salt mixture wet particles. The extra water absorbed by NH4NO3 wet particles made the percolation through nanofiber layer easier, so that mass gain would be greater.

Figure 7–8. SEM images of filter samples loaded with pure NH4NO3 wet particles (left) and wet m((NH4)2SO4):m(NH4NO3)=20%:80% mixture particles (right) at 60% RH

130 7.1.4 Conclusions

To better understand ambient hygroscopic salt particles on air filters loading performance, (NH4)2SO4 and NH4NO3 mixture particles with various mixing ratios were used to simulate ambient hygroscopic salt particles. A nanofiber layer coated cellulose filter was used as the test filter, which has the similar structure as the one used in the single species salt particle loading (Section 6.2). It was found that mass gains of dry salt mixture particles were complex and depend on mixture particle composition, loading RH, and particle dehydration history. The mass gain complexity is caused by the salt mixture hygroscopicity, which is proven complicated and unpredictable. As aforementioned, ERH of NH4NO3 is very low or doesn’t exist, and metastable droplets caused by incomplete efflorescence may bring uncertainties in the salt mixture states. Results of wet salt mixture particles is a little simpler and they are comparable and less composition, RH dependent.

This study reveals the complexity of the (NH4)2SO4 and NH4NO3 mixture on filter loading at different RH, and it offers insight of lab-simulated ambient particles filter loading performance. Based on current experiment results, wet salt mixture particles could offer a conservative filter holding capacities estimation, while it cannot represent dry particles filter loading pressure evolution curves. With the aid of the Global Air

Pollutant/Meteorological Condition Database introduced in Chapter 4, a location specified

(NH4)2SO4 and NH4NO3 mixing ratio could be found, and the location specified filter loading tests could be achieved to better predict the holding capacities of candidate filters.

131 7.2 Soot with Organic Compounds Coating 7.2.1 Introduction

Soot is a major constituent of ambient air pollutants. It accounts for higher percentage in urban area, as it is generated from combustions. Transportation, industry, biomass burning, and other combustion processes could generate soot particles. Soot consists mainly of solid carbonaceous particles with small amount of other organic compounds from incomplete combustion. For example, diesel soot exhaust consists of solid carbon particles and volatile or soluble organic compounds (Kittelson 1998). Soot particles could act as cloud condensation nuclei (CCN) and ice nuclei (IN), and they may also have adverse effect on people’s health as they are in inhalable size range. For ambient air filters, soot removal performance and filter loading has been studied with lab-generated soot particles (Godoy and Thomas 2020, Kim et al. 2009b, Sun et al. 2019). It was found that pressure curves of the pure soot loading has a linear relationship with mass loading on

HEPA filters, and Godoy and Thomas (2020) studied the RH effect during soot particles loading and on soot loaded filter samples. It was found that the loading RH does not affect soot loading as soot particles are non-hygroscopic, while the pressure of loaded soot particles declines at lower RH restructure when it undergoes a higher conditioning RH.

The aforementioned two studies were carried out with fresh lab-generated soot, which is considered as hydrophobic (Kireeva et al. 2010, Popovicheva et al. 2008a,

Popovicheva et al. 2008b, Tritscher et al. 2011, Weingartner et al. 1995). It has been studied extensively that sulfuric acid and some organic compounds can alter the hygroscopicity of soot particles in both laboratory and field studies. (Dalirian et al. 2018, Ebert et al. 2002,

Huang et al. 1994, Kireeva et al. 2010, Köllensperger et al. 1999, Leung et al. 2017, Li et 132 al. 2017, Liu et al. 2020, Mikhailov et al. 1996, Popovicheva et al. 2008a, Popovicheva et al. 2008b, Saathoff et al. 2003, Schnitzler et al. 2014, Tritscher et al. 2011, Weingartner et al. 1995, Weingartner et al. 1997, Wu et al. 2018, Yuan et al. 2019, Zhang and Zhang 2005,

Zhang et al. 2008) It was reported that aging process of soot particles could improve hygroscopicity (Li et al. 2017, Liu et al. 2020, Weingartner et al. 1995). In the meantime, soot particles shrinkage at elevated RH was also reported for both with and without organic compounds coating (Ebert et al. 2002, Huang et al. 1994, Ma et al. 2013, Tritscher et al.

2011, Weingartner et al. 1995, Zhang et al. 2008). Both the hygroscopicity and morphology change of soot particles with RH would affect ambient air filters performance. It would be worthwhile to study filter loading performance with aged soot particles at different RH, which is close to the actual working conditions of ambient air filters.

The most common aging method for soot particle study is utilizing a smog chamber, whereas it is not a continuous flow system. It is good for soot particle study, while does not fit filter loading study. For example, in Leung et al. (2017), it took more than 1h to activate the aging process, which cannot be applied to filter loading testing experiments.

Another soot particle coating method was described with continuous flow coating process without aging chamber (Dalirian et al. 2018, Xue et al. 2009). Lab-generated soot particles would be introduced into a heated organic compounds reservoir where organic compounds were vaporized. Organic compounds vapor would condense on the soot particles upon cooling to room temperature. This aging method fits filter loading experiments and would be used here with some modifications.

133 7.2.2 Experimental Method and Material

Similar to other filter loading tests introduced before, LAEAF Test Rig was used in this study with RH control function at room temperature. Atomizer was not used in this study, and the homemade soot generator (Section 2.1) was used together with the organic compounds coating apparatus introduced in Section 2.2. Two organic compound carrying flow rates were used in this study to compare the effect of the amount of the organic compounds on filter loadings. The amount of the organic compounds was controlled so that homogenous condensation would not happen. Two organic compounds, glutaric acid and oleic acid, were planned to be coated on soot particles separately. Table 7–1 lists the chemical and physical properties of glutaric acid and oleic acid.

Table 7–1. Chemical and Physical Properties of glutaric acid and oleic acid Molar Melt Boiling Water DRH mass Density point point solubility (g/mol) (℃) (℃) Glutaric 1.429 g/cm3 639 g/L at 85% 132.11a 95 ~ 98 a 200 a acid at 15 ℃ a 23.9 ℃ a ~90% b Oleic 0.895 g/cm3 282.46c 13 ~ 14 c 360 c Insoluble c acid at 25 ℃ c a: (National Center for Biotechnology Information 2020b), b: (Parsons et al. 2004), c:(National Center for Biotechnology Information 2020a)

It could be found that glutaric acid is hydrophilic while oleic acid is hydrophobic.

Glutaric is a byproduct of the fossil fuel or biomass burning. It was also produced from unsaturated hydrocarbons and fatty acids (Kawamura and Yasui 2005, Li et al. 2013). Oleic acid is generated from cooking, especially meat cooking (He et al. 2004, Rogge et al. 1991).

The selection of organic compounds is mainly based on their hygroscopicity and atmospheric relevance. The coated soot particles could represent hydrophobic or hydrophilic coated soot pollutant in the atmosphere. 134 In this study, uncoated and coated soot particles were loaded on filter samples at

30%, 60%, 90%, and 95% RH. Based on previous studies, soot shrinkage happens when surrounding RH is close to saturation, so that 90% and 95% were selected. The air filter used in this study is the same as the one used in Section 7.1. And it was also loaded to 4 inches of water at 5.3 cm/s face velocity. The filter handling process was same as descriptions in Section 1.6, so that moisture absorbed in filter samples could be eliminated.

It should be noticed that mass loading includes organic compounds coated on soot particles.

Soot particle morphology can be described by mass fractal dimension. McMurry et al. (2002) and Park et al. (2003) developed DMA-APM-CPC technique to characterize soot structure and density by particle mass m and mobility diameter dm.

6푚 6퐴푑 퐷푓푚 6퐴 푚 (퐷푓푚−3) 휌푒푓푓 = 3 = 3 = 푑푚 휋푑푚 휋푑푚 휋

ρeff is the effective density and Dfm is mass fractal dimension. Mass fractal dimension could be used to describe soot structure. In this study, the soot particle morphology would be characterized by mass fractal dimension.

7.2.3 Results and Discussion

Due to COVID-19 university closure, this study was not completed as planned.

Only glutaric acid coated soot particles loading was studied with few data points unfinished.

However existing results can provide some understandings for current filter testing and suggestion for future study.

Mass of pure soot and coated soot particles with DMA selected mobility diameter was measured by aerosol particle mass analyzer (APM). Effective density of given particle 135 mobility diameter could be calculated according to the equation above. Figure 7–9 lists existing effective densities of pure soot and glutaric acid coated soot. There were two air flow rates to bring glutaric acid vapor into the system, 2 LPM and 4 LPM, which could adjust the amount of the glutaric acid coating. Since fresh soot particles are fractal-like shape, their effective densities declining with larger particle mobility diameter. The power- law fitting could be used to infer mass fractal dimensions accordingly.

0.6

0.5

)

/ 0.4

(

n 0.3

i t 0.2 c 30% RH Soot

f 30% RH Soot + GA@2LPM

E 30% RH Soot + GA@4LPM 0.1 90% RH Soot 90% RH Soot + GA@2LPM 90% RH Soot + GA@4LPM 95% RH Soot + GA@4LPM 0 50 100 200 Mobility Diameter (nm)

Figure 7–9. Effective densities of pure soot and glutaric acid coated soot particles with different mobility diameters

Table 7–2 lists mass fractal dimensions of all cases in Figure 7–9. It could be found that mass fractal dimension is positively correlated with amount of the glutaric acid coating and RH. The glutaric acid coating results agrees with Xue et al. (2009), and authors found that soot particles shrank when glutaric acid coating was applied and this restructuring

136 could be boosted by elevated RH (90%). The pure soot restructuring was not significant.

There are several possible reasons: as mentioned in the introduction, pure soot is hydrophobic and not sensitive to the RH change; pure soot restructuring was insignificant and only local compaction could happen when condensed water is limited (Ma et al. 2013).

These finding indicating a saturation and oversaturation should be tested to in the future to achieve more compacted soot particles. However, the compaction induced by glutaric acid coating is significant enough. The amount of the glutaric acid and the coated particle

EC/OC ratio was not tested due to time constrain, which should be quantified in the future.

Table 7–2. Mass fractal dimensions of pure soot and glutaric acid coated soot at different RHs RH Pure soot Soot+GA@2LPM Soot+GA@4LPM 30% 2.137 2.193 2.575 90% 2.198 2.173 2.734 95% / / 2.851

To compare pure soot particles and coated soot particles visually, a filter sample was firstly loaded with pure soot particles and followed by glutaric acid coated soot particles. The filter sample was bent to damage the soot layer before SEM image was taken.

Figure 7–10 shows clear differences between pure soot particles and glutaric acid coated soot particles. Pure soot particles are fractal-like structure while the glutaric acid coated soot particles are more compact clump-like structures. The restructuring induced by coating process could improve filter holding capacity by reducing the total area. The structure change occurred before soot particles deposited on the filter, which is different from the effect of RH on deposited salt particles mentioned in previous chapters.

137

Figure 7–10. SEM images of a filter sample loaded with pure soot in the bottom and glutaric coated soot on the top

Mass Gains of pure soot particles and glutaric acid coated soot particles at different loading RHs are shown in Figure 7–11. It could be found that the RH effect on mass gain is not significant. These results agree with mass fractal dimension findings that soot restructuring is insignificant when condensed water is limited. It should be noticed that the mass gain in Figure 7–11 is the total mass of the soot and glutaric acid coating. As mentioned above, the amount of the glutaric acid coating was not measured. However, the mass gains of 4LPM glutaric acid flow is much more than that of 2LPM one. With the aid of the mass fractal dimension and SEM images above, it could be inferred that glutaric coating could improve filter holding capacity when the coating amount reach a certain level.

According to filed measurements, the aged soot in the atmosphere is covered with various pollutant to restructured to embedded soot (Li et al. 2017, Wu et al. 2018, Yuan et al. 2019).

Therefore, the glutaric acid coated soot particles could simulate the aged soot in the 138 atmosphere to a certain extent. Further particle analysis should be conducted so that the percentages of the soot particles and coating could be identified.

Soot Soot + GA@2LPM 2.0 Soot + GA@4LPM

)

(

1.5

n 1.0

M 0.5

0 30% 60% 90% 95% Loading RH Figure 7–11. Mass Gains of pure soot and glutaric coated soot at different loading RH

7.2.4 Conclusions

This study used soot particles coated with lab-generated common atmospheric organic compound to simulate the aged soot in ambient air. Effective densities and mass fractal dimensions of pure soot particles and glutaric acid coated soot particles were characterized with DMA-APM-CPC technique. Glutaric acid coating and elevated RH can restructure pure soot particles to a more compact structure. The simulated aged soot particles were loaded to nanofiber coated filter media to 4 inches of water. Mass gains of organic compounds coated soot particles is higher than that of pure soot particles.

According to field studies, ambient aged soot particles are more compact than fresh soot, which were simulated in this study. Previous filter loading testing used fresh lab-generated

139 soot particles to represent ambient soot. However, the ambient soot aging process restructures soot particles, which could result in a higher mass loading. Though only glutaric acid coating was tested, this study implies that to better simulate ambient soot particles in labs, aging process should be considered. RH effect on the aged soot particles is worthwhile to study, with both hydrophobic and hydrophilic organic compounds. Some other treatments that may prompt the hygroscopicity, such as UV and Ozone, could be added in the future to better simulate aging process in a continuous flow system to test air filters.

140 Chapter 8: Effect of RH Change on Pressure Drop of Hygroscopic Particles Loaded Cellulose Filter Media

This chapter is in preparation: Pei, C., Ou, Q., and Pui, D. Y. H., Effect of relative humidity change on pressure drop of loaded cellulose filter media with hygroscopic deposits

8.1 Introduction

Cellulose media is a common material for air filters, because it is inexpensive and easy to process. It is commonly used in many air-cleaning applications, including gas turbine filters, engine intake filters, HVAC system filters. Whereas there are some shortcomings that hinder the use of cellulose filter media, such as low initial efficiency, relatively high differential pressure, and limited holding capacity. Nanofiber technology development highlights the potential of improving air filters with sub-micron fibers, whose diameter is normally less than 1 µm. When nanofibers are coated on the cellulose substrate, the performance of overall cellulose air filters could be improved. The filter initial differential pressure is reduced due to thin fibers and less drag force. The dense nanofiber coating layer improves the initial filtration efficiency significantly. The smaller pore size and greater fiber surface area contribute to the efficiency improvement. In addition, the holding capacity is also improved. With the aid of the nanofiber layer, surface filtration is achieved, which forms a high-efficiency particle cake and linear pressure drop. In this study, the nanofiber coated cellulose filter will be studied.

In Chapter 5, 6, and 7, temperature and RH effect on air filter performance were studied, however they were all conducted at a constant RH, which is not realistic in actual filter operation. Based on the conclusion of Chapter 5, air filters operating RH is dominant

141 in air filter performance comparing to the operating temperature. The RH effect on air filter performance should be studied in a more realistic way other than a constant RH. From

Chapter 4, Global Air Pollutant/Meteorological Condition Database, it can be found that

RH seasonal variation is significant in some places. Whereas the seasonal variation is a gradual process that RH will not change abruptly. Air filter operating RH can be regarded as relatively stable in the time scale of one month or one season. However, there are some scenarios that the RH will change rapidly. For example, thunderstorms can raise the ambient RH to 100%, and the RH will revert to close to previous level afterwards.

Commercial aircraft cabin is normally conditioned to 20% RH to prevent metal from rusting, which is less than the comfortable RH of a person. When the aircraft is parked on the ground with the outside hot and humid air entering the cabin, the powerful cooling system will cool the outside air down, so that condensation could happen and cause fogging in the cabin, which can bring cabin RH to 100%. The effect of this kind of RH spike on air filters should be studied to better understand the filter performance change after such RH spike.

The study of effect of the RH variation on the filter performance is limited.

Montgomery et al. (2015a) studied the impact of the relative humidity on the loaded HVAC filter with hygroscopic (NaCl) and non-hygroscopic (Al2O3) particles. It was found that the elevated RH could cause the irreversible pressure reduction and efficiency drop when the

HVAC filter is loaded with hygroscopic particle. When the HVAC filter is loaded with non-hygroscopic particles, the similar phenomenon was not observed. Authors speculated the small growth factor increase could change the dendrite morphology when the RH is below its DRH. Joubert et al. (2011) performed similar experiments and explained as the

142 absorption of moisture from the high RH air by cake particles, which could restructure the dendrites and create preferential air passageways. However, their studies limited elevated

RH to below DRH of NaCl. The filter performance of elevated RH above the DRH was not covered. Wang et al. (2020) studied electret filter under the RH varying condition with

KCl deposition. It was found that the differential pressure of electret filter drops when the filter is treated at the RH higher than the loading RH, which agrees with conclusions of

Montgomery et al. and Joubert et al. Wang et al. also treated the loaded filter sample with lower RH than the loading RH and found the differential pressure stayed unchanged. Both of studies loaded and treated filters below the DRH of testing salts, which cannot apply to the scenarios mentioned before. In addition, as emphasized previously, the challenging salt species and states are critical when the test filter is designed to remove ambient particles.

In this study, KCl, (NH4)2SO4 and NH4NO3 are used to load nanofiber coated cellulose filters in their wet or dry states. Both conditioning RH below and above the loading RH are tested so that different scenarios are simulated. The loading and conditioning RH cover from below the ERH to above the DRH of challenging salts, which is a comprehensive test matrix to simulate the actual operation conditions.

8.2 Experimental Method and Material

This study was carried out on the LAEAF Test Rig. The test temperature was at the room temperature, and the RH control function was employed. The test system was described in detail in Chapter 1. A solenoid valve is added on the compressed air line as the shut-off valve to atomization (Figure 1–6). The atomization will stop when test filters

143 reach preset pressure. This study used the same filter media as Chapter 5 and Section 6.2

(Figure 5–2). The filter substrate is cellulose media and the nanofiber coating is made of nylon.

In this study, filter samples were firstly loaded to 4 inches of water (≈1000 Pa) at the loading RH (X %) and the face velocity is 5.3 cm/s. Atomizer shut-off valve would stop the compressed air and salt particle generation was halted. The system RH would automatically switch to the conditioning RH (Y %) and the test filter flow rate remain the same. Since two filter samples cannot reach 4 inches of water at the same time, only one sample holder on the LAEAF Test Rig was used in this study. The criterion of conditioning termination is the differential pressure change rate is less than 0.1 inches of water/hour.

Figure 8–1 is the flowchart of the loading and conditioning processes. Only one-step RH change was performed in this study.

Load test filter at X% RH until 4 in. H2O Pressure Drop

Shut down atomizer while keep the test filter flow

Condition test filter at Y% RH

Stop until Pressure Drop (Terminal Pressure) instantaneous decreasing rate less than 0.1in. H2O/hr

Figure 8–1. Loading and conditioning processes of the test filter sample

144 According to Section 1.4, the state of a hygroscopic salt particle depends on the surrounding RH and hydration history. Figure 8–2 shows four possible states A-D of a salt particle. For any given loading RH, there are two possible conditioning RH. For example, when the filter sample is loaded at A, it could be conditioned at B or D. This RH variation may or may not cause the salt particles state change. From A to B, salt particles remain in dry state, but with higher chance to absorb more moisture; whereas from A to D, dry salt particles will absorb moistures from surrounding air and become salt droplets. The possible loading and conditioning RH process is listed in Table 8–1. It should be noticed that A, B,

C, D represent salt states rather than exact RH. A could be any RH below the ERH, B and

C could be any RH below the DRH while above the ERH, and D could be any RH above the DRH. In this study, all the possible processes were studied with KCl, (NH4)2SO4, and

NH4NO3, when it is feasible. As discussed in Section 1.4 and Section 6.2, only wet state pf

NH4NO3 can be achieved with the current system (Dougle et al. 1998, Lightstone et al.

2000). Therefore, only C and D are available for NH4NO3 and the processes can only happen between C and D, both in wet states.

145 6 Wet State 5 D

o 4

F C

h 3

o r 2 G Dry State A B 1 Efflorescence RH Deliquescence RH 0 0 20 40 60 80 100 Relative Humidity (%)

Figure 8–2. Illustration of possible states of loading and conditioning processes (KCl and (NH4)2SO4)

Table 8–1. All possible process from loading RH to Conditioning RH Loading RH Conditioning RH State Change A Dry => Dry A B Dry => Dry D Dry => Wet B Dry => Dry B A Dry => Dry D Dry => Wet C Wet => Wet C A Wet => Dry D Wet => Wet D Wet => Wet D C Wet => Wet A Wet => Dry

146 8.3 Results and Discussion 8.3.1 KCl Loaded Filter Sample

According to the conclusions from Chapter 6 and previous studies, for hygroscopic salts filter loading, the higher the loading RH, the less resistance of the salt cake. For a given amount of the deposited salt, the surrounding RH growth can lead to a differential pressure reducing. Figure 8–3 is the differential pressure curve during conditioning. It only shows the conditioning curves of KCl loaded filter samples at 30% loading RH. The horizontal axis is the time after salt particle generation was halted. The RH of air flow through the filter sample started to switch to the conditioning RH and stabilize quickly. It can be found that the time took for reaching a stabilized pressure drop for different conditioning RH varies from 2000 seconds to 12000 seconds. The larger the difference between loading and conditioning RH, the shorter the time needed for stabilization. This phenomenon was also observed in Montgomery et al. (2015a) and Wang et al. (2020). The possible explanation is the higher the conditioning RH, the more moisture can be absorbed, so that dendrite restructuring is easier to take place with more condensed water.

It should be noted that the differential pressure also declined when the conditioning

RH is the same as the loading RH. This was also observed by Montgomery et al. (2015a), when NaCl was loaded on the HVAC filters. During the filter loading, there could be two competing mechanism in the pressure increasing. The first one is the particle deposition process that can cause pressure growth. While the deposition structure may not be the most stable structure under the influence of the RH, and the particle cake may restructure afterwards. However, the pressure reduction was not observed in Wang et al. (2020). It is because the amount of salt particles deposited on the electret media was limited, and 147 dendrites were not formed on filter fibers. So that dendrite restructuring could not take place. Joubert et al. (2011) found the pressure unchanged when NaCl was loaded and conditioned at 5% RH, which is very low, and the influence of moisture is very limited.

) 3.40

2 3

(

o r 2

D 2.15

s 1.56

s 30%=>30% e 1 r 30%=>60%

P 0.75 30%=>75% 30%=>90% 0 0 5,000 10,000 Time after loading (s) Figure 8–3. KCl loaded filter samples pressure stabilization during the RH conditioning process

The terminal pressures of other loading RH and conditioning RHs are shown in

Figure 8–4. Solid points represent dry salt particle loadings and open points represent wet salt particle loadings. When filter samples were loaded with dry KCl particles at 75% RH, the terminal pressure of 30% conditioning RH is very close to 4 inches of water. While the terminal pressure of 90% conditioning RH dropped to about 2 inches of water. The overall trend of dry particle loading is the higher the conditioning RH, the lower the terminal pressure.

148 4

) Wet

2 Particle

H Region

n 3

(

s s 2

l Loading RH

n 30% i 1 Dry m 75%-Dry r Particle Multi-state

T 75%-Wet Region Region 90% 0 0 20 40 60 80 100 Conditioning RH (%) Figure 8–4. KCl terminal pressures of different loading and conditioning RHs

To understand the salt cake structure changes when dry salt particles loaded filter samples are conditioned at higher RH, SEM images were taken. Figure 8–5 is the SEM images of KCl loaded filter sample that was loaded at 30% RH and conditioned at 90%

RH. It could be found that the salt cake is a very open structure comparing to the filter sample without conditioning (Figure 6–14). Nanofiber coating layer can be observed on the right-side magnified image. The differential pressure reduction to almost clean filter pressure drop can be explained by this SEM image and dendrite restructuring shown in

Figure 6–3. 90% RH is above the DRH of KCl, the salt particles exhibit a coalesced state, that nano particles are dissolved and combined to micrometer size salt particles. When the conditioning RH is lower than 90%, capillary condensation can also introduce water between salt particles, which would help dentrite restructuring (Figure 6–3). However, the cake structure was retained in Figure 8–5, which may be due to the limited amount of the water absorbed at 90%. The conditioning at 100% RH will be discussed in Section 8.3.2 for (NH4)2SO4 loaded filter sample.

149 For wet particle loading, the differential pressure reduction is not significant as dry particle. The 75% wet KCl loaded filter samples holded 4 inches of water when they were conditioned at 30% and 75%. The terminal pressure reduction was observed at 90% conditioning RH. The wet particle loaded filter sample conditioning will be discussed in

Section 8.3.2 as well.

Figure 8–5. SEM images of KCl loaded filter sample that was loaded at 30% RH and conditioned at 90% RH. Left: 1000X; Right: 5000X.

8.3.2 (NH4)2SO4 Loaded Filter Sample

The terminal pressures of (NH4)2SO4 loaded filter samples at different loading and conditioning RHs are shown in Figure 8–6. In general, the trend is similar to the KCl case.

Two processes will be discussed. The first one is filter samples that were loaded with wet particles and conditioned at lower RH. The second process is filter samples that were conditioned at 100% RH. For the second process, both dry and wet salt particle loaded filter samples will be discussed.

150 4

) Wet

O 2 Particle

n 3 Dry i Region

(

e Particle

u Region

s s 2

r Loading RH

l 30%

i 45%-Dry 1

m Multi-state

r 45%-Wet

e Region T 75% 90% 0 0 20 40 60 80 100 Conditioning RH (%)

Figure 8–6. (NH4)2SO4 terminal pressures of different loading and conditioning RHs

As described in Section 6.2.3.1, wet particle deposition on nanofiber coated cellulose filter media can hardly accumulate in the direction perpendicular to the filter media and a film will be form on the top of nanofiber layer. When it was conditioned at the same RH as the loading RH, there are possibility that salt can percolate through the nanofiber layer or even damage the nanofiber layer. When these conditions happened, the terminal pressure would drop. When the conditioning RH is below the loading RH, the terminal pressure could hold at 4 inches of water. As shown in Figure 8–7, water in the film evaporated at 30% conditioning RH, while retain the film structure so that the differential pressure held at 4 inches of water. The same condition can also be seen when the conditioning RH is 45% RH.

151

Figure 8–7. SEM images of (NH4)2SO4 loaded filter sample that was loaded at 90% RH and conditioning at 30% RH

Figure 8–8 shows the process that filter samples were conditioned at 100% RH. It is clear that there is limited amount of salt particles on the nanofiber layer on both dry and wet particle loaded filter samples. The terminal pressures in Figure 8–6 is close to a new filter sample. It is highly possible that deposited particles absorb water at 100% RH. The amount of water absorbed at 100% RH can dissolve salt particles so that salt solutions can percolate into filter sample to reduce the flow resistance.

152

Figure 8–8. SEM images of (NH4)2SO4 loaded filter sample that was loaded at 45% RH and conditioning at 100% RH. Left: dry particle loading; Right: wet particle loading

8.3.3 NH4NO3 Loaded Filter Sample

NH4NO3 loaded filter sample was also studied. NH4NO3 is the wet particle loading condition. The overall trend that the higher the conditioning RH, the lower the terminal pressure still holds. There is a terminal pressure discrepancy when the conditioning RH is

60%. The terminal pressure of 30% loading RH is only about a half of 60% loading RH.

One of the possible reasons is that the amount of salt deposited on filter media can also affect the terminal pressure. In Figure 6–11, the volume loading of 60% RH NH4NO3 is about twice of that of 30% RH. Though salt deposition film can percolate into the filter, the extent is limited by the amount of the water absorbed. When the conditioning RH is

100% (Figure 8–8), almost all salt can percolate into the filter; while when the conditioning

RH is 60%, the absorbed water is limited.

153 4

)

(

s 2

n i 1

r Loading RH

T 30% 60% 0 0 20 40 60 80 100 Conditioning RH (%)

Figure 8–9. NH4NO3 terminal pressures of different loading and conditioning RH

Similar to other wet particles, when NH4NO3 was loaded at 60% RH, and conditioned at 30% RH, the film could also form as seen in Figure 8–10. Comparing to wet

(NH4)2SO4 particles, NH4NO3 particles coalescence more, but droplets boundaries are still distinct. Even without ERH, the RH drop can cause water evaporation so that the salt deposition retains its shape and maintain the pressure drop across the filter media.

Figure 8–10. SEM images of NH4NO3 loaded filter sample that was loading at 60% and conditioning at 30% RH

154 8.4 Conclusions

This study focused on the pressure variation of hygroscopic salt particles loaded filter samples when the surrounding RH changes. KCl, (NH4)2SO4, and NH4NO3 dry and wet particles were loaded on the nanofiber coated cellulose filter media at the loading RH to 4 inches of water. The filter media was then conditioned at conditioning RH. The terminal pressure curves and terminal pressure were recorded.

It was found that for dry particle deposition, the higher the conditioning RH, the lower the terminal pressure. When the conditioning RH is below salt’s DRH, the reduction of differential pressure is mainly caused by the capillary condensation discussed in Section

6.1.3.1, and the condensed water prompts dendrite restructuring. The overall pressure of loaded filter sample reduced after dendrite restructuring. When the conditioning RH is above salt’s DRH, the deliquescence water could prompt a further differential pressure reduction. When the conditioning RH is 100%, dissolved salt can percolate through the nanofiber layer which would cause a significant pressure reduction, to where the filter pressure is close to a new filter. The conditioning at higher RH process could reduce hygroscopic particle loaded filter pressure, whereas it is not clear whether this process can be used to prolong the filter life or a method to “recover” used air filters. The second round loading at lower RH should be done to further investigate the possibility of extending the filter life.

For wet particles, the elevated conditioning RH can prompt pressure reduction by introducing more water to salt deposit. However, the lowered conditioning RH would not cause a pressure reduction. The reason is that the water in the liquid film evaporated at

155 lower RH while the film structure was retained so that differential pressure remained unchanged. This finding is important to alert users that dry air blowing may “freeze” wet salt deposit on the filter media so that the pressure will stay unchanged. The results can also be used to explain the filter pressure change after a RH spike mentioned in the introduction. However, this phenomenon may help in reverse pulse cleaning application, where the solidified particle deposition can be peeled off easily.

This study broadens understandings of RH change effect on the hygroscopic particles loaded filters. It covers the condition that RH is above the DRH of salts and the condition of wet particle loaded filters. The next step would be cycling RH effect on the filter loading, which is closer to the actual conditions. Also, the possibility of utilizing the hygroscopic particle to prolong filter life or clean filters could be considered. Since only pressure change was studied here, the efficiency variation should also be studied in the future.

156

Part 3: COVID-19 Related Study: Mask and Respirator Filtration Study

COVID-19 was characterized as a pandemic by WHO on March 11, 2020, and it is a global challenge for the world since then. Center for Filtration Research is one of the leading filtration research centers in the world, and several COVID-19 related mask and respirator filtration studies were conducted at the Center for Filtration Research. Chapter 9 compared common materials that have the potential to be used as alternative masks for the public in daily protection. Chapter 10 aimed to evaluate the effect of decontamination of the commercial and alternative respirator and mask materials from the filtration perspective.

Both two chapters were collaborated with Dr. Qisheng Ou and share the first authorship of the published or submitted papers.

158 Chapter 9: Alternative Face Masks Made of Common Materials for General Public: Fractional Filtration Efficiency and Breathability Perspective

This section has been published: Pei, C.1, Ou, Q.1, Kim, S. C., Chen, S-C., and Pui, D. Y. H. (2020). Alternative Face Masks Made of Common Materials for General Public: Fractional Filtration Efficiency and Breathability Perspective Aerosol and Air Quality Research 20 doi:10.4209/aaqr.2020.07.0423

9.1 Introduction

COVID-19 was characterized as a pandemic by World Health Organization (WHO) on March 11, 2020, and it is a global challenge. As an infectious disease that is mainly transmitted through droplets, the appropriate respiratory protection could protect the healthy population from infection (National Academies of Sciences Engineering Medicine

2020, WHO 2020a). Respiratory protection is especially critical for frontline healthcare workers, who have to contact COVID-19 patients directly. When one wears a mask, the mask can help reducing virus spreading by capturing most of emitted droplets during speaking, sneezing, and coughing (Anfinrud et al. 2020, Davies et al. 2013, Greenhalgh et al. 2020, Rengasamy et al. 2010, van der Sande et al. 2008); on the other hand, the mask can offer effective protections by filtering virus-laden particles in the air (Johnson et al.

2009, Lai et al. 2012, Leung et al. 2020b, Wang and Yu 2020, Zuo et al. 2013). Therefore, the priority is to ensure that frontline healthcare workers are well protected. However, the

COVID-19 pandemic has led to the personal protective equipment (PPE) shortage, especially respirators and masks. Therefore, various efforts have been paid to allocate available respirators and masks inventory for frontline healthcare workers, including decontamination (Liao et al. 2020, Ou et al. 2020). 159 As COVID-19 is infectious, it is critical to curb the spread the virus to protect the healthcare system from overload, so-called flattening the curve. It has been found that asymptomatic infected people and pre-symptomatic infected people could be potentially infectious sources (CDC 2020b, National Health Commission and State Administration of

Traditional Chinese Medicine 2020). On the other hand, it is equally important for the general public to wear masks or equivalent face coverings to slow down the spread of the virus (CDC 2020c, National Health Commission 2020). As many states are re-starting and social activities are recovering in the United States, many governors across the country issued orders requiring face coverings when social distancing is unable to maintain

(Mendelson 2020). WHO updated its advice on the masks use on June 5, which encourages the general public masking when physical distancing cannot be maintained (WHO 2020b).

However, it has been an ongoing debate about whether universal masking is necessary. As early as Mar 2nd, 2020, there is a correspondence on the Lancet, discussing the necessity of wearing masks in community settings and mentioning that WHO was not recommending it due to lack of evidence. Authors emphasized that “absence of evidence of effectiveness should not be equated to evidence of ineffectiveness, especially when facing a novel situation with limited alternative options” (Leung et al. 2020a). A similar concern was also raised by Morawska and Cao (2020) where authors called on considering the possible airborne transmission. On July 6th, 239 scientists signed an open letter to the

WHO to address the airborne transmission of COVID-19, which is published in Clinical

Infectious Diseases (Morawska and Milton 2020). Later, WHO announced in a press conference that WHO plans to update its guidelines. This discussion brings up the question

160 that how respirators and masks can protect wearers from the possible airborne transmission of COVID-19.

Whereas, considering the current desperate shortage of commercially available masks and respirators, the use of the common materials to substitute masks for the general public can be the only option, such as scarves and bandanas. Also, the Centers for Disease

Control and Prevention (CDC) offers the tutorial for homemade masks (CDC 2020c). It has been studied that homemade masks could not only protect the others from the wearer by reducing the amount of the respiratory droplets exhaled from the wearer but also offer protection to the wearer from inhaling virus attached particles (Anfinrud et al. 2020, Bae et al. 2020, Davies et al. 2013, Howard et al. 2020, Konda et al. 2020, Leung et al. 2020a,

MacIntyre et al. 2015, Mueller et al. 2020, Prather et al. 2020, Rengasamy et al. 2010, van der Sande et al. 2008, Zhao et al. 2020). However, common fabrics, such as T-shirts, cotton cloth, were the most often used materials, which is not as good as commercial masks (CDC

2020d). How the general public could make good use of common materials to improve their homemade masks? In this work, we tested several common materials besides common fabrics to help the general public to protect themselves and people they interact with. It is worth noting that conclusions drawn from this study can offer guidance for the general public under the current pandemic. They can also be generalized to do-it-yourself (DIY) the alternative air pollution protection gears when a commercial respiratory protection equipment is not available.

161 9.2 Experimental Method and Material

Respirators and masks efficiency and breathability are key performance factors, associated with the filtering effectiveness and the breathing resistance of a respirator or a mask. In this study, the fractional filtration efficiency and the pressure drop across the filter media were measured.

Air filtration is not as simple as sieving, and there are several filtering mechanisms, including inertial impaction, interception, diffusion, electrostatic deposition, etc. With different filtration mechanisms particle size less than 0.1 µm or particle size in the range of 0.3 - 1 µm particles, overall fractional filtration efficiency curve will be a U-shape (Ou et al. 2017, Spurny 1986, Yang 2012). The particle filtration efficiency strongly depends on the particle size(Ou et al. 2017, Spurny 1986). The fractional filtration efficiency is defined as the number of particles of a given size penetrating the filter medium, divided by the total number of particles challenged on it.

The fractional filtration efficiency was reported in this study for comparing the performance of common materials with commercial respirators/masks and to screen potential candidate materials. The particle size with the least efficiency of the filter is the most penetrating particle size (MPPS), which depends on filter material itself as well its operating conditions. For the filter without electrical charges (so-called the mechanical filter, e.g., glassfiber HEPA media, engine intake filter media), the MPPS is about 0.1 - 0.3

µm. For the filter with electrical charges (so-called the electret filter, e.g., some furnace filters, N95) the MPPS is around 0.1 µm or less. Common household materials that are not designed for filtration can have MPPS larger than 0.3 µm. Fractional filtration efficiency

162 reported in this study was from 0.03 µm to 1 µm, which covers possible MPPS of most of materials tested and offers the size-resolved material efficiency.

The pressure drop of filter media is a measure of the breathing resistance. The pressure drop across a filter media is proportional to the face velocity with which the fractional filtration efficiency is also correlated. Hence, to compare common materials fairly, the same face velocity should be used. In this study, we tested all materials at 10.5 cm/s. This face velocity is calculated from 85 LPM flow rate and 135 cm2 filtration area.

The flow rate of 85 ± 4 LPM is used by the National Institute for Occupational Safety and

Health (NIOSH) to certify N95. The filtration area of 135 cm2 is the lower end filtration area of commercial N95s. The lower filtration area results in a higher face velocity, which adversely affects the filtration efficiency. Hence, efficiency measurements in this study are conservative, representing the worst conditions that may occur. In this study, we used a differential pressure gauge (Dwyer, Michigan City, IN) to measure the pressure drop across the testing filter media. To ensure the stable and accurate filter face velocity and flowrate, several mass flow controllers (MKS Instruments, Inc. Andover, MA) controlled the flow in the fractional filtration efficiency measuring system.

There are different test standards for respirators and masks. NIOSH tests respirators according to 42 CFR part 84 (Office of the Federal Register 1995). The challenging particle is neutralized NaCl, whose count median diameter is 0.075 ± 0.020 µm and the geometric standard deviation is less than 1.86. The testing flow rate is regulated to 85 ± 4 LPM for one respirator. For the surgical mask, ASTM and FDA have different testing protocols for particle filtration efficiency and bacterial filtration efficiency. In general, the challenging particle of particle filtration efficiency is latex spheres, and the face velocity is defined in

163 a broad range from 0.5 cm/s to 25 cm/s. FDA guidance document for surgical mask 510(k) calculates the particle efficiency by 0.1 µm latex spheres (FDA), while ASTM F2299 claimed a particle size range from 0.1 to 5.0 µm (testing by monodisperse particles) (ASTM

International. 2017). As for bacterial filtration efficiency, the FDA follows ASTM F2101 to test surgical masks(ASTM International. 2019). Staphylococcus aureus bacteria are used with the mean particle size at 3.0 ± 0.3 µm. The testing flow rate is 28.3 LPM. This study used the NIOSH protocol as a reference.

Electron micrographs of SARS-CoV-2 particles show that they are generally spherical, and their diameter is around 0.06 to 0.14 µm (Zhu et al. 2020). The respiratory generated droplets could be several micrometers or larger, which will then evaporate to a smaller nucleus (Anfinrud et al. 2020, Bourouiba et al. 2014, Bourouiba 2020, Gralton et al. 2011, Morawska 2006, Morawska et al. 2009, Morawska and Cao 2020, Scharfman et al. 2016). Therefore, the fractional filtration efficiency in this study (0.03 µm - 1 µm) covered the size range for both the MPPS of common filter media and the SARS-CoV-2 particle cluster size range. The material filtration efficiency at its MPPS or SARS-CoV-2 size could offer a conservative evaluation of the material used. It is worth noting that as aforementioned, the fractional filtration efficiency curve is a U-shape, even the largest particle size used in this study is only 1 µm, the fractional filtration efficiency of several micrometers’ particle could be inferred to be higher than the efficiency at 1 µm. Therefore, if the MPPS is confirmed, the lowest efficiency is found.

This study did not test whole respirators/masks due to limited resources. Common materials tested in this study were not sewn, so all testing samples are cut to a 40 mm diameter disc to fit the filter holder. As aforementioned, the testing face velocity was kept

164 the same as the whole respirators/masks testing, so that the efficiency and pressure drop measurement are kept consistent. The filter holder used in this study hold the media by tightening the outside mounting nut, so it could adapt different material thickness and seal around the sample.

NaCl particles were selected as the challenging particle to align with the NIOSH testing protocols. NaCl submicron particles were generated by atomizing its solutions and then dried by a diffusion dryer. In this study, we measured the fractional efficiency with monodisperse NaCl particles, hence it is important to keep the particle number concentration high enough to get statistically meaningful efficiency measurement in the size range in this study (0.03 µm - 1 µm). A pneumatic atomizer (TSI 3076, Shoreview,

MN) was used in this study, with 0.4 % volume concentration NaCl solution and 30 PSI filtered compressed air.

The efficiency of filter media may change after challenging particles loaded on it, which could be mitigated by reducing the amount of the particles deposited on the filter media. However, the number concentration at the upstream should be high enough so that the efficiency measurement is statistically meaningful. Since we are measuring the fractional filtration efficiency, which allows us to measure the efficiency of a certain particle size at one time. By allowing only monodisperse particles with the size being measured to reach the filter, the loading effect could be mitigated, and the number concentration is high enough at the same time. An Electrostatic Classifier (TSI 3080,

Shoreview, MN) with a long Differential Mobility Analyzer (TSI 3081, Shoreview, MN) classified the NaCl particle to the desired size. A Po-210 Staticmaster® Alpha Ionizer

(NRD, LLC. Grand Island, NY) neutralizer brought the challenging particles to the

165 Boltzmann charge equilibrium. The test filter was then challenged by neutralized monodisperse NaCl particles with a certain amount of dilution air to meet the face velocity requirement. An Ultrafine Condensation Particle Counter (UCPC, TSI 3776, Shoreview,

MN) was connected to the upstream and downstream of the filter holder by a pneumatic switching three-way valve to measure corresponding concentrations. Testing apparatus is shown in Figure 9–1 The fractional filtration efficiency is calculated by the formula below.

퐶 퐶 퐹푟푎푐푡푖표푛푎푙 퐹푖푙푡푟푎푡푖표푛 퐸푓푓푖푐푖푒푛푐푦 = 1 − 푑표푤푛 / 푏푙푎푛푘,푑표푤푛 퐶푢푝 퐶푏푙푎푛푘,푢푝

퐶푑표푤푛 and 퐶푢푝 are the UCPC concentration measurement at the upstream and downstream of the filter holder, respectively. 퐶푏푙푎푛푘,푑표푤푛 and 퐶푏푙푎푛푘,푢푝 are concentration measurements without the test material in the filter holder, accounting for the system particle loss. Theoretically, Filter penetration could be calculated from the ratio of the downstream particle concentration and the upstream concentration. Hence the filtration efficiency is calculated from one minus the penetration. However, in practice, there are some system loss in measurement. When filter sample is not in place, the system loss could be measured. The upstream and downstream sampling length, geometry, and port location could cause slight difference in 퐶푏푙푎푛푘,푑표푤푛 and 퐶푏푙푎푛푘,푢푝 concentration readings. This formula is incorporated with system loss.

166

Figure 9–1. Schematic of the fractional efficiency measurement setup

Due to the time and resources constraint, all results reported in this article were the results from a single measurement. The standard deviation showing on fractional filtration efficiency plots are calculated from the particle concentration variation at the upstream and downstream based on error propagation. The purpose of this study is to offer a reference for the general public to select better alternative mask material. The sample-to-sample variation will not jeopardize final recommendations, besides, variations between the material the general public used, and the material tested cannot be compensated by more replicates.

9.3 Results And Discussion 9.3.1 Commercial Respirators and Masks

To study the efficacy of common materials filtering out the particles and droplets carrying SARS-CoV-2, available commercial respirators and masks were tested as the references in this study. As one of the best filtration research centers in the world, the

Center for Filtration Research at the University of Minnesota received many test requests,

167 and we also collected available masks and other common materials to screen potential face- covering candidates. Some masks received were not in its original package, so their information was insufficient, such as brands, efficiency ratings, etc. For simplicity, we named 10 masks with letters (Mask A to K). Besides 10 flat masks, a shaped KN95 respirator was received. 3M™ 8210 N95, 3M™ 1820 procedure mask, and Kimberly-

Clark™ 47090 procedure mask were tested from the lab PPE stockpile. This makes a total of 14 respirators/masks tested as the reference, which is representative of commercial respirators and masks.

The fractional filtration efficiency and breathing resistance of 14 respirators/masks are shown in Figure 9–2. Fractional filtration efficiency curves could be categorized into two groups, electret media (blue shade) and non-electret media (orange shade). The lowest efficiency of the electret media mask is about 60% at 0.03 µm. The 3M™ 8210 N95, KN95, and Mask A are all above 90% efficient. The overwhelming majority of electret masks fractional efficiencies fall in 70% ~ 85%. The MPPSs of all electret media are below 0.1

µm. It could be found the fractional filtration efficiency of all electret media increases with greater particle size. The Mask F has the poorest performance among all electret, while the fractional filtration efficiency of the Mask F at 1 µm reaches around 85% and will continue to increase beyond 1 µm. The non-electret media masks fractional efficiencies are significantly lower than that of the electret media mask and their MPPSs are about 0.3 -

0.4 µm. The fractional filtration efficiency at MPPS is as low as 20%, while the fractional filtration efficiency increases towards both smaller and larger particles, showing a U-shape.

The fractional filtration efficiency of the Mask K increases to 40% at 1 µm particle size and it will continue to grow at larger sizes. The overwhelming breathing resistance of

168 respirators and masks are in the 30 Pa to 100 Pa range, and Mask H is extremely low with only 5.4 Pa. The breathing resistance of all tested respirators and masks are in the acceptable range for normal breath.

To compare the common material to the commercial respirators and masks, the

3M™ 8210 N95, 3M™ 1820 procedure mask, and Mask J will be plotted as references in common material fractional filtration efficiency and breathing resistance figures (Figures

9–3 to 9–5 shown below). They represent the N95 level respirators, electret media masks, and non-electret media masks.

Figure 9–2. Fractional filtration efficiency and breathing resistance of commercial respirators and masks

169 9.3.2 Furnace Filters

Furnace filters are common material that are designed to filtering the dust, pet dander, allergen, bacteria, and even virus in the room air. In this study, we collected four types of furnace filter: Filtrete™ MPR 1900, 2200, and 2800, and Arm & Hammer™ Air

Filter. All four furnace filters are pleated and marketed as electret media filter. Filtrete™

MPR 1900 and 2200 are rated as MERV 13, this could be confirmed from their fractional filtration efficiency curves in Figure 9–3. Filtrete™ MPR 2800 is rated as MERV 14 and has the highest efficiency (ASHRAE 2017). For both Filtrete™ MPR 1900, 2200, and 2800, their fractional efficiencies are higher than that of the 3M™ 1820 procedure mask, a mask that is commonly used in hospital settings. As designed for handling a large amount of air circulation in residential or commercial buildings, the breathing resistances of single layer

Filtrete™ MPR 1900, 2200, and 2800 are significantly lower than the reference respirator/mask as shown in the bar chart in Figure 9–3. Multiple-layer performance of collected furnace filters was also evaluated besides the single-layer measurements. The 5- layer Filtrete™ MPR 1900, 2200, and 2800 perform comparably to 3M™ 8210 N95 respirators in terms of efficiency and breathing resistance. It can be found that the pressure drop of 5-layer media is about 10 times of the single layer media, with the reason being the thickness of the stacked multi-layer filter media is reduced due to the pressing seal nature of the filter holder. The stacked fluffy media is compressed so that its permeability is reduced. The fractional filtration efficiency curve of 2-layer Arm & Hammer™ Air Filter is slightly lower than that of the 3M™ 1820 procedure mask, while much better than that of the Mask J. Its 1 µm particle efficiency is above 80% and the efficiency could be higher

170 for even larger particle size. Its breathing resistance is only 10.3 Pa, about 1/6 of the 3M™

1820 procedure mask.

100% 3M™ Filtrete™ MPR 1900 - 5 layers N95 Filtrete™ MPR 2200 - 5 layers 90% Filtrete™ MPR 2800 - 5 layers Filtrete™ MPR 1900 Filtrete™ MPR 2200 80% 3M™ Filtrete™ MPR 2800 1820 ARM & HAMMER™ Air Filter - 2 layers P-Mask 3M™ 8210 N95 70% 3M™ 1820 Procedure Mask Mask J 60%

e i 50% Mask J

f Breathing Resistance

f 100

)

(

s 40% / 80 75.7 m 73.1

c 68.7

0 61.3

1 60 56.9

@ 50.2 30% e

s s 40

20% n

e 20

f 9.8 10.3

D 4.5 5.0 0 k 10% 0 0 0 5 5 5 2 5 s J 0 0 0 × × × × 9 a k 9 2 8 r N s 1 2 2 0 0 0 e 0 M a R R R 0 0 0 lt 1 0 M 9 2 8 i 2 2 P P P 1 2 2 F 8 8 M M M R R R ir 1 P P P A ™ ™ ™ ™ M M M M ™ te te te ™ 3 M re re re ™ ™ ™ R 3 lt lt lt e e e E 0% i i i t t t M F F F re re re ilt ilt ilt M F F F A 0.02 0.05 0.1 0.2 0.5 1 H & M R Particle Mobility Diameter (µm) A Figure 9–3. Fractional filtration efficiency and breathing resistance of furnace filters

9.3.3 Vacuum Cleaner Filters

Vacuum cleaner filters are another common air filter. We tested three grades of pleated vacuum cleaner filters and one vacuum cleaner filter bag. The fractional efficiency and breathing resistance are shown in Figure 9–4. The Ridgid® 5-Layer Allergen Vacuum

Cleaner Filter could offer protection even better than N95 respirator. However, its breathing resistance is about 10 times of the N95, which is impractical to be used as an alternative respirator/mask material. If it was molded as a tight-fitting respirator, just like the commercial N95 mask, the breathing resistance would be twice of the breathing resistance requirement (343 Pa) indicated in 42 CFR Part 84. On the other hand, if this

171 material was used to make a loose-fitting mask, the overall efficacy of the mask will be largely jeopardized since a large portion of the inhaled air would leak through the non- sealed edge rather than flow through the media because of differential pressure balancing.

With a similar fractional filtration efficiency as the procedure mask, the breathing resistance of the Ridgid® Fine Dust Vacuum Cleaner Filter and the Ridgid® High- efficiency Dust Bag is about triple and double of the procedure mask. However, the efficiency of Ridgid® 1-Layer Pleated Paper Filter is even worse than the Mask J, although breathing resistance is slightly lower than that of Mask J.

100% ® 3M™ RIDGID 5-Layer Allergen Filter RIDGID® 3-Layer Fine Dust Filter N95 ® 90% RIDGID High-Efficiency Dust Bag RIDGID® 1-Layer Pleated Paper Filter 3M™ 8210 N95 80% 3M™ 1820 Procedure Mask Mask J 3M™ 70% 1820 P-Mask 60%

c Mask J

e i 50%

f E Breathing Resistance 700 40% 651.8

)

(

600

c 500

5 30% .

1 400

s 300

20% r

210.3

a 200

i t 149.3

f 100 73.1 f 68.7 10% i 57.7 61.3

D 0 r r g r 5 k J e e a e 9 s k y y B y a s a a t a N a -L -L s -L 0 M 5 3 u 1 1 0 M ® ® D ® 2 2 0% ® 8 8 ID ID ID 1 ID ™ G G G M ™ 0.02 0.05 0.1 0.2 0.5 1 ID ID G ID 3 M R R ID R 3 R Particle Mobility Diameter (µm) Figure 9–4. Fractional filtration efficiency and breathing resistance of vacuum cleaner filters

9.3.4 Common Household Materials

We also tested some common household materials to assess their potential as alternative respirator/mask materials. The comparison could offer a filtration efficiency 172 and breathing resistance perspective on candidate common material of homemade masks.

We tested several kinds of material, including sterilization wrap, household fabrics, paper products, which are representatives of materials that have been widely mentioned recently as DIY mask materials. The results are shown in Figure 9–5.

We tested two sterilization wraps that are designed for sterilized instrument wrapping, Halyard H600 and CardinalHealth™ CH600. The sterilization wraps were designed for blocking bacteria while allowing gas penetration. They are composed of multiple layers of spunbond and meltblown materials; both are widely used as air filtration media. Therefore, they share a similar structure with air filters, which helps in particle filtration. Electric charges were applied on Halyard H600 to enhance particle capture capabilities, according to the manufacturer, which is one of the reasons that the fractional filtration efficiency of Halyard H600 is higher than that of CardinalHealth™ CH600 which does not have electric charges according to our tests. The breathing resistance of Halyard

H600 and CardinalHealth™ CH600 are about 2 and 3 times of that for 3M™ 1820 procedure mask, respectively.

Four common fabrics were tested to offer the scientific support in material selections: 2-layer Thinsulate™, and 5-layer bed sheets, T-shirts, and Swiffer® Sweeper®.

Their fractional filtration efficiency curves are comparable, and MPPSs are around 0.3 µm with 50% efficiency at MPPSs. Thinsulate™ is intended to offer insulation while keeping the moisture penetration, which is achieved by highly porous microfibers, an essential microstructure of air filters. The efficiency of 2-layer Thinsulate™, therefore, could be comparable to that of 5-layer the other three materials. Swiffer® Sweeper® is electrostatically charged so that it can pick up and retain small dust and debris when used

173 to swipe off household surfaces. It filters out particles like an electret media, but the charge level on it is not as strong as commercial electret filters (e.g., respirators, face masks, furnace filters, etc.) so that its MPPS is still 0.3 µm, which is also partially attributed to its larger fiber size than most commercial electret filtration materials. 5-layer bedsheets and

T-shirts tested in this study have similar efficiencies, which are ~ 50% efficient at 300 nm, suggesting that stacking sufficient layers of similar fabric materials are necessary to ensure desired respiratory protection to the wearer, while simply pulling the collar band above the nose is not recommended. From the breathing resistance perspective, the breathing resistance of Thinsulate™ and Swiffer® Sweeper® is only about 75% and 50% of that for

3M™ 1820 procedure mask, respectively. Their efficiency could be further improved by adding more layers, where the increased material thickness can be a concern. Additionally, known as an excellent thermal insulator, the advantage of Thinsulate™ in its original duty now pose a disadvantage for it being used as a face mask material. Exhaling through mouth and nose is one of the major routes for the human to expel the body heat, where good thermal insulation is not welcome. As the easiest accessible materials in a household, 5- layer bedsheets and T-shirts tested in this study have significantly higher breathing resistance than their procedure mask reference, making them less recommended from breathability and facial-leak points of view. Paper products are also discussed widely as the candidates for alternative masks. The efficiencies of a 5-layer shop towel and kitchen towel are in between of the 3M™ 1820 procedure mask and Mask J. Again, the multiple- layered configuration is the key for adequate self-protection. One of the concerns of using paper products is their breathing resistance, which is high from the measurement and may increase further after gradually picking up moisture from human exhalation. Coffee filters,

174 though the only material in this group with a “filter” in its name, is the last material to be considered as an air-filtering face mask, owing to its combination of exceptionally low efficiency and higher-than-average breathing resistance.

100% 3M™ Halyard H600 Sterilization Wrap N95 Shop Towel - 5 layers 90% Thinsulate™ - 2 layers T-Shirt - 5 layers Swiffer® Sweeper® - 5 layers 80% Bedsheet - 5 layers Kitchen Towel - 5 layers 3M™ CardinalHealth™ CH600 70% 1820 Sterilization Wrap P-Mask Coffee Filter - 2 layers 60% 3M™ 8210 N95 3M™ 1820 Procedure Mask

c Mask J

e i 50% Mask J

E Breathing Resistance 500

) 40% a 433.9

(

s / 400

1 30% 300

r 231.1

s s 200 185.8 187.9

e r 158.9 P 151.3 153.4 20%

e 100 r 73.1 68.7 e 61.3

f 48.2 i 27.4 10% D 0 0 5 2 5 5 5 5 0 2 5 sk J 0 × × × × × × 0 × 9 a k 6 l t ® t l 6 r N s H e ™ ir r e e H e 0 M a d w e h e e w C lt 1 0 M r o t s p h o i 2 2 a T la - e s T ™ F 8 8 ly u T e d e 1 a p s e n lth e 0% o n w B e a ff ™ H h i ® S h o M ™ S h tc e 3 M T r i lH C 3 fe K a 0.02 0.05 0.1 0.2 0.5 1 if in w rd S a Particle Mobility Diameter (µm) C Figure 9–5. Fractional filtration efficiency and breathing resistance of common fabrics

9.3.5 Discussion of The Figure of Merit

The above discussions about fractional filtration efficiency and breathing resistance of common materials recommended the candidates for the alternative mask materials.

However, it is still hard to compare two materials if one has high efficiency, high breathing resistance, and the other one has low efficiency and low breathing resistance. The figure of merit is therefore introduced to compare two materials under the same standards. The figure of merit (Brown 1993) of filter material can be calculated as 175 ln(1 − 퐸) 퐹푂푀 = − ∆푃

where E is the filter efficiency and ΔP is the differential pressure. The figure of merit represents the level of particle capturing a filter media can achieve per unit of flow resistance penalty, which accounts for the efficiency and breathing resistance (differential pressure) concurrently, so that it can be used to compare the efficiency-resistance tradeoff among all the materials tested. Figure 9–6 shows the figure of merit of the selected materials.

The first tier consists of four materials carrying electrostatic charges. The figure of merit of Filtrete™ MPR2200 and MPR 2800 are even higher than 3M™8210 N95. It should be noticed that Filtrete™ MPR2800 ranked higher than Filtrete™ MPR2200 in terms of efficiency, but its figure of merit is slightly lower than that of Filtrete™ MPR2200 due to its higher breathing resistance. The other two materials, Swiffer® Sweeper® (5 layers) and Thinsulate™ (2 layers), have a similar figure of merit as 3M™ 1820 procedure mask.

All four first-tier materials are the most-recommended candidates from the filtration point of view. The second tier includes another four materials: Halyard H600 sterilization wrap,

RIDGID® 3-layer Fine Dust Filter, and T-Shirt (5 layers) have a similar figure of merit, which is comparable to that of the Mask J, a non-electret commercial mask. These results show that the quality of those three common materials as good as the Mask J. It could be found that the efficiencies of the above three materials are much higher than that of the

Mask J, at least 30 percentage points (5-layer T-shirt) at MPPS. Therefore, the comparable figure of merit to Mask J is mainly because of the high breathing resistance of three common materials. The figure of merit of 5-layer bedsheet is lower than the other three,

176 the likely reason is that the breathing resistance of the bedsheet we tested was about double that of the T-shirt. The inadvisable material is coffee filters, which is ranked lowest in both efficiency plot and figure of merit plot. Even it is called coffee filter, it is designed to separate the ground coffee from water, having hundreds of times larger size than the SARS-

CoV-2 virus or other air-borne pathogens a mask needs to filter out. Although counter- intuitive at first glance, it needs to be recognized that furnace filters and vacuum filters may work well as alternative mask materials, but coffee filters do not work well.

Filtrete™ MPR 2200 Filtrete™ MPR 2800 Swiffer® Sweeper® - 5 layers Thinsulate™ - 2 layers Halyard H600 Sterilization Wrap RIDGID® 3-Layer Fine Dust Filter 10−1 3M™ T-Shirt - 5 layers N95 Bedsheet - 5 layers Coffee Filter - 2 layers 3M™ 3M™ 8210 N95

)

1 - 1820 3M™ 1820 Procedure Mask

( P-Mask Mask J

e −2 r 10

F Mask J

10−3

0.02 0.05 0.1 0.2 0.5 1 Particle Mobility Diameter (µm) Figure 9–6. The figure of merit of common materials

9.3.6 Some Discussions

Some of the recommended materials can be cut and sewn to a mask, such as T- shirts, bedsheets. However, some material is hard to be processed in the same way, such as 177 furnace filters, Swiffer® Sweeper®. A combination of a cloth mask and a filter material insert is therefore recommended. A cloth mask could be made following various instructions online (CDC 2020d). The face-covering part can be made into a double-layer design with a slit on the inner side, where a filter material can be cut and inserted to boost the filtration efficiency. This makes the use of those difficult-to-sew materials, such as furnace filters, possible and it can also reduce the material consumption by using small pieces only covering the size of the insert pocket.

This study evaluated common materials that have the potential to be made into alternative masks for the general public under the current COVID-19 pandemic as an emergency option. It should be noticed that this study does not endorse the safety of using the studied material as an alternative mask. For example, fiber shedding could be potential harm to wearer, especially people who have underlying respiratory conditions. Many of the tested materials in this study were not intended for respiratory protection so that potential fiber shedding was not tested by their manufacturers or certification agencies.

With the tide of DIY masks using alternative materials, some manufacturers announced the notice regarding altering their products to mask use. For example, 3M does not recommend altering the furnace filter to other usages (3M 2020); Halyard claims that they cannot recommend the off-label use of the sterilization wrap (Halyard 2020); Ridgid has a similar notice that their vacuum filter and vacuum filter bag is not designed as a respiratory protection material (RIDGID 2020).

In this study, different commercial respirators, masks, and common materials were all tested with a sealed filter holder and the same face velocity, to ensure a fair performance comparison of the materials themselves. When a filter material is made into a respirator or

178 mask, even the tight-fitting respirator design experiences various degrees of a facial leak, due to design faults, the variation of wear’s facial profiles, and improper use. Such leak is more substantial on loose-fitting face mask designs, which most DIY masks are. Face leak causes a certain portion of air inhaled not filtered by mask materials, resulting in lower overall efficacy than the material efficiency as reported in this study, the degree of which varies largely among mask shape/style designs and material’s breathing resistance For similar mask designs, the pressure-driven face leak is severer on masks made of materials with higher resistance, which emphasizes another reason why breathing resistance is treated equally important to filtration efficiency in this study, other than the consideration of breathability and wearer’s comfort. The figure of merit is a well-accepted indicator of efficiency-resistance balance in filtration society, but its correlation with the overall efficacy of a face mask (with face leak counted) is not fundamentally or experimentally studied. Although a comprehensive assessment of overall protection efficacy of various mask materials and designs is not an easy task, given the large variations from various factors, not to mention the fit of a respirator/mask itself is subject dependent, a case study showing the comparison of material efficiency and overall efficacy, and its association with breathing resistance and other factors, it is urgently needed under current pandemic.

Since the particles used in this study can not only represent the virus-laden particles but also, more importantly, the particulate air pollutant. The most concerning particulate pollutant is PM2.5 whose aerodynamic particle size is less than 2.5 µm. The particle size studied in this study falls into this size range. Hence, the conclusion of particle filtration efficiency and breathability of different common materials could also be applied when the general public DIY their masks to prevent particulate air pollutants. This is valuable for

179 the countries suffering from air pollutions, especially developing countries with limited access to commercial masks. Under the current pandemic, the SARS-CoV-2 laden particle could cause acute respiratory disease, while the particulate air pollutant can also cause chronic respiratory diseases. According to many epidemiologic studies, the positive association between respiratory diseases and atmospheric pollutants has been proven. It is equally important for the general public to be protected from both the pathogens and ambient pollutants. The conclusions of this study can guide the general public when the commercial respiratory protection equipment is unavailable under a pandemic or a heavy air pollution episode.

9.4 Conclusions In this study, we tested commercial respirators and masks, furnace filters, vacuum cleaner filters, and common household materials. We evaluated the materials’ fractional filtration efficiency and breathing resistance, which are primary factors affecting respiratory protection. Neutralized monodisperse NaCl particles were used to measure the fractional efficiency. Breathing resistances were also compared at the same face velocity.

To compare the efficiency-resistance tradeoff, the figure of merit of each tested common material was also calculated. Filter media with electrostatic charges (electret) is recommended due to its high efficiency with low flow resistance; multiple-layer household fabrics and sterilization wraps are acceptable materials; a coffee filter is inadvisable due to its low efficiency. The outcome of this study can not only offer guidance for the general public under the current pandemic but also suggest the appropriate alternative respiratory protection materials under heavy air pollution episodes.

180 Chapter 10: Evaluation of Decontamination Methods for Commercial and Alternative Respirator and Mask Materials – View from Filtration Aspect

This chapter has been published: Ou, Q.1, Pei, C.1, Kim, S. C., Abell, E. and Pui, D. Y. H. (2020). Evaluation of decontamination methods for commercial and alternative respirator and mask materials – view from filtration aspect Journal of Aerosol Science 150: 105609. doi:10.1016/j.jaerosci.2020.105609 10.1 Introduction

The COVID-19 pandemic has led to a severe shortage of personal protection equipment (PPE) worldwide (WHO 2020e). According to World Health

Organization(2020c), the novel coronavirus (SARS-CoV-2) is transmitted primarily through human respiratory droplets generated by coughing or sneezing(Leung et al. 2020b,

National Academies of Sciences Engineering Medicine 2020), either through direct contact when virus-containing droplets reach the nose, mouth, eyes of a person, or indirectly when a person touches virus-contaminated objects or surfaces and then touches his/her eyes, nose, or mouth. A respiratory protection equipment is therefore essential for frontline healthcare workers who may have direct contact with patients carrying COVID-19 virus. Nevertheless, the scarce inventory of N95 respirators and surgical masks in hospitals has been reported

(Klompas et al. 2020). In light of the urgent needs and the exhausted national stockpile of

PPEs, although not designed for, the possible decontamination and reuse of disposable filtering facepiece respirators (FFRs) becomes necessary (CDC 2020a).

While tight-fitting respirators provide better protection against airborne sub-micron particles (Ollier et al. 2019, Zhou and Cheng 2017), they are generally recommended to be reserved for healthcare workers during current pandemic (WHO 2020d). Loose-fitting

181 masks, such as surgical masks and procedure masks, are designed mainly to be used as one-way protection, capturing large particles or droplets exiting from the wearer and preventing them from being spread to the environment. Surgical masks often have a fluid- resistant outer layer, which also protect the wearer against large droplets, splashes, or spray of bodily or other hazardous fluids. While being reported effective in preventing spread of influenza virus (Johnson et al. 2009), these loose-fitting masks are not typically considered as proper respiratory protection, since they do not provide the wearer with a reliable level of protection from inhaling small airborne particles, which can either penetrate through mask materials (insufficient capturing efficiency) (Chen and Willeke 1992) or leak through the gap between mask perimeter and wearer’s face (loose-fitting) (Oberg and Brosseau

2008). Nonetheless, due to the severe shortage of FFRs and the recommendations of general publics to wear face-coverings when other social distancing measures are difficult to maintain (Feng et al. 2020, WHO 2020d), loose-fitting face masks now become a major measure of respiratory protection for general publics, first responder workers, and even some frontline healthcare workers (mainly surgical masks in some low risk settings). While performance of respirators are well evaluated and many studies on respirator decontamination methods are reported (CDC 2020a, N95DECON 2020), the information of filtration performance and possible decontamination and reuse strategy for face masks is scarce.

SARS-CoV-2 is a large size virus with a total size of ~ 0.06 to 0.14 µm (Lim et al.

2016, Zhu et al. 2020). Human exhalation droplets can be several micrometers or larger

(Anfinrud et al. 2020), but shrink while traveling in air due to water evaporation. Dried droplet nuclei are usually less than 5 µm in diameter (Hsiao et al. 2020), and therefore can

182 remain suspended in air for a longer period of time, and travel a further distance away (Yu et al. 2004). Efficiency of filter media is a strong function of particle size and filtration velocity, as various filtration mechanisms (e.g. diffusion, interception, impaction, electrostatic capturing, etc.) work effectively at different size and velocity (Brown 1993,

Hinds 2012, Ou et al. 2017), which yields a least efficient size, namely most-penetrating particle size (MPPS). Under the range of normal operation conditions, most mechanical filters have a MPPS between 100 and 300 nm, while electret filters, which most commercial respirators and face masks are made of, have smaller MPPS typically below

80 nm. In United States, respirators are approved by National Institute for Occupational

Safety and Health (NIOSH), by testing well-sealed respirator at a flowrate of 85 L/min using charge neutralized polydisperse sodium chloride (NaCl) particles with a count median diameter (CMD) of 0.075 ± 0.02 μm and a geometric standard deviation (GSD) of less than 1.86 (NIOSH 2019). A mass integrated efficiency is then determined by a light scattering based photometer. Surgical masks, differently, are cleared by the Food and Drug

Administration (FDA), by reviewing data supplied by the manufacturers that is collected from third-part independent testing laboratories. A wide range of testing flowrate or face velocity is implemented for testing face masks, and different sometimes confusing efficiencies (e.g., PFE, BFE, VFE, etc.) are reported (Rengasamy et al. 2017). In view of the lack of a universal standard for testing respirators and various face masks side-by-side, in our work, fractional efficiency measurement was conducted which provides efficiency at a series of particle sizes from 30 to 400 nm, covering the MPPSs of all samples tested.

Various disinfection methods have been tested for decontaminating filtering facepiece respirators, and a comprehensive review of these studies is available at

183 N95DECON website, a scientific consortium for data-driven study of N95 filtering facepiece respirator decontamination (N95DECON 2020). Among all these methods, US

Center for Disease Control and Prevention (CDC) (2020a) has identified three most promising ones: vaporous hydrogen peroxide (VHP) (Battelle 2016, Bergman et al. 2018), ultraviolet germicidal irradiation (UVGI) (Lindsley et al. 2015, Viscusi et al. 2009), and moist heat (Bergman et al. 2018). Steam treatment (Fisher et al. 2011, Heimbuch et al.

2011) and liquid hydrogen peroxide (Bergman et al. 2018) are listed as promising methods with some limitations. All these methods are evaluated for respirators only and are mainly targeted for reuse by healthcare workers. In our study, we attempted to extend the scope a little further to include the decontamination of loose-fitting face masks and some alternative materials that have been used for making respirators/masks under this crisis, and the reuse purpose for both healthcare workers and general publics. Because of that, among all applicable decontamination methods being reported to be efficient in deactivating COVID-19 and other pathogens, we excluded methods that require sophisticated equipment and professions (e.g. VHP), and chose UVGI in view of its wide availability in hospitals and small clinics and the heat treatments (both dry oven heat and water steaming) for simple household applications. Filtration performance were evaluated to assess if these decontamination methods can be implemented without compromising the function of respirators/masks. Although it has been well reported that isopropanol (IPA) cannot be used to disinfect respirators as it removes charges so that lead to filtration deterioration, its possible application for disinfecting non-electret alternative respirator material was evaluated in this study.

184 10.2 Experimental Method and Material 10.2.1 Decontamination Methods

For the ultraviolet germicidal irradiation (UVGI) method, the respirator/mask samples were hung between a Xenex LightStrike Germ-Zapping Robots and a UV reflective wall, both < 1m away from the samples. The UV-C light from the source machine and the reflected illumination irradiated the samples for 5 minutes. For the oven heating method, 77oC (170oF) was chosen because it is the lowest temperature setting for most household ovens, and Covid-19 virus is reported deactivated at 70oC (Chin et al. 2020). At the target temperature, the respirators/masks were placed on a stack of coffee filters inside the oven without touching any metal surfaces to prevent thermal damage and heated for 30 minutes. For the steam heat treatment method, the respirators/masks were placed on the rack of a steamer with boiling water for 30 minutes. For isopropanol (IPA) soaking, test samples were soaked in 75% IPA-water solution for 10 min and let dry overnight.

10.2.2 Filtration Performance Testing

A round piece of filtration materials with 40 mm diameter was cut from each sample and secured in a stainless-steel filter holder with gaskets sealing around the edge on both sides to prevent air leakage. The breathing resistance of the materials was determined by measuring the differential pressure drop across test samples under testing filtration velocity of 10.5 cm/s, which is equivalent to 85 L/min across a filtration area of

135 cm2. 85 L/min is the standard flowrate at which NIOSH certifies N95s, and 135 cm2 is the lower end of filtration surface area of typical commercial N95 respirators. The choice of lower end value is to ensure our measurement represents the worst scenario, as filtration

185 performance deteriorates with increasing filtration velocity. As a reference, 3M 8210 N95 respirator has an effective surface area of 150 to 170 cm2, depending on how it is sealed for standard NIOSH N95 certification test.

For particle filtering efficiency, a method called fractional efficiency measurement was used. As depicted in Figure 9–1, NaCl nanoparticles generated from a Collision-type atomizer (TSI 3076, Shoreview, MN) were employed as challenging aerosol for fractional efficiency measurement. The test particles were first passed through a diffusion dryer to remove excess moisture from atomizer solution and bring the humidity level of the aerosol below deliquescence relative humidity of NaCl. They were then sent into a Po-210

Staticmaster® Alpha Ionizer (0.5 mCi, NRD, LLC. Grand Island, NY) neutralizer and a

DMA, which act as a size selector and only allow NaCl particles with a narrow mobility range to exit. To save time, the NaCl solution concentration was not changed during the measurement, but the number mode was kept at slightly below 65 nm to minimize the multiple charge effect in mobility classification for large particles. The size-selected particles were then passed through another Po-210 neutralizer (0.5 mCi) to bring the charge level of these charged particles to a Boltzmann equilibrium charge distribution (Liu and

Pui 1974, Wiedensohler 1988), a steady state distribution best representing the charge level of ambient aerosols. This second neutralizer is particularly important for testing respirators and masks, as most commercial units employ electret media which show much higher efficiency on small (<100 nm) charged particles. The importance of testing with neutralized

(Boltzmann equilibrium) aerosol was overlooked by a recent publication with similar topics (Konda et al. 2020), in which un-conditioned aerosols from a mechanical nebulizer was employed, whose inherent high charge level would cause overestimation of filtration

186 efficiency. The neutralized test particles were then mixed with dilution air and sent into filter holder, whose number concentrations upstream and downstream of test sample were measured by a Condensation Particle Counter (TSI 3776, Shoreview, MN) with the assistance of a pneumatic three-way switching valve. The fractional efficiency (FE) of the

퐶푑푛 퐶푏푙푎푛푘,푑푛 sample was determined as 퐹퐸 = 1 − / , where 퐶푢푝 and 퐶푑푛 are the number 퐶푢푝 퐶푏푙푎푛푘,푢푝 concentrations of NaCl particles at upstream and downstream of the sample, and 퐶푏푙푎푛푘,푢푝 and 퐶푏푙푎푛푘,푑푛 are the number concentrations of NaCl particles at upstream and downstream of the filter holder when no sample presents. The calculated FE accounts for particle transportation loss in the test system. FE was determined for 8 sizes spanning from 30 to

400 nm, which covers the MPPS of both electret and mechanical filtration materials.

10.3 Results and Discussion

Three commercially available respirator and mask samples were selected as the major data set of this study, namely 3M 8210 N95 respirator, Halyard 48207 surgical mask, and 3M 1820 procedure mask. Additionally, 2 other materials that are reported being used for making respirators/masks under Covid-19 pandemic were selected. One is Halyard

H600 one-step fabric sterilization wrap, which was first proposed as alternative mask material by a team at University of Florida who has a pending patent on this (University of

Florida 2020). The other material, named as EX101, was donated by Cummins Inc. to

University of Minnesota as part of a project to supply masks to Minnesota’s M Health

Fairview network (Businesswire 2020, University of Minnesota 2020).

187 All the 3 commercial respirator and masks and Halyard H600 fabrics have electrostatic charges on their fiber surface, which help to boost filtration efficiency without causing extra breathing resistance. However, these charges may be completely or partially removed by some sterilization methods, which could negatively impact the selection of decontamination and reuse strategy for these materials. EX101 is a pure mechanical filtration material with no electric charge on it. It was made of polymer fibers: PBT polyester, Nylon 6-6, and PET polyester in different layers.

10.3.1 Base Performance Before Decontamination

Filtration efficiency and breathing resistance of all the five materials are measured first without any decontamination treatment was conducted, which is summarized in Figure

10–1(a). Among all five materials, N95 respirator material and EX101 media have the highest level of efficiency, greater than 95% over the entire size range between 0.03 to 1.0

µm. All three commercial respirator and masks show similar trends of increasing efficiency with particle size, suggesting they all possess high level of electrical charge on their fiber surface. Unlike N95 respirator having efficiency greater than 95%, surgical mask and procedure mask tested have efficiency around 80-90% and 70-80% respectively in the range of 0.03 to 0.4 µm. Their efficiency increased sharply after 0.4 µm, to greater than

95% at 1.0 µm, attributed to the improved capture by interception and inertia impaction.

Since the efficiency keeps increasing and stays high after 0.4 µm for all 5 materials, only the results between 0.03 and 0.4 µm are reported in the rest part of the paper. EX101 has a

MPPS at ~ 0.1 µm, which is typical for pure mechanical filter with small fiber diameter at

188 this face velocity (10.5 cm/s). The shortcoming of achieving N95 level of efficiency without an assistance from electrostatic charge is the highest breathing resistance (323 Pa

@ 10.5 cm/s) among all, which though is still below NIOSH’s limit of 343 Pa for an N95 respirator. This breathing resistance can be further reduced by making masks with larger effective filtration area. H600 sterilization wrap has the lowest efficiency and the second highest breathing resistance among all. It is a surprise that its efficiency is only marginally lower than 3M procedure mask, given that this material is not originally designed as a filtration material, which is mainly because it has certain level of electrical charge on its fibers. This is known by comparing its filtration with that of a 75% isopropanol-soaked sample as discussed later. It is worth mentioning that the sterilization wrap is more plastic than most non-woven face masks, potentially providing better facial fit than loose-fitting masks, although this advantage is partially offset by its higher breathing resistance (causing more pressure driven facial leak). All the three commercial respirator and masks have flow resistance around 50-75 Pa, which agrees with previous measurement (Kim et al. 2015). A recent study (Konda et al. 2020) measured significantly lower flow resistance at as low as

10-15 Pa may suggest leaks in their test apparatus.

For a better comparison among all the five materials, their quality factor (Q) is determined by

ln (1−퐸푓푓) Q = − , ∆푃 where Eff is the fractional efficiency at certain size, while ΔP is the differential pressure drop (breathing resistance). Quality factor represents a tradeoff between efficiency and resistance of a filtration material under a specified operating condition (e.g., face velocity, temperature, etc.), with a higher value preferable. It is clearly seen from Figure 10–1(b), 189 that 3M 8210 respirator has much higher quality factor than the other materials as it is one of the highest efficient and the least resistant material among all five tested. With the assistance of electrostatic capturing, both surgical mask and procedure mask have decent quality factor that is higher than typical filtration materials at this face velocity. As a pure mechanical material, the quality factor of EX101 media is among the top tier and is even slightly higher than H600 sterilization wrap which has electrical charge on it. The non- electret nature of EX101 makes it more tolerant to various decontamination methods, which will be demonstrated later.

100% 3M™ 8210 Halyard 48207 0.1 95% Surgical Mask 3M™ 1820 Procedure Mask 90% Halyard H600

]

Sterilization Wrap 1 - 0.05

] EX101

[

85% i

f i Breathing Resistance

400

]

f 80%

[ E 350 323.0

i c 300

0 0.02

250

75% @

r 200

u 159.0

r 150

t 100

n 73.0 72.7

70% e

r 51.2

f 50

D 0 0.01 0 1 1 7 0 0 0 (a) 2 0 k 2 k 0 1 (b) 8 2 s 8 s 6 p 8 a 1 a H ra X 4 E ™ d M ™ M rd W 65% M r l e a n 3 a a M r y o ly ic 3 u l ti g d a a a r e H z 0.1 1 H u c ili 0.1 1 S ro r P te Particle Mobility Diameter [µm] S Particle Mobility Diameter [µm] Figure 10–1. Fractional efficiency (a), breathing resistance, and quality factor (b) of 3M 8210 respirator, Halyard 48207 surgical mask, 3M 1820 procedure mask, Halyard H600 sterilization wrap, and Cummins EX101 material, before any decontamination treatments

10.3.2 Decontamination by Isopropanol Alcohol

It is well reported that isopropanol alcohol and soap can cause filtration degradation of N95 respirators (Viscusi et al. 2009), and the purpose of this set of test is mainly to confirm if similar degradations can be found on other face mask materials and alternative respirator/mask materials. As shown in Figure 10–2, after only one treatment, significant 190 decay in efficiency was found in all the materials except for EX101 which is a non-electret media. For all four electret media, the degree of efficiency degradation increases with particle size, resulting in MPPS shift towards larger size from below 30 nm (except for sterilization wrap at 100 nm) before the treatment to 150-300 nm afterwards. The shift of

MPPS towards larger size is a strong indication that the drop in efficiency level is due to charge loss rather than structural change or integrity degradation. Another evidence of this conclusion is the unchanged breathing resistance level, which is also shown in Figure 10–

2. Exposure to organic solvents such as isopropanol, acetone, xylene, toluene, and benzene has been reported to increase charge mobility in polypropylene fibers, subsequently reducing the electrostatic charge on the fibers and their ability of capturing particles (Jasper et al. 2006, Kim et al. 2009a).

Although not as significant as others, EX101 also shows a slight drop in efficiency with an unchanged MPPS at 100 nm. At this size, it is 90.5% efficient after treatment comparing to 95.7% of untreated media. The change is likely from the loss of small fibers when IPA soaks into the material’s micro-structures or evaporates from its fiber surface. It is worth noting that although the IPA soaked EX101 sample has an efficiency below 95% at MPPS (100 nm), it may still pass N95 certification for two reasons: (1) photometer based

NIOSH N95 certification test puts more weight on the efficiency at ~ 300 nm; and (2) the sample would have been tested under a low face velocity if a respirator was made with greater than 135 cm2 effective filtration area, which is the case for most commercial respirators. For further justification, an integrated mass efficiency was calculated by assuming a lognormal test aerosol distribution with a count median diameter at 75 nm and

191 a geometric standard deviation of 1.86. The calculated mass weighted efficiency was

95.18%.

Instead of soaking, a gentler treatment was then implemented by spraying 75% isopropanol-water solution onto media surface using a small garden sprayer. Almost identical level of efficiency degradations was found on all four electret materials, suggesting that the small amount of IPA droplets generated is sufficient to almost fully discharge all electrical charges on electret fiber surface. This result suggests that attentions should be paid to avoid direct contact of alcohol liquid or vapor with respirators and face masks, especially in hospitals and clinics where alcohol or other disinfection agents are used routinely. One example of that is adjusting/touching respirators/masks immediately after using hand sanitizer. Unlike IPA soaked case, no efficiency degradation was observed on IPA sprayed EX101 sample, indicating the nanofibers are strong enough to retain their structure when contacting with small amount of IPA droplets.

192 100% 3M™ 8210 N95 3M™ 8210 N95 IPA Soak 90% 3M™ 8210 N95 IPA Spray Halyard 48207 Surgical Mask Halyard 48207 Surgical Mask IPA Soak 80% Halyard 48207 Surgical Mask IPA Spary 3M™ 1820 Procedure Mask 70% 3M™ 1820 Procedure Mask IPA Soak 3M™ 1820 Procedure Mask IPA Spray Halyard H600 Sterilization Wrap

] 60%

- Halyard H600 Sterilization Wrap IPA Soak

[

y Halyard H600 Sterilization Wrap IPA Spary

n 50% EX101

i EX101 IPA Soak

E EX101 IPA Spray 40%

]

[ Breathing Resistance

s / 400

m Untreated 30% c 338 323

5 IPA Soak 312

0 300

1 IPA Spray

e 20% r 200 159 u 148 146

e r 100 81 82 P 73 73 74 66 l 51 59 49

a 10% i

e 0

f 5 1

f 7 0 0 i 9 0 2 0 0 N 2 k 8 k 6 p 1 D 8 s 1 s a X 0 4 a a H r E 0% 1 M M d W 2 rd l ™ l r 8 a a M a a n y ic 3 ic ly io 0.02 0.05 0.1 0.2 0.5 ™ l g g a t M a r r H a 3 H u u liz S S ri Particle Mobility Diameter [µm] te S Figure 10–2. Fractional efficiency and breathing resistance of the 5 filter materials before and after IPA treatments (soaking or spraying)

10.3.3 Decontamination by UVGI

Within the spectrum of ultraviolet light (10-400 nm wavelength), ultraviolet-C

(100-280 nm) has the highest disinfection efficacy as UV-C is absorbed by DNAs and

RNAs of microbes and viruses which in turn induces changes in DNA/RNA structures(Lindblad et al. 2020). It is reported that it is more difficult to disinfect virus on respirators than on a flat object surface, and a UV-C dose of 1000 mJ/cm2 is necessary to disinfect (≥3-log inactivation) viruses similar to SARS-CoV-2 on N95 FFRs (Mills et al.

2018). The exact UV-C dose in one treatment cycle in this study is not available, but it is estimated to be sufficiently higher than 1000 mJ/cm2. As shown in Figure 10–3, no systematic efficiency drop was found on either of three commercial respirators or face

193 masks after up to 10 cycles. Since the filtration performance testing in this study is a destructive method, to save precious PPEs, only one sample was prepared for each decontamination method and treatment cycle combination. The small fluctuations (within

2% for N95s, 11% for surgical mask, and 8% for procedure mask) observed can be mainly attributed to variation among different pieces of samples. N95s were reported to keep integrity and fit at higher dose of UV-C light (Heimbuch et al. 2011, Lindsley et al. 2015).

Despite of sample variation, which is shown as efficiency fluctuation evenly over size range of 30-400 nm, no efficiency decay associated with charge loss can be identified, which would otherwise show higher efficiency reduction in larger sizes. This results are in agreement with previous experiments (Bergman et al. 2018, Viscusi et al. 2009, Viscusi et al. 2011), suggesting UVGI can be an option to decontaminate respirators and face masks under the shortage caused by Covid-19 without causing degradation in filtration.

As one of the standard disinfection methods used in hospitals, UVGI generally has larger availability than other recommended respirator decontamination options (e.g., VHP) for hospitals. Question however remains open about UV-C’s ability of penetrating deeply into melt-blown materials, which is made of polypropylene, a UV absorber. The amount of penetration was found to vary largely among N95 models (Fisher and Shaffer 2011).

Higher UV-C dose may be necessary to disinfect other pathogens that are less susceptible.

Both concerns require higher UV-C dose, leading to longer irradiation period. Some studies

(Heimbuch and Harnish 2019, Mills et al. 2018) demonstrated residual virus on respirator straps after UVGI treatment (likely due to the factor that the straps twist so as be shielded from UV-C irradiation), suggesting a subsequent disinfection of straps are necessary. The

194 consideration of UV shading also requires that respirators and masks cannot be stacked during the treatment, which further limits the decontamination throughput of this method.

100% 100% (a) (b) 95% 98% 90%

]

- - 85%

[ 96% [

n n 80%

f 94% f

E E 75% UVGI Treatment UVGI Treatment 3M™ 8210 N95 Untreated 70% Halyard 48207 Surgical Mask Untreated 92% 3M™ 8210 N95 1 cycle Halyard 48207 Surgical Mask 1 cycle 3M™ 8210 N95 3 cycles Halyard 48207 Surgical Mask 3 cycles 65% 3M™ 8210 N95 5 cycles Halyard 48207 Surgical Mask 5 cycles 3M™ 8210 N95 10 cycles Halyard 48207 Surgical Mask 10 cycles 90% 60% 0.02 0.05 0.1 0.2 0.5 0.02 0.05 0.1 0.2 0.5 Particle Mobility Diameter [µm] Particle Mobility Diameter [µm] 100% 120 (c) UVGI Treatment (d) Untreated

3M™ 1820 Procedure Mask Untreated ] 1 cycle 95% a 3M™ 1820 Procedure Mask 1 cycle P

[ 3 cycles

100

3M™ 1820 Procedure Mask 3 cycle / 5 cycles

90% 3M™ 1820 Procedure Mask 5 cycle m c 10 cycles

5 3M™ 1820 Procedure Mask 10 cycle . 80

]

85% 1

[

n 80% r 60

r E 75%

l 40

70% n

f 20

f 65% i

60% 0 0.02 0.05 0.1 0.2 0.5 3M™ 8210 N95 Halyard 48207 3M™ 1820 Particle Mobility Diameter [µm] Surgical Mask Surgical Mask Figure 10–3. Fractional efficiency (a-c) and breathing resistance (d) of N95 respirator, surgical mask, and procedure mask after multiple UVGI treatment cycles

10.3.4 Decontamination by Thermal Treatments

Among all potential decontamination methods, thermal treatment is the most applicable one for the general public to implement in household setting, because of its simple procedure and the high availability of various ways of heat generation at home. For the same reason, methods requiring sophisticated humidity control are not evaluated due 195 to its relatively low feasibility at home, although there is evidence showing that higher humidity will better inactivate viruses similar to SARS-CoV-2 (McDevitt et al. 2010). A dry heat temperature of 77oC (170oF) was chosen as it is the lowest temperature setting of most household ovens, and a recent evidence showed that dry heat at 70oC for 30 min can achieve 1.9-log activity reduction, which further increased to 3.3-log if the treatment duration was extended to 60 min (Fischer et al. 2020). Each treatment cycle was chosen to be 30 min in this study considering a slightly higher temperature was used and the relatively low exposure risk for general publics. As shown in Figure 10–4 (a1-a6), dry heat at 77oC did not cause any observable efficiency degradation of the 3 commercial respirator/mask materials after 10 cycles. Similar to UVGI method, small fluctuations were observed which can be attributed to sample variation rather than any systematic efficiency deterioration for the reason explained above already. A slight change of MPPS from below 30 nm to 50 nm was found on 3M 8210 N95 respirators, implying a subtle loss on charge level which is not sufficient to yield any practically meaningful performance degradation. For sterilization wrap, after the first treatment, ~10-15% of reduction was observed at sizes greater than 100 nm, suggesting certain degree of charge loss occurred at high temperature. Further dry heat treatment for multiple cycles (up to 10 cycles) did not seem to cause further filtration degradation. Similar to sterilization wrap, EX101 experienced slight (<5%) efficiency drop after first dry heat treatment, but further reduction was not observed after multiple cycles

(up to 10). No MPPS change was observed on EX101, indicating the efficiency drop is likely due to small amount of structure change (likely loss of nanofiber).

Although a precise control of humidity in thermal treatment is not an easy implementation in household, steam heat treatment can be a feasible alternative high

196 humidity option which keeps the moisture level near or above saturation. Recent studies demonstrated SARS-CoV-2 (Kumar et al. 2020) and bacteriophage MS2 (Massey et al.

2020) viruses can be deactivated by autoclave or microwave generated steam in shorter treatment time (≤ 15 min). As illustrated in Figure 10–4 (b1-b6), although 3M 8210 N95 respirator retained its efficiency after 10 cycles, efficiency decay was observed on surgical mask (after 5 cycles), procedure mask (after 10 cycles), and sterilization wrap (after 1 cycle). The efficiency reduction is more pronounced at larger sizes, causing shift of MPPSs from below 30 nm to ~ 200 nm, which is a strong indication of charge loss during the treatments. The almost unchanged breathing resistance provided further evidence that change of material integrity or micro-structure is not the major source of efficiency degradation. Polypropylene is hydrophobic, but water can condense on polypropylene fiber surface under super-saturation, which could potentially deteriorate surface charge by an unknown mechanism that requires more study. Accelerated charge degradation on corona- charged polypropylene fibers has been reported when conditioned at higher humidity levels

(Motyl and Łowkis 2006). Similar to dry heat case, steam heat induced small (<5%) efficiency drop on EX101 samples, which is likely attributed to loss of nanofibers.

Besides the filtration efficiency of the material, another important property of a respirator or mask is the fit, which measures the seal between the respirator/mask and the wearer’s face. An annual fit test is required when a healthcare worker needs to wear a tight- fitting respirator, such as a 3M 8210 N95 respirator. The quantitative fit tests were performed using a TSI PortaCount® Pro+ 8038 by a specific researcher in this study. The fit factor, defined as the ratio of ambient particle concentration to the particle concentration inside the respirator, should be equal to or above 100 to pass the test. The fit tests were first

197 performed with a new 3M 8210 N95 respirator and then performed after 1, 3, 5, and 10 cycles of 77oC dry heat treatments with the same respirator. A second 3M 8210 N95 respirator was fit tested after 0, 1, 3, 5, and 10 cycles of steam heat treatment. As shown in

Table 10–1, dry heat treatment deemed safe for the integrity and the fit of the respirator, while the steam heat treatment caused the respirator fail in fit test after 5 treatment cycles.

It is worth noting that although all the fit tests were performed with the same person wearing the respirator in order to minimize variation of wearers, the results reported here should be considered as subject-dependent. Different fit factors should be expected if the tests were performed on a different wearer, even if with the same respirator. The impact of thermal treatment on respirator fit also varies by respirator model (Anderegg et al. 2020,

Bergman et al. 2018, Fischer et al. 2020, Kumar et al. 2020). Loose-fitting face masks were not fit tested in current study. Although a fit test is not required, the overall efficacy of a face mask’s respiratory protection is strongly affected by its ability to minimize facial leak around edges. With more loose-fitting masks are used by frontline healthcare workers as

(sometimes the only) respiratory protection PPE, more study is urgently needed to quantify the fit factor of these face masks over a wide particle size range in order to provide a comprehensive assessment of their respiratory protection efficacy.

In summary, in terms of retaining both filtration efficiency and fit (tested for respirators only), dry heat seems to be a better option over steam heat treatment, while evidences (Fischer et al. 2020, Fisher et al. 2011, Heimbuch and Harnish 2019, Kumar et al. 2020, Lore et al. 2012, McDevitt et al. 2010) suggested steam heat or moist heat (50-

85% RH) disinfect SARS-CoV-2 or similar viruses at shorter exposure time. Respirators and masks can be heat treated in an enclosed rigid heat compatible container with small

198 amount of water added (less than 1 ml for 1.25 quart container size) to help maintain a favorable 60-85% RH (Anderegg et al. 2020). It is suggested to keep the water amount low to avoid steam generation which may negatively impact efficiency and fit. The authors claim this method is scalable, so it can also be potential decontamination options at medium or large quantities after the viral inactivation of SARS-CoV-2 on FFRs under the given conditions can be confirmed.

Table 10–1. Quantitative fit testing results of the new N95 respirator and after oven and steam heat treatment cycles. Oven Treatment Cycles Exercise New 1 3 5 10 Normal Breathing 200+ 200+ 200+ 200+ 200+ Deep Breathing 200+ 200+ 200+ 200+ 200+ Head Side to Side 200+ 200+ 200+ 200+ 200+ Head Up and Down 200+ 200+ 200+ 200+ 200+ Talking 135 134 124 170 125 Grimace Excl. Excl. Excl. Excl. Excl. Bending Over 200+ 200+ 151 197 200+ Normal Breathing 200+ 200+ 200+ 200+ 200+ Overall Fit Factor 188 188 177 195 185

Moist Heat Treatment Cycles Exercise New 1 3 5 10 Normal Breathing 200+ 200+ 191 109 141 Deep Breathing 200+ 200+ 200+ 184 179 Head Side to Side 200+ 200+ 92 43 51 Head Up and Down 200+ 165 101 72 99 Talking 135 86 80 56 58 Grimace Excl. Excl. Excl. Excl. Excl. Bending Over 200+ 144 136 47 37 Normal Breathing 200+ 157 180 112 50 Overall Fit Factor 188 152 124 70 66

199 100% 100% (a1) (a2)

98% 90%

] ]

- -

[ 96% [ 80%

y y

c c

n n

e e

i i

c c

i i

f f f 94% f 70%

E E Dry Heat Treatment Dry Heat Treatment 3M™ 8210 N95 Untreated Halyard 48207 Surgical Mask Untreated 92% 3M™ 8210 N95 1 cycle 60% Halyard 48207 Surgical Mask 1 cycle 3M™ 8210 N95 3 cycles Halyard 48207 Surgical Mask 3 cycles 3M™ 8210 N95 5 cycles Halyard 48207 Surgical Mask 5 cycles 3M™ 8210 N95 10 cycles Halyard 48207 Surgical Mask 10 cycles 90% 50% 0.02 0.05 0.1 0.2 0.5 0.02 0.05 0.1 0.2 0.5 Particle Mobility Diameter [µm] Particle Mobility Diameter [µm] 100% 100% Dry Heat Treatment (a3) Dry Heat Treatment (a4) 3M™ 1820 Procedure Mask Untreated Halyard H600 Sterilization Wrap Untreated 3M™ 1820 Procedure Mask 1 cycle Halyard H600 Sterilization Wrap 1 cycle 90% 3M™ 1820 Procedure Mask 3 cycles 90% Halyard H600 Sterilization Wrap 3 cycles 3M™ 1820 Procedure Mask 5 cycles Halyard H600 Sterilization Wrap 5 cycles 3M™ 1820 Procedure Mask 10 cycles Halyard H600 Sterilization Wrap 10 cycles

] ]

- -

[ 80% [ 80%

y y

c c

n n

e e

i i

c c

i i

f f f 70% f 70%

E E

60% 60%

50% 50% 0.02 0.05 0.1 0.2 0.5 0.02 0.05 0.1 0.2 0.5 Particle Mobility Diameter [µm] Particle Mobility Diameter [µm]

]

100% a 400

(a5) P Untreated (a6)

[

s 1 cycle

m 3 cycles

c 300 98% 5 5 cycles

1 10 cycles

]

[ 96% 200

f a 100 94% i

Dry Heat Treatment r

EX101 Untreated f

200 100% 100% (b1) (b2)

98% 90%

] ]

- -

[ 96% [ 80%

y y

c c

n n

e e

i i

c c

i i

f f f 94% f 70%

E E Steam Heat Treatment Steam Heat Treatment 3M™ 8210 N95 Untreated Halyard 48207 Surgical Mask Untreated 92% 3M™ 8210 N95 1 cycle 60% Halyard 48207 Surgical Mask 1 cycle 3M™ 8210 N95 3 cycles Halyard 48207 Sursgical Mask 3 cycles 3M™ 8210 N95 5 cycles Halyard 48207 Sursgical Mask 5 cycles 3M™ 8210 N95 10 cycles Halyard 48207 Sursgical Mask 10 cycles 90% 50% 0.02 0.05 0.1 0.2 0.5 0.02 0.05 0.1 0.2 0.5 Particle Mobility Diameter [µm] Particle Mobility Diameter [µm] 100% 100% Steam Heat Treatment (b3) Steam Heat Treatment (b4) 3M™ 1820 Procedure Mask Untreated Halyard H600 Sterilization Wrap Untreated 3M™ 1820 Procedure Mask 1 cycle Halyard H600 Sterilization Wrap 1 cycle 90% 3M™ 1820 Procedure Mask 3 cycles 90% Halyard H600 Sterilization Wrap 3 cycles 3M™ 1820 Procedure Mask 5 cycles Halyard H600 Sterilization Wrap 5 cycles 3M™ 1820 Procedure Mask 10 cycles Halyard H600 Sterilization Wrap 10 cycles

] ]

- -

[ 80% [ 80%

y y

c c

n n

e e

i i

c c

i i

f f f 70% f 70%

E E

60% 60%

50% 50% 0.02 0.05 0.1 0.2 0.5 0.02 0.05 0.1 0.2 0.5 Particle Mobility Diameter [µm] Particle Mobility Diameter [µm]

]

100% a 400

(b5) P Untreated (b6)

[

s 1 cycle

m 3 cycles

c 300 98% 5 5 cycles

1 10 cycles

]

[ 96% 200

f a 100 94% i

Steam Heat Treatment r

EX101 Untreated f

92% EX101 1 cycle D 0 EX101 3 cycles 5 7 0 0 1 9 0 2 0 0 N 2 k 8 k 6 p 1 EX101 5 cycles 8 s 1 s a X 0 4 a a H r E 1 M M d W EX101 10 cycles 2 rd l ™ l r 8 a a M a a n 90% y ic 3 ic ly io ™ l g g a t M a r r H a 0.02 0.05 0.1 0.2 0.5 3 H u u liz S S ri te Particle Mobility Diameter [µm] S Figure 10–4. Fractional efficiency and breathing resistance of all five filter materials after dry (a) and steam (b) heat treatments.

201 10.3.5 Autoclave Treatment on Masks Made of Sterilization Wrap

Since first introduced as a N95 alternatives, the blue sterilization wrap has been well adopted to make face masks over many hospitals and clinics in United States. As discussed above, although not really N95 grade, this material does provide good protection with efficiency similar to 3M procedure masks at about doubled breathing resistance.

Unused and worn masks made of this sterilization wrap were provided by a local clinic, with all samples underwent an autoclave treatment at 132oC for 30 min after the last use

(if any) to minimize possible transmission of SARS-CoV-2 virus. Worn masks were used in real clinic activities and were treated with autoclave after each work shift and were then reused by the same person. 3 masks each were randomly selected after being wore and autoclaved for 1 or 2 times. Selected unworn masks were autoclaved up to 4 times as control groups. As can be seen in Figure 10–5, autoclaved unworn masks had slight efficiency deterioration after only 1 treatment, mainly in large sizes, due to charge loss.

Multiple treatments caused further degradation, but the impact seemed to fade out after 2 treatments, with very similar efficiencies observed for samples with 2, 3, and 4 autoclave treatments. The largest efficiency reduction after 4 treatments were within 20%.

On the other hand, worn masks showed much severer efficiency degradation even after only one worn-autoclave cycle. The fact of no significant change in breathing resistance and the shift of MPPS towards larger size both suggested the filtration deterioration was from charge loss. The efficiency of the 3 masks worn by 1 shift reduced by 18, 19, and 35% respectively at 300 nm. After 2 shifts, the reductions further developed to 23, 36, and 43% respectively. The worst sample of 2-shift cases (#1) showed similar efficiency to the IPA-soaked sterilization wrap (the same material the mask is made of), 202 indicating this sample lost almost all the electric charge on it. Respirators/masks used in real world may be contaminated by saliva and mucus from human body or chemicals and particles from ambient environment. These foreign objects may negatively impact the efficiency of electret media by neutralizing or shading their electrical charges (Ji et al. 2003,

Mahdavi et al. 2015, Raynor and Chae 2004), which cannot be recovered by decontamination. Another possible mechanism causing charge loss is the moisture (from human exhalation) condensation on filter fibers. The subsequent water molecule evaporation from fiber surface could potentially lead to electron exchange, which neutralizes the electric dipoles gradually over repetitive exhalation (condensation) and inhalation (evaporation) cycles. This hypothesis may also help to explain the mild filtration degradation after multiple steaming treatment cycles. Although testing of more field-used respirators/masks is needed to confirm this observation and further investigation is necessary to validate our hypothesis or any other potential mechanisms, the result from limited field-used samples here rises a question that if decontamination tests conducted on unworn respirators and masks are sufficient to provide appropriate representation of their real-world counterparts being worn and subsequently decontaminated for multiple times.

Various decontamination methods may be proven to not cause filtration degradation by themselves even after 10-20 cycles. The normal human wearing activities, however, may cause irreversible filtration deterioration that cannot be recovered by these decontamination methods. If this is proven true, it will negatively impact the current decontamination and reuse strategy of respirators and masks and may largely limit the number of times a respirator or mask can be reused.

203 100% Unworn + N Autoclave Treatment(s) 0 Treatment 1 Treatment 90% 2 Treatments 3 Treatments 4 Treatments 80% Worn 1 shift + 1 Autoclave Treatment #1 #2 #3 70% Worn 2 shifts + 2 Autoclave Treatments

]

- #1

[

y #2

n 60% #3

i IPA Soaked

E IPA Soaked 50% Breathing Resistance

] 200

P 177

[ 159168

/ 163 157 m 150 148 c 146 150 141 141

5 139 40% . 0 129

r 100

30% P

i 50

D 0 20% Unworn + Worn 1 shift +Worn 2 shifts + 0.02 0.05 0.1 0.2 0.5 N Autoclave 1 Autoclave 2 Autoclave IPA Soaked Treatment(s) Treatment Treatments Particle Mobility Diameter [µm] Figure 10–5. Filtration performance of unworn and worn H600 sterilization wrap masks after one or multiple cycle(s) of autoclave treatment. IPA soaked same material is included for reference purpose

10.4 Conclusions

This research studied the effect of various decontamination methods on the filtration performance of five selected commercial and alternative respirator/mask materials. Both UVGI for 5 min and dry heat at 77oC for 30 min were found not to cause observable performance degradation after up to 10 cycles on commercial respirators/masks tested. Steam heat treatment for 30 min was found to cause charge loss on 3M 1820 procedure mask and Halyard 48207 surgical masks, which led to efficiency deterioration after 5 or 10 cycles. Although similar efficiency decay was not observed on 3M 8210 N95 respirators, the respirators’ fit factor decreased with steam heat treatment cycles and 204 dropped below 100 (fail) after 5 cycles. As an easily accessible method, heat treatment can be implemented by general publics in household setting, but our results suggested to keep the moisture level below saturation if masks made of electret materials are decontaminated.

Both IPA soaking and spraying caused significant performance deterioration of all electrically charged materials, including the three commercial respirators/masks and

Halyard H600 sterilization wrap, by largely removing charges from material surface. No change of material integrity or structure was observed on any combination of filter materials and decontamination methods tested, suggesting the stability of electrical charge is the major factor (together with facial fit) that limits the ability of a material being decontaminated and reused for multiple times. Non-electret filter materials (e.g. EX101), therefore, pose their advantages of wider compatibility with a variety of decontamination methods so more attention can be put on retaining the fit and the functionality of other auxiliary components (e.g., straps, nose piece, etc.).

Fractional efficiency measurement approved to be a better method for respirator/mask performance evaluation than any size integrated methods (e.g. NIOSH N95 certification test), by providing size-resolved efficiency information (Rengasamy and

Eimer 2012). When used for evaluating decontaminated respirators/masks, this method is capable of distinguishing efficiency degradation due to charge loss from that caused by structure or integrity deterioration, which is not otherwise possible by a size integrated method.

The current study focused on the filtration performance of the filter materials, with fit testing conducted only on N95 respirators with selected decontamination methods. With more loose-fitting masks being used by frontline healthcare workers, assessing the facial

205 fit performance of these face masks, which is equally important to material filtration testing, has become urgently needed and requires more immediate efforts. Moreover, the limited results on sterilization wraps in this study implied that normal human wearing activity could potentially cause irreversible charge loss and performance deterioration of electret materials, which has been overlooked by most respirator/mask decontamination and reuse studies. Decontamination tests on real-world used samples, together with the material durability evaluation especially after extended reuse and multiple donning/doffing cycles, are necessary to fully assess current decontamination and reuse strategy of respirators and face masks.

206 Chapter 11: Summary and Future Work

This dissertation focuses on the effects of temperature and RH on filter loading by simulated atmospheric aerosols and COVID-19 related mask and respirator filtration study.

There are two objectives of this dissertation: the first one is to fill the gap between lab filter testing and actual filter operation; the second one is to help curb the COVID-19 spreading from filtration perspective. Part 1 and Part 2 focus on the first objective, Part 3 covers the second objective.

11.1 Effects of Temperature and RH on Filter Loading by Simulated Atmospheric Aerosols

The structure of the first objective of this dissertation is shown in Figure 11–1. It is aimed to study the discrepancy between lab filter testing and real filter operation and select better filters.

Part 1 and Part 2 cover the methodology development, and research experiments, respectively. Chapter 1 and Chapter 2 introduce the experimental apparatus used in this dissertation. Chapter 3 is about a novel filter performance monitoring system, which can report filter performance in real-time. Chapter 4 covers a database that can provide both historical pollution and meteorological information of the city of interest so that air filters can be recommended accordingly with the knowledge gained from this dissertation.

207

Figure 11–1. Structure of the first objective of this dissertation

Chapter 1 and 2 set the experimental apparatus foundation of this dissertation. To study the effects of temperature and RH on filter loading, a Lab-simulated Ambient

Environment Air Filter Test Rig was built so that both temperature and RH could be well controlled for filter testing in the laboratory. To simulate atmospheric aerosols, inorganic hygroscopic salts ((NH4)2SO4, NH4NO3), soot and organic compounds coated soot were employed. The techniques to generate atmospheric aerosols were established, including dry and wet salt particles, fresh and aged soot, and mixture particles. In previous studies, dry NaCl or KCl salt particles were used in filter testing. However, hygroscopic (NH4)2SO4 and NH4NO3 salt particles response at different RHs was carefully considered here so that both dry and wet salt particles were employed for filter testing at various RHs. In previous

208 studies, only fresh soot was used in filter testing. Whereas, aged soot filter testing was performed here, which can simulate ambient filtration scenario better.

Chapter 5 focuses on the temperature effect on filter loading. The testing temperature covers from 5 ℃ to 50 ℃ and the testing RH covers from 30% to 75%. It is found that temperature is a minor factor affecting the filter life. KCl and (NH4)2SO4 particles were tested in their dry state, and NH4NO3 was in its wet state, which represented the different particle states in the atmosphere. From experiment results, RH plays a significant role in affecting the filter life. In contrast, the temperature is a minor factor.

This confirms that filter testing RH should be well controlled while filters can be tested at room temperature. Consequently, the following studies were performed at room temperature.

Chapter 6 covers filter loading of the single-species particle at constant RH. In this chapter, conventional cellulose filter media and nano-fiber coated cellulose filter media were studied. It is found that in general, the higher the loading RH, the longer the filter life.

While the filter life of dry and wet particles is reversed on two types of filter media due to the formation of the liquid film on the nanofiber layer. When the majority of challenging particles are dry, the nano-fiber coated filter can last longer; while the conventional filter can hold more wet particles. It is learned that filter selection should be cautious, and the operating RH and challenging particles’ state should be considered.

Chapter 7 studied filter loading of the multiple-species particles at constant RH. In

Chapter 7, (NH4)2SO4-NH4NO3 mixture and organic compounds coated soot were studied at constant loading RH. It is found that mass gains of salt mixture particles were complex and depend on mixture particle composition and states, loading RH, and particle

209 dehydration history. The mass gain complexity is caused by the salt mixture hygroscopicity, which has been proven complicated and unpredictable. This result suggests that filter loading test should consider filter operating locations so that (NH4)2SO4-NH4NO3 mixing ratio can change accordingly to better simulate filter operation. According to field study, aged soot in the atmosphere is more compact than fresh soot. It is found that the test filter can last longer when it is challenged by organic compounds coated soot, which is lab simulated aged soot. This study implies that to better simulate ambient soot in lab filter testing, the aging process should be considered. Chapter 7 reveals the complexity of filter loading with particles that are similar to the ambient particulate pollutants at different loading RHs and suggests the necessity of testing filters based on their operating conditions.

Chapter 8 investigated the effect of RH change on the hygroscopic loaded filters, which is similar to the actual operating conditions since it is not realistic to have a constant

RH during ambient filter operation. It is found that for dry particle deposition, the higher the conditioning RH, the lower the terminal pressure due to moisture absorption induced restructuring. For wet particles, the elevated conditioning RH can prompt pressure reduction by introducing more water to salt deposits. However, the lowered conditioning

RH would not cause a pressure reduction. This study broadens the understandings of RH change effect on the hygroscopic particles loaded filters. Also, the possibility of utilizing the hygroscopic particle to prolong filter life or clean filters could be considered.

The ultimate goal of this objective is to recommend filters based on operating locations. Whereas it is still challenging to achieve filter selection and optimization based on all research mentioned above. Therefore, in Chapter 4, a Global Air

Pollutant/Meteorological Condition Database was established. For a city of interest, this

210 database can report historical pollutants speciation and concentration, historical monthly average temperature and RH, and pollutants concentration trend in past years. Before making decisions, the local filter operation conditions could support filter selection and optimization. Air filters can be customized in the future to accommodate different needs.

Meanwhile, even with the customized filter for different locations, the actual filter performance is still not well known. Chapter 3 solves this problem by prototyping a Smart

Sensor for Filter Performance Monitoring System. With such a system, real-time filtration efficiency, differential pressure across the filter, and operating temperature, RH are all available to closely monitor the filter performance. This system was tested on an indoor air purifier to validate the concept and it was justified that such a filter performance monitoring system is effective and convenient. This system could be customized to fit all air filter applications, which can open the black box of air filter operation. With millions of system installations, massive filter operation data could be received. With the advantage machine learning tool and big data analysis, filter manufactures could understand the direction of the optimization clearly. End-users can stay assured that filtered air qualifies for their needs, and occupants’ health and equipment are well protected as desired.

Both the Global Air Pollutant/Meteorological Condition Database and Smart

Sensor for Filter Performance Monitoring System could help close the loop of this study

(Figure 11–1). The discrepancy between the conventional filter testing and the actual filter operation is minimized by studies in this dissertation. With more stringent pollution control in past decades, sub-micron particulate pollutants are concerned. This dissertation extends the understandings of filter testing and operating with sub-micron particulate pollutants at

211 different operating conditions, which will benefit both the scientific research and the filtration industry.

11.2 COVID-19 Related Mask and Respirator Filtration Study

Chapter 9 compared common materials that have the potential to be used as alternative masks for the public in daily protection to slow down COVID-19 spread. Both commercially available respirators and masks were evaluated to establish references for fractional efficiency and breathability. Furnace filters, vacuum cleaner filters, common household materials were tested in the same method. It is found that electret media has a good balance between filtration efficiency and breathability, which could be an emergency option at the current COVID-19 pandemic. Besides, stacking several layers of fabrics could also offer protections while maintaining reasonable breathing resistance. The conclusions of this study are not only useful during the current pandemic, but also applicable to severe pollution episodes while respirators or masks are not available.

Chapter 10 aimed to evaluate the effect of decontamination of the commercial and alternative mask and respirator materials from the filtration perspective. This study was initiated by the desperate shortage of PPE of health care workers, and there were

212 suggestions of reusing respirators after decontamination. But there were also concerns that decontamination may jeopardize the filtration capability of respirators or masks. In this study, commercial and alternative mask and respirator materials underwent several common decontaminations, including isopropanol alcohol, ultraviolet germicidal irradiation, thermal treatment, and autoclave treatment, were tested for their fractional efficiencies and breathabilities. It is found that ultraviolet germicidal irradiation and dry heat treatment cannot cause observable efficiency decay for 10 cycles. However, isopropanol alcohol and wet heat (steaming) could lead to efficiency deterioration. The possible reason is the loss of charge on electret media since the similar efficiency decay was not observed on a mechanical filter media. An N95 respirator fit factors evolution after treatment cycles was also studied, which agrees with the media study that dry heat can maintain efficiency while moist heat may lead to efficiency deterioration. It is also found that normal breathing may lead to charge loss and subsequently efficiency decay, but further investigation is needed.

11.3 Future Work

Though this dissertation drives the lab filter testing closer to the actual filter operating condition, it could be improved in the following aspects:

1. This dissertation only tested conventional cellulose filter and nanofiber coated

cellulose filter. Another popular filter media, electret filter media, was not

tested. Luckily, several researchers are working on the electret media (Li et al.

2020, Wang et al. 2020). While the salt they chose was still limited to KCl and

213 NaCl. Different particle states and RH with (NH4)2SO4 and NH4NO3 should be

tested on the electret media to achieve a filter testing closer to the actual filter

operating. From the experience in the Part 3, moisture variation may lead to

charge loss on electret media, therefore such phenomenon should be considered

when electret media is tested at different RHs.

2. Limited efficiency measurements were achieved in this dissertation. Efficiency

evolution during the loading process at different RHs is also worthwhile to

study. For example, the wet NH4NO3 efficiency curve never reached 100% in

Section 6.1. Hence the extra caution is needed in the efficiency change during

the loading process.

3. In Chapter 7, soot was only coated with hydrophilic glutaric acid due to

university closure due to COVID-19 pandemic. The hydrophobic oleic acid

coating should be also conducted to compare the effect of coating

hygroscopicity on the filter performance. The testing RH close or above 100%

with both hydrophilic and hydrophobic coating should also be included. And

some other treatments that may prompt the hygroscopicity, such as UV, Ozone,

could be added in the future to better simulate the aging process in a continuous

flow system to test air filters.

4. In Chapter 8, only a single step RH change was performed. It is worthwhile to

study the filter efficiency and loading after the RH change, especially when

pressure drop declines. RH change could be a potential method to prolong the

filter life. Another future work that could be tested is the reverse pulse cleaning

process on filters that were loaded at high RH but conditioned at lower RH. The

214 solidified salt layer on the top of the filter may be peeled off easier than the high

RH salt paste.

For the COVID-19 related study, there are several studies worth studying:

1. The overall efficacy of masks when it is worn on a person. All studies presented

in this dissertation only tested the filtration efficiency of mask and respirator

materials. For tight-fitting respirators, the overall efficacy could be estimated

from media efficiencies when the wearer passed the fit testing. While it is

impossible to estimate the overall efficacy of the loose-fitting masks due to

leaks between the mask and wearer’s face. It is unclear that which one of the

two is more important to provide respiratory protection: the media efficiency or

the seal around wearer’s face. The answer is vital for loose-fitting masks

development.

2. The effect of the normal breathing on the electret media in mask and respirator.

The RH of the exhaled air is very high, while the inhalation process draws

ambient air, normally a lower RH air, through respirators or masks. The

alternation of the RH may cause water vapor condensation and evaporation on

the electret media. The cycling process may cause charge loss on the electret

media, which leads to efficiency deterioration. If normal breathing could cause

the efficiency decay, the reuse of respirators and masks should be reconsidered,

since the decontamination process cannot recover the filtration efficiency.

Therefore, it is worthwhile to study this topic.

215 Bibliography 3M (2020). "Coronavirus and Important Things to Know About Airborne Particles." Retrieved Jun 2, 2020, from https://www.filtrete.com/3M/en_US/filtrete/home- tips/full-story/~/coronavirus-and-important-things-to-know-about-airborne- particles/?storyid=d69e7735-c02c-46d2-9c8a-ae23372934ac. Adams, J. R. and Merz, A. R. (1929). "Hygroscopicity of Fertilizer Materials and Mixtures" Industrial & Engineering Chemistry 21(4): 305-307. doi:10.1021/ie50232a003 Ahn, K. H., Kim, S. M., Jung, H. J., Lee, M. J., Eom, H. J., Maskey, S. and Ro, C. U. (2010). "Combined use of optical and electron microscopic techniques for the measurement of hygroscopic property, chemical composition, and morphology of individual aerosol particles" Anal Chem 82(19): 7999-8009. doi:10.1021/ac101432y Anderegg, L., Meisenhelder, C., Ngooi, C. O., Liao, L., Xiao, W., Chu, S., Cui, Y. and Doyle, J. M. (2020). "A scalable method of applying heat and humidity for decontamination of N95 respirators during the COVID-19 crisis" PLoS One 15(7): e0234851. doi:10.1371/journal.pone.0234851 Anfinrud, P., Stadnytskyi, V., Bax, C. E. and Bax, A. (2020). "Visualizing Speech- Generated Oral Fluid Droplets with Laser Light Scattering" N Engl J Med 382(21): 2061-2063. doi:10.1056/NEJMc2007800 ASHRAE (2017). ASHRAE 52.2-2017 Method of Testing General Ventilation Air- cleaning Devices for Removal Efficiency by Particle Size. Atlanta, GA. ASHRAE (2020). "ASHRAE Position Document on Infectious Aerosols", ASTM International. (2017). F2299/F2299M-03(2017) Standard Test Method for Determining the Initial Efficiency of Materials Used in Medical Face Masks to Penetration by Particulates Using Latex Spheres. West Conshohocken, PA;, ASTM International.,. ASTM International. (2019). F2101 − 19 Standard Test Method for Evaluating the Bacterial Filtration Efficiency (BFE) of Medical Face Mask Materials, Using a Biological Aerosol of Staphylococcus aureus, ASTM International.,. Bae, S., Kim, M. C., Kim, J. Y., Cha, H. H., Lim, J. S., Jung, J., Kim, M. J., Oh, D. K., Lee, M. K., Choi, S. H., Sung, M., Hong, S. B., Chung, J. W. and Kim, S. H. (2020). "Effectiveness of Surgical and Cotton Masks in Blocking SARS-CoV-2: A Controlled Comparison in 4 Patients" Ann Intern Med 173(1): W22-W23. doi:10.7326/M20-1342 Barthazy, E., Stetzer, O., Derungs, C., Saathoff, H. and Lohmann, U. (2006). Water uptake of soot particles emitted from a Jing-CAST soot generator. 12th Conference on Cloud Physics. Battelle (2016). "Final report for the bioquell hydrogen peroxide vapor (HPV) decontamination for reuse of N95 respirators.",

216 Bergman, M. S., Viscusi, D. J., Heimbuch, B. K., Wander, J. D., Sambol, A. R. and Shaffer, R. E. (2018). "Evaluation of Multiple (3-Cycle) Decontamination Processing for Filtering Facepiece Respirators" Journal of Engineered Fibers and Fabrics 5(4): 33- 41. doi:10.1177/155892501000500405 Boukhriss, M., Zhani, K. and Ghribi, R. (2013). "Study of thermophysical properties of a solar desalination system using solar energy" Desalination and Water Treatment 51(4-6): 1290-1295. doi:10.1080/19443994.2012.714925 Bourouiba, L., Dehandschoewercker, E. and Bush, John W. M. (2014). "Violent expiratory events: on coughing and sneezing" Journal of Fluid Mechanics 745: 537-563. doi:10.1017/jfm.2014.88 Bourouiba, L. (2020). "Turbulent Gas Clouds and Respiratory Pathogen Emissions: Potential Implications for Reducing Transmission of COVID-19" JAMA 323(18): 1837-1838. doi:10.1001/jama.2020.4756 Brown, R. C. (1993). Air Filtration: An Integrated Approach to the Theory and Applications of Fibrous Filters, Elsevier Science. ISBN:9780080412740 Businesswire (2020). "Cummins and DuPont Partner to Use Filter Technology to Supply Critical Material for N95 Respirator Masks." from https://www.businesswire.com/news/home/20200406005502/en/Cummins- DuPont-Partner-Filter-Technology-Supply-Critical. Butt, H. J. and Kappl, M. (2009). "Normal capillary forces" Adv Colloid Interface Sci 146(1-2): 48-60. doi:10.1016/j.cis.2008.10.002 Cao, Q., Pui, D. Y. H. and Lipiński, W. (2015). "A Concept of a Novel Solar-Assisted Large-Scale Cleaning System (SALSCS) for Urban Air Remediation" Aerosol and Air Quality Research 15(1): 1-10. doi:10.4209/aaqr.2014.10.0246 Cao, Q., Huang, M., Kuehn, T. H., Shen, L., Tao, W.-Q., Cao, J. and Pui, D. Y. H. (2018a). "Urban-scale SALSCS, Part II: A Parametric Study of System Performance" Aerosol and Air Quality Research 18(11): 2879-2894. doi:10.4209/aaqr.2018.06.0239 Cao, Q., Kuehn, T. H., Shen, L., Chen, S.-C., Zhang, N., Huang, Y., Cao, J. and Pui, D. Y. H. (2018b). "Urban-scale SALSCS, Part I: Experimental Evaluation and Numerical Modeling of a Demonstration Unit" Aerosol and Air Quality Research 18(11): 2865-2878. doi:10.4209/aaqr.2018.06.0238 CDC (2020a). "Decontamination and Reuse of Filtering Facepiece Respirators." Retrieved Aug 1, 2020, from https://www.cdc.gov/coronavirus/2019-ncov/hcp/ppe- strategy/decontamination-reuse-respirators.html. CDC (2020b). "Clinical Questions about COVID-19: Questions and Answers." Retrieved May 26, 2020, from https://www.cdc.gov/coronavirus/2019- ncov/hcp/faq.html#Transmission.

217 CDC (2020c). "Use of Cloth Face Coverings to Help Slow the Spread of COVID-19." Retrieved May 26, 2020, from https://www.cdc.gov/coronavirus/2019- ncov/prevent-getting-sick/diy-cloth-face-coverings.html. CDC (2020d). "How to Make Cloth Face Coverings." Retrieved May 31, 2020, from https://www.cdc.gov/coronavirus/2019-ncov/prevent-getting-sick/how-to-make- cloth-face-covering.html. Chang, S.-Y. and Lee, C.-T. (2002). "Applying GC-TCD to investigate the hygroscopic characteristics of mixed aerosols" Atmospheric Environment 36(9): 1521-1530. doi:10.1016/s1352-2310(01)00546-5 Chen, C.-C. and Willeke, K. (1992). "Aerosol penetration through surgical masks" American Journal of Infection Control 20(4): 177-184. doi:10.1016/s0196- 6553(05)80143-9 Cheng, Y.-H. and Tsai, C.-J. (1998). "Factors Influencing Pressure Drop through a Dust Cake during Filtration" Aerosol Science and Technology 29(4): 315-328. doi:10.1080/02786829808965572 Cheng, Z., Luo, L., Wang, S., Wang, Y., Sharma, S., Shimadera, H., Wang, X., Bressi, M., de Miranda, R. M., Jiang, J., Zhou, W., Fajardo, O., Yan, N. and Hao, J. (2016). "Status and characteristics of ambient PM2.5 pollution in global megacities" Environment International 89-90: 212-221. doi:10.1016/j.envint.2016.02.003 Chin, A. W. H., Chu, J. T. S., Perera, M. R. A., Hui, K. P. Y., Yen, H.-L., Chan, M. C. W., Peiris, M. and Poon, L. L. M. (2020). "Stability of SARS-CoV-2 in different environmental conditions" The Lancet Microbe 1(1): e10. doi:10.1016/s2666- 5247(20)30003-3 Cohen, M. D., Flagan, R. C. and Seinfeld, J. H. (1987a). "Studies of concentrated electrolyte solutions using the electrodynamic balance. 1. Water activities for single-electrolyte solutions" The Journal of Physical Chemistry 91(17): 4563-4574. doi:10.1021/j100301a029 Cohen, M. D., Flagan, R. C. and Seinfeld, J. H. (1987b). "Studies of concentrated electrolyte solutions using the electrodynamic balance. 2. Water activities for mixed-electrolyte solutions" The Journal of Physical Chemistry 91(17): 4575-4582. doi:10.1021/j100301a030 Cohen, M. D., Flagan, R. C. and Seinfeld, J. H. (1987c). "Studies of concentrated electrolyte solutions using the electrodynamic balance. 3. Solute nucleation" The Journal of Physical Chemistry 91(17): 4583-4590. doi:10.1021/j100301a031 Contal, P., Simao, J., Thomas, D., Frising, T., Callé, S., Appert-Collin, J. C. and Bémer, D. (2004). "Clogging of fibre filters by submicron droplets. Phenomena and influence of operating conditions" Journal of Aerosol Science 35(2): 263-278. doi:10.1016/j.jaerosci.2003.07.003 Cziczo, D. J., Nowak, J. B., Hu, J. H. and Abbatt, J. P. D. (1997). "Infrared spectroscopy of model tropospheric aerosols as a function of relative humidity: Observation of

218 deliquescence and crystallization" Journal of Geophysical Research: Atmospheres 102(D15): 18843-18850. doi:10.1029/97jd01361 Dalirian, M., Ylisirniö, A., Buchholz, A., Schlesinger, D., Ström, J., Virtanen, A. and Riipinen, I. (2018). "Cloud droplet activation of black carbon particles coated with organic compounds of varying solubility" Atmospheric Chemistry and Physics 18(16): 12477-12489. doi:10.5194/acp-18-12477-2018 Davies, A., Thompson, K.-A., Giri, K., Kafatos, G., Walker, J. and Bennett, A. (2013). "Testing the efficacy of homemade masks: would they protect in an influenza pandemic?" Disaster medicine and public health preparedness 7(4): 413-418. doi:10.1017/dmp.2013.43 Dougle, P. G., Veefkind, J. P. and ten Brink, H. M. (1998). "Crystallisation of mixtures of ammonium nitrate, ammonium sulphate and soot" Journal of Aerosol Science 29(3): 375-386. doi:10.1016/s0021-8502(97)10003-9 Ebert, M., Inerle-Hof, M. and Weinbruch, S. (2002). "Environmental scanning electron microscopy as a new technique to determine the hygroscopic behaviour of individual aerosol particles" Atmospheric Environment 36(39-40): 5909-5916. doi:10.1016/s1352-2310(02)00774-4 Edgar, G. and Swan, W. O. (1922). "The Factors Determining the Hygroscopic Properties of Soluble Substances. I. The Vapor Pressures of Saturated Solutions" Journal of the American Chemical Society 44(3): 570-577. doi:10.1021/ja01424a014 EPA (1998). PM2.5 Mass Weighing Laboratory Standard Operating Procedures for the Performance Evaluation Program, European Committee for Standardization (2012). EN 779:2012 Particulate air filters for general ventilation. Determination of the filtration performance. FDA (2004). "Guidance for Industry and FDA Staﬀ. Surgical Masks - Premarket Notiﬁcation [510(K)] Submissions; Guidance for Industry and FDA." Retrieved May 26,2020, from https://www.fda.gov/regulatory-information/search-fda- guidance-documents/surgical-masks-premarket-notification-510k-submissions. Feng, S., Shen, C., Xia, N., Song, W., Fan, M. and Cowling, B. J. (2020). "Rational use of face masks in the COVID-19 pandemic" The Lancet Respiratory Medicine 8(5): 434-436. doi:10.1016/S2213-2600(20)30134-X Fischer, R. J., Morris, D. H., van Doremalen, N., Sarchette, S., Matson, M. J., Bushmaker, T., Yinda, C. K., Seifert, S. N., Gamble, A., Williamson, B. N., Judson, S. D., de Wit, E., Lloyd-Smith, J. O. and Munster, V. J. (2020). "Assessment of N95 respirator decontamination and re-use for SARS-CoV-2" medRxiv 2020.04.11.20062018 doi:10.1101/2020.04.11.20062018 Fisher, E. M. and Shaffer, R. E. (2011). "A method to determine the available UV-C dose for the decontamination of filtering facepiece respirators" J Appl Microbiol 110(1): 287-295. doi:10.1111/j.1365-2672.2010.04881.x

219 Fisher, E. M., Williams, J. L. and Shaffer, R. E. (2011). "Evaluation of microwave steam bags for the decontamination of filtering facepiece respirators" PLoS One 6(4): e18585. doi:10.1371/journal.pone.0018585 Freney, E. J., Martin, S. T. and Buseck, P. R. (2009). "Deliquescence and Efflorescence of Potassium Salts Relevant to Biomass-Burning Aerosol Particles" Aerosol Science and Technology 43(8): 799-807. doi:10.1080/02786820902946620 Gao, Y., Chen, S. B. and Yu, L. E. (2006). "Efflorescence relative humidity for ammonium sulfate particles" J Phys Chem A 110(24): 7602-7608. doi:10.1021/jp057574g Ge, Z., Wexler, A. S. and Johnston, M. V. (1996). "Multicomponent Aerosol Crystallization" Journal of Colloid and Interface Science 183(1): 68-77. doi:10.1006/jcis.1996.0519 Ge, Z., Wexler, A. S. and Johnston, M. V. (1998). "Deliquescence Behavior of Multicomponent Aerosols" The Journal of Physical Chemistry A 102(1): 173-180. doi:10.1021/jp972396f Godoy, C. and Thomas, D. (2020). "Influence of relative humidity on HEPA filters during and after loading with soot particles" Aerosol Science and Technology 54(7): 790- 801. doi:10.1080/02786826.2020.1726278 Gralton, J., Tovey, E., McLaws, M. L. and Rawlinson, W. D. (2011). "The role of particle size in aerosolised pathogen transmission: a review" J Infect 62(1): 1-13. doi:10.1016/j.jinf.2010.11.010 Greenhalgh, T., Schmid, M. B., Czypionka, T., Bassler, D. and Gruer, L. (2020). "Face masks for the public during the covid-19 crisis" BMJ 369: m1435. doi:10.1136/bmj.m1435 Greenspan, L. (1977). "Humidity fixed points of binary saturated aqueous solutions" Journal of Research of the National Bureau of Standards Section A: Physics and Chemistry 81A(1): 89-96. doi:10.6028/jres.081A.011 Gupta, A., Novick, V. J., Biswas, P. and Monson, P. R. (1993). "Effect of Humidity and Particle Hygroscopicity on the Mass Loading Capacity of High Efficiency Particulate Air (HEPA) Filters" Aerosol Science and Technology 19(1): 94-107. doi:10.1080/02786829308959624 Gupta, D., Kim, H., Park, G., Li, X., Eom, H. J. and Ro, C. U. (2015). "Hygroscopic properties of NaCl and NaNO₃ mixture particles as reacted inorganic sea-salt aerosol surrogates" Atmos. Chem. Phys. 15(6): 3379-3393. doi:10.5194/acp-15-3379-2015 Guttikunda, S. K. and Calori, G. (2013). "A GIS based emissions inventory at 1 km × 1 km spatial resolution for air pollution analysis in Delhi, India" Atmospheric Environment 67: 101-111. doi:10.1016/j.atmosenv.2012.10.040 Gysel, M., Weingartner, E. and Baltensperger, U. (2002). "Hygroscopicity of aerosol particles at low temperatures. 2. Theoretical and experimental hygroscopic

220 properties of laboratory generated aerosols" Environ Sci Technol 36(1): 63-68. doi:10.1021/es010055g Hajra, M. G., Mehta, K. and Chase, G. G. (2003). "Effects of humidity, temperature, and nanofibers on drop coalescence in glass fiber media" Separation and Purification Technology 30(1): 79-88. doi:10.1016/s1383-5866(02)00134-x Halyard (2020). "Can I Use Halyard Sterilization Wrap for DIY Masks?" Retrieved Jun 2, 2020, from https://www.halyardhealth.com/covid-19/diy-masks.aspx. Haynes, W. M. (2014). CRC Handbook of Chemistry and Physics, 95th Edition. ISBN:9781482208689 He, K., Yang, F., Ma, Y., Zhang, Q., Yao, X., Chan, C. K., Cadle, S., Chan, T. and Mulawa, P. (2001). "The characteristics of PM2.5 in Beijing, China" Atmospheric Environment 35(29): 4959-4970. doi:10.1016/s1352-2310(01)00301-6 He, L.-Y., Hu, M., Huang, X.-F., Yu, B.-D., Zhang, Y.-H. and Liu, D.-Q. (2004). "Measurement of emissions of fine particulate organic matter from Chinese cooking" Atmospheric Environment 38(38): 6557-6564. doi:10.1016/j.atmosenv.2004.08.034 Heimbuch, B. and Harnish, D. (2019). Research to Mitigate a Shortage of Respiratory Protection Devices During Public Health Emergencies. Heimbuch, B. K., Wallace, W. H., Kinney, K., Lumley, A. E., Wu, C. Y., Woo, M. H. and Wander, J. D. (2011). "A pandemic influenza preparedness study: use of energetic methods to decontaminate filtering facepiece respirators contaminated with H1N1 aerosols and droplets" Am J Infect Control 39(1): e1-9. doi:10.1016/j.ajic.2010.07.004 Hinds, W. C. (2012). Aerosol Technology: Properties, Behavior, and Measurement of Airborne Particles, Wiley. ISBN:9781118591970 Howard, J., Huang, A., Li, Z., Tufekci, Z., Zdimal, V., van der Westhuizen, H.-M., von Delft, A., Price, A., Fridman, L., Tang, L.-H., Tang, V., Watson, G. L., Bax, C. E., Shaikh, R., Questier, F., Hernandez, D., Chu, L. F., Ramirez, C. M. and Rimoin, A. W. (2020). "Face Masks Against COVID-19: An Evidence Review" Preprints 2020: 2020040203. doi:10.20944/preprints202004.0203.v2 Hsiao, T.-C., Chuang, H.-C., Griffith, S. M., Chen, S.-J. and Young, L.-H. (2020). "COVID-19: An Aerosol’s Point of View from Expiration to Transmission to Viral- mechanism" Aerosol and Air Quality Research: 905-910. doi:10.4209/aaqr.2020.04.0154 Hu, D., Qiao, L., Chen, J., Ye, X., Yang, X., Cheng, T. and Fang, W. (2010). "Hygroscopicity of Inorganic Aerosols: Size and Relative Humidity Effects on the Growth Factor" Aerosol and Air Quality Research 10(3): 255-264. doi:10.4209/aaqr.2009.12.0076 Huang, P.-F., Turpin, B. J., Pipho, M. J., Kittelson, D. B. and McMurry, P. H. (1994). "Effects of water condensation and evaporation on diesel chain-agglomerate

221 morphology" Journal of Aerosol Science 25(3): 447-459. doi:10.1016/0021- 8502(94)90063-9 IQAir (2020). "World's most polluted cities 2019 (PM2.5)." from https://www.iqair.com/world-most-polluted-cities. ISO (2014). ISO 5011:2014 Inlet air cleaning equipment for internal combustion engines and compressors — Performance testing. ISO (2016). ISO 16890:2016 Air filters for general ventilation. Jasper, W., Hinestroza, J., Mohan, A., Kim, J., Shiels, B., Gunay, M., Thompson, D. and Barker, R. (2006). "Effect of xylene exposure on the performance of electret filter media" Journal of Aerosol Science 37(7): 903-911. doi:10.1016/j.jaerosci.2005.06.008 Ji, J. H., Bae, G. N., Kang, S. H. and Hwang, J. (2003). "Effect of particle loading on the collection performance of an electret cabin air filter for submicron aerosols" Journal of Aerosol Science 34(11): 1493-1504. doi:10.1016/s0021-8502(03)00103-4 Jing, B., Tong, S., Liu, Q., Li, K., Wang, W., Zhang, Y. and Ge, M. (2016). "Hygroscopic behavior of multicomponent organic aerosols and their internal mixtures with ammonium sulfate" Atmos. Chem. Phys. 16(6): 4101-4118. doi:10.5194/acp-16- 4101-2016 Jing, L. (1999). Standard combustion aerosol generator (SCAG) for calibration purposes. 3rd ETH Workshop “Nanoparticle Measurement”, ETH Hönggerberg Zürich. Johnson, D. F., Druce, J. D., Birch, C. and Grayson, M. L. (2009). "A Quantitative Assessment of the Efficacy of Surgical and N95 Masks to Filter Influenza Virus in Patients with Acute Influenza Infection" Clinical Infectious Diseases 49(2): 275- 277. doi:10.1086/600041 Joubert, A., Laborde, J. C., Bouilloux, L., Callé-Chazelet, S. and Thomas, D. (2010). "Influence of Humidity on Clogging of Flat and Pleated HEPA Filters" Aerosol Science and Technology 44(12): 1065-1076. doi:10.1080/02786826.2010.510154 Joubert, A., Laborde, J. C., Bouilloux, L., Chazelet, S. and Thomas, D. (2011). "Modelling the pressure drop across HEPA filters during cake filtration in the presence of humidity" Chemical Engineering Journal 166(2): 616-623. doi:10.1016/j.cej.2010.11.033 Kawamura, K. and Yasui, O. (2005). "Diurnal changes in the distribution of dicarboxylic acids, ketocarboxylic acids and dicarbonyls in the urban Tokyo atmosphere" Atmospheric Environment 39(10): 1945-1960. doi:10.1016/j.atmosenv.2004.12.014 Kim, J., Hinestroza, J. P., Jasper, W. and Barker, R. L. (2009a). "Effect of Solvent Exposure on the Filtration Performance of Electrostatically Charged Polypropylene Filter Media" Textile Research Journal 79(4): 343-350. doi:10.1177/0040517508090887

222 Kim, J. H., Roberge, R. J., Powell, J. B., Shaffer, R. E., Ylitalo, C. M. and Sebastian, J. M. (2015). "Pressure drop of filtering facepiece respirators: How low should we go?" Int J Occup Med Environ Health 28(1): 71-80. doi:10.13075/ijomeh.1896.00153 Kim, S. C., Wang, J., Shin, W. G., Scheckman, J. H. and Pui, D. Y. H. (2009b). "Structural Properties and Filter Loading Characteristics of Soot Agglomerates" Aerosol Science and Technology 43(10): 1033-1041. doi:10.1080/02786820903131081 Kireeva, E. D., Popovicheva, O. B., Khokhlova, T. D. and Shoniya, N. K. (2010). "Laboratory simulation of the interaction of water molecules with carbonaceous aerosols in the atmosphere" Moscow University Physics Bulletin 65(6): 510-515. doi:10.3103/s0027134910060159 Kittelson, D. B. (1998). "Engines and Nanoparticles: A Review" Journal of Aerosol Science 29(5-6): 575-588. doi:10.1016/s0021-8502(97)10037-4 Klompas, M., Morris, C. A., Sinclair, J., Pearson, M. and Shenoy, E. S. (2020). "Universal Masking in Hospitals in the Covid-19 Era" N Engl J Med 382(21): e63. doi:10.1056/NEJMp2006372 Köllensperger, G., Friedbacher, G., Kotzick, R., Niessner, R. and Grasserbauer, M. (1999). "In-situ atomic force microscopy investigation of aerosols exposed to different humidities" Fresenius' Journal of Analytical Chemistry 364(4): 296-304. doi:10.1007/s002160051340 Konda, A., Prakash, A., Moss, G. A., Schmoldt, M., Grant, G. D. and Guha, S. (2020). "Aerosol Filtration Efficiency of Common Fabrics Used in Respiratory Cloth Masks" ACS Nano 14(5): 6339-6347. doi:10.1021/acsnano.0c03252 Kumar, A., Kasloff, S. B., Leung, A., Cutts, T., Strong, J. E., Hills, K., Vazquez-Grande, G., Rush, B., Lother, S., Zarychanski, R. and Krishnan, J. (2020). "N95 Mask Decontamination using Standard Hospital Sterilization Technologies" medRxiv: 2020.2004.2005.20049346. doi:10.1101/2020.04.05.20049346 Lai, A. C., Poon, C. K. and Cheung, A. C. (2012). "Effectiveness of facemasks to reduce exposure hazards for airborne infections among general populations" J R Soc Interface 9(70): 938-948. doi:10.1098/rsif.2011.0537 Langlet, M., Nadaud, F., Benali, M., Pezron, I., Saleh, K., Guigon, P. and Metlas- Komunjer, L. (2011). "Kinetics of Dissolution and Recrystallization of Sodium Chloride at Controlled Relative Humidity" KONA Powder and Particle Journal 29(0): 168-179. doi:10.14356/kona.2011019 Laskina, O., Morris, H. S., Grandquist, J. R., Qin, Z., Stone, E. A., Tivanski, A. V. and Grassian, V. H. (2015). "Size matters in the water uptake and hygroscopic growth of atmospherically relevant multicomponent aerosol particles" J Phys Chem A 119(19): 4489-4497. doi:10.1021/jp510268p Lee, C.-T. and Hsu, W.-C. (2000). "The Measurement of Liquid Water Mass Associated with Collected Hygroscopic Particles" Journal of Aerosol Science 31(2): 189-197. doi:10.1016/s0021-8502(99)00048-8

223 Leung, C. C., Lam, T. H. and Cheng, K. K. (2020a). "Mass masking in the COVID-19 epidemic: people need guidance" Lancet 395(10228): 945. doi:10.1016/S0140- 6736(20)30520-1 Leung, K. K., Schnitzler, E. G., Jäger, W. and Olfert, J. S. (2017). "Relative Humidity Dependence of Soot Aggregate Restructuring Induced by Secondary Organic Aerosol: Effects of Water on Coating Viscosity and Surface Tension" Environmental Science & Technology Letters 4(9): 386-390. doi:10.1021/acs.estlett.7b00298 Leung, N. H. L., Chu, D. K. W., Shiu, E. Y. C., Chan, K.-H., McDevitt, J. J., Hau, B. J. P., Yen, H.-L., Li, Y., Ip, D. K. M., Peiris, J. S. M., Seto, W.-H., Leung, G. M., Milton, D. K. and Cowling, B. J. (2020b). "Respiratory virus shedding in exhaled breath and efficacy of face masks" Nature Medicine 26(5): 676-680. doi:10.1038/s41591- 020-0843-2 Li, J. J., Wang, G. H., Cao, J. J., Wang, X. M. and Zhang, R. J. (2013). "Observation of biogenic secondary organic aerosols in the atmosphere of a mountain site in central China: temperature and relative humidity effects" Atmos. Chem. Phys. 13(22): 11535-11549. doi:10.5194/acp-13-11535-2013 Li, K. W., Chen, L. H., Han, K., Lv, B. A., Bao, K. J., Wu, X. C., Gao, X. and Cen, K. F. (2017). "Smog chamber study on aging of combustion soot in isoprene/SO2/NOx system: Changes of mass, size, effective density, morphology and mixing state" Atmospheric Research 184: 139-148. doi:10.1016/j.atmosres.2016.10.011 Li, S., Chen, D.-R., Zhou, F. and Chen, S.-C. (2020). "Effects of relative humidity and particle hygroscopicity on the initial efficiency and aging characteristics of electret HVAC filter media" Building and Environment 171: 106669. doi:10.1016/j.buildenv.2020.106669 Li, W., Sun, J., Xu, L., Shi, Z., Riemer, N., Sun, Y., Fu, P., Zhang, J., Lin, Y., Wang, X., Shao, L., Chen, J., Zhang, X., Wang, Z. and Wang, W. (2016a). "A conceptual framework for mixing structures in individual aerosol particles" Journal of Geophysical Research: Atmospheres 121(22): 13,784-713,798. doi:10.1002/2016JD025252 Li, X., Gupta, D., Eom, H.-J., Kim, H. and Ro, C.-U. (2014). "Deliquescence and efflorescence behavior of individual NaCl and KCl mixture aerosol particles" Atmospheric Environment 82: 36-43. doi:10.1016/j.atmosenv.2013.10.011 Li, Y. J., Liu, P. F., Bergoend, C., Bateman, A. P. and Martin, S. T. (2016b). "Rebounding hygroscopic inorganic aerosol particles: Liquids, gels, and hydrates" Aerosol Science and Technology 51(3): 388-396. doi:10.1080/02786826.2016.1263384 Liao, L., Xiao, W., Zhao, M., Yu, X., Wang, H., Wang, Q., Chu, S. and Cui, Y. (2020). "Can N95 Respirators Be Reused after Disinfection? How Many Times?" ACS Nano 14(5): 6348-6356. doi:10.1021/acsnano.0c03597 Lide, D. R. (2005). CRC Handbook of Chemistry and Physics, 86th Edition, Taylor & Francis. ISBN:9780849304866 224 Liew, T. P. and Conder, J. R. (1985). "Fine mist filtration by wet filters—I. Liquid saturation and flow resistance of fibrous filters" Journal of Aerosol Science 16(6): 497-509. doi:10.1016/0021-8502(85)90002-3 Lightstone, J. M., Onasch, T. B., Imre, D. and Oatis, S. (2000). "Deliquescence, Efflorescence, and Water Activity in Ammonium Nitrate and Mixed Ammonium Nitrate/Succinic Acid Microparticles" The Journal of Physical Chemistry A 104(41): 9337-9346. doi:10.1021/jp002137h Lim, Y. X., Ng, Y. L., Tam, J. P. and Liu, D. X. (2016). "Human Coronaviruses: A Review of Virus-Host Interactions" Diseases 4(3): 26. doi:10.3390/diseases4030026 Lindblad, M., Tano, E., Lindahl, C. and Huss, F. (2020). "Ultraviolet-C decontamination of a hospital room: Amount of UV light needed" Burns 46(4): 842-849. doi:10.1016/j.burns.2019.10.004 Lindsley, W. G., Martin, S. B., Jr., Thewlis, R. E., Sarkisian, K., Nwoko, J. O., Mead, K. R. and Noti, J. D. (2015). "Effects of Ultraviolet Germicidal Irradiation (UVGI) on N95 Respirator Filtration Performance and Structural Integrity" J Occup Environ Hyg 12(8): 509-517. doi:10.1080/15459624.2015.1018518 Liu, B. Y. H. and Pui, D. Y. H. (1974). "Equilibrium bipolar charge distribution of aerosols" Journal of Colloid and Interface Science 49(2): 305-312. doi:10.1016/0021-9797(74)90366-x Liu, H., Qi, L., Liang, C., Deng, F., Man, H. and He, K. (2020). "How aging process changes characteristics of vehicle emissions? A review" Critical Reviews in Environmental Science and Technology 50(17): 1796-1828. doi:10.1080/10643389.2019.1669402 Liu, Y., Yang, Z., Desyaterik, Y., Gassman, P. L., Wang, H. and Laskin, A. (2008). "Hygroscopic behavior of substrate-deposited particles studied by micro-FT-IR spectroscopy and complementary methods of particle analysis" Anal Chem 80(3): 633-642. doi:10.1021/ac701638r Lore, M. B., Heimbuch, B. K., Brown, T. L., Wander, J. D. and Hinrichs, S. H. (2012). "Effectiveness of three decontamination treatments against influenza virus applied to filtering facepiece respirators" Ann Occup Hyg 56(1): 92-101. doi:10.1093/annhyg/mer054 Ma, X., Zangmeister, C. D., Gigault, J., Mulholland, G. W. and Zachariah, M. R. (2013). "Soot aggregate restructuring during water processing" Journal of Aerosol Science 66: 209-219. doi:10.1016/j.jaerosci.2013.08.001 MacIntyre, C. R., Seale, H., Dung, T. C., Hien, N. T., Nga, P. T., Chughtai, A. A., Rahman, B., Dwyer, D. E. and Wang, Q. (2015). "A cluster randomised trial of cloth masks compared with medical masks in healthcare workers" BMJ Open 5(4): e006577. doi:10.1136/bmjopen-2014-006577 Mahdavi, A., Haghighat, F., Bahloul, A., Brochot, C. and Ostiguy, C. (2015). "Particle loading time and humidity effects on the efficiency of an N95 filtering facepiece

225 respirator model under constant and inhalation cyclic flows" Ann Occup Hyg 59(5): 629-640. doi:10.1093/annhyg/mev005 Massey, T., Borucki, M., Paik, S., Fuhrer, K., Bora, M., Kane, S., Haque, R. and Baxamusa, S. H. (2020). "Quantitative form and fit of N95 filtering facepiece respirators are retained and coronavirus surrogate is inactivated after heat treatments" medRxiv: 2020.2004.2015.20065755. doi:10.1101/2020.04.15.20065755 McDevitt, J., Rudnick, S., First, M. and Spengler, J. (2010). "Role of absolute humidity in the inactivation of influenza viruses on stainless steel surfaces at elevated temperatures" Appl Environ Microbiol 76(12): 3943-3947. doi:10.1128/AEM.02674-09 McMurry, P. H., Wang, X., Park, K. and Ehara, K. (2002). "The Relationship between Mass and Mobility for Atmospheric Particles: A New Technique for Measuring Particle Density" Aerosol Science and Technology 36(2): 227-238. doi:10.1080/027868202753504083 Mendelson, L. (2020). "Facing Your Face Mask Duties – A List of Statewide Orders, as of May 26, 2020." Retrieved May 26, 2020, May 26, 2020, from https://www.littler.com/publication-press/publication/facing-your-face-mask- duties-list-statewide-orders. Miguel, A. F. (2003). "Effect of air humidity on the evolution of permeability and performance of a fibrous filter during loading with hygroscopic and non- hygroscopic particles" Journal of Aerosol Science 34(6): 783-799. doi:10.1016/s0021-8502(03)00027-2 Mikhailov, E. F., Vlasenko, S. S., Kiselev, A. A. and Ryshkevitch, T. I. (1996). "Restructuring of soot particles: Experimental study" Journal of Aerosol Science 27: S711-S712. doi:10.1016/0021-8502(96)00428-4 Mills, D., Harnish, D. A., Lawrence, C., Sandoval-Powers, M. and Heimbuch, B. K. (2018). "Ultraviolet germicidal irradiation of influenza-contaminated N95 filtering facepiece respirators" Am J Infect Control 46(7): e49-e55. doi:10.1016/j.ajic.2018.02.018 Montgomery, J. F., Green, S. I. and Rogak, S. N. (2015a). "Impact of Relative Humidity on HVAC Filters Loaded with Hygroscopic and Non-Hygroscopic Particles" Aerosol Science and Technology 49(5): 322-331. doi:10.1080/02786826.2015.1026433 Montgomery, J. F., Rogak, S. N., Green, S. I., You, Y. and Bertram, A. K. (2015b). "Structural Change of Aerosol Particle Aggregates with Exposure to Elevated Relative Humidity" Environ Sci Technol 49(20): 12054-12061. doi:10.1021/acs.est.5b03157 Morawska, L. (2006). "Droplet fate in indoor environments, or can we prevent the spread of infection?" Indoor Air 16(5): 335-347. doi:10.1111/j.1600-0668.2006.00432.x Morawska, L., Johnson, G. R., Ristovski, Z. D., Hargreaves, M., Mengersen, K., Corbett, S., Chao, C. Y. H., Li, Y. and Katoshevski, D. (2009). "Size distribution and sites 226 of origin of droplets expelled from the human respiratory tract during expiratory activities" Journal of Aerosol Science 40(3): 256-269. doi:10.1016/j.jaerosci.2008.11.002 Morawska, L. and Cao, J. (2020). "Airborne transmission of SARS-CoV-2: The world should face the reality" Environ Int 139: 105730. doi:10.1016/j.envint.2020.105730 Morawska, L. and Milton, D. K. (2020). "It is Time to Address Airborne Transmission of COVID-19" Clin Infect Dis doi:10.1093/cid/ciaa939 Motyl, E. and Łowkis, B. (2006). "Effect of air humidity on charge decay and lifetime of PP electret nonwovens" Fibres and Textiles in Eastern Europe 14(5): 39-42. Mueller, A. V., Eden, M. J., Oakes, J. J., Bellini, C. and Fernandez, L. A. (2020). "Quantitative Method for Comparative Assessment of Particle Filtration Efficiency of Fabric Masks as Alternatives to Standard Surgical Masks for PPE" doi:10.1101/2020.04.17.20069567 N95DECON (2020). "A scientific consortium for data-driven study of N95 filtering facepiece respirator decontamination." from https://www.n95decon.org. National Academies of Sciences Engineering Medicine (2020). Rapid Expert Consultation on the Effectiveness of Fabric Masks for the COVID-19 Pandemic (April 8, 2020). Washington, DC, The National Academies Press. National Center for Biotechnology Information (2020a). "PubChem Compound Summary for CID 445639, Oleic acid." Retrieved Aug 14, 2020, from https://pubchem.ncbi.nlm.nih.gov/compound/Oleic-acid. National Center for Biotechnology Information (2020b). "PubChem Compound Summary for CID 743, Glutaric acid." Retrieved August 14, 2020, from https://pubchem.ncbi.nlm.nih.gov/compound/Glutaric-acid. National Health Commission (2020). "Guidelines for wearing masks for public." Retrieved May 26, 2020, from http://www.nhc.gov.cn/jkj/s3577/202003/0a472cc09e744144883db6a74fe6e760.s html. National Health Commission and State Administration of Traditional Chinese Medicine (2020). “Diagnosis and Treatment Protocol for Novel Coronavirus Pneumonia” NIOSH (2019). "Determination of particulate filter efficiency level for n95 series filters against solid particulates for non-powered, air- purifying respirators standard testing procedure (STP) Procedure No. TEB-APR-STP-0059", https://www.cdc.gov/niosh/npptl/stps/pdfs/TEB-APR-STP-0059-508.pdf O'Brien, F. E. M. (1948). "The Control of Humidity by Saturated Salt Solutions" Journal of Scientific Instruments 25(3): 73-76. doi:10.1088/0950-7671/25/3/305 Oberg, T. and Brosseau, L. M. (2008). "Surgical mask filter and fit performance" American Journal of Infection Control 36(4): 276-282. doi:10.1016/j.ajic.2007.07.008

227 Office of the Federal Register (1995). Approval of Respiratory Protective Devices, Code of Federal Regulations, Part 84, Title 42 Ollier, K., Leppänen, M., Wu, B., Yermakov, M., Newman, N. C., Reponen, T. and Grinshpun, S. A. (2019). "Inhalation Exposure and Respiratory Protection of Home Healthcare Workers Administering Aerosolized Medications (Simulation Study)" Aerosol and Air Quality Research 19(5): 937-944. doi:10.4209/aaqr.2018.10.0366 Onasch, T. B., Siefert, R. L., Brooks, S. D., Prenni, A. J., Murray, B., Wilson, M. A. and Tolbert, M. A. (1999). "Infrared spectroscopic study of the deliquescence and efflorescence of ammonium sulfate aerosol as a function of temperature" Journal of Geophysical Research: Atmospheres 104(D17): 21317-21326. doi:10.1029/1999jd900384 Orr, C., Hurd, F. K. and Corbett, W. J. (1958). "Aerosol size and relative humidity" Journal of Colloid Science 13(5): 472-482. doi:10.1016/0095-8522(58)90055-2 Ou, Q., Pei, C., Chen, S.-C. and Pui, D. Y. H. (2014). “HVAC and Engine In-take Filter Loading by Bimodal Aerosols” CFR Meeting Ou, Q., Stanley, N. and Pui, D. Y. H. (2015). “Effect of Relative Humidity on Filter Loading with ISO fine test dust” CFR Meeting Ou, Q., Maricq, M. M. and Pui, D. Y. H. (2017). "Evaluation of metallic filter media for sub-micrometer soot particle removal at elevated temperature" Aerosol Science and Technology 51(11): 1303-1312. doi:10.1080/02786826.2017.1349871 Ou, Q., Pei, C., Kim, S. C., Abell, E. and Pui, D. Y. H. (2020). "Evaluation of decontamination methods for commercial and alternative respirator and mask materials – view from filtration aspect" Journal of Aerosol Science 150: 105609. doi:10.1016/j.jaerosci.2020.105609 Park, K., Cao, F., Kittelson, D. B. and McMurry, P. H. (2003). "Relationship between Particle Mass and Mobility for Diesel Exhaust Particles" Environmental Science & Technology 37(3): 577-583. doi:10.1021/es025960v Parsons, M. T., Mak, J., Lipetz, S. R. and Bertram, A. K. (2004). "Deliquescence of malonic, succinic, glutaric, and adipic acid particles" Journal of Geophysical Research: Atmospheres 109(D06212): n/a-n/a. doi:10.1029/2003jd004075 Peckhaus, A., Grass, S., Treuel, L. and Zellner, R. (2012). "Deliquescence and efflorescence behavior of ternary inorganic/organic/water aerosol particles" J Phys Chem A 116(24): 6199-6210. doi:10.1021/jp211522t Pei, C., Ou, Q. and Pui, D. Y. H. (2019a). "Effect of relative humidity on loading characteristics of cellulose filter media by submicrometer potassium chloride, ammonium sulfate, and ammonium nitrate particles" Separation and Purification Technology 212: 75-83. doi:10.1016/j.seppur.2018.11.009 Pei, C., Ou, Q., Yu, T. and Pui, D. Y. H. (2019b). "Loading characteristics of nanofiber coated air intake filter media by potassium chloride, ammonium sulfate, and ammonium nitrate fine particles and the comparison with conventional cellulose

228 filter media" Separation and Purification Technology 228: 115734. doi:10.1016/j.seppur.2019.115734 Pei, C., Ou, Q., Kim, S. C., Chen, S.-C. and Pui, D. Y. H. (2020). "Alternative Face Masks Made of Common Materials for General Public: Fractional Filtration Efficiency and Breathability Perspective" Aerosol and Air Quality Research 20 doi:10.4209/aaqr.2020.07.0423 Pei, C., Ou, Q. and Pui, D. Y. H. (2021). "Effects of temperature and relative humidity on laboratory air filter loading test by hygroscopic salts" Separation and Purification Technology 255: 117679. doi:10.1016/j.seppur.2020.117679 Philip, S., Martin, R. V., van Donkelaar, A., Lo, J. W.-H., Wang, Y., Chen, D., Zhang, L., Kasibhatla, P. S., Wang, S., Zhang, Q., Lu, Z., Streets, D. G., Bittman, S. and Macdonald, D. J. (2014). "Global chemical composition of ambient fine particulate matter for exposure assessment" Environmental Science & Technology 48(22): 13060-13068. doi:10.1021/es502965b Popovicheva, O., Persiantseva, N. M., Shonija, N. K., DeMott, P., Koehler, K., Petters, M., Kreidenweis, S., Tishkova, V., Demirdjian, B. and Suzanne, J. (2008a). "Water interaction with hydrophobic and hydrophilic soot particles" Phys Chem Chem Phys 10(17): 2332-2344. doi:10.1039/b718944n Popovicheva, O. B., Persiantseva, N. M., Tishkova, V., Shonija, N. K. and Zubareva, N. A. (2008b). "Quantification of water uptake by soot particles" Environmental Research Letters 3(2): 025009. doi:10.1088/1748-9326/3/2/025009 Prather, K. A., Wang, C. C. and Schooley, R. T. (2020). "Reducing transmission of SARS- CoV-2" Science 368(6498): 1422-1424. doi:10.1126/science.abc6197 R Core Team (2019). R: A Language and Environment for Statistical Computing, R Foundation for Statistical Computing. Vienna, Austria. http://www.R-project.org/. Raynor, P. C. and Chae, S. J. (2004). "The long-term performance of electrically charged filters in a ventilation system" J Occup Environ Hyg 1(7): 463-471. doi:10.1080/15459620490467783 Rengasamy, S., Eimer, B. and Shaffer, R. E. (2010). "Simple Respiratory Protection— Evaluation of the Filtration Performance of Cloth Masks and Common Fabric Materials Against 20–1000 nm Size Particles" The Annals of Occupational Hygiene 54(7): 789-798. doi:10.1093/annhyg/meq044 Rengasamy, S. and Eimer, B. C. (2012). "Nanoparticle penetration through filter media and leakage through face seal interface of N95 filtering facepiece respirators" Ann Occup Hyg 56(5): 568-580. doi:10.1093/annhyg/mer122 Rengasamy, S., Shaffer, R., Williams, B. and Smit, S. (2017). "A comparison of facemask and respirator filtration test methods" J Occup Environ Hyg 14(2): 92-103. doi:10.1080/15459624.2016.1225157

229 Ribeyre, Q., Charvet, A., Vallières, C. and Thomas, D. (2017). "Impact of relative humidity on a nanostructured filter cake – Experimental and modelling approaches" Chemical Engineering Science 161: 109-116. doi:10.1016/j.ces.2016.12.013 RIDGID (2020). "RIDGID’s important notice." Retrieved Jun 2, 2020, from https://www.ridgid.com/us/en/vacuum-filters. Rockland, L. B. (1960). "Saturated Salt Solutions for Static Control of Relative Humidity between 5° and 40° C" Analytical Chemistry 32(10): 1375-1376. doi:10.1021/ac60166a055 Rogge, W. F., Hildemann, L. M., Mazurek, M. A., Cass, G. R. and Simoneit, B. R. T. (1991). "Sources of fine organic aerosol. 1. Charbroilers and meat cooking operations" Environmental Science & Technology 25(6): 1112-1125. doi:10.1021/es00018a015 RStudio Team (2019). RStudio: Integrated Development Environment for R, RStudio, PBC. Boston, MA URL http://www.rstudio.com/. Saathoff, H., Naumann, K. H., Schnaiter, M., Schöck, W., Möhler, O., Schurath, U., Weingartner, E., Gysel, M. and Baltensperger, U. (2003). "Coating of soot and (NH4)2SO4 particles by ozonolysis products of α-pinene" Journal of Aerosol Science 34(10): 1297-1321. doi:10.1016/s0021-8502(03)00364-1 Scharfman, B. E., Techet, A. H., Bush, J. W. M. and Bourouiba, L. (2016). "Visualization of sneeze ejecta: steps of fluid fragmentation leading to respiratory droplets" Exp Fluids 57(2): 24. doi:10.1007/s00348-015-2078-4 Schnitzler, E. G., Dutt, A., Charbonneau, A. M., Olfert, J. S. and Jager, W. (2014). "Soot aggregate restructuring due to coatings of secondary organic aerosol derived from aromatic precursors" Environ Sci Technol 48(24): 14309-14316. doi:10.1021/es503699b Snider, G., Weagle, C. L., Martin, R. V., van Donkelaar, A., Conrad, K., Cunningham, D., Gordon, C., Zwicker, M., Akoshile, C., Artaxo, P., Anh, N. X., Brook, J., Dong, J., Garland, R. M., Greenwald, R., Griffith, D., He, K., Holben, B. N., Kahn, R., Koren, I., Lagrosas, N., Lestari, P., Ma, Z., Vanderlei Martins, J., Quel, E. J., Rudich, Y., Salam, A., Tripathi, S. N., Yu, C., Zhang, Q., Zhang, Y., Brauer, M., Cohen, A., Gibson, M. D. and Liu, Y. (2015). "SPARTAN: a global network to evaluate and enhance satellite-based estimates of ground-level particulate matter for global health applications" Atmospheric Measurement Techniques 8(1): 505- 521. doi:10.5194/amt-8-505-2015 Snider, G., Weagle, C. L., Murdymootoo, K. K., Ring, A., Ritchie, Y., Stone, E., Walsh, A., Akoshile, C., Anh, N. X., Balasubramanian, R., Brook, J., Qonitan, F. D., Dong, J., Griffith, D., He, K., Holben, B. N., Kahn, R., Lagrosas, N., Lestari, P., Ma, Z., Misra, A., Norford, L. K., Quel, E. J., Salam, A., Schichtel, B., Segev, L., Tripathi, S., Wang, C., Yu, C., Zhang, Q., Zhang, Y., Brauer, M., Cohen, A., Gibson, M. D., Liu, Y., Martins, J. V., Rudich, Y. and Martin, R. V. (2016). "Variation in global chemical composition of PM2.5: emerging results from SPARTAN" Atmospheric Chemistry and Physics 16(15): 9629-9653. doi:10.5194/acp-16-9629-2016 230 Spurny, K. R. (1986). "On the filtration of fibrous aerosols" Science of the Total Environment 52(3): 189-199. doi:10.1016/0048-9697(86)90119-1 Sun, J., Liu, L., Xu, L., Wang, Y., Wu, Z., Hu, M., Shi, Z., Li, Y., Zhang, X., Chen, J. and Li, W. (2018). "Key Role of Nitrate in Phase Transitions of Urban Particles: Implications of Important Reactive Surfaces for Secondary Aerosol Formation" Journal of Geophysical Research: Atmospheres 123(2): 1234-1243. doi:10.1002/2017jd027264 Sun, Y., Zhuang, G., Wang, Y., Han, L., Guo, J., Dan, M., Zhang, W., Wang, Z. and Hao, Z. (2004). "The air-borne particulate pollution in Beijing—concentration, composition, distribution and sources" Atmospheric Environment 38(35): 5991- 6004. doi:10.1016/j.atmosenv.2004.07.009 Sun, Z., Liang, Y., He, W., Jiang, F., Song, Q., Tang, M. and Wang, J. (2019). "Filtration performance and loading capacity of nano-structured composite filter media for applications with high soot concentrations" Separation and Purification Technology 221: 175-182. doi:10.1016/j.seppur.2019.03.087 Tang, I. N. (1976). "Phase transformation and growth of aerosol particles composed of mixed salts" Journal of Aerosol Science 7(5): 361-371. doi:10.1016/0021- 8502(76)90022-7 Tang, I. N. (1980). Deliquescence properties and particle size change of hygroscopic aerosols In: Generation of aerosols and facilities for exposure experiments. K. Willeke and N. Kagakkai, Ann Arbor Science. Tang, I. N. and Munkelwitz, H. R. (1993). "Composition and temperature dependence of the deliquescence properties of hygroscopic aerosols" Atmospheric Environment. Part A. General Topics 27(4): 467-473. doi:10.1016/0960-1686(93)90204-c Tang, I. N. and Munkelwitz, H. R. (1994a). "Aerosol Phase Transformation and Growth in the Atmosphere" Journal of Applied Meteorology 33(7): 791-796. doi:10.1175/1520-0450(1994)033<0791:Aptagi>2.0.Co;2 Tang, I. N. and Munkelwitz, H. R. (1994b). "Water activities, densities, and refractive indices of aqueous sulfates and sodium nitrate droplets of atmospheric importance" Journal of Geophysical Research 99(D9): 18801-18808. doi:10.1029/94jd01345 Tang, I. N. (1996). "Chemical and size effects of hygroscopic aerosols on light scattering coefficients" Journal of Geophysical Research: Atmospheres 101(D14): 19245- 19250. doi:10.1029/96jd03003 Tang, I. N., Tridico, A. C. and Fung, K. H. (1997). "Thermodynamic and optical properties of sea salt aerosols" Journal of Geophysical Research: Atmospheres 102(D19): 23269-23275. doi:10.1029/97jd01806 ten Brink, H. M. and Veefkind, J. P. (1995). "Humidity dependence of the light-scattering by ammonium nitrate" Journal of Aerosol Science 26: S553-S554. doi:10.1016/0021-8502(95)97184-g

231 Thomson, W. (1871). "LX. On the equilibrium of vapour at a curved surface of liquid" The London, Edinburgh, and Dublin Philosophical Magazine and Journal of Science 42(282): 448-452. doi:10.1080/14786447108640606 Tritscher, T., Jurányi, Z., Martin, M., Chirico, R., Gysel, M., Heringa, M. F., DeCarlo, P. F., Sierau, B., Prévôt, A. S. H., Weingartner, E. and Baltensperger, U. (2011). "Changes of hygroscopicity and morphology during ageing of diesel soot" Environmental Research Letters 6(3): 034026. doi:10.1088/1748-9326/6/3/034026 University of Florida, Department of Anesthesiology (2020). "Mask alternative." from https://anest.ufl.edu/clinical-divisions/mask-alternative/. University of Minnesota, College of Science & Engineering. (2020). "U of M researchers design respirator mask prototypes from filter technology." from https://cse.umn.edu/college/news/u-m-researchers-design-respirator-mask- prototypes-filter-technology. van der Sande, M., Teunis, P. and Sabel, R. (2008). "Professional and home-made face masks reduce exposure to respiratory infections among the general population" PLoS One 3(7): e2618. doi:10.1371/journal.pone.0002618 Viscusi, D. J., Bergman, M. S., Eimer, B. C. and Shaffer, R. E. (2009). "Evaluation of five decontamination methods for filtering facepiece respirators" Ann Occup Hyg 53(8): 815-827. doi:10.1093/annhyg/mep070 Viscusi, D. J., Bergman, M. S., Novak, D. A., Faulkner, K. A., Palmiero, A., Powell, J. and Shaffer, R. E. (2011). "Impact of three biological decontamination methods on filtering facepiece respirator fit, odor, comfort, and donning ease" J Occup Environ Hyg 8(7): 426-436. doi:10.1080/15459624.2011.585927 Walsh, D. C., Stenhouse, J. I. T., Scurrah, K. L. and Graef, A. (1996). "The effect of solid and liquid aerosol particle loading on fibrous filter material performance" Journal of Aerosol Science 27: S617-S618. doi:10.1016/0021-8502(96)00381-3 Wang, M., Zhu, T., Zhao, D., Rubach, F., Wahner, A., Kiendler-Scharr, A. and Mentel, T. F. (2018). "Cloud condensation nuclei activity of CaCO3 particles with oleic acid and malonic acid coatings" Atmospheric Chemistry and Physics 18(10): 7345- 7359. doi:10.5194/acp-18-7345-2018 Wang, Q. and Yu, C. (2020). "The role of masks and respirator protection against SARS- CoV-2" Infection control and hospital epidemiology 41(6): 746-747. doi:10.1017/ice.2020.83 Wang, Y., Lin, Z. and Zhang, W. (2020). "Interaction between high humidity and hygroscopic aerosol on the filtration performance of electret media" Particulate Science and Technology: 1-13. doi:10.1080/02726351.2020.1799125 Weingartner, E., Baltensperger, U. and Burtscher, H. (1995). "Growth and structural change of combustion aerosols at high relative humidity" Environ Sci Technol 29(12): 2982-2986. doi:10.1021/es00012a014

232 Weingartner, E., Burtscher, H. and Baltensperger, U. (1997). "Hygroscopic properties of carbon and diesel soot particles" Atmospheric Environment 31(15): 2311-2327. doi:10.1016/s1352-2310(97)00023-x Weingartner, E., Gysel, M. and Baltensperger, U. (2002). "Hygroscopicity of aerosol particles at low temperatures. 1. New low-temperature H-TDMA instrument: setup and first applications" Environ Sci Technol 36(1): 55-62. doi:10.1021/es010054o Wexler, A. and Hasegawa, S. (1954). "Relative humidity-temperature relationships of some saturated salt solutions in the temperature range 0 degree to 50 degrees C" Journal of Research of the National Bureau of Standards 53(1): 19-26. doi:10.6028/jres.053.003 Wexler, A. S. and Seinfeld, J. H. (1991). "Second-generation inorganic aerosol model" Atmospheric Environment. Part A. General Topics 25(12): 2731-2748. doi:10.1016/0960-1686(91)90203-j WHO (2020a). "Coronavirus disease (COVID-19) Pandemic." Retrieved May 26, 2020, from https://www.who.int/emergencies/diseases/novel-coronavirus-2019. WHO (2020b). "Coronavirus disease (COVID-19) advice for the public: When and how to use masks." Retrieved Jun 10, 2020, from https://www.who.int/emergencies/diseases/novel-coronavirus-2019/advice-for- public/when-and-how-to-use-masks. WHO (2020c). "Modes of transmission of virus causing COVID-19:implications for IPC precaution recommendations." Retrieved May 26,2020, from https://www.who.int/publications-detail/modes-of-transmission-of-virus-causing- covid-19-implications-for-ipc-precaution-recommendations. WHO (2020d). "Advice on the use of masks in the context of COVID-19 Interim guidance." Retrieved 5 June, 2020, from https://apps.who.int/iris/bitstream/handle/10665/332293/WHO-2019-nCov- IPC_Masks-2020.4-eng.pdf. WHO (2020e). "Shortage of personal protective equipment endangering health workers worldwide." Retrieved Aug 1, 2020, from https://www.who.int/news- room/detail/03-03-2020-shortage-of-personal-protective-equipment-endangering- health-workers-worldwide. Wiedensohler, A. (1988). "An approximation of the bipolar charge distribution for particles in the submicron size range" Journal of Aerosol Science 19(3): 387-389. doi:10.1016/0021-8502(88)90278-9 Wu, L., Li, X. and Ro, C. U. (2019). "Hygroscopic Behavior of Ammonium Sulfate, Ammonium Nitrate, and their Mixture Particles" Asian Journal of Atmospheric Environment 13(3): 196-211. doi:10.5572/ajae.2019.13.3.196 Wu, Y., Cheng, T., Liu, D., Allan, J. D., Zheng, L. and Chen, H. (2018). "Light Absorption Enhancement of Black Carbon Aerosol Constrained by Particle Morphology" Environmental Science & Technology 52(12): 6912-6919. doi:10.1021/acs.est.8b00636 233 Xu, J., Imre, D., McGraw, R. and Tang, I. (1998). "Ammonium Sulfate: Equilibrium and Metastability Phase Diagrams from 40 to −50 °C" The Journal of Physical Chemistry B 102(38): 7462-7469. doi:10.1021/jp981929x Xue, H., Khalizov, A. F., Wang, L., Zheng, J. and Zhang, R. (2009). "Effects of Coating of Dicarboxylic Acids on the Mass−Mobility Relationship of Soot Particles" Environmental Science & Technology 43(8): 2787-2792. doi:10.1021/es803287v Yang, C. (2012). "Aerosol Filtration Application Using Fibrous Media—An Industrial Perspective" Chinese Journal of Chemical Engineering 20(1): 1-9. doi:10.1016/s1004-9541(12)60356-5 Yang, F., Tan, J., Zhao, Q., Du, Z., He, K., Ma, Y., Duan, F., Chen, G. and Zhao, Q. (2011). "Characteristics of PM2.5 speciation in representative megacities and across China" Atmospheric Chemistry and Physics 11(11): 5207-5219. doi:10.5194/acp- 11-5207-2011 Yang, L., Pabalan, R. T. and Juckett, M. R. (2006). "Deliquescence Relative Humidity Measurements Using an Electrical Conductivity Method" Journal of Solution Chemistry 35(4): 583-604. doi:10.1007/s10953-005-9015-8 Yao, X., Chan, C. K., Fang, M., Cadle, S., Chan, T., Mulawa, P., He, K. and Ye, B. (2002). "The water-soluble ionic composition of PM2.5 in Shanghai and Beijing, China" Atmospheric Environment 36(26): 4223-4234. doi:10.1016/s1352-2310(02)00342- 4 Yu, I. T., Li, Y., Wong, T. W., Tam, W., Chan, A. T., Lee, J. H., Leung, D. Y. and Ho, T. (2004). "Evidence of airborne transmission of the severe acute respiratory syndrome virus" N Engl J Med 350(17): 1731-1739. doi:10.1056/NEJMoa032867 Yuan, Q., Xu, J., Wang, Y., Zhang, X., Pang, Y., Liu, L., Bi, L., Kang, S. and Li, W. (2019). "Mixing State and Fractal Dimension of Soot Particles at a Remote Site in the Southeastern Tibetan Plateau" Environmental Science & Technology 53(14): 8227- 8234. doi:10.1021/acs.est.9b01917 Zhang, D. and Zhang, R. (2005). "Laboratory Investigation of Heterogeneous Interaction of Sulfuric Acid with Soot" Environmental Science & Technology 39(15): 5722- 5728. doi:10.1021/es050372d Zhang, Q., Jimenez, J. L., Canagaratna, M. R., Allan, J. D., Coe, H., Ulbrich, I., Alfarra, M. R., Takami, A., Middlebrook, A. M., Sun, Y. L., Dzepina, K., Dunlea, E., Docherty, K., DeCarlo, P. F., Salcedo, D., Onasch, T., Jayne, J. T., Miyoshi, T., Shimono, A., Hatakeyama, S., Takegawa, N., Kondo, Y., Schneider, J., Drewnick, F., Borrmann, S., Weimer, S., Demerjian, K., Williams, P., Bower, K., Bahreini, R., Cottrell, L., Griffin, R. J., Rautiainen, J., Sun, J. Y., Zhang, Y. M. and Worsnop, D. R. (2007). "Ubiquity and dominance of oxygenated species in organic aerosols in anthropogenically-influenced Northern Hemisphere midlatitudes" Geophysical Research Letters 34(13): L13801. doi:10.1029/2007gl029979 Zhang, R., Khalizov, A. F., Pagels, J., Zhang, D., Xue, H. and McMurry, P. H. (2008). "Variability in morphology, hygroscopicity, and optical properties of soot aerosols 234 during atmospheric processing" Proceedings of the National Academy of Sciences 105(30): 10291. doi:10.1073/pnas.0804860105 Zhao, M., Liao, L., Xiao, W., Yu, X., Wang, H., Wang, Q., Lin, Y. L., Kilinc-Balci, F. S., Price, A., Chu, L., Chu, M. C., Chu, S. and Cui, Y. (2020). "Household Materials Selection for Homemade Cloth Face Coverings and Their Filtration Efficiency Enhancement with Triboelectric Charging" Nano Lett 20(7): 5544-5552. doi:10.1021/acs.nanolett.0c02211 Zhou, Y. and Cheng, Y. S. (2017). "Evaluation of N95 Filtering Facepiece Respirators Challenged with Engineered Nanoparticles" Aerosol and Air Quality Research 16(1): 212-220. doi:10.4209/aaqr.2015.06.0399 Zhu, N., Zhang, D., Wang, W., Li, X., Yang, B., Song, J., Zhao, X., Huang, B., Shi, W., Lu, R., Niu, P., Zhan, F., Ma, X., Wang, D., Xu, W., Wu, G., Gao, G. F., Tan, W. and China Novel Coronavirus Investigating Research Team (2020). "A Novel Coronavirus from Patients with Pneumonia in China, 2019" N Engl J Med 382(8): 727-733. doi:10.1056/NEJMoa2001017 Zuo, Z., Kuehn, T. H. and Pui, D. Y. H. (2013). "Performance evaluation of filtering facepiece respirators using virus aerosols" American Journal of Infection Control 41(1): 80-82. doi:10.1016/j.ajic.2012.01.010

235 Appendix

This appendix contains copies of copyright permissions granted by the owners of the journals where the work is published.

236 237

238