1 Quantifying use of lethal ZnCl2 on Black Lives Matter demonstrators by United States Homeland Security 1 2 Juniper L. Simonis1 3 4 1DAPPER Stats, Portland, Oregon, United States of America 5 simonis@dapperstats.com 6 ORCID: 0000-0001-9798-0460 7 8 Short title: US Homeland Security uses lethal gas on demonstrators 9 mailto:simonis@dapperstats.com 2 ABSTRACT 10 11 Law enforcement’s use of chemical weapons is a threat to human and environmental health, 12 exemplified during 2020 racial justice protests in Portland, Oregon, USA. In July, US Department of 13 Homeland Security (DHS) agents used an exceptionally toxic and unknown weapon to quell free speech in 14 support of Black lives and against federal presence. With significant help from the community, I combined 15 first-hand accounts, videos and photos of munitions, primary literature, and analytical chemistry to identify 16 the weapon as gaseous ZnCl2 from Hexachloroethane (HC) “smoke” grenades. Using hierarchical Bayesian 17 methods, I estimated that DHS deployed 26 (25 – 30; 95% CI) HC grenades. Given the toxicity of ZnCl2, 18 that many canister could have killed hundreds of people. Although no fatalities were reported, the exposed 19 population experienced acute, delayed, and persistent health issues. DHS’s wanton use of ZnCl2 will have 20 lasting impacts and was identified through a community standing up for racial justice. 21 22 INTRODUCTION 23 24 “The use of poison in any manner, be it to poison wells, or food, or arms, is wholly excluded from modern 25 warfare. He that uses it puts himself out of the pale of the law and usages of war.” 26 General Orders No. 100, Article 70, signed President Abraham Lincoln 18631 27 28 Following the killing of George Floyd by Minneapolis police officers in Minnesota on 25 May 2020, 29 Black Lives Matter (BLM) protesters took to the streets around the world to demand justice2. In solidarity 30 with Minneapolis, BLM demonstrations in Portland, Oregon began on 27 May3 and continued for over a 31 hundred days, only interrupted by hazardous wildfire smoke4,5. In response to these protests, city, county, 32 state, and federal law enforcement agencies deployed a variety of chemical weapons such as 2-33 chlorobenzalmalononitrile (CS), oleoresin capsicum (OC) and caustic munitions smokes2,4-7, building upon a 34 legacy of chemical weapons usage by Portland Police Bureau2. Indeed, since the start of the 2020 BLM 35 3 protests, Portland has not only had the most number of publicly reported instances of police violence at BLM 36 protests among US cities (403 of 1,281; 31.5%) but also over a third of all such instances involving chemical 37 weapons (180 of 522; 34.4%)6, despite only having 0.2% of the total US population (Fig.1)6,8. Seattle, 38 Washington (1.2 times the population of Portland) had the next most number of incidents involving chemical 39 weapons by police at BLM protests: 40, which is less than 25% of Portland’s 180 such incidents (Fig.1)6 40 41 Chemical weapons were outlawed for American wartime use by Abraham Lincoln via the Lieber 42 Code in 18631 and internationally in the Hague Conventions of 1899 and 19079,10 and the Geneva Protocol of 43 192811, yet they have a long history of domestic use by law enforcement against civilians to quell unrest12-14. 44 This is despite chemical weapons being fundamentally indiscriminate, often deployed against specifications, 45 and lethal12-14. In the United States, the use of chemical weapons exacerbates systemic inequities and limits 46 constitutionally protected expression of speech and assembly2,7,14. In Portland, Oregon, law enforcement has 47 a well-documented history of only deploying chemical weapons to prevent free speech in support of Black 48 lives and racial justice; no such actions are taken against gatherings of recognized white supremacist hate 49 groups2, even when occurring on the same day15. 50 51 During the second half of July, as interest in Portland’s BLM protests was resurging and focused 52 particularly on federal presence (Fig. 2; Dataset S1)4, agents of the US Department of Homeland Security’s 53 (DHS) new Protecting American Cities Task Force (PACTF)16,17 deployed deadly gaseous Zinc Chloride 54 (ZnCl2) via Hexachloroethane (HC) Smoke grenades (Figs. 3, S1-S24) during Operation Diligent Valor18. At 55 the time, ZnCl2 was not a familiar weapon to demonstrators nor were HC cans among any munitions 56 recovered in the previous two months of protests (Fig. S25). And as such, concerning new heath symptoms 57 (severe headache, sudden vomiting, chemical skin rash, loss of hair and nails, etc.) that began appearing at the 58 same time could not be attributed to a causal agent. Indeed, it took an incredible effort (summarized herein) 59 by frontline journalists, scientists, community leaders, legal observers, medics, and protesters to document the 60 munitions so that mass severe health symptoms could be definitively tied to HC used by DHS, and produced 61 4 by Defense Technology (DT), a subsidiary of The Safariland Group (Fig. S26). Hampering this connection 62 were [1] DHS’s lack of admission to using HC during disclosure of their arsenal to the City of Portland, when 63 answering journalists, and in response to Freedom of Information Act (FOIA) requests (E. Holmes, personal 64 observation)19,20 and [2] DT’s failure to mention ZnCl2 or its human and environmental health impacts on its 65 product specification page (Fig. S26) or Safety Data Sheet21. 66 67 Hexachloroethane is a munitions “smoke” developed in the early 1930s by the US Army Chemical 68 Warfare Service that was understood by the mid-1940s to be a poisonous chemical agent22-24 and which has 69 since been replaced throughout the US Armed Services25. HC itself is listed as hazardous by the International 70 Agency for Research on Cancer, and the US Environmental Protection Agency, Department of 71 Transportation, Occupational Safety and Health Administration, American Conference of Governmental 72 Industrial Hygienists, National Institute for Occupational Safety and Health, and National Toxicology 73 Program26 and has significant human and environmental health consequences24. A more dire result of the use 74 of HC grenades, however, is that they produce a high volume (> 75% of all products w/w) of gaseous ZnCl2, 75 a lethal vapor, during the focal reaction27-29: 76 77 𝐶2𝐶𝑙6 + 3𝑍𝑛𝑂 → 3𝑍𝑛𝐶𝑙2 + 𝐴𝑙2𝑂3 + 𝐶 (Eq. 1). 78 79 Indeed, despite being called Hexachloroethane Smoke grenades, the goal in using them is to produce gaseous 80 ZnCl2, which refracts light and thus creates a “smoke”. Additionally, due to the high heat energy of the 81 reaction (1000 – 2000 C), many noxious gaseous byproducts are created, most notably carbon monoxide 82 (CO), phosgene (COCl2), hexachlorobenzene (C6Cl6), tetrachloroethene (C2Cl4), carbon tetrachloride (CCl4), 83 hydrogen chloride (HCl), and chlorine (Cl2)27-29. 84 85 Hundreds of cases of toxicity from HC smoke have been documented across the intervening decades 86 since its development, showing a range of significant symptoms including immediate dyspnea, coughing, 87 5 lacrimation, chest pain, vomiting, nausea, and mucosal irritation; delayed and prolonged inflammation of skin 88 and internal organs as well as tachycardia; chronic genotoxicity of the bronchial epithelium; and an average 89 fatality rate of 14% among case clusters30. HC smoke has further significant effects on the environment, 90 including defoliation and long-term reduction in tree growth31,32 and stunted development, scale deterioration, 91 skeletal weakness, and bioaccumulation in fish33-35. This is of particular note, as HC grenades were deployed 92 in the catchment of Portland’s untreated stormwater system that outfalls directly into nesting and rearing 93 habitats of salmonids. Given the lethality of its products, the wanton use of HC by DHS in Portland is 94 alarming and warrants significant further investigation. 95 96 The goal of this study was therefore to quantify use of HC by DHS to provide a basis for estimating 97 human and environmental impacts. To accomplish this, I combined multiple data streams of observations 98 (visual/photographic confirmations, recovered canisters) on HC use into a single Bayesian hierarchical 99 model36-38 fit using a Gibbs sampler39. I then sampled the protest environment (soil, plants, clothing, 100 canisters, ground, tent) for signatures of HC use (Zinc, Hexachloroethane, other chlorinated hydrocarbons) 101 using standard analytical chemistry methods. Such an exercise would not be necessary if DHS were to release 102 actual chemical weapons deployment data. Given the lack of transparency regarding chemical weapons use by 103 all law enforcement agencies in Portland, however, including retrieval of canisters to prevent identification 104 and shooting those who touch canisters40,41 estimation is a critical starting point on the road to understanding 105 the scope and scale of HC’s impacts. 106 107 RESULTS 108 109 Over the course of July 2020, DHS deployed an estimated 26 (25 - 30, 95% posterior interval) 110 grenades of hexachloroethane in the focal protest area in downtown Portland Oregon, specifically in the 111 immediate vicinity of the Wyatt Federal Building and Hatfield Federal Courthouse (Fig. 5, Table 1). Twenty 112 grenades were recovered, five more were observed being deployed by agents but not recovered (Figs. S1-S24, 113 6 Table S1), and 1 (0 – 5, 95% posterior interval) was estimated to not have been observed or recovered. The 114 estimated rate of HC grenade deployment (λ) by DHS during July was 0.12 grenades per hour of federal 115 agents on the street (0.03 – 0.39, 95% posterior interval; Fig. 5, Table 1). The rate of recovery (ρ) was 0.73 116 (0.57 – 0.85, 95% posterior interval), notably higher than the visual observation rate (ν; 0.50, 0.32 – 0.68, 95% 117 posterior; Table 1). 118 119 The Gibbs sampler efficiently sampled and effectively searched the joint posterior distribution (Eqs. 120 2-4). Convergence was high among the parallel chains: the potential scale reduction factors (psrf, a.k.a. 121 Gelman-Rubin statistic)36 being all ~1.0 (Table 1). All parameters exhibited very small MCMC 122 autocorrelations (~0.0) and had resultingly large effective sample sizes (Table 1). 123 124 Translation of the total estimated HC deployment to ZnCl2 gas produced18,26 using published lethal 125 doses42 and weights43 shows that hundreds of fatalities could have occurred (median: 235, 95% posterior 126 interval: 156 – 306), although there was large uncertainty due to imprecise LD50 measurements and weight 127 variation among individuals. While the canisters were deployed outside, which certainly prevented many 128 deaths, diffusion was limited by crowds of thousands of people (Fig. 2), closed tree canopies, cars, and tents 129 (Simonis, personal observation). Indeed, the off-gassing ZnCl2 presented significant risks to individuals in the 130 vicinity as evidenced by high levels of zinc in environmental samples (Dataset 2)42; immediate45,46, delayed47,48, 131 and chronic49,50 health symptoms; and odors detectable miles away51. 132 133 Of particular note from the environmental chemistry samples was a hexachloroethane canister 134 recovered after deployment by DHS (Figs. 3(b,c,d), 4, S16, Video S1), from which I sampled solid residue. 135 Ion Chromatography and Gas Chromatography/Mass Spectrometry (GC-MS) identified that the residue was 136 27% Zinc w/w and contained hexachloroethane, identifying that the munition was not fully spent (Dataset 137 S3). The grenade also contained tetrachloroethene, benzene, toluene, phthalic anhydride, Chromium, and 138 Lead (Dataset S2), all of which pose significant human and environmental health risks24,30. The spread of 139 7 ZnCl2 through the protest area and beyond was shown through all environmental samples having significant 140 concentrations of Zinc (Dataset S2). Perhaps the most notable of which was the organic vapor filter worn by 141 a medic outside of the plume on the far side of the protest area (~300 yd or 275 m from the main release 142 points) which contained Zinc (made gaseous as ZnCl2), yet no Chromium or Lead (neither of which were 143 made gaseous), as well as phthalic anhydride, toluene, and xylene (Dataset S2). 144 145 DISCUSSION 146 147 Under ideal conditions in a wide-open field at night, the concentration of ZnCl2 produced by a 148 typical HC grenade is high enough that an unmasked individual 200 yd (three city blocks in Portland, 183 m) 149 from detonation has a maximum of 24 min of safety before significant acute health symptoms appear22. An 150 individual a 1,000 yd (0.9 km) away is still at risk and only has 2.5 h22 It is unclear how ZnCl2 dissipates 151 through a densely-gassed, tree-lined urban landscape within a river valley like Portland52, but reported signs 152 and symptoms indicate that it spread widely, entered the stormwater system that flows to the Willamette 153 River, and passed through or around protective equipment worn by journalists, protesters, medics, legal 154 observers, and bystanders45-51. 155 156 The impact of ZnCl2’s novelty cannot be overstated, as both veteran and newer demonstrators, press, 157 medics, and legal observers were unprepared for this weapon specifically. Virtually all existing chemical 158 weapons seen prior to HC’s use produced liquid or solid particles, despite being called “tear gas”, that could 159 be removed from air using particle filters such as respirators. As such, many individuals had insufficient 160 filtration to remove gaseous ZnCl2. Only those with filters designed for gases were able to limit inhalation, as 161 evidenced by the Zinc found on the medic filter (Dataset S2) and first-hand accounts of individuals with gas 162 masks being less affects (Simonis, personal observation). Even when using a properly fitting gas mask, 163 however, ZnCl2 is absorbed into the body via dermal uptake22-24,30. Further, given its capacity to 164 bioaccumulate and cause delayed severe inflammation responses, ZnCl2 exposure is measured cumulatively 165 8 over 10 d22,30, a significant departure from other presently used chemical weapons12-14. Despite these life-166 threatening differences with HC, the public was never informed of DHS’s use of the weapon by any agency 167 (federal or otherwise). Indeed, DHS has continued to deny using HC (E. Holmes, personal observation), 168 despite overwhelming evidence to the contrary (Figs. S1-24, Video S1, Dataset S2)19,20 and even a more recent 169 (29 October 2020) deployment of HC by DHS in front of a property rented by Immigrations and Customs 170 Enforcement (ICE) in South Portland53,54, which again caused severe health symptoms in the exposed 171 population52. 172 173 As a highly mobile and poisonous gas that lacks an odor itself, ZnCl2 poses a significant risk to 174 humans as well as the environment22,30,31,35. Building upon a legacy of resistance to excessive police force2,3, a 175 community of protesters, activists, journalists, legal observers, and scientists standing up for Black lives 176 documented its use and are just beginning to understand its impacts on the residents and environment of 177 Portland. Human health and environmental impact studies are urgently needed to grasp the full impact of 178 DHS’s literal salting of the earth using Hexachloroethane smoke grenades in Portland, OR. 179 180 METHODS 181 182 Incidents of Police Brutality at 2020 BLM Protests 183 184 I collected all incidents from the Police Brutality 2020 database6, which collates publicly reported 185 events of police violence associated with the 2020 Black Lives Matter (BLM) protests. I used their RESTful 186 API (https://api.846policebrutality.com/api/incidents) via R v3.6.355 with the httr v1.4.156 and jsonlite 187 v1.6.157 packages. I collected all incidents during the 25 May to 31 December 2020 interval, and specifically 188 filtered Portland, Oregon incidents. I also filtered the total and Portland incidents by chemical weapons tags 189 using entries marked “gas”, “marking-round”, “pepper-ball”, “pepper-spray”, “spray”, “tear-gas”, or “tear-190 gas-canister”, and then simply divided to determine ratios. 191 9 192 Portland Protest Data 193 194 I used timestamps from videos and augmented them with time-stamped or time-noting tweets as 195 needed, to calculate the amount of time each night during July 2020 that the federal agents were outside of 196 their buildings58,59. Individual jaunts by federal agents are detailed in Dataset S3 and collated daily-level data 197 are in Dataset S1. I used crowd size estimates provided by the news and social media aggregating site The 198 Recompiler, which has a project (RE: Portland) that has collected documents of the Portland BLM protests 199 since the end of June 2020, prior to the arrival of the DHS agents59. On nights where a range of crowd sizes 200 was given, the midpoint was used. See Dataset S1 for data used in analyses. 201 202 Munition Identification 203 204 Protesters, concerned civilians, medics, legal observers, trash cleaners, scientists, and neighbors have 205 been collecting munitions (e.g., Figs S1-25) since the beginning days of the protest. I leveraged this 206 specifically to address the uncertainty around Hexachloroethane (HC) usage by connecting with the network 207 of individuals already collecting and documenting munitions and notifying them of the particular can types of 208 interest. In addition, I put out specific calls publicly for submissions of photos via Twitter and the Chemical 209 Weapons Research Consortium website (https://chemicalweaponsresearch.com) via secure email and secure 210 form, which have yielded dozens of submissions, but no additional HC cans beyond those included (Simonis, 211 personal observation). The avenues remain open to submissions and I will update the Dataset used here with 212 any further HC cans. I also watched dozens of hours of footage58,59 and read through aggregated news and 213 tweets59 to investigate potential other deployments. Because HC cans are so distinctive when deploying and 214 afterwards (Figs. 3, S1-S25, but note Fig. S9; Video S 1), it was possible to retroactively evaluate 215 documentation and collections of munitions to enumerate the HC cans. If a can spewed sparks, off-gassed 216 white/grey/black “smoke”, and glowed and burned hot and long (~2 minutes), it was considered an HC can, 217 10 due to the distinctive nature of its incendiary aspects (Fig. S27). Further, if any recovered can was so corroded 218 to be illegible and was of the distinctive size of the HC cans (Figs. S1-24), it was considered as such. In the 219 instance of the grenade in Fig. S9, it was identified as HC due to its clear labeling. 220 221 Bayesian Model 222 223 Having evaluated a large volume of photographic, video, and print media, I identified deployments of 224 hexachloroethane (HC) grenades and recovery of munitions during July 2020 (Figs. S1-S24). I also estimated 225 the time federal agents were out of their buildings and crowd size for each day from the media compilation 226 (Fig. 2, Datasets S1,S3). I combined the visual confirmation of HC deployment and recovery of HC canisters 227 with the time federal agents were out of their buildings using a hierarchical Bayesian model to infer the 228 underlying unknown number of canisters deployed by the Department of Homeland Security (DHS) on a 229 given day (𝑑𝑖 ) and over all days (𝐷 = ∑ 𝑑𝑖 )36-38, where the 𝑖 subscript represents day. 230 231 The hourly rate of deployment for that day (𝜆𝑖) is a log-linear (to handle Poisson response) function 232 of the raw intercept (𝜆∗) and stochastic error term (ε𝑖), and is weighted by the time DHS agents were on the 233 street/out of their buildings each night (FT𝑖)37. The number of canisters deployed each day is then a Poisson 234 distribution with rate 𝜆𝑖 FT𝑖 truncated at the minimum by the known cans deployed on that day (𝑐𝑖): 235 236 ε𝑖 ~ Normal(0, 𝜎 2) 237 𝜆𝑖 = 𝑒 𝜆∗+ε𝑖 (Eq. 2). 238 𝑑𝑖 ~ Poisson(𝜆𝑖 FT𝑖 )c𝑖 239 240 Deployed grenades were then subjected to each detection process via Binomial distributions: observation 241 (regardless of recovery) is governed by rate 𝜈 to give daily observed cans 𝑜𝑖 and by recovery (regardless of 242 observation) by rate 𝜌 to generate daily recovered cans 𝑟𝑖. The processes are joined using a third, constrained 243 11 Binomial describing the number of grenades that were both observed and recovered (𝑜𝑟𝑖) by applying the 244 recovery process to observed grenades and capping the number at the total grenades recovered37,38. Both rates 245 are fit on the logit scale: 246 247 𝑜𝑖 ~ Binomial(𝜈, 𝑑𝑖 ) 248 𝑟𝑖 ~ Binomial(𝜌, 𝑑𝑖 ) 249 𝑜𝑟𝑖 ~ Binomial(𝜌, 𝑜𝑖 ) 𝑟𝑖 (Eq. 3). 250 𝜈 = logit−1(𝜈∗) 251 𝜌 = logit−1(𝜌∗) 252 253 This model therefore assumes no false positives, a fair baseline assumption, given the distinctive burn pattern 254 and resulting canister (Figs. 2, S1-24)37. I used generally uninformative priors on the raw scales: 255 256 𝜆∗ ~ Normal(0,1) 257 𝜈∗ ~ Normal(0,1) (Eq. 4). 258 𝜌∗ ~ Normal(0,1) 259 𝜎 ~ Uniform(0,100) 260 261 I fit the model using JAGS (Just Another Gibbs Sampler, v4.2.0)39 via the runjags v2.0.4-6 package60 262 in R55. I used four MCMC chains with varying starting values for parameters and ran each for 10,000 263 adaptation, 100,000 burn-in, and 1,000,000 final samples thinned to 10,000 per chain to total 40,000 samples 264 across chains. I evaluated chain convergence using the autocorrelation, sample size adjusted for 265 autocorrelation, and Gelman-Rubin statistic36 for each parameter. All code is included within Datasets S5,S6. 266 267 I then converted the estimated number of cans deployed each day to the potential number of human 268 fatalities from ZnCl2. A standard Military Style can contains 19 oz of HC mix Type C24, there are 28.4 g in an 269 12 oz, and assuming no loss of mass, 1 g of Type C mix generates 1 g products. ZnCl2 constitutes 0.764 w/w of 270 all products29, which translates to 412.3 g ZnCl2 per grenade. It is difficult to gauge specifically the lethal dose 271 or concentration of ZnCl2, given the limited data on humans and multiple modes of uptake (inhalation, orally, 272 dermally)24,27. Thus, for a simple approximation, I used a log-normal distribution based on the eight studies 273 included in PubChem that report LD50 values for mammal models (24, 58, 91, 200, 330, 350, 1100, 1260 274 mg/kg)42, which has a log-scale mean of 5.36 and standard deviation of 1.39, generating a back-transformed 275 mean of 555 mg/kg. For human sizes, I used the most recent (2017-2018) US National Health and Nutrition 276 Examination Survey with available data and combined the data across genders to construct a log-normal 277 distribution with a back-transformed mean of 83 kg (log-scale mean: 4.42, sd: 0.22). Thus, an average LD50 is 278 46 g/person and an average grenade contains enough ZnCl2 to kill 9.0 people. To incorporate the LD50 and 279 weight uncertainty in estimating the number of potential fatalities, I treated the total ZnCl2 from each MCMC 280 iteration as a resource pool and individuals consumed the ZnCl2 represented by their personal LD50 values, 281 which were drawn separately for each iteration. 282 283 All code used is included within Datasets S5 and S6, including full R session information and package 284 versions. 285 286 Chemical Analyses 287 288 I collected 11 environmental samples from a variety of sources around the areas of hexachloroethane 289 deployment (Fig. 4, Dataset S2): 290 291 [1] Medic Filter: filter medium from a NIOSH Organic Vapors DMA 6001 filter set worn by a medic 292 only on 2020-07-27, 2020-07-28, 2020-07-29 in the area of SW 4th and Main. Medic only brought 293 out mask when chemical weapons were used and always positioned themselves outside of the visible 294 plume to treat individuals as they came out. 295 13 [2] HC Can: dust/particle residue from inside Defense Technology Hexachloroethane (HC) Smoke can 296 deployed and recovered post “completion” on 2020-07-28 night into 2020-07-29 (Figs. 3, S16; Video 297 S1). 298 [3] A’s Backpack: Cut out from a black Jansport backpack that was worn by a protester the night of 299 2020-07-23 and prepped for sampling thereafter. 300 [4] 3rd Avenue and Salmon Street Plants: shrub within the fence at the Federal Courthouse and Tree 301 at the corner of Lownsdale, samples taken 2020-07-27 night after a bleach smell was noticed and 302 2020-07-28 during the following daytime. 303 [5] Lownsdale Surface Soil at 3rd Avenue and Salmon Street: Scoop of topsoil from the NE corner 304 of the park taken 2020-07-28 midday. 305 [6] SW 3rd Avenue: samples of paper and other refuse on the street in front of the Federal courthouse 306 on 3rd near Salmon from immediately after a bleach smell was noticed 2020-07-27 into 2020-07-28. 307 [7] E’s Shirt: water taken from a soak of a shirt worn by a protester on 2020-07-26 into 2020-07-27, with 308 noticeable bleach-like smell and visible loss of coloration. 309 [8] Green Smoke Can: dust/particle residue from inside Defense Technology Green Smoke canister 310 deployed and recovered post “completion” on 2020-07-28 into 2020-07-29. 311 [9] S’s Leggings: water taken from a soak of leggings worn by protester recovering spent canisters 2020-312 07-28 into 2020-07-29. 313 [10] Witches’ Tent: passive sample taken from existing cotton rounds, paper towels, etc that were 314 present in a medical tent in Lownsdale (run by medics known as The Witches) the night when the 315 tent reeked of bleach 2020-07-2661. 316 [11] Spicy Bucket Scrape: residue scraped from inside of a Home Depot 5-gallon bucket used to cover 317 smoke and gas canisters during 2020-07-27 and 2020-07-28 nights. 318 319 14 Samples were stored frozen in 1-L glass jars until submitted to Specialty Analytical in Clackamas, Oregon. 320 Each sample was tested using standard EPA methods for volatile organic compounds (SW8260D, E8260D); 321 semi-volatile organic compounds (E8270E); and Zinc, Chromium, and Lead (SW 6020B) (Dataset S2). 322 323 DATA AVAILABILITY 324 325 All data used herein are included in Datasets S1 to S6. 326 327 ACKNOWLEDGEMENTS 328 329 This work would not have been possible without a large community standing up despite the wanton 330 use of chemical weapons to say that Black Lives Matter. Front-line journalists including Alissa Azar, Garrison 331 Davis, Robert Evans, Mariah Harris, Laura Jedeed, Jacob Hanning, Melissa Lewis, Sergio Olmos, Mac Smith, 332 Tuck Woodstock and many individuals who have chosen to remain anonymous due to continuing tensions 333 with law enforcement provided invaluable documentation. Additional contributions to the dossier are cited in 334 the Supplementary Information. Substantial thank you to the Don’t Shoot Portland team for documenting, 335 researching, and organizing around use of chemical weapons in Portland; Sarah Riddle for life-cycle 336 documentation; The Recompiler Magazine for aggregated protest news; and Eric Greatwood for particularly 337 useful standardized video footage. Mason Fidino gave helpful feedback on the model and Nelson Hairston Jr. 338 and Sandy Simonis provided editing support. 339 340 REFERENCES 341 342 1. F. Lieber & A. Lincoln. The Lieber Code. General Order No. 100. Article 70. Available at 343 https://avalon.law.yale.edu/19th_century/lieber.asp (1863). Accessed 1 October 2020. 344 2. A. Morman, Z. Williams, D. Smith., & A. C. Randolph. Riot control agents: systemic 345 https://avalon.law.yale.edu/19th_century/lieber.asp 15 3. reassessment of adverse effects on health, mental stability, and social inequities. Don’t Shoot PDX, 346 Portland, OR. Available at https://www.dontshootpdx.org/wp-347 content/uploads/2020/06/DSPFinal-RCAreport4SocialChange-AM.AR_.ZW_.DS-.pdf (2020). 348 Accessed 1 October 2020. 349 4. @ProudBulba. Tweet. https://twitter.com/ProudBulba/status/1265896197866176512. Accessed 1 350 October 2020. 351 5. E. Campuzano, B. Hebert, B. Nakamura, D. Killen, & M. Graves. 100 Days of Protests in Portland. 352 The Oregonian, 8 September 2020. https://www.oregonlive.com/portland/2020/09/100-days-of-353 protests-in-portland.html. Accessed 1 October 2020. 354 6. A. Zielinski. Good Morning, News: Smokey Skies, Police Overtime Costs, and 100 Days of Portland 355 Protests. Portland Mercury, 8 September 2020. 356 https://www.portlandmercury.com/blogtown/2020/09/08/28807957/good-morning-news-357 smokey-skies-police-overtime-costs-and-100-days-of-portland-protests. Accessed 1 October 2020. 358 7. PB2020 Team. Police Brutality During the 2020 George Floyd Protests. Available at 359 https://github.com/2020PB/police-brutality. Accessed 8 January 2021. 360 8. E. Weizman, R. Trafford, I. J. Brown, E. Farooqui, L. Karmode, Z. Ioannidis, L. Brown, L. Starr, N. 361 Elçi, S. Nankivell, C. Varvia, O. Ferwati, C. Godart, N. Waters, A. Mossou, G. Fiorella, E. Alder, R. 362 Barron, N. Boyle, R. Cavanaugh, A. Dhillon, M. Ficarelli, K. Mager, M. Nieto, I. Piper, S. Reeves-363 Dyer, & A. Weizman. Police Brutality at the Black Lives Matter Protests. Forensic Architecture. 364 https://forensic-architecture.org/investigation/police-brutality-at-the-black-lives-matter-protests. 365 Accessed 8 January 2021. 366 9. United States Census Bureau. Population Estimates V2019 (July 1, 2019). 367 https://www.census.gov/quickfacts/fact/table/US/PST045219. Accessed 8 January 2021 368 10. Hague Peace Conference. Declaration on the Use of Projectiles the Object of Which is the Diffusion 369 of Asphyxiating or Deleterious Gases. Available at https://avalon.law.yale.edu/19th_century/dec99-370 02.asp (1899). 371 https://www.dontshootpdx.org/wp-content/uploads/2020/06/DSPFinal-RCAreport4SocialChange-AM.AR_.ZW_.DS-.pdf https://www.dontshootpdx.org/wp-content/uploads/2020/06/DSPFinal-RCAreport4SocialChange-AM.AR_.ZW_.DS-.pdf https://twitter.com/ProudBulba/status/1265896197866176512 https://www.oregonlive.com/portland/2020/09/100-days-of-protests-in-portland.html https://www.oregonlive.com/portland/2020/09/100-days-of-protests-in-portland.html https://www.portlandmercury.com/blogtown/2020/09/08/28807957/good-morning-news-smokey-skies-police-overtime-costs-and-100-days-of-portland-protests https://www.portlandmercury.com/blogtown/2020/09/08/28807957/good-morning-news-smokey-skies-police-overtime-costs-and-100-days-of-portland-protests https://github.com/2020PB/police-brutality https://forensic-architecture.org/investigation/police-brutality-at-the-black-lives-matter-protests https://www.census.gov/quickfacts/fact/table/US/PST045219 https://avalon.law.yale.edu/19th_century/dec99-02.asp https://avalon.law.yale.edu/19th_century/dec99-02.asp 16 11. Hague Peace Conference. Laws and Customs of War on Land (Hague IV). Article 23. Available at 372 https://avalon.law.yale.edu/20th_century/hague04.asp (1907). 373 12. Geneva Protocol. Protocol for the Prohibition of the Use in War of Asphyxiating Gas, and of 374 Bacteriological Methods of Warfare. Available at 375 https://avalon.law.yale.edu/20th_century/geneva01.asp (1928). 376 13. H. Hu, J. Fine, P. Epstein, K. Kelsey, P. Reynolds, & B. Walker. 1989. Tear gas - harassing agent or 377 toxic chemical weapon? The Journal of the American Medical Association 262:660-663 (1989). 378 14. R. A. Greenfield, L. N. Slater, M. S. Bronze, B. R. Brown, R. Jackson, J. J. Iandolo, & J. B. Hutchins. 379 Microbiological, biological, and chemical weapons of warfare and terrorism. The American Journal of the 380 Medical Sciences 323:326-340 (2002). 381 15. D. Kaszeta. Restrict use of riot-control chemicals. Nature 573:27-29 (2019). 382 16. R. Devereaux. Proud Boys Rally Fizzled But Portland’s Cops Went on the Attack. The Intercept, 28 383 September 2020. https://theintercept.com/2020/09/28/portland-proud-boys-rally-journalists/. 384 Accessed 1 October 2020. 385 17. D. J. Trump. Executive Order on Protecting American Monuments, Memorials, and Statues and 386 Combating Recent Criminal Violence. 26 June 2020. Available at 387 https://www.whitehouse.gov/presidential-actions/executive-order-protecting-american-388 monuments-memorials-statues-combating-recent-criminal-violence/. Accessed 1 October 2020. 389 18. Department of Homeland Security. DHS Announces New Task Force to Protect American 390 Monuments, Memorials, and Statues. 1 July 2020. Available at 391 https://www.dhs.gov/news/2020/07/01/dhs-announces-new-task-force-protect-american-392 monuments-memorials-and-statues. Accessed 1 October 2020. 393 19. K. Flanigan. Docs: Homeland Security’s Portland protest mission called “Operation Diligent Valor”. 394 KOIN. 23 July 2020. https://www.koin.com/news/protests/docs-homeland-securitys-portland-395 protests-mission-called-operation-diligent-valor/. Accessed 1 October 2020. 396 https://avalon.law.yale.edu/20th_century/hague04.asp https://avalon.law.yale.edu/20th_century/geneva01.asp https://theintercept.com/2020/09/28/portland-proud-boys-rally-journalists/ https://www.whitehouse.gov/presidential-actions/executive-order-protecting-american-monuments-memorials-statues-combating-recent-criminal-violence/ https://www.whitehouse.gov/presidential-actions/executive-order-protecting-american-monuments-memorials-statues-combating-recent-criminal-violence/ https://www.dhs.gov/news/2020/07/01/dhs-announces-new-task-force-protect-american-monuments-memorials-and-statues https://www.dhs.gov/news/2020/07/01/dhs-announces-new-task-force-protect-american-monuments-memorials-and-statues https://www.koin.com/news/protests/docs-homeland-securitys-portland-protests-mission-called-operation-diligent-valor/ https://www.koin.com/news/protests/docs-homeland-securitys-portland-protests-mission-called-operation-diligent-valor/ 17 20. C. Vaughn. “Highly Toxic” Smoke Grenades Deployed at Portland Protests. Portland Tribune, 20 397 October 2020. https://pamplinmedia.com/pt/484862-390719-sinister-smoke-pwoff. Accessed 8 398 January 2021. 399 21. S. Lerner. Federal Agents Used Toxic Chemical Smoke Grenades in Portland. The Intercept, 10 400 October 2020. https://theintercept.com/2020/10/10/portland-tear-gas-chemical-grenades-401 protests/. Accessed 8 January 2021 402 22. Defense Technology. Safety Data Sheet. 1083: Military-Style Maximum Smoke HC Grenade. 2015. 403 http://sds.chemtel.net/webclients/safariland/finished_goods/Defense%20Technology%201083%2404 0-%20Military-Style%20Maximum%20Smoke%20HC%20Grenade.pdf. Accessed 8 January 2021 405 23. H. Cullumbine. The toxicity of screening smokes. BMJ Military Health 103:119-22 (1957). 406 24. E. W. Novak, L. G. Lave, J. J. Stukel, & S. Miller. A revised health risk assessment of the use of 407 hexachloroethane smoke on an army training area. US Army Construction Engineering Research 408 Laboratory. Technical Report N-164 (1987). 409 25. J. C., Eaton, R. J. Lopinto, & W. G. Palmer. Health effects of hexachloroethane (HC) smoke. US 410 Army Biomedical Research and Development Laboratory. Technical Report 9402 (1994). 411 26. R. Blau & N. Seidner. Hexachloroethane obscurant replacement. ATK Propulsion Systems. Strategic 412 Environmental Research and Development Program. Report TR031214 (2012). 413 27. C. Smith-Simon, J. M. Donohue, & C. Eisenmann. Toxicological profile for hexachloroethane. US 414 Department of Health and Human Services, Public Health Service, Agency for Toxic Substances and 415 Disease Registry (1997). 416 28. F. S. Dainton & K. J. Ivin. The pyrolysis of hexachloroethane. Transactions of the Faraday Society 43:295-417 299 (1947). 418 29. W. L. Archer. Hexachloroethane. In M. Grayson & C. Eckroth, eds. Kirk-Othmer Encyclopedia of 419 Chemical Technology, 3rd ed., Vol. 5. pp. 738-740. John Wiley & Sons, New York, NY (1979). 420 30. A. P. Shaw, J. S. Brusnahan, J. C. Poret, & L. A. Morris. Thermodynamic modeling of pyrotechnic 421 smoke compositions. ACS Sustainable Chemistry and Engineering 4:2309−2315 (2016). 422 https://pamplinmedia.com/pt/484862-390719-sinister-smoke-pwoff https://theintercept.com/2020/10/10/portland-tear-gas-chemical-grenades-protests/ https://theintercept.com/2020/10/10/portland-tear-gas-chemical-grenades-protests/ http://sds.chemtel.net/webclients/safariland/finished_goods/Defense%20Technology%201083%20-%20Military-Style%20Maximum%20Smoke%20HC%20Grenade.pdf http://sds.chemtel.net/webclients/safariland/finished_goods/Defense%20Technology%201083%20-%20Military-Style%20Maximum%20Smoke%20HC%20Grenade.pdf 18 31. A. El Idrissi, L. van Berkel, N. E. Bonekamp, D. J. Z. Dalemans, & M. A. G. van der Heyden. The 423 toxicology of zinc chloride smoke producing bombs and screens. Clinical Toxicology 55:167-174 424 (2017). 425 32. M. C. Sadusky, J. M. Skelly, M. Simini, R. T. Checkai, & R. S. Wentsel. Hexachloroethane obscurant: 426 assessing tree foliage injury. Environmental Toxicology and Chemistry 12:685-694 (1993). 427 33. C. Hartl, S. St. George, O. Konter, L. Harr, D. Scholz, A. Kirchhefer, & J. Esper. Warfare 428 dendrochronology: trees witness the deployment of the German battleship Tirpitz in Norway. 429 Anthropocene 27:100212 (2019). 430 34. K. A. Davidson, P. S. Hovatter, & R. H. Ross. Water quality criteria for hexachloroethane. Oak 431 Ridge National Laboratory. Report ORNL-6469 (1988). 432 35. J. W. Nichols, J. M. McKim, G. J. Lien, A. D. Hoffman, & S. L. Bertelstein Physiologically based 433 toxicokinetic modeling of three waterborne chloroethanes in rainbow trout (Oncorhynchus mykiss). 434 Toxicology and Applied Pharmacology 110:374-389 (1991). 435 36. A. Salvaggio, F. Marino, M. Albano, B. M. Lombardo, S. Saccone, V. Mazzei, & M. V. Brundo. Toxic 436 effects of Zinc Chloride on the bone development in Danio rerio (Hamilton,1822). Frontiers in 437 Physiology 7:153 (2016). 438 37. A. Gelman & D. B. Rubin. Inference from iterative simulation using multiple sequences. Statistical 439 Science 7:457-511 (1992). 440 38. J. K. Kruschke. Doing Bayesian Data Analysis: A Tutorial with R, JAGS, and Stan. Academic Press, San 441 Diego, CA, USA (2015). 442 39. N. T. Hobbs & M. B. Hooten. Bayesian Models: A Statistical Primer for Ecologists. Princeton University 443 Press, Princeton, NJ, USA (2015). 444 40. M. Plummer. JAGS: Just Another Gibbs Sampler. v4.2.0. Available at 445 https://sourceforge.net/projects/mcmc-jags/files/JAGS/4.x/. (2016). 446 41. Johnnthelefty. Tweet. https://twitter.com/Johnnthelefty/status/1286590669482389504. Accessed 1 447 October 2020. 448 https://sourceforge.net/projects/mcmc-jags/files/JAGS/4.x/ https://twitter.com/Johnnthelefty/status/1286590669482389504 19 42. J. Levinson. Donavan La Bella, man shot by US Marshals in Portland, to be released from hospital. 449 Oregon Public Broadcasting, 25 July 2020. https://www.opb.org/news/article/donavan-labella-portland-450 condition-shot-head-federal-officer/. Accessed 1 October 2020. 451 43. National Center for Biotechnology Information. PubChem Compound Summary for CID 5727, 452 Zinc chloride. Available at https://pubchem.ncbi.nlm.nih.gov/compound/Zinc-chloride. Accessed 1 453 October 2020. 454 44. Centers for Disease Control. National Health and Nutrition Examination Survey 2017-2018 455 Questionnaire Data. Available at 456 https://wwwn.cdc.gov/nchs/nhanes/Search/DataPage.aspx?Component=Questionnaire&CycleBeg457 inYear=2017. Accessed 1 October 2020. 458 45. City of Portland, Oregon. Environmental Services Releases Results of CS Gas Residue Sampling in 459 City Stormwater Pipes; $20,000 Penalty to Feds. 10 September 2020. 460 https://www.portland.gov/bes/news/2020/9/10/environmental-services-releases-results-cs-gas-461 residue-sampling-city-stormwater. Accessed 1 October 2020. 462 46. T. Woodstock. Tweet. https://twitter.com/tuckwoodstock/status/1284013510700097537. Accessed 463 1 October 2020. 464 47. O. K. Smith. Tweet. https://twitter.com/livkittykat/status/1287274511629336576. Accessed 1 465 October 2020. 466 48. A. Azar. Tweet. https://twitter.com/R3volutionDaddy/status/1287688005130391552. Accessed 1 467 October 2020. 468 49. G. McKelvey. Tweet. https://twitter.com/GregoryMcKelvey/status/1287495472983052288. 469 Accessed 1 October 2020. 470 50. J. Noelle. Tweet. https://twitter.com/JesykaNoellex3/status/1308535486609321985. Accessed 1 471 October 2020. 472 51. C. Smith. Tweet. https://twitter.com/alsoconnor/status/1285054272841773057. Accessed 1 473 October 2020. 474 https://www.opb.org/news/article/donavan-labella-portland-condition-shot-head-federal-officer/ https://www.opb.org/news/article/donavan-labella-portland-condition-shot-head-federal-officer/ https://pubchem.ncbi.nlm.nih.gov/compound/Zinc-chloride https://wwwn.cdc.gov/nchs/nhanes/Search/DataPage.aspx?Component=Questionnaire&CycleBeginYear=2017 https://wwwn.cdc.gov/nchs/nhanes/Search/DataPage.aspx?Component=Questionnaire&CycleBeginYear=2017 https://www.portland.gov/bes/news/2020/9/10/environmental-services-releases-results-cs-gas-residue-sampling-city-stormwater https://www.portland.gov/bes/news/2020/9/10/environmental-services-releases-results-cs-gas-residue-sampling-city-stormwater https://twitter.com/tuckwoodstock/status/1284013510700097537 https://twitter.com/livkittykat/status/1287274511629336576 https://twitter.com/R3volutionDaddy/status/1287688005130391552 https://twitter.com/GregoryMcKelvey/status/1287495472983052288 https://twitter.com/JesykaNoellex3/status/1308535486609321985 https://twitter.com/alsoconnor/status/1285054272841773057 20 52. A. Gupta. Tweet. https://twitter.com/arunindy/status/1288012683543900160. Accessed 1 October 475 2020. 476 53. R. Evans. Tweet. https://twitter.com/IwriteOK/status/1286585525365768193. Accessed 8 January 477 2021. 478 54. G. Davis. Tweet. https://twitter.com/hungrybowtie/status/1321719869662916610. Accessed 8 479 January 2021. 480 55. J. Simonis. Tweet. https://twitter.com/JuniperLSimonis/status/1321896995607334912. Accessed 8 481 January 2021. 482 56. R Core Team. R: A language and environment for statistical computing. R Foundation for Statistical 483 Computing, Vienna, Austria. Available at https://www.R-project.org/ (2020). 484 57. H. Wickham. httr: Tools for Working with URLs and HTTP. R package version 1.4.1. 485 https://CRAN.R-project.org/package=httr (2019). 486 58. J. Ooms. The jsonlite Package: A Practical and Consistent Mapping Between JSON Data and R 487 Objects. arXiv:1403.2805 [stat.CO] https://arxiv.org/abs/1403.2805 (2014). 488 59. E. Greatwood. Veteran for Peace. https://www.facebook.com/EricGreatwood/videos. Accessed 1 489 October 2020. 490 60. The Recompiler. Re: Portland. https://pdx.recompilermag.com/about. Accessed 1 October 2020. 491 61. M. J. Denwood. runjags: An R Package Providing Interface Utilities, Model Templates, Parallel 492 Computing Methods and Additional Distributions for MCMC Models in JAGS. Journal of Statistical 493 Software 71:1-25 (2016). 494 62. J. Hanning. Tweet. https://twitter.com/Cascadianphotog/status/1287714834893565952. Accessed 1 495 October 2020. 496 63. Google Search Term Trends. https://trends.google.com/trends/explore?geo=US-OR-497 820&q=protests,black%20lives%20matter,federal. Accessed 1 October 2020. 498 64. W. S. Cleveland, E. Grosse & W. M. Shyu. “Local regression models” Chapter 8 in Statistical Models in 499 S, J. M. Chambers & T. J. Hastie, Eds. (Wadsworth & Brooks/Cole, 1992). 500 https://twitter.com/arunindy/status/1288012683543900160 https://twitter.com/IwriteOK/status/1286585525365768193 https://twitter.com/hungrybowtie/status/1321719869662916610 https://twitter.com/hungrybowtie/status/1321719869662916610 https://twitter.com/JuniperLSimonis/status/1321896995607334912 https://www.r-project.org/ https://cran.r-project.org/package=httr https://arxiv.org/abs/1403.2805 https://www.facebook.com/EricGreatwood/videos https://pdx.recompilermag.com/about https://twitter.com/Cascadianphotog/status/1287714834893565952 https://trends.google.com/trends/explore?geo=US-OR-820&q=protests,black%20lives%20matter,federal https://trends.google.com/trends/explore?geo=US-OR-820&q=protests,black%20lives%20matter,federal 21 TABLES 501 502 Table 1. Statistical fit results and diagnostics from the Bayesian estimator of hexachloroethane grenade use. 503 Lower 95 Median Upper 95 Mean SD Mode MC err MC % of SD SS eff AC 1000 psrf D 25 26 30 25.953 1.34 25 λr -3.425 -2.113 -0.859 -2.149 0.66 0.003 0.5 38820 -0.003 1.00 νr -0.752 0.006 0.744 0.008 0.38 0.002 0.5 39526 0.001 1.00 ρr 0.247 0.978 1.701 0.985 0.37 0.002 0.5 40000 0.002 1.00 σ 0.000 1.335 8.021 2.427 3.81 0.034 0.9 12730 0.013 1.00 r indicates untransformed (raw) scale 504 MCMC diagnostics are not included for D, a state variable, just the parameters. MC err: MCMC standard 505 error; MC % of SD: MCMC standard error as a percentage of the Standard Deviation of the posterior 506 distribution; SS eff: Effective Sample Size; AC 1000: Autocorrelation at 1000 MCMC steps (AC 10 for the 507 thinning interval of 100 used); and psrf: potential scale reduction factor (Gelaman-Rubin statistic; 27). 508 22 FIGURES 509 510 511 512 Figure 1. Total and chemical-weapons-based incidents of police brutality during the George Floyd Black 513 Lives Matter 2020 protests through the end of the year for all US cities with at least 10 incidents6. City 514 population sizes9 are depicted via the points and the secondary y-axis (see also Dataset S4). 515 516 23 517 Figure 2. Top: time series of general interest (normalized Google Search Term Trends63) for the Portland 518 Metro Area in “protests” (orange), “Black Lives Matter” (red), and “federal” (purple) during the 2020 George 519 Floyd Black Lives Matter protest period to 1 September. Bottom: July-focused time series of crowd size 520 (purple) and the number of minutes federal agents were out (red) each night. Lines in both portions were fit 521 using local polynomial regression (loess)64. Grey box points connected by grey line show the number of 522 hexachloroethane (HC) grenades used each night, based on observations and collections combined. 523 524 525 526 527 528 24 529 530 Figure 3. Hexachloroethane (HC) / Zinc Chloride smoke grenade canisters: (a) Unexploded ordnance clearly 531 marked as “Military Style Maximum Smoke HC” from “Defense Technology” (Fig. S26); (b) HC ordnance 532 off gassing Zinc Chloride mid-deployment; (c) Same canister from (b) after reaction stopped, showing 533 charred remains of the label that matches the canister in (a); and (d) three exploded HC canisters, including 534 the one from (b) and (c) in the middle. Photos (a) and (d) are from the author, (b) and (c) are from Sarah 535 Riddle and used with permission. The canister in (b), (c), and (d) was sampled in Dataset S2 (Fig. 4, S16), and 536 its deployment is shown in Video S1. 537 25 538 539 Figure 4. Sample locations and pictures for the 11 environmental chemistry samples taken around the 540 downtown area of Portland, OR. [1] Medic Filter, [2] HC Can (Video S1), [3] A’s backpack, [4] 3rd Ave and 541 Salmon St Plants, [5] Lownsdale Surface Soil, [6] SW 3rd Avenue, [7] E’s Shirt, [8] Green Smoke Can, [9] S’s 542 Leggings, [10] Witches’ Tent, [11] Spicy Bucket Scrape. Blue lines on the street map show the streets drained 543 by the stormwater system that empties unfiltered into the Willamette river. 544 545 26 546 547 Figure 5. Posterior distributions for the number of HC grenades deployed (left) and the rate of grenade 548 deployment (grenades per hour) (right). 549