Mitigating greenhouse gas (GHG) emissions from the transportation sector is crucial for achieving climate goals in the United States and globally. In response, policymakers worldwide have committed to accelerating the transition from internal combustion engine vehicles to electric vehicles (EVs) (1-3). Diverse regional policies supporting EV transition have been adopted in the United States (1). California, which is the largest market for new cars in the United States, has set a goal of putting 5 million EVs on the road and achieving 100% zero-emission vehicle sales for new passenger cars and light trucks by 2035 (4). This unprecedented large-scale transportation electrification will necessitate the establishment of a vast network of EV chargers globally. Public EV chargers mitigate the constraints of home charging (5), thereby alleviating drivers' concerns about limited driving range (commonly known as "range anxiety"), and provide crucial support for the widespread adoption of EVs in multi-dwelling housing that may lack home charging capability. With the massive investments through the Inflation Reduction Act and the Infrastructure Investment and Jobs Act, the U.S. government has ambitious plans to allocate $7.5 billion toward the advancement of EVs and charging infrastructure (6). This commitment includes the establishment of an widespread network of 500,000 EV charging stations by 2030 (6). Among different types of public EV chargers, Level 3 Direct Current Fast Charging (DCFC) is especially critical for mitigating range anxiety compared to low-power Level 1 and medium-power Level 2 chargers, particularly for unplanned trips or emergencies. DCFCs greatly reduce charging time, thereby eliminating one of the major barriers toward wider adoption of EVs (7, 8).
The dramatic expansion of DCFCs accelerates the transition to clean transportation, which is expected to mitigate climate change and improve air quality. However, their environmental and public health impacts have been poorly understood. Previous studies have primarily focused on optimization of power sources and clean energy grid integration to reduce the carbon footprint of EV lifecycles (9, 10). Studies have also evaluated their impacts on reducing GHG emissions and improving urban air quality (11), and on residential property values (12). Conversely, environmental and public health concerns have been raised about the deployment of DCFCs. For example, studies have indicated that EV chargers produce electromagnetic radiation during the charging process (13), but reported minimal health risks overall (14). Additionally, the operation of DCFC stations may release heavy metals, such as nickel, lead, and lithium (15). Recently, some claims have been made on social media about DCFC stations using diesel engines for power (16), raising questions about the energy cleanness and potential air pollutant emissions, but these claims lack empirical data support. To determine whether DCFCs emit air pollutants and to understand potential emission mechanisms, we collected 24-hr filter samples at DCFC stations across 47 different cities in the Los Angeles County and conducted real-time sampling at one typical DCFC station.
Distribution of EV chargers
Fig. 1a illustrates the distribution of operational EV fast chargers across the United States as of November 2023, totaling 36,000 units (17). While some stations were located along interstate highways, the majority were clustered in major metropolitan areas, especially along the West and East Coasts, corresponding with major population centers. As shown in Fig. 1b, California led with 9,900 EV fast chargers, followed by Texas with 2,200 and Florida with 2,100, respectively. The DCFC network is expected to undergo a rapid expansion from 2023 to 2028, as facilitated by the Infrastructure Investment and Jobs Act. This is illustrated by the darker orange bars for each state in Fig. 1b.
Among all counties in the United States, Los Angeles County had the highest number of EV fast chargers in operation, totaling 1,938 units. On average, Los Angeles County had about six individual chargers per DCFC station. Out of a total of 313 operational DCFC stations, we identified 113 that meet our sampling criteria, which included specific requirements for charging output power and the presence of power cabinets. We then randomly selected 50 DCFC stations across 47 cities in Los Angeles County to perform field measurements (Fig. 1c and Methods). For comparison purposes, we also sampled eight gas stations and four urban background sites across Los Angeles County during the same sampling period.
PM2.5 concentrations at DCFC stations
We collected integrated 24-hr filter samples for PM2.5 (particulate matter with aerodynamic diameters equal to or less than 2.5 μm) at the power cabinet at each of the 50 randomly selected stations. Fig. 2 shows a typical DCFC station in Los Angeles County, alongside PM2.5 sample collection at one of the power cabinets. This DCFC station was located at a business plaza, surrounded by residential neighborhoods (Fig. 2a). It was equipped with four power cabinets and 16 fast charging stalls, connected by electric wires (Fig. 2b). To collect 24-hour PM2.5 samples, we set up an Ultrasonic Personal Aerosol Sampler (UPAS) to the top of one of the power cabinets at each station, approximately 2 meters above the ground (Fig. 2c). A similar experimental setup was used for each of the 50 sampling sites. Detailed information about the number and output power of EV fast chargers at each station can be found in the supplementary materials, Table S1.
Fig. 2d presents the 24-hr mean PM2.5 concentrations at each of the 50 sampled DCFC stations, ranging from 7.3 µg m-3 to 39.0 µg m-3, with 46% of the sites exceeding the 24-hr World Health Organization (WHO) Global Air Quality Guidelines (AQGs), set at 15 µg m-3. As shown in Fig. 2e, the mean ± SD and median (IQR) PM2.5 concentrations were 15.2 ± 6.3 µg/m³ and 14.1 (6.7) µg/m³ for DCFC sites, 11.9 ± 4.7 µg/m³ and 11.8 (7.3) µg/m³ for gas stations, and 8.6 ± 3.8 µg/m³ and 9.3 (7.1) µg/m³ for urban background sites. The Mann-Whitney U test shows substantially higher PM2.5 concentrations at DCFC stations compared to those at the urban background sites (p = 0.02). In comparison, PM2.5 concentrations at U.S. Environmental Protection Agency (EPA) outdoor air monitoring stations across Los Angeles County during the same sampling period were (7.5 ± 4.3) µg m-3 and 6.7 (5.6) µg m-3, respectively, slightly lower than our measurements at the urban background sites. For comparison, we also surveyed PM2.5 levels at 10 Level 1 and Level 2 EV charging stations and found no elevated concentrations compared to urban background locations.
Potential mechanisms for particle emissions at DCFC stations
Given the unprecedented nature of these findings and the absence of previous literature, we investigated four potential particle emission and formation mechanisms associated with the specific characteristics of DCFC stations. These stations use converters inside power cabinets to switch Alternating Current (AC) to Direct Current (DC) and deliver power to EV batteries. This process results in the production of excessive heat (18), requiring thermal management of either air cooling or liquid cooling technologies to keep the temperature of power cabinet components within an operational range.
The first potential mechanism pertains to combustion, motivated by the lack of empirical data on whether some DCFC stations use diesel engines for power. The second possible mechanism involves nucleation due to corona discharge, given the potential presence of high voltage and electromagnetic field at DCFC stations. The third mechanism is condensation, since coolants used in power cabinets may vaporize due to high temperatures and undergo a condensation process to generate aerosols. The final possible mechanism is resuspension, which occurs when particles previously deposited on surfaces inside power cabinets are reintroduced into the atmosphere through mechanical disturbances.
To test each of these potential particle emission mechanisms, we conducted extensive real-time measurements of PM2.5 mass concentrations, particle size distribution, and other pollutants at the station shown in Fig. 2b over a period of 13 days for a total of approximately 15 hours. Throughout this extensive sampling period, we also monitored ambient temperature and EV charging activities at each fast charger. Fig. 3 presents the time series plots of measured air pollutant concentrations over a 5-hr period, along with temperature and charging load, which is defined as the percentage of EVs charging relative to the total number of available fast chargers at the station. As shown in Fig. 3a, we observed frequent peaks of PM2.5 reaching approximately 40 µg m-3. EV charging activity data are also presented in Fig. 3a. Surprisingly, there appears to be no direct correlation between the number of EVs being charged at the station and PM2.5 concentrations. A lag effect, however, seems to exist; for example, PM2.5 levels increased around 9:00 am following a period of intense EV charging activity that started at 8:30 am. As discussed later in this paper, we hypothesize that EV charging activity is only one of several factors influencing the observed particle emissions.
To test the first mechanism (i.e., combustion), we measured carbon monoxide (CO) and carbon dioxide (CO2) concentrations (Fig. 3b), indicators of combustion, which were found to be stable and comparable to urban background levels (19). We further conducted a breakpoint detection analysis to identify moments when changes in air pollutant concentrations exceeded expected variability. We observed 52 breakpoints in PM2.5 concentrations (10.4 ± 5.0 per hour), whereas no breakpoints were detected for CO and CO2. This suggests the absence of combustion emissions indicating that DCFC stations are unlikely to rely on diesel-powered generators.
To test the second hypothesized mechanism (i.e., nucleation due to corona discharge), we measured ozone (O3) concentrations along with ambient temperature (Fig. 3d). This is because corona discharges can create reactive species such as O3, which can then participate in chemical reactions in the air, leading to particle formation (20). As shown in Fig. 3d, O3 concentrations gradually increase overtime from below 10 ppb in the early morning to over 30 ppb by noon, coinciding with the rise in ambient temperature. This is typical for a good air quality day in Los Angeles (21). Like CO and CO2, there are no breakpoints for O3, which suggests that the corona discharge mechanism is unlikely.
To test the other two potential mechanisms (i.e., condensation and resuspension), we measured the particle size distribution in the range of 0.54 to 20 μm. Fig. 3c shows time-resolved particle size distributions as a contour plot, where the x-axis represents the same period when other pollutant data were collected. The y-axis is the particle size, and the color intensity indicates the normalized particle number concentration (dN/dLogDp) for a given size at a given time. Particle sizes up to 1 μm are presented in Fig. 3c, as larger particles were rarely detected. As shown in Fig. 3c, the particle concentration and size distribution vary greatly over time, with the dominant sizes in the submicron size range (from 0.5 to 1 µm). Additionally, there is an interesting pattern of decreasing particle mode diameter and generally increasing particle concentrations over time.
To test if these particles originate from the condensation of glycols (i.e., mechanism no. 3), which are most commonly used in DCFC cooling systems due to their excellent heat transfer properties (22), we measured particle volatility using our previously developed c-Air monitor (23, 24). This device employs computational microscopy and holography and has been used in our previous studies on electronic cigarette (e-cig) aerosols (23, 24). Fig. 4 illustrates the particle size distributions and their volatility for samples collected from three different sites: DCFC station (Fig. 4a), gas station (Fig. 4b), and urban background site (Fig. 4c). Similar to the data presented in Fig 3c, particles detected at the DCFC station are predominantly in the submicron size range. Fig. 4d-f show the volatility of the particles determined by analyzing the size changes over time, as demonstrated in the images of the typical captured particles. For instance, Fig. 4e shows the volatilization process of a detected volatile particle at a gas station, where the phase of the particle vanished at t = 2 s, indicating the evaporation of the sample. Overall, among all collected particles from the DCFC station, gas station, and urban background site, those from the DCFC station have the lowest percentage of volatile particles (i.e., 10%). This contrasts sharply with e-cig particles. It is well established that liquid-based e-cig particles that are formed through the condensation of e-liquids primarily consisting of propylene glycol (PG) and vegetable glycerin (VG) show extremely high volatility (24). Thus, condensation is unlikely to be the mechanism for these observed submicron particles.
This leads us to the last mechanism, resuspension, which, at the time of writing, is the most likely mechanism responsible for the observed particle emissions, for the following reasons. First, a large variety and amount of PM from both natural and anthropogenic sources, such as pollen and vehicular exhaust, along with non-exhaust emissions (e.g., brake, tires, and road wear) (25) are continuously released into the atmosphere. These PM can deposit on various surfaces, providing an “infinite” source of particles to be resuspended (26). Thus, particle accumulation on the internal surfaces of power cabinets is highly likely. In fact, some power cabinets use filters to prevent ambient PM from entering the system (27, 28). Second, power cabinets employ either air cooling or liquid cooling technologies to reduce system heat, both utilizing mechanical fans to dissipate the heat into the ambient air (29). This could create forced convection and generate turbulence, detaching the settled particles and causing them to resuspend in the atmosphere (30). Finally, an intriguing observation from Fig. 3c is the mode diameter of detected particles, which decreases from about 0.75 µm to approximately 0.6 µm over the 5-hr sampling period. This pattern aligns with the particle resuspension theory, which indicates that smaller particles are more difficult to detach from surfaces due to the size-dependent nature of the forces involved (31). Specifically, the particle resuspension process is governed by adhesive forces, which are proportional to the particle diameter (d), and removal forces, which, in the case of DCFC, are primarily driven by air current and are proportional to d2. This observed shift in particle mode diameter suggests that larger particles are more easily resuspended first as the DCFC station begins operation in the morning, likely because a substantial amount of dust has accumulated overnight. As the charging activity intensifies and becomes more continuous, smaller particles are also resuspended, but at a later stage.
It should be noted that dust resuspension is a complex process that depends on a range of variables, including surface type, dust loading, particle size, air velocity, relative humidity, and electrostatic fields (32, 33). In the case of DCFC stations, this process is likely influenced by intertwined factors such as the amount and size of particles accumulated on the internal surface of power cabinets, charging activities, fan speed, ambient temperature, and relative humidity among others. While our data support the notion that particle resuspension is a potential mechanism, given the unprecedented nature of these observations, future studies are warranted to study these influencing factors and characterize the chemical compositions of these particles to confirm this.
Societal implications
While no study we are aware of has specifically documented particle resuspension due to DCFC stations, extensive literature exists on road dust resuspension (26), particularly in recent studies focusing on non-exhaust emissions from mobile sources (25). Exposure to these resuspended particles has been linked to various health issues, including respiratory and cardiovascular diseases (34). With the shift toward EVs, which do not produce exhaust emissions, the role of road dust in urban air pollution is expected to become more important. This highlights the need for addressing particle resuspension in air quality management to protect public health, as society moves rapidly toward vehicle electrification (35, 36). Our research suggests that DCFC stations could be a notable source of particle resuspension, thus expanding the scope of sources considered in urban PM pollution studies.
With the U.S. government’s target to establish 500,000 EV chargers by 2030, and with the rapid ongoing EV adoption worldwide, the anticipated increase in DCFC stations raises concerns about potential environmental and health impacts of this unexpected particle emission on a population scale. Implementing regulations and engineering controls in the design and construction of future DCFC stations is crucial to effectively mitigate particle emission from power cabinets. For example, while some DCFC stations have filters in the air inlet to prevent PM from entering the unit, manufacturers could also add filters to prevent resuspended particles from re-entering the atmosphere. Moreover, strategic site selection, emphasizing the importance of placing DCFC stations at a distance from schools, nursing homes, and densely populated residential areas to minimize potential risks could also reduce exposures among vulnerable and underserved populations. In addition, future studies should investigate the physicochemical properties and toxicity of these particles to confirm emission mechanisms and better understand their potential health risks. More broadly, these findings emphasize the need for continuous monitoring of major energy transitions that may result in unintended risks to public health before major investments are made.