Paper prepared for the Second Symposium on the History of Aerosol Science, Portland, OR, October, 2001 The Emergence of the Bimodal Distribution Concept Rudolf B. Husar CAPITA, Box 1124, Washington University, St. Louis, MO, 63130. rhusar@me.wustl.edu Introduction In a nutshell, the bimodal distribution concept states that the atmospheric aerosol mass is distributed in two distinct size ranges, fine and coarse; each aerosol mode having characteristic size distribution, chemical composition and optical properties. Furthermore, implicit in the bimodal model is that the scientific and regulatory treatment of these modes can be dealt with separately since each mode is governed by its own sources, transformation, transport, and removal processes and is associated with different effects. With the introduction in 1997 of the PM2.5 national air quality standard, the bimodal distribution concept has gained full acceptance in air quality management as well. According to (Friedlander, 2000), the concept of multimodal mass (or volume) distributions has dominated our understanding of atmospheric aerosol dynamics for the last thirty years. Being a story of successful science-policy interaction, the history of the bimodal distribution seems suitable for chronicling, particularly since indeed it has a long history: It took about thirty years to move from the scientific conception to the air quality management implementation! The emergence and resilience of the bimodal distribution concept was evidently facilitated by the confluence of three parallel developments: 1) development of complete and continuous size spectrum measurement technologies; 2) the introduction of aerosol dynamics as the scientific principle for the explanation of atmospheric aerosol pattern; 3) unprecedented synergistic collaboration by key researchers in the field. This summary is structured to describe the above three developments. Following a brief review of the ‘Junge’ distribution, the paper roughly mirrors the content of the three papers by the University of Minnesota group on the ‘1969 Pasadena Smog Caper’ describing the Minnesota Aerosol Analyzing System (MAAS), (Whitby et al. 1972a); the size distribution of the Los Angeles smog (Whitby et al., 1972b) and the physical mechanisms governing the dynamics of the Los Angeles aerosol (Husar et al., 1972, Husar and Whitby, 1973) and related papers 1. Some aspects of the bimodal distribution history were previously reported during in the first History of Aerosol Science volume (Preining and Davis, 2000) by several authors (Friedlander, 2000; Hidy, 2000; Sem and Whitby, 2000). 1 Several of the figures are reproduced here in original form, as I have plotted during the 1969 Pasadena experiments. In fact, I have painted myself into this chronicle with shameless regularity. In order to reduce the damage, from this pretentious, self-centered reporting, I have placed most of the self-referencing into the footnotes. 1 Scientific Background: The Junge Distribution The modern science of atmospheric aerosols began with the pioneering work of Christian Junge who performed the first comprehensive measurements of the size distribution and chemical composition of atmospheric aerosols. The nature of the particulate matter arising from these measurements is reviewed and summarized in a series of papers, Junge (1953; 1955; 1957; 1958; 1961) as well as in his book on “Air Chemistry and Radioactivity” (Junge, 1963). Based on tedious and careful size distribution measurements performed over many different parts of the world, Junge and co-workers have observed that there is a remarkable similarity in the gathered size distributions: they follow a power law function over a wide range of particle size from 0.1 to over 20 m in particle radius. The inverse power law exponent of the number distribution function ranged between 3 and 5 with a typical value of 4. This power-law form of the size distribution became known as the Junge distribution atmospheric aerosols. When plotted as a volume distribution the Junge distribution is a horizontal line between about 0.1 and 20 m (Figure 1). The data indicated that below and above this size range the volume distribution function dropped off. In the 1960s the physical mechanisms that were responsible for producing these similarities in the atmospheric aerosol size spectra were not known, although it was clear that homogeneous and heterogeneous nucleation, coagulation, sedimentation and other removal processes were all influential mechanisms. In particular, it was unclear which combination of these mechanisms is responsible for maintaining the observed similarity of the size spectra. Sheldon Friedlander tackled the science i.e. the explanation of atmospheric aerosols along two fronts. In a series of articles based on dimensional analysis and similarity theory, Friedlander (1960a, 1960b, 1961) has shown that IF coagulation and sedimentation were the dominant physical mechanisms, then a significant part the aerosol spectrum could follow the power law shape. In essence, the implication of this theory was that the observed quasi-stationary size distribution could be the result of balancing aerosol production (nucleation) and removal (sedimentation) with coagulation transferring particles from the nuclei to the sedimentation size ranges. Friedlander (1965) and Friedlander and Wang (19) also pursued another path by introducing the concept of self-preserving size distributions for dynamic aerosol systems. According to the selfpreserving distribution concept, in an aging aerosol system, the size distribution asymptotically approaches a self-similar shape. In other words, when aging size spectra are properly normalized by total number and volume, the consecutive size distributions follow a universal, self-preserving shape that is characteristic for that aerosol system. In a numerical solution of the coagulation equation, Hidy (1965) has indeed found a confirmation of the self-preserving size distribution shape for coagulating aerosols. The lack of complete, dynamic size distribution data covering the entire 0.01-10 m size range of atmospheric aerosols in the 1960s have precluded a detailed verification of these theories. That changed when Clark and Whitby (1967) reported an extensive set of complete size distribution measurements in Minneapolis using the newly developed Minnesota Aerosol Analyzing System (MAAS) (see next section). The new aerosol size distribution data were welcomed by the atmospheric aerosol research community and sparked a vigorous debate 2 regarding the applicability and the causality of the Junge distribution. Clark and Whitby (1967) themselves spoke in favor Friedlander’s ideas but Junge (1969) swiftly responded that the ‘selfpreserving’ distritibution of a coagulating aerosol can be ruled out as an explanation of the atmospheric power-law data with negative slope4: since the slope of self-preserving coagulating aerosol size distribution was much steeper. Also, Junge argued, that since thermal coagulation is a weak mechanism for changing the size range above 0.1 m, there must be other stronger mechanisms that govern the atmospheric aerosol dynamics. He named homogeneous and heterogeneous gas-particle conversion, and cloud scavenging as candidate processes. However, he has declined to speculate on the relative roles and magnitudes of these other processes. Figure 1. Volume distribution functions of several atmospheric aerosol types (Junge, 1969) In a further significant statement, Junge also re-emphasized his earlier findings, that the slope of the power law distributions can be rather different, depending on the aerosol type as reproduced here in Figure 1 from (Junge, 1969). For example, the volume spectra of upper tropospheric aerosols has a volume peak at around 0.1 m radius, while marine aerosols show a broad peak at 1-10 m. Only the aerosols observed in the continental boundary layer adhere to the flat Junge distribution with 4 over the 0.1-100 m radius range. Finally, Junge offered an alternative explanation for the flat Junge spectrum: Continental aerosol arise from many different aerosol sources, each having a narrow size range but of different mean size in the range 0.1-100 m. When these heterogeneous aerosols are mixed in the atmosphere, they form a broad size spectrum that can be represented by a broad log-normal distribution which resembles the typical flat Junge distribution2. This was roughly the state of understanding in 1969 as the research groups were preparing the Los Angeles smog aerosol study in Pasadena, California. 2 It is remarkable how close Junge came to a full explanation of the of the atmospheric aerosol dynamics in 1969, yet how far he was from actually completing the story. 3 Instrumentation Developments: Minnesota Aerosol Analyzing System (MAAS) Until the late 1960s aerosol size distribution was obtained using impactors and filter measurements, electron microscopy and other time consuming techniques that precluded continuous monitoring of atmospheric aerosol dynamics. The total number of measured size spectra that covered the 0.1-10 m range was very limited, probably in the range of 50-100 spectra. That has changed rapidly when in the mid 1960s Kenneth Whitby and Benjamin Liu at the University of Minnesota Particle Technology Laboratory developed or adapted a suite of instruments for near continuous in situ monitoring of atmospheric aerosol size spectra, Whitby et al., (1972a). It consisted of an electrical mobility analyzer, an optical counter and a condensation nuclei counter. The heart of the Minnesota instrument package was the electrical mobility analyzer covering the size range 0.008–0.5 m. Atmospheric particles were passed through a diffusion charger and subsequently separated by the electric mobility. The current resulting from the charged particles in a given mobility range was used as a measure of aerosol concentration in that range. The instrument relied on the fact that the electrical mobility of charged particles was monotonically decreasing with increasing particle size as shown in Figure 2. This allowed particle size segregation in the size range 0.008-0.5 m. A history of the electrical mobility analyzer is given in Sem and Whitby (2000). Figure 2. Electrical mobility as function particle diameter. Note the distribution of mobility due to nonuniform charging. 4 Ken Whitby and Ben Liu have also made significant improvements in the size resolution of commercial optical particle counters. In the commercial counters a wide jet of atmospheric aerosols was passed through the entire illuminated volume and the scattered light pulse from the individual particles was classified by a multi-channel pulse height analyzer. Whitby and Liu have introduced the “sheath” air inlet such that the aerosol particles were passed through a uniformly illuminated center of the light beam. The result was a significant improvement in the size resolution. The availability of monodisperse polystyrene latex (PSL) beads in the size range 0.3-2.0 m allowed accurate size calibration of the optical counters. The performance of the optical counter was also evaluated by theoretical values arising from Mie calculations. The MAAS was not without problems. The most uncertain part of the measured size distributions was in the size range 0.3-0.5 m, which was at the “edge” of the size ranges for both electrical mobility analyzer and the optical counter. In fact, at about 0.4 m neither instrument could be trusted to provide high quality data. In case of the electrical mobility analyzer the particle electrical mobility curve becomes virtually size independent, as discussed earlier Figure 2. Hence, particle misclassification in that size range was rather likely. On the other hand, the optical counter was at the lower limit of its size detection. The light pulses generated by 0.4 m particles were barely above the photomultiplier noise and it was plausible that some of the noise was registered as particle signal in the 0.4 m range. Repeated field calibrations of the optical counter with polystyrene latex (PSL) beads provided some comfort that the OPC data are real. Another source of uncertainty was that the Royco model 220 had a light scattering sensor at 90 degree from the light source. It was known from Mie calculations and experimental data that at that angle the amount of light scattering is highly dependent on particle shape. Hence, if a significant fraction of the smog aerosol was composed of irregularly shaped soot particles, then the PSL calibration would be inappropriate. Figure 3. The Minnesota Aerosol Analyzing System (MAAS) prior to shipment to the 1969 Pasadena Study. 5 The Minnesota Aerosol Analyzing System, MAAS, as prepared for the 1969 Pasadena experiment is shown in Figure 3. The white cabinet houses one of the first TSI commercial mobility analyzers (Whitby Aerosol Analyzer). The Royco optical counter, equipped with the sheath air inlet along with the multi-channel pulse height analyzer is house in a separate cabinet. The suite of MAAS instruments was completed by the General Electric condensation nuclei counter for continuous monitoring of the total nuclei count, i.e. particles greater than 0.01 m. The analog and digital data from these MAAS and other meteorological instruments were gathered by a modern data acquisition system in 20 minute intervals. 3 As a result a total of 350 complete size distributions were recorded during the three-week Pasadena study, September 320, 1969. The digital data recorder was a teletype machine equipped with the punch tape unit that recorded the digital data as holes in a continuous paper tape4. This data acquisition and recording system allowed swift computer processing of the size distributions and other monitoring data. In fact, the prompt availability of the size distribution data to the other collaborating research groups allowed synergistic collaboration between the Minnesota and other groups5. The 1969 Pasadena Study: A Stimulating Environment The 1969 Pasadena Smog Caper was conceived and conducted by three pioneers of atmospheric aerosol science K. T. Whitby, S. K. Friedlander and P. K Mueller6. Whitby was a professor of mechanical engineering at University of Minnesota, with keen interest in developing aerosol instrumentation, and measuring the atmospheric aerosol size distributions. Friedlander a professor of chemical and environmental engineering at the California Institute of Technology was the ‘theoretician’ working on the explanations of the measured size distribution data using principles of aerosol dynamics. Peter Mueller, head of the Air Industrial Hygiene Laboratory in Berkley was the chemical expert, for both aerosols and gases with deep experience in chemical measurements7. The measurements were conducted in September 1969 at the Keck Laboratory of Caltech, Pasadena, California. 3 Ken Whitby was justifiably very proud of the modern data acquisition system of MAAS. He talked about it more frequently than about the aerosol instrumentation. 4 In 1969, data editing was rather similar to the currently used cut-and-paste graphic user interfaces. The difference was that in 1969 cut meant cutting the paper punch tape with scissors, replacing the bad data with a new punched tape section and pasting the new tape segment with the appropriately perforated sticky tape. 5 For example, Dave Ensor and Bob Charlson could immediately work on calculating aerosol light scattering from the size distribution data and compare those to their four wavelength nephelometer data. 6 I had the good fortune of being the PhD student of Ken Whitby and postdoctoral associate of Sheldon Friedlander during the time when the bimodal distribution concept has emerged and matured. Obviously, in the right place at the right time in the right company. 7 Prima donnas , collaboration Whitby mused about this as well. 6 Significant factors in the success of the 1969 Pasadena study were the high ‘visibility’ and the stimulating intellectual environment. The leaders of the project Whitby, Friedlander Mueller provided clear guidance. The Pasadena Aerosol project has also attracted several (then) junior scientists e.g. Bob Charlson, George Hidy who eagerly contributed their instruments and rich set of ideas while the monitoring was in progress (e.g. Figure 4)8. Figure 4. Scene from the 1969 Pasadena Study: Bob Charlson of University of Washington (center) leading an animated discussion next to the new four-wavelength nephelometer, the ‘coffin’. Throughout the experimental period, Sheldon Friedlander as the host, has invited many visiting dignitaries, including Arie Haagen-Smit, the “discoverer” of the Los Angeles smog and chairman of the California Air Resources Board. This high visibility and stimulating company has provided ample impetus to analyze and “show-off” the collected data and to discuss the features of the size distribution with the participants and visitors – virtually as it happened. The collective frame of mind of the Minnesota group was particularly upbeat 9. (Figure 5). There was no question in minds of the Minnesota contingent, that something really interesting will be discovered about the Los Angeles smog aerosol, but what will that be? 8 For us graduate students this was an intense learning experience on how “science” is conducted, how scientist operate. I never forget a scene with Bob Charlson, probably the one pictured in Fig. 5. Bob arrives from the airport, enters the lab and immediately looks at the strip charts of his four-wavelength nephelometer. First, it showed a large spread of the scattering and then abruptly the scattering in the four wavelengths overlapped. Eureka! Bob turns to me and says: ‘Lets write a paper together comparing his nephelometer data with our sizes distributions’. It took full 5 minutes between his entrance and the plan to publish! 9 In fact, I talked Ken Whitby and Ben Liu into posing with the fake headline “U of Minn. Solves L.A. Air Pollution”. Little did we suspect that 30 years later ‘U of Minn.’ helped solving the national air pollution problem. 7 Figure 5. Ken Whitby (left) and Ben Liu of University of Minnesota pose in front to the Minnesota Aerosol Analyzing System (MAAS) in Pasadena. The LA Smog Data: The Bimodal Distribution Since the MAAS data acquisition and recording were completely automatic it freed much of our time for data displays, special experiments, and stimulating discussions with the participants and visitors. One of the size distribution displays posted on the bulletin board (Figure 6) early in the measurement period was the sequence of “Typical daily changes in spectra.”10 It showed a strong increase of 0.01-0.02 m size range during the day compared to the late night hours. So, we were looking for aerosol dynamics, we have observed aerosol dynamics from the first day on. 10 If Ken Whitby was there when I posted this chart, he would have given me a real lecture on statistics. Two or three days of data are NOT enough to derive as “typical” spectra. 8 Figure 6. Posting of early smog size distribution data, plotted as number, surface and volume spectra. Note the evidence of increased daytime concentrations. Once the size distributions were plotted routinely on daily basis, a persistent feature of the LA smog aerosol was appearing: a sharp peak in the volume distribution at about 0.3 m. It was even evident in log-log plots that we have used at that time (Figure 6). The sharpness of the volume peak was rather puzzling since it deviated rather significantly from our expectation of a broad Junge type distribution. But was the persistent volume peak at 0.3 m ‘real’ or ‘instrumental’? Throughout the measurement period the Minnesota group and others have discussed the possibility that the sharp peak was due to instrumental artifacts. This was rather plausible since the volume peak occurred at the “edge” of the size ranges for both electrical mobility analyzer and the optical counter as discussed in the previous section. A certain level of comfort arose from the fact that in most days the concentration of 0.1-0.2 m size particles co-varied with 0.60.8 m size range. Since these ranges were well within the sensitive range of the respective instruments, it indicated that there was indeed a single aerosol mode in that size range. Nevertheless, this measurement uncertainty in the 0.3-0.5 m size range was the single most troublesome aspect of the Pasadena measurements11. Other aspects of the dynamic size distribution data and the special experiments on aerosol formation and decay are discussed in the next section on aerosol dynamics. 11 In fact, during my thesis work in 1970 I have spent over half of the time on re-calibrating the electrical mobility analyzer with emphasis on the 0.1-0.5 m size range. In doing so, the experience and the guiding wisdom of Ken Whitby was invaluable. In fact, as still vividly remember his baseball analogy “In baseball it is perfectly OK to miss two third of the balls, as long as one hits one third. In science, he said, there is a ten-to-one rule. If you miss once, it takes ten good hits to fix the missed hit.” This was his way of telling that there is no way we should miss the existence of this fine mode peak. The mobility distribution chart in Figure 2 was a key result of that work. 9 Stretching the plot. Around 1970, while searching the literature for my thesis, a ran across a paper by Berry (1967) who made the strong case that for coalescing (and coagulating) aerosol systems, one should plot the volume distribution on semi-log paper: dV/dlogD. In such a plot, the area under the volume distribution curve is proportional to the volume concentration. For depicting dynamic aerosol processes this is advantageous, since it shows, in a quantitative way, the shifts in particle size as well as gain or loss in mass. Equally important to us at that time, the linear volume scale accentuated the peaks of the volume distribution which made the bimodal distribution much more apparent than in the more flat log-log plot. Ken Whitby has immediately accepted the use of the semi-log plot and from there on all the volume spectra were published in the semi log form as shown in Figure 7 . (Whitby et al., 1972b, Husar et al., 1972)12. Figure 7. Normalized ‘grand average’ size distribution of the LA smog with clear indication of the bimodal volume distribution. The influential processes included nucleation, coagulation, condensation and sedimentation. The Grand Average Los Angeles and Minneapolis Spectra. Probably the most succinct depiction of the bimodal distribution concept is the normalized ‘grand average’ size distribution plot shown in Figure 2 taken from the summary paper Whitby et al., 1972b)13. Figure 7 shows the grand averages for LA, including the number, surface and volume weighing of the smog aerosol spectra, the different subranges, as well as the dominant mechanisms that govern the 12 Throughout the years, I was rather proud of this ‘unique contribution’ to the discovery of the bimodal distribution. It was not until Friedlander (2000) has brought to our attention the fact that Junge (1963) has used that type of a plot, probably for the same reasons as we did! It was just one of the many re-discoveries. 13 Whitby was always keen on statistics and such ‘grand averages’ were key part of his analytical toolbox. 10 aerosol dynamics in each subrange. The semi-log plot clearly conveys the bimodal volume distribution with peak in the 0.1-1.0 m size range. It also indicates the rise of the mass spectra above 2m but the data abruptly end at 6 m since in the Royco optical counter reached only up to that size. The grand averages for Minneapolis, when presented in the same way, showed a similar pattern including the volume peak in the 0.1-1.0 m size range. However, the story of the bimodal distribution was not quite complete. The lingering questions were: Was there an equally sharp coarse particle mode and at what size? How do the two modes compare? Is the bimodality specific to LA and Minneapolis? Completing the Bimodal Distribution Story: The Coarse Particle Mode. Around 1970-71, Whitby has completed the bimodal distribution story by collecting and analyzing several size distribution data sets arising from different locations, times, and sampling methods. Whitby presented the summarized data in Figure 14 in Whitby et al. (1972b). The broad range of data provided strong evidence that the bimodal distribution is ‘real’ i.e. it occurs as a ubiquitous feature of atmospheric aerosols in general. The lower, fine particle mass mode was consistently between 0.1-1.0 m but it was shifting considerably between 0.1 and 0.8 m depending on location and time. All the data have indicated a saddle point between the modes in the 1-3 m size range. The assembled data showed a distinct coarse particle peak but the characteristic size of the coarse particle mode tended to increase with increasing coarse mass. With this, the bimodal distribution as a new aerosol model was established to Whitby’s satisfaction and the summary paper on the bimodal distribution (Whitby et al.,1972) was ready for publication. The rest, as they say, is history. 11 The Explanation of the Bimodal Distribution: Aerosol Dynamics The Whitby et. al., (1972b) paper described the bimodal distribution but the causal processes and mechanisms were not elaborated. Key scientific issues have not been tackled: What is the explanation for the bimodality? What are the key driving mechanisms each of the modes? The introduction of the bimodal concept has allowed swift progress in establishing the governing mechanisms throughout the size spectrum, including the nuclei and accumulation modes. It was no longer necessary to lump the entire size range (0.1-20 m) into a single aerosol dynamic system. Dynamics of the Coarse Mode In the summer of 1970 Whitby and Husar have participated in another aerosol study in Ft. Collins, Colorado. While continuously monitoring the ambient aerosol, they observed the changing size distributions during the passage of a dust event with a strong increase of coarse particles and simultaneous decrease of fine particle mass. Figure 8. Measured size distribution of dust in Colorado showing a strong coarse particle peak. It is compared to the average LA smog aerosol. The dusty aerosol spectra for Fort Collins is compared to the grand average LA aerosol in Figure 8. Such observations have confirmed that the dynamics of the fine and coarse mode aerosols tended to vary independently. This is a key consequence of the bimodal model, since the modes were governed by different sources, formation, transport, and removal mechanisms. 12 By the 1970s, the scientific literature has well established that coarse sea salt particles were the result of mechanical breakup and dispersion and that sedimentation was the key removal mechanism (e.g. Blanchard, 1955). It was also generally accepted that wind blown dust mass in the coarse particle size range, well above 1.0 m (e.g. Gilette, 1972). With the physical processes and the dynamics of the coarse mode understood, in principle, no further effort was invested in the coarse mode. Besides, suitable instrumentation for the measurement of large particles (>10 m) was not available. Dynamics of the Fine Particle Spectra The remaining scientific issues pertained to the dynamics of sub-micron size particles. As seen in the grand average Figure 7, the fine particle mode is at very different location depending on the weighing of the size distribution. The mode of the fine particle number concentration was about 0.01 m, the surface area peaked at 0.2 m, while the volume at 0.3 m. It became evident early on during the observations that the dynamics of the number concentration was virtually decoupled from the dynamics of the fine particle volume spectra. For this reason, the fine particle mode could be sub-divided into a ’nucleation’ range (0.01-0.1 m) and an ‘accumulation’ range. The level of physico-chemical understanding in these subranges was considerably weaker since the sources, transformation and removal processes were only partially established. Furthermore, the nature of the interaction between the nucleation and accumulation mode was unclear. Ever since the landmark work of Haagen-Smit (1952) on the chemistry and physiology of the Los Angeles smog, it was clear that visible smog particles are largely due to photochemical reactions and gas-particle conversion processes. The subsequent aerosol studies (e.g. Leighton, 1961; Renzetti and Doyle, 1960; Stevenson et al., 1965; Goetz et al., 1961; Goetz and Pueschel 1965; Goetz and Pueschel, 1967) have revealed that the formation of light scattering aerosols was driven by the concentration of gaseous precursors. Goetz and Pueschel (1967) have also shown that in the aerosol-rich Los Angeles basin, the gas to particle conversion process occurs mostly by heterogeneous nucleation on existing particles rather than homogeneous nucleation. These studies have revealed a very complex interaction between the precursor gases (SO2, NOx, hydrocarbons, and water vapor). As Goetz and Pueschel (1967) point out, “the available results did not permit even a preliminary interpretation of the very complex [positive and negative] reaction pattern involved in smog aerosol formation”. In particular, the evolving size distribution of aging photochemical aerosol was not on hand. This was the topic of the third paper in the Minnesota trilogy on the physical mechanisms governing the dynamics of Los Angeles aerosol (Husar et al., 1972) 14. Given the extensive aerosol dynamics data set collected during the 1969 Pasadena Study it was possible to infer the governing mechanism based on the observations rather than from purely 14 In preparing for the Pasadena aerosol study, Whitby has frequently expressed his anticipation that the project will reveal the dynamic size changes of aging photochemical aerosols. Hence, obtaining and interpreting such data was definitely not an accident. I was fortunate that by the time of the Pasadena experiments, I had learned enough about Whitby’s size distribution ideas and Friedlander’s aerosol dynamics approach, so that I could tie the two together – during the in Pasadena study and later in finishing my PhD theses work. 13 theoretical considerations. The key issues at hand were the specific roles of nucleation, coagulation and heterogeneous gas particle conversion. Dynamics of the Nuclei Subrange: Coagulation The role of coagulation was the easiest to evaluate and quantify based on the ambient data, laboratory experiments, and numerical simulation15. Coagulation was found to be a significant mechanism for the removal of particles in the size range below 0.1 m. Night after night, the smog particle concentration in 0.01-0.1 m size range declined by an order of magnitude or more with the corresponding reductions in the total number concentration. The diurnal pattern of this size range was simulated numerically by solving the coagulation equation with initial conditions corresponding to the daytime spectra. The results of such simulation have shown that the decline of nighttime concentrations in 0.01-0.1 m size range could be explained by coagulational mass transfer to the “accumulation” mode in the 0.1-1.0 m size range. ?????Hence, it was established that the total number concentration of particles was dominated by primary emissions (presumably from automobile), some photochemical nucleation during the midday hours followed by decay due to coagulation over the night hours. Condensation Establishing the dynamics of the accumulation mode aerosols focused on three factors: 1) primary emissions into that subrange; 2) coagulation transfer from the nucleation mode; 3) condensational growth by heterogeneous condensation of photochemically produced vapors. Coagulation mass transfer from 0.01-0.1 m size range was quickly eliminated as a significant factor for the accumulation mode. Based on Monte Carlo coagulation simulations the amount of mass transferred by coagulation into the accumulation mode range was found to be insignificant. The role of primary emissions for the accumulation mode was also assessed to be of weak significance for the Pasadena aerosol. This was based on the observation that the mass in the accumulation mode had a peak during the midday hours, while the various tracers of primary emissions (NO, condensation nuclei) showed a morning peak and an evening peak associated with rush hour traffic. Thus, the remaining explanation for the midday growth focused on the heterogeneous conversion of condensable species. The sharp mid-day accumulation of the aerosol volume is illustrated in Figure 9a. I shows a virtually constant-volume mean size as the fine particle volume increased in the morning and decreased in the afternoon. This was a puzzling pattern since condensational growth of particles tend to increase the mean particle size. This paradox was resolved by a set of condensational growth calculations reproduced in Figure 9b. In this case, the initial size distribution was taken to be the measured spectra at 10:00 AM, and subsequently grown by diffusion limited condensation. An important constraint in the condensational growth equation was the assumption 15 Besides, this was the topic of mine Ph.D. dissertation “Coagulation of Knudsen Aerosols”, so I was most comfortable with the topic, Husar, 1971. 14 of a critical particle size of 0.09 m. Particles below this size range were taken to be below the “Kelvin cutoff” and therefore not subject to condensational growth. Figure 9. Measured (left) and simulated (right) growth of photochemical smog aerosols. The simulation assumes condensation of precursor gases onto existing particles. A comparison of the observed and simulated condensational growth pattern revealed that in both cases the volume mean diameter has remained constant but the magnitude of the peak has increased. In other words, the mass has accumulated without much change in the mean size – hence the term ‘accumulation’ mode. The explanation of the paradox was that for a broad initial size distribution, the small particle range growth faster than the large ones. In other words, the abundance of 0.1 m particles more than replenished the particles growing out of the 0.3 m sub-range. Special Studies on Aerosol Dynamics During the 1969 Pasadena Study several aerosol dynamics experiments were carried out to further explore the dynamics of the Los Angeles aerosol. In the “big bag” experiment a large plastic balloon of 50 m3??? was rapidly filled with ambient aerosols. Subsequently, the aging size distribution in the bag was monitored to measure the aging aerosol. It was confirmed that coagulation rapidly reduced the concentration of total number of particles. The fact that even in the presence of sunlight the number concentration has decayed by coagulation indicated that in the presence of smog aerosol, homogeneous nucleation was not taking place. Repeating the experiment but filling the bag with aerosol-filtered air yielded dramatically different results. The total number concentration in 0.01-0.03 m sub-range rapidly increased within 15 minutes from zero to over 105 particles/cm3. Clearly, the gaseous precursors in the “big bag” illuminated by sunlight were self-nucleating. Repeating the self-nucleation experiments after sunset yielded no nucleation. This confirmed that the nucleation of gaseous precursors was driven by solar radiation. Conversely, this special experiment confirmed that in the presence of 15 smog particles the self-nucleation was suppressed and that the gas–particle conversion must have occurred onto the existing particles. In another special experiment the smog aerosol was artificially humidified by passing air by a wet surface which increased the relative humidity from 25% to 86%. The measured aerosol volume concentration increased from 26 to 38 m3/cm3 as a result of humidification. However, the mean particle size did not change. This experiment implied that a considerable fraction of the smog aerosol in the sub-micron range was hygroscopic. Therefore, relative humidity was an important meteorological parameter influencing the dynamics of the smog aerosol. At that time, this was new information since previous reports by Cadle at al. (1950) and others indicated that smog aerosol was ‘dark brown gummy, water-insoluble organic material’. Based on the field observations, special experiments and numerical simulation, the following conclusions were reached regarding the Los Angeles smog aerosol dynamics (Husar at al., 1972; Husar and Whitby, 1973). The driving force for the photochemical gas-particle conversion in the accumulation mode were gaseous photochemical reactions producing condensable species. In the presence of foreign nuclei, the concentration of these species was relaxed by deposition on existing particles (heterogeneous nucleation). The growth of the accumulation mode aerosol was consistent with diffusional condensation. The amount of surface area required for heterogeneous nucleation was found to increase with conversion rate. In the polluted Los Angeles atmosphere the growth on foreign nuclei dominated the gas-particle conversion rate and this condensational growth was largely responsible for the growth in the ‘accumulation’ mode. Coagulation tended to limit the lifetime of particles below 0.1 m in size but it did not influence the concentrations in the accumulation mode. The smog aerosol was found to be hygroscopic. Summary In the view of this undoubtedly biased chronicler, the most significant factor in the emergence of the bimodal aerosol model was Whitby’s new aerosol measuring system for dynamic, in situ size distribution measurement over three decades of particle size. It provided the observational evidence of the ubiquity of the two mass modes. The aerosol chemistry data organized first by Peter Mueller and subsequently analyzed by Friedlander and coworkers showed that the fine and coarse mass modes were chemically distinctly different. In fact, the differences in the chemical composition of different aerosol types is the basis for Friedlander’s chemical mass balance method (Miller et al., 1972; Heisler et al., 1973). The aerosol optical data by Bob Charlson and co-workers confirmed that the accumulation mode was responsible for the scattering of visible light (Ensor et al., 1972). In the context of the bimodal size distribution history, it is ironic how close Junge was not only to observe but also to explain the remarkable regularities of atmospheric aerosols. He clearly identified aerosol mass modes for specific aerosol types (e.g. upper tropospheric fine particle mode and marine coarse mode). The deviation between the Junge and the Whitby models was in the shape of the continental aerosol spectra, composed of a mixture of different aerosol types with varying size distributions. Junge indicated a broad flat distribution in the range 0.1-200 m 16 for the mixture of the different continental aerosol types. On the other hand, the Whitby distribution for the same continental aerosol mixture showed two distinct modes16. The second factor was the use of aerosol dynamics as introduced and promoted by Sheldon Friedlander. Applying these analytical tools helped Husar and others to explain the observed fine particle dynamics and hence provided a robust scientific support for the bimodal concept based on accepted physical mechanisms. These mechanisms are the basis of the current regional dynamically coupled gas-aerosol models that are the key tools for the implementation of air quality management programs. Thirdly, a major factor in the discovery, promotion and the general acceptance of the bimodal distribution concept was the unprecedented collaborative spirit of the visionary organizers of the 1969 Pasadena Aerosol Study and other participants. That collaborative and supportive spirit created a most stimulating environment for the collective creation of the new atmospheric aerosol paradigm. References Berry, E. X. Cloud droplet growth by collection. J. Atmos. Sci., 24, 688-701, 1967. Blanchard, D. C. Bubble formation and modification in the sea and its meteorological significance. Tellus, 9, 145-158, 1957. Cadle, R. D., Rubin, S., Glassbrook, C. I. and Magill, P. L. Identification of particles in Los Angeles smog by optical and electron microscopy. Arch. Industr. Hyg. Occup. Med., 2, 1-18, 1950. Clark, W. E. and Whitby, K.T. Concentration and size distribution measurements of atmospheric aerosols and a test of the theory of self-preserving size distributions. J. Atmos. Sci., 24, 677687, 1967. Ensor, D. S., Charson, R. J., Ahlquist, N. C., Whitby, K T. Husar, R. B. and Liu, B. Y. H. Multiwavelength nephelometer measurements in Los Angeles smog aerosol-I, Comparison of calculated and measured lighy scattering. J. Colloid Interface Sci., 39, 242-., 1972. Friedlander, S. K. On the particle size spectrum of atmospheric aerosols, J. Meteor., 17, 373-374, 1960a. Friedlander, S. K. Similarity considerations for the particle size spectrum of a coagulating, sedimenting aerosol, J. Meteor., 17, 479-483, 1960b. Friedlander, S. K. Theoretical considerations for the particle size spectrum of the stratospheric aerosol. J. Meteor., 18, 753-759, 1961. Friedlander S. K. The similarity theory of the particle size distribution of the atmospheric aerosol. Proc. First Natl. Conf. Aerosols, Liblice, 8-13 October 1962, 115-130, 1965. Friedlander, S. K. and Wang, C. S. The self-preserving particle size distribution for coagulation by Brownian motion. J. Colloid Interface Sci., 22, 126-132, 1966. Friedlander, S. K. A brief history of aerosol dynamics. In “History of Aerosol Science” Verlag der Oesterreichischen Akademie der Wissenschaften (Preining O. and Davis E. J. Eds.), Vienna, 2000. A few good data points in the saddle range between the modes (1-3 m) would have been probably enough to reveal to Junge the bimodal nature of the continental aerosol. 16 17 Gilette, D. A. Measurement of aerosol size distribution and vertical fluxes of aerosols on land subject to wind erosion. J. Appl. Met., 11, 977-987, 1972. Goetz, A., Preining, O. and Kallai, T. The metastability of natural and urban aerosols. Geofis. pura Appl., 50, 67-80, 1961; Goetz, A. and Pueschel, R. The effect of nucleating particulates on photochemical aerosol formation. J. Air Pollut. Control Assoc., 15, 90-95, 1965; Goetz, A. and Pueschel, R. Basic mechanisms of photochemical aerosol. Atmos. Environ., 1, 287-306, 1967. Haagen-Smit, A. J. Chemistry and physiology of Los Angeles smog. J. Ind. Eng. Chem., 44, 1342-1346, 1952. Heisler, S. L., Friedlander, S. K. and Husar R. B. The relationship of smog aerosol size and chemical element distributions to source characteristics. Atmos. Environ., 7, 633-649, 1973. Hidy, G. M. On the theory of the coagulation of noninter-acting particles in Brownian motion. J. Colloid Interface Sci., 20, 123-144, 1965. Hidy, G. M. Characterizing atmospheric aerosols in California. In “History of Aerosol Science” Verlag der Oesterreichischen Akademie der Wissenschaften (Preining O. and Davis E. J. Eds.), Vienna, 2000. Husar R. B. “Coagulation of Knudsen Aerosols”, Ph.D. thesis, University of Minnesota, 1971. Husar, R. B., Whitby, K. T., and Liu, B. Y. H. Physical mechanisms governing the dynamics of Los Angeles aerosol. J. Colloid Interface Sci., 39, 211-224, 1972. Husar, R. B., and Whitby, K. T. Growth mechanisms and size spectra of photochemical aerosol. Environ. Sci. Technol., 7, 241-247, 1973. Junge, C. E. Die Rolle der Aerosols und der gasformigen Beimengungen der Luft im Spurenstoffhaushalt der Troposhere. Tellus, 5, 1-26, 1953, Junge, C. E. The size distribution and aging of natural aerosols as determined from electrical and optical data on the atmosphere, J. Meteor., 12, 13-25, 1955. Junge, C. E. Remarks about the size distribution of natural aerosols. In “Artificial Stimulations of Rain”, New York, Pergamon Press, pp3-16, 1957. Junge, C. E. Atmospheric Chemistry. Advances in Geophys., 4, 1-108, 1958. Junge, C. E., Chagnon C. E. and Mason, J. E. Stratospheric aerosols, J. Meteor., 18, 81-108, 1961. Junge, C. E. Comments on “Concentration and size distribution measurements of atmospheric aerosols and a test of the theory of self-preserving size distributions”, J. Atmos. Sci., 26, 603608, 1969. Junge, C. E. “Air Chemistry and Radioactivity”, Academic Press, New York, 1963. Leighton, P. A. “Photochemistry of Air Pollution”, Chapter IX, Academic Press, New York, 1961 Miller M. S., Friedkander S. K. and Hidy G. M. A chemical element balance for the Pasadena aerosol. J. Colloid Interface Sci., 39, 165-176, 1972 Preining, O. and Davis, E. J. (Eds.) “History of Aerosol Science” Verlag der Oesterreichischen Akademie der Wissenschaften,Vienna, 2000. Renzetti, N.A. and Doyle, G. J. Photochemical aerosol formation in sulfur dioxide hydrocarbon system. Air Wat. Pollut. Int. J., 2, 327-345, 1960. Sem, G. J. and Whitby, E. R. Kenneth Thomas Whitby: A pioneer of aerosol characterization. In “History of Aerosol Science” Verlag der Oesterreichischen Akademie der Wissenschaften (Preining O. and Davis E. J. Eds.), Vienna, 2000. 18 Stevenson, H. J. R. Sanderson, and D. E. Altshuller, A.P. Formation of photochemical aerosols, Air Wat. Pollut. Int. J., 9, 367-375, 1965. Whitby, K. T., Liu, B. Y. H., Husar, R. B., and Barsic, N. J. The Minnesota Aerosol Analyzing System used in the Los Angeles Smog Project, J. Colloid Interface Sci., 39, 136-164, 1972a. Whitby, K. T., Husar, R. B. and Liu, B. Y. H. The aerosol size distribution of Los Angeles smog. J. Colloid Interface Sci., 39, 177-204, 1972b. 19