Formation and Transformation of Metastable Double Salts from the

Ammonium nitrate (AN) and ammonium sulfate (AS) are ubiquitous components of atmospheric aerosols. Thermodynamic models predict formation of pure (AN ...
0 downloads 0 Views 909KB Size
Download PDF
Environ. Sci. Technol. 2007, 41, 8077–8083

Formation and Transformation of Metastable Double Salts from the Crystallization of Mixed Ammonium Nitrate and Ammonium Sulfate Particles TSZ YAN LING AND CHAK K. CHAN* Department of Chemical Engineering, Hong Kong University of Science and Technology, Clear Water Bay, Kowloon, Hong Kong

Received June 13, 2007. Revised manuscript received August 18, 2007. Accepted September 18, 2007.

Ammonium nitrate (AN) and ammonium sulfate (AS) are ubiquitous components of atmospheric aerosols. Thermodynamic models predict formation of pure (AN and AS) and double salts (3AN · AS and 2AN · AS) for the AN/AS system. Because of the high supersaturation at which a droplet crystallizes, metastable crystal formation is possible. In this study, the identity of the crystals formed from the crystallization of equimolar AN/ AS mixed droplets was investigated in an electrodynamic balance coupled with a Raman spectroscopic system. Raman spectra of bulk AN/AS double salts possibly formed in this system are first reported for comparison with the single particle Raman results. The double-salt 3AN · AS, not predicted from thermodynamics, was observed in the freshly crystallized single particles. The degree of metastability can be different among several crystallization processes of the same particles. The metastablesalt3AN · ASgraduallytransformedintostable2AN · AS, and the rate of such transformation increased with increasing relative humidity. This study illustrates the possibility of occurrence of metastable salts in atmospheric aerosols.

Introduction Atmospheric particles are known to play an important role in affecting radiative forcing and climate (1). Ammonium nitrate (AN), ammonium sulfate (AS), and protons (H+) are the predominant inorganic components in atmospheric aerosols (2). There have been many laboratory studies on the thermodynamics, phase transition, and hygroscopicity of their mixtures. Atmospheric SO42-/NO3-/NH4+/H+ particles can be either solid or in aqueous state, depending on the relative humidity (RH) history of their immediate environment (3). Thermodynamic models, for example, the aerosol inorganic model (AIM) (4, 5) and the UHAERO model (6), have been developed to predict the phase, water content, and solid formation of aerosols with various ionic compositions and at different RH and temperature. In the SO42-/NO3-/ NH4+/H+ system, complex salt formation is possible upon crystallization from solutions (4). Martin et al. (7) summarize the crystals formed in the SO42-/NO3-/NH4+/H+ system as predicted by the AIM and represent the results in form of an * Corresponding author phone: (852) 2358-7124; fax: (852) 23580054; e-mail: [email protected]. 10.1021/es071419t CCC: $37.00

Published on Web 10/26/2007

 2007 American Chemical Society

isothermal phase diagram. For the subsystem containing mixtures of AN and AS only, the two pure salts (AN and AS) and the two double salts, 3(NH4NO3) · (NH4)2SO4 (3AN · AS) and 2(NH4NO3) · (NH4)2SO4 (2AN · AS), are predicted to form, depending on the AN/AS mixing ratios. While equilibrium thermodynamic models are available, there have been laboratory aerosol studies reporting the formation of metastable salts, not accounted for in the thermodynamic predictions. For example, inorganic salts like Na2SO4, LiClO4 (8), and NaClO4 (9) and water soluble organic salts like sodium formate and sodium acetate (10) form metastable anhydrates and hydrates that are not predicted by the bulkphase thermodynamics. When a solution contains more than one solute, in addition to metastable hydrate/anhydrous salts, metastable crystals with a different chemical identity may form. Colberg et al. (11) characterized the phase transformation of H2SO4/ NH3/H2O droplets and found that droplets with an ammonium-to-sulfate ratio of 1 (corresponding to the composition of ammonium bisulfate (AHS)) did not crystallize into AHS, which is the product predicted by the AIM. Instead, (NH4)3H(SO4)2 was formed. Braban and Abbatt (12) and Badger et al. (13) used an aerosol flow-tube FTIR system to study the phase transition of mixtures of AS with malonic/ humic acid. On the basis of the shifting of the NH4+ absorption peak, they proposed the formation of an AS–humic acid complex upon crystallization, which converts into pure AS before deliquescence. In a series of aerosol flow-tube FTIR experimental studies, Martin and co-workers (7, 14, 15) examined the crystallization properties of the SO42-/NO3-/NH4+/H+ system. Schlenker et al. (14) described the dependence of the types of crystals formed on the overall aqueous composition. Most of their results are consistent with the predictions of the AIM. However, they also observed the formation of metastable crystals. In some cases, part of the salts chemically transformed into the thermodynamically stable forms at higher RH (but below the deliquescence RH, DRH). These results further support the possibility of metastable crystal formation, and the proposition that RH affects the transformation of metstable components. Single-particle Raman spectroscopy has been found to be useful in the characterization of the properties of levitated particles, namely, the size and composition of droplets (16–18), the hydration states of solid particles (8, 19) and the phase-transition process (20, 21). Other examples of its application, including monitoring particle reactions, are available in refs 22–25. The methodology built upon our previous experience using an electrodynamic balance/Raman (EDB/Raman) system to study the hygroscopic properties, molecular structures, and phase transformations of aerosols (26–28) and, more recently, the heterogeneous reactions of organic particles (29, 30). In the present study, such a system was applied to characterize the crystals formed from a supersaturated AN/AS solutions and to observe the transformation of the metastable particles in situ. The effects of RH on the transformation were also examined. To identify the multiple salts formed in single-particle crystallization experiments reference Raman spectra were first acquired for the bulk salts which are possibly formed in this system, that is, pure AN and AS, as well as 3AN · AS and 2AN · AS. The results contribute the first report of the Raman spectra of the isolated double salts 3AN · AS and 2AN · AS. VOL. 41, NO. 23, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

8077

Methods The experiments involved both bulk and single-particle Raman characterizations. The reference Raman spectrum of each salt that can possibly form in this system was determined with bulk samples, and such spectra were then applied to determine the chemical identity of the crystals formed in the single particles. Bulk Studies. While pure AN and AS crystals are readily available, the double salts possibly formed in the AN/AS system are not commercially available. They were thus prepared in bulk following the methods used in ref 31 and characterized by ion chromatography (IC) for their compositions and X-ray diffraction (XRD) (32) for their crystallographic properties. The NO3-/SO42- ratios of the salts collected clearly distinguished the two double salts. The XRD results (not shown) were comparable to the data reported in Smith et al. (34) and Coates and Woodard (35). Details of the double-salt preparation, IC, and XRD characterizations can be found in the Supporting Information. Single-Particle Studies. Single AN/AS ) 1:1 (molar ratio) aqueous droplets were levitated in an electrodynamic balance (EDB) (29). A trapped droplet, typically 20–60µm in diameter, was first crystallized at RH ≈ 30%, which is similar to the crystallization RH (CRH) (27–30%) reported in ref 14. Upon droplet crystallization, the RH of the EDB was adjusted and fixed between the CRH and DRH, and the chemical components of the particle were subsequently monitored by Raman spectroscopy at regular intervals of about 1–2 h, to monitor its transformation, until changes in the Raman signatures ceased. The particle was then subjected to RH > 85% until it had deliquesced completely. It was then recrystallized by the same procedures already described. The recrystallized particle was exposed to a different RH, and its chemical composition as a function of time was again monitored. Mass measurements did not show any detectable mass loss during the experimental period of a few days, indicating that no significant evaporative loss of volatile species had occurred. Additional experimental details can be found in the Supporting Information. Raman Characterizations. The Raman spectroscopy system coupled to the EDB consisted of a 5 W argon ion laser (Coherent I90-5) and a 0.5 m monochromator (Acton SpectraPro 500) attached to a CCD (Andor Technology DC420-OE). The resolution of the spectra was about 1.2–1.3 cm-1. The integration time for each spectrum was 60 s (10 frames, each with an accumulation time of 6 s). Raman spectra of the bulk samples, including the synthesized double salts and the pure AN and AS (used as purchased, without further purification), were obtained using the same Raman system. A few micrograms of each sample placed inside a quartz tube (1 mm diameter) was inserted into the center of the EDB where the excitation laser was focused for the Raman measurements.

Results and Discussion Figure S1 in the Supporting Information shows the bulk and single-particle Raman spectra of AS crystals. The sulfate peaks at ∼980 and 450 cm-1 of the bulk sample and single particles of AS showed no differences, while the 615 cm-1 peaks differed a little in line shape. This is probably caused by the difference in crystal morphology of the samples. Overall, the bulk and single-particle Raman spectra of AS were almost identical, supporting the use of the bulk spectra as a reference in the analysis of the single-particle results. Raman Characterization of the Bulk Salts. The four crystals, possibly formed in mixed AN/AS solutions, predicted by the AIM are pure AS, pure AN, 3AN · AS, and 2AN · AS, depending on the overall composition. The Raman spectra of these four bulk samples are shown in Figure 1. Nitrate 8078

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 23, 2007

FIGURE 1. Reference Raman spectra of bulk samples of AN, 3AN · AS, 2AN · AS, and AS. Intensities of sulfate peaks at ∼450 and 620 cm-1 are about 1/4 and 1/7 of that at 980 cm-1, while intensity of nitrate peak at ∼720 cm-1 is about 1/8 of that at ∼1050 cm-1.

FIGURE 2. Expanded view of the example Raman spectra of three freshly crystallized particles. Contributions of ASpure (peak at 975 cm-1) and ASdouble (peak at 980 cm-1) were estimated and shown in gray lines. The corresponding I1053/I1046 ratios are also indicated. Expanded view of the reference Raman spectra in Figure 1 was included for comparison. Relative intensities of the peaks of the original spectra can be referred to caption of Figure 1. shows Raman signals at ∼720 and 1050 cm-1, while sulfate signals are observed at ∼ 450, 620, and 980 cm-1. No distinct ammonium Raman peaks were observed in the 400–1100 cm-1 range. Pure AN and pure AS show only the nitrate and sulfate peaks, respectively, while the double-salt spectra exhibit both nitrate and sulfate features at various intensities. Because the peaks at Raman shift below 800 cm-1 were much weaker than those above, the rescaled spectra are shown in the bottom of Figure 2. Nitrate has a weak peak at ∼720 cm-1 and a strong peak at ∼1050 cm-1. However, the detailed features of these peaks were different in the spectra of pure AN, 3AN · AS, and 2AN · AS. The position of the strong sulfate signal at ∼980 cm-1 distinguishes the sulfate in pure AS from that in the double salts. Table 1 summarizes the spectral characteristics of the bulk salts. AN can exist in a number of polymorphic forms (36), which may have different Raman spectral features. Phase IV is the most stable form at 25 °C, and the transformation from phase IV to other polymorphic forms does not take place below 35 °C (33). Because these measurements were conducted at room temperature, the influence of the polymorphic solid-phase transformation was not considered. Detailed analysis of the spectra of the bulk samples is available in the Supporting Information. The IR spectra of AN/AS double salts have been obtained previously by the deconvolution of a batch of spectra (14), but this study was the first to obtain the Raman spectra of the isolated double salts. The spectral differences observed in the nitrate and sulfate peaks proved most useful for analysis of the single-particle Raman results. We also performed FTIR characterization of the bulk samples. Overall, our mea-

TABLE 1. Summary of Raman Spectral Features of (1) Bulk Samples and (2) 3AN · AS-Rich and 2AN · AS-Rich Single Particlesa peak location

bulk pure AN

bulk 3AN · AS

717 cm-1 (nitrate signal)

no shoulders around the main peak

two shoulders around the main peak

∼980 cm-1 (sulfate signal) ∼1050 cm-1 (nitrate signal)

NA

symmetric peak at 980 cm-1 asymmetric peak at 1053 cm-1

peak location

symmetric peak at 1046 cm-1

3AN · AS-rich particles

717 cm-1 (nitrate signal)

two shoulders around the main peak

∼980 cm-1 (sulfate signal)

a smaller shoulder on the higher wavenmuber side

∼1050 cm-1 (nitrate signal)

single peak at 1053 cm-1, large I1053/I1046 ratio

bulk 2AN · AS a single shoulder on the lower wavenumber side of the main peak symmetric peak at 980 cm-1 double peaks at 1046 and 1053 cm-1, with peak intensity ratio (I1053/I1046) equal to 1.1

bulk pure AS NA symmetric peak at 975 cm-1 NA

2AN · AS-rich particles a single shoulder on the lower wavenumber side of the main peak a dominant shoulder on the higher wavenmuber side increased dominance of 1046 cm-1 peak for double peaks, small I1053/I1046 ratio (close to 1.1)

a I1053/I1046 ) 1.1 is regarded as a feature of 2AN · AS. A deviation from this value means the Raman signal originates from samples other than 2AN · AS or mixtures of 2AN · AS with other salts.

surements are, in general, consistent with those of Schlenker et al. (14), except for some small differences in the locations of the nitrate split peaks at around 830 cm-1. Experimental procedures and results are given in the Supporting Information. Raman Characterization of AN/AS Mixed Single Particles. Experiments were performed with seven single particles, undergoing a total of 15 crystallizations: two particles went through one crystallization, three went through two crystallizations, and another two particles underwent three and four crystallizations, respectively. The discussion of the single-particle Raman results is organized as follows: (1) chemical identity of freshly crystallized particles based on the results of the 15 crystallization events and (2) chemical transformation of the crystals and its dependence on RH based on the two particles undergoing three and four crystallization events. Chemical Composition of Freshly Crystallized Particles. The AIM model predicts the formation of 2AN · AS and pure AS in equal amounts during the crystallization from an equimolar AN/AS solution. According to the model, no 3AN · AS should be formed. However, in these experiments, Raman features of both 3AN · AS and 2AN · AS, as well as pure AS, to maintain the overall equimolar AN/AS composition, were observed in the freshly crystallized particles. The signatures of 3AN · AS and 2AN · AS are observed in various degrees among 15 crystallizations. Three example single-particle Raman spectra are shown in Figure 2, corresponding to the three particles containing (1) the weakest 3AN · AS signal (relative to 2AN · AS), (2) a moderate signal, and (3) the strongest signal observed upon crystallization. These Raman spectra were taken within 15 min after crystallization of the droplets. The strong nitrate signal at ∼1050 cm-1 was an indicator for the chemical nature of nitrates. It was assumed that whenever a particle contained more than one salt, their Raman signals would superimpose linearly. From Figure 2, strong signals at 1053 cm-1 are accompanied with a weaker peak at 1046 cm-1 (Figure 2, traces 1 and 2) or a shoulder on the low-wavenumber side (Figure 2, trace 3). The peak shapes do not match with any of the reference Raman spectra,

suggesting that a mixture of nitrate salts was present. In the case of Figure 2, trace 3, the solid particle probably consisted mainly of 3AN · AS, whose reference spectrum also shows a single slightly asymmetric peak at 1053 cm-1. The shoulders around the 717 cm-1 peak also support this suggestion. Figure 2, traces 1 and 2 show obvious double peaks with different relative intensity at ∼1050 cm-1. The strong peak at 1053 cm-1 indicates the presence of 3AN · AS, while the weaker peak at 1046 cm-1 might represent either 2AN · AS or pure AN. If this were from pure AN alone, there should be shoulders of roughly equal dominance on both sides of the nitrate peak at 717 cm-1 because AN does not give any shoulders around 717 cm-1 and the shoulder characteristics should reflect the features of 3AN · AS alone. A detailed examination of the shoulders of the 717 cm-1 peak reveals a dominance of the shoulder on the lower-wavenumber side compared to the other side, suggesting that the minor peak at 1046 cm-1 indeed originated from 2AN · AS, which has a single shoulder on the lower-wavenumber side in its reference spectrum. In addition, if AN had been present in a significant amount, the main peak would have been much more dominant compared to the shoulders. Therefore, the single peak at ∼1053 cm-1 (signal of 3AN · AS) was apparently superimposed on the double peaks (with I1053/I1046 ) 1.1) of 2AN · AS. Dependent on the relative amount of 3AN · AS and 2AN · AS, the I1053/I1046 ratios were observed to take different values, but they were always larger than 1.1 for all 15 freshly crystallized particles. A large I1053/I1046 ratio implies that the particle was relatively rich in 3AN · AS, while a particle with a ratio close to 1.1 was relatively rich in 2AN · AS. The above analysis on the relative abundance of the double salts, based on the nitrate peaks, can be further supported by the sulfate peaks at ∼980 cm-1. Because the overall ratio of AN to AS remained 1:1, AS should have been present both as pure AS (ASpure) and in double salts (ASdouble). The ratio ASpure/ASdouble ) 2 when all nitrate was present as 3AN · AS, and ASpure/ASdouble ) 1 when all nitrate was in the form of 2AN · AS. Therefore, 1 < ASpure/ASdouble < 2 for particles containing both double salts. A 3AN · AS-rich particle had more ASpure than one containing less 3AN · AS (or 2AN · ASrich). The reference spectra showed sulfate signals at 975 VOL. 41, NO. 23, 2007 / ENVIRONMENTAL SCIENCE & TECHNOLOGY

9

8079

and 980 cm-1 for ASpure and ASdouble, respectively. In Figure 2, the freshly crystallized particles show an asymmetric Raman sulfate peak at ∼980 cm-1, with a shoulder on the high-wavenumber side. This indicates that both ASpure and ASdouble were present and that their signals are superimposed. The growth of the shoulder (shown from the deconvolution of the sulfate peak) on the higher wavenumber side from a 3AN · AS-rich particle to a 2AN · AS-rich particle is in agreement with the evidence from the nitrate signals. Table 1 summarizes the peak characteristics for 3AN · AS-rich and 2AN · AS-rich particles, which would be the indicators for following the transformation of the metastable component into the stable form. Figure 3a shows the I1053/I1046 ratios of all 15 freshly crystallized particles. Assuming the residual signal after subtraction from the 2AN · AS signal, that is, I1053 – 1.1 × I1046, arose from 3AN · AS, the relative amount of 3AN · AS, 2AN · AS, and pure AS is also estimated in Figure 3b. The results suggest that freshly crystallized particles contained both 2AN · AS and 3AN · AS in various amounts. Errors in the calculations of the ratios can be referred to the caption of Figure 3. Since the same procedures were applied in crystallizing the particles, differences in the composition of the freshly crystallized particles could not have been caused by experimental variations. We assume that the differences were the result of the random nature of nucleation events. The AIM predicts that 2AN · AS and pure AS are the thermodynamically stable crystals formed from equimolar AN/AS solutions. 3AN · AS should therefore be a metastable component. The results thus clearly show that freshly crystallized particles can have various degree of metastability. Differences in the crystals formed from supersaturated solutions among several crystallization events have been reported in some previous studies (8, 11). Therefore, the varying metastability observed in these experiments is plausible, and the results may represent typical relative amounts of 3AN · AS in particles crystallized from supersaturated equal molar AN/AS solutions. 3AN · AS to 2AN · AS Transformation and Its Dependence on RH. Although the freshly crystallized particles differed in their initial compositions, there was a consistent tendency for metastable 3AN · AS-rich particles to transform into thermodynamically more-stable 2AN · AS-rich particles. The rate of this transformation was found to be RH dependent. A typical series of spectra obtained at a fixed RH ) 55% as a function of time is shown in Figure 4. From crystallization at hour 0, the 3AN · AS-rich particle can be seen to transform into a 2AN · AS-rich one, as revealed by the changes in the three spectral features as listed in Table 1. The solid particles freshly formed from the crystallization of a droplet showed a strong peak at 1053 cm-1, which also represents a large initial I1053/I1046 ratio, and shoulders on both sides of the main peak at 717 cm-1, indicating that it was 3AN · AS-rich. As can be seen from Figure 4, the 1046 cm-1peak, a distinctive signal for 2AN · AS, gradually developed (i.e., the I1053/I1046 ratio decreased) with time. This indicates that the chemical transformation 4AS + 2(3AN · AS) f 3(2AN · AS) + 3AS was taking place. This conclusion is also supported by comparison of the Raman signal changes at ∼720 and 980 cm-1, although these signals are too weak for quantitative analysis. The degeneration of the shoulder on the higherwavenumber side of the nitrate peak at ∼720 cm-1 indicates the disappearance of 3AN · AS, while the shoulder on the lower-wavenumber side at the end of the transformation suggests that 2AN · AS was the product formed. The growth of the shoulder on the higher-wavenumber side of the sulfate signal at ∼980 cm-1 also suggests the proposed transformation, because more ASdouble would have been formed from ASpure during the transformation. Even though AN might also have contributed to the peak at 1046 cm-1, 3AN · AS would not be expected to transform 8080

9

ENVIRONMENTAL SCIENCE & TECHNOLOGY / VOL. 41, NO. 23, 2007

FIGURE 3. (a) I1053/I1046 ratios of 15 freshly crystallized particles. The dashed lines separate experiments with the same particle undergoing repeated crystallization processes. Each point represents the average of six spectra collected continuously. Errors for I1053/I1046 were estimated by different methods for values larger and smaller than 2.4. (1) Upper and lower bounds of errors for I1053/I1046 > 2.4 were estimated by I1053/I1047 and I1053/ I1045, respectively, representing the possible errors in incorrectly identifying the location of the shoulder contributed by 2AN · AS. (2) Two peaks can be clearly identified for I1053/I1046 > 2.4, and the errors were represented by the standard deviation of the I1053/I1046 values from six Raman spectra taken. Variations in I1053/I1046 < 2.4 are small, and thus the error bars are not clearly visible for most data in the figure. (b) Triangular plot of the composition of the freshly crystallized particles. Labeled points (1–6) correspond to I1053/I1046 ratios equal to 6.3, 3.8, 2.4, 1.9, 1.6, 1.42, and 1.39 shown in panel a. Because the overall molar ratio of AN/AS remained as 1:1, the plot of the molar ratio AS/ 2AN · AS/3AN · AS lies on the straight line with AS/2AN · AS/ 3AN · AS ) 0.5:0.5:0 and 0.67:0:0.33 as the end points. into AN in these experiments because the shoulder features at ∼720 cm-1 suggest the dominance of 2AN · AS, not AN. Besides, the transformation to AN would require the formation of more ASpure to maintain the overall material balance, that is, 4AS + 2(3AN · AS) f6AN + 6AS, and this is not consistent with the observed sulfate peak changes. The following discussions on the dependence of RH on chemical transformation will focus on the changes in the I1053/I1046 ratio, which were obvious and readily quantified for representing the relative extent of 3AN · ASf2AN · AS transformation. Other changes in spectral features followed the same trend as the I1053/I1046 ratio, as shown in Table 1.

FIGURE 4. Series of Raman spectra of a crystallized particle obtained at RH ) 55% showing the chemical transformation as a function of time. The peaks are normalized by the 420 cm-1 sulfate peak, which showed little spectral change during the transformation. Contributions of ASpure and ASdouble were estimated and are shown as gray lines.

FIGURE 5. Plot of I1053/I1046 ratio as a function of time for particles (a) 1 and (b) 2. Because there can be differences in the initial I1053/I1046 ratios among several crystallization processes, to facilitate the comparison of the data, the times with comparable ratios were artificially designated as time ) 0. The period between the actual initial I1053/I1046 and the recalibrated zero time was plotted in the negative time domain. The chemical transformation was monitored at RH ) 35, 55, and 65% for experiment 1 and RH ) 30, 45, 55, and 65% for experiment 2. Variations in I1053/I1046 in experiment 1 are small, and thus the error bars are not clearly visible for most data in the figure. Figure 5a and b show the I1053/I1046 ratio as a function of time at several RHs for experiments with two particles. Because the time scale of the transformation differed for the two particles, the results are presented separately. In each experiment, the same particle was subjected to several crystallization–deliquescence cycles to eliminate any particle

size effects on the transformation. The size of the crystallized particles in experiments 1 and 2 was estimated to be 20–60µm through microscopic analysis. The RH clearly affected the rate of the 3AN · ASf2AN · AS transformation. This is apparent by comparison of the decreases in the I1053/I1046 ratio with time at different RHs. Generally, a higher RH accelerated the transformation. It has been proposed that higher RH values promote the adsorption of water vapor onto the crystal surface, which enhances ion mobility (14). The observed RH dependency of the transformation rate supports this hypothesis. It should be noted that the initial rates were not compared because the difference in degree of metastability may also affect the transformation rate. Hence, we focus on the effects of RH on the overall transformation using the same “initial state” of a fixed I1053/I1046 ratio, instead of the initial rate. The transformation ended when I1053/I1046 stabilized at about 1.2 and 1.1 for experiments 1 and 2, respectively. The stabilized particles contained mainly 2AN · AS (and pure AS), and possibly trace amounts of 3AN · AS. In experiment 1, the I1053/I1046 ratio increased from 3.7 at time ) 0 to 4.1 at 18 h at RH ) 35%. This was not regarded as evidence of an increase in the amount of metastable 3NA · AS present because the increase is within the error. For experiment 2, the I1053/I1046 ratio stabilized at ∼1.2 for RH ) 45/55%, which is different from that observed at RH ) 65%. The possibility that the transformation was still on going cannot be disregarded, and the rate may have been so slow that it could not be observed within the time span of the experiments. Schlenker et al. (14) reported the formation of 3AN · AS and pure AS for freshly crystallized AN/AS ) 1:1 particles, but they did not observe any chemical changes when the particles crystallized at RH ) 1% nor when the particles crystallized at RH ) 1% were passed through an RHcontrolled cell at 52% for 3.5 min Although the particles studied in Schlenker et al.’s experiments (14) were around 300 nm in diameter (compared to 20–60 µm in these experiments), their residence time of a few minutes within the RH-controlled cells, as compared to hours in this case, might not have been enough to allow the detection of any slow chemical changes. However, these results match the observations of this study that the freshly crystallized particles are 3AN · AS dominated, instead of 2AN · AS as predicted by the AIM. While Schenkler et al. (14) did not detect chemical changes in particles crystallized from equal molar AN/AS solutions, they did observe the transformation of 3AN · AS into 2AN · AS in particles crystallized from AN/AS ) 2:3 solutions. They also observed other transformations in particles containing H+ when the crystallized particles were exposed to a higher RH (