Photochemistry of aqueous pyruvic acidElizabeth C. Griffitha,b, Barry K. Carpenterc, Richard K. Shoemakera, and Veronica Vaidaa,b,1

aDepartment of Chemistry and Biochemistry and bCooperative Institute for Research in Environmental Sciences, University of Colorado Boulder, Boulder, CO80309; and cPhysical Organic Chemistry Centre, Cardiff University, Cardiff CF10 3AT, United Kingdom

Edited* by Josef Michl, University of Colorado Boulder, Boulder, CO, and approved May 23, 2013 (received for review February 18, 2013)

The study of organic chemistry in atmospheric aerosols and cloudformation is of interest in predictions of air quality and climatechange. It is now known that aqueous phase chemistry is importantin the formation of secondary organic aerosols. Here, the photo-reactivity of pyruvic acid (PA; CH3COCOOH) is investigated in aque-ous environments characteristic of atmospheric aerosols. PA iscurrently used as a proxy for α-dicarbonyls in atmospheric modelsand is abundant in both the gas phase and the aqueous phase(atmospheric aerosols, fog, and clouds) in the atmosphere. The pho-toreactivity of PA in these phases, however, is very different, thusprompting the need for a mechanistic understanding of its reactivityin different environments. Although the decarboxylation of aqueousphase PA through UV excitation has been studied for many years, itsmechanism and products remain controversial. In this work, photol-ysis of aqueous PA is shown to produce acetoin (CH3CHOHCOCH3),lactic acid (CH3CHOHCOOH), acetic acid (CH3COOH), and oligomers,illustrating the progression from a three-carbon molecule to four-carbon and even six-carbon molecules through direct photolysis.These products are detected using vibrational and electronic spec-troscopy, NMR, and MS, and a reaction mechanism is presentedaccounting for all products detected. The relevance of sunlight-ini-tiated PA chemistry in aqueous environments is then discussed inthe context of processes occurring on atmospheric aerosols.

atmospheric chemistry | sunlight-initiated chemistry

Predictive models of air quality and climate change require un-derstanding of the role that organic compounds play in aerosol

interactions with radiation (1, 2). Volatile organic compounds inthe atmosphere oxidize with the products of this chemistry, con-tributing to the formation of secondary organic aerosols (SOAs) (3,4). Current model predictions, however, do not agree with findingsof field measurements and highlight the importance of under-standing organic reaction mechanisms, which are expected to affectthe oxidizing capacity of the atmosphere, the formation of SOAs,and climate-relevant aerosol properties (1–10). Organic polymershave been identified as important contributors to SOAs. Thesepolymeric species do not result from gas phase reactions (11) butrather appear to be products of organic reactions in condensed(aqueous) phase, where high-molecular-weight compounds, poly-mers, and oligomers can be generated (1, 5, 12–15). Recent resultshave mainly addressed radical reactions in aqueous environments(5, 7, 16), with only a few using direct photolysis (13, 17). The needfor understanding organic reaction mechanisms in aqueous sol-utions representative of aqueous media in the atmosphere (aero-sols, cloud droplets, fog, and rain) has been recognized (1, 2, 7, 18).In aqueous environments with sufficiently high organic con-

centrations, polyfunctional carbonyl compounds react to formhigh-molecular-weight species, polymers, and oligomers, yet thereaction mechanisms responsible for these aqueous phase reac-tions remain elusive targets for investigation (5, 6, 13, 19–21).This investigation provides fundamental laboratory results andmechanisms for the photochemistry of pyruvic acid (PA) in anaqueous solution. PA, an oxidative product of isoprene, is abun-dant in the atmosphere in gas and aerosol phases, and it has beenused in models as a proxy for atmospheric α-dicarbonyls (3, 22–24). PA undergoes decarboxylation in the gas and aqueous phasesby means of different mechanisms. Decarboxylation of gas phasePA has been seen to occur thermochemically, through IR multi-

photon pyrolysis, as well as through UV and visible excitation (13,25–36). The main removal pathway of gas phase PA from thetroposphere is UV photolysis, with reaction with OH radicalproviding a minor contribution (31). The photochemistry (directphotolysis) of PA in an aqueous solution remains controversial inthe literature, and it is the focus of the present laboratory study.Photodecarboxylation of aqueous PA is facilitated through ex-

citation of its n→π* band, with an absorption maximum of 321 nm(37). This decarboxylation has been known for many years, but themechanism and resulting products have been controversial (13,37–39). Earlier studies performed by Leermakers and Vesley (37,38) report 3-hydroxy-2-butanone (acetoin) as the only isolableproduct from the aqueous photolysis of PA. The photochemicalproducts were further found to be medium-dependent, with thephotochemistry progressing through two different pathways: de-carboxylation and reduction. In water, photochemistry of PAprogressed through a photodecarboxylation pathway, resulting insignificant CO2 production, as well as acetoin as the major aque-ous phase product detected. In organic solvents, such as methanol,chloroform, and ethyl ether, Leermakers and Vesley (37) did notdetect CO2 with photolysis, and thus suggest that the photo-chemistry is instead progressing through a reduction pathway, withthe dimer dimethyltartaric acid found to be the major product inall three of these solvents. Closs and Miller (39) then contend thatthe production of acetoin in aqueous solution is facilitated by thetransition through α-acetolactic acid, a β-keto acid that is knownto decarboxylate thermally to acetoin (40). Under similar experi-mental conditions many years later, Guzman et al. (13) reportedno detection of acetoin; instead, they saw two oligomers of PA(dimers) as the major products of this photochemistry, one ofwhich (“Product A”) is the same oligomer (dimethyltartaric acid)detected by Leermakers and Vesley (37) in other organic solvents.In the current work, the photochemistry of PA in aqueous solu-tion is reinvestigated to provide a mechanism for this reactionconsistent with the observations reported here and supported bythe literature.

ResultsBoth aqueous and gas phase products were detected using FTIRand UV spectroscopy, NMR, and MS. The photoinduced decar-boxylation of aqueous PAwas confirmed through IR spectra takenof the gas produced after 2 h of photolysis, as seen in Fig. 1. Theintense features between 2,200 and 2,400 cm−1 are due to the CO2produced. A closer look at the region between 1,000 and 2000 cm−1

reveals features due to an additional compound being released intothe gas phase concurrent with decarboxylation (Fig. 1, Inset),buried beneath the residual water vapor lines. This second gasphase component is identified as acetoin through comparison withthe gas phase spectrum of pure acetoin (shown in red in Fig. 1,Inset), which was obtained in this laboratory. [Acetoin was initially

Author contributions: E.C.G. and V.V. designed research; E.C.G. and B.K.C. performedresearch; R.K.S. contributed new reagents/analytic tools; E.C.G. and R.K.S. analyzed data;and E.C.G., B.K.C., and V.V. wrote the paper.

The authors declare no conflict of interest.

*This Direct Submission article had a prearranged editor.1To whom correspondence should be addressed. E-mail: vaida@colorado.edu.

This article contains supporting information online at www.pnas.org/lookup/suppl/doi:10.1073/pnas.1303206110/-/DCSupplemental.

11714–11719 | PNAS | July 16, 2013 | vol. 110 | no. 29 www.pnas.org/cgi/doi/10.1073/pnas.1303206110

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 11

, 202

0

identified by its distinctive odor by one of the authors (B.K.C.).] It isimportant to note that it is difficult to detect acetoin both in the gasphase using FTIR and in an aqueous solution using NMR due tothe interference of water.The NMR spectra (Fig. 2 A and B) clearly demonstrate the

presence of both acetoin and lactic acid in solution after photolysis.The 1H and 13C NMR of aqueous PA after photolysis yieldsa complex mixture of products, as shown in Fig. 2B; however, two-dimensional NMR not only provides the chemical shift informationof the 1H and 13CNMR, but also the key through-bond correlationspresent between coupled protons [two-dimensional gradient-se-lected homonuclear correlation spectroscopy (gCOSY) in Fig. 2A]and between protons and carbons via multiple-bond correlations[gradient-selected heteronuclear multiple bond correlation usingadiabatic pulses (gHMBCad) in Fig. 2B] for each compound. Byacquiring the same 2DNMRdata using samples of aqueous acetoin(shown in green in Fig. 2) and lactic acid (shown in blue in Fig. 2),the direct correspondence of the 1H and 13C chemical shifts, as wellas the identical through-bond correlations between those reso-nances with signals present in the photolyzed PAmixture, confirmsthe presence of both of these compounds in the photolyzed prod-uct. The two-dimensional gradient-selected heteronuclear single-bond correlation spectrum using matched adiabatic pulses (2D-gHSQCad), with acetoin resonances assigned, is available in Fig.S1. Acetic acid was also identified in the mixture via the chemicalshifts of the methyl group (1H and 13C: 2.07 and 20.6 ppm, re-spectively) and the carboxyl carbon shift of 176.8 ppm via directgHMBC correlation to themethyl resonance at 2.07 ppm (circled inFig. 2B). These shifts correspond to values from known libraryspectra (NMR prediction software; ACD Labs), and the full sets ofdata used to identify acetic acid are presented in Fig. S2.Oligomers of PA, observed previously in the work performed

by Guzman et al. (13), were also detected here using both elec-trospray ionization MS (Fig. S3) and diffusion ordered spectros-copy (DOSY) NMR (Fig. 3). [The authors recognize the formalmisuse of the term “oligomer” but are using this term in con-gruence with the published literature (13). The oligomer productshere are formed in a termination step, through reaction betweentwo radicals.] After 2 h of photolysis, the mass spectrum of thephotolyzed mixture shows a major product peak at 177 m/z,consistent with Guzman et al.’s Product A (dimethyltartaric acid),as well as a minor peak at 175m/z, the mass of which is consistentwith their oligomer “Product B” (13). The DOSY data confirmthe presence of oligomers, primarily consisting of dimers of var-ious forms, which represent a majority of the components of thephotolyzed mixture. Using the sample containing the mixture ofphotolysis products, the calibrated diffusion coefficient (D) for

most of the NMR signals between 1 and 3 ppm is measured to bebetween 6.5 and 6.9 (×10−10 M2/s). The well-resolved signalsassigned to acetoin (1H quartet, 4.41 ppm) show D = 10.2 × 10−10

M2/s, lactic acid (1H quartet, 4.36 ppm) gives D = 9.7 × 10−10 M2/s,and acetic acid (1H singlet, 2.07 ppm) gives D = 11.2 × 10−10 M2/s(DOSY data, Fig. 3). For verification of the assignments of acetoinand lactic acid, and to confirm the diffusion measurements, iden-tical DOSY experiments were performed using separate aqueoussamples of 0.04 M acetoin and 0.04 M lactic acid. These experi-ments yieldedD= 10.0–10.2× 10−10M2/s for acetoin and 9.3–9.6×10−10 M2/s for lactic acid, closely matching the values for eachcompound measured in situ in the photolysis mixture (Fig. S4).Comparing the D values for acetoin and lactic acid with the Dvalues for themajority photolysis products with lower D values, theratio is found to be in the range of 1.4:1–1.6:1 (Dacetoin/lactic acid/Dmajority product ratio). Doubling the size/mass of the molecule (i.e.,either a homodimer or heterodimer of the small molecules) shouldlower the measured D value by approximately √2 depending onthe molecular geometry, suggesting that the majority of the pho-tolysis products are approximately twofold the size of the in-dividual molecules already identified (i.e., acetoin, lactic acid) (41,42). These D values indicate that the major products of photolysisof PA are small oligomers. Although no structural informationwas obtained directly in this work, the size and masses of theseoligomers are consistent with the two oligomer products firstidentified by Guzman et al. (13).Finally, Fig. 4 presents a UV-visible spectrum of aqueous PA

after 1 h of photolysis (black solid line), along with the under-lying components of the spectrum shown in broken lines. After 1h of photolysis, there is still some amount of unphotolyzed PAremaining in the spectrum (scaled experimental spectrum of PAshown as a red broken line in Fig. 4). Acetoin is also seen as anaqueous phase product (scaled experimental spectrum shown asa blue broken line in Fig. 4). The addition of these two com-ponents, however, does not account for the entire absorptionbetween 250 and 400 nm shown in Fig. 4. The residual absorp-tion is shown in green (broken line in Fig. 4). This absorptioncannot be due to the oligomer dimethyltartaric acid, because it islacking in a UV chromophore in the wavelength range of in-terest. Rather, this residual absorption is consistent with anotherdimer product containing a UV chromophore. A UV-visible ab-sorption spectrum was also taken after 2 h of photolysis and ispresented in Fig. S5. After 2 h of photolysis, no new absorbers areapparent in the spectrum, but a further decrease in PA contri-bution is observed, as well as an increase in the contribution dueto acetoin and the oligomer product, consistent with the pro-gression of the same reaction scheme.

Fig. 1. Infrared spectrum of collected gas over0.1 M aqueous PA after photolysis for 2 h. (Inset)Acetoin features (broad peaks beneath the sharpresidual water lines) between 1,000 and 2,000 cm−1

(black) with the pure acetoin gas phase spectrum (red)overlaid for comparison. AU, absorbance units.

Griffith et al. PNAS | July 16, 2013 | vol. 110 | no. 29 | 11715

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 11

, 202

0

DiscussionAldehydes and, to a lesser degree, ketones are known to hydratein aqueous solution (43–50). It is known that PA is reversiblyhydrated in aqueous solution to its gem-diol and that the equi-librium is both pH- and temperature-dependent, with lower pHand lower temperature both favoring the hydrate (51, 52). Inaqueous solution at 298 K, ∼35% of PA exists in its keto form,with the majority existing as its gem-diol (13, 45, 46). The ketoform contains a UV chromophore, which can be excited in thenear-UV state to induce photolysis. UV irradiation of aqueousPA is known to result in phosphorescence from a 3(n,π*) state(53). This triplet is presumed to be involved in the aqueous phasephotochemistry of PA. Ab initio calculations (36) have suggestedthat the S1 state of PA is 1(n,π*) and that this is higher in energythan both the T1 and T2 states, which are found to be 3(n,π*) and3(π,π*), respectively. Assuming these calculations to be at leastqualitatively correct, it seems likely that the UV photochemistry

of PA with light of λ > 300 nm proceeds according to the se-quence S0 → S1 → T2 → T1. This intersystem crossing to thetriplet surface is markedly different from the gas phase photo-chemistry, which occurs solely on the singlet surface to generatemethylhydroxycarbene (28–30, 54). However, in the aqueousstudy presented here, PA is seen to react from its T1 state toproduce acetoin, lactic and acetic acids, and oligomers, as de-scribed below and depicted in Scheme 1.CBS-QB3 calculations suggest that the T1 state of PA can

abstract a hydrogen atom from the carboxyl group of PA with aclassical activation enthalpy of 2.7 kcal/mol. The reaction is foundto be endothermic by 0.9 kcal/mol at the same level of theory.The corresponding calculation cannot be carried out for hydro-gen atom transfer from the carboxyl of the PA hydrate becausethe resulting carboxyl radical is found to be unbound: It losesCO2 with zero activation barrier. In accord with this finding,the computed transition structure for H-atom abstraction by

A

B

Fig. 2. Superimposed spectra of aqueous 500-MHz NMR spectra of PA afterphotolysis (red), 0.1 M acetoin (green), and 0.1 M lactic acid (blue). 2D-gCOSYwith WET suppression (A) and 2D-gradient gHMBCad (B). The circled peak in B isassigned as acetic acid (further evidence for acetic acid is provided in Fig. S2).

A

B

Fig. 3. DOSY spectrum of photolysis products of PA. Projection of the dif-fusion dimension (A) and full 2D DOSY plot showing all components (B).

11716 | www.pnas.org/cgi/doi/10.1073/pnas.1303206110 Griffith et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 11

, 202

0

PA T1 from PA hydrate has one very long C–C bond (Fig. S7).When an intrinsic reaction-coordinate calculation is carried outat the B3LYP/CBSB7 level (which is used for geometry optimi-zation in the CBS-QB3 composite method), it reveals that thereaction is a concerted hydrogen atom transfer and decarbox-ylation. Hence, the products would be a triplet radical pair withsubstantial separation between the spins, as shown in reaction 1of Scheme 1. The overall reaction with the gem-diol has a slightlyhigher classical barrier (4.3 kcal/mol) than the PA (T1) + PA (S0)reaction (used in ref. 13); however, unlike that reaction, it isexothermic by 22.1 kcal/mol, according to the CBS-QB3 calcu-lations. Thus, although it is conceivable that the photoexcited T1

PA molecule could abstract a hydrogen atom from either aground-state (S0) PA molecule in its keto or gem-diol form, thereaction with the gem-diol form is used here due to its excess insolution as well as its favorable exothermicity and concertedH-abstraction and decarboxylation.

The radical pair shown in reaction 1 of Scheme 1 will be gen-erated in a solvent cage, but geminate reactions between the rad-icals will presumably not occur until the pair has undergoneintersystem crossing to a singlet-coupled state. In competition withthis, escape of the radicals from the cage is likely. Hence, one mustconsider several possible radical–radical reactions. Two of thesereactions, resulting in products detected in this work, are shown inreactions 2 and 3 of Scheme 1. The product of reaction 2 is a hy-drate of acetolactic acid. Its dehydration should be thermody-namically favorable because, unlike PA hydrate, the product is notan α-keto acid. Decarboxylation of acetolactic acid, known to occurwithin minutes at 25 °C in aqueous solution (39, 40), would yieldacetoin, a principal product of aqueous PA photolysis. This se-quence is depicted in reactions 4a–4c of Scheme 1. Finally, reaction3 of Scheme 1 is a disproportionation that generates lactic andacetic acids, both of which are observed products of aqueous PAphotolysis here (NMR results, Fig. 2). The alternative dispropor-tionation (SI Additional Radical–Radical Reactions) would generateground-state PA and the hydrate of acetaldehyde, which is alsoa known product of PA photochemistry (37) but was not detectedhere. In addition, there could be other recombination and dis-proportionation reactions occurring (SI Additional Radical–RadicalReactions). These may be kinetically important, but they are notsignificant in terms of product identification because they generateproducts already considered in Scheme 1 or products that are notobserved here, and hence they are not depicted explicitly.In addition to acetoin and lactic and acetic acids, small oligomers

(dimers) were observed in this work, in the previous work byGuzman et al. (13) in aqueous solution, and by Leermakers andVesley (37) in organic solvents. Reaction 5 of Scheme 1 showsthe generation of 2,3-dimethyltartaric acid, presumably in bothmeso- and DL diastereomers, from the reaction between twomolecules of the prevalent radical intermediate (PA•). Thissame oligomer has been identified in earlier studies (13, 37), andit is detected in the present work as a significant product ofaqueous PA photochemistry using MS (Fig. S3). This dimerproduct is also consistent with the approximate size of oligomersdetected using DOSY NMR (Fig. 3). Although at least one otherdimer product is detected in this work, as described in Results,

Fig. 4. UV-visible absorption spectrum of aqueous PA after 1 h of photolysis(black), with scaled underlying components shown in broken lines. PA isshown in red, acetoin is shown in blue, and remaining absorption due tooligomer product is shown in green.

Scheme 1. Summary of the proposed mechanism for aqueous PA photochemistry.

Griffith et al. PNAS | July 16, 2013 | vol. 110 | no. 29 | 11717

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 11

, 202

0

because no structural information is available, it is not includedin the mechanism presented here.One key feature of ourmechanism is the concerted hydrogen atom

transfer and decarboxylation shown in reaction 1 of Scheme 1.According to our CBS-QB3 calculations, the radical H3C–C(OH)2–CO2• resulting from H-abstraction from the gem-diol of PA isunbound; it decarboxylates with zero barrier. This results in theconcerted H-abstraction and decarboxylation proposed here. Bycontrast, the radical H3C–CO–CO2•, resulting from H-abstractionfromaPAmolecule in its keto form, is found to be bound at the samelevel of theory. This emphasizes the need for water in the solutionphase photodecarboxylation chemistry presented here (reac-tions 1–4 in Scheme 1). The concerted H-atom transfer anddecarboxylation of reaction 1 of Scheme 1 require the presenceof the gem-diol, which is formed only in aqueous environments.There are a few key points to emphasize concerning this aqueous

phase mechanism. First, it is important to note that CO2 is producedin aqueous solution concurrent with the formation of acetoin andlactic and acetic acids but is not required in the production of theoligomer dimethyltartaric acid. In our mechanism, the formation ofdimethyltartaric acid (reaction 5 of Scheme 1) does not require wateror necessarily the production of CO2, merely a good H-atom donorto produce the radical intermediate PA•. Leermakers and Vesley(37) originally proposed this to explain their observation of photo-reduction in organic solvents (in contrast to photodecarboxylation),proposing H-abstraction facilitated by reaction of the T1 state of PAwith a solvent molecule, resulting in PA• formation followed bya self-reaction to form dimethyltartaric acid. Our results are con-sistent with this assertion. It is also interesting to note that gas phasephotolysis does not require the presence of water for decarboxylation(37), whereas solution phase decarboxylation chemistry does. This islikely due to the fact that gas phase photochemistry of PA occurspredominantly on the singlet surface, with PA directly decarbox-ylating in a unimolecular reaction to form a methylhydroxycarbene,followed by H-atom transfer to form acetaldehyde and vinyl alcohol(28, 54). In water, however, the chemistry occurs on the triplet sur-face, proceeding bimolecularly through reaction of one molecule ofPA in the T1 state with one molecule of PA in its gem-diol form.Water appears to be playing a role in energy flow from the singlet-excited state to the triplet surface, in contrast to the favored singletchemistry in the gas phase. Once the photoexcited PAmolecule is onthe triplet surface, water again plays a role in allowing a bimolecularreaction specifically with the gem-diol form of PA (the predominantform in aqueous environments).This photochemistry of PA in aqueous solution is also known

to be pH-sensitive, with the protonated form of PA decarbox-ylating with a yield nearly 20-fold that of the anionic form (py-ruvate) at higher pH values (37), an observation that is consistentwith the mechanism presented here. The presence of the acidhydrogen atom is essential to the decarboxylation chemistry,confirming the first abstraction: the decarboxylation reaction ofScheme 1. It is interesting to note that this pH effect is exactlythe opposite of what has been found to occur for reaction withOH (16), a competing removal pathway for PA in the tropo-sphere. This indicates that in aqueous environments, the domi-nance of either direct photolysis or OH reaction in the removal ofPA may itself be pH-dependent.Finally, it has been observed previously that UV irradiation (λ >

300 nm) of aqueous PA glass at 77 K results in the formation oftriplet-coupled radical pairs with a separation of ≥5 Å between thespins, which is accompanied by ultrafast decarboxylation (55). Theexplanation offered for the last observation is long-range electrontransfer between the S0 and T1 states of PA, resulting in a triplet-coupled radical ion pair. The radical cation component of this pairis proposed to be responsible for the decarboxylation (13, 55). Themechanism presented in this work provides an alternative expla-nation for this cryogenic aqueous glass photochemistry. During theconcerted H-abstraction and decarboxylation of reaction 1 inScheme 1 (transition state structure depicted in Fig. S7), the

carbons that carry the highest spin density in the products arealready separated by 5.5 Å in the transition structure. It seemsthat this mechanism could provide an alternative explanationof the EPR spectrum, resulting in similar spin separations, andthe ultrafast decarboxylation seen in the aqueous glass photo-chemistry of PA, in lieu of the radical ion pairs previouslyproposed (55).

Atmospheric ImplicationsThe contrast between the chemistry performed by UV photolysis ofgaseous and aqueous phase PA highlights the importance of athorough understanding of chemistry as a function of themolecule’senvironment for inclusion in atmospheric models. This work showsthat photochemical processes in the aqueous phase of aerosols re-sult in different products than in the analogous gas phase chemistry,thus affecting aerosol composition and optical properties differ-ently. The bulk ocean is slightly basic, causing the progression of thisreaction scheme near the ocean surface (where UV light still pen-etrates) to be unlikely due to the presence of PA predominantly inits anionic form. Atmospheric aerosols, however, are known toexperience a wide range of pH values throughout their atmosphericlifetime, and can reach acidic pH values at which this chemistry willbecome dominant (56). These aerosol particles have the addedadvantage of concentrating the reactant species through selectiveevaporation of water, resulting in high concentrations of reactantmonomers (57–61). Therefore, although this chemistry is unlikely tobe an important contributor to bulk ocean chemistry, it will beimportant in acidic aqueous atmospheric aerosols. In addition, theproduction of acetoin, which partitions both to the aqueous phaseand the gas phase, may contribute to the seeding of new particles.Finally, the direct photolysis of aqueous PA presented here, as wellas the known competing OH reaction pathway, results in the for-mation of oligomers, a unique product to the aqueous phasechemistry (compared with the gas phase chemistry), which havebeen implicated in SOA formation (1, 5, 12, 14).

Materials and MethodsPA (98%) was purchased from Sigma–Aldrich and was distilled under re-duced pressure before its preparation to a final concentration of 0.1 M indistilled water, a typical concentration for atmospheric aerosol particles(5, 6). Acetoin (≥92%, natural) was also purchased from Sigma–Aldrich,and it was prepared to a final concentration of 0.04 M in distilled waterwithout further purification. Similarly, lactic acid (racemic, meets USPharmacopeia testing specifications, purchased from Sigma–Aldrich) wasprepared to a final concentration of 0.04 M in distilled water withoutfurther purification.

Aqueous PA was photolyzed using a 450-W Xe arclamp (Newport) for1–2 h, while kept chilled at 4 °C in a temperature-controlled water bath. AXe lamp was used due to the similarity of its output to the solar spectrum.The photolyzed PA solution was allowed to return to room temperaturebefore any subsequent analysis was performed.

UV-Visible Spectroscopy. UV-visible spectra were obtained of aqueous samplesusing a USB2000miniaturefiberoptic UV-visible spectrometer fromOceanOptics.

IR Spectroscopy. The gas that evolved during photolysis of aqueous PA wascollected over the solution during 2 h of photolysis. Then, after the so-lution was allowed to return to room temperature, a gas phase IR spectrumwas obtained of the evolved gaseous products using the internal com-partment of a Bruker Tensor 27 FTIR spectrometer, purged with dry houseair. Spectra were collected at a resolution of 1 cm−1 and were averagedover 50 scans.

NMR. NMR spectra were acquired using a Varian INOVA 500-MHz NMR spec-trometer operating at 499.60 MHz for 1H observation. All samples were pre-pared with a volume of 0.8 mL in 5-mm NMR tubes. A total of 0.050 mL of99.9% (mole % deuteration of the deuterated solvent) D2O was added toprovide a field-frequency lock, making the final solvent 95% (vol/vol) H2Oand 5% (vol/vol) D2O. To characterize the key components of the photolysisproducts fully, a combination of standard high-resolution NMR experiments,such as 1D 1H NMR, 2D gCOSY, gHMBCad, gHSQCad, and high-resolution

11718 | www.pnas.org/cgi/doi/10.1073/pnas.1303206110 Griffith et al.

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 11

, 202

0

2D-DOSY were performed using WET suppression of the H2O resonance(details of the WET method are described in SI Materials and Methods). Ad-ditional NMR method information is included in SI Materials and Methods,along with supporting NMR data (Figs. S6 and S8).

Computation. Electronic structure calculations used the CBS-QB3 compositemethod (62), as implemented in the Gaussian 03 suite of programs (63).

Cartesian coordinates and energies of stationary points are included inDataset S1.

ACKNOWLEDGMENTS. E.C.G. and V.V. acknowledge funding for this workfrom the National Science Foundation (Grant CHE-1011770). E.C.G. alsoacknowledges a National Aeronautics and Space Administration Earth andSpace Science Graduate Fellowship and a Marian Sharrah Graduate Fellow-ship from the University of Colorado, Boulder.

1. Ervens B, Turpin BJ, Weber RJ (2011) Secondary aerosol formation in cloud dropletsand aqueous particles (aqSOA): A review of laboratory, field and model studies. At-mos Chem Phys 11(21):11069–11102.

2. Perraud V, et al. (2012) Nonequilibrium atmospheric secondary organic aerosol for-mation and growth. Proc Natl Acad Sci USA 109(8):2836–2841.

3. Ervens B, et al. (2008) Secondary organic aerosol yields from cloud-processing ofisoprene oxidation products. Geophys Res Lett 35(2):L02816.

4. Kanakidou M, et al. (2005) Organic aerosol and global climate modelling: A review.Atmos Chem Phys 5(4):1053–1123.

5. Tan Y, Lim YB, Altieri KE, Seltzinger SP, Turpin BJ (2012) Mechanisms leading tooligomers and SOA through aqueous photooxidation: Insights from OH radical oxi-dation of acetic acid and methylglyoxal. Atmos Chem Phys 12(2):801–813.

6. Lim YB, Tan Y, Perri MJ, Seitzinger SP, Turpin BJ (2010) Aqueous chemistry and its rolein secondary organic aerosol (SOA) formation. Atmos Chem Phys 10(21):10521–10539.

7. Monod A, Carlier P (1999) Impact of clouds on the tropospheric ozone budget: Directeffect of multiphase photochemistry of soluble organic compounds. Atmos Environ33(27):4431–4446.

8. Ervens B, et al. (2003) CAPRAM 2.4 (MODACmechanism): An extended and condensedtropospheric aqueous phase mechanism and its application. J Geophys Res D Atmos108(D14).

9. Blando JD, Turpin BJ (2000) Secondary organic aerosol formation in cloud and fogdroplets: A literature evaluation of plausibility. Atmos Environ 34(10):1623–1632.

10. Monod A, Poulain L, Grubert S, Voisin D, Wortham H (2005) Kinetics of OH-initiatedoxidation of oxygenated organic compounds in the aqueous phase: New rate con-stants, structure-activity relationships and atmospheric implications. Atmos Environ39(40):7667–7688.

11. Fuzzi S, et al. (2006) Critical assessment of the current state of scientific knowledge,terminology, and research needs concerning the role of organic aerosols in the at-mosphere, climate, and global change. Atmos Chem Phys 6:2017–2038.

12. Wonaschuetz A, et al. (2012) Aerosol and gas re-distribution by shallow cumulus clouds:An investigation using airborne measurements. J Geophys Res D Atmos 117(D17):D17202.

13. Guzman MI, Colussi AJ, Hoffmann MR (2006) Photoinduced oligomerization ofaqueous pyruvic acid. J Phys Chem A 110(10):3619–3626.

14. Carlton AG, Wiedinmyer C, Kroll JH (2009) A review of secondary organic aerosol(SOA) formation from isoprene. Atmos Chem Phys 9(14):4987–5005.

15. De Haan DO, et al. (2009) Secondary organic aerosol formation by self-reactions ofmethylglyoxal and glyoxal in evaporating droplets. Environ Sci Technol 43(21):8184–8190.

16. Ervens B, Gligorovski S, Herrmann H (2003) Temperature-dependent rate constantsfor hydroxyl radical reactions with organic compounds in aqueous solutions. PhysChem Chem Phys 5(9):1811–1824.

17. Donaldson DJ, George C, Vaida V (2010) Red sky at night: Long-wavelength photo-chemistry in the atmosphere. Environ Sci Technol 44(14):5321–5326.

18. Buszek RJ, Francisco JS, Anglada JM (2011) Water effects on atmospheric reactions. IntRev Phys Chem 30(3):335–369.

19. Monod A, Doussin JF (2008) Structure-activity relationship for the estimation ofOH-oxidation rate constants of aliphatic organic compounds in the aqueous phase:Alkanes, alcohols, organic acids and bases. Atmos Environ 42(33):7611–7622.

20. El Haddad I, et al. (2009) In-cloud processes of methacrolein under simulated conditions—Part 2: Formation of secondary organic aerosol. Atmos Chem Phys 9(14):5107–5117.

21. Liu Y, et al. (2012) Aqueous phase processing of secondary organic aerosol fromisoprene photooxidation. Atmos Chem Phys 12(13):5879–5895.

22. Carlton AG, Turpin BJ, Lim H-J, Altieri KE, Seitzinger S (2006) Link between isopreneand secondary organic aerosol (SOA): Pyruvic acid oxidation yields low volatility or-ganic acids in clouds. Geophys Res Lett 33:L06822.

23. Kawamura K, Kasukabe H, Barrie LA (2010) Secondary formation of water-solubleorganic acids and alpha-dicarbonyls and their contributions to total carbon and wa-ter-soluble organic carbon: Photochemical aging of organic aerosols in the Arcticspring. J Geophys Res D Atmos 117(D27):D21306.

24. Nguyen TB, et al. (2010) High-resolution mass spectrometry analysis of secondaryorganic aerosol generated by ozonolysis of isoprene. Atmos Environ 44(8):1032–1042.

25. Plath KL, Takahashi K, Skodje RT, Vaida V (2009) Fundamental and overtone vibra-tional spectra of gas-phase pyruvic acid. J Phys Chem A 113(26):7294–7303.

26. Vesley GF, Leermakers PA (1964) Photochemistry of alpha-keto acids and alpha-ketoesters. 3. Photolysis of pyruvic acid in vapor phase. J Phys Chem 68(8):2364–2366.

27. Larsen MC, Vaida V (2012) Near infrared photochemistry of pyruvic acid in aqueoussolution. J Phys Chem A 116(24):5840–5846.

28. Takahashi K, Plath KL, Skodje RT, Vaida V (2008) Dynamics of vibrational overtoneexcited pyruvic acid in the gas phase: Line broadening through hydrogen-atomchattering. J Phys Chem A 112(32):7321–7331.

29. Yamamoto S, Back RA (1985) The photolysis and thermal decomposition of pyruvicacid in the gas phase. Can J Chem 63(2):549–554.

30. Saito K, Sasaki G, Okada K, Tanaka S (1994) Unimolecular decomposition of pyruvicacid—And experimental and theoretical study. J Phys Chem 98(14):3756–3761.

31. Mellouki A, Mu YJ (2003) On the atmospheric degradation of pyruvic acid in the gasphase. J Photochem Photobiol A Chem 157(2-3):295–300.

32. Taylor R (1987) The mechanism of thermal eliminations. 23.1 The thermal de-composition of pyruvic acid. Int J Chem Kinet 19(8):709–713.

33. Rosenfeld RN, Weiner B (1983) Energy disposal in the photofragmentation of pyruvicacid in the gas-phase. J Am Chem Soc 105(11):3485–3488.

34. Horowitz A, Meller R, Moortgat GK (2001) The UV-VIS absorption cross sections of thealpha-dicarbonyl compounds: Pyruvic acid, biacetyl and glyoxal. J Photochem Pho-tobiol A Chem 146(1-2):19–27.

35. Murto J, Raaska T, Kunttu H, Rasanen M (1989) Conformers and vibrational spectra ofpyruvic acid—An ab initio study. Theochem 59:93–101.

36. Dhanya S, et al. (2003) Dynamics of OH formation in photodissociation of pyruvic acidat 193 nm. J Chem Phys 118(22):10093–10100.

37. Leermakers PA, Vesley GF (1963) Photochemistry of alpha-keto acids and alpha-ketoesters. 1. Photolysis of pyruvic acid and benzoylformic acid. J Am Chem Soc 85(23):3776–3779.

38. Leermakers PA, Vesley GF (1963) Photolysis of pyruvic acid in solution. J Org Chem 28(4):1160–1161.

39. Closs GL, Miller RJ (1978) Photo-reduction and photodecarbosylation of pyruvic acid—Applications of CIDNP to mechanistic photochemistry. J Am Chem Soc 100(11):3483–3494.

40. Ronkainen P, Brummer S, Suomalai H (1970) Diacetyl and formic acid as decompositionproducts of 2-acetolactic acid. Acta Chem Scand 24(9):3404–3406.

41. Crutchfield CA, Harris DJ (2007) Molecular mass estimation by PFG NMR spectroscopy.J Magn Reson 185(1):179–182.

42. Valencia DP, Gonzalez FJ (2011) Understanding the linear correlation between dif-fusion coefficient and molecular weight. A model to estimate diffusion coefficients inacetonitrile solutions. Electrochem Commun 13(2):129–132.

43. Wolfe S, Kim CK, Yang K, Weinberg N, Shi Z (1995) Hydration of the carbonyl group—Atheoretical study of the cooperative mechanism. J Am Chem Soc 117(15):4240–4260.

44. Krizner HE, De Haan DO, Kua J (2009) Thermodynamics and kinetics of methylglyoxaldimer formation: A computational study. J Phys Chem A 113(25):6994–7001.

45. Buschmann HJ, Dutkiewicz E, Knoche W (1982) The reversible hydration of carbonylcompounds in aqueous solution. 2. The kinetics of the keto-gem-diol transition. PhysChem Chem Phys 86(2):129–134.

46. Buschmann HJ, Fuldner HH, Knoche W (1980) Reversible hydration of carbonylcompounds in aqueous solution. 1. Keto-gem-diol equilibrium. Phys Chem ChemPhys 84(1):41–44.

47. Bell RP, Rand MH, Wynnejones KMA (1956) Kinetics of the hydration of acetaldehyde.Transactions of the Faraday Society 52(8):1093–1102.

48. Hsieh YH, et al. (2005) Hydration of the carbonyl group—Acetic acid catalysis in theco-operative mechanism. Can J Chem 83(6-7):769–785.

49. Axson JL, Takahashi K, De Haan DO, Vaida V (2010) Gas-phase water-mediatedequilibrium between methylglyoxal and its geminal diol. Proc Natl Acad Sci USA 107(15):6687–6692.

50. Vaida V (2011) Perspective: Water cluster mediated atmospheric chemistry. J ChemPhys 135(2):020901.

51. Pocker Y, Meany JE, Nist BJ, Zadorojn C (1969) Reversible hydration of pyruvic acid. I.Equilibrium studies. J Phys Chem 73(9):2879–2982.

52. Maron MK, Takahashi K, Shoemaker RK, Vaida V (2011) Hydration of pyruvic acid toits geminal-diol, 2,2-dihydroxypropanoic acid, in a water-restricted environment.Chem Phys Lett 513(4-6):184–190.

53. Arnett JF, Larson DB, McGlynn SP (1973) Absorption and emission spectroscopy ofpyruvic acids and pyruvate esters. J Am Chem Soc 95(23):7599–7603.

54. Schreiner PR, et al. (2011) Methylhydroxycarbene: Tunneling control of a chemicalreaction. Science 332(6035):1300–1303.

55. Guzmán MI, Colussi AJ, Hoffmann MR (2006) Photogeneration of distant radical pairsin aqueous pyruvic acid glasses. J Phys Chem A 110(3):931–935.

56. Levinger NE, Rubenstrunk LC, Baruah B, Crans DC (2011) Acidification of reverse mi-cellar nanodroplets by atmospheric pressure CO2. J Am Chem Soc 133(18):7205–7214.

57. Ellison GB, Tuck, Vaida V (1999) Atmospheric processing of organic aerosols. J Geo-phys Res 104(D9):11633–11642.

58. Dobson CM, Ellison GB, Tuck AF, Vaida V (2000) Atmospheric aerosols as prebioticchemical reactors. Proc Natl Acad Sci USA 97(22):11864–11868.

59. Gilman JB, Tervahattu H, Vaida V (2006) Interfacial properties of mixed films of long-chain organics at the air-water interface. Atmos Environ 40(34):6606–6614.

60. Tervahattu H, et al. (2005) Fatty acids on continental sulfate aerosol particles. J Ge-ophys Res D Atmos 110(D6):D06207.

61. Donaldson DJ, Vaida V (2006) The influence of organic films at the air-aqueousboundary on atmospheric processes. Chem Rev 106(4):1445–1461.

62. Montgomery JA, Frisch MJ, Ochterski JW, Petersson GA (1999) A complete basis setmodel chemistry. VI. Use of density functional geometries and frequencies. J ChemPhys 110(6):2822–2827.

63. Frisch MJ, et al. (2003) Gaussian 03, Revision C.02 (Gaussian, Inc., Wallingford CT).

Griffith et al. PNAS | July 16, 2013 | vol. 110 | no. 29 | 11719

CHEM

ISTR

Y

Dow

nloa

ded

by g

uest

on

Feb

ruar

y 11

, 202

0