The automatic generation of reduced mechanisms for tropospheric chemistry modelling

The automatic generation of reduced mechanisms for tropospheric chemistry modelling

Atmospheric Environment 34 (2000) 1563}1574 The automatic generation of reduced mechanisms for tropospheric chemistry modelling D.J. Fish* Department...

230KB Sizes 1 Downloads 79 Views

Atmospheric Environment 34 (2000) 1563}1574

The automatic generation of reduced mechanisms for tropospheric chemistry modelling D.J. Fish* Department of Meteorology, University of Reading, Earley Gate, PO Box 243, Reading, RG6 6BB, UK Received 23 February 1999; received in revised form 16 August 1999; accepted 3 September 1999

Abstract An automatic hydrocarbon oxidation system that uses structure activity relationships to automatically calculate how aliphatic hydrocarbons breakdown in the atmosphere is described. The program includes functions to calculate average reaction rates and products for use in a reduced mechanism based on functional groups. Chemical concentrations from the reduced mechanism compare well with results from the regional atmospheric chemistry mechanism, RACM, for clean and polluted boundary layer scenarios. The main advantage of this approach is the ease of generating di!erent reduced mechanisms for di!erent emissions scenarios, such as for di!erent countries, or to take account of reductions in the emissions of particular VOCs. ( 2000 Elsevier Science Ltd. All rights reserved. Keywords: Air quality modelling; Lumped mechanism; Tropospheric ozone

1. Introduction Reduced chemical mechanisms, such as the regional acid deposition model (RADM2, Stockwell et al., 1990), and the carbon bond mechanism (CBM-IV, Gery et al., 1989), are widely used in chemistry and transport models to represent the many thousands of chemical reactions actually occurring at the large number of points in the 3D model domain. The main limitation of these reduced mechanisms is that they have been developed manually. For example, lumped mechanisms like RADM-2 and its successor, the regional atmospheric chemistry mechanism, (RACM, Stockwell et al., 1997), lump related compounds together into a single chemical species which reacts at an average rate to give average products. If the mechanism is applied to a region with di!erent proportions of individual volatile organic compounds (VOCs) then the average rates and products are less valid. Similarly, carbon-bond mechanisms are based on the reactions of functional groups, but have not been system-

E-mail address: d.j."[email protected] (D.J. Fish)

atically developed to take account of interactions between the di!erent functional groups in a molecule. An alternative way to generate reduced chemical mechanisms has recently been developed by Carslaw et al. (1998), who applied mechanism reduction techniques to the master chemical mechanism (Jenkin et al., 1997) to systematically eliminate unnecessary chemical species and reactions. Like Carslaw et al., the aim of the work described in this paper is to develop a reduced mechanism systematically, taking the full chemistry into account. However, the approach taken is quite di!erent. The aim of this work is to develop a reduced mechanism based (like CBM-IV) on functional groups, but taking account of the di!erent chemical environments in which each functional group can occur. In order to do this it was necessary to consider all the VOCs that are emitted and all the VOCs formed when the emitted compounds break down. In this work, structure}activity relationships (SARs) have been used to describe how aliphatic alkanes and alkenes break down in the atmosphere. These structure}activity relationships can be automatically evaluated for virtually any non-cyclic aliphatic compound using object-oriented programming (in C##) to provide the computational equivalent of a physical ball and stick model of organic compounds.

1352-2310/00/$ - see front matter ( 2000 Elsevier Science Ltd. All rights reserved. PII: S 1 3 5 2 - 2 3 1 0 ( 9 9 ) 0 0 4 1 3 - 6

1564

D.J. Fish / Atmospheric Environment 34 (2000) 1563}1574

Once arbitrary molecules can be represented and created on the computer, rules for their chemical reactions can be input and degradation mechanisms containing many thousands of reactions generated. The structure of this paper is follows. The use of C## to represent molecules and their reactions is described in Section 2, and the structure-activity relationships used are listed in Section 3. The program keeps track of each time particular functional groups, such as CH groups, 3 react in order to calculate average rate constants and products, as described in Section 4. The resulting reduced mechanism is listed in Section 5 and results from this mechanism are compared with results from the RACM in Section 6. Advantages and disadvantages of the method, and possible applications, are summarised in Section 7.

ecule to work out which carbon atom belongs in each product fragment, just as a chemist would look along the sequence of carbon atoms in the line structure to do this. Each carbon atom object has a &type', which corresponds to the number of carbon atoms to which it is bonded. CH groups, CH OH groups and CHO groups 3 2 are all of type &primary', as they are bonded to one other carbon atom. A CH or CHOH group, bonded to two 2 other carbon atoms, is classi"ed as &secondary'. Carbon atom objects also have a &status', such as &alkyl', &carbonyl', &hydroxyl' or &alkene' corresponding to their functional group. A collection of carbon atom objects, bonded together by pointers, comprises a molecule object. 2.2. How the program works

2. The C## hydrocarbon automatic oxidation system The fundamental principle underlying the C## hydrocarbon automatic oxidation system (CHAOS) is that it should be able to work out the degradation mechanism for a particular VOC in the same way that a chemist would. 2.1. How molecules are represented First, the system needs to be able to represent the chemical structure of the VOCs considered } non-cyclic alkanes and alkenes and their degradation products. A chemist might represent such compounds using a physical ball and stick model, or by drawing line structures. Using C##, a language designed for object-oriented programming (Stroustrup, 1991), the CHAOS represents each carbon atom in a molecule as an object. Objects are constructs that include both data (e.g. which other atoms the carbon atom is bonded to) and functions which act on that data (e.g. to convert an oxy radical to an aldehyde by ejecting hydrogen, or to work out the rate of reaction with OH). The objects representing carbon atoms are joined together by pointers, which act as the computational equivalent of chemical bonds. A pointer is a variable whose value is the location of an address in memory. Each carbon atom object contains one pointer to the address of each carbon object that it is bonded to. A bond between carbon atoms A and B is thus represented by two pointers: one from A to B and one from B to A. Using this representation it is possible for the CHAOS to work out the e!ect of reactions such as the decomposition of an oxy radical, e.g. CH CH C(CH )O.CH PCH CH .#CH COCH . 3 2 3 3 3 2 3 3 Once it has identi"ed which carbon}carbon bond to break, the CHAOS can follow pointers along the mol-

A list of VOCs is input to the CHAOS as character strings which represent the VOCs by their condensed formula, e.g. &CH3CH2CH(CH3)CH3'. A constructor function then constructs the corresponding molecule object by stepping along this character string and interpreting it. In this case, the function would create 5 carbon atom objects, bond them together in the correct way and set each of their statuses to &alkyl'. Once a molecule object has been correctly created it is sent to a function called react, which works out how that molecule will react. The function works by stepping through the array of carbon atoms in the molecule to see which carbon atom or atoms governs the reactivity. In the case of iso-pentane, react will recognise that each alkyl group is of comparable reactivity and that its only reaction is with OH. In order to investigate the e!ect of each possible reaction pathway, react makes "ve identical copies of the isopentane molecule object, using a copy constructor. Each alkyl group in turn is then allowed to react with OH to form an alkyl radical which immediately reacts with oxygen to form a peroxy radical. The CHAOS also keeps track of the fraction of isopentane which reacts each way, by comparing the rate constants (calculated using structure}activity relationships) for each reaction. Having accomplished a reaction, react recursively calls itself to consider the reactions of this new intermediate. The CHAOS is currently unable to di!erentiate between carbon atoms in identical environments, such as the two of the methyl groups in isopentane, and it follows the reactions of each group separately. This time when react looks along the chain of carbon atoms in the peroxy radical it recognises that the peroxy functional group is by far the most reactive, and will dominate the reactivity of this intermediate. As described in Section 3.3, the CHAOS considers three reactions of peroxy radicals: reaction with HO to form a peroxide, 2 reaction with NO to form an oxy radical or an organic nitrate, and reaction with other peroxy radicals. This time CHAOS makes several copies of the molecule, and

D.J. Fish / Atmospheric Environment 34 (2000) 1563}1574

allows one to react in each way. Typical concentrations of reactants such as OH (0.1 pptv), NO (1 ppbv), HO (10 2 pptv) and other organic peroxy radicals are speci"ed, in order to estimate the fraction of peroxy radicals which react in each way. Note that in the reduced mechanism, described in Section 4, the concentrations of OH, NO and HO vary } the above concentrations are only used 2 as a guide to predict what compounds are likely to form during VOC degradation so that the chemical reactions of these secondary VOCs can be taken into account in the reduced mechanism. Sensitivity studies for the plume scenario discussed in Section 6 showed that results were not a!ected by a factor of 10 increase in the assumed NO mixing ratio. However, when the NO mixing ratio was decreased by a factor of 10 mixing ratios of O , OH, HO , NO and 3 2 x PAN were unaltered, while mixing ratios of ROOH, H O , HCHO, aldehydes, ketones, NO and HNO 2 2 3 3 changed by around 10% on average. Results from the CHAOS thus show a small sensitivity to the assumed mixing ratios, but no reduced mechanism can be expected to simulate the whole range of tropospheric conditions. The CHAOS thus generates many thousands of chemical reactions as it works out how the input VOC breaks down. The degradation process is stopped if the molecule is oxidised to molecules like CO, CO , or HCHO, which 2 contain only one carbon atom and whose oxidation would be represented explicitly in a reduced chemical mechanism. The degradation is also stopped if the amount of a certain product formed falls below a very small fraction (typically 1]10~5) of the amount of VOC originally input. The CHAOS was designed to generate reduced mechanisms for use in regional models, and is thus concerned with the reactions that occur within a few days of the VOC being emitted. Consequently, the timescale for each reaction is estimated and only the fraction of the reactant that would react in the time available is converted to the product of that reaction. The initial time period (typically 5 days) is speci"ed at the start and the amount of time remaining is reduced following a slow reaction. For example, if an alkane reacted with OH and the e-folding lifetime was 1 day (1!exp(!5/1)) of the alkane would react during the 5 day period, but in the CHAOS all of the alkane is assumed to react instantaneously with OH after 1 day, leaving 4 days left for the reaction products to react. Although in reality each compound would react gradually, this approach captures the essential di!erence in behaviour between compounds that react fully during the "ve-day period and those which do not react signi"cantly. The CHAOS is thus able to calculate how any noncyclic aliphatic VOC breaks down in the atmosphere, using the structure}activity relationships described in Section 3. This information is then used to calculate

1565

a reduced mechanism, since rate constants and product distributions can be extracted from the mechanism and averaged to represent the overall e!ect of the reactions occurring.

3. Structure}activity relationships The idea of this work is to use structure}activity relationships to describe all the reactions in the degradation of all hydrocarbons larger than methane and ethene, rather than just to estimate rate constants that have not been measured. Consequently, the SARs will be described in some detail below. 3.1. Reactions with OH 3.1.1. Alkanes The site-speci"c attack of alkanes by OH was modelled using the structure}activity relationships described by Kwok and Atkinson (1995), in which the rates of H abstraction from primary, secondary and tertiary alkyl groups are described by equations of the form "k f (X) f Y k (X~CH2 ~Y) 4%# where k is the (temperature-dependent) group rate 4%# constant for the reactions of secondary alkyl groups with OH, and f (X) and f (Y) are factors which describe the e!ect of the substituents X and Y on the rate constant. These structure-activity relationships allow rate constants to be predicted to within a factor of two of experimental values for a wide range of compounds. The resulting alkyl radicals are immediately converted to peroxy radicals by the addition of oxygen. 3.1.2. Nitrates, peroxides, aldehydes and ketones A wide range of compounds } nitrates, peroxides, alcohols, aldehydes and ketones } are formed during the breakdown of alkanes. The rate of reactions of these compounds with OH are also calculated using the structure}activity relationships of Kwok and Atkinson (1995), including the abstraction of H atoms from OH, CHO and OOH functional groups. When OH abstracts a hydrogen atom from a carbon compound with a hydroxyl, peroxide or nitrate substituent the resulting alkyl radical is allowed to decompose rather than form a peroxy radical, e.g. CH CH OH#OHPCH C z HOH 3 2 3 PCH CHO#HO . 3 2 3.1.3. Alkenes and conjugated dienes The site-speci"c addition of OH to alkenes was modelled using the structure-activity relationships described by Peeters et al. (1994). Rate constants k , 13*.

1566

D.J. Fish / Atmospheric Environment 34 (2000) 1563}1574

k and k describe the rate of addition of OH to 4%# 5%35 alkenes to form primary, secondary and tertiary alkyl radicals. These alkyl radicals are assumed to react immediately with O to form peroxy radicals. 2 When OH adds to conjugated alkenes some of the resulting alkyl radicals can be stabilised by resonance. Rate constants for the formation of these resonance canonicals were also taken from Peeters et al. (1994). For example, when OH adds to isoprene CH "C(CH )CH"CH #OHP 2 3 2 CH OH!C z (CH )CH"CH % 2 3 2 CH OH!C(CH )"CH!CH z 2 3 2 the resulting generic rate constant is labelled k , 5%35@13*. because there is resonance between tertiary and primary alkyl radicals.

or a carboxylic acid as follows. H COO#H OPHCOOH#H O (99.5%) 2 2 2 HCHO#H O (0.5%) 2 2 RCHOO#H OPRCOOH#H O (37.5%) 2 2 PRCHO#H O (62.5%) 2 2 R COO#H OPR C(O)#H O . 2 2 2 2 2 Two types of decomposition reaction to form OH were considered. CH OO and RCHOO were allowed to react 2 as follows: RCHOOPRO #CO#OH 2 so an alkyl group such as CH is converted to a peroxy 2 radical like CHO . R COO biradicals were allowed to 2 2 react by the hydroperoxide channel (e.g. Atkinson et al., 1995)

3.2. Reactions with O3 The reaction of alkenes with ozone is of comparable importance to their reaction with OH in the troposphere. Generic rate constants for the reactions of alkenes with di!ering degrees of substitution were taken from Table 1 of Jenkin et al. (1997). Alkenes react with ozone to form a molozonide, which rapidly decomposes to form a carbonyl compound and an excited Criegee biradical (e.g. Atkinson and Carter, 1984), for example RCH"CH #O Pmolozonide 2 3 Pa(HCHO#RCHOO)#(1!a)RCHO #CH OO 2 For 1-alkenes, laboratory studies have measured values of a between 0.32 and 0.65, consistent with opposing electronic and steric e!ects (Grosjean and Grosjean, 1996). Fewer experiments have been carried out on internal alkenes, and in this work molozonides have been assumed to decompose to form equal amounts of the two possible pairs of products. The excited Criegee biradicals can react in one of three ways: they can be stabilised by collision with an air molecule, forming a stabilised Criegee biradical; they can decompose generating OH radicals; or they can decompose without forming OH radicals. The proportions of Criegee biradicals reacting in each way were taken from the fates of symmetric alkenes in Table 9 of Jenkin et al. (1997), using the entries for the symmetrical alkenes CH "CH for CH OO, RCH"CHR for RCHOO 2 2 2 and R C"CR for R COO. 2 2 2 Stabilised Criegee radicals were assumed to react exclusively with water to form either a carbonyl compound

R }CH }C.(R )}O}O.PR CH"C(R )}OOHP 1 2 2 1 2 R CH"C(R )}O.PR CHO .C(O)R 1 2 1 2 2 The program automatically calculates whether each of the possible alkyl groups attached to the Criegee group can eject a hydrogen and assumes equal probability if both can. If the group that surrenders a hydrogen is not a straightforward alkyl group then the appropriate alternative reaction is allowed, e.g. if a carbon atom bonded to a nitrate group forms an alkyl radical it will eliminate NO to form a carbonyl compound, and so on. 2 Finally, two types of decomposition not forming OH were considered, again following Jenkin et al. (1997): CH OOPCO#H O 2 2 and RC(OO)RPR #R #CO (50%) 1 2 2 PR !R #CO (50%). 1 2 2 3.3. Reactions of peroxy radicals Peroxy radicals were allowed to react with NO, HO 2 and other peroxy radicals. Typical mixing ratios of NO, HO , and methyl, primary, secondary, tertiary and acyl 2 peroxy radicals are speci"ed in order to calculate the fraction of each peroxy radical that reacts in each way. Peroxy radicals were assumed to react with HO to form 2 peroxides, and rate coe$cients for alkyl peroxy radicals were calculated using the expression k(RO #HO )"3.0]10~13][1!exp(!0.34n)] 2 2 ]exp (1250/¹),

D.J. Fish / Atmospheric Environment 34 (2000) 1563}1574

where n is the number of carbon atoms in the molecule (Jenkin et al., 1997). Acyl peroxy radicals were assumed to react with HO at the same rate as CH C(O)O , again 2 3 2 following Jenkin et al. (1997). The available laboratory data for the reactions of peroxy radicals with NO is contradictory: one study suggests that the rate constant is independent of the size of the peroxy radical (Eberhard and Howard, 1997), while another suggests that the rate constant decreases exponentially with size (Sehested et al., 1993). In this work rate constants for the reaction of alkyl peroxy and acyl peroxy radicals with NO were calculated using the expressions k(RO #NO)"2.54]10~12 exp(360/¹), 2 k(RC(O)O #NO)"8.10]10~12 exp(270/¹) 2 which were taken from Seakins et al. (1999), based on the measurements of Eberhard and Howard (1997). The reaction of RO with NO has two pathways 2 RO #NOPRO#NO 2 2 RO #NO#MPRONO #M 2 2 and the fraction of the peroxy radical reacting in each way was calculated following the rules in Carter and Atkinson (1989). Acyl peroxy radicals were also allowed to react with NO to form PAN in a reversible reaction, 2 again using the rate of reaction of CH C(O)O . 3 2 Rate constants and product distributions for the reactions of peroxy radicals with other peroxy radicals were calculated using the simpli"ed method of Madronich and Calvert (1990). Typical mixing ratios of CH O , 3 2 }CH O , }CHO , }CO and C(O)O are speci"ed. 2 2 2 2 2 3.4. Reactions of alkoxy radicals Alkoxy radicals were allowed to react with O , to 2 decompose and to isomerise. Rate constants for the reactions of primary and secondary alkoxy radicals with O to form aldehydes and ketones were taken from 2 Atkinson (1997). Rates of decomposition were calculated as the sum over all decomposition pathways of the rate

1567

for the ejection of a CH , }CH , }CH, }C, hydroxy-alkyl, 3 2 or acyl group from a primary, secondary or tertiary alkoxy radical, again following Atkinson (1990), as more recent formulations using the enthalpy change resulting from the decomposition are not general enough to apply to any alkoxy radical that might be formed. Rate constants for isomerisation were also taken from Atkinson (1997). Isomerisation rate constants were assumed to depend only on the nature of the group from which a hydrogen atom was extracted, with primary alkyl groups releasing hydrogen atoms less readily than secondary alkyl groups. A "rst isomerisation reaction was immediately followed by a second since, as described by Eberhard et al. (1995), hydrogen atoms are readily abstracted from the resulting hydroxyl-substituted alkyl groups.

4. Calculation of a reduced mechanism Before extracting a reduced mechanism from the CHAOS it was necessary to choose the functional groups that would form the basis of the reduced mechanism. This is somewhat subjective, and it was decided to represent primary, secondary and tertiary alkyl groups (}CH , 3 }CH and }CH), peroxy radicals and oxy radicals separ2 ately, and to include the main functional groups in secondary VOCs. The functional groups used are listed in Table 1. The representations of most functional groups should be relatively self-explanatory, but }C"CH rep2 resents terminal alkenes, while }C"C-represents internal alkenes. Preliminary work suggests that it is possibile to generate an even more reduced mechanism with just one class of alkyl group, peroxy radical and alkyl radical, and this will be investigated in future work. A reduced mechanism for the degradation of a single VOC can be generated as follows. The CHAOS is run for that VOC and each time one of the functional groups in Table 1 reacts the contribution of that reaction to the average reactions of the functional group is noted. Reactions fall into two types: simple bimolecular reactions for which there are only two possible products (e.g.

Table 1 Functional groups on which the reduced chemical mechanism is based Alkyl groups

Peroxy radicals

Oxy radicals Carbonyl groups

Acyl peroxy

PAN

CH 4 }CH 3 }CH 2 }CH

CH O 3 2 }CH O 2 2 }CHO 2 }CO 2

CH O. 3 }CH O. 2 }CHO. }CO.

}CO . 3

}CO NO 3 2

HCHO }CHO }C(O)

Peroxides

Carboxylic acids

CH OOH 3 }OOH

HCOOH }COOH

Alkenes

}C"CH 2 }C"C-

1568

D.J. Fish / Atmospheric Environment 34 (2000) 1563}1574

}CH O #HO P}CH O H#O ) and reactions which 2 2 2 2 2 2 can lead to the formation of several products or to di!erent products for the same functional group in di!erent molecules (e.g. the reaction of internal alkenes with O ). 3 For the simple bimolecular reactions the system just needs to take the averate rate constant for each occurrence of that reaction, weighted by the fraction of the initial VOC that reacts that way. For the reactions with several possible products, the system also has to calculate the fraction of each possible product formed on average. Once average reaction rates and products have been calculated for each VOC in the emissions inventory of interest overall average reactions rates and products are calculated by weighting the reactions of each VOC according to the amount of VOC emitted. The temperature dependence of the averaged rate constants was calculated by running the CHAOS for atmo-

spheric temperatures of 273 and 298 K and "tting the resulting rate constants to obtain the Arrhenius parameters for each reaction. The fractional amounts of each product formed at 298 K were adopted for all temperatures. The CHAOS does not include photolysis, and photolysis reactions were manually added to the reduced mechanism. This is not thought to be a signi"cant limitation, as absorption cross-sections and quantum yields are often not as well known as rate constants, so assumptions about photolysis rates have to be made in all chemical mechanisms. 5. Listing of mechanism The following inorganic compounds were used in the reduced mechanism: N and O , which were not integ2 2 rated; the oxygen compounds O, O(1D) and O ; the 3

Table 2 Bimolecular reactions used in the reduced mechanism! Reaction

A

O(1D)#O PO#O 2 2 O(1D)#N PO#N 2 2 O(1D)#H OP2OH 2 OH#O PHO #O 3 2 2 OH#H PHO # H O 2 2 2 HO #O POH#O 2 3 2 OH#HO PH O# O 2 2 2 HO #HO PH O # O 2 2 2 2 2 OH#H O PHO # H O 2 2 2 2 OH#HNO PH O #NO 3 2 3 OH#HNO PH O #NO #O 4 2 2 2 HO #NOPOH# NO 2 2 NO#O PNO #O 3 2 2 NO #O PNO #O 2 3 3 2 NO#NO P2NO 3 2 CO#OHPHO (# CO ) 2 2 HCHO#OHPCO# HO 2 HCHO #NO PCO#HNO #HO 3 3 2 CHO#NO PCO # HNO 3 3 3 CH "CH #NO P!CH O #-CH 2 2 3 2 2 3 CH #OHPCH O #H O 4 3 2 2 CH O #NOP CH O#NO 3 2 3 2 CH O #HO P CH OOH#O 3 2 2 3 2 CH O #NO P CH O#NO 3 2 3 3 2 }CH O #NO P-CH O#NO 2 2 3 2 2 }CHO #NO P-CHO#NO 2 3 2 }CO #NO P-CO#NO 2 3 2 CH O#O P HCHO#HO 3 2 2 +CH3 1OHP-CH2 O2 1H2 O +CHO1OHP-CO3 1 H2 O +CH2 O2 1HO2 P-OOH1H2 O +CHO2 1HO2 P-OOH1H2 O +CO2 1HO2 POOH 1H2 O

3.2]10~11 1.8]10~11 2.2]10~10 1.9]10~12 7.7]10~12 1.4]10~14 4.8]10~11 * 2.9]10~12 * 1.5]10~12 3.7]10~12 1.8]10~12 1.2]10~13 1.8]10~11 * 8.6]10~12 5.8]10~16 1.4]10~12 3.3]10~12 2.3]10~12 4.2]10~12 3.8]10~13 1.2]10~12 1.2]10~12 1.2]10~12 1.2]10~12 7.2]10~14 2.39]10~12 5.26]10~12 2.04]10~13 2.23]10~13 2.38]10~13

E/R }67 }107 0 1000 2100 600 }250 160 !360 !240 1370 2450 !110 !20 0 1860 2880 1765 }180 }780 0 0 0 0 1080 848 }445 }1232 }1288 }1261

!Reactions in bold type were calculated using the CHAOS. Reactions marked * have more complicated rate expressions which were taken from Atkinson et al. (1997).

D.J. Fish / Atmospheric Environment 34 (2000) 1563}1574

1569

Table 3 Bimolecular reactions with fractional products used in the reduced mechanism. Reactions in bold type were calculated using the CHAOS Reaction

A

E a

CH "CH #O PHCHO#0.44CO#0.25HO #0.37COOH#0.12OH 2 2 3 2 C"CH #NO P0.35CH #0.35CH O #0.65CH #0.65CHO 2 3 2 2 2 3 2 }C"C!#NO PCH#0.8CHO #0.2CO 3 2 2 OOH#OHP0.32CH O #0.18HCHO#0.43OH#0.22CH O #0.04CHO#0.20C(O)#0.03NO 3 2 2 2 2 CH OOH#OHP0.65CH O #0.35HCHO#0.35OH 3 3 2 HO #NO P0.3HNO #0.7NO #0.7OH 2 3 3 2 ethyne#OHP1.3HCHO#0.35COOH#0.35CO#HO }0.65NO#0.65NO 2 2 CH O #CH O P0.8HO #1.4HCHO#0.6CH 3 2 3 2 2 4 }CH O #-CH O P1.2CH Oz#0.4-CHO#0.4}CH 2 2 2 2 2 3 CH O #}CH O P0.75HCHO#0.5HO #0.5CH Oz#0.25-CHO#0.25-CH #0.25CH 3 2 2 2 2 2 3 4 CH O #-CHO P0.75HCHO#0.5HO #0.25-CHOz#0.25-C(O)#0.25-CH #0.25CH 3 2 2 2 2 4 CH O #-CO P1.0HCHO#0.7HO #0.7-COz#0.3-CH 3 2 2 2 -CH O #-CHO P0.6CH Oz#0.6CHOz#0.2-CHO#0.2-CH #0.2C(O)#0.2-CH 2 2 2 2 2 3 }CH O #-CO P0.8CH Oz#0.8COz#0.2-CHO#0.2-CH 2 2 2 2 CH O #-CO zP1.0HCHO#0.7HO #0.7-CH O #0.3-COOH 3 2 3 2 3 2 }CH O #-CO zP0.8CH Oz#0.8CH O #-0.2-CHO#0.2-COOH 2 2 3 2 3 2 }CHO #-CO zP,)8CHOz#0.8CH O #-0.2-C(O)#0.2-COOH 2 3 3 2 }CO #-CO zP1.0CH Oz#1.0CH O 2 3 2 3 2 }CO z#-CO zP2.0CH O 3 3 3 2 CH2 1OHP 0.25HO2 10.71-CHO2 10.04CH2 O10.25-CHO CH1OHP 0.013HO2 10.98-CO2 10.013-C(O) +C5CH2 1OHP 0.87CH3 10.122CH2 10.13CH2 O2 10.82CHO2 10.05-CO2 +C5C-1OHP 0.96CH2 10.039CH10.93CHO2 10.072CO2 +C5CH2 1O3 P 0.35OH10.03HO2 -0.16-CH3 -0.16-CH-0.03-CH10.18HO2 10.18CH3 O2 1 0.14CH2 O2 10.03CHO2 10.59CO10.5HCHO10.51-CHO10.04C(O)10.23COOH +C5C-1O3 P0.61OH10.06HO2 -0.60CH3 -0.12CH2 10.23HO2 10.53CH3 O2 1 0.19CH2 O2 10.52CO10.97-CHO10.18C(O)10.10-COOH +CH2 O2 1NOP 0.97NO2 10.03CHO10.97CH2 O +CHO2 1NOP 0.78NO2 10.22CH2 10.78-CHOz +CO2 z1NOP 0.94NO2 10.06CH10.94COz +CO3 z1HO2 P 0.71CH3 10.29COOH10.29O3 +CO3 z1NOP 1.0NO2 20.61CH3 -0.27CH2 10.04CH10.04HO2 10.58CH3 O2 10.26CH2 O2 10.04CHO2 10.03CHO10.07C(O)10.08CO3 z CH2 Oz1O2 P +0.13CH3 -0.01CH2 10.97HO2 10.30HCHO10.84CHO-0.02C(O)10.02CO3 z +CHOz1O2 P-0.07CH3-0.10CH2 -0.02CH10.71HO2 10.06CH2 O2 10.01CHO2 10.01CO1 0.07HCHO10.43CHO10.40C(O)10.20CO3 z COz1O2 P-0.34CH3 -0.49CH2 -0.04CH10.03HO2 10.32CH3 O2 10.48CH2 O2 10.04CHO2 1 0.01CO10.01HCHO10.88C(O)10.12CO3 z +CHO2 1-CHO2 P 1.2-CHOz10.4C(O)10.4-CH2 +CO2 1-CO2 P 2-COz +CHO2 1-CO2 P 0.8CHOz10.8COz10.2-C(O)10.2-CH

9.1]10~15 1.79]10~13 8.64]10~13 3.2]10~12 3.2]10~12 3.5]10~12 2.6]10~13 4]10~13 4]10~13 4]10~13 5.1]10~14 4.9]10~15 5.1]10~14 4.9]10~15 1.55]10~12 1.55]10~12 9.8]10~14 1.9]10~14 6]10~12 2.61]10~12 1.36]10~12 3.6]10~11 6.28]10~11 1.02]10~17

2580 450 !450 !190 !190 0. 820 0 0 0 0 0 0 0 0 0 0 0

hydrogen compounds H , OH, HO , H O and H O; 2 2 2 2 2 the nitrogen species NO, NO , NO , N O , HNO , 2 3 2 5 3 HNO and HONO; and CO. Rates of chemical reactions 4 of inorganic species, reactions of methane and its degradation products, and reactions of compounds with NO 3 were taken from Atkinson et al. (1997). The more important peroxy radical interactions were also included in the reduced mechanism following Madronich and Calvert (1990). These reactions and averaged reactions of func-

1.87]10v16

267 !135 0 0 0 0

2.54]10~12 2.54]10~12 2.54]10~12 4.24]10~13

!360 !360 !360 !1046

8.00]10~12 3.93]10~12 2.13]10v8

!275 1663 3892

3.21]10~6

4472

1.6]10~15 6.0]10~17 3.1]10~16

0 0 0

tional groups that have been output from the C# # program are listed in Tables 2 and 3. It should be remembered that this mechanism is valid for the particular mixture of VOCs input to the C# # program and that the mechanism can thus be optimised for di!erent regions (e.g. the US versus Europe) or di!erent atmospheric conditions. Finally, photolysis reactions are listed in Table 4, and the rates of uni- and termolecular reactions are listed in Table 5.

1570

D.J. Fish / Atmospheric Environment 34 (2000) 1563}1574

6. Comparison of the reduced mechanism with the regional atmospheric chemistry mechanism The performance of the reduced mechanism output from the C# # program has been tested by comparison with results from the Regional Atmospheric Chemistry Model (M. Karl, personal communication). Both models were run for the PLUME-1 scenario from the intercomparison of Kuhn et al. (1998) except that aromatic compounds, which are not represented in the CHAOS at present, were omitted. Results from RACM simulations with and without aromatic compounds were very similar, partly because oxidant formation is NO -limited in this x scenario. Essentially this scenario represents the polluted boundary layer and includes emissions of NO, NO , CO, 2 CH , and 53 di!erent hydrocarbons. 4

Table 4 Photolysis reactions included in the reduced mechanism O 3 O 3 H O 2 2 NO 2 NO 3 NO 3 HNO 3 HNO 4 HONO HCHO HCHO }CHO }C(O) OOH CH OOH 3 N O 2 5

P P P P P P P P P P P P P P P P

O#O 2 O(1D)#O 2 2OH NO#O NO#O 2 NO #O 2 NO #OH 2 HO #NO 2 2 OH#NO 2HO #CO 2 H #CO 2 CO#HO #CH2O 2 2 CO .#CH O 3 2 2 CH O.#OH 2 CH O.#OH 3 O #2NO 2

In general, there is good agreement between results from the two mechanisms, particularly given the large spread obtained in the intercomparison of Kuhn et al. (1998). The concentrations of ozone predicted by the two models are compared in Fig. 1(a). The increase in ozone concentration predicted by CHAOS is 8% lower than that from RACM, and the cause of this di!erence is not known. CHAOS also predicts lower concentrations of OH (Fig. 1(b)) than RACM. HO and H O (Fig. 1(c)) con2 2 2 centrations are in excellent agreement, but the CHAOS predicts lower peroxide (ROOH) concentrations than RACM. This di!erence can be attributed to the di!erent rate constants for the reaction of peroxy radicals with NO. In the RACM, the rate constant for the reaction RO #NO is 4]10~12 cm3 s~1 at all temperatures for 2 the peroxy radicals HC3P, HC5P and HC8P (formed from alkanes and related compounds which have been lumped on the basis of their reactivity with OH). In CHAOS the rate constants are larger: 8.86]10~11 cm3 s~1 at 288 K (see Section 3.3) so peroxy radicals react with NO rather than with HO to form 2 peroxides. The concentrations of carbonyl compounds calculated by the two mechanisms are shown in Fig. 2. CO, HCHO and several aldehydes and ketones are emitted in this simulation, but their formation by VOC oxidation is also important and there was a large spread (50% for HCHO) in the concentrations of carbonyl compounds calculated by models in the Photocomp intercomparison. In this comparison, concentrations of CO, HCHO and other aldehydes are in good agreement, while CHAOS forms less ketone than does RACM. The amount of ketone formed in the two simulations provides an interesting example of the advantage of the automatic hydrocarbon oxidation system. In RACM,

Table 5 Unimolecular and termolecular reactions included in the reduced mechanism! Reaction

F c

k300 0

n

O#O PO 2 3 OH#OHPH O 2 2 OH#NOPHONO OH#NO PHNO 2 3 HO #NO PHNO 2 2 4 NO #NO PN O 2 3 2 5 CO .#NO PCO NO 3 2 3 2 CH "CH #OHPCH O #CH 2 2 2 2 3

0.6 0.6 0.6 0.6 0.6 0.6 0.6 0.7

5.64]10~34 6.2]10~31 7.0]10~31 2.6]10!30 1.8]10~31 2.2]10~30 9.7]10~29 7.0]10~29

2.8 1 2.6 3.2 3.2 3.9 5.6 3.1

k300 =

m

2.6]10~11 1.5]10~11 2.4]10~11 4.7]10~12 1.5]10~12 9.3]10~12 9.0]10~12

0 0.5 1.3 1.4 0.7 1.5 0.

A

B

2.1]10~27 2.7]10~27 9]10~29

10900 11000 1400

! Rates of termolecular reactions are given by k "k (¹)[M]/(1#k [M]/k (¹))FM1`*-0'10 (k0 (T)*M+@k= (T))+2N~1, where f 0 0 = c k (¹)"k300(¹/300)~n and k (¹)"k300(¹/300)~m. Rates of the reverse reactions, i.e. unimolecular decomposition are given by 0 0 = = k "k /(Aexp(-B/T)). A and B parameters are given only for those compounds whose unimolecular decomposition is included. " &

D.J. Fish / Atmospheric Environment 34 (2000) 1563}1574

1571

Fig. 1. Mixing ratios of (a) O , (b) OH, (c) H O and (d) peroxides from RACM (solid line) and the CHAOS (dashed line) for the plume 3 2 2 scenario from Kuhn et al. (1998) but without aromatic compounds.

Fig. 2. Mixing ratios of (a) CO, (b) HCHO, (c) aldehydes and (d) ketones for the same case as Fig. 1.

1572

D.J. Fish / Atmospheric Environment 34 (2000) 1563}1574

Fig. 3. Mixing ratios of (a) NO , (b) NO , (c) PAN and (d) HNO for the same case as Fig. 1. x 3 3

one source of ketone is the reaction HC5#OHP0.75 HC5P#0.25 KET #0.25 HO #H O 2 2 where HC5 is the lumped species representing alkanes, alcohols, esters and alkynes with OH rate constants between 3.4]10~12 and 6.8]10~12 cm3 s~1. Isopropanol accounts for 25% of HC5, when aggregation factors are taken into account (Stockwell et al., 1997). Isopropanol reacts with OH as follows, generating acetone and HO 2 CH CHOHCH #OHPCH CzOHCH #H , 3 3 3 3 2 CH CzOHCH #O PCH COCH #HO , 3 3 2 3 3 2 thus accounting for the 25% ketone formation from the reaction of HC5 with OH. However, although isopropanol is a major component in the emissions inventory used to generate the RACM mechanism, it is not emitted in the Photocomp intercomparison. Consequently CHAOS should be more accurate in this case because the reduced mechanism has been automatically generated for the actual emissions used in this scenario. When the CHAOS mechanism was modi"ed to include 25% ketone formation in the reaction of CH2 groups with OH

higher ketone concentrations were obtained, in excellent agreement with RACM. The concentrations of NO , NO , peroxy acyl nitrate x 3 (PAN) and HNO calculated by the two models are 3 shown in Fig. 3. Concentrations of NO , NO and PAN x 3 are in excellent agreement, while CHAOS calculates smaller HNO concentrations, probably because of its 3 lower OH concentrations. After development, the ability of the CHAOS mechanism to simulate the chemistry occurring under clean conditions was checked using the LAND scenario from Kuhn et al. (1998). The CHAOS mechanism performed well, with calculated concentrations of ozone, OH, NO , x NO and H O agreeing well with the values from the 3 2 2 RADM.

7. Conclusions and Discussion A C# #hydrocarbon automatic oxidation system, capable of representing non-cyclic aliphatic VOCs and using structure}activity relationships to work out how they react in the atmosphere, has been developed. A reduced mechanism for tropospheric chemistry, containing 44 chemical species and 89 chemical reactions has been extracted from the CHAOS. It may be possible to further

D.J. Fish / Atmospheric Environment 34 (2000) 1563}1574

reduce this mechanism by lumping all peroxy radicals, alkoxy radicals and so on into a single species rather than having separate primary, secondary and tertiary radicals, by reducing the number of inorganic species, or by applying mechanism reduction techniques. Results from the reduced mechanism are in good agreement with results from the Regional atmospheric chemistry mechanism for the polluted troposphere scenario including emissions of 53 di!erent VOCs. However, some di!erences are apparent. There are many possible explanations for these di!erences, which are small compared with the spread of results from models participating in the intercomparison of Kuhn et al. (1998). However, one factor is that the RACM was developed using a di!erent emissions inventory to that used in the intercomparison, while the CHAOS used the VOC emissions from the intercomparison to generate the reduced scheme. This is a cause of di!erences in calculated ketone concentrations, as RACM erroneously produces ketone when HC5 (which includes isopropanol) reacts with OH, while isopropanol was not emitted in the intercomparison. The main advantage of the CHAOS is the ease with which a reduced mechanism that is appropriate for particular VOC emissions (characteristic of a particular region, for example) can be generated. The resulting reduced mechanism can also be easily modi"ed to simulate changes in emissions as a result of say legislation that restricts the usage of particular VOCs. There are many possible applications of the CHAOS, which will be explored in future work. The CHAOS could in theory be used to calculate photochemical ozone creation potentials, POCPs (Derwent et al., 1996). To calculate the POCP for a compound, the CHAOS would be used to generate a reduced mechanism with normal and perturbed amounts of that compound. The reduced mechanism would then be run for the two cases and the amount of ozone formation compared. If successful in a trajectory model, the reduced mechanism could be used in a more detailed chemistry and transport model to calculate POCPs using a more detailed representation of meteorology. Note that the CHAOS produces a reduced mechanism that is appropriate for one particular mix of VOCs, and that the reduced mechanism would be unable to simulate the e!ects of di!erent VOC emissions in di!erent areas accurately (e.g. over a forest compared with over an urban area). However, all reduced mechanisms su!er from this problem, and many emission inventories use the same VOC mixture in all areas. It is also planned to investigate the oxidation of biogenic VOCs using the automatic hydrocarbon oxidation system. The system can already simulate the degradation of conjugated alkenes, and it would be interesting to investigate how well it simulates the degradation of isoprene. In the future, the system will be modi"ed to treat cyclic compounds, allowing the system to simulate

1573

the degradation of terpenes such as a-pinene. Another possible application is the incorporation of rules relating structure and volatility, in order to predict the amount of secondary organic aerosol formed when terpenes and other hydrocarbons react in the atmosphere. Finally, when contrasting this method of developing reduced mechanisms with the mechanism reduction techniques of Carslaw et al. (1998) it is interesting to note that 90% of the reactions in the master chemical mechanism, from which the reduced mechanism is derived, were derived from structure activity relationships. The main philosophical di!erence between the two approaches is whether it is best to retain reactions explicitly or to lump similar reactions together. Lumping will probably result in a more highly reduced mechanism, while the explicit approach may be more accurate.

Acknowledgements I thank Bill Stockwell and Matthais Karl for providing results from the RACM model, and Nic Carslaw, Mike Pedder and Alison Tomlin for useful discussions.

References Atkinson, R., 1990. Gas-phase chemistry of organic compounds: a review. Atmospheric Environment 24A, 1}41. Atkinson, R., 1997. Atmospheric reactions of alkoxy and b-hydroxyalkoxy radicals. International Journal of Chemical Kinetics 29, 99}111. Atkinson, R., Baulch, D.L., Cox, R.A., Hampson, R.F., Kerr, J.A., Rossi, M.J., Troe, J., 1997. Evaluated kinetic, photochemical and heterogeneous data for atmospheric chemistry. 5. IUPAC subcommittee on gas kinetic data evaluation for atmospheric chemistry. Journal of Physical Chemistry Reference Data. 26, 521}1011. Atkinson, R., Carter, W.P.L., 1984. Kinetics and mechanisms of the gas-phase reactions of ozone with organic compounds under atmospheric conditions. Chemical Review 84, 437}470. Atkinson, R., Tuazon, E.C., Aschmann, S.M., 1995. Products of the gas-phase reactions of O with alkenes. Environmental 3 Sciences Technology 29, 1860}1866. Carslaw, N., Jacobs, P.J., Pilling, M.J., 1998. Modelling OH, HO and RO radicals in the marine boundary layer: 2. 2 2 Mechanism reduction and uncertainty analysis, submitted to Journal of Geophysical Research, for publication. Carter, W.P.L., Atkinson, R., 1989. Alkyl nitrate formation from the atmospheric photooxidation of alkanes: a revised estimation method. Journal of Atmospheric Chemistry 8, 165}173. Derwent, R.G., Jenkin, M.E., Saunders, S.M., 1996. Photochemical ozone creation potentials for a large number of reactive hydrocarbons under European conditions. Atmospheric Environment 30, 181}199.

1574

D.J. Fish / Atmospheric Environment 34 (2000) 1563}1574

Eberhard, J., Howard, C.J., 1997. Rate coe$cients for the reactions of some C to C hydrocarbon peroxy radicals with 3 5 NO. Journal of Physical Chemistry 101, 3360}3366. Eberhard, J., Muller, C., Stocker, D.W., Kerr, J.A., 1995. Isomerisation of alkoxy radicals under atmospheric conditions. Environmental Science and Technology 29, 232}241. Gery, M.W., Whitten, G.Z., Killus, J.P., Dodge, M.C., 1989. A photochemical kinetics mechanism for urban and regional scale microcomputer modelling. Journal of Geophysical Research 94, 12925}12956. Grosjean, E., Grosjean, D., 1996. Gas-phase reaction of ozone with 1-alkenes. Atmospheric Environment 30, 4107}4113. Jenkin, M.E., Saunders, S.M., Pilling, M.J., 1997. The tropospheric degradation of volatile organic compounds: a protocol for mechanism development. Atmospheric Environment 31, 81}104. Kwok, E.S.C., Atkinson, R., 1995. Estimation of hydroxyl radical rate constant for gas-phase organic compounds using a structure-reactivity relationship: an update. Atmospheric Environment 29, 1685}1695. Kuhn, M., et al., 1998. Intercomparison of the gas-phase chemistry in several chemistry and transport models. Atmospheric Environment 32, 693}709.

Madronich, S., Calvert, J.G., 1990. Permutation reactions of organic peroxy radicals in the troposphere. Journal of Geophysical Research 95, 5697}5715. Peeters, J., Boullart W., Van Hoeymissen, J., 1994. Site-speci"c partial rate constants for OH addition to alkenes and dienes, In: P.M. Borrell et al. (Eds.) The Proceedings of EUROTRAC symposium '94. SPB Academic Publishing, Netherlands. Sehested, J., Neilson, O.J., Wallington, T.J., 1993. Absolute rate constants for the reaction of NO with a series of peroxy radicals in the gas phase at 298K. Chemistry Physics Letters 213, 457}460. Stockwell, W.R., Kirchner, F., Kuhn, M., Seefeld, S., 1997. A new mechanism for regional atmospheric chemistry modelling. Journal of Geophysical Research 102, 25847}25897. Stockwell, W.R., Middleton, P., Chang, J.S., Tang, X., 1990. The second generation regional acid deposition model chemical mechanism for regional air quality modelling. Journal of Geophysical Research 95, 16343}16367. Stroustrup, B., 1991. The C## programming language, Addison Wesley Publishing Company, Reading, MA. ISBN 0-201-53992-6.