Predicting the Reactivity of Phenolic Compounds with Formaldehyde. II

PredictingtheReactivityofPhenolicCompoundswith Formaldehyde.II.ContinuationofanAbInitioStudy

TohruMitsunaga,1 AnthonyH.Conner,2 CharlesG.Hill,Jr.3 1 FacultyofBioresources,MieUniversity,Tsu514-8507,Japan 2 USDAForestService,ForestProductsLaboratory,Madison,Wisconsin53705 3 DepartmentofChemicalEngineering,UniversityofWisconsin,Madison,Wisconsin53705

Received20March2001;accepted20December2001 Publishedonline23July2002inWileyInterScience(www.interscience.wiley.com).DOI10.1002/app.10926

ABSTRACT:Phenol–formaldehyderesinsareimportant hydeasfunctionsoftime.Thereactionrateconstantsvaried Ϫ Ϫ Ϫ adhesivesusedbytheforestproductsindustry.Thephe- overawiderange(approximately10 2 to104 Lmol 1 h 1). noliccompoundsintheseresinsarederivedprimarilyfrom Anestimateofthereactivityperreactivesiteonthephenolic petrochemicalsources.Alternatesourcesofphenoliccom- ringwasdeterminedbydividingtheratebythenumberof poundsincludetannins,lignins,biomasspyrolysisprod- reactivesites.Atomicchargesforeachphenoliccompound ucts,andcoalgasiﬁcationproducts.Becauseofvariationsin werecalculatedbyabinitiomethodsattheRHF/6-31ϩG theirchemicalstructures,thereactivitiesofthesephenolic leveloftheoryusingtheCHelpGmethod.Thechargeper compoundswithformaldehydevaryinquitesubtleways. reactivesitewasestimatedbysummingthechargesatallthe Previously,itwasdemonstratedthatthereactivityofa reactivesitesonthephenolicringanddividingbythenum- numberofphenolswithformaldehydeinnonaqueouscon- berofreactivesites.Astrongcorrelationwasobserved ditionscouldbecorrelatedwithchargescalculatedforreac- betweenthereactivityperreactivesiteandtheaverage tivesitesonthearomaticring(Conner,A.H.JApplPolym chargeperreactivesite.©2002WileyPeriodicals,Inc.JAppl Sci2000,78,355–363).Westudiedthereactivityofalarger PolymSci86:135–140,2002 numberofphenoliccompoundswithformaldehydeinan aqueoussolutionusingsodiumhydroxideasthecatalyst. Reactionratesweredeterminedfrommeasurementsofthe Keywords:chemicalcomputation;abinitio;phenolics;form- concentrationsofthephenoliccompoundsandformalde- aldehyde;adhesives

INTRODUCTION Sprung6 investigatedthereactionsofaseriesof methylphenolswithformaldehyde.Hiskineticmea- Toutilizephenolicadhesivesystemsmoreeffectively surementswerebasedsolelyontherateofthedisap- andtodevelopnewphenolicadhesives,itisimportant tounderstandthereactionsofphenoliccompounds pearanceofformaldehyde.Asexpected,differencesin withformaldehyde.Todate,analyticalstudiesonphe- thereactivitiesofthisseriesofphenoliccompounds nolicadhesiveshaveconcentratedmainlyonkinetic dependedonsubtledifferencesinthechemicalstruc- 7 studies.1–5 Thesestudiesinvolvenotonlythecalcula- ture.Conner demonstratedthattherelativeratesof tionofreactionratesbutalsocomplexprocessesfor thesereactionscouldbecorrelatedwithelectrostatic isolationandidentiﬁcationofintermediates,aswellas chargesatreactivepositionsinthephenolicringcal- reactionproducts.Computationalchemistrymethods culatedusingabinitiomethods.Becauseoflimitations allowanalysisofreactionmechanismsandprediction ontheanalyticalinstrumentsinuseatthetimeSprung ofthereactivitiesofchemicalstartingmaterials. conductedhisstudy,itwasnotclearwhethereither Therefore,computationalchemistrymightbeusedto formaldehydeorthephenoliccompoundswereun- predictthereactivitiesofphenoliccompoundswith dergoingreactionsotherthanthoseinvolvedinhy- formaldehydeandtherebyprovidenewinsightinto droxymethylation.Moreover,Sprung’skineticdata thereactionmechanisms.Suchinformationwouldbe werecollectedinnonaqueoussystemsratherthanin usefulindevelopingstrategiesfortheformulationand theaqueous-basedsystemstypicallyencounteredin cureofphenolicadhesives.Thisinsightwouldalso industrialapplicationsofphenolicadhesives.Because servetodecreasethetimeneededfordevelopmentof oftheselimitationsontheearlierworkandtheindus- newadhesivesystems. trialsigniﬁcanceofphenol–formaldehydeadhesives, weemployedaqueous-basedsystemstoinvestigate thereactionsofformaldehydewithalargerseriesof Correspondenceto:T.Mitsunaga. phenoliccompounds(TableI).Thesephenoliccom- JournalofAppliedPolymerScience,Vol.86,135–140(2002) poundsincludedmostofthephenolsinvestigatedby ©2002WileyPeriodicals,Inc. Sprung. 136 MITSUNAGA, CONNER, AND HILL

TABLE I Phenolic Compounds, Their Reaction Rates with Formaldehyde, and the Charges at the Reactive Aromatic Sites as Calculated at the RHF/6-31؉G Level of Theory Using CHelpG Total Average k charge Average k charge Ϫ Ϫ Ϫ Ϫ Compound Abbreviation (L mol 1 h 1 Ϯ SE) (e) (L mol 1 h 1) (e)

2-Methylphenol anion 2MP 0.21 Ϯ 0.04 Ϫ0.853 1.10E Ϫ 01 Ϫ0.427 3-Methylphenol anion 3MP 1.80 Ϯ 0.36 Ϫ1.844 6.00E Ϫ 01 Ϫ0.615 4-Methylphenol anion 4MP 0.15 Ϯ 0.02 Ϫ0.936 7.70E Ϫ 02 Ϫ0.468 2,3-Dimethylphenol anion 23MP 1.28 Ϯ 0.21 Ϫ1.025 6.39E Ϫ 01 Ϫ0.513 2,4-Dimethylphenol anion 24MP 0.39 Ϯ 0.02 Ϫ0.412 3.91E Ϫ 01 Ϫ0.412 2,5-Dimethylphenol anion 25MP 1.51 Ϯ 0.29 Ϫ1.136 7.54E Ϫ 01 Ϫ0.568 2,6-Dimethylphenol anion 26MP 0.12 Ϫ0.359 1.18E Ϫ 01 Ϫ0.359 3,4-Dimethylphenol anion 34MP 0.61 Ϯ 0.30 Ϫ1.149 3.07E Ϫ 01 Ϫ0.575 3,5-Dimethylphenol anion 35MP 6.94 Ϯ 1.80 Ϫ2.229 2.31E ϩ 00 Ϫ0.743 2,3,5-Trimethylphenol anion 235MP 4.00 Ϯ 0.66 Ϫ1.345 2.00E ϩ 00 Ϫ0.673 2-Hydroxymethylphenol anion 2HMP 0.22 Ϯ 0.04 Ϫ0.954 1.12E Ϫ 01 Ϫ0.477 3-Hydroxymethylphenol anion 3HMP 0.23 Ϯ 0.07 Ϫ1.688 7.78E Ϫ 02 Ϫ0.563 Phenol anion P 0.15 Ϯ 0.03 Ϫ1.445 5.13E Ϫ 02 Ϫ0.482 Resorcinol anion R 423 Ϯ 106 Ϫ2.118 1.41E ϩ 02 Ϫ0.706 Resorcinol dianion RD Ϫ2.496 Ϫ0.832 5-Methylresorcinol anion 5MR 6907 Ϯ 1632 Ϫ2.457 2.30E ϩ 03 Ϫ0.819 5-Methylresorcinol dianion 5MRD Ϫ2.885 Ϫ0.962 3-Methoxyphenol anion 3OMP 16.77 Ϯ 1.20 Ϫ2.091 5.59E ϩ 00 Ϫ0.697 Phloroglucinol anion Ph 18,369 Ϯ 2655 Ϫ2.827 6.12E ϩ 03 Ϫ0.942 Phloroglucinol dianion PhD Ϫ3.141 Ϫ1.047 5-Methoxyresorcinol anion 5OMR 3069 Ϯ 670 Ϫ2.694 1.02E ϩ 03 Ϫ0.898 5-Methoxyresorcinol dianion 5OMRD Ϫ3.048 Ϫ1.016 3,5-Dimethoxyphenol anion 35OMP 1668 Ϯ 286 Ϫ2.746 5.56E ϩ 02 Ϫ0.915

EXPERIMENTAL acetonitrile in 30 min were used for the reaction of phenol and the reactions of methylphenols and me- Reaction of phenols with formaldehyde thoxyphenols, respectively. The eluants were detected Reactions of the phenolic compounds with formalde- by UV absorbance at 273 nm. The amount of each hyde were conducted in a three-necked flask fitted phenolic compound present was calculated using cal- with a condenser and a thermometer. The reactions ibration curves describing the relation between the were conducted at 30°C. The phenols, 2 mmol, and concentration and peak area for that compound. Each formaldehyde, 2 mmol, were dissolved in a 20% aque- phenolic compound formed during the reaction was ous dimethylformamide (DMF) solution with stirring. assumed to be characterized by the same relative re- A sufficient 10% sodium hydroxide solution was sponse as that of the starting phenolic compound. added such that the pH equaled the pKa of the phenolic compound. The solvent volume was adjusted to give a solids concentration of 1%. Determining formaldehyde by the hydroxylamine hydrochloride method Analysis of the reaction mixtures by high- performance liquid chromatography (HPLC) The concentration of formaldehyde remaining in the reaction mixture was determined by the hydroxyl- Samples of the reaction mixture were taken at various 8 times after the reaction was initiated. The amount of amine hydrochloride method. One milliliter of the the phenolic compound in the sample was analyzed sample was taken from the reaction mixture and using a Hewlett-Packard 1050 series chromatograph poured into a weighing jar containing 3 mL of a 0.02N containing an Inertsil ODS-3 column (25 ϫ 0.46 cm). hydrochloric acid solution. The solution was adjusted The mobile phase consisted of acetonitrile/0.01% to pH 4 with a 0.02N sodium hydroxide solution, and aqueous trifluoroacetic acid (TFA). An elution gradi- 3mLofa0.5N hydroxylamine hydrochloride solution ent of 5–45% acetonitrile in 30 min was used for the was added to form hydrochloride. After stirring the analyses of the products of the reactions of formalde- solution for 10 min, the sample was then back-titrated hyde with phloroglucinol, resorcinol, 5-methylresor- to pH 4 with a 0.02N sodium hydroxide solution using cinol, and 5-methoxyresorcinol; corresponding gradi- an autotitrator FMS-201 (Fluid Management Systems, ents of 10–45% acetonitrile in 25 min and 30% to 60% Inc.). REACTIVITY OF PHENOLIC COMPOUNDS. II 137

TABLE II is generally higher than is the corresponding pKa in pK Values and Dissociation Constants 12 a water. The pKa values for phenol measured in water of the Phenolic Compounds and a 20% DMF solution were 9.98 and 10.31, respec- pK Phenolic a tively. These results indicate that DMF has only a 12 compound Literature This study Ka small effect on the extent of ionization of phenol and, Ϫ by inference, is presumed to have a minimal effect on Phenol 10.0 10.3 4.9E 11 the extent of ionization of the other phenols. Resorcinol 9.4 9.7 1.9E Ϫ 10 Phloroglucinol 9.2 6.0E Ϫ 10 The base-catalyzed hydroxymethylation of phenol 2MP 10.2 10.6 2.5E Ϫ 11 by formaldehyde in a dilute aqueous solution is gen- 3MP 10.4 4.3E Ϫ 11 erally considered a second-order reaction.14 Thus, the Ϫ 4MP 10.2 10.6 2.5E 11 reaction follows the general rate expression 23MP 10.9 1.1E Ϫ 11 Ϫ 24MP 10.9 1.1E 11 ϭϪ ͓ ͔͓ ͔ 25MP 10.7 2.0E Ϫ 11 dP/dt k P F (1) 26MP 10.2 10.9 1.1E Ϫ 11 34MP 10.7 2.0E Ϫ 11 where P is the concentration of the phenolic com- 35MP 10.5 3.0E Ϫ 11 F Ϫ pound, and , the concentration of formaldehyde at 235MP 11.1 7.9E 12 any time t. Equation (1) can be rearranged as shown: 2HMP 10.1 7.9E Ϫ 11 3HMP 10.1 7.9E Ϫ 11 5MR 9.9 1.2E Ϫ 10 dP/͓P͔͓F͔ ϭϪkdt (2) 30MP 9.9 1.2E Ϫ 10 50MR 9.4 4.3E Ϫ 10 Ϫ By plotting the data in the form of the left-hand side of 35OMP 9.6 2.5E 10 eq. (2) versus dt, one obtains a linear plot with slope k. This method was used to determine the rate constants (k) for reactions of phenolic compounds with formal- Computational method dehyde. It should be noted that the rate constant de- Optimized starting structures for each of the phenolic termined in this fashion represents the sum of the rate compounds were obtained using HyperChem9 at constants for each of the reactive sites on the aromatic

RHF/PM3. The structures were further optimized at ring. The average rate constant per reactive site (kave) the RHF/6-31ϩG level of theory using Gaussian 98.10 was calculated by dividing the value of k by the num- The ﬁnal, optimized structures were used to calculate ber of reactive sites. This procedure gives an estimate ChelpG (ref. 11)-based charges obtained using Gauss- of the reactivity of the phenolic compound with form- ian 98. aldehyde on an individual reactive-site basis. Values

of kave for each of the phenolic compounds used in this study are shown in Table I. RESULTS AND DISCUSSION In terms of frontier molecular orbital theory,13 the The reaction between phenol and formaldehyde in an reaction between phenol and formaldehyde is envi- alkaline solution leads to the introduction of a hy- sioned as involving the highest occupied molecular droxymethyl group onto the aromatic nucleus at po- orbital (HOMO) of the phenoxyanion and the lowest sitions ortho and para to the phenolic group. The reac- unoccupied molecular orbital (LUMO) of formalde- tion corresponds to an electrophilic aromatic substitu- hyde. The HOMO of the phenoxyanion is distributed tion. Under alkaline conditions, phenol forms a at the o- and p-positions of the aromatic ring. The phenoxyanion. A phenoxyanion is generally consid- LUMO of formaldehyde is located on the carbon atom. ered to be the phenolic form that reacts with electro- In addition, chemical computational methods predict philic compounds like formaldehyde. Consequently, that the carbon atom of formaldehyde bears a positive the reactions of the phenolic compounds (Table I) with charge, while the o- and p-positions of the aromatic formaldehyde were conducted at conditions such that ring bear negative charges. Because there are no ap- each phenolic compound studied was characterized parent steric factors to prevent a reaction, one predicts by the same concentration of its corresponding anion, that formaldehyde will react with phenol at the o- and that is, at a pH equivalent to the pKa of that individual p-positions, as was observed experimentally. Thus, the phenolic compound. electrons in the HOMO orbital of the phenoxyanion An aqueous solution containing 20 % (w/w) DMF are in effect shared with the LUMO orbital of formal- was needed for completely dissolving the methylphe- dehyde to form a bond, giving the o- and p-hydroxym- nols and phloroglucinol. Hence, this solvent system ethylphenol derivatives. was used in all our studies. The pKa values of the The reaction of formaldehyde with other phenolic phenolic compounds determined in the indicated compounds can be envisioned as taking place in a aqueous DMF solution are shown in Table II. The pKa similar fashion. By establishing a correlation between determined in a solution containing an organic solvent the average reactivity (kave) for each compound in a 138 MITSUNAGA, CONNER, AND HILL

Figure 1 Semilog plot of the average rate constant versus average CHelpG charge calculated at the RHF/6-31ϩG level of theory for the reactions of formaldehyde with various phenolic compounds. The compounds containing two and three phenolic groups are assumed to exist in solution as monoanions. series of phenolic compounds with their respective expects the positively charged carbon of formalde- calculated average-charge per reactive site (qave), a hyde to be capable of reacting at these three sites. method can be devised to predict the reactivity of However, because a methyl group is attached at C2, formaldehyde with a wide variety of phenolic com- access of formaldehyde to this site is more sterically pounds. hindered than is its access to the C4 and C6 positions. The general method used to determine the number As a result, it is assumed that reaction of 25MP is of reactive sites, to estimate the average charge per possible only at two reactive sites, namely, at C4 and reactive site, and to account for gross steric hinderance C6. The average charge per reactive site (qave) for can be illustrated from calculations at RHF/6-31ϩG 25MP can be calculated by summing the charges at the on the 2,5-dimethylphenol anion (25MP) using the unsubstituted carbons on which the HOMO orbital is CHelpG method for calculation of the charges. The located and dividing by the number of reactive sites. total of all the atomic charges equals Ϫ1, as required In the present case, the average charge is thus for an anion bearing a net charge equal to that of a [(Ϫ0.482) ϩ (Ϫ0.654)]/2 ϭϪ0.568. single electron. Large negative charges are located at A plot of the log of kave versus the average charge on Ϫ Ϫ Ϫ the C2 ( 0.209), C4 ( 0.482), and C6 ( 0.654) posi- a reactive site qave is shown in Figure 1. The correlation tions of the aromatic ring. The HOMO orbital is also coefﬁcient (R2) for this plot is 0.82. In light of the fact distributed on these same sites. Consequently, one that values of kave cover a range of six orders of REACTIVITY OF PHENOLIC COMPOUNDS. II 139

Figure 2 Semilog plot of the average rate constant versus average CHelpG charge calculated at the RHF/6-31ϩG level of theory for the reactions of formaldehyde with various phenolic compounds. The compounds containing two and three phenolic groups are assumed to exist in solution as dianions.

Ϫ magnitude (10 2 to 104), this represents a good corre- correlation that was observed indicates the predomi- lation between kave and qave. nate role that charge considerations play in determin- The average charges used in the construction of ing the reactivity of formaldehyde with phenolic com- Figure 1 were based on the assumption that all the pounds. In addition, the method indicated above can phenolic compounds existed in solution as monoan- be used to predict the relative reactivities of phenolic ions. However, in the case of the di- and triphenolic compounds with a variety of different chemical struc- compounds, one should also consider that dianionic tures. Thus, the rates of reaction of resorcinol and forms of these compounds probably also exist under phloroglucinol relative to phenol are on the order of the reaction conditions used in this study. An alterna- 4000 and 180,000, respectively. These relative reactiv- tive plot of the data in the form of kave versus qave ities are consistent with the experimental observations (calculated using average charge values for the dian- that a phenol–formaldehdye resin must be cured at ions of the di- and triphenols) is shown in Figure 2. elevated temperatures while a resorcinol––formalde- This approach results in a slightly stronger correlation hyde resin cures at room temperature and that resins (R2 ϭ 0.90) between the experimental data and the made from compounds containing a phloroglucinolic calculated charges. substructure (e.g., pine tannins) cure almost instanta- These correlations of the experimental data with the neously. calculated value are not perfect because several other factors can also affect reactivity (e.g., steric hindrance). CONCLUSIONS These other factors have not been adequately ac- 1. Rate constants for the reactions of a variety of counted for in the present study. However, the strong phenolic compounds with formaldehyde in an 140 MITSUNAGA, CONNER, AND HILL

aqueous solution under alkaline conditions 6. Sprung, M. M. J Am Chem Soc 1941, 63, 334–343. were obtained from experimental measure- 7. Conner, A. H. J Appl Polym Sci 2000, 78, 355. ments. These data represent the largest avail- 8. Meyer, B. Urea–Formaldehyde Resins; Addison-Wesley: Read- able database for the reactivities of different ing, MA, 1979; pp 123–126. phenolic compounds with formaldehyde. 9. HyperChem 5.1; Hypercube, Inc., Gainesville, FL. 2. A strong correlation was obtained between kave 10. Frisch, M. J.; Trucks, G. W.; Schlegel, H. B.; Scuseria, G. E.; Robb, determined experimentally for the reactions of M. A.; Cheeseman, J. R.; Zakrzewski, V. G.; Montgomery, J. A., phenolic compounds with formaldehyde and Jr.; Stratmann, R. E.; Burant, J. C.; Dapprich, S.; Millam, J. M.; qave determined by calculations at the RHF/6- Daniels, A. D.; Kudin, K. N.; Strain, M. C.; Farkas, O.; Tomasi, J.; 31ϩG level using CHelpG. Barone, V.; Cossi, M.; Cammi, R.; Mennucci, B.; Pomelli, C.; 3. The best correlation between kave and qave was Adamo, C.; Clifford, S.; Ochterski, J.; Petersson, G. A.; Ayala, obtained when di- and triphenols were assumed P. Y.; Cui, Q.; Morokuma, K.; Malick, D. K.; Rabuck, A. D.; to exist in solution as their respective dianions. Raghavachari, K.; Foresman, J. B.; Cioslowski, J.; Ortiz, J. V.; Baboul, A. G.; Stefanov, B. B.; Liu, G.; Liashenko, A.; Piskorz, P.; REFERENCES Komaromi, I.; Gomperts, R.; Martin, R. L.; Fox, D. J.; Keith, T.; Al-Laham, M. A.; Peng, C. Y.; Nanayakkara, A.; Gonzalez, C.; 1. Aldersley, J. W.; Hope, P. Angew Makromol Chem 1972, 24, Challacombe, M.; Gill, P. M. W.; Johnson, B.; Chen, W.; Wong, 137–153. M. W.; Andres, J. L.; Gonzalez, C.; Head-Gordon, M.; Replogle, 2. Astarloa-Aierbe, G.; Echeverria, J. M.; Egiburu, J. L.; Ormaetxea, E. S.; Pople, J. A. Gaussian 98, Revision A.7; Gaussian, Inc.: M.; Mondragon, L. Polymer 1998, 39, 3147–3153. Pittsburgh, PA, 1998. 3. Dijkstra, R., DeJonge, J.; Lammers, M. F. Trav Chim 1962, 81, 11. Breneman, C. M.; Wiberg, K. B. J Comp Chem 1990, 11, 361–373. 285–296. 12. Abe, I.; Ono, H. Mokuzai Gakkaishi 1980, 26, 686–692. 4. Grenier-Loustalot, M. F.; Larroque, S.; Grenier, P.; Leca, J.-P.; 13. Clark, T. A Handbook of Computational Chemistry; Wiley: Bedel, D. Polymer 1994, 35, 3046–3054. New York, 1985; pp 133–138. 5. Grenier-Loustalot, M. F.; Larroque, S.; Grande, D.; Grenier, P. 14. Zavitsas, A. A.; Beaulieu, R. D.; Leblanc, J. R. J Polym Sci Part Polymer 1996, 37, 1363–1369. A-1 1968, 6, 2533–2540.