1、Computational Study of Hydrogen Bond Interactions in WaterClusterOrganic Molecule ComplexesEduardo Romero-Montalvo and Gino A. DiLabio*Cite This: J. Phys. Chem. A 2021, 125, 33693377Read OnlineACCESSMetrics & MoreArticle Recommendations*sSupporting InformationABSTRACT: We analyzed the interactions p
2、resent in complexes that acetone,azomethane, dimethylamine, dimethyl ether, methyl acetate, and oxirane form with 39different (H2O)nclusters (n = 110). A random generation of configurations and asubsequent screening procedure were employed to sample representative interactions.Using quantum chemical
3、 computations, we calculated the associated binding energies,which range from 0.19 to 10.76 kcal/mol at the DLPNO-CCSD(T)/CBS level. Itwas found that the binding energies can be understood in terms of various factors,including the water cluster size, the nature of the organic molecule, and the type
4、ofhydrogen bond donor. We find that the most stable complexes often arise from acombination of a strong hydrogen bond plus a secondary interaction between theorganic molecule and the water cluster.1. INTRODUCTIONWater can be used as a reaction medium in organic chemistry.Diels and Alder mixed furan
5、and maleic anhydride in hotwater, obtaining a cycloaddition adduct with an increasedendo-selectivity compared to their organic solvent counter-part.1However, it was not until the 1980s that a systematic andpioneering series of papers by Breslow and co-workersestablished the paradigm of water as usef
6、ul and relevant inorganic chemistry. In their research, it was demonstrated thatH2O as a solvent increases the reaction rate of DielsAlderreactions 700-fold, compared to reactions in organic solvents,mainly due to hydrophobic effects.25Breslows seminalcontributions set the foundations of aqueous org
7、anic chemistry(AOC) as a new field of research that is highly active today.68One of the key areas of activity is developing chemicalprocedures that reduce organic solvent usage in favor ofH2O;9,10the nontoxicity, reusability, and low cost of waterperfectly align with the principles of green chemistr
8、y.11It isexpected that further expanding research in AOC will lead tosignificant discoveries in chemistry.Understanding hydrogen bonding (HB) interactions be-tween organic molecules and water molecules is central tounravel the poorly understood mechanisms of AOC. Forexample, it has been proposed tha
9、t a group of acceleratedreactions in aqueous emulsions, called “on water” reactions,occurs due to HB stabilization of transition states relative toreactants.12Also, countless chemical processes relevant tobiology involve water clusterorganic substrate interac-tions.1319In light of the importance of
10、such interactions,the present work focuses on their accurate description from thequantum mechanical perspective.HB interactions within water clusters have been extensivelystudied over the past decades. It is a well-known fact thatcooperative effects2024are prominent in these interac-tions,2529and th
11、at they increase with the number of watermolecules in the cluster. In addition to the cluster size, thelocal environment of hydrogen bonds is relevant; the nature ofthe HB donor and acceptor and their neighboring moleculeshas been used to describe noncovalent interactions withinwater clusters of sma
12、ll size.30,31However, to the best of ourknowledge, a comprehensive work on organic moleculeinteractions with water clusters of distinct sizes has not yetbeen explored. Some computational studies have shown thatdensity functional theory (DFT) can reproduce experimentalresults in AOC reactions.12,32,3
13、3However, most quantumchemistry models utilize only a few water molecules torepresent the complex traits of HB in these systems,completely neglecting the role of water cluster size inaccurately depicting hydrogen bonding interactions withorganic molecules. For these reasons, in this work, we explore
14、the interaction strengths of water clusters of varying sizes withorganic HB acceptors. In the process of our study, wedeveloped a new benchmark data set of hydrogen bonding fororganic moleculewater cluster complexes.Received:February 14, 2021Revised:March 29, 2021Published: April 20, 2021Articlepubs
15、.acs.org/JPCA 2021 American Chemical Society3369https:/doi.org/10.1021/acs.jpca.1c01377J. Phys. Chem. A 2021, 125, 33693377Downloaded via UNIV OF PRINCE EDWARD ISLAND on May 15, 2021 at 17:02:25 (UTC).See https:/pubs.acs.org/sharingguidelines for options on how to legitimately share published articl
16、es.2. COMPUTATIONAL DETAILSAn accurate quantum chemical description of water systemscan only be achieved for small water clusters.34,35Theselimitations demand that we approximate our molecular systemswith small models. Despite their reduced size, model systemscan often produce insightful information
17、 for a variety ofchemical problems. For our study, we wanted to understandthe behavior of binding energies between organic moleculesand water clusters of increasing size. For this purpose, weretrieved the water clusters produced by Temelso and co-workers36that range in size from the monomer to the d
18、ecamer.A summary of the structures studied in this paper is shown inTable 1, and an example of them is displayed in Figure 1.Temelso et al. obtained these structures at the RI-MP237,38/aug-cc-pVDZ39level of theory, making them appropriate forbenchmarking purposes. We also selected the following seve
19、norganic molecules as representative of species that are presentin AOC reactions: acetone, dimethyl ether, oxirane, methylacetate, azomethane, dimethylamine, and trimethylamine,which we preoptimized at B3LYP4042/6-31+G* usingGaussian 16 (g16).43This selection was based on theirsimilarity to organic
20、substrates involved in on waterreactions.44We used these organic molecules as probes tostudy the HB donating capabilities of OH groups in smallwater clusters. For the sake of simplicity, we will often refer tothese organic molecules as probes in this work.The complexes formed by the interactions bet
21、ween theprobes and water clusters of various sizes could take on anyone of a large number of configurations: as water cluster sizesincrease, so does the number of available OH donor groupswith which the acceptors can interact. To explore differentcomplexes, we fixed the coordinates of the water clus
22、ters andallowed the preoptimized probes to freely move around thewater clusters using a random generator of coordinates toobtain 2000 structures per water clusterprobe complex,followed by an approach with intermediate computationalexpense steps to refine the data set. The module genmer fromthe Molcl
23、us software45was employed to generate the 2000configurations for each complex. Next, we performed single-point (SP) calculations on these geometries and energy rankedthem using the universal force field (UFF) force-field46andeliminated the 1000 least stable complexes (step I). In step II,we computed
24、 the single-point energy of the remaining 1000compounds using HF-D3(BJ)/MINIs-ACP and reordering thestructures based on electronic energy. The 500 least stablestructures were removed from the list. Step III involvedcalculating the single-point energies of the remaining 500complexes using B3LYP-D3(BJ
25、)/6-31+G(d,p)-BSIP, reorder-ing the structures by energy, and eliminating the 250 leaststable structures. Step II employs atom-centered potentials(ACPs) that are designed to mitigate the effects of incompletecorrelation and incomplete basis sets associated with the HF-D3(BJ)/MINIs approach.47Step II
26、I uses a form of ACPscalled basis set incompleteness potentials (BSIPs) designed toreduce the effects of incomplete basis sets used withconventional density functional theory (DFT)-based meth-ods.48Steps IIII employed g16, and the process issummarized in Figure 2.Following the foregoing screening, a
27、 set of 250 config-urations per water clusterprobe complex was obtained. Theseenergy-ranked configurations were divided into 10 binscontaining 25 conformers each. These groups of 25 complexgeometries were also energy ranked; bin 1 contains the 25most stable systems, bin 2 the following 25 most stabl
28、e ones,and so on. This approach led us to sample meaningfulconfigurations involving not only strong HBs but also weakerTable 1. Names of the Water Clusters Considered in ThisWorkaclustersize, nlabels1monomer22Cs33UUD, 3UUU44Ci, 4PY, 4S455CA-A, 5CA-B, 5CA-C, 5CYC, 5FR-A, 5FR-B, 5FR-C66BAG, 6BK-1, 6BK
29、-2, 6CA, 6CB-1, 6CB-2, 6CC, 6PR77BI1, 7BI2, 7CA1, 7CA2, 7CH1, 7CH2, 7CH3, 7HM1, 7PR1,7PR2, 7PR388D2d, 8S499D2dDD, 9S4DA1010PP1, 10PP2aThe structures of all oligomers of water shown here can be seen inFigures 1 and 2 of ref 36.Figure 1. Hexamer (6CC) of water. Color codes: red: oxygen andblue: hydrog
30、en. The dashed lines indicate HB interactions.Figure 2. Summary of the screening process followed in selecting themost stable complexes analyzed in this work. We applied thisprocedure to each of the possible 273 complexes (7 probes interactingwith 39 water clusters).The Journal of Physical Chemistry
31、 Apubs.acs.org/JPCAArticlehttps:/doi.org/10.1021/acs.jpca.1c01377J. Phys. Chem. A 2021, 125, 336933773370interactions. We then selected the most stable complex fromeach bin to get 10 structures per complex and reoptimized thegeometries of the probes within the complexes at B3LYP-D3(BJ)4042,4952/6-31
32、+G*-BSIP48using g16; in all cases,the structures of the water clusters were kept fixed at theTemelso geometries.36This approach has the advantage ofusing high-level geometries; however, since they are fixedstructures, the donating OH group in the water clusters willnot relax and, hence, the binding
33、energies we calculate will allbe upper bounds. The resulting structures constitute the set ofcomplexes for our benchmark calculations. We obtained 2730complex structures in total.From our 2730 geometries, we removed the redundantgeometries that arose from configurations that converged toidentical ge
34、ometries when reoptimizing the probes in thecomplexes. For doing so, the COMPARE feature, asimplemented in the program Critic2,53,54was employed. Thefinal data set contains 2376 nonredundant geometries. Ourscreening method ensured that strong interactions from thedangling OH to the probes were prese
35、nt and other weakerinteractions such as those involving CH HB donors anddispersion interactions. The molecular systems were visualizedusing the software Chemcraft.55Reference binding energy calculations were calculated usingDLPNO-CCSD(T)56with aug-cc-pVDZ and aug-cc-pVTZbasis sets39(represented as D
36、Z and TZ, respectively). Astrategy similar to the one presented in ref 57 was applied forthe extrapolation to the complete basis set (CBS) limit: wecalculated the counterpoise (CP) corrected binding energies(BECP)58,59of the complexes asEEEEEBE(complex)(H O) )(R)(R)(R)nCPcomplexcomplexcomplexcomplex
37、2complexcomplexcomplexRRR=+(1)where the general nomenclature, Egeometrybasis(X), represents theelectronic energy of species X (R for probe, (H2O)nfor a watercluster of size n, and complex for the water clusterprobeaggregates), at the geometry specified by the subscript andusing the basis set identif
38、ied by the superscript. For example,the term EcomplexR(R) represents the electronic energy of a probeat its geometry in a complex using the probe basis sets. Notethat eq 1 can be written alternatively asEEBE(complex)(R)CPintCPdefR=+(2)inwhichEintCP(complex)=Ecomplexcomplex(complex)Ecomplexcomplex(H2
39、O)n) Ecomplexcomplex(R) and the deformation energyof the probe, EdefR(R), equals EcomplexR(R) ERR(R). Thedeformation energies of the water clusters (Ecomplex(H2O)n(H2O)n) E(H2O)n(H2O)n(H2O)n) are zero because we are using frozengeometries for those structures throughout. The noncoun-terpoise (non-CP
40、) binding energies are defined byEEEBE(complex)(H O) )(R)nnonCPcomplexcomplex(H O)(H O)2RRnn22= +(3)The ORCA60package was used for the DLPNO-CCSD(T)calculations. Eqs 2 and 3 were evaluated for both basis sets, DZand TZ. A two-point extrapolation (DZTZ) to the CBS limit,based on the El= A + B/(l + 1/
41、2)4formula, was applied.61Inthis equation, A represents the extrapolated energy, B is aconstant, and l = 2 and 3 for DZ and TZ, respectively. We alsocomputed the average of the CP and non-CP approaches asBEavei= (BECPi+ BEnonCPi)/2, with i = DZ, TZ, and DZTZ.We defined our benchmark binding energies
42、 as the average ofthe CP and non-CP energies at the CBS extrapolation limit (i= DZTZ). Previous research has shown that CP BEs tend toconverge to the complete basis set limit from above, while thenon-CP BEs converge from below.21,57Therefore, averagingthe two quantities generally results in quicker
43、convergence tothe CBS limit. These average binding energies show a quickconvergence to the CBS limit for most noncovalentinteractions studied herein. A selected example of theconvergence of BEaveiis depicted in Figure 3; since BECPandBEnonCPconverge to the same value at the CBS limit, BEaveiwillmore
44、 rapidly converge to the complete basis set limit than theindividual counterpoise and noncounterpoise binding energies.However, the convergence of the average of the CP and non-CP BEs is not always ideal, i.e., when the CP and non-CP BEsdo not converge from above or below, respectively. Figure 4show
45、s one of those examples in which the binding energies of acomplex formed between azomethane and the hexamer 6CCare displayed. For this example, the non-CP binding energiesshow little variation with the basis set. Nevertheless, more than97% of the 2376 BEs we calculated show better convergence ofthe
46、averaged BEs than the CP and non-CP BEs. The completeset of graphs showing the CBS extrapolation limit for allcomplexes is included in the Supporting Information (SI).With our benchmark binding energy calculations in hand, weassessed the performance of a cross-section of densityfunctional theory met
47、hods6270and plotted the results inFigure 5. This graph shows the mean absolute error (MAE) foreach DFT approach; B3LYP-D3(BJ)/6-31+G*-BSIP, whichwe used as part of our procedure to generate the complexstructures, has a 0.44 kcal/mol MAE value, which suggests thatthe method was suitable for the geome
48、try optimization step ofour calculations. We also note that BLYP-D3(BJ)/6-31+G*-BSIP and M05/6-31+G* reproduce the binding energiesmost closely among all of the DFT approaches assessed(MAEs of 0.25 and 0.26 kcal/mol, respectively).Figure 3. Two-point extrapolation to the CBS extrapolation limit foro
49、ne of the binding energies between the decamer 10PP2 and methylacetate. Our benchmark binding energies are the average of the CPand non-CP binding energies at the CBS extrapolation limit (DZTZ). Color codes: red: oxygen, blue: hydrogen, and magenta: carbon.The dashed lines indicate HB interactions.T
50、he Journal of Physical Chemistry Apubs.acs.org/JPCAArticlehttps:/doi.org/10.1021/acs.jpca.1c01377J. Phys. Chem. A 2021, 125, 3369337733713. RESULTS AND DISCUSSIONA diverse set of water clusterprobe configurations wasobtained from the process described in the previous section.Figure 6 shows all bench
