Rationalization of the Color Properties of Fluorescein in the Solid State: A Combined Computational and Experimental Study

Abstract Fluorescein is known to exist in three tautomeric forms defined as quinoid, zwitterionic, and lactoid. In the solid state, the quinoid and zwitterionic forms give rise to red and yellow materials, respectively. The lactoid form has not been crystallized pure, although its cocrystal and solvate forms exhibit colors ranging from yellow to green. An explanation for the observed colors of the crystals is found using a combination of UV/Vis spectroscopy and plane‐wave DFT calculations. The role of cocrystal coformers in modifying crystal color is also established. Several new crystal structures are determined using a combination of X‐ray and electron diffraction, solid‐state NMR spectroscopy, and crystal structure prediction (CSP). The protocol presented herein may be used to predict color properties of materials prior to their synthesis.


Introduction
Organicp igments are of great importance to society. They are used in the production of photovoltaic materials, [1] optical data storaged evices, [2] anda lso in the coloration of plastics [3] and in the textile industry.W ith the current rapid population growth there is an ever-increasing demandf or pigmentst ailoredf or very specific applications. [4] The ability to design new pigment materials with desirable properties, however,r equires ad eep theoretical understanding of the underlying physical processes that involvet he interaction of light with crystalline materials. The development of computational methods, particularly plane-wave electronic structure calculations, [5] coupled with constantly increasing available computer power have allowed researchers to studys olid-state phenomenar esponsible for av ariety of applications,p articularly in the areas of semiconductingm aterials, [6,7] organic photovoltaics, [8] solid electrolytes, [9] lithium-ion batteries, [10] magnetic materials, [11] surface chemistry,and catalysis. [12] The situation with pigments,however,i sd ifferent:m any commerciallyi mportantm aterials are used withouta ny detailed understanding of the structuralf eatures responsible for their optical properties. Determination of the crystal structure of apigment would be the first step in understanding the behavior of the material; once the structurei s known, band structure calculations can be performed to shed light on the opticalp roperties of the material. Despite the recenta dvancementsi nt he methodso fc rystal structure determinationf rom powder X-ray diffraction (PXRD)a nd crystal structure prediction (CSP), [13] the structures of many industrially important pigmentsr emain unknown or are determined decadesa fter the beginning of industrial production. [14,15] The lack of computational studiesm akes the development of new pigment materials largely at rial and error process, at rend that must be changed to fulfil society's needs for new advanced optical materials.
Current developments in periodic DFT band structure calculationsh ave enabled the modeling of optical absorption and reflection spectra,a sw ell as establishing the electronic density of states (DOS)o fm aterials. [16] The ability to perform such calculations not only provides the opportunity to understand the optical properties of existingp igments( and other materials), but also to predict optical properties of new materials before beingp repared.
We describe hereinh ow recent developments in both experimentala nd computational methods can be utilized to characterize the optical solid-statep roperties of the model pigment fluorescein. Thes tudy servesa sa ne xample of how computationals tudies supplement experiments in providing the most detailedu nderstanding of the complexo ptical solid-state properties of pigments, as well as providing ag eneral strategy for the computational development of new organic pigments and other materials.
The optical properties of organic pigments dependn ot only on the moleculars tructure of the compound but also on the associated crystal packing:d ifferent polymorphs of the same compound often show considerable variation in color as aresult of differences in molecular conformation or intermolecular interactions. [17,18] An archetypal example of such as ystem is the compound ROY,w hich produces aw ide variety of polymorphsw ith colors ranging from red to orange and yellow. [18,19] Even greaterd iversity may be achieved by producing multicomponent cocrystals in which coformer molecules electronically interactw ith the organic chromophore, [20] often causing large changes in the band gap and, therefore, color. [21] Furthermore, the use of cocrystals in opticalm aterials is not limited to the modification of color:r ecent work in our group has shown that cocrystallization is also as uccessful methodf or modifyingt he luminescent properties of organic compounds. [22,23] During studies of the solid-state behavior of organic chromophores, our attention was drawn to the pigmentf luorescein. This compound is mostw idely used in the water-soluble disodium-salt form. Therefore, the solid-state behavior of the neutralc ompound has not received equivalent attention and the availablei nformation is often ambiguous and incomplete. [24,25] Nonetheless, the behavior of crystalline fluorescein represents ac urious case of an interplay of molecular tautomerism [26] and the effects of crystal packing.
The fluorescein (fls)m olecule exists in three tautomeric forms ( Figure 1): the quinoid form (flsQ), the zwitterionic form (flsZ), and the lactoid form (flsL). In the solids tate, flsQ is reported to form ar ed powder, [24] for which the crystal structure has been determined using PXRD. [27] The flsZ tautomer,o nt he other hand, produces ay ellow solid, for which the crystal structure has not been reported. The third tautomer, flsL,h as not been crystallizedi np ure form, although the crystals of this hypothetical solid are expected to be colorless by analogy to ar elatedl actoid compound diacetylfluorescein ( Figure 2). [28] Despite the inability to obtain crystals of pure flsL,s everal structures containing the lactoid tautomer in the form of acetone, [29] methanol, [30] and 1,4-dioxane [25] solvates have been produced. Descriptions of the colors of thesesolvates in the literature, however,a re rather ambiguous as different authors characterize them as colorless, [24] amber-yellow, [29] or even orange. [30] It was also suggested [30] that the yellow coloration of the flsL solvate crystals may be causedb yp artial loss of solvent molecules from the crystal surface. The resulting "free" flsL molecules then convert into the yellow flsZ form, thus generating ac olored layer on the surface of the crystals. Our own attempts to produce the solvates of fluorescein have shownt hat the color of these materials depends on the particle size:p owdered materials display significantly brighter colors thanl arger crystals( Figure3), suggestingt hat the color is indeed formed within as urface layer.H erein we present ac ombined experimental and computational investigation that confirms the existence of the zwitterion surface layer and offers possible reasons for its formation.
Recently,o ur group hasr eported three new cocrystals of flsL. [31] These cocrystalsw ere prepared using mechanochemical methodsa nd displayed colors ranging from yellow to green. It was considered instructive to establish whether the mechanism of color generation in the cocrystals of flsL is the same as in the solvates (namely,t hrough the formation of the flsZ surface layer), or whether the presence of another molecule (i.e. cocrystal former) in the crystal lattice plays an active role in the optical properties of the resulting multicomponent form.
In this study,w eu se av ariety of experimental andt heoretical techniques to obtain acomplete understanding of the optical behavior of fluorescein in the solid state. The previously unreported crystal structures of pure flsZ alongw ith dioxane solvates of flsL are determined using ac ombination of X-ray diffraction, electron diffraction, [32][33][34][35] and CSP.T he identity of the three tautomeric forms of fluorescein is further established   www.chemeurj.org through 13 Cs olid-state NMR spectroscopy and calculations of NMR chemical shielding. Finally,t he existence of the zwitterion surfacel ayer on the flsL-containing crystalsi sc onfirmed by UV/Vis spectroscopy,a nd insight into the mechanism of its formation is given. Additionally,t he role of the coformers in the opticalp roperties of fluorescein cocrystals is established with the aid of band structure calculations. Most importantly,h owever,t his study will demonstrate how the synergyo f( standard and state-of-the-art) experimental andc omputational techniques can be used to obtain the best understanding of the properties of pigment materials. We believe that the protocol demonstrated herein not only provides ab etter understanding of color generation in pigments, but also demonstrates that it is possible to engineer new materials with desirable properties.

Results and Discussion
To perform the calculations of optical properties of materials it is first necessary to determine the corresponding crystal structures. The structures of diacetylfluorescein, [28] flsQ, [27] flsL acetone solvate, [29] and cocrystals with acridine, phenanthridine, and pyrazine [31] have been previously reported. The preparation of unsolvated flsZ and flsL:dioxane hemisolvate have been reported withouts tructure determination, [24] while the flsL:dioxane hemipentasolvate has not previously been reported.
Crystal structure determination of new fls crystal forms flsL:dioxane hemipentasolvate The flsL:dioxane hemipentasolvate was crystallized from solution andt he structure was determined using single-crystal Xray diffraction.T his material crystallizes in the monoclinic space group P2 1 /n having one molecule of flsL and2 .5 molecules of 1,4-dioxane in the asymmetric unit. The potential hydrogen bondingg roups of dioxane molecules in this crystal structure are rather poorly utilized:t wo of the dioxane molecules each use one of their oxygen atoms to form OÀH···O hydrogen bonds with hydroxy groups of flsL.T he third dioxane molecule, whichi sl ocated on an inversion center, only interacts through weak CÀH···O interactions with the other dioxane molecules (Figure 4c). It is, therefore, not surprising that the fluoresceind ioxane hemipentasolvate converts to the dioxane hemisolvate within several days, even at room temperature.

flsL:dioxane hemisolvate
The flsL:dioxane hemisolvate was obtained by heating the hemipentasolvate at 80 8Cf or 15 min. The desolvationp rocess is accompanied by significant structural rearrangements that cause the crystalst os hatter.A saconsequence, structure determination had to be performed using powderX -ray diffraction (ford etails, see Section 4i nt he Supporting Information).
The flsL:dioxane hemisolvate crystallizes in the triclinic unit cell, space group P1.T he asymmetric unit contains one flsL molecule and half of ad ioxane molecule located on an inversion center.B oth oxygen atoms of the dioxanem olecule are hydrogen-bonded to flsL through OÀH···O interactions. Furthermore, fluorescein moleculesf orm centrosymmetric dimers through OÀH···O(carbonyl) interactions, which were not present in the hemipentasolvate structure ( Figure 4b). Overall, the hemisolvate structure offers am ore even balance between the number of hydrogen-bond donors and acceptors, which is reflectedi navery high temperature for complete desolvation (150 8C).

flsL:acetone monosolvate form II
The flsL:acetone monosolvate was obtained by evaporating an acetone solution of fluorescein. While initial crystallization experiments resultedi nt he formation of the previouslyr eported solvate structure (Form I), [29] later attempts to reproduce the materialm ostly resulted in the formation of an ew polymorph of the solvate (Form II).
The crystal structure of Form II was determined from X-ray powderd ata. The materialc rystallizes in the monoclinic P2 1 /c space group with one flsL and one acetone molecule in the asymmetricu nit. The principal intermolecular interactions are the OÀH···O(acetone) and OÀH···O(flsL,c arbonyl) hydrogen bonds ( Figure 4d). Careful inspection of the structure revealed that the new solvate polymorph is isostructural with the flsL:pyrazine cocrystal, whereby the acetonem olecules are replacing pyrazine in the cocrystal structure (SupportingInformation, Figure S19).
Making as lurry of Form II resulted in the conversion to Form I, suggesting that Form Ii st he thermodynamically stable polymorph. This was further supported by the solid-state DFT (PBE + G06) energy calculations, whichs howedt hat the lattice energy of Form II is 11.2 kJ mol À1 higher than that of Form I.

Crystal structure of flsZ
The zwitterionic form of fluorescein( flsZ)w as crystallized by rapidly quenching an aqueous alkaline solution of fluorescein disodiums alt with acetic acid. As ar esult of such rapid crystallization the product was obtained as av ery fine powder.T he small particles ize of the materialr esulted in broad peaks in the X-ray powder pattern, which made direct structure determinationf rom powder data difficult. To elucidate the crystal structurew eh ave studied the crystal energy landscapeo fp ossible flsZ structures using the crystal structure prediction (CSP) methods described previously. [36][37][38] The CSP calculations [39] generate as et of trial crystal structures that ag iven molecule can form;t hese are lattice energy minimized and ranked by their lattice energy. [40] Analysis of the predicted structures provides information about the most prevalent intermoleculari nteractions and supramolecular synthons [41] responsible for the formation of the crystal.The experimentals tructure may then be determined by validating the predicted structure against experimental data, such as X-ray powderp atterns, [37,39] TEM electron diffraction data, [33,34] or solid-state NMR [44,45] spectra.
X-ray powder patterns were calculated using the software package CCDC Mercury [46] for each of the low energy (within 15 kJ mol À1 of the global minimum) predicted crystal structures and the assignment of the experimental structure was performed by comparing the calculated and experimental patterns (for X-ray powder pattern comparisons, see the Supporting Information, Figures S31-S34).T his analysis revealed the structurer anked third, 6.6 kJ mol À1 above the globalm inimum in lattice energy,a st he most likely candidate fort he observed crystal structure. This structurala ssignment was subsequently validatedb ys uccessful Rietveld refinement [47] of the predicted structure against the experimental pattern. While the PXRD analysisd emonstrates that predicted structure number 3i st he major crystalline component of yellow fls,t he diffraction pattern also indicates that an amorphous component mayb e presenti nt he bulk material (Supporting Information, Figure S8).
The correctness of the crystal structure determination was furthers upported by performingT EM analysiso nt he flsZ powderp articles and determining which of the predicted structures were consistent with the resulting electron diffraction patterns.S tructure number 3w as found to be the best match to the TEM data from the low energy predicted crystal structures. Moreover, the experimentale lectron diffraction patterns were as atisfying match to patterns simulated from the CSP-PXRD-derived crystal structure of flsZ ( Figure 5). The combined assignment based on both PXRD and TEM electron diffraction data provides ah igh level of confidence in the CSP structure determination.
The principal intermoleculari nteractionsp resent in the flsZ crystal structure (and which are also present in the other lowenergy predicted structures) are charge-assisted OÀH···O hydrogen bonds between the hydroxy protons andt he carboxylate oxygen atoms of the neighboring flsZ molecules (Figure 4a).
Careful inspection of the predicted structures revealed that the lowest-energy structure of flsZ showed ac lose similarity with the experimentally reported structure of flsQ.A no verlay of the two structures ( Figure 6a)s hows av ery close alignment of heavy atom positions. The observed quinoid and predicted zwitterionic crystal structures are relatedb ys witching of the hydrogen positioni nt he COOH···O (carbonyl) intermolecular hydrogen bond (Figure 6b). In fact, an attemptt op erform aD FT geometry optimization of the experimental flsQ crystal structure using the PBE functional [48] led to ap roton shifta nd transition of the molecule into the flsZ tautomeric form. The most likely reasonf or this incorrect modeling of the hydrogen bondingi st he known feature of the semilocal DFT functionals to underestimate, or even completely suppress, the energy barriers for hydrogen bond proton transfers,e ven when dispersionc orrection is applied. [49] The incorrect description of hydrogen bonding by the PBE functional may also cause certain errors in the energy ranking of the predicted structures. The accuracyo ft he calculations could have been improved by using ah ybrid functional such as B3LYP [50] or PBE0, [50,51]   althought he cost of performing such calculations with the plane-wavebasis set would be prohibitive.

Crystal structure prediction for unsolvated flsL
Avariety of crystallization methods were appliedi na na ttempt to crystallize flsL without the presence of guest molecules in its crystal structure. All experiments, however,r esulted in the formationo fc rystalline flsQ or flsZ,o ri nt he crystallization of the solvated forms of flsL.D esolvation of flsL:acetone monosolvate Form II was shown to produce an amorphousp hase that subsequentlyc rystallizes into flsQ (Supporting Information, Figure S25). The apparent difficulty to crystallize pure flsL led us to investigate this phenomenon from at heoretical perspectiveb yp erforming crystal structure prediction to assess the possible crystal packing of pure flsL.C omparison of the crystal energy landscapes formed by flsL and flsZ tautomers revealed that the lowest energy predicted structure of flsL is 21.1 kJ mol À1 higher in lattice energy then the experimental structureo fflsZ (for completec rystal energy landscapeo f both tautomeric forms, see the Supporting Information, Figure S30). Such al arge energy difference between the two forms indicates that crystalline flsL is thermodynamically unstable with respectt oflsZ in the solid state. The calculated energy differencei sw ell outside the energetic range of observed polymorphism, where lattice energy differences are typically well under 10 kJ mol À1 , [52] explaining the difficulty of crystallizingthe lactoid tautomer.

Solid-state NMR measurements and chemical shift calculations
The methodo fs olid-state NMR hasa cquired considerable importancea sacomplementary structuralt echnique to X-ray diffraction.I th as been used to study hydrogen bonding, [53] ion mobility, [54] and static and dynamic disorder in solids. [55,56] Althoughs olid-state NMR is invaluable as ap urely experimental technique, results obtained by this method can be greatly reinforced through the use of NMR chemical shielding calculations. The latter has been greatly facilitatedb yt he development of the gaugei ncluding projector augmented waves( GIPAW) [57] method, currently implemented in severalp lane-wave packages including CASTEP. [58] GIPAW calculations can display an accuracy of 1-2 ppm for modeling 13 CNMR spectra. [44,59] Such ahigh accuracyofNMR calculations establishes it as apowerful technique for assignment of predicted structures to experimental data, [60] and makes it particularly effective for distinguishing tautomeric forms. [61] Since the hydrogen bonding in the crystal structures of flsQ and flsZ differs only in the position of the hydrogen atom, interconversion of the two tautomeric forms in the solid state may occur as ar esult of ap rotons hift across the hydrogen bond. Since crystal structures of both tautomers have been determined using PXRD,amethod that is not sensitivet od eterminationo fh ydrogen atom positions, solid-state NMR measurements were performed in order to unambiguously assign the moleculart automers to the different color forms of fluorescein.
Solid-state 13 CNMR spectra were recordedf or the samples of pure red and yellow fluorescein( assumed to contain flsQ and flsZ forms respectively)a sw ella sf or four solvates of flsL (both polymorphs of acetonem onosolvate, dioxane hemisolvate andd ioxaneh emipentasolvate). The solvate materials were readily assigned to contain the flsL tautomer based on the NMR signal at 85-88ppm, which could only correspond to the quaternary carbona tom in the lactone ring. Such ah ighfield signal cannot correspond to either flsQ or flsZ since these molecules contain only sp 2 -hybridized conjugated carbon atomsw ith signals furtherd ownfield. The spectra of flsQ and flsZ,o nt he other hand, display close similarityt hat makes the assignmento ft hese tautomers more difficult. In order to resolve this ambiguity,a nd aid the full assignment of NMR spectra, CASTEP GIPAW calculationsw ere performed on the crystal structures of the correspondingm aterials.T he calculated NMR tensors were related to the experimental chemical shifts using the linear regression procedure described in the Supporting Information.
To perform the calculations the crystal structures of red and yellow fluorescein were geometry-optimized, with hydrogen atom position constrained to represent either the flsQ or the flsZ tautomer,a nd the NMR parameters were computed for each. The calculated chemical shifts were then compared to the experimental spectra andi tw as unambiguously shown that the red and yellow forms contain flsQ and flsZ tautomers, respectively (Figure 7).
During the analysis of the 13 CNMR spectra of the flsL solvates it was noticedt hat the chemical shift of the carbonyl carbon in the dioxane hemipentasolvate (168.47 ppm) is approximately 3ppm lower than that in the dioxane hemisolvate and acetonem onosolvate (172.54 and 172.62 ppm, respectively). Inspection of the corresponding crystal structures (Figure 4b-d) suggests that the variations in chemical shifts are caused by the differences in hydrogen bonding:int he acetone monosolvate and dioxane hemisolvate the carbonyl oxygen is hydrogen-bonded to the hydroxy group of another flsL www.chemeurj.org molecule, while in the dioxane hemipentasolvate such ahydrogen bond is absent. This effect of hydrogen-bondingo nt he NMR chemical shift is very well modeled by the GIPAW calculation:t he carbonyl shifts calculatedu sing the regression equation are 167.77, 171.97, and 171.45ppm for the dioxaneh emipentasolvate, dioxane hemisolvate, and acetone monosolvate Form I, respectively.Thisresult illustrates the excellent accuracy of the GIPAW method in modeling the effect of structuralf eatures such as hydrogen bonds on the NMR parameters.

Opticalproperties of fls tautomers in the solid state
Once the agreement between the tautomers ands olid forms of fluorescein had been established, the origins of the color of these solid forms remained to be explained. Out of the three availablet automeric forms only two were isolated in the solid state in their pure form:t he red fluoresceinc ontaining molecules of flsQ and yellow fluorescein, flsZ.T he molecule of flsL, as the calculations have shown, cannotf orm at hermodynamically stable crystal structure without other guest molecules present. Nonetheless, the lactoid form produces an umber of multi-component solids including dioxane hemi-and hemipenta-solvates,a cetone and methanol monosolvates as well as cocrystalswith acridine, phenanthridine, and pyrazine.
The unifying structural feature of all the known crystal forms of flsL is that the guest molecules in these structures act as hydrogen bond acceptors interacting with the hydroxy groups of fls.T hese materials are colored yellow,w ith the exception of the pyrazine cocrystal,w hich is colored green.B yc ontrast, as imilar compound, diacetylfluorescein,c rystallizes as as inglecomponent colorless solid in its lactoid form. Therefore, apossible explanation for the generation of color in the lactoid fluorescein samples may be the electronic interactions of flsL with the coformerm olecules. Coloredc ocrystalsc onsisting of components that are colorless in their pure form have previously been reported, [20] althought hese normally contain coformers with extended conjugated systems. It is very unlikely that the presenceo fadioxaneo ra cetonem olecule in the crystal structures would lead to adramatic change in color.
Another possible explanation for the origin of color is the formationo fayellow flsZ layer on the surface of the lactoid crystals. The main argument in favor of this hypothesis is the apparent color dependence on the particle size:s amples with smaller particles ize have greaters urfacet ob ulk ratio and display brighter colors.
As af irst step in the analysiso ff luorescein optical properties, the solid-state UV/Vis spectra of all availables olid forms were recorded and the corresponding band-gaps were determined.T he value of the band gap determines the energy cutoff below which the light photons will be reflected. Therefore, knowledge of the band-gap is critical for understanding the color properties of the material. Theoretical density of states calculations were performed using the codeO ptaDOS [62,63] in parallelw ith the experimentalm easurements. The calculated band gaps are noticeably lower than the experimental values (Figure 9), which is ak nown behavioro fs emilocal DFT functionals such as PBE. Most importantly,h owever,t he calculated valuess howg ood correlation with experiment meaning that theoretical band structure can provide important insights into the opticalp roperties of fluorescein crystal forms.O verall, the band gaps of the materials increase in the order flsQ < flsZ < flsL < diacetylfluorescein (a full summary of the measured band gaps is given in the SupportingI nformation, Table S9 and FiguresS41-S50).
The band gaps of flsQ and flsZ are consistentw ith their observed colors, but this is not the case for solids containing flsL, where the majority of materials have band gaps well above the visiblel ight energy range (1.5-3.3 eV). Materials with such high band gaps are expected to be white, as is demonstrated by the white solid of diacetylfluorescein. The band gaps of both dioxanes olvates of flsL are essentially the same as that for diacetylfluorescein. However,t hese solvates display ab right yellow colori np owder form. The only material that has as ufficiently low band gap to reflect ac ertain proportion of visible light is the acridine cocrystal.
Detailed analysis of the spectra of lactoid fls solid forms revealed that they all contain aw eak yet reproducible feature corresponding to the electronic transition of 2.30 AE 0.06 eV (Figure 8b), which is very similar to the band gap of pure flsZ. This observation suggests that an impurity of zwitterionic fluoresceini sp resent in the lactoid samples. Visible-light spectroscopy,h owever,c annot distinguish between as urface layer of the zwitterion impurity and randomly distributed im- Figure 8. a) The correlation between computed and experimental band-gaps of the flsL solid forms. b) Ta uc plot [66] constructed from the reflectance spectrum of the flsL:dioxane hemisolvate. The linear regions crossingt he abscissa at 2.3 and 3.9 eV correspond to the band gaps of the zwitterion impurity and the bulk material, respectively.
Chem.E ur. J.2016, 22,10065 -10073 www.chemeurj.org purity within the bulk structure in the form of crystal defects. This ambiguity has been resolved by dissolving the surface layer of the flsL:dioxane hemipentasolvate single crystal:t he originally yellow crystal was placed in am ixture of dioxane with silicon oil. Observation of this crystal under am icroscope has shown completel oss of color ( Figure 9) thus proving that the zwitterion impurity is concentrated on the crystal surface. It is noteworthy that the yellow color is rapidly restored when the crystal is taken out of solution and exposed to air.
Another solid for which it has been possible to remove the zwitterion surface layer is the pyrazinec ocrystal of flsL.T his cocrystal is preparedi nam echanochemical process that produces the yellow-green powder.A dding an excess of pyrazine into the grinding jar, however, resulted in the formation of ag reyish powder with reduced flsZ surface content ( Figure 10). Both the green and the grey materialh ave identical crystal structures, as confirmed by PXRD,w ith the green product having significantly greatera mount of flsZ impurity at the surface (Supporting Information, Figures S46 and S47). The connectionb etween the lack of ac olored surface layer and the use of an excesso fp yrazinei nt he synthesis of the cocrystal is so far not entirelyu nderstood. The color of the material is affected by factorss uch as frequency of grinding, duration of the experiment anda mount of liquid added to the reaction mixture. Investigation of all these factorsl ies outsidet he scope of this study.Adetailed experimentals tudy is, however,u nderway and its results will be reported separately.
The combination of solid-state UV/Vis measurements with analysis of flsL:dioxane hemipentasolvate and flsL:pyrazine crystal surfaceb ehavior has allowed us to concludet hat the origin of the color of the lactoid fluorescein crystals is the yellow surface layer that consists of flsZ molecules. To fully understandt he opticalp roperties of the materials, however, the contribution of the bulk crystal structure has to be considered. Amongt he available crystal forms of flsL there is ac onsiderable variation in the experimentally measured band gaps. More specifically,c ocrystals of flsL with highly conjugated coformers (pyrazine, acridine, andp henanthridine) display significantly lower band gaps than the dioxanes olvates. This observation clearlys uggests that cocrystal coformers have ap ronounced effect on the optical properties of the materials. The available experimental techniques, however,d on ot provide an explanation for this effect. The computational analysiso ft he band structure of the material, on the other hand, can readily provide such information.
The mostd irect way to establisht he role of the coformer in the generation of the band gap is to calculate the partial density of states (PDOS) alongside the full DOS calculation. The PDOS analysis presents am eanso fp artitioningt he full DOS into contributionsf rom different chemical species present in the crystal structure. In the present study,t he band structure was partitioned into contributions from fluorescein and the coformer molecules. The fls PDOS was then compared to the full DOS in order to establish whether the coformer molecularo rbitals are involved in the formation of the cocrystal frontier bands (HOMO or LUMO). The outcome of the calculation was in excellent agreement with the experimental observations: materials with the highest observed band gaps (acetone solvate and both dioxane solvates) showed no contribution to the frontier bands from the coformer molecules. Density of states analysis of the cocrystals involving pyrazine, phenanthridine, and acridine, on the other hand, revealed that the LUMO of these materials is fully localized on the coformer molecules:t he highly conjugated coformer molecules have LUMO p-orbitals that are lower in energy then the LUMO of flsL. Therefore, the presence of such coformers lowerst he band gap of the material by 0.66, 0.56, and0.33 eV for acridine, pyrazine, and phenanthridine, respectively (Supporting Information, FiguresS57-59). It should be noted, however,t hat the band gaps of the cocrystals, althoughl ower than those of the solvates, still lie outside of the visible light range. It is, therefore, unlikely that coformer molecules have as ignificant effect on the color properties of the cocrystals of flsL.T he color is mainly determined by the presence or absence of the flsZ surface layer.The calculations nonethelessconfirm that cocrystallization may be used as am ethod of altering the band gap of am aterial, thus alteringi ts optical and electronic properties.

Formation mechanism of the flsZ surface layer
The combinationo fe xperimental and computational studies has allowed us to establish the role of the zwitterion surface layer in the color properties of the lactoid fls cocrystalsa nd solvates.T og ain ac omplete understanding of the phenomenon, however, we need to establisht he reasons for the formation of this surface layer.E xperimentsw ith gray samples of flsL:pyrazine cocrystal have shown that the yellow surface  www.chemeurj.org layer gradually appearsw hen in contact with air.T he observed effect is unlikely to be caused by an interaction with air oxygen, since the transition of flsL to flsZ is not an oxidation process. Most likely,t he transformation is facilitated by water present in the air.I ndeed, our experiments have shown that the crystals remain grey for days when stored under zero humidity conditions (in ad esiccator filled with phosphorus pentoxide as ad rying agent). Exposure of the same material to 98 %r elative humidity,h owever, leads to the formation of the surfacelayer within 30 min.
Dissociation of cocrystalsc atalyzed by air humidity has previously been reported. [64,65] The processb egins with condensation of the water vapor on the crystal surface, where it leads to partial dissolution of the material and the formation of saturated solution of fls in water.I na queous solution the flsL molecules will form an equilibrium with the flsZ tautomer.S ince flsZ hasalower lattice energy than flsL,i ti sl ikely to be the least soluble speciesp resenti ns olution and is expected to crystallize preferentially,g enerating ac olored layer on the crystal surface.

Conclusion
Ac ombination of experimental and computational methods was used to rationalize the crystallographic behavior and color properties of fls tautomeric forms in the solid state. Three crystal structures have been determined: flsZ (yellow form) and two solvates of flsL with 1,4-dioxane. The structure of flsZ has been determined using ac ombined CSP/PXRD approach and furthers upported by TEM analysis. The determination of this crystal structure gives yet another example where ac ombination of experimental techniques such as PXRD, [42,43] TEM, [33,34] and solid-state NMR [60] with CSP provest ob ee xceptionally useful.
The previously reported crystal structure of flsQ [27] and the new structure of flsZ have both been determined using PXRD, at echnique insensitive to the positionso fh ydrogen atoms. To fully resolve any ambiguities in the assignment of these two tautomeric forms,s olid-state NMR spectroscopy combined with DFT GIPAW calculations were employed. The NMR studies fully supported the generally accepted assignment of flsQ and flsZ as the red and yellow forms,respectively.
Solid-state UV/Vis spectroscopy was used to measuret he band gaps of the materials and correlate them with the observed colors. The band gaps of all lactoid samples were consistently high, comparable to the band gap of aw hite solid of diacetylfluorescein, suggesting that the fluorescein samples should also be colorless. More detailed analysis of the spectra of the lactoid forms, however,r evealed the presence of an impurity with ab and gap of approximately 2.3 eV,c onsistent with the gap of pure flsZ.Further experimentswith larger crystals of flsL dioxane hemipentasolvate showed that the zwitterion impurity is concentrated on the crystal surface. Evidence has also been found that the formation of this layer is catalyzed by water presenti nt he air.W ater condensing on the crystalsdissolves the surface layer,forming asaturated solution in which the lactoid andz witterionic forms of fluorescein are in equilibrium. The zwitterionic form, which has al ower lattice energy and is, therefore, less soluble, crystallizes out of the solution and generates the yellow coating on the surface of the crystals.
The theoretical band structure calculations allowed us to establish the role of crystal coformers in the modification of color properties of cocrystals. It was shown that the coformers with extended p-conjugation have aL UMO lower than that of flsL.I ncorporation of such ac oformer into the crystal structure therefore leads to the lowering of the band gap of the material, the effect being most pronounced for the acridine cocrystal.
In this work we have applied ac ombination of experimental and computational methods to characterize the effects of tautomerisma nd crystal packing on the optical properties of fluorescein. We expect that, with the ever-increasing computing power and the progress in the development of computational methods, computational studies of organic solids will soon be standard procedure in materials research and in the development of new materials with targeted properties.