You are on page 1of 8


1900, 62, 615-622


Reaction Detection in Liqukl Chromatography; Krull, I. S., Ed.; Chromatographic Science Series; Marcel Dekker: New York, 1986; Vol.

(4) Haddad. P., unpublished work, 1989. (5) Jandik, P.; Li, J., unpublished work, 1989. (6) Small, H.; Stevens, T. S.; Bauman, W. C. Anal. Chem. 1976, 47, 1601.

Nile Red as a Solvatochromic Dye for Measuring Solvent Strength in Normal Liquids and Mixtures of Normal Liquids with Supercritical and Near Critical Fluids
Jerry F. Deye and T. A. Berger*
Hewlett-Packard, P.O. Box 900, Route 41 and Starr Road, Auondale, Pennsylvania 19311-0900

Albert G. Anderson
Central Research and Development Department, E. I. du Pont de Nemours and Co., Wilmington, Delaware 19880-0328

Nlle Red was used as a solvatochromk probe to measure the solvent strength of 85 pure llqulds and the resutts were compared to solvent strength scales from the better known Reland Et(Bo) dyes. Nlle Red Is also soluble In m e chardt’s pure supercrltkai flulds and In polar mixtures of supercrttical fluids and modlflers. The solvent strength of mlxtures of organlc modners In CO,, Freon-13, and Fr-23 was measured as a function of modlfler concentratlon wlth Nile Red and other solvatochromlc dyes. Nlle Red exhlblts large shlfts in the wavelength of Its absorption maxlmum allowing subtle changes In solvent strength wlth fluid composition to be quantlfled, and It Is stable In very strong acids. Relatlve scales of solvent strength comparlng different modtflers added to C02 are presented and compared to a wlde range of normal liquids.

function of composition according to the equation (5)

p‘ = 4 a p a + 4hpb



INTRODUCTION The polarity range of solutes that can be separated by supercritical fluid chromatography (SFC) can be greatly extended by the addition of polar modifiers to supercritical fluids. It has been widely repqrted (1-4) that the initial small additions (less than 1%) of polar modifiers (like methanol) to supercritical fluids (like carbon dioxide) significantly decrease the chromatographic retention of solutes on packed columns including those with a bonded stationary phase. Such large changes in retention caused by small changes in modifier concentration appear to be inconsistent with bonded phase chromatography (BPC). It is often assumed that bonded phases do not adequately cover silica surfaces and residual silanol groups produce long retention and very poor peak shapes. This can be viewed as a competition between a bonded phase retention mechanism and an adsorption retention mechanism which, in the absence of a modifier, favors the adsorption mechanism. Polar modifiers are assumed to decrease retention and improve peak shapes by covering the silanol sites (Le. blocking the adsorption mechanism). It is uncertain whether, upon site deactivation, retention follows a bonded phase or adsorption mechanism. Retention in BPC is normally a function of a mobile phase polarity parameter, P’, which in turn is usually a linear

where 4, and 4 b are the volume fractions of mobile phase components a and b, and Paand P b are the solvent polarity parameters for the two pure components. These polarity parameters are based on experimental solubility data like those reported by Rohrschneider (6). Partition ratios (k’Jare not necessarily linearly related to P’ but retention changes are generally limited to a 2.2-fold decrease in k’for a change in P’by 1.0 unit (7). If the P’values for carbon dioxide and methanol were 0 and 5.1, then the addition of 1%methanol to carbon dioxide should decrease retention by only a few percent compared to retention using pure carbon dioxide. Instead, much larger changes in retention are common. We recently observed that the addition of 1%methanol to carbon dioxide decreased the retention of chrysene on a SA column by 2.4 times (k’= 17.9 in pure COz to 7.6 in 1%methanol). Such large retention changes suggest that eq 1is inadequate for predicting retention in packed column SFC. Adsorption chromatography is characterized by large changes in k’with small additions of polar modifiers and thus, at least superficially, appears to offer a better explanation of retention in packed column SFC than BPC. In the standard theoretical treatment (8),retention of solutes is related to the adsorption energy of the mobile phase solvent on the stationary phase through the equation log K O = log V,

+ a(SO- A,€”)

where K O is the distribution coefficient for the adsorption of the solute, V, is the adsorbent surface volume per weight, a is a measure of the activity of the adsorbent, So the sample adsorption energy on the standard-activity adsorbent, A, the surface area covered by a molecule of the mobile phase solvent, and t o the adsorption energy of the mobile phase solvent per unit area of the standard adsorbent. In binary fluids, the mobile phase solvent strength, tab, tends to be dominated by the more polar component, b

0 1990 American Chemical Society

VOL. The rest of the work reported here was recently done with a Hewlett-Packard (HP) Model 8450A photodiode array spectrophotometer.(. Unfortunately. Only very small concentrations of water could be added to ternary mixtures of water. Attempts to rationalize the solvent behavior of binary mixtures with chromatographic behavior (9. For the most recent work involving modified supercritical fluids.Freon-13. For example. EXPERIMENTAL SECTION Chemicals and Instrumentation. on the ends of the flow path with Kel-F washen compressed between the windows and the cell M y .except for Reichardt’sE. This added water (either purposely added or absorbed from room air) was most likely responsible for small differences between the 1983 and some of the 1988 data. A Brownlee Labs microsyringe pump was used to pump the pure no framework for measuring these deviations other than through the general equation for the net energy of adsorption (AE) AE = E. SFC grade carbon dioxide in aluminum cylinders . the dyes employed often produce only minor wavelength shifts which make small changes in solvent strength difficult or impossible to observe. We use the term “solvent strength” as a measure of the interaction between solutes and solvent in a fluid phase.1. the nonlinear solute-solvent interactions (clustering) in the mobile phase. no special precautions were taken to exclude water. In the standard theory. At the same time. not stationary phase. We interpret this as increasing deviations from regular solution behavior. Extreme care was taken to exclude water from the solvents. which was synthesized according to procedures in the literature (13). acid-base.. using solvatochromic dyes. The present work employs Nile Red to provide detailed measurements of the mobile phase solvent strength of mixtures of polar modifiers in C02.. structure 11. MARCH 15.E. Instead.101 usually contain an implicit assumption that the mixtures behave as regular solutions. Early unpublished reports established that pure carbon dioxide ( C 0 2 )was in the same polarity range as hexane (32.. Flow from the two pumps was dynamically mixed inside the HP 1084. The windows were sealed. shifts only 3.).Ed represents mobile phase interactions independent of the stationary phase. They also point out some unexpected solution characteristics of mixtures of polar modifiers and supercritical fluids. respectively. and nb is related to molecular size. E. Reichardt’s Et(%)scale has been compared to tab values from adsorption chromatography (6) with “good” correlation. solutes may be surrounded by a solvation sphere containing a high local concentration of the modifier. In addition. 29-31) on pure supercritical fluids. Mobile phase interactions with the solute are generally ignored. using a pyridinium betaine dye. We believe that polar modifiers strongly interact with column packing materials. Anderson in 1983 with an IBM 9430 spectrophotometer. In spite of this.616 ANALYTICAL CHEMISTRY. and the other dyes were purchased from Aldrich. using a variety of carefully chosen dyes. there . The adsorption maxima and the resulting calculated values of transition energies for Nile Red reported in Table I were collected by A. leak tight. In the chemical engineering literature. where each term represents dimensionless partial molar free energies and the subscripts x and s refer to sample and solvent. A high-pressure flow cell was made of 316 stainless steel. Note that in this context “solvent strength” is strictly related to solute (dye)-solvent interactions whereas in adsorption chromatography “solvent strength“ is really an elution strength and normally has nothing to do with interactions in the mobile phase. ion-dipole interactions. respectively. Et(. once adsorption measurements are made available. 62. the adsorption maximum of which shifts depending on the polarity of the solvent in which it is dissolved. effects. Even in mixtures containing low bulk concentrations of modifier. and hydrogen bonding. Kamlett and Taft and co-workers (14-16) have developed separate scales. which are generally accepted in chemical engineering.E. related to surface phenomena. to separately account for polarizability. Strong interactions in the mobile phase tend to produce deviations in retention in adsorption chromatography which cannot be explained by mobile phase adsorption (8).4 nm when the solvent is changed from tetrahydrofuran to methanol (17 ) . structure I. 1990 and pure components. Fused silica windows were 6 mm thick and were transparent to UV to below 190 nm. If true.5 mm diameter with a 1cm long path length. a widely used dye in measurements on Kamlet and Taft’s A* scale. .Ed) independent of chromatographic retention. and Freon-23.and N20 with changing pressure. solvent strength measurements obtained by using solvatochromic dyes may provide the means of correcting for deviations between retention and the adsorption isotherm for the solvent on the stationary phase. Alternately.. they measured the solvent strength of C 0 2containing several concentrations of methanol.. It is also well-known that mobile phase interactions become more important as the difference in the polarity of the solvent components increases (8). solvents were of chromatographic grade or better. NO. There is still considerable debate as to the value or general applicability of solvatochromic solvent strength scales (Le. 31) reported the solvent strength of NH3. + mEal . This suggests that a substantial part of the nonlinear behavior of chromatographic systems using supercritical fluids mixed with polar modifiers may be due to mobile phase. Yonker et al. clustering of polar modifier molecules around polar solute molecules (18-28) in a large excess of a supercritical fluid is widely accepted. Carbon dioxide and methanol possess vastly different solvent strengths on all solvent strength scales and should produce extreme differences in retention in either BCP or adsorption chromatography. but we doubt such interactions represent a form of classic adsorption chromatography. Retention is. Nb is the mole fraction of b in the mixture. have been virtually ignored in discussions of retention in packed column SFC. thus. refs 11and 12). the absorbance maximum of 2-nitroanisol. modifier. C 0 2 .). The flow path was 1. The effect of water was checked by adding water to the modifiers. Reichardt (13) developed a solvent strength scale. Such clustering becomes more pronounced as the difference in polarity between the supercritical fluid and the modifier is increased. this has important implications on retention mechanisms in both packed column SFC and liquid chromatography. A plot of cab vs the mole fraction of b yields an adsorption isotherm for b. independent of any stationary phase interactions. and a and 1 refer to adsorbed and liquid phases. but a t the least they provide an approximate correlation to solvent polarity.33). Several solvatochromic studies have been performed (17. even though acidic and alcoholic solvents appear to produce a substantially different relationship between Et(30) and cab than other solvents. they attributed dramatic shifts in retention accompanying small additions of modifier to a modification of packed column stationary phases through adsorption of the modifier (15) onto “active sites”. Nile Red. They concluded that mobile phase solvent strength did not change dramatically w i t h small additions of polar modifiers. and carbon dioxide.e. 6. Solvatochromic dye measurements offer the possibility of directly measuring some function related to P’or alternately to mobile phase free energies (i. (17. Dyes were purchased commercially. and more detailed studies corroborate and expand on this finding. This raises the possibility that retention may folow a BPC type mechanism but with a nonlinear relationship between solvent strength and composition. A Hewlett-Packard Model 1084 liquid chromatograph with an adaptor kit for SFC was used to pump pure supercritical fluids. was purchased from Eastman Kodak. Ed .

4 518.0 541.1 37.44 benzonitrile 52.55 pentadecafluorooctanoic acid 45. This required the use of dyes exhibiting large solvatochromic wavelength shifts. and hydrogen bonding interactions.2.81 54.3-pentafluoropropanol 45. permanent dipoles.4 542.2 508.6 39.l-trifluoroacetone thiophene acetone tetrahydrofuran benzene ethylene glycol dimethyl ether toluene ethyl acetate isobutylamine furan perfluorobenzene 1. nm 541.0 543.5 38.6 480. E.71 1. VOL.8 504.2 548.58 aniline 50. In addition.4 518.2 556.42 54.6 627.2 37.95 57. .86 49.88 53.0 45.2 543.2 542. The systematic approach of Kamlet and Taft (14-16) appears to offer the best means of comparing the solvent strength of modifier-supercritical fluid mixtures to normal liquids and deconvoluting the contributions from dispersive forces.(. We were primarily interested in measuring relatively subtle changes in solvent strength caused by small changes in solvent composition.47 pentafluoropropionic acid 45. of the solvatochromic band calculated by E = 28591.0 535.0 35.4-dioxane tert-butylamine C0.87 52.1 44.8 628.2-trifluoroethanol 48.2 521.N-dichloroethane chlorobenzene methylene chloride ethyl acetoacetate phenylacetylene acetonitrile anisol fluorobenzene l.4 523. 2 nm spectral bandwidth between 400 and 800 nm).31 ethanolamine 51.9 57.34 53.02 53.16 acetic acid 51. Kamlet and Tuft's x* Scale.2 545.65 pentadecafluorooctanol 51..5 33. Since the system of Kamlet and Taft could not provide the desired sensitivity.0 50. (a) is EX3. PA.56 iodobenzene 52.6 547.44 butyrolactone 52.9 OTransition energy.50 57.47 trifluoroacetic acid 45.2 34.0 34.63 acetophenone 52.8 628.78 53.ANALYTICAL CHEMISTRY.50 heptafluorobutyric acid 2.2 42.34 53.) in Normal Liquids" solvent 1 2 3 4 5 6 7 ~ N R ) 617 wavelength.3-dimethyl-2-imidazolidinone 52.5 609.93 55.0 497.8 555.3.5 41.3 30. '. Comparison of Nile Red Transition Energies to Reichardt's E .9 45.2 544.35 55. 1270 psig 52.2 Et(30) 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 pentafluorophenol (40 "C) 44.5 536.83 52.4 34. acid-base.3 547.2 540.8 540. 62.26 53. NO.4 568. was obtained from Scott Specialty Gases.3 39.15 I-butanol 52. s )family of pyridinium betaine 60 c 57. The dyes had to be reasonably soluble in a wide range of pure solvents and supercritical fluids.36 55. 6.6 530.45 55.2 535.p Diathyl athrr ' I- '.9 37.9 38.9 37. Edgemont.2 33.2 37.8 557.58 53.0 41.7 539.0 34.2 511.5 527.1 37.6 520. ( .0 550.6 536.5 536.6 487.15 55.0 516.4 48.42 53. 24 "C.29 53.1 36.2 494.2 50.9 54. nm 637.2 63.0 558.8 46.62 55.44/(nm) in kcal/mol.2 538.65 48.5 55 t - cc'4\T(.14 53.N-dimethylformamide N. Finding an appropriate dye or family of dyes became a bigger issue than expected.19 54.49 m-cresol phenol (45 "C) 47.8 566.87 58.2.6 51.46 2.75 pyridine 52.21 water hydrogen cyanide (15 "C) 48.98 methanol 52. our initial measurements with several of these dyes produced inadequate resolution of solvent strength differences and further work was dropped.0 600. Comparison of transftion energies calculated from the wavelengths of maximum absorption of Nlle Red and EN301 solvatochromic dyes in a number of common liquids.2 533. Plumsteadville.43 water/methanol 1:l (v/v) benzyl alcohol 50.36 heptafluorobutanol 49.5 56.89 53. other alternatives were evaluated.26 formamide 50.N-diethylamine tert-butyl methyl ether m-xylene diethyl ether C02/methanol 95:5 (v/v) carbon tetrachloride triethylamine tributylamine cyclohexane n-hexane pentane COz.5 45.24 56.8 39.79 43.19 55.12 59.6 40.05 ethanol/water 8020 (v/v) 51.6 42.4 529.7 53. Unfortunately.4 515.99 54.0 565.29 53.5 544.4 36.3 51.0 536.63 59.l.6 Et(30) solvent 44 45 46 47 48 49 50 51 52 53 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78 79 80 81 82 83 84 85 86 87 N.54 wavelength.5 43.9 50.5 634. MARCH 15. )and E.1 53.21 nitrobenzene 52.4 548.51 50 47.16 58.68 56.1 55.5 560.9 51.) value calculated in ref 27.78 53.methanol 9O:lO (v/v) tetramethylethylenediamine N. .N-dimethylacetamide 1-methyl-2-pyrrolidinone 2-propanol carbon disulfide chloroform bromobenzene 2-methyl-2-propanol 1.25 1-propanol 52.0 41.51 1-octanol 52.4 627.3 40.6 545.0 42.02 59.2 578.2 545.2 35.6 56.2 564.5 530.6 514.63 propionic acid 52.6 484.3.51 ammonia (-34 "C) N-methylformamide 51.6 525.6 531.1 33.93 47.06 quinoline 52.8 46.93 56.0 628.56 hexafluoro-2-propanol 46.90 53.5 35.0 537.1 37.4 37. 1990 Table 1.3 46.02 dimethyl sulfoxide 52.0 585. Reichardt's ( 1 3 .85 formic acid 45.2 57.6 531.4 506.35 51.90 53.13 ethanol 52.09 45.8 536.64 54.51 ethylene glycol 50. Freons were obtained from Du Pont through Melchior h s t r o n g .2 602.4 41.94 55. Reichardt's Dyes. PA.3 46.2-trichloroethanol 48.5 516.41 56. Kosower's dye (34) was not soluble in a wide enough range of solvents.6 585.9 35.0 48.6 549.0 549.4 32.2-dichloroethane N.6 32.0 511.18 53.0 491.2 590. Choosing Dyes.2.71 nitromethane 52.42 53.7 44.2 579.8 50.82 2. available instrumentation consisted of a spectrophotometer with limited wavelength resolution (1nm spectral bandwidth between 200 and 400 nm..4 483.4 530.0 593. 52.40 sulfolane 52.4 502.4 542.5- 45 I 30 35 40 45 50 55 60 E t mo) Figure 1.

but we found little difference in the solubilitiesof and Et(. Although not ideal.The Nile Red absorption maxima and transition energies in a large number of solvents are presented in Table I. composition was varied while spectra were continuously monitored. in carbon dioxide. Neither of the latter dyes is stable in acidic media. single spectral bands were observed. MARCH 15. . Two values of Et(30) for methanol are presented as half shaded circles in Figure 2. structure 111. Subsequent purging of the solutions with either helium or nitrogen resulted in the rapid return of the color.. and spectra with usable levels of absorption could be obtained. Nile Blue A a limited number of solvents as a solvatochromic probe of solvent basicity in polar aprotic media (42. making it somewhat difficult to be certain of the wavelength of maximum absorption. Wavelength measurements were converted to transition energies in kilocalories per mole by dividing 28592 by wavelength in nanometers (34. Bubbling COzthrough solutions of either of these dyes in polar solvents. Results in Fluids Used in SFC.26. 35) and Et(2s)(37) derivatives were purported to be more soluble in COz than Et(30). Values of E ( N R ) vs Et(30) for mixtures of methanol in carbon dioxide.. Some of the data in Table I are also presented graphically. Et(26) Ili.618 ANALYTICAL CHEMISTRY.05) (47)permits observation of its solvatochromic peak in acidic media (48). This contrasts with the hypsochromioc shift of Reichardt’s E. In very acidic. and the more polar value was obtained from an older.. Note that the data on the left and right extremes of Figure 1 allow parallel lines to be drawn with an offset occurring in the region where hydrogen bonding and donor/acceptor characteristics are important. However. resulted in the rapid loss of visible color. has been the most popular 0 i. especially those related to Nile Blue A (39). As the experiment progressed and increasing concentrations of the more polar solute was added. Nile Red NH. provides a clear picture of the relative strengths of the solvents. The C02 data were obtained by using very dry methanol. and solvent polarity.) vs Et(3o) curve from Figure 1. The lower dashed line is the normal solvent E. although its solvatochromic behavior has been noted. while bases made color reappear. The wavelength of maximum absorption of Nile Red was measured in a number of common liquids used as solvents or modifiers in both LC and packed column SFC.. This dye exhibits large shifts in the wavelength of maximum absorbance with solvent polarity (>160 nm shift between carbon dioxide and trifluoroaceticacid). 43). Nile Red has found substantial use as a lipophylic dye for staining fatty tissues (38). used bottle of methanol. New bands appeared but could be differentiated from the band of interest. Nile Red was previously used with Nile Red is fluorescent in some solvents. 6. maintained nearly constant color for months. The figure shows that Et(. This behavior led to uncertainty in interpreting the results. Freshly opened bottles gave the same results as scrupulously dried modifier. in Figure 1. it has not been examined extensively. 62. The absorbance maximum of Nile Red does not appear to shift significantly when subjected to hydrogen bonding solvents. such as methanol. The addition of acids also caused the color to disappear. Et(30) solvatochromic dye due. R = t Bu. and tends to precipitate from solution as modifier concentration becomes small.(. 1990 dyes including Et(30).where values of E(m) for Nile Red and Reichardt’s Et(. RESULTS AND DISCUSSION Measurements in Normal Liquids. R = Phenyl. Freshly opened bottles tended to produce slightly less polar readings than older bottles that had been opened many times. Nile Red. absorption bands of the protonated form of Nile Red appear. VOL. Nile Red provides the best available combination of characteristics for our requirements. to the fact that it exhibits the largest known shifts in absorption maxima with changes in solvent polarity. have also been noted (4043). NO. strong acids. Freon-13. because it is more stable in acidic media. Because of its photochemical stability and strong fluorescence peak (which is also solvatochromic). in contrast. by itself. The less polar value was obtained by using methanol from a freshly opened bottle.00 i 0. The solvatochromism and fluorescence of oxazine dyes. In solutions of bases.) are plotted against each other. Nile Red transition energies are designated E ( y ) . Nile Red is uniquely stable in extremely acidic media and is not susceptible to a loss of molar absorptivity in the presence of acids like Reichardt’s dye. such as trifluoroacetic acid. . We concluded that these dyes are very susceptible to the presence of traces of water. The two values represent the extremes observed over several months using methanol with no special preparation. From these data it appears that Nile Red. in part. The solvatochromic absorption bands of the regular form of Nile Red are still present but are sometimes not very obvious. Fluorescence emission band($ overlap the solvatochromic absorption bands of the protonated form of Nile Red to create what appears to be a triplet in the absorption spectra. Neither scale. At the beginning of the experiment.but.) is extremely sensitive to hydrogen bonding solvents. However. Solutions of Nile Red in neat. Its low basicity (pK. considerable concem has been expressed over this dye’s susceptibility to solvent hydrogen bonding power (16. we did observe a major decrease in the molar absorptivity of both and Et(30) when COzwas added to many modifiers. Nile Red has been suggested as a laser dye (46). solution acidity. = 1. Values for Reichardt’s Et(30)and Et(26) are also presented for comparison. The penta(t-butyl) (29. Composition was varied by making mixtures of a low polarity solvent and the high polarity solvent of interest. 11. and Freon-23 are presented in Figure 2. R IV. Unfortunately. that Nile Red is much less sensitive to hydrogen bonding and proton donor/acceptor solvents. It also produces a bathochromic wavelength shift consistent with stabilization of the excited state in p-r* or T-T* electronic transitions and comparable with the T* scale of Kamlet and Taft (16). significant concentrations of carbonic acid could not form or dissociate.36)and it produces a hypsochromic shift with increasing solvent polarity in direct contrast to the family of dyes suggested is not very soluble in pure C02 by Kamlet and Taft (14-16). #en water was excluded.. as previously observed by others (36). but the intensity of absorption for Et(m) was low. The numbers next to the curves are volume percents of methanol mixed into the various fluids. Nile Red tends to lose color over several days. Methanol Modifier in Carbon Dioxide and Freons. is somewhat more useful than Et(30) or Et(zoin measuring the solvent strength of acids and other very polar solvents. R - 0.structure IV. polar solvents. Small concentrations of water purposely added to the modifier gave small spectral shifts similar to those observed with “old” bottles of modifier.49). the band due to the normal form of the dye was observed to shift. Nile Red is soluble in solvents at the extremes of polarity of interest. To avoid confusion and to deconvolute the overlapping bands.

the more dramatically the solvent strength of the mixture shifts.1 57. then it is probable that there will also be a strong association between the polar modifier and the polar stationary phase. The Nile Red transition energies of mixtures of methanol. The difference in behavior between the mixtures containing hexane and carbon dioxide can be attributed to the more limited compressibility of hexane compared to carbon dioxide (1419).9 57. causing clustering in the mobile phase. Pooling of the modifier on the stationary phase causes a change in the polarity and/or volume of the stationary phase.4 54. and tetrahydrofuran in carbon dioxide are listed in Table I1 and shown in Figure 4.9 52. The similarity between the curves in Figures 2 and 3 gave some reassurance that the effeds observed were realistic measurements of solvent strength.7 53. On either solvent strength scale (E(*) or Et(3o)) the polarity of the pure fluids follows the order Freon-13 < C02 < Freon-23. Mixtures of methanol in Freon-13 are always less able to elute polar solutes than mixtures of methanol in carbon dioxide at the same concentration of modifier. local composition and bulk composition are probably the same. however. 1.3 55. The relatively more linear behavior of 2-propanol in hexane. Instead. for reference. Other Modifiers i n Carbon Dioxide.6 56. respectively. shown in Figure 3.9 57.6 53.ANALYTICAL CHEMISTRY. the more intense the clustering and the more nonlinear the relationship between solvent strength and composition. but for mixtures of methanol with three supercritical fluids. In Figure 4. This can be explained by using the concept of clustering (17-28). If plots like Figures 2 or 3 are linear.8 54. VOL. Hexane is very nonpolar while 2-propanol is moderately polar.4 57.7 52.6 57. We believe this is due to intense interactions between the modifier and the relatively polar stationary phases we usually use in packed column SFC (i. The results in Figure 2 are very informative. Horizontal lines connect these two extremes.6 56. NO.7 55. Wavelengths of Absorption Maxima and Calculated Transition Energies for Mixtures of Common Modifiers with Carbon Dioxide (at 25 "C) ~ N R ) MeOH % CHBCN kcall mol CH2Cl2 THF kcall mol nm modifier 1" kcall nm mol nm nm 488 494 498 500 504 512 522 532 540 548 554 kcall mol 2" 3a 4" 5 10 20 40 60 80 100 58. methanolFreon-13 mixtures appear more polar than methanol-COz mixtures of similar composition (on either solvent strength scale). MARCH 15.\ \ \ \ \ 50 55 60 E f (30) Figure 2.4 53. Thus. With greater than 12% methanol added to Freon-13. the base-line noise increased dramatically.2 54. acetonitrile. Mixtures of 2-propanol in hexane were also evaluated with Nile Red to determine whether the behavior of methanol in nonpolar supercritical fluids is perceptibly different in comparison to mixtures of normal liquids. . cyanopropyl or diol bonded phases on silica). The larger the difference in solvent strength between the two pure fluids (supercritical fluid and modifier). the vertical lines on the extreme left and right represent the solvent strengths of pure carbon dioxide and pure methanol. 1990 619 Table 11.9 53.8 53.1 492 496 504 516 524 530 538 58.1 490 494 498 508 518 528 58. polar molecules tend to cluster together forming a local environment that is much more polar than the bulk composition would suggest.2 51. At low concentrations they do not. 2-Propanol in Hexane. 6. suggesting the fluid broke down into two phases.4 56. methylene chloride.2 56. Lower dashed line is same as line in Fig. The greater the difference in polarity between the components. as in Figure 1.e. Polar molecules are only sparingly soluble in Freon-13 and tend to be rejected by it. The top three '\ '. Hexane and 2-propanol were chosen because they have one of the greatest differences in polarity possible among binary pairs of miscible. None of the dyes is soluble in pure Freon-13.3 57.8 54. Unpublished chromatographic results have demonstrated to us that these pure fluids follow the same order in their elution strength of many solutes from a wide range of stationary phases. Comparlson of energies. form a second phase. \ 50 t 30 35 40 45 .1 Collected at 29 "C. The addition of methanol as modifier appears to shift the solvent strength of the mixtures dramatically. We could not produce mixtures of over 12% methanol in Freon-13 at temperatures between room temperature (25 "C) and 80 "C and over a wide range of pressures. compared to alcohols in carbon dioxide suggests that there is less clustering of the 2-propanol. The solvent strength measurements summarized in Figure 2 do not precisely parallel the elution strengths of these mixtures in chromatography. The results.7 55. so the position of the unmodified fluid is not shown in Figure 2.5 54. 62.8 54.6 496 504 512 522 526 532 538 57. indicate similar but less dramatic behavior compared to the behavior observed for mixtures in supercritical fluids (see Figure 2). If the supercritical fluid essentially rejects the modifier.7 55. normal liquids.

57.06 . These wavelength changes are similar in magnitude to shifts caused by small changes in fluid composition as shown in Figure 5. for a pure fluid at constant density. Apparently.23 25 7' THF 80 I? 20 40608 0 5 1 3 20 40 60 CH3CN 5 IO lp 20 5080 20 I 40 60 80 I / I . This result suggests that negative temperature programming of constant composition binary mixtures should produce increasing solvent strength which agrees with general experience in capillary SFC.620 ANALYTICAL CHEMISTRY. Effect of temperature on absorbance maxima of Nile Red in mixtures of carbon dioxide and methand.35 58. Much larger changes in k ' are achievable in LC by changing fluid composition. Effects of Temperature and Pressure. 62. in volume percent. Pressure.the nonlinearity is a mobile phase phenomenon involving the solute and the modifier whereas in adsorption chromatography the nonlinearity is due to a mobile phase-stationary phase interaction. We also observed only modest shifts in the wavelength of Nile Red absorption with changing temperature.95 . tetrahydrofuran (THF). Several authors have shown (52. 51).87 to 0.76 \ \ \ \ \ 490 .e. I co I 2 59 58 MeOH 1 60 57 E(NR) 56 55 54 53 52 51 .56. VOL. Very little change in solvent strength was observed when pressure was increased from 76 bar (1100 psi) to 241 bar (3500 psi) at temperatures from 25 to 40 OC.584 25 45 I 30 35 TEMP. The density vs temperature and pressure relationships of methanol-carbon dioxide mixtures were not known when these measurements were taken. They do not necessarily represent large changes in solvent strength and can only cover a small fraction of the full range of solvent strengths available with normal liquids. I+?-----. Hyatt (29) reported similar mobile phase solvent strengths at sub. Figure 5.0 . This means that 10% methanol in carbon dioxide has a solvent strength one might normally expect from 50% methanol in carbon dioxide (assuming ideal solutions).18 E ( ~ ~ ) -<.62 ' . and acetonitrile (ACN). there are no abrupt changes in chromatographic retention as temperature is increased from sub. Such large variations in k'. on the bottom line the number 10 appears about half way between the vertical lines representing pure carbon dioxide and pure methanol. NO.531 that. x 5% . Such observations need to be put in perspective.-$ i 56. It is also interesting to observe that solvent strength decreases when the fluid temperature is changed from subcritical to supercritical. have led many people to equate large changes in density with large changes in solvent strength. Transition energies obtained by using Nile Red for mixtures of carbon dioxide and common modifiers. KCAL/ MOLE Figure 4. Thus.are small compared to changes achievable in LC. Experiments were undertaken to determine the range of solvent strengths achievable with density variations compared to composition changes using Nile Red transition energies.58. I 2 34 5 I IO I I / I I 20 I 4 0 I 60 80 I I . The numbers above the tick marks on the horizontal lines are the concentration of the more polar modifier in the less polar solvent. horizontal lines are broken by intermediate vertical lines representing the solvent strength of pure Freon-23. FREON. The wavelength of maximum absorbance changes no more than . a change in solvent from hexane to methanol (through multistep substitutions) can result in a change in solute retention by as much as lo9 or 1Olo times. subsequent measurements of density using a "U" tube densitometer shows that an increase in temperature from 24 to 50 "C (at 130 bar) decreases density from 0. The numbers next to the curves represent the volume percent o f methanol in carbon dioxide. 2-propanol and hexane. Temperature. However. achievable using only changes in the density of common supercritical fluids. the dye molecules are surrounded by a solvation sphere that has a higher concentration of -modifier than is present in the bulk. The data are presented in Figure 4 in a format similar to the format used to express elution strength (cab) information in adsorption chromatography (50). However. One operating mode in SFC involves density programming where partition ratio's ( k ? can be changed by a factor of at least lo3times (i. MARCH 15. Thus. 6. 40 45 50 30 35 40 45 50 55 60 'C E t(30) Figure 3. Comparison as in Figures 1 and 2 but for mixtures o f two normal liquids. in Figure 4._ 57. achievable with changes in density. 1990 60 I"""' \ 5. In adsorption chromatography.and supercritical temperatures of a number of Kamlet and Taft (14-16) dyes.69 g/cm3. the apparently large changes in k'in supercritical.

We also wish to thank the Du Pont Company for allowing us to publish the Nile Red measurements in Table I.. Wong. 1985. Binary mixtures of a supercritical fluid and a polar modifier appear to allow much wider variation in mixture solvent strength through composition adjustment than is possible by using binary pairs of normal liquids. R. J. 2nd ed.4 821 B X METHANOL 1 - 1. I n Supercrltical Fluids: Chemical and Engineering Applications. J. M. and experience a subsequently higher polarity.. 1983. L. T. 8 . Chromatogr. R. 2239. Fluid Phase Equlllb. 667. Kim.1976. P. 119-147. 73. 1983. MARCH 15. M. density programming is likely to produce very subtle changes in solvent strength and will likely be most useful in differentiating very similar molecules. R.8 0 . 1986. P. 1887. K.. R. 25(1). L. 1962. Chapter 7. DC. (21) Deiters. (9) Mourier. K.. J. Such wide variations allow nonlinearities between solvent strength and composition to be much more apparent. 1990 A 1. ACKNOWLEDGMENT We thank Professor Keith Johnston of the Chemical Engineering Department of the University of Texas and Dr. but substituting E(NR) for methanol concentration. The wavelength of maximum absorbance was primarily determined by the amount of modifier added and was relatively insensitive to the pressure or density of the fluid. Chapter 3. Results similar to those reported for pure carbon dioxide. P. 109. or “local composition”.7 g/cm3). (13) Dlmroth. L. 1982. Chem.3. KCAL/MOLE - 55 Flgure 8. Clearly. Chem. Reichardt.. are not being accounted for when log k’ is plotted against E(NR). M. Am. Cooper. R. C. L. M. D. Acta Chem. (25) Nitsche.. CONCLUSIONS The retention mechanism in many packed column SFC separations is still a matter of debate. solubility of such solutes is controlled by the identity and concentration of the modifier and the nature of the clustering phenomenon. K. 835. J. Chromatogr. 6027-6036. Prauznitz. Rosset. Such findings are consistent with the observation that density changes produce much smaller changes in retention in SFC than composition changes produce in LC. S. R. Bohimann. Am. 36 39-82. W. (18) Kim. 6. U. 25. In supercritical fluids containing polar modifiers. Chem.or “clustering” means that polar solutes are surrounded by a solvation sphere containing a substantially higher modifier concentration.2 - 1. Johnston.. The possibility that solvent mixtures produce nonlinear changes in solute-solvent interactions vs composition has not been widely considered in chromatography even though discrepancies between retention and composition have .. 4-5. Taft. were also observed with mixtures of modifiers in carbon dioxide. J. 1985. Plots of log k’vs percent methanol are nonlinear. 123-129. Azevedo. isoleucine (squares).. W. R. 1974. 1973. W. T. W. SOC. AIChEJ. Yonker. 346-348.. Such mobile phase effects must have an impact on chromatographic retention. Scand. Slepmann. R. VOL. FluidPhase Equillb. T. J. J. 57.. K. Anal. but the exact nature of such effects has not yet been determined. as shown in Figure 6b. 537-554. Chem.. W. S. R. C. Scand... Prausnttz. (7) Snyder. 1963. Such clustering produces very nonlinear solvent strength vs composition profiies and suggests a poasible alternative to or an enhancement of previous models of retention. D. P. Suppan. This leads to a drop in solubility of polar solutes in the mixture (27). Abboud. T. J. K. R. R. but the number of clusters decreases substantially. 7391-108. (12) Kamlet. likely related to modifier-stationary phase interactions. Chem. 1981. 351-360. R. P. The clustering of polar modifier molecules has been shown to produce local regions of high solvent strength. P’. M. J. Eds. J.5. 62. 99. 2886-2894. Retention of the phenykhiohydantoin derivatives of proline (diamonds). 13. princ@ks of Adsorption Chromatography. P. Paulaltis. (11) Sjostrom. J. Grelbrokk. There is a reasonable likelihood that some aspects of both BPC and adsorption chromatography impact retention in normal phase.4 1. If adsorption of the mobile phase solvent onto the stationary phase is important in retention. Preliminary Correlation with Chromatography.. the equations for adsorption chromatography must be extended to quantify specific solute-solvent effects in the mobile phase. (16) Kamlet. Intmdwtkm to MDdem UquM Chrometography. The minimal effect of pressure on the apparent solvent strength of the mixtures suggests that chromatographic retention of polar solutes will likely be more a function of compositon than density when polar modifiers are added to supercritical fluids. and a-aminobutyrlc acid (circles) on a cyanopropyl column as a function of (a) methanol modifier concentration and (b) solvent strength measured by using Nile Red... L. K. Taft.. Kamlet. been discussed (49). Hewlett-Packard Applkatlon Note AN800-2. 98. as shown in Figure 6a for three PTH amino acids. Preliminary correlation of retention with b. P. (24) Hu. The separation of phenylthiohydantoin (PTH) derivatives of amino acids have been extensively studied in our laboratory (54) as a function of methanol concentration in carbon dioxide... Chimia 1982. G. (6) Rohrschnelder. Squires. Anal. System density primarily determines the nature of the modifier cluster and thus has only a secondary effect on retention adjustment (a limited range of solvent strength adjustment). Fluid Phase Equilib.8 0. E. SOC. M.. M. Chapter 8. Such “association”. Atlas of Chromatograms. S. 1986.1977. R. 367. M. Tan. S. Am. Wold. (23) Mathlas. (2) Gere. J. Chromatogr.. (15) Taft. 6 -4 . J. The spectrophotometer resolution (2 nm) is on the same order of magnitude as the observed spectral shifts. 6. Chromtographia 1988. 92. at least at high densities (>0. (14) Kamlet. 377-383. R.120. Alternately. D. LITERATURE CITED (1) Blllie. T. 1979. 55-60. A. 1. Kalkwarf.. W. P.. (3) Taylor. M. M. Smith. Residual nonlinearities in Figure 6b suggest that other important interactions.. Marcel Dekker: New York.. L. (5) Snyder.. 1241. NO. (19) Johnston. packed column SFC. F. L. Sassiat. (17) Frye. 353. 1883.. 36(9). 835..0 log h’ 0. If packed column SFC follows a mechanism characteristic of BPC. E. than expected from bulk composition. prhrate communication. Plots of the same retention data. 33(10). D. Chem. (8) Snyder. (4) Yonker. 611-628. 1983.ANALYTICAL CHEMISTRY.. Chromatogr. produce substantially improved linearity. corresponding to a change in E ( N R ) of less than 0. 45. Fluidphase EquHlb.. 1968. 1603-1611. J. G.John Wiley and Sons: New York. Smith. J. 61-75. Soc. the equation for solvent polarity. Fluidphase Equillb. American Chemical Soclety: Washington. Y. This is consistent with previous observations (18) that the concentration of modifier inside clusters increases as the density decreases. C. Clement Yonker of Battelle Northwest Labs for helpful discussions. 1978. 2-4 nm. W. 1987.. must be modified to account for the difference between bulk and local composition. Copeman. 25-32. (22) Whltlng. 0 0 25 5 75 IO 96 METHANOL 59 56 57 66 E(NR). 9 . Kkkland. The intensity of interaction between solutes and modifier in the clusters actually increases. Justus LWigs Ann.-S. Scl. (10) Smith.. (20) Johnston. M. J. 1987. Acta Chem. J . 223-230. Solvatochromic shift measurements using Nile Red indicate a strong association between the dye and polar modifiers. Caude. 98. L. L. 1987. T. the primary mobile phase interactions experienced by polar solutes should be with clusters of the modifier.

F. the extraction system is. D.. U. Soc. H. Wong. Chem.. J . 62. Conn’s Biob@C~?/ Stalns. >99%) was purified to remove all traces of acidic impurities. b 48. 324.1 M acetate solution. E.. Fluid Fhase Eguilib.. D. 93. D. DTsch. Chem. pp 372-430.. Am. RECEIVED for review June 27. M. Czech. Thls Is the first report of the lanthanldes extracting as hydroxide complexes. 1888. J . M. Chem. 1989.. 46. (32) Bente. S. I. 1984. Stock solutions containing lanthanum(III). and ytterbium(II1) were prepared from chloride salts (Alfa Inorganics. D. D. 295-314.. D. Central Research and Development Department. 1804. Org.. Am. M. 1968. 1370. Ges. 3255-3263. rlvallng those for common acidic organophosphorus extractants In the TOPO concentration range 0. R. D. M. E. Extraction of Tervalent Lanthanides as Hydroxide Complexes with Tri-n -octylphosphine Oxide Ted Cecconie and Henry Freiser* Strategic Metals Recovery Research Facility. Cbem. J. Harbusch-Gornert. 3022-3026. G. T. Infrared spectra were obtained with a PerkinElmer 1800 Fourier transform infrared spectrometer. 1987. auxiliary complexing agents. 3 )led us to examine systems such as quinolin-%thiol and dithizone in the presence of TOPO for selected Ln3+.. 1985. Soc. Ouw. Justus LieMg. Chem. E. A. Uhlmann.7 ppm.. L. Collect. A. G. M. 49.nTOPO. 1917.. 60.. Dtsch. S. 45268. M. INTRODUCTION The use of monodentate neutral phosphorus compounds in metal ion extraction is widely known (I). A.. 0. Swaid.. (27) Dobbs.. 34-41... 9th ed. L. W.. crystallize overnight: mp. 33. D. 38. J. Leffler. praseodymium( I I I ) . F. Anal. 1989. The acetate-washed solution was equilibrated 12 times with an equal volume of deionized water to remove any acetic acid extracted by TOPO ( 5 .P. The extraction selectivity for thls system increases with increasing TOPO concentration. Paul: Weaver. europkrm(III).B. Chem.. MD. Saunders. 1887. R. H. pp 133-143..025-0. Opt. On observing the surprising extent of extraction in the absence of the S-containing ligand. 1966. (33) Deye.. T. University of Arizona. Chem. holmium(III). unpublished work. 50. Res. Chem. J. unpublished work. A. A. 721. Ed. the most selective for lanthanlde separations to date. R. 99. qll& 31PNMR (CDCI. (28) Walsh.. E. Lindby. terbium(III). terbium(III). The report of extraction of lanthanides and actinides by sulfur ligands in the presence of tri-n-octylphosphine oxide (TOPO) ( 2 . we decided to examine this unusual phenomenon further.. Berger. Hetzer. 56-65. 27. Anal.5 M) TOPO solution in chloroform 8 times with an equal volume of a 0.J. Chem. Chem. G. 873. 1471-1472. Chromatugraphy 1987. 0. The lanthanides were found to extract as hydroxide complexes of the form Ln(OH). (31) Yonker. R. G. E. J. G. europium(III). employing manganese as an intemal standard. 38.. 8 1971.100 M and surpassing them at higher concentratkns. Commun. Mcmanigill. to remove the acidic impurities (4). 18. 1876. T. F. F. These compounds generally function as adductants.. F. 3261. (38) H. When TOPO concentrations of 0. 1990. The lanthanide concentration remaining in the aqueous phase was determined directly. 884. Department of Chemistry. 1983. 2(5). M. 50-52 “C. M. Herzbaum. J. 451. M. 575-587. Chem. Am. D. Frye. with the resulting oily material placed in a vacuum desiccator to (P=O). R. 26. Doran.9 Ann. (36) Fowler. Kalkwarf. W. Smith. IR (KBr). 90. TOPO (courtesy of American Cyanamid. Distribution Studies. and ytterbium( I I I ) from dilute chloride solutions into chloroform solutions contalnlng trkn-octylphosphlne oxide (TOPO) was examined. V. Harry. 48. Ed.50/0 0 1990 American Chemical Society . Soc. The equilibrium pH of the aqueous phase was measured after phase separation. Applied Science Publishers: Essex. 1973. Ber.622 Anal. 6er. Anal.Ashraf-Korassani. Taylor. K.R. Ind. J. Schneider. 105. 1898. private communication. Chloroform (AR grade) was washed 3 times with deionized water prior to use. van Wassen. overall.. Board. Chem. Gas Chromatugraphy 1972.. 260. praseodymium(III). The concentration of lanthanide extracted into the organic phase was determined by back extraction into - 0003-2700/90/0362-0622$02. Angew. F. The Williams and Wlkins Co. K. 3267.). (29) Hyatt. Lillie. T. Anal. 3581. Chem. Johnston. Donohue. Bastlng. J. 1980. J.. C. C. R.: Baltimore. HewlettPackard. Accepted December 15. I. 1973. Arizona 85721 The extraction behavior of lanthanum( I I I). Chem. (39) Relssig. 38. 1871.. D. 107. Katritzky. 289. du Pont de Nemows. Stuzka.. J . (26) Figueras. 1974. J . Tucson. Kolling. a time found adequate for equilibrium to be reached. Lahlere. (40) Mohlau. 6 ) . Anderson. (37) Andersen. R. Rutherford. 90. J. 1977. J . 19. which enhance the extractability of coordinatively unsaturated chelates by replacing water molecules remaining coordinated to the complexed metal ion. Ikonomou. (41) Kehrmann. 1907. SOC. J . 37. L. J . Engl. J . 194. Sci. Anal.1989. M. HCl/NaCl) containing the lanthanides at 10”’ M was equilibrated with an equal volume of a TOPO solution in chloroform for at least 15 min. Int. R. J. Chem.25 M or greater are employed.. Chem. J. M. This document is contribution number 4980 from Du Pont’s Central Research Department. P. This purification involved equilibrating a concentrated (0. m t n .9%). Our hypothesis of the formation and extraction of a lanthanide hydroxide complex was so unexpected that we took particular care to examine and reject alternate explanations. (30) Slgman. such as di(n-octy1)phosphoric acid. 472. 1983. 5097-5101. A.1. D. Soc. H. Eng. 460. Kehrman. Cbromatogf. hdmium(III). 91. pH = 5. 1976. The phosphorus level of a l l except the first of the eight washes was below the inductively coupled plasma limit of detection for P ( 5 X lo4 M). Simanek. 3253. Chem. 55.. 31PNMR spectra were obtained at 250 MHz with a Bruker WM-250 spectrometer.622-625 (42) (43) (44) (45j (46) (47) (48) (49) (50) (51) (52) (53) (54) Thorpe. P. Common. Chem. 1986. An unbuffered solution (p = 0. Kollina. V. Chem. Lauer. EXPERIMENTAL SECTION Apparatus. (34) Kosower. Deye. Schafer. Justus Liebigs Ann. Phosphorus and lanthanide concentrations were determined with a Perkin-Elmer 6500 ICP spectrometer. Metal distribution results were obtained by inductively coupled plasma atomic emission spectrometry (ICP-AES)analysis.. (35) Reichardt. R. Reagents. Qes. D. The final TOPO solution was placed in a Rotovap to remove the chloroform solvent. M y s . J. 1424. Dorsey. W.. which are found in commercial sources of TOPO. Davis. 1958. F.