Download as pdf or txt
Download as pdf or txt
You are on page 1of 8

Journal of Colloid and Interface Science 231, 351358 (2000) doi:10.1006/jcis.2000.7164, available online at http://www.idealibrary.

com on

The Dielectric Function for Water and Its Application to van der Waals Forces
Raymond R. Dagastine, Dennis C. Prieve, and Lee R. White1
Center for Complex Fluids Engineering, Department of Chemical Engineering, Carnegie Mellon University, Pittsburgh, Pennsylvania 15213 Received April 21, 2000; accepted August 11, 2000

The dielectric response, (i ), for water (which is required in Lifshitz theory to calculate the van der Waals interactions in aqueous systems) is commonly constructed, in the absence of complete spectral data, by tting a damped-harmonic-oscillator model to absorption data. Two sets of parameters for the model have been developed corresponding to different constraints: Parsegian and Weiss (J. Colloid Interface Sci., 1981, 81, 285) and Roth and Lenhoff (J. Colloid Interface Sci., 1996, 179, 637). These different representations of the dielectric response lead to signicant differences in the van der Waals force calculated from Lifshitz theory. In this work, more recent and complete spectral data for water were compiled from the literature and direct integration of the Kramers Kronig relations was used to construct a new (i ) for water at 298 K. This approach also allows a number of different types of spectral measurements (such as infrared spectroscopy, microwave resonance techniques, and x-ray inelastic scattering) in the compilation of absorption data over a large frequency range (on the order of 8 to 10 decades in frequency). A KramersKronig integration was employed to construct the real and imaginary parts of (), (), and () for water from the different spectral measurements before calculation of (i ) from its integral denition. The resulting new (i ) is intermediate between the ParsegianWeiss and Roth Lenhoff representations of (i ), does not use a model, and treats the conversion of absorption data as rigorously as possible. We believe the (i ) from the present work is the most reliable construction for use in van der Waals force calculations using Lifshitz theory. The extension of the (i ) construction to other temperatures is also discussed. C 2000 Academic Press Key Words: water; dielectric function; KramersKronig; van der Waals; Lifshitz theory.

netic eld. On a macroscopic scale, the interaction becomes a many-body problem, where the electromagnetic eld is limited to specic frequencies dependent on the dielectric properties and geometry of the materials involved. The modern theory for predicting van der Waals forces in colloidal systems, Lifshitz theory, is a continuum theory derived from quantum electrodynamics (1, 2). The magnitude of the nonretarded van der Waals force is proportional to the product of the normalized differences of the dielectric permittivity, (), of the two interacting materials from that of the intervening medium, 13 23 , where
jk

j j

k k

[1]

1. INTRODUCTION

The van der Waals force is ubiquitous in colloidal dispersions and always attractive between like materials and therefore is one of the most common causes of dispersion destabilization. The van der Waals force originates from the interaction of the induced dipole uctuations of molecules via the electromag-

1 To whom all correspondence should be addressed at; Department of Chemical Engineering, Carnegie Mellon University, Doherty Hall, 5000 Forbes Avenue, Pittsburgh, PA 15213.

The product of the differences needs to be evaluated over the full spectrum of frequencies, which requires extensive characterization of all the materials. Due to absorption, the dielectric function, (), takes on complex values for real frequencies and is a highly nonmonotonic function of frequency. Using the mathematical principles of analytic continuation, this nonmonotonic and complex-valued function of real frequency, , can be mapped into a monotonic and purely real function of imaginary frequency, i . The use of Lifshitz theory did not become prevalent until Parsegian and Ninham (3) and others (4, 5) developed a model for the dielectric function evaluated at imaginary frequencies, (i ) (see below), from available absorption spectra, by tting to a series of damped harmonic oscillators. Parsegian and Weiss (PW) (6) made detailed constructions for several substances (including water) that have been widely used in calculations of van der Waals forces. Roth and Lenhoff (RL) (7) have more recently presented a newer dielectric representation for water by retting the same data used by Parsegian and Weiss. There are signicant differences between the force calculated using these different representations for water; for example, the Hamaker constants for two polystyrene half spaces across water using the PW and RL constructions of (i ) for water are 1.41 and 0.897 1019 J, respectively. The prevalence of water in colloidal and biological systems makes accurate values for its dielectric properties critical in applications ranging from radiation transport in the atmosphere (8) to measuring the water content of soils (9), as well as in the calculation of van der Waals forces in these systems, and motivates the present study.
0021-9797/00 $35.00
Copyright C 2000 by Academic Press All rights of reproduction in any form reserved.

351

352

DAGASTINE, PRIEVE, AND WHITE

TABLE 1 A More Modern Data Set for the Adsorption Spectra of Water Compiled From the Literature (1924)
Frequency region Microwave IR Visible UV Quantity measured () and () k () n () and k () Im(1/ ()) Spectroscopic method Wave guide resonance tech. FTIR FTIR and refractometry Inelastic X-ray scattering

and n are the sampling frequencies of the dielectric function, k (i ). The prime indicates that the rst term in the summation has half weight. The expression for the n equals zero term in the presence of electrolyte is given in A.1. The dielectric permittivity, (), evaluated at a real frequency, , has a real and imaginary part denoted by and respectively, viz., () = () + i (), [4]

Historically, in the use of harmonic-oscillator models to construct the dielectric spectra of water it was necessary to interpolate in regions where no experimental measurements were available. Pashley (10) compared the use of the integral denition of dielectric function to the use of harmonic-oscillator models, but the available spectral measurements were sparse and the analysis did not account for errors introduced by using different types of spectroscopic measurements. Recently, a more complete spectral data set has become available for water over a large enough frequency range for us to construct an accurate dielectric representation for water directly from its integral definition. Table 1 shows the data used in the present study and the frequency regions of the data sets. In the present work we have integrated different spectral measurements over the largest frequency range possible to construct (i ) without the use of models by employing an iterative approach to calculate the dielectric permittivity via the appropriate KramersKronig (KK) relations as discussed below.
2. EXPRESSIONS FOR VAN DER WAALS INTERACTION AND DIELECTRIC PERMITTIVITY

and is related to the memory function for the polarization of the material over all prior times (13). The real part is related to the transmission characteristics of the material, while the complex part is related to the absorption characteristics. Dielectric permittivity is related to other experimentally measurable quantities including refractive index, n (), and absorption coefcient, k (), () = n 2 () k 2 () () = 2n ()k (), [5]

loss function (the imaginary part of 1/ ()), where 1 = () and reectance = R 1/2 ei = (n () 1) + ik () , R (n () + 1) + ik () [7] ()2 () () i , 2 2 + () () + ()2 [6]

The retarded interaction energy per unit area between two innite half spaces, 1 and 2, separated by an intervening third medium, 3, is (1, 11, 12) E 123 ( L ) = where the Hamaker function is 3 A123 ( L ) = kT 2
n =0 rn

where R is the magnitude of the reectance and is the phase angle. Each of these quantities is related to its conjugate pair through a KramersKronig (KK) relation, which is a mathematical expression of the principle of causality applied to the linear response theory that connects polarization to applied eld. For example, the real and imaginary parts of () are related by (14) () = 1 + 2 P

A123 ( L ) , 12 L 2

[2]

dx
0 0

x (x ) x 2 2

x ln{[1

13

23 e

2 () = P ]

x 2 ( (x ) 1) dx , x 2 2

[8]

jk

[1 13 23 ex ]} d x , sk s j j sk k s j = , jk = , s + s s j k k j k + sj

[3]

where P denotes the principal value of the integral. If one of the components of () is available on 0 < < , it is possible to calculate the other component by means of this relation. The quantity (i ) used in Lifshitz theory can also be expressed in this fashion, viz. (10) (i ) = 1 + 2

2 = x2 + sk

2 n L 2 ( k 3 ), c 2 L n 3 2 nkT , n = , rn = c h

dx
0

x (x ) . x2 + 2

[9]

k (i n ),

where c is the speed of light in a vacuum, k is the Boltzmann - is Plancks constant divided by 2 , constant, T is temperature, h

If () is available on 0 < < , then (i ) can be calculated from [9]. When absorption data over the entire frequency regime is not available, oscillator models have been used to construct approximate representation for (i ).

WATER DIELECTRIC FUNCTION APPLICATION TO VAN DER WAALS FORCES

353

3. DAMPED HARMONIC OSCILLATOR MODEL: RESOLVING TWO FITS TO THE SAME DATA

Parsegian and Ninham (15, 16) were the rst to t a series of damped harmonic oscillators to piecewise spectral data of several types in order to model the complete polarization response of a material. The general form for () is
n

() = 1 +
k =1

dk + 1 i k

n 2 j =1 j

fj , [10] i g j 2

where the rst sum consists of Debye terms to account for dipole moment relaxations (dk is the peak height and k is the relaxation time). The second sum consists of several damped harmonic oscillators ( f j is the oscillator strength, j is the resonance frequency, and g j is the bandwith). The analytic continuation of this expression to imaginary frequencies, i , is
n

FIG. 1. The real part of the dielectric permittivity, (), in the UV and visible region. The circles are the actual experimental data tted by Parsegian and Weiss and by Roth and Lenhoff.

(i ) = 1 +
k =1

dk + 1 + k

n 2 j =1 j

fj ; + 2 + gj

[11] the UV region. RL (7) relaxed this constraint to obtain a better t to the UV region, while keeping the same IR and microwave peaks. The results of both ts are compared to the experimental data used by PW and RL in Figs. 1 and 2. It is clear that forcing agreement with refractive index data in the visible region has a deleterious effect on the quality of the t to the data in the UV region. It is interesting to note that using the oscillator model approach for water employs six peaks in the UV region, with 18 parameters, but obtaining a denitive t to the spectral data is still difcult. The motivation for our revisiting the water spectral data is their importance in the calculation of van der Waals forces. As we will now show, improving the t in the UV region at the expense of other regions improves predictions of the nonretarded Hamaker constant at the expense of predictions of retardation effects (11). Equation [3] contains a sum over an innite number

() and (), calculated from [10], and (i ), calculated from [11], satisfy the KramersKronig relations [8] and [9]. PW (6) and RL (7) t the damped harmonic oscillator model form for () and () to the same experimental data (see Table 2). The data required to calculate () and () are not complete: e.g., the experimental absorption coefcient, k (), was available in the IR region, but n () was not. Both n () and k () are required to calculate () and () from Eq. [5]. Thus, it was necessary (e.g., see (17)) to calculate refractive index from a KK integration in which data (beyond the experimentally measured region) were either extrapolated or neglected. A similar problem occurs in the UV region where the phase angle, (), is required to convert reectance data, R (), to () and () via Eqs. [5] and [7]. For example, in the highfrequency region (where no experimental data were available) the data were extrapolated, causing errors in () of 20% at some frequencies (18). This increases the possibility of error in the () data beyond experimental error in the measurements. To partially compensate for these errors, PW applied the constraint that () must match experimentally measured refractive index data in the visible region where k () is approximately zero and () = n 2 () (6). The resulting model has one microwave term, ve oscillator terms for the IR region, and six terms for
TABLE 2 The Adsorption Data Set From the Literature (4, 17, 18, 28), Fitted by Both Parsegian and Weiss (PW) and Roth and Lenhoff (RL)
Frequency region Microwave IR Visible UV Quantity measured () and () k () n () and k () R () Spectroscopic method Dielectric dispersion tech. IR IR and refractometry reectance

FIG. 2. The imaginary part of the dielectric permittivity, (), in the UV and visible region. The circles are the actual experimental data tted by Parsegian and Weiss and by Roth and Lenhoff.

354

DAGASTINE, PRIEVE, AND WHITE

FIG. 3. The contribution to the van der Waals interaction from the different frequency regimes for two polystyrene half spaces across water with a 0.5 mM salt as a function of separation distance.

FIG. 5. The Hamaker function for two water half spaces across a vacuum as function of separation distance using the representations for (i ) by Parsegian and Weiss, Roth and Lenhoff, and the one developed for the present work.

of frequencies, but each frequency region (IR, visible, and UV) does not contribute equally to the sum (11). Moreover, the relative importance of each range varies with the separation distance. Figure 3 shows the percentage contribution of the summation in Eq. (3) from the zero-frequency term, IR, visible, and UV regions for the interaction of two polystyrene half spaces across water with 0.5 mM sodium chloride using the PW representation for (i ) (the curves are virtually identical if the RL representation is used instead). The UV region provides the largest contribution to the sum at small separations but as the separation distance increases the other spectral regions become more important. This consideration is important in deciding which dielectric representation to use in van der Waals calculations. The effect on the calculated van der Waals forces of using the different dielectric representations over a large range of separation distances is illustrated in Fig. 4 for two polystyrene half

spaces across water with 0.5 mM of a binary electrolyte (with a negligible effect on the refractive index of the solution) added to account for some screening of the interaction, since this is the case for real systems. The material choice can be somewhat arbitrary for this illustration; polystyrene was chosen because the dielectric function is well characterized (6) and it is a prevalent material in colloidal dispersions. Regardless of material choice, it is important to compare the effect of the dielectric function at different separation distances, not just zero separation, A(0). From Figs. 1 and 2 the PW (6) representation is a better t in the visible region. Accordingly, the van der Waals calculation using the PW representation should be more reliable at large separation distances, where Fig. 3 shows that the visible region is more important then the UV. Similarly, the calculation which used the RL t (7) should be more reliable at shorter separation distances because that dielectric representation has a more accurate t of the UV region. A second illustration is provided in Fig. 5, where innite half spaces of water are separated by a vacuum. While this example may not be as pertinent for experimental dispersions, it is a comparison between the Hamaker functions without the need of choosing a third material that may have some error associated with its dielectric spectra. Again, the differences in (i ) as a function of frequency correspond to differences in the A( L ) as a function of separation distance.
4. NEW DATA SET AND THE KRAMERSKRONIG INTEGRATION

FIG. 4. The Hamaker function for two polystyrene half spaces across water with 0.5 mM salt as a function of separation distance using the representations for (i ) by Parsegian and Weiss, Roth and Lenhoff, and the one developed for the present work.

Since 1981, spectral data for water have become available over a larger, more complete frequency range. Our approach is to construct () and () directly from their KK relations without the use of a model like [11] for interpolation. The experiments summarized in Table 1 span the frequency range from 108 to 1018 rad/s. In the microwave region, () and () are directly

WATER DIELECTRIC FUNCTION APPLICATION TO VAN DER WAALS FORCES

355

available from wave guide spectrometry (19, 20). Fourier transform infrared spectroscopy yields the absorption coefcient in the infrared and visible regions, (21, 22), which is supplemented by refractive index measurements in the visible region (23). The loss function was measured in the UV region in a recent inelastic X-ray scattering study by Hayashi et al. (24). In particular, this last study extends the experimental data to 200 eV, well beyond the 25 eV available in 1981. Above 200 eV, the absorption of water is virtually zero. The experimental error ranges from 1 to 3% over most of the range. The far UV and X-ray regions may have somewhat larger errors (around 510%), but the overall magnitude of () or loss function data are already near zero at these frequencies. Construction of (i ) from the integral denition in Eq. [9] requires conversion of several different spectroscopic data into (). In the IR region, the absorption coefcient, k (), was measured, but the values of n () reported by (21, 22) were calculated from the KramersKronig integral relation between n () and k (). Like the other KK integrations, this one requires k () over the entire frequency range; outside the region where k () was measured, k () was either extrapolated or neglected. Thus the n () data reported by them are only estimates. The possible error in the values of n () makes conversion of the IR spectral data to () via Eq. [5] difcult. The same difculty is encountered in the UV region where the loss function was measured (24) and the Re(1/ ) is required to convert the absorption measurement to (). Because of the importance of the UV region in the calculation of van der Waals forces, great care was taken to convert Im(1/ ) to (), which we will now describe. In the iterative procedure all of the spectral data are converted to one quantity, in this case Im(1/ ()), and then a KK relation, Eq. [14], is used to calculate its counter part, Re(1/ ()). This process is repeated and the errors from converting different measured are removed by using the integration result to rene the conversion of the different types of spectral data to one quantity. The iterative procedure (described below) for obtaining Re(1/ ) and Im(1/ ) is summarized by Fig. 6. Im(1/ ) is directly measured in the UV by dielectric loss experiments. In the IR and

VIS, Im(1/ ) is calculated from the absorption coefcient and refractive index: Im 2nk 1 = 2 . () (n + k 2 )2 [12]

One complication is that experimental measurements of the refractive index are only available in the VIS; in the IR, n must be calculated. In the rst iteration we use the values of n calculated by (21, 22); in subsequent iterations, we use the value calculated from Re(1/ ) (see Eq. [15]). In the microwave, Im(1/ ) is calculated from measurements of () and (): Im 1 = () () . ()2 + ()2 [13]

Having Im(1/ ) for all frequencies, we are in a position to calculate Re(1/ ) from the KK relation for 1/ (); Re 2 1 = 1+ P ()

dx
0 0

x Im 1 (x ) , x 2 2

2 1 = P Im ()

x 2 (Re 1 (x ) 1) dx . x 2 2

[14]

Care must be taken in performing the integration of Eq. [14]. Details of the numerical integration procedure are given in A.2. With Re(1/ ) known, we can calculate new values for n in the IR from Re 1 () = n2 k 2 , (n 2 + k 2 )2 [15]

FIG. 6. Flow diagram for the iterative integration of loss function. The analogous procedure for () was used over the entire experimental frequency range.

using experimental values for k (). Having new values for n () in the IR, we calculated all the other quantities above and iterated until n () in the IR and Re(1/ ) converged to fractional change less than 0.01. The values of n () were compared to independently measured experimental values for refractive index in the visible region as a consistency check of the integration as well. It was immediately evident that using the new UV data in this way produced a systematic over-prediction of n () compared to the experimentally measured values in the visible region. A sensitivity study of the integration showed that the loss function in the UV region determines the values of Re(1/ ) in the visible region, and therefore n () as well. A systematic error in the loss function data will cause a systematic error in Re(1/ ) in the visible region. A detailed analysis of the inelastic x-ray scattering data and private communications with the investigators showed that the likely cause of the error in the UV data was the normalization of the scattering results (the area under the inelastic x-ray scattering structure factor was scaled to a theoretical value predicted from a quantum mechanical argument (24, 25)). In our reanalysis of the UV data the area under the loss function was renormalized by scaling the area with a constant chosen so that

356

DAGASTINE, PRIEVE, AND WHITE

pling frequency with a corresponding n value from Eq. [3] of 1183 for T = 298 K). Since the summation in Eq. [3] may require one to two thousand terms for nonconducting materials, one may require (i ) beyond this region. Note from [9] that, for large (26), (i ) 1 + p
2

[17]

where the plasma frequency, p is given by


2 = p

x
0

(x ) d x .

[18]

FIG. 7. The real and imaginary parts, function for water.

() and

(), of the dielectric

the KK integration gave results consistent with the n () measured in the visible region. The constant scaling factor resulted in a 10% change of the area under the loss function curve, which is not unreasonable with regards to the vagaries of the original scattering data analysis. With this rescaling procedure, the position and relative size of the absorption peaks in the UV region were not affected. The iterative procedure for calculating Re(1/ ) from Im(1/ ) was accurate at higher frequencies, but the magnitude of 1/ in the microwave region is on the order of the experimental error and integration errors. The behavior of Im(1/ ) in the microwave region had a negligible effect on the integration results at higher frequencies, so convergence of the 1/ integration was only required in the visible, near IR, and UV, regions. The results of the KK integration on 1/ in the UV region were converted via relations Eqs. [5] and [7] into (), as were the k () data in the IR region. Then, including the () microwave data, an analogous iterative procedure was performed on () and () over the entire experimental frequency range. This was continued until () gave the same results for consecutive iterations over the entire frequency range. A nal consistency check was used after iteration to verify that the () data did cover the physically signicant frequency range, viz., (0) = 1 + 2

A t of the calculated (i ) values around 2.9 1017 rad/s to the form of Eq. [17] yields p = 3.48 1016 rad/s comparing favorably with the analytic result of Eq. [18]: p = 3.68 1016 rad/s. Obviously the asymptotic [17] form has set in before the cutoff in the numerical calculations and may be used for frequencies greater than 2.9 1017 rad/s. The (i ) from the present work is compared to (i ) from the PW and RL constructions in Fig. 8. The present work is closer to the PW construction in the visible and IR and closer to the RL construction in the UV region. The overall deviation of the present work from the other two representations is due to the use of the direct integration technique rather than curve tting spectral data to an oscillator model and iteration of a newer, more extensive data set. The calculation of the van der Waals energy with the new (i ) representation for water produces a result intermediate between the calculations using the PW and the RL representations for (i ) as shown in Fig. 4. The agreement of the van der Waals calculation using (i ) from this work with that using the PW construction at large separations is consistent with the agreement of the (i )s in the visible and IR spectral regions. Similarly the agreement in the calculation of the van der Waals energy

dx
0

(x ) , x

[16]

where (0) is the dielectric constant. The real and imaginary parts of () shown in Fig. 7 are the results of this iterative melding of the disparate data sets.
5. (i ) AND DISPERSION CALCULATIONS

The construction of (i ) from the () integration results follows from Eq. [9] using a linear interpolation between data points integrated on a logarithmic scale as outlined in A.2. We calculate (i ) to an upper frequency of 2.9 1017 rad/s (a sam-

FIG. 8. The value of (i ) from the works of by Parsegian and Weiss and Roth and Lenhoff compared to the present work. The value for n corresponds to sampling frequency from Eq. [3].

WATER DIELECTRIC FUNCTION APPLICATION TO VAN DER WAALS FORCES

357

between the present work and that using the RL construction for closer separation distances is consistent with the agreement of the (i )s in the UV region. We see the same type of agreement for the two water half spaces separated by a vacuum in Fig. 5. The new spectral data set for water and the unbiased construction of (i ) in the present study make the present calculation the most reliable of the three.
6. TEMPERATURE EFFECTS

The most accurate way to calculate (i ) at other temperatures is by the use of extensive absorption measurements at those temperatures, but we offer an alternative method (which is not as rigorous as the approach outlined above) to estimate how (i ) will change with temperature when these data are not available. The dipole moment relaxation in the microwave region is a strong function of temperature due to hydrogen bonding, while the electronic and molecular vibrations at higher frequencies are much less temperature dependent. The rst sampling frequency for the van der Waals force calculation from Eq. [3] is well beyond the microwave region, so the effect of temperature change has only a small impact on the van der Waals force. For the higher frequency region, the ClausiusMossotti theory (27) is a widely accepted mean eld theory approximation that correlates molecular polarizability to bulk properties such as density, viz., (, T ) 1 4 (, T ) (T ) = , 3 (, T ) + 2 [19]

FIG. 10. An illustrative calculation of the Hamaker constants, A(0), for two temperature-insensitive PS half spaces and two temperature-insensitive silica half spaces, both separated by water with 0.5 mM NaCl over a range of temperatures. The dashed lines correspond to the silica Hamaker constant.

The approach in this work was to utilize this scaling with density to estimate (i ) at other temperatures. The low-frequency () and () data were t to the Debye model from Eq. [10], () = 1 +
o1 , 1 i

[20]

where o is the dielectric constant and is the dipole moment relaxation time. The low-frequency data were then subtracted from the (i ) data set according to the form (i ) = 1 +
o1 1 +

where (T ) is density. Using available data for the refractive index in the visible region as a function of temperature (23) and the temperature dependence of density (23) we have calculated (, T ) as shown in Fig. 9 at three wavelengths. There we note that molecular polarizability is nearly constant over a large temperature range thus verifying the ClausiusMossotti scaling of (, T ) with density (T ).

[21]

to yield a microwave-free (i ) spectrum for water at 298 K. Hopefully, the microwave-free data set can be temperature corrected by the density scaling of the ClausiusMossotti formula (i ) 1 (i ) + 2 =
Tnew

(Tnew ) (Tref )

(i ) 1 (i ) + 2

,
Tref

[22]

FIG. 9. Molecular polarizability at three wavelengths in the visible region, 404.66, 589.32, and 706.52 nm.

where Tref is the reference temperature, 298 K, and Tnew is the new temperature. Once the microwave-free (i ) is constructed at the new temperature the Debye model can be used to add the microwave relaxation back, given that the parameters o (Tnew ) and (Tnew ) are available from the literature (20, 23). A data le with sampling points at n (T = 298.16 K) for a microwave-free (i ) in the form of ( (i ) 1)/ ( (i ) + 2) is available from the authors. To illustrate the effect of temperature on the van der Waals force, the Hamaker constants, A(0), were calculated over a range of temperatures for two PS half spaces across water with 0.5 mM NaCl and two fused silica half spaces across water at the same salt concentration in Fig. 10. This is purely for illustrative purposes. The temperature dependence of (i ) for water was taken

358

DAGASTINE, PRIEVE, AND WHITE

into account according to the procedure outlined above, but the (i ) construction for PS was taken from PW and treated as a temperature-independent property. The construction of (i ) for silica (from (26)) was also treated as temperature-independent. Even though these are not a strictly realistic calculation they do provides some insight into the importance of temperature dependence in van der Waals forces across water.
7. CONCLUSIONS

consecutive data points containing the singularity were t to a quadratic form where the principle value integral is more sensitive to interpolation and experimental error. The regions beyond the measured data were t to asymptotic tails based on a Debye term from Eq. [10] at low frequencies and a high-frequency form of the oscillator term from Eq. [10] at the high-frequency end of the data. There was no noticeable effect on the results from evaluating the asymptotic tails, but they were included for completeness.
REFERENCES
1. Lifshitz, E. M., Sov. Phys. JETP 2, 73 (1956). 2. Dzyaloshinskii, I. E., Lifshitz, E. M., and Pitaevskii, L. P., Adv. Phys. 10, 165 (1961). 3. Ninham, B. W., and Paregian, V. A., J. Chem. Phys. 52, 4578 (1970). 4. Gingell, D., and Parsegian, V. A., J. Theor. Biol. 36, 41 (1972). 5. Nir, S., Rein, R., and Weiss, L., J. Theor. Biol. 34, 135 (1972). 6. Parsegian, V. A., and Weiss, G. H., J. Colloid Interface Sci. 81, 285 (1981). 7. Roth, C. M., and Lenhoff, A. M., J. Colloid Interface Sci. 179, 637 (1996). 8. Irvine, M. W., and Pollark, J. B., Icarus 8, 324 (1968). 9. Zolotarev, V. M., Mikhailov, B. A., Aperovisch, L. I., and Popov, S. I., Opt. Spectrosc. 27, 430 (1969). 10. Pashley, R. M., J. Colloid Interface Sci. 62, 344 (1977). 11. Mahanty, J., and Ninham, B. W., Dispersion Forces, Academic Press, London, 1976. 12. Prieve, D. C., and Russel, W. B., J. Colloid Interface Sci. 125, 1 (1988). 13. Landua, L. D., and Lifshitz, E. M., Electrodynamics of Continuous Media, Pergamon, Reading, MA, 1960. 14. Bassani, F., and Altarelli, M., in Handbook of Synchrotron Radiation (E. Koch, Ed.), p. 463, Elsevier, New York, 1983. 15. Parsegian, V. A., and Ninham, B. W., Nature 224, 1197 (1969). 16. Parsegian, V. A., and Ninham, B. W., Biophys. J. 10, 664 (1970). 17. Kislovskii, L. D., Opt. Spectrosk. 7, 201 (1959). 18. Heller Jr., J. M., Hamm, R. N., Birkhoff, R. D., and Painter, L. R., J. Chem. Phys. 60, 3483 (1974). 19. Schwan, H. P., Sheppard, R. J., and Grant, E. H., J. Chem. Phys. 64, 2257 (1976). 20. Kaatze, U., J. Chem. Phys. 108, 4545 (1989). 21. Bertie, J. E., and Zhida, L., Appl. Spectrosc. 50, 1047 (1996). 22. Hale, G. M., and Querry, M. R., Appl. Opt. 12, 555 (1973). 23. Lide, D. R., (Ed.), Handbook of Chemistry and Physics, 74th ed., CRC Press, London, 1994. 24. Hayashi, H., Watanabe, N., and Udaawa, Y., J. Chem. Phys. 108, 823 (1998). 25. Wang, J., Tripathi, A. N., and Smith Jr., V. H., J. Chem. Phys. 101, 4842 (1994). 26. Hough, D. B., and White, L. R., Adv. Colloid Interface Sci. 14, 3 (1980). 27. Born, M., and Wolf, E., Principles of Optics. 5th ed., Pergamon, Reading, MA, 1975. 28. Buckley, F., and Maryott, A. A., Tables of Dielectric Data for Pure Liquids and Dilute Solution, U.S. National Bureau of Standard Circular 589, Sup. of Documents, Washington, D.C., 1958.

The present study provides the most accurate construction of (i ) for water along with () and () over the widest frequency range available. While not as easy to use computationally as the oscillator model used by previous workers, the calculation of (i ) can be reduced to interpolation of an (i ) data le for 2.9 1017 rad/s and the evaluation of a simple formula [17] for > 2.9 1017 . A copy of that data le with sampling points at n (T = 298.16 K) is available from the authors, as well as a data le for use at temperatures other than T = 298.16 K (http://www.cheme.cmu.edu/jcis/).
A. APPENDIX

A.1. Zero Frequency Term The expression for the n = 0 term in Eq. [3] with the presence of electrolyte is, from (11), 3 A(0)123 ( L ) = kT 4 jx j3 = jx +
0 3s 3s

x ln{[1

13

23 e

]} d x [23]

s 3 = x 2 + 4( L )2 ,

where is the Debye length for the system. Terms for n > 0 are unchanged. A.2. Integration The KK relation was integrated on a logarithmic scale to properly weight the contributions from the 10 decades of frequency spanned by the data. The singularity of the integrand at x = in Eq. [14] makes numerical integration of the discrete data set difcult, so an interpolation form was assumed between adjacent data points. The integral was evaluated analytically using the form of the interpolation and summed over the entire frequency range. The integral was divided into three parts centered around the singular point. A linear interpolation form was employed in the regions before and after the singularity. The three

You might also like