Modernization of the van Deemter Equation for Chromatographic Zone Dispersion

Stephen J. Hawkes Oregon State University, Corvailis, OR 97331 In the mid-1950's, a group of Dutch chemical engineers undertook a study of zone dispersion in chromatography. Starting with the equations of Lapidus and Amundson ( I ) they made some approximations and assumptions and arrived a t the equation ( 2 ) Svmbols Used
A = eddy diffusion or anastomosis term B = coefficient of longitudinal diffusion e = concentration C = coefficient of mass transfer d = diameter. or deoth D = coefficient of m.nlecular diffusion end = end~effects f = Giddings compressibility factor 9(pfl: - l)(p?lp2- 1)/8(p!lp~- 112 fn = function of h = column dispersivity ("plate height") i = dames-Martin comoressihilitv factor h = column capacity ratio t$t, = Ku,lu, K = partition coefficient, c$c, 1 = length of empty tube containing sample before injection L = column length p = pressure q = factor describing shape of liquid phase t = time u = linear velocity, crn of column traversed per second G = average linear velocity = Llt,
/J

This has come to be known as the "van Deemter equation." While this title involves some injustice to previous workers (see the review by Giddings) (3a),the naming was inevitable since the equation was produced during the early development of gas chromatography and was seized upon, used, and quoted in that dramatic period of chromatographic history. The van Deemter equation was the basis of the many improvements in column design that were realized as the various parameters were optimized. I t is now quoted in sophomore texts and physico-chemical research papers as the definitive equation. Often it is given in the simplified form

= ""lllrn~

11 i- ~~l~tvrrutiatc thnr it hay he;wne fossllii.cJ i n [hi- way. 1> hr euuation rt.prt.scntrd the I~r.>r rhlr c c ~ l d I,, - . i d 111 IlrLri. hut thk ensuingdecades have shown that chromatographic dispersion is better represented as
h = Blu t Cu
= Blu

Subscripts ads = adsorbed e = column f = film i = a t inlet of column 1 = liquid stationary phase m = mobile phase o = a t outlet of column p = particle s = stationary phase

(3)
(4)

+ (C, + C,)u

This form is as rigorously correct as current theory allows, is no more complex than van Deemter's form, and lends itself to discussion of the geometrical parameters in the last (C,u) term. It also avoids fossilizing a precise form of the Cmu term which is still poorly understood. The van Deemter equation omitted the C,u term because the columns of the period contained so much stationary liquid that Cmu was negligible compared to the first (C,u) term. However, once the equation was widely understood and more sensitive detectors became available, it became customary to use less stationary phase so as to reduce drand hence CI so that C, became relatively more important. From that period in the early 1960's it has been necessary to include C,u in discussions of zone dispersion. Usefulness of the Equation Values of the parameters in eqn. (5) are seldom known with anv precision. Its main usefulness is therefore in aualitative diicussions of the effect of various parameters on column dih , and the comoromises that must be made in s~ersivitv. column design and in controlling temperature and flow rate. Equation ( 5 ) can also be manipulated to give equations for resolution or analysis time. Column design, temperature, and flow rate may then be chosen to optimize the desired performance. The van Deemter equation has also been used to measure

Greek Svrnbols

y = obstruction factor
7 = viscosity X = proportionality factor A A d , d = fraction of the mobile phase that is outside the porous particles o = standard deviation of distances of molecules of a component from zone center o = proportionality constant C,D,ldt

DI and Dm,quite often by naively using eqn. (1). The use of eqn. (5) makes the problems in experimental design much clearer and shows that such measurements are unlikely to be reliable.
The Meaning of h

cessive equilihratio& in a series of "theoretical plates%epresenting successive segments of the column and called h the "height equivalent to a theoretical plate" or "HETP." This was a logical development of Martin's previous work on multiple liquid-liquid extraction, and the theory of distillation. They used the model to show that peaks should be Gaussian in the limit of a large number of plates, derived a relation between the number of plates and resolution, and Volume GO Number 5 May 1983
393

". ,, , . .

;;, : . ::,
:
,

have nractical sienificance but are not represented in eqn. (5) and it is a matter of pedagogical judgment whether they are better discussed in terms of a more detailed and complex equation or simply as qualitative variations of eqn. ( 5 ) .T h e full equation would he

. I I I _

; : \

The variance, c2, of me distribution of me molecules of a single substance about

showed that resolution varied with the square root of the column length. This was about the limit of usefulness of this model. I t has given us the term "plate height" symbolized by h. This is a most unfortunate inheritance because later models have proved more tractable, easier to comprehend, and less deceptive. The danger of the model is in the mental concept that it conjures. If a column consisted of a series of plates, then joining two columns together would produce a number of olates eaual to the sum of the dates in the separate columns. ,\n malyst using this conceln imd joining an efiit icnl n h m n 11, an intffirienl one i i dijappointed t o find rhnt the number .. of plates is reduced and the resolution is worse. Klinkenhurg and Sjenitzer introduced a better model in 1956 when they showed that h = cZ/L where oZis the variance of the distribution of the molecules of a single substance about their mean position in the column or thin layer plate (i.e., 2d21number of molecules, in the figure). and L is the distance the zone of substance has moved. In ~ u h u n n chromatt~yraphy.I. is usunlly the column lmgth. 'l'hii detinirim i j more u3eful than the plate height in every way and has rendered the plate concept obsolete^ Pedagogically it is superior because u2 is directly related to the important experimental phenomena i t describes, viz., the spreading of the zone, the width of the chromatographic peak and hence the resolution. Theoretical plates relate to these phenomena only indirectly in a manner needing mathematical treatment. Theoretically it is superior because i t relates directly to the underlying physical processes through the Einstein equation (38), u2 = 2Dt, where D is the dispersion coefficient and t is the elapsed time. This has formed the basis of the random walk and the nonequilihrium treatments of chromatography. The use of h = u2IL as a definition of h eliminates the conceptual error of additive plate numbers described above, but the name "plate height" or "height equivalent to a theoretical plate" or "HETP" perpetuate the error. The term "plate" should now be dropped. For the remainder of this paper h will be referred to as the "column dispersivity." Unfortunately, no such neat name has occurred to the author to replace the popular "plate number"
N
= L2/02

The A Term, "Eddy Diffusion" Much discussion and research has eventually shown that the A term is not useful in discussion of modern chromatography. The confusion surrounding it arises from several causes including conflicting definitions of the phenomena it represents, the ill-defined effect of diffusion on its magnitude, and the fact that in one of its definitions the effect is real at verv high velocities. The term "eddv diffusion" is used bv chemical engineers to describe a phenomenon related to turbulence, bu<turhulence is a tvne of flow that prohablv never occurs in chromatographii kactice. Van ~ e e m t e it r al. (2) used the term to describe a process occurring in pockets of mobile fluid between particles acting as mixing cells. Later workers used the term to describe the broadening of a chromatographic zone by the variety of short, long, fast, and slow paths by which molecules could migrate through the packing. Increasingly precise measurements of h showed that there was no measurable additive A term. Hence, the turbulence and the mixing cell models became untenable. The effect of the variety of flow paths remained and became known as the "multipath term" or as "anastomosis" (a term borrowed from physiology where it refers to the complicated interconnections of blood canillaries ( 4 ) ) . The mukipath model is necessarily correct. A variety of paths must spread out the molecules and increase h. However, Giddings showed in a series of papers (5-7) that the effect is offset bv diffusion and is therefore flow-dependent. Most studengfind this concept difficult so it is explained more fully as follows. In the absence of diffusion a molecule entering the column is locked into a sinale meandering flow path through the packing. Any other molecule enter;ng at the same point will follow the same path and all such molecules will take exactly the same time to reach the end of the column. Molecules enterin~ at a different point in the column cross-section will and take a longer or shorter time to follow a different leave the column. This variety of paths causes the molecules to disperse over a length of column which is independent of the mobile phase velocity. However when there is diffusion in the mobile phase, a molecule migrates from one flow path to another, spending only a short time in each. When the flow rate is very slow, every molecule samples every flow path a large number of times so that every molecule has a similar pattern of flow, and the molecules are not significantly dispersed by this multipath effect. Thus, the multipath term tends to zero as velocitv tends to zero. At very fast flow rates been represented by the equatidn

In default of a better name, I have used the name "plate number" in class and indicated that the name has no conceptual significance but is an accident of history. For the future then, the plate concept should be discarded and replaced by h = 021L or in non-uniform columns by h = du2/dL. Detailed Treatment The remainder of this paper will he devoted to a discussion of the reasons that eqn. (1) should nolonger be taught and that eqns. (3)-(5) should he preferred and to a discussion of the underlying phenomena in terms that can he transmitted to sophomores. There are some aspects of zone broadening which
394

often called the "coupling equation," a term suggested by Giddings. (5) This has often been reproduced in textbooks but it should not be. The equation gives presumably correct values at u = 0 and u = but there is much experimental data to show that it does not do so a t any point between these two extremes. Even the extreme values cannot be confirmed exnerimentallv because the eauation is based on the assumption hf laminar flow whereas at high velocities turbulence becomes imnortant. Moreover. the unpredictable dependence (12,13) of longitudinal diffusion on velocity makes low-velocity data

Journal of Chemical Education

uninterpretable. Thus, eqn. (7) gives no more information than the qualitative explanation that precedes i t but is deceptive because it seems to do so. There have been many attempts to provide better equations than (7) by Giddings (3b), by Knox and his co-workers (8-lo), and by numerous chemical engineers whose work is reviewed by Cluff and Hawkes (11). Knox was able to get good correlation between theory and exneriment for unretained samples, but no equation serves for realistic samples with realistic retention times. This work will not he reviewed here because the thesis of this section is that no quantitatme treatment of this concept should be presented to students of analytical chemistry. I t is of relevance to chemical engineers working with high velocity fluid streams through reactor beds but it makes no significant rontribution to zone hroadenine at the velocities used in - - - ~ ~ - chromatography. It was hoped &ring the 1960's that high velocities would he used in liauid chromatoerauhv - - - to obtain prepared fast results, and the discoverythat h - A as u the way for such work. However, the 1970's have shown that i t is more productive to concentrate on short, very efficient columns ouerated at velocities that uroduce minimum h. The muitipath effect should he presented qualitatively with no supporting equations to deceive by oversimplification. I t should he incorporated in the vague last term of eqn. (51, and will account for the variation of mobile uhase mass transfer C, (discussed later) with velocity. I t can,in fact, he discussed , rather than as a separate effect. under C

Liquid Stationary Phase Mass Transfer, Cl

Van Deemter et al. (2) derived the expression

The form of this expression is correct hut the factor 8/n2 is wrong. Moreover, CI increases when the liquid is coated on an a d ~ o ~ b esupport. nt Van Deemter's explanation for the 8/n2 was incomplete. It started with the equation
dcldt = (a2/4)(D~/d?)(c, - KCI)

The B l v Term, Longitudinal Diffusion .

The van Deemter (2) expression 2yDJu or its abbreviation Bju is essentially correct. There is both theoretical (3c) and experimental (12)evidence that the obstruction factory decreases at very low velocities, but the experimental evidence is contradictory (13). Since this term is well explained in textbooks, it will not he discussed here in further detail.
The Cv Term, Mass Transfer

where c, and c, were the concentrations in the gaseous and the liquid phases, and K was the partition coefficient. The equation was said to he derived from "the simplified solution of the general eauation for diffusion into a laver" auuarentlv assuming that "the average concentration (in the layer) does not differ too much from the concentration at the surface." No reference was supplied. Diffusion into a layer is discussed at length by Crank (14) and the equations he supplies do involve n2, hut they donot obviously lead to the above equation. Even so, there would be no reason to distrust the 8/n2 had not two separate workers, Giddings ( 3 4 and Golay (15) hoth derived the constant as 2/3. Giddings derived the 2/3 from nonequilihrium theory, and Golay derived it from an electrical analog using the telegraphers equation. They hoth obtained and comthe constant %. ." As hoth derivations are c o m ~ l e t e ~ ~ prehensihle, the 2/3 must he accepted as a correction of van Deemter's 81x2 (unless a full derivation of ean. 191 is ever published). For a uniform film then

However, the liquid film is never uniform. I t is held in the

The mass transfer term arises because eauilihrium between the stationary and mohile phases is not reached instantaneously. Consequently, material in the mobile phase tends to he swept forward ahead of the average position of the zone while material in the stationary phase lags behind. Similarly, material in fast flow paths is swept ahead of material in slow paths. The more quickly a system tends toward equilibrium the smaller this effect will he. Specifically, the effect is reduced hy fast diffusion in the mobile phase moving material from one flow path to another and to the stationary phase, and by fast diffusion in a liquid stationary phase moving material to the interface with the mohile phase. If these processes were infinitely fast, the material in the stationary phase would not lag behind that in the mobile phase and there would he no spreading from this cause. The faster the mobile phase moves the worse this effect becomes. so the effect is urouortional to the velocity and the term is usually written a s - ~ u : ~ h discussion e of the k l t i p a t h term ahove surrests that this strict uroportionality will cease at high velocities in packed columns bkcause thepart of Cu that is due to a variety of flow paths will be reduced to the constant term A. This deviation is probably negligible over the range of velocities in which chromatography is actually practiced and should be ignored in elementary treatments. I t is usual to divide the C term into two parts representing mass transfer in the stationary and mobile phases thus The stationary phase mass transfer C, may he on an adsorbent or in a liquid stationary phase, but as simple adsorptiondesorption kinetics are very fast. They make a negligible contribution to C,. We shall therefore discuss only mass transfer in the liquid phase, CI.

Furthermore, even if the support surface were uniform i t would support only a few monolayers of uniform film and the rest would run out of the column. Giddings has allowed for the non-uniformity of the liquid layer by replacing the Zl3 with a "shape factor" q; thus (3e),

where dc is now the debth of the non-uniform film a t its the non-uniform film. It normally varies from 2/ls f i r sphe;ical particles (as in a bead of resin where df is the radius) tozjs for a uniform film. I t is possihle (3f) to conceive shapes that give q outside the range 2/1:, to 2/1. When the liquid is coated on smooth glass beads and accumulates a t the contact points, then y = % 2 , and if it were held in pores that were wide at the to these unlikely circumstance^. When there is a variety of different shapes of pores, then q is the mean of the individual values of q for the various types of pore, weighted by the volume of liquid in each type of pore i3e)
.

s=x s j
i .

'J j
U,
~

(121

where uj is the volume of stationary phase in pores with shape q j , and ul is the total volume of stationary phase. Thus if a small fraction of the liquid accumulates in badly shaped blobs, the effect on h is small K ~ ~ . v f tilere r , i: \I.I ,111uther ~ . ~ , I I I I > ~ I ~ ...\ II~ .uri~t,.e .,I~.
>upp,,rti~~: .I - l : ~ l i ~ m . Iiqu.~l ~ r \ II~II.I~ i w e s w r ~ IN l v td<t,rhenl ~~nhvrn.i.t t l lirlud ~ nil1 l i l ~ tw r t i t , . S , ~ n i ~ ~ I~ lI c L , I ~ C. i II nijy I~

Volume 60

Number 5

May 1983

395

diffuse through the liquid and be adsorbed a t the bottom of it on the support surface. The liquid phase nonequilibrium is then very much increased over simple solution in the liquid. Giddings has derived a correction factor (3g) to allow for adsorption below a uniform film, but there is no suitable expression for non-uniform films. I t is evident from his derivation that a multiplying factor that is a function of the ratio of adsorbed to dissolved sample c,d,/ci must be used. For a uniform film the multiplying factor is 1 2/(1 c,/c,d,) 3(c,dJc/cl)/(l cl/c,d,). This function approaches 3C,d$~/cl when cad, >> cl so the effect can be dramatic. The important practical consequence is that support surfaces should he no more adsorbent than is necessary to allow the stationary phase to wet them. A final complication is that all the above equations are derived assuming that the liquid surface is in direct contact with flowing mohile phase. When it is held in the pores of particles, the sample must pass through stagnant mohile phase in these pores before it reaches the stationary phase. In this case, CI is increased ( I 7) by a factor f l

This same logic shows that in an open tubular column C, d2 and Golav (15) and Giddinas (3h) have shown by different approaches that for a round column
a

..

fl

for calculations. However, it must not he pushed too far. I t assumes a parabolic flow profile which is true only with perfectly smooth walls and will be wildly inaccurate for 0.005-in. steel columns, judging from Drew and Bens' micrograph (16). The various flow vaths in a column are inevitably out of equilibrium with each other, having different concentrations in each flow stream. The disequilibrium is offset by diffusion between the flow streams, and C, is, therefore, proportional to the sauare of the distance between them. The distance is usually H few particle diameters and is assumed to he proportional to d,. This is again written However, the assumption is only approximate. I t assumes that as the particle diameter is reduced, the arrangement of the particles (the "packing structure") remains unaltered. This is not realistic. Even identical tubes and particles can he packed only moderately reproducihly. Changes in particle diameter will make further changes in the structure. Moreover, the very small particles now used in HPLC are packed by a different technique ("slurry packing") from larger particles with the intention of producing more uniform packing. T o complicate matters further, the packing structure changes radically near the wall of the column so that to maintain a uniform oackine structure as d, chances, the column diameter d, must change in direct proportion. This does not happen in oractice so C,...deoends in a comvlex and poorly understood . has done much to elucimanner on d, and d,. Knox's date this prohlem (8-10) hut the resulting equations fail when applied to retained solutes. The packing structure is also strongly dependent on the uniformity of the particle diameters so that meticulously size-graded particles give lower C, than packings with a wider range of particle sizes (211.

mass of sample dissolved (or adsorbed) total mass dissolved (or adsorbed) and in stagnant mohile phase his correction factor is usually close t o unity and i t can he omitted from elementary discussions.

Moblle Phase Mass Transfer, C, The complexities of the mohile phase in a packed column have so far defied exact analvsis. By contrast, the causes of in all branches of chroprovements in column matography. Briefly, h increases with the square of the particle diameter d;, the square of the column diameter d:, the ratio of d, to the diameter of the coil of a coiled column, and is affected by the geometry of the column, the uniformity of the vackinp, and the efficiency of the connections from the c o l n k to the injection system and the detector. Particle Diameter, d, There are several reasons why the particle diameter is im~ o r t a n tthese : are related to starnant mobile phase within the parrick 2nd to the tlmv patrcrn III.IU.F~II the pi~rtia.l~i. A t ro.lilihriun~ t11ccmccn1r;iti~m o r .anil~lt.in the Iluwiny mohilebhase would be the same as the concentration in the stagnant mohile phase held in the pores of the particles. In a real column this equilibrium cannot he achieved. Molecules in the pores lag behind those in the flow stream for a time that depends on the distance they must diffuse to leave the particle. This distance is proportional to the particle diameter dp and thus C, a d;. For spherical particles the dependence is given (3h) by

Column Diameter, d,

any but the most advanced &dents since this term is only one of several C, terms, most of which cannot he analyzed exactly. However, the qualitative concept C, a d i bas practical importance and should he taught at all levels. The particle must he reached by diffusion from the flow stream. The greater the distance between the particles, the greater is the width of the flow stream through which the sample must diffuse. This distance is proportional to the di-

experimentally that in a column packed with the light diatomaceous earth "Celite" there is a velocitv gradient across .the column, i.e., the gas velocity is faster on one side of the column than on the other and varies uniformly across the column. Using this model be showed that C, varied with d: and with the square of the slope of the velocity profile. He further showed that a great deal of previously published data was compatible with this model. Other workers have found different flow profiles. Huyten et al. (23) found a symmetrical profile in a 3-in. diameter column with fastest flow at the wall. This gives C , a d:. The parabolic flow profile in open tubular columns also leads to C, = da as shown above. I t is unlikely that trans-column variations are so imvortant in modern HPLC columns with the advanced technology now used. Indeed. it is common exoerience that ' I ~ i n .diameter columns givebetter efficiencythan '18-in. in HPLC making the exponent of d, negativd. Coil Diameter When a column is coiled, the flow path on the inside of the coil is shorter than on the outside: as both have the same pressure drop the flow is faster on the inside and gives rise to trans-column noneauilibrium. I t is for this reason that liauid chromatography coiumns are usually straight. The quanti-

iBoften written

C ,

= wd;/D,

where w is an ill-defined constant.

Journal of Chemical Education

tative effect increases with the column diameter, d f , and decreases with the coil diameter, dfoil.Giddings has derived the equation (241

where in this case u and d,,il are measured at the column center. Littlewood (22) has confirmed the dependence on d:/dloil using a different derivation. This useful equation may he quoted with some confidence although it has not been verified experimentally. It is useful in column design so long as the following warning is added. When the coil diameter is very small the curvature directs the flow stream across the column. This phenomenon is called "secondary flow," and it reduces C, in the same way as molecular diffusion across the column, and thus, improves column efficiency. Moulijn et d. (25) quote the condition for this phenomenon to occur in an empty tube as when q is the kinematic viscosity of the mobile phase. It may he assumed that this phenomenon never occurs in regular chromatoeranhic but verv " . aractice. . - tiehtlv - " coiled onen tubular columns have been shown to give improved efficiency (25), and it is uossible that it mav become chromatoeranhic aractice - . . in the fiture.
"Infinite Diameter Columns"

In gas chromatography the sample is often evaporated in the injection system and a finite rate of evaporation spreads the zone into an exponential distribution. The extent of this distribution in space increases with gas velocity so it appears in the C term. However, it is not reduced hy diffusion, so it is . , As well as not part of the C, term but contributes to C broadening the peak, it also gives i t an exponential tail. If a valve system is used so that the sample is confined in an empty tube before injection, then the C, term resulting from that empty tuhe may be overlaid with a contrihution due to the finite volume of the sample. If h is defined as 02/Land if the distribution due to injection of a finite volume of length 1 is given by u2 = 12/12,then the contrihution to h is 12/12L. If the sample tuhe has a smaller cross-section than the haction of the sectional area of the column that is occupied hy mohile phase, then this term must be corrected hy a factor equal to the ratio of the two areas. This could appear as an additive term in the equation for h since it is unaffected by velocity or diffusion and could be considered as a justification for the historical A term as in eqn. (2). There are four reasons for not doing this:
(1) It would continue the present confusion over the A term, which would continue to be attributed to "eddy diffusion." (2) T h e equation is usually used to describe phenomena in t h e column, which this is not. (31 In a well-designed apparatus it is negligible and the equation would give a n unrealistic impression of the dependence of h on u. This would be the more serious because it complicates the algebra when the equation for h is used to predict optima or to evaluate conditions a t optima. (4) T h e interaction of this term with the C, term produced by nonequilibrium in the sample tube is not known, and it is conceivable that the overall phenomenon may be better rep, . resented as a contribution to C

The zone broadeninr caused bv the chance in wackine arranied by injecting the sample into the center of the column cross-section and having the column sufficiently wide that the sample will not he carried to the walls (by diffusion or by the flow pattern) by the time it reaches the end of the column. The outlet system is arranged so that mobile phase near the walls does not enter the detector. Since the sample does not reach the wall region it experiences the same flow profile as it would in a column of infinite diameter. Knox and Saleem 19) this kind of . . first investieated u column in 1969, named it the "infinite diameter column", and demonstrated reductions in plate height up to twofold at high velocities. This has been confirmed by DeStefano and Beachell(26.271. However, there is no data in regular - "iournals on such columns for velocities near those giving minimum h such as are now most used in liquid chromatography. Extrapolation of Knox and Saleem's data (9) suggests that they may he worse than regular columns at these lower velocities. However, such columns have been sold commercially so there is presumably evidence of their practical value.
End-effects

I t seems better to make a break with a term that is mired in confusion and to present 12/12L in a separate discussion.
Corrections for Gas Compressibility

In gas chromatography the mobile phase expands as it moves through the column and encounters lower pressures. This results in a continuous increase in linear velocity u through the column. I t also results in a continuous increase in diffusivity D , As both are inversely proportional to pressure, D,lu remains contant along the column, at least for an ideal gas. The effect of these variations was elucidated by Giddings (31).An elementary exercise in calculus yields
h

=Eg+C,,fi+C,o
U

(19)

where

Zone svreadine also occurs in the iniection svstem and to a lesser extent in the connections from the column to the detector. "End-effects" occurring in these regions have been discussed by many authors of which the most thorough is urobahlv Sternbere (28) whose mathematical treatment is beyond t h e training of most chemists. Even more complex treatments have been offered hy Gill and his coworkers, e.g., in reference (29). The principal cause of the broadening is the flow profile across the various tubes through which the sample must flow to and from the column. This is usually more uneven than in the column itself and so causes significant contrihution to the C, term. The longer the column, the less important this becomes, so the contribution to C, is inversely proportional to the column length. A practical consequence of this is that connectine tuhes should be as narrow as wossihle or. if the design of &e instrument prevents this, theyshould be packed with fine elass - heads of about the same diameter as the nartitles of column packing.

A,, C,, are R and C, at the column outlet

= 9(p?lp:.- l)(p:lp: = jz(p?lp:

1)/8ipyp: - 112

+ 1112

This is aless detailed version of eqn. (6). Expressions have also been derived for non-ideal gases by Schettler et al. (32).
Summary Equations

To simplify the algebra used in optimization schemes a number of simplified equations have heen used. Equations (3) and (4) are the simplest and in the opinion of the writer are also the hest. For the higher velocities that characterized liquid chromatography a decade ago, the equation
h = Euz
Volume 60
Number 5

(20) May 1983

was often used where x varied from 0.3 to 0.7. This ignored the Blu term which probably was negligible. Now that liquid chromatography is most often carried out a t velocities close to the velocity giving minimum h , this form is no longer useful. Some theoreticians have extended this by adding a term to account for those mass transfer terms that are simply proportional to u to give the equation
h = Blu + Cu + Eoi13 (21) The origins of this equation are rather obscure. The earliest use seems to be by Done, Kennedy, and Knox in 1972 (33). They give the last term as Eun and suggest that n = 0.33 is the best value for velocities such that ud,lD, = 10 to 100. This is much higher than the minimum of the hlu plot that represents velocities that are now usual in modern liquid chromatography, so higher values than can now he expected. The value of n = 0.33 comes from an earlier paper by Knox and but reference to the paper shows that it refers to Parcher (9), a form of the coupling equation

N = L2/02= Llh

becomes smaller. Giddings ( 3 j ) has defined this condition approximately as

m>> 1
so that the equation is certainly inapplicable when N < 100 and douhtful when N < 1000. Moreover, the flow pattern in the regions near the two ends of the column is different from that in the main bodv of the column. This has been theoretically elucidated by Gill and Sankasubramanian 129) for emutv . . tubes, hut there is no guid:u~re, oinhi, I1 theauthtx iidwarv, t o suggest what lel~yth o i a vurkrd cAlmn hus ~~nrc:r)rcicmtarirc flow pro:ilea. On thc hasis of the work on empty tubes, bowever,-it would seem unwise to auulv the van Deemter equation to columns that are less than 50times as long as they are wide. These limitations do not affect the qualitative usefulness of the dispersion equation, hut unfortunately, a number of workers have attempted to use it to derive kinetic constants using columns that were too short to justify its use. I t is therefore important to make students aware of these limitations and prevent future errors of this kind. Literature Cited
11) Lapidus, L., and Amundson, N. R., J Pkya. Ckem., ,%, 984 (1%2). 12) Van Deemler, J. d., Zuiderweg, F. J.. and Klinkenberg, A., Chem En& Sci,5, 271

which requires some simplification to reduce to eqn. (20). In

engineerrug data involving both laminar and turbulent flow. This was first used in chromatoeraohv .. . . hv . Huber (35). .In rquation similar I,, t:qn ,221 was deriwd ihmreticull? fur unrttained c h i c - ; IIV Elmwith and [.in ~ . . r ! l.i thilt reduces to the form
h = Alu

+ Bl(1 + Cu-'I3) + Du2I3+ Eu

This was shown to provide a better fit to experimental data for unretained solutes in packings of glass beads than other published equations. While this was a real step toward understanding zone dispersion, it is improbable that i t will represent the real world of retained solutes. Moreover, it was based on Pfeffer and Happel's model (40) which showed that the mass transfer coefficient is proportional to u"3 a t values of the Peclet number ud,/D, (known in chromatographic nomenclature as the reduced plate height) greater than 30 and a porosity of 0.4. Values of ud,lD, are seldom that high in chromatographic practice, although they would occur a t the very high velocities at which simple versions of the van Deemter eauation break down. It would-be unwise to introduce most students to semiempirical equations such as these. There mav be occasion in group in optimization of thin-layer chromatography (36). Otherwise, eqns. (3) and (4) are to be preferred, perhaps supplemented with a qualitative description of the deviations to be expected a t higher velocities and the uncertain effect of these deviations at lower velocities. Unfortunately, they have begun to appear in textbooks without any explanation of their origin or their very limited usefulness. Limitations of the van Deemter Equation Whatever form of the van Deemter equation is used, it is strictly applicable only to infinitely long columns. The approximations involved in its derivation become progressively more important as the plate number N defined by

I821 Schett10r.P. D.,Eikelberger.M.,snd Giddings, J. C., Anal. Chern.,39,146 (19671. 1331 Done. J. N.. Kennedv. G. JLand Knur. J. H.. "GaqChromatoeraohv 1972." S C Pcrrv

136) I361 I371 138) (39) (40)

Huber,J. F. K., J. Chiornat"g. S c i . 7.85 (1968). Guiochun, G..Brersolle,F.,and Siouffi, A , J. Chiornotag. Sci., 17,368 (1979). Msrfin,A.d. P.,and Syn8e.R. L. M., Bioihem. 11.35,1358(1941). Einstein. A,. Ann. der Phyaik, 17.549 (1905). Horunth, C., andLin. H..J Chromatog., 126.401 (1976). Preffer,~. and , J . , A I. c h . E J., 10,605 (1964).

398

Journal of Chemical Education