/

:;,
JOURNAL OF RESEARCH of the National Bureau of Sta ndards-A. Physics and Chemistry
Vol. 67A , No.6, November- December 1963
I
>
I

Adsorption, Desorption, Resorption

W illiam V. Loebenstein
(July 31, 1963)

The eomplete characterization of batch adsorption from solution, desorption, and

related phe nomena have been interpreted in the light of a general equation. The forward
and reverse adsorption-rate constants and the adsorptive capacity comprise the on I)'
parameters. Where adsorption alone is of importance and the desorption-rate constant can
be negl ected, a simplified form of the theory results in a special equation which may suffice for
most ad sorption purposes . In either case, the characteristic parameters are determinable
from t he data and serv e as crit er ia for comparing similar sys tems. The theory has been
confirmed by the data of variou s investigators taken froll1 the l iterature. The parameters
derived from column adsorption are in agreement with th e corresponding batch-derived
parameters. The limita t ion s as lVell as the capabilities of the theory are prese nted; bu t
eve n where deviation s fro m th e ass um ed mod el ex ist, th e results are useful.

~\ 2 . Two-Parame ter Batch Equation

1. Introduction
In earlier publications [1, 2)1 the basis was estab- The two-parameter batch adsorpLion equation pre-
lished for characterizing adsorption from solution in viously dcrj"cd by integrating the irreversible rate
terms of just two parameters, namely; the adsorptive equation was shown to bc:
capacity peL' gram of adsorbcnt, qo, and the specific
adsorption-rate constant, k The values of corrr.-
j .
- ("~qO- l)eoklt
sponding parameters derived from batch and from q 1-e l eo
column adsorption were shown to be in substantial (1)
qo Wqo _(lVqO- l) Coklt
agreement with one another, respectively. The - - -e Veo
two-parameter equations are based on the assump- VCo
tions that the adsorption step is monomolecular, where:
I. irreversible, and rate controlling. Further more,
the assumption of a uniform surface is implicit in q= the amount of solute adsorbed per gram of the
the t·heoretical treatment, since the differential adsorb en t at any time, t;
equations used in the derivations are essentially of go = the maximum value q would have if all of the
I,A the Langmuir type. Although these requirements adsorption sites were filled;
may not be completely applicable in any given c= the instantaneous solute concentration whose
instance, t.he equations are still useful insofar as imtial value is co;
they provide an approximation of the characteristic W = the weight of the adsorbent;
I' parameters which may not be obtainable by other V = the volume of the solution; and
k = the specific adsorption-rate constant.
means. The present paper deals, to a considerable j

extent, with the treatment of data which fall in this

category. Methods are available ror obtaining values of the
In the more general case where reversibility m ust parameters go and kl which give an optimum fit of eq
be reckoned with, but otherwise subject to the same (1) to the experimental data ill t he general case where
limitations mentioned, a three-parameter batch W, V, CO, and t may all vary front point to point.
adsorption equation has been derived [2] which At best, however, they are cumbersome, and short-
includes the desorption rate constant k 2 • For the cut methods will certainly be preferred wherever they
first time a means is afforded for predicting dcsorp- can be used.
tion into solvent as well as adsorption from solution One such method was worked out [2] for the special
with equal facility. Perhaps even more interesting case where both liV/V and Co are held constant.
is the phenomenon of resumed sorption or "resorp- Under these conditions q val u,s, qj and q2, are deter-
tion" following the interruption of an initial mined corresponding to times tl and t 2, respectively,
adsorption or desorption step. Should an initial such that t2 = 2t1 • It was then shown that:
adsorption process, for example, be interr upted and
the resumption preceded by a deliberate and
sufficient lowering of the concentration, the theory
predi?ts a change to desorption in agreement with (2)
expenence.
I" 1 Figures in brackets indicate the literature references at the end of this paper.
: ~
615
and of t. The solu tion concentration is measured when t
takes on the values 2 : i, 2i, .1, and 2.1. By using the
same type of reasoning which led to eqs (2) and (3)
from eq (1 ), it is possible to show from eq (6) that
(3)
2q7J1;l - (2qi-q2i) (M 2_N 2) = q7Q2" (9)

Equation (10) can be written by inspection, J,

Use was made of eqs (2) and (3) in an example
taken from published data of Dryden and Kay [3]
for the adsorption of aqueous acetic acid on a steam- -<
activated coconut carbon. Good agreement was since it differs from eq (9) only in the subscripts.
obtained from three independent determinations of Equations (9) and (10) constitute a set of simul- ", I
go and k 1 . This agreement would not have resulted if taneous equations in two unknowns, M and
the neglected desorption rate constant had been (M2 - HZ) , for which the solution is easily obtained. I
Once these quantities have been found, (lvl +N ) and
~
appreciable.
Another special case whose derivation and solution (M- N) can readily be solved for use in eq (6) .
are very similar to the aforementioned method occurs Back substitution of one experimental point is
when t is constant providing that a second condition sufficient for the determination of k 1 • Equation
is satisfied. This is the requirement that two (6) can then be used to predict q for all values of t.
values of q can be found , say, (Jl[(Wj V)t, (co)!] and In the event that only the value of qo is desired
Q2[(W/ V) 2, (co) 2] such that (Wj V )2= 2(Wj V)1 and in a given instance, it is only necessary to determine
(co)2 = 2(co)!. Under these conditions the recurring (M2- N2) from eqs (9) and (10) for use in eq (S).
The quantity (lvI2 - N2) is given (according to
quantity TVVqo which appears so prominently in eq (1) Cramer's rule) by the ratio:
Co
, -I
remains unchanged and it follows that
2q1 - (2 Q,-q2i)1
1
2q~ - (2q j- q2j) .
(4) Consider the example (Dryden and Kay's Run
#201) already cited for the two-parameter equation
for purposes of illustration and comparison. The
and amoun ts of acetic acid adsorbed per gram of charcoal
are repeated in table 1 corresponding to the measured
solu tion concentrations. If, arbitrarily, i and .7 are
taken as 10 min and Vi min, respectively, it follows
(5) that
q,=O.3533 qj= 0.4033
q2i= 0.4667 q2j= O.5033. -'" I
3 . Three-Parameter Equation for Batch , i

Adsorption Substitution of these values in eqs (9) and (10)

results in M2_N2 = 0.63965 so from eq (S), '1
It has also been shown [2] that where it is desired
to retain the desorption rate constant, k2' in batch qo = 0.627 meq.g - l. \>" 1
adsorption, the integrated equation takes the form:
To continue with the illustration,
(M - N)-q M- N - 2 (y) Nklt (6) M = 0.S4S. Thus, M-N= 0.566 and M + N = 1.130.
(M+ N)-q M+N e By substitu ting qo and M back into eq (7), it can be
...,
verified that k 2/k 1= O.OOI47. This conflI'ms that I
where M and N are defined as: the desorption rate constant is, indeed, very small , I
compared with k1 . One of the adsorption points,
(7) say, q= 0.3533 for t = 10 may now be substituted
back in eq (6) to solve for k 1 • The two rate constants
are
k2= O.00524 min- 1
2 Sioce the three independent parameters m nst be evaluated Irom the experi-
Here again the parameters qo, kl' and k2 can be readily m ental pOints, no less than three pOints arc required. The use 01 fonr points
estimated from a single batch adsorption experiment affords a degree 01 flexibility in the choice of data. The only restriction is that
ir'j although it is permissible for 2i=j. In the latter instance, the minimum of
in the special case where q is determined as a function three points would, of courso , result.

616
 and eq (6) , in t his particular ins tance, r educes to : ca tion of eqs (4) and (5), t he ,'alue f 2.024 mM·g- I
was selected for q2 corresponding to (W jV )2= 0 .04
0.566 - q 0. 501 e-O.0604t . g·cm- 3 (i. e., W 2= 2.0 g) and the cOl'l'esponding value
1.130- q of (coh was, con sequently, 0.3 779 mM ·cm- 3 • N o
m easurement was avn,ilfl,ble fo1' q l corresponding to
T A B LE 1. Data of batch adsorpti on R un #201 0.1 890 mM ·cm- 3 for (co) I and (W / VL = 0.02 g·cm- 3 •
(Dryden and Kay) Howe\rer , a conventional plot of the data by Burgers
Acelic acid:
V= I00 m]
Coconu t charcoal:
W =3.0g
showed very little sca t ter of th e points, h en ce an
00=0.0306 meq- ml-! Sieve si?c: (8 on 10) interpola tion was m ade between the close neighborin g
T emperature: 30 °C values r esulting in ql ~ 1.683 mM .g-I .
Stirring rate: 400 rp m

2. Batch adsorption data f or acetic acid on "Norit"

r q
T A BLE
(W . G . Burgers)
I E aeh de term in at ion was carried out at 25°C w ith V =50 em ' a n d / = 2.0 h r bu t
r min m eq·ml-I meq'rr' wlth initial concentra.tions an d ad sorben t weigh ts as iJl dicated. '
10.0 0. 0200 0.3533
15. 0 . 0185 . 4033
20. 0 .0166 . 4667 q (o bl)
30. 0 .01 55 .5033 Co
45. 0 . 01 47 .5300
60.0 . 01 40 .5533 W = 1.0 g W =2.0g W = 3.0 g

1nJ.' 1·cm- 3 mA[ 'rr' m M' g-1 mA[' rr'

It should b e po inted ou t thfl,t the value of 0.627 0.0105 0.354 0.224
. 0323 .725 .551
-obtained h er e fol' qo is about 10 p ercent higher than .0489 .908 .737
the corresponding valu e previously cited [2] rOT' t he .0546 0.655
simpler case where irreversibility was assumed. .0643 1. 030 .876
.OS8! 1.199
This (presen t) value is, moreover, consistent with
indep endent go determinations of 0.666 and 0.641 ror . 0917
. 1091
1. 240 I. 124
I. 056
{;ombinations of i= 15 with .1= 2i = 30 and lor i = 10 . 1796 1. 654 I. 514 I. 396
with .1= 30, r esp ectively . The v alue or 3.57 Jor kl .21 1.777 1. 646
{;ompares well with the corresponding value from th e .2588
.3146
1.869
I. 957
1. 750
1.896
1.671
two-p ammeter equa tion.
> .3779 2.082 2.024 I. 950

;, 4 . Application of Batch Adsorption Theory

for Different Values of C o E q ua tion (4) yielded qo~2 .1 and this was r etain ed
for use wi th the three-param eter equ ation . While a
A 1944 publicalion by W . G . Burgers [4] fl,fford ed somewhat lo w~r \'alue t han 6.0 was ob tained from
the oppor tunity to tes t the applicftbility of eq (1) eq (5) for let, this value was tested along with an
and/or eq (6) to t he case wher e the initial concen- estimate for k2 such t hat le2!le l~ 0.0 2.
tr ation differ ed from b atch to b atch. Al though lengthy op t imizing techniques ar e ava il-
Acetic acid was ad orbed on pulverized "N ori t" able for obtaining fl, " b est" fi t of the p arame ters in
charcoal at 25 °0 with con tinuous agi tation for eqs (6- 8) , no improve men t was sough t in this appli-
p eriods of 2 hr. The volume of solution was h eld cation. The purpose was t o sh ow t h at the agreement
constan t at 50 cm 3 while t he weigh t of t he carbon is r eason ably close between calculated a nd exp eri-
and the initial concen tration of acid were varied in mental values of q despiLe the use of these r ound ed
acC01'dfl,nce with table 2 which also lists th e co1'1'e- off first estimates of qo, le i, and k z. This comparison
spondinO' v alues for the amounts of ace ti c acid is shown in the last t wo columns of t able 3. The
adsorbed per gram. The experin1ental data of table first t wo columns identify t he p oin ts, whil e t he
2 ar e reproduced directly from the fu'st two columns intermediate columns list t he values co mputed for
of Burgers' Tableau I , Tableau II, et Tableau III. the component parts of eqs (6- 8) for each
In th e estima tion of qo and lei through the appli- determi nfl,tion.
T ABLE 3. Adsorpti on calculations fr om the data of table 2
E st imated values o[ 2.1, 6.0, andO.02 [or the parametersqo, kl , a nel k,/k" respectively, were used in th e fittin g o[ eq (6) .

Co lV/ V ,\ f2-N ' 0'1 M:-N M+N 2( ~) Nk,t q (c alc) q ( ob~)

- - - - -- - - - ---
> m .\[ ·cm- 3 (J·cm- 3 mA['g-' m]o, [ ·(/- I

0.2588 0.060 9.0573 3.3869 I. 8333 4.9405 2.237 1.60 I. 671

. 1796 .060 6.285 2.713 1.677 3.7'19 1. 492 I. 45 1. 396
.3779 .060 13.2258 4.3643 I. 951 6.777 3.475 1. 91 1. 950

.0643 . 040 3.3758 2.103 1. 079 3. 127 0.983 0.78 0.876

.3146 .010 16.517 5.232 1. 937 8.527 3. 163 1. 87 1. 896
.2188 .0·10 11. 487 4. 035 1. 845 6.225 2.102 1. 68 1. 646

.0·189 .020 5.1345 2.773 I. 175 4.37 1 0.767 0. 72 0.908

.2188 .020 22.974 7.020 I. ~9 1 12.149 2.462 I. 75 1. 777
.OSS! .020 9.25! 3.752 1. 551 5.950 I. 055 1.11 1.199

G99 3~2 - 63-- 61 7

5. Analysis and Comparison of Parameters T A BLE 4. Characteristic parameters determined fr om column
adSO"ption ex periments
from Column and Batch Adsorption Results arc based on t ile st atic-beel elata (C. E. Drydcn) for acetic acid on coconut
charcoal at 30°C fitted to cq (11).
An example has ftlready b een gi \Ten in an earlier
paper [2] of the application of the present adsorption Run Sieve Co 17 x (10 kl
theory to a b at.ch adsorption run described by Dry- -No.- size ----
den and Kay [3]. That run was p ftrt of a wealth meg·ml- I rnl ·min- g meq·g- I
1 ml·meq-l.min- 1
·of experim en tftl data contain ed in the Ph.D. Thesis 18
19
(8-10)
(8- 10)
0. 10
.10
21. 8
35.0
59.8
59.2
0.850
.676
0. 769
.575
of C. E . Dryden [5]. An in terpr etation of the re- 20 (8-10) .10 6.99 18.8 .928 .677
sults of that data in thc light of the present theory 2t (8-10) . 10 3.32 18.9 1. 176 .379
36 (8-10) . 10 4.88 59.4 1. 333 .398
can now be readily made. Dryden's experim ents (12- 16) .10 9.18 18. I 1. 259 1. 124
53
·consisted of some 20-odd column adsorptions (s tatic 55 (12- 16) . 10 3.49 18. 5 L 330 0.867
56 (12- 16) . to 22.9 18.4 1. 023 1. 845
bed experimen ts), a somewhat shor ter series of 57 (12-16) . 10 23.2 59.1 1. 363 1. 259
·column desorptions, nearly 20 bfttch adsorptions, 58 (12- 16) . 10 64.2 59.1 1. 057 2.259
and 7 batch desorptions. All runs were carried out 60 (12-16) .31 3.81 17.9 2.100 0.664
61 (12-16) . 31 45.4 60.6 1. 279 .690
at 30 °C usin g ftcetic acid together with a steam-
ftcti \"i1ted coconu t carbon. 75
76
(24-30)
(24-30)
.
.to
to 37.7
57.8
55.6
17.0
1. 542
1. 336
2.798
5.682
In the column adsorption experiments, a 4-fold i7 (24-30) . to 25.2 17.3 1. 625 3.059
variation in column height was used from run to 78 (24- 30) . to 9.53 16.9 1. 583 2.628
79 (24- 30) . to 64.6 55.5 1. 358 3.115
nm. Extreme values of volume-flow rate varied
83 (40- 60) . 10 13.9 15.4 1. 470 11. 569
over a 20-fold nmge. Four U .S. Standard sien 84 (40- 60) . 10 66.6 15.5 1. 406 22.33
sizes of chftrcoal werc used ranging from (8 on 10) 90 (40- 60) .10 25.3 15.6 1. 443 9.673

to (40 on 60) . Two levels of initial acetic acid 91 (40- 60)

92 (40- 60)
.31
.31
13. 1 15.6 2.2 10 3.642 , ~ '

63.6 15.5 1.935 5. 697

con centration ·wcre employed; namely, 0.10 Nand
0.3 1 N.
In the b atch adsorption experiments, the ra te of
1 -
agitation was varied from 0 to 400 rpm of magnetic
stirring. The sie\'e fractions range from (8 on 10) a. Agreement With Theory
'I
to (80 on 100) in five steps. W ater-we tted charcoal The most significant result which is immediately
was compared with the customary initially dry e,-iden t from table 4 is the degree of agreement
ma terial. The two levels of initial acetic acid con- among the computed yalues of qo. The spread in
-centration used were 0.03 Nand 0.10 N . sieve sizes correspond s to a range in mean particle
diameter frolTl about 360 to nearly 2200 microns.
5 .1 Parame te rs Derived from Column Adsorption The initial concentm tion ,' aries oyer three-fold.
Runs The Yclocity of flow ranges from 3.32 ml·min- 1 to
64.6 wl ·min- l , The weight of adsorbent yaries
For each of the column adsorption experi ments a between about 15.5 g and 60 g. Yet, notwithstand-
semilogarithmic plot of (co /c) - 1 again st throughput, ing the int,crplay of these factors, for the results of
y , was made in order to det(,l'mine the characteristic the 18 run s at which co= O.10, there yielded a mean
parameters, 1.0 and k 1. This is in accordance with 9.0 yalue of 1.264 mcq·g- 1 wiLh a standard deyiation
the equation of 0.263.
In (~-1) = k1?OX _ k1~OY (11)
h. Anomalous Effects . I
c V V A closer scrutiny of table 4 discloses se,'eral
in terrstin g fac ts. The excellent dat[1 mak e it
which had been deriyed and t('sted in earlier work possible to discern "second order" effects whi ch
[1] . In eq (11), V is the Yolume-,-clocity; x is the canno t be in terpreted in the light of the prese nt
)
mass of adsorbent upstream from the point at which simplified theory. ,
effluent is coll ected; and y is the throughput or A comparison of Run #55 with #6 0, of #83 with
cumul atiye volume of solution which has passed #91, and of #84 with #92 suggest tha t the effect of
that point sin ce the start. a three-fold increase in the initial conc ntration , Co,
The initial (low throughput) points of the cunes other things b eing equal, resulted in an increase
were not used in fittin g to the linear requirement of in 9.0 of about one-and-one-h alf-fold. This can )
eq (11). The substitution of solu tion for the water readily be explained as a departure from idealized
used to settle the columns is not, strictly speaking, Langmuir behavior. The L angmuir model implies
a piston-displacement-like process. Consequently, a un iform surface . If this is only approximated, <
Lhc early values of c should be abnormally low re- t hen the number of adsorbing sites eft measure of
sulti ng in initially hi gh ;-alues for (co /c) - 1. This 1.0) whic h could be capable of participating in the
phenomenon has been obseI"\'ed in other work [2], case of a much greater initial concentration of
as well. T he yalues of th e parameters 1.0 an d lei solu te would include some portions of the surface
consequently determ ined from the final points of req uiring higher activation energies. If, mor('oYel",
each plo t arc shown in table 4 along with the condi- adsorption proceeded more slowly at these latter
t ions applicable to each run. sites, it would result in a lower overalllc, value. o
i

618
 j-------------

'.:> 15 I ,t ncl (3) to determine go and lei :

0.060[0.258 (3. 268) - 1]

go= 0.060 (3. 268) - 0.490 + 0.258 0.262 m eq·g- I
iO
In [ 0.262- 0.245 (0.( 562) J
0.262 - 0.245
k1= (0.0306) (30)0 - 0.H562) 8.50ml·meq- l. m in - l .

Jf one had cbosen th e 15 and 80-min poinLs in sLead,

the computation for go would Jliwe he en :
0.0493[0.245(3.268) - 1]
I O L-______ ~ ______ ~ ________ ~ ______ ~ ______ ~
qo= 0.0493(3.268) - 0.444 + 0.245 0.259 m eq ·g -l
) o 500 1000 1500 2000 2500
MEAN PARTICLE DI AM ETER, fL ] n general, t h e greater time intenTals were con-
sistently chosen an d were considered mosL reli able.
FI GU RE 1. Dependence oj parameters on particle size.

rrhc parameters wete computed [rom col umn adsorption. Each poin t on both TABLE 5.- Data oj batch adsorption Hun #221
c urves rcprcsellts the average or throe to five indi vidual determinations. Ihllls (C. E. Dryden )
with esse ntially the same in it.ial concentrations wl!rc used, namely , 0.10 mCf]·ml- 1 Acelic acid: Cocon ut cha rcoal:
\1=100 JIll 11'= 10.0 g •
co=0.0301) JIlCq'JIll - l S ievc size: (8 on 10)
'rcm peraLure: 30 ':I e
Stirring rate : 400 rpm

Another "second order" eO'ect appears to be

f\) prese nt in consid erin g flow rates. Other Lhings - ------
. being equal, an in crease in \relo city res ults in it lower min meqoml- I meq·y-I
value for qo and ,1 higher n du e for kl as can b e seen 10.0 0.0100 0.20G
I from table 4. This behavior wou ld be expec ted 15.0 .0084 .222
20.0 .00i5 .231
1 based on diffusio n considerations wh ich have been 30.0 .OOGI .245
entirely neglected in the development of the present 45.0
no. 0
.0053
.0048
. 25~
.258
simplified theory . The greater the fl ow rate be- 75.0 .00 14 .262
.0040 .206
com es, t he more d ifficul t it is to insure equal access
of solute to all or t he absorbi ng sites. Thus, t he
computed nLlue of ([0 based upon experiments tLt the
hig her flo w rates would be underestimated. This rea- W hile it is possible to u Lilize eq (9) and (1 0)
oning is consisten t with o\' erestimated valu es for lei. for dete rmining t he parameter in accordance with
The ntri ation in s ie l'e size (par ticle dianwter ) has t he ge ner al adsorption equatio n as previously
almost a negligible efl'ect upon ([0 alt houg ll its i Ll- illus trated, this pracLice is only r equir ed when the
fluence on lei is quite pronounced. These co mpari- deso rp tion r ate co nsta nL, /c2, is appreciable relative
sons are portrayed . quite st rikin gly in . fig ure l. Lo k l . The simpler method shown h ero will b e
Phenomena such as t he ver y sillall dependence of qo preferred wh er ever i t ca n be used.
on par ticle size are of p ar ticular interest in co nfirming In like mann er , values for go and lei were deter-
t he physical signifi cance of the deri \Ted pammeters. mined for all or t he b atch experiments. These
Capacity for adsorption, like surface ar ea, is a results are group ed so as to bring out most eIrectively
quantity measurable only at the molecular level. the possible inHuence of each of t he factors studied
The process of subdiyiding a highly porou s particle such as rate of stirring, sieve size, etc.
\ creates very little additional surface not already a . Initially Dry Versus Prewetted AdsDrbent
accessible to a molecule. A few b atch runs were described in Dryden' s
Thesis [5] in which the adsorbent h ad been pre-
5.2 Parameters Derived From Batch Adsorption Runs soaked in water prior to con tact with the acetic
acid solution . It was hoped to ascertain whether
The quantities qo and kl were calculated from the presoaking h ad any effect upon t he adsorp tion.
> data or each of t he batch adsorption runs reported. It now appears clear, in light of t he present theory,
r In general, eq,s (2) f1 nd (3) were employed for this that t he prewetted adsorb ent gave rise Lo qo and lei
purpose, t he fa ct haying b ee n established t hat t he values which fell in lin e with those from th e initi ally
desorption rate co ns tant lez was negligibl y small com- dry adsorbent. These r~sulLs are s hown i n table 6.
pared with k1 • Th e volume of the solutions were 100 ml a nd the
As an example to illustra te Lhe procedure , the data ini tial conce ntr ation of acetic acid was 0.03 meq.ml- I.
and calculations Jor Batcll #22 1 are typical. Th ese Th er e was a sligh t dilu tion effect caused by the
data are given in table 5. The 30 and 60-min points water contained in th e presoaked samples as reflected
correspondi ng to 0.245 and O. 258 m eq·g-I for ql and b.v th e in crease in V a nd decrease in Co. However ,
Q2, respec[,i voly, -were selected Jor u e witb eqs (2) t his was limi ted to 10 percent in all cases a nd is seen

619
to have a minor effect at most compared with the T A BLE 8. Batch comparison : effect of stilTing rate
influence of changes in TV. 1 "~100ml lV~3 . 0g
The values of qo and of k1' of course, should b e Co =O.03 mCQ·ml- i rremperature: 30 °C
-constant if the ideal conditions assumed in the
derivation of the theory were closely approximated. R un rrin g I
No. 1Stirate QO

The observed trend , attributable to the increase in

the lIr/v ratio , is undoubtedly caused by a depar ture &-a
from these conditions. Sieve s ize: (12 o n 16)

T A 131"E 6. Batch comparisons: the consequences of pre welling

and the ejJect of varying the amount of adsorbent
208
207
rpm
o
120
rneq·g-l
0. 522
. 541
ml·meq -! ·mill - 1
2.97
6.16
1
Sievc size: (8 on 10) Sti rring rate: 400 rpm 206 400 . 562 4.77
rr cm perature: 30 °0
8- b
Sieve size: (8 on 10)
1
Run Ini tial IV 1" Co go kl
No. state
----- ------ 203 o 0.590 1. 26
9 ml meg .ml- 1 m eq·g-l ml·meq-l·1ni n- 1 209 150 . 498 14.6
222 d ry 1.0 100 0. 03 0.726 2.19 201 400 .565 3.74
236 w et 2.48 108 .027 .504 ~l. 06
201 dry 3.0 100 .03 .565 3.74
238 wet 4.01 107 .026 . 474 4.45
220
221
dr y
dry
5.0
10.0
100
100
. 03
.03
.421
. 262
5.26
8.50 d. Effect of Initial Concentration , '

The influence of Co on th e r esults of b atch adsorp -

b. Effect of W /V tion are s trikingly similar to those for column.
Although only two b atch runs were made at co = 0.10 ,
The effect of W /V is equivalent to the effect of W these are sufficient for c.omparison purposes . 'fables
in this work since V was held constant at 100 mI. 9- a and 9- b compare th ese b atches (Runs #202 and
(Runs #236 and #238, itlone, hitd sli ghtly higher #215) with other lmtch run s which d iffered only j,
values b ecause of presoaking.) T ables 6 a nd 7 show with respect to ini tial concen tration. The qo values
the r esults or increasing adsorb en t weight. The of 1.425 and 1. 201 meq .g- l ob tained in Runs #202 t
I
essential dis tinction between the two tables is the and #215, respectively, compare well wi th 1.264
ieve sizes although these differences are not pro- m eq.g- l, tbe itverage of the 18 column Tuns pre-
nounced because the sizes ar e close toge ther. A viously compu ted for the sitm e initial concentration.
more sear ching comp/trison or the effect of sieve sizes VVhile kl seems to b e more sensitive to variations in
is taken up later. Th e important poin t here is that conditions than does qo, its magnitude is also con-
the value d etermined for the parameter qo decr eases sistent "vi th the corresponding colulnn resul ts. -"
with increasing IV. Both tn,bl es confirm that a It is interes ting tbitt a three-fold increase in in-
10-fold change in W results in abou t a 3-fold change itiitl concentration from 0.03 to 0.10 m eq .ml- 1 (as
in qo. The paritmeter kl is also affect ed by a change seen in tables 9- it and 9- b) resulted in neitrly a
in lIV. As qn decreas es, kl increases. It is about three-fold increase in qo. However, at higher initial
twice itS sensitive as qo, moreover, to changes in IV. concentrations the effect was much less pronounced.
This can be seen from t able 4 by comparing Run
T A BLE 7. Batch comparison: effect of W / V #55 with Run #60; Run #83 with Run #91; and Run
V= lOO m l St irring r at e: 400 rpm #84 with Run #92. In each of these comparisons
co=0.03 ru cq ·rul - 1 Sieve sizc: (12 on 16) where factors other than Co were essentially con- .. I
T emperatm c: 30 °C stant, the initial c.oncentration increased from 0.10
to 0.31 meq.ml - 1; yet t he increase in qo WllS limited
Run No. w gO kl
- - - - - - - - - ------ to about 50 percent.
g meq·q- l ml·meq-I, min- 1
217- 8- 9 l.0 0.672 2.64 TABLE 9. Batch co mpaTison: effect of c~
206 3.0 .562 4.7i
210 5.0 .462 5. 19 j "=100ml Stirring ratc: 400 rp m
216 10.0 .2 71 15.9 1'cmpcratuJ'c: 30 ° C

c. Effect of Stirring Rates

Run No. 1 co gO

9-a
The rate of s tirring was varied in three steps from W= 3.0g Sieve size: (8 on 10)
o to 400 rpm within each of two sets of experiments.
The sets differed from one another in regard to sieve meq·ml- 1 [ 11leQ,g-1 ml·meg- I .mi'n- l

size. The res ults are shown in tables 8- a and 8- b. 2,01

202 I 0. 03
.10
0. 57
1. 425 I 3.74
0.819
,Vithin each set there is no apparen t correlation of
parameters wIth stirring rate. The observed spread 9- b
S ieve s ize: (12 on 16)
1V= 5.0 g
in values or go are entirely within experimental error.
Tb e sitme is true for k1' except perhaps for the unusu- (
itlly high value obtained in Run #209. No reason 210
215 I 0.03
. 10
0.462
1. 201
I 5. 19
0.905 I
.;;:
~an b e found for this singular anomaly.

620
 e. Effect of Sieve Size fe2 . Furthermore, it would b e extreillely desirable
~'he influence of partide ize on Lhe parameters to be able to use on ly one equation for both adsorption
del'llTed frOIll th e batch ~Ldso l'pLio n experim ents a nd desorption.
conftrms the findings of the colum n runs. Very The difl"erence beLween Lbe two p rocesses should
b e r eflected only in Lhe bo und ary cond itions. Tn
, . little, if any, change in qo is eloid ent from table 10- b ,
alt hough t he siel' e size ranges in fiv e steps from (8
Oll 10 ) Lhrough (8 0 on 100 ), other fa ctors being con-
the clerivtttio n of Lhe ~ld s orpLion eq uati on Lhe initi al
conditions r eq ui red all or LI1C acl sorbab le species to
he in the solution phase. Co nvrrscly , ror desol'p t ivu
stan t. At t.he s~me time, howe\' er, the accompany-
mg \' alue of kl mcreased markedly with decreasino' the adsorb able species ini Li~t n:v lI'o uid bc entirely
particle size. T~bles 10- a , 10- c, and lO- d sho \~ in the adsorbed phase . To procced 011(> s tcp furlhrl" ,
t he sa me lack of dependency of qo although only it might be stipulated that bot h ~td so rpt ion and
two runs were available for comparison in each case. desorption should be considered , rro1l1 Lhis poinL of
In a preceding paper [2], the adsorbent im' oll' ed view, as special cases of an ini t ifLl co ndi Lion wllCre
was a senice bone char which had been subject ed some of the adsorbable species m ay pxis t in olu lion
to J~umerous cy:cl~s of adsorption, partial desorption , while the r emainder is adsorbed. Th e process which
drYll1g , and k~l l1l~g. Its prior history may have would subseq u ently tfLke place migh t either be acl-
been reflected In Its dependence of qo on sie l'e size sorption or clesorp tion , depending upon the levels
in. co ntrast . with the prese nt st udy. This very of t he interrelated variables. These various concepts
pomt was dIscussed in some detail at that time. m ay b e r eco nciled by use of the term "resorpli on "
to def! ne Lhis resumed sorption process.
TAB IJE 10.- Batch co m pa?'ison: e.D·ect oj particle si ze In the original derivation of eq (6), the qUi\,ntity
Co = 0.03 rneq · rn j- l Sti rri ng rate: 400 rpll l
Co was defined as the co ncen tntL ion or lhe ads orb ~lble
Te mperatu re: 30°(' solu te b erore ~l n y ad sorp tion Imd tftkell phtce. !for
t he gcneral casc (applic!1ble as well in t.he original
-,.' -Hun ~0. 1 Sicve size gO
CRSC), Co should be r edefi ned as foll ows:
(
lOr a lI'=l.Og
r' Co = the COllccn Lration th at would E'xisl aL any Lim o
i r all of t h e ttd sorbable species were ass L1 1n ed
me,,· a- I ml·meq- 1. min- 1
to be ill the sol u Lion phasc.
~
222 (8 011 10) 0.72G 2. 19
217- 8- 9 ( 12 on 16) . 672 2.64

lO- b 11 '= 3.0 g

Two new symbols can now be defined as Cl and ill to'
co rrespond to the concentration a nd Lh e ~LJll O llnt
20 1 (8011 10) 0.565 3. 74
ad sorbed per g J'nm , r es pecLively , whi ch e,is t at the
206
212
( 12 on 16)
(24 on 30)
..\62 'J. 7i onsct of it sorp tion process .
.608 15.22
213 (40 on 60) .567 71.5 Si nce the cO JlservtlLion eq uaLi on h olds under n,ll
214 (80 0 11 100) .653 110.5 condiL ioJls, it foll ows here Lhat
lO- e W =5.0 g

l:, (12)
220
210
I ((8011 10)
12 on 16)
0.42 1
. 462
I 5,26
5. 19

r lO- cl 1F=1O.0 g and th e gen e ml rorm of Lhe inLegrated equation

becomes
221
216
I ((812011on10)16) 0. 262
. 271
I 8. 50
15. 9
(M - N )- q (J 3)
(M + N) - q
6. Adsorption- Desorption- Resorption
i ::\1uch has been written in the preceding sections while Ai and il,{2 - N2 r etain t h eir defin it ions as given
of t his paper regarding the limitations of the present byeqs (7) and (8), respectively.
theory. Examples ha I'e b een gi loen and comparisons It is seen that the only difference boLween eq (13)
made showing the exte nt of departure from ideal and eq (6) is the appearance of qr in l111nlE'rator and
conformi ty with. t he model assumed, although denominator of the c.oeffi cient 01' e. Refer ence to
plausible explanation s were offered for most of the eq (J 2) confirm s that for an adsorption process IJI = O'
ob en' ed discrepancies. D espite these shortco min gs, and Co= Cr . U nder these condi tions eq (13) reduces
the theory hfLS much to recommend it including to eq (6) as a spec ial case. For a desorptio n process
applicatio ns whi c h hal"e not heretofore been dis- where the adsorbent containing adsorbtlte is added
cussed . One s uch application is in desorp tion. I t to pure solvent, CI vanishes and eq (12) shows that
is clear , of course, that t he simplified t wo-p~Ll"ameter VCO / W = qI' Obviously, in al1~T C~lse, it is always
eq uation cannot be used in this applicfLtion , sin ce tru e that IJo '2.r]L. Sin ce qI is differen t from zero in
it neglects entirely the desorption rate co nstant, this instance (desorption ), eq (13) would apply.

621
 6.1. Desorption experiment . Next, it is ascertained by use of eq (8)
tha t N = 1.338 . The coefficient of t in the exponent
The consequ ence of sub tracting qr from t he of eq (13) can now be determined as well as th e
numerator and denominator of the coefficient of e factor :
in eq (13 ) can impart a nega tiye value to this factor
which immedia tely iden tifies the process fLS one of (M - N) - qr
desorp tion. It is instructi \'e to consider th e batch
desorp tion da t fL of table 11 as an example of th e (M + N) - qI
use of eq (13 ) in this cfLp acit y . The table con tains
t he data collected by Dryden in Run #224 . Accordingly , t he desorpt ion equation reduces to:
Amcoal
oun___________
t of acctic ac_______________
id p reaclsorbed on t hc__char-
____ __ __ 7.26 meq 9·800 -q= _ 0.283 - 0. 136t.
'W eigh t of wc t charcoaL ___________________ _ 8. 036 g 3.476- q e ..f)
vVeigh t of d ry charcoal (W ) ________________ _ 5. 22 1 g
D iffe re nce (assum ed to be excess wa ter) ___ ___ 2. 82 ml It can be seen tha t as t b ecomes large, the righ t side
Wa te r a dcled _____________ ____ ____________ _ 100. 00 ml
Tota l wa ter prese nt , (l ')-------------------- 102. 82 rnl of the equfL tion approaches zero . Therefore, th e
limiting value of q must be 0.800 in agreem ent
_ 7.26 _ 39 ._]. _ 7.26 with table 11. At the other extrmne the value of
Qr - 5 .221 - 1. 1 meq ·g , cO-- I02 .82 q predicted for 1 min is 1.33 compared with 1. 190
as seen from the table. The remaining desorp tion
= 0.07061 m eq·ml - 1 . experiments r eported b y Dryden [5] yield resul ts in
substan tial agreem ent wit h the example gil-en here.
While i t is possible, analy tically , to soh-e the de-
sorp tion equa tion usin g fL m ethod based on the sam e 6.2. Resorption
principles as in the case of adsorp tion, it is con-
siderably more involved. It is extrem ely sensitive The rem ark able yersa tility of eq (13 ) canno t b e
both to t he accuracy of each of the t hree or four fully apprecia ted un t il som e examples of r esorption
m easured points used, as well as to the sligh test are considered . Fortuna tely, it is not necessary t o
departure from the assumed model. For these redesign additional exp erimen ts to illustrate these
r easons t h e usefuln ess of this m ethod for determin- applications.
ing t he param eters is purely academic. For t he first exa mple, consider desorp tion Ru n
#2 24 just discussed. The desorp tion equation pre-
T ABLE 11. D ala oj balch des ol' pli on R un #224 dicts q= 0.81 m eq· g- l for t = 30 min . Suppose t hat
(C. E . D r yden ) after desorptioll h ad p rogressed for 10 min , th e
Acet ic acid: Coconut charcoal: process were halted by physically sep ara tin g t h e
St irring rate: 400 rpnl Sie,'e s ize: (8 on 10) adsorben t from solu tion for an indefinite period of
T emperature: 30 °C t ime. Ultima tely , adsorben t and solution could b e
r ecombined , t hus perm it tin g t he desorp t ion process
t o b e r esumed. R efer ence to t able 11 discloses
that when t= lO min , 0 .9 12 m eq· g- I is t h e obsenred
min
1.0
meq'ml- 1
0. 0102
meq-rr l
1. 190 value of q which , consequen tly , would become the l
2. 0
3. 0
. 014
.0169
1. 115
1. 058
n ew valu e for qI in the r esorp tion pr ocess. N either
4.0 .0188 1. 020 M nor N would change, sin ce the weigh t, volume,
5. 0 .0201 0. 989
i .5 .0221 . 955 concentra tion , etc., were no t altered . The new
10. 0
15. 0
.0243
. 0266
.912
.86i
coefficien t of t he exponentilLl in eq (13 ) would be:
20. 0 . 02i 5 .849
30.0 .0282 .835
45.0 . 0293 .814 0.800 - 0. 912
60.0 . 0299 .802 - 0. 0437
3.476 - 0.912

while the only change in the exponen t, itself, would

For the example used in this illustration , a reason- be the substitution of (t - 10 ) for t. Almost by
ably fair agreem ent with the desorption data can inspection, therefore, the new resorption equation
be obtained using the approxima te yalues: could be \vrit ten :
qo= 2.0 m eq· g - l 0. 800 - q - 0.0437 e- 0. 136 (t - 1O) .
k 1= 1.0 ml·m eq- l.m in - 1 3.476 - q
k 2 = 0.045 min -

It is noteworthy t h at the magnitude of each of

I. The 30 min point is again calculated to be 0.81
m eq· g- l, in agreemen t with the original desorption
equa tion.
.'
these param eters is consistent with corresponding The same trea tment can be applied to inter-
values derived from adsorption. Equation (13) rupted adsorption . Consider the illustration given
can now be evalua ted. It is first determined by earli er in connection with table 1. If the adsorption
cq (7 ) that 1\1= 2.1 38 under the conditions of the h ad been interrupted after having been allowed to
622
 I

proceed for , say, 20 min, and the amount adsorbed condition corresponded to a q valuc of 0.802 meq·g-l
pel' gram at that t ime were considered the new as indicated in the last column of table 12. In his
initial conditions; what would the resultant re- second batch before d ilution, Co was 0 .1376 mcq.ml- 1
sorption equation become ? Again, M and N would while after dilution , it l'eycrLed back to 0.06880
be unchanged, but now qr (instead of being zero as meq·ml- 1 . The final co nccnLn1Lion or t he sccond
at the b eginning of t h e original experiment ) would batch after dilution cOlTcsponded Lo 1'f = 0.8 16 m eq·g- l.
take on the new value of 0.4667. Immediately, the If the prese nt theory f1ppl ic Lo Freu ncllich's
I resorption equation in that instance could be written: experiment, it ou ght to b e possibl e to ass ig n r eason-
able values to the three parameters, qo, lei , and le2 ,
I 0.S66 - q + 0.lS0 e-0.0604(/-20) consistent with results already di sc usscd for similar
~ 1.130-q systems under s ubstantially the sa me ('0 nci.iLi.ons. If
it is estimated that
i where the coefficient + 0.lS0 is determined from

0.S66 - qr
1.130-qr

according to eq (13). The fact that the coefficient

0.lS0 remains greater than zero shows that th e the sorption equa tions applicable to both batch es <11'e
resorption in this case is a n adsorption process. determined as follows:
Jf arter 20 min in this sam e illu stration , thc
solution had b een diluted by adding water until its T A J3LE 12. F1'eundlich' 8 exp eriment
volume, \1, b eca me, say, 500 mJ the situation would
have changed considerably. 'While qr would still 11' V
- - - - " - - -- ------ -
?; be 0.4667, the initial concentration Cr would now ml hr meq.ml- 1 meq.y- l
Icirst batch
I b ecome (0.01 66) / S 01' 0.00333 mecpnl- 1 . The n cw Ini tial statc _________________ 1.0 100.0 0 0.06880 0
\) Final st ate ________________ __
valu e of Co according to eq (12 ) would b e 0.0061 3. 1.0 100.0 20.5 .06078 0.802
It would now b e possi ble to recompute ill! from eq Second batch
Hefore dil ution :
(7) and then to r edetermine N from eq () . The Ini tial s(alc _________________ 1.0 50.0 0 .1376 0
Final state __________________ 1.0 50.0 21. 0
new values would be M = 0.947 ; N = 0.506. T JIO Aftcr dilu ti on:
resorption equation would then reduce t o I n itial state _________________ 1.0 100.0 21.0
Final state _____________ _____ 1.0 100.0 22.0 .0064 0.816

0.441 - q -0 .0264e -o. 0217(1 - 20)

1.453 -q For the first batch:
and since the coefficient is now negati\'e, tbe res umrd
process would have ch anged from adsorption Lo 1\1[=2100 [ 0.02 + 0.06880 (1.000+
1.0 0O.068S0
100X X 1.07 )J = 4.975
desorption. Clearly, if the degree of dilution h ad
been but slight, the resorption would haye continu ed
( as an adsorption process but to a diminished extent. an d
It is instrnctive to selec t the final illustration
(
r, from an experiment cited b y Burgers [4] in refel'l'in g },J.2- N2= (l.07) (6.880) = 7.362 so, N = 4.170
to a paper by Freundlich [6] published nearly 60
years ago. Freundlich compared two batch adsorp- hence,
tion runs using 19 of blood charcoal as the adsorbent
M_ ~lI,T )
in each run and acetic acid as the adsorbate. The
t~ J.l;1+ N =+ 0.0880.
I second run used twice the initial concentration, but
only half of the volume. However, after a reasonably
long period of time , the second batch was diluted with Finally, the adsorption equalion takes thc form :
an equal volume of water and stirring was continu ed
for an additional hour- presumably lon g enou gh to 0. 80S - q O. 0880e -2. 502t
re-establish equilibrum . Both runs ended under 9.145 - q
comparable conditions, yet the final solution con-
centration was slightly lower in the second experiment which may be solyed for I'f when t= 20.S hI' to give
than in the first. Freundlich ignored the diffcrence q= 0.80S meq.g- l compared with 0.802 in table 12,
and used the illustration to prove the r eversible Fc1' the second batch- before dilution :
I nature of adsorptio n . It should be possible in
I ~

I light of the present theory to Te-examine t he data

F quantitatively in an attempt to account for the .L\l1= 50
2 [ 0.02 + 0.1376 (l.000+
l . OSO
O X l.07)J = 4.475
X O.1376
observed discrepa,ncy.
-;:" Freundlich's measurements are shown in table 12 .
For his first batch, the initial value of c was also Co, M 2-N2= (1.07)(6.880)= 7.362, as before; but now
l;> sin ce all of t he acetic acid was in solu tion. The final N =3. SS9
I,
.r 623
 \ ,

· . M -N kl ' an~ k2 whose valu es best fit the general adsorption

so m t hIS lllstance M + N=0.1l40. Therefore, the equatIOn, eq (6). /

In the special case where the desorption rate con-

adsorption equation applicable to this case b ecomes: stant .can ?e neglect,ed, a sirTlplified two-parameter
equatIOn IS adequate for adsorption. Short-cut
0.916 - q methods have been found for evalua ting the param-
8.034-q eters from the data:
(a) when the adsorption is a function of time
This would r equire that at the time of dilution; or 1
namely, when t = 21.0 hr, the value of q would have (b) when the adsorption is a function of both 1

been 0.916 meq.g- l (although it was not actually W jV and co.

measured). Values of the parameters can also be determined 1

for the general case where the adsorption data are j

For the second batch- after clilut1:on: The addition time dependent.
of 50 ml of water would have the effect on lv! and N The characteristic p arameters determined from
of causing them to revert back to the values 4.975 batch adsorpti.on are in agreement with e?ITesponding J
and 4.170, resp ectively, which applied to the first values deternuned from column adsorptIOn.
batch. This is e\'ident from the definitions of J..([ Some deviations in the results have been observed
and 0.£ M2 _N2 in eqs (7) and (S) . The only diffrr- in c.ertain instances and can be explained in terms of
en.c~ IS th~t qr would now bo 0.916 meq.g- l where
a shght departure from the theoretical model.
ongmnlly It was zero . Therefore the coefficient of Even .the general adsorption equation, eq (6) , can
the exponential becomes: ' b ~ cO.l1s1dered as a special case of eq (13) which
diffenng on~y in initial conditions but u tilizing th~
0.S05 - 0.916 same set . ~f paramet~rs , ' yill, in fact, predict with
-0.0135 equal f~cJhty desor.rtIOn , mterrup t()cl sorption, and
9.145 - 0.916 sequentIal combmatIOns of adsorption and desorption
as the case may be.
which, being n egative, m.eans desorption. The
final re~orption equation can therefore b e written by
lllspectIOn: 8. References
0.S05 - q [1] vV. V. Loebenstein, Proc. of Fifth T ech. Sess. on Bone
- o. 01 35e -2. 5021. Char, 253 (1957).
9.145 - q [2] W . V . Loebenstein , J. l1es . N B S 66A (Phys. and Chem .)
No.6, 503 (1962) .
The final condition after dilution and resorption was [3] C. E. Dr yden and W . B . Kay, Ind . Eng. Chem. 46 2294 -·1
reached 1 hI' later. By substitution of t = 1.0 in (1954). '
[4] W. G. Bmgers, R ec. Tra,·. Chim. 63, 46 (1944).
thi~ egll~tion , it is found that q= 0.SI4 m eq.g-l [5] C. E. Dryd en,. Ph .D. t hesis: Th e kin etics of sorption of
whlCh IS ll1 good agreement with 0.S16 in table 12. acetIc a CId III water-carbon syste ms (The Ohio State
Ul1l VerSlty, 1951) .
[6] H. Freundli ch, Z. Ph ysik. Chem. 57, 385 (HJ07) .
7. Summary
B atch adsorption from solution can be character- (P aper 67 A6- 24S)
ized and interpreted in terms of tbe parameters qo,

" I

I
.\0

624