THE PI-ITSIC-IL CHEMISTRY OF lIEI1IBR-IXES KITH PARTICULAR REFERESCE TO T H E ELECTRIC-IL BEI-I+41T710ROF lIElIBRA4SESOF POROUS CHAR-4C‘I‘ER. I1 THE SATURE OF THE “DRIED” COLLODIOX ~IEIIBRASE. SOMECL-IIEENT ~IEJIBRASE THEORIES ASD THEIR LIIIITATIOS~ ILIRL SOLLSEII Departnzeiit of Phiisiology, U n i v e r s i t y of J l i n n e s o t a , Jliitneapolis 14, Jlinnesota Receiaed September 6 , 1941 I. ITTHODUCTIOS
Current. theories concerning the nature of nieiiihraiie3 are primarilj- *.penneability theories” focusFed on the question of the penetration of n r i o u i solute: through membranes. l l u c h less attention has been paid to the solrent in contact ivith membranes, t o its state in the latter, and t o its fate during variou; nienibrane It, is impossible t o obtain a really clear pictuir of the nnture arid the function of a membrane vithout some insight, into the latter que5tions; nevertheless. it, i.5 necessary a t present to start investigations concerning the applicability of the various membrane theories to specific cases, lieginning with the coni-entionn! concept3 . In the subsequent sections of this paper variou:: physicochemical approaches are used t o elucidate the nature of collodion membranes in general, particular attention being paid t o the much-investigated though still highly controversial dried collodion membrane. These studies \\-ere originally undertaken in order to be able to ansn-er a number of specific questions concerning the functional behavior of collodion membranes. The present’ treat’nient attempts to put the several isolated facts together in a co1nprehensiT-emanner. I n the past, membrane phenomena hare too frequently been treated in the abstract without considering their exact’ structural and kinetic bases. The es1 Presented in abstract before the Division of Colloid C’hemistry at the 10Sth 3Ieeting of the hnierican Chemical Society, S e w Torli City, September 14, 1944. 2 One can easily conceive of membranes of homogeneous-phase type which are permeable t o the solute hut impermeable t o the solvent. The osmotic phenomena t o be expected under these conditions have never been discussed in a sysieniaiic manner, though this r~-ouldo h viously be desirable. A satisfactory kinetic theory of the osmotic processes \vliich occur across membranes of porous character must co6rdinate the general concept of the activities of solvent and solute in solution and the influence of hydrostfliic pressure on these activities Tvitli the iollowing f a c t : K i t h “eeniipermeable” membranes the Ii>-tlro.staticpressure (under erl:~ilihriuni conditions) cornpensates the osmotic pressure of the solution, in t h e c branes of relatively high poro?iry in the pores of which the nornial hydrod:,-nnmic !sn-.s seem t o hold true, as n-cll as n-it11 nienibrmes the pores of Tvliich are of molecular size a n d thus outside the range of the nornial Iiydrodynaniic~l : t n x . Tlie author hopes t o h p :~h!c t r 4 present :i tiisc’iissiiin (if tliis n i u t t r r i i i thc iic:ir future joiiii!! .A it!i 1 ) r . 31. 13. T-issclii.!,. 171
172
K.iRL SOLIAXR
perimental results reported vill, it is hoped, demonatrate that it is pos3ible to investigate such phenomena as specific physicochemical problems and t o establish or eliminate the possibility of one or another mechanism operating to bring about observed effects. 11. ISDIVIDU-IL DIFFERESCES I N ‘(DRIED” COLLODIOT 3IE\IBR%XES
In ,z preceding paper (12) it u a s indicated that the majority of investigators seem t o be inclined to the vien. that dried collodion membrane. owe their characteristic properties to their microheterogeneous structure, Le., to their porous character, though other prominent investigators prefer the viwv that dried collodion membranes should be considered to be closely similar to, if not identical with, oil phases. Experiments n-hich could throw decisive light on this question 11 ere therefore highly deqirnble. There is one well-known observation which can thron considerable light on the mechanism of the electrolyte permeability of dried collodion membranes (16). Dried collodion membranes cast in the conventional m y from the same solution give widely varying concentration potentials. This effect is particularly pronounced with membranes prepared from electrochemically inactive preparations. Though they are of fairly uniform thickness, different membranes may show concentration potentials which vary by more than 100 per cent, the potentials across the individual niembranes being very constant and easily reproducible over long period.. I t can be shon n readily that this effect is not due to any cracks or other imperfections of some of the membranes. The homogeneous-phase theory cannot conceivably account for this variability in electromotive behavior. -111 specimens of a homogeneous phase must of necesity hare the same properties. The homogeneous-phase theory must assume equiralence or a t lemt statistical equivalence of all actual or virtual molecular interstices (the only ones n-hich exist in a homogeneous phase) and liken ise equivalence of a11 the characteristic groups which determine the electrochemical behavior of the memlxme. d structural theory, howerer, can easily explain the observed facts; differences in aggregation and orientation of molecules and micelles could easily account for the uncontrollable variation in the preparation of the individual membranes. -4ccording to such a theory one must think of the membranes as having a “micellar-structural” constitution as opposed to the molecular-structural constitution of a perfect crystal or 8 liquid, the emphasis being more on the nature of the interstices than of the structure-forming physical elements. 111. E S P E R I I I E S T S TO .U‘PR.IIBE
T H E REL.iTIVE MERITS O F THE HONOGENEOUS-
PH.-ISE THEORY ASD T H E STRlSCTUR.\L THEORY AS l P P L I E D TO DRIED COLLODIOS I\IE3IRR.%TES 7 7
1he ioiloiving line of reasoning has also served as a basis for experiments designed to be helpful in a decision betn-een t,he homogeneous-phase theory and the micellar-st.ructura1 theory as applied to the dried collodion membrane (16). -4 niimber of dried collodion niembranes are prepared under stnndnrdizecl conditions
PHYSICAL CHEUISTRT O F M E J I B R d S E S .
11
173
in the conventional manner on the inside of test tubes froi:i the same solution $11' tlectrochemically inactive collodion (state I). The mean characteristic concentration potential across a representative group of these menibranes is determined. S e x t , the inembranes are "activated" by osidation (state 111, as described pre~.-iou4y (12, 13), and the mean concentration potential of a representative nuniber 0: thew menihranes is determined. The remaining majority of the tnembranes in s t n t e I I is dissolved in the same solvent mixture 32s used originally. Driad collodion iiiemhranes are prepared from this solution (state 111); 2.5 before. the concentration potentials across these memhranes are measured and the mean -,-alne noted as in the former cases. The concentration potentiah ac:'o.si the membranes in states I, 11, and I11 are compared. The homogeneous-phase t,heory would predict that the properties of the nienibrnnes In state I1 and state I11 are substantially identical, since the properties of ;iri interplia,se irhich behaves like a homogeneous phase .should be independent of S o rearrangement of molecules or micelles, brought about' by an\ulcl be able t o alter the properties of membranes prepared from the Game material. The micellar-str~icturaltheory n-ould predict a distinct difference between state I1 and ctate 111. ,Iceording to this vien-, the dissociable groups which are introcluced 1 - j ~oxidation are mainly confined to certain more accessible structura? elements, i t . , the surfaces of those particles n-hich forin the pore n d h . When the membrme is dissolved, t,he elements whicli e cause of the characteristic properties of state I1 are mised with the .i\-hole of the menibrane material. If new membranes nre now prepared from this solution (state 111).the concentration of the active groups a t the accessible points is nccewwily rer1:iced :is cornpared with stnte 11. In order to demonstrate the nnticipntetl effect iiioit cleclrlJ-, r!ie Acti:-ation ~ h o u l i lbe ,st)oppedjust at the time when maximum or nenrlj- ninsi trntlon potentia!s are obtained. Too thorough osiil:ij.ioii mw.t ne i o diminution in the espected effect,. I n two 3eries of experiments seventy membrane; \yere c n 3 t \,state 1 reristic concentration potentials :icross ten iif theni. selected at randoil1. were memured, t!ie nicnn being 2T.5 and 30.4 inTT., respectively. T!ie t n - 1 2 halchei of .seventy membranes each were oxidized to a different degree i t a t e 111. Tn-enty of t!ie.se membranes were selected at random from e;icli b:itch: and t h e me:m chi:acterist'ic concentration potential across thein irn.: tieteminecl: x i t h the one Geries it was 49.9, n-ith the other series 51.3 m ~ - Tlie . fifty remaining membrane.+ of each hatch were dissolved in the same solvent mistwe lvhicti had lieen used for preparation of the original membranes, and dried collodion memhrnne were prep:uetl in exactly the same manner as originally j$tate 111). I n the one 5erie.s t,hirteeri such membranes were obtained, in the other. ,series twenty. Tlie rnea:~ characteristic concentration potential of the membranes in state I11 iws 26 m y . in the series which had undergone less oxidation, and 3 i . i mv. in the more highly oxidized series. I n the first series the characteristic concentration potentials in Ztntrl T x x l state 11 are identical n-it3hinthe limit< of rsperiment.al nntl .dtntisticni
error; in the case oi the more highly oxidized series the concentration po1en:isi in srate 111. though significantly higher than in state I , is much lon-er than ir, state 11. In a n nnalogoul; t'hirtl qerics the inellibraries Tvere oxidized very strongly. In agreement .i\-ithexpectation, the average concentration potential of the membranes in state I11 in this series n-as rather high, about 49 Inx-. Since the properties of n honiogeneous phase should not be influenced Lj- :i rearrangement of its constituent particles, these results strongly indicate that the hoinogeneaui-phase theory cannot, hold true in the case of the dried collodion rnembrane. The experinieiital results agree excellently n-ith the prediction5 of I,he n~icellar-structuraltheor>- and seem to provide clear evidence that the characteristic behavior of dried collodion membranes in solutions of strong inorganic e1ectrcllyte.s i~ due to the micellar character of its electrochemicallg importar:: inter4ces.
b
C
FIG.
1
2
Two s i i ~ p l enmiihrnne structures; highly sclienintizetf IT. THE THICIiSESS EFFECT
Prior investigators 11n\-ealn-ays niade the assumption, tacitly or esplicitljp, that thicker :xembrane.s sho~!-the same characteristic behax-ior as thinner membrane: of the same geneid character, all processes across them, of course, occurring more ~101vIy. Thus it has been generally assumed that the characteristic concentration potential acros,5 dried collodion membranes must be independent of i h e ~ h i e k n e soi t,he membrane (20). Since, as v:e shall see below, the Cc)Ilceiltrfi:ion ai t h e e!ectricallj- important tiissociable groups is very small, the ycestioii arwe w h e ~ h e rthere iq a siificient number of active groups available to supply encli oi ?he piisihle pa,thn-nysarross the membrane \\-ith at least one such :ictive g r o ~ p . I i the critical groltps are really scarce, the chance of this latte: occurrelice. gerLerail-qienliing. ;hcniid tliminidi i\-ith decreasing thickness of the memI:i.ane. ~!;oughthis need not iiece. rily he the case ( I 7). 2 present in highly schematic manner t v o possible ways of vi.+ -i\)utio:i of t h e nctivr: groiips and the t,hicliness effect. We 3.c-
PHYSICAL CHEJII8TRP OF 3IElIBR.ISE5.
I1
175
~ ; i r n pi n~ agreement
with the best' a,wilable information, that a cli.ssocialk pri.~!ip m r r o ~ pore r prevent3 the passage of an ion of the w x e .sign. hi?allun-2 t h e 'gt3 oi oppositely cliargetl ions; pores of proper dimension-; 1i:rving licj .sucl; c*nr.,rgocIgroups allow the passage of both anions :ind cLition-. In the figure; !lie membxne substance is indicated b~ striation; spots n+icli cnrry fixed charge tiiv;.< prevent the passage of ions of the oppxite sign :ire denoted hj- nsteriik-. i g w c 1 ,>lion-sthe simplest conceivable strwtare of a porou; :wnh:sne. Figure In. representing a m i t thicknew of membrane. s h o w ei-ery other pore i i l o ~ k dior the passage of ions, the ratio of the free to the geoiiietricdly po.+ib!si ~:)at!i.,~~:~~~-.+ thus being 0.5. I t is easy t o see that the niimher of pathn-ayi free to rhe permeation of ions of the m i i e sign as that of t.Iie fisetl disociabic group. decren5e.e in a. geometrical series as the thickness of the nienibrane increases in r i ~ l :irlthnzetical series. The theoretically posdile niasinimn value of the coiice:i:r:iticuj potential could he reached Ti-ith s w h nienibrme. if they are iufficiently Thick. H,x::e~.-er, this kind of very simple beliai-io:. i. 1))- 110 means necessarily inherent In inenibranei: irhich have a structure; mother povibi!ity is .3!ion-n in figure 2. Figure 2a represents :i unit layer of membrane, 50 per cer?t of the p r e s 5eiag i t l o c k ~ d: figure 2b shon-s a possible assembly oi tu-o such lnyers which dhii-ii iree cross c.onnection of tlic pores betn-een tn-o adjacenr membrane !aye:.'. if :in VIectrolyt e iliffiwe5 through such a menibrace, t,lii.s space i d .soon Le fillcti n.i~l: m electrolyte solution, the concentration of IT-hich i:: .mien-here ??etn-.ee:i :l:e ;.~)nce;itrn:ic~nsof the solutions on the tn-o sicles of the nienhrane. It i . ~ -?e, m t l ?\-e have also tlevi.1opt.d the idea in detail (17). that) the potenrial i n e : i ~ h m oi e this character is independent, of the n1unher of unit !n~-er.< from d1ic.i: ne is: built u p >i.v.>it is intlcpendeiit of t l i c le+ or' the mernhranc. +w thnt the cross connections betn-een :it port?' influence :he kh;t\-i- oi pojsiliilicellsr-structural theory nius ereforc 1): condpred to be com\.;it11 dependence or independence of the cci!ictxt2rcltion po:ent inl U ~ I W cknesa. The lioniogencous-phnse ti!Por,v :\-ixild unciei 311 circumt independence of the concentration potcntinl of the tliicknesq of The niembrruie. Since. as .slid :ibove, the concentration potcnti:tl :ic'ross clrierl co!h i i r m e i cii identii:al thickness varies greiitly, the question of t,he esi :hickr?e+ efiect could only be decided hy the u3e of .sta ically sigcifican: :iiiiyLh i . ? ci!' inexhrancs. We therefore have i i s d , as required by the circumstances. 'dp t o twciity membranes t o determine the average characteri.stic concentratioii pxential arising across membranes of different t,hicknes prepnrerl from col1otiiorr 1.nrioi1s clcct~rocheinicnlactivities.
jr-I ,z
?:ic-
7 7
176
IiARL SOLLSER
Bag-4aped nienibranes ere prepared in the conventional manner on the inside of test tubeq. I n order to obtain membranes of different thickness, collodion Solution* of various concentrations ( 2 , 4 , 6 , 8 ,and 10 per cent) in the same solvent mixture were used. To obtain very thick membranes we have in a few instances superimposed several layers of collodion on top of one another. Figure 3 gives the mean characteridic concentration potentials obtained xith the varioiimembrane.. The reiults of the experiments represented in figure 3 may be summarized afollows: The concentration potential is a function of the membrane thicknes uitb all the collodion preparations which were investigated. With membranes prepared from electrochemically active collodion, membranes of 15 to 20 p thickness give potential values which closely approach the thermodynamically possible maxjmiim. T i t h inactive collodion preparations the potential raluez obtained
A Boker Collodion
Oxidized Collodion
Membrane Thickness m Microns
FIG 3 The correlation betiyem the thickness of various dried collodion iiienibrnries and the characteristic concentmtion potential 0 1 chloride across them
AY potassium chloride-0 01 S potassium
with thin membranes %re lo\\ , they increase with increasing thickness of the membranes and reach constant values 11 ell below the thermodynamically possible maximum \\hen 20 to 30 I.( thick; still thicker membranes do not yield significantly higher concentration potentials; there is no indication that the thermodynamically possible maximum value could he obtained with membranes of any thickness. The assumption that the concentration potential is independent of membrane thickne.. doer not therefore hold true in the case of the dried collodion membrane?. Theqe results nre incompatible n-ith the predictions of the homogeneouqp h a v theory; they are compatible n ith the micellar-structural theory. Hon ever neither of the two simple pictures of membrane structure repreqented in figure? 1. and 2 fully agrcei: n i t h the results. Though I\ ith active collodion preparations the thermodynamically possible maxirnun~value can be approached -with increasing membrane thickness, the sh:ipc Jf fhe t h i c k i i ~ s ~ - p o t c n t i scurve l does not fit thc predictions of the theory. f
PHYSICAL CHEbIISTRY O F JIEblBR.1SES.
I1
1177
With inactive preparations there is first an increase in the potmiial n-ith increasing membrane thickness qualitatively (though not quantitatively), as can be predicted on the basis of figure 1. With still thicker membranes, hon-ever, a constant potential value is reached, as predicted in a qualitative manner by the assumptions n-hich underlie figure 2. The real situation is ob1-iously more complicated, somewhat in betn-een the two extremes represented by figures 1 and 2. I n a tentative way one may explain the observed facts on the basis of the assumption of highly active surface layers 11hich nit11 thicker membranes enclose zones of 101~.specific activity. This picture, howevr, has not yet been proven experimentally; it must therefore be considered to be of an entirelv hypothetical nature (17). Y, WATER UPTAKE A S D SK'CLLIXG OF COLLODIOS bIEMBK1SEB
It is impossible t o ha\ e a clear picture of the qtructure of a membrane if it is unknon n whether it is permanent and rigid or IThether it undergoes changes n hen it is immersed in different solutions. Water and aqueous solutions of inorganic and organic compounds are generally believed to exert no specific influence on collodion membranes, but this point has never been proven definitelj-. Concentrated solutions of most organic compounds are knon-n to sivell collodion membranes appreciably. 3Iichaclis (9, 20) ashumes that the dried collodion membrane behaves as a rigid porous structure 11hich shows no trace of swelling; water and aqueous solutions are supposed to enter the membrane by filling preformed pores. Manegold ( 6 ) , 11ho has carried out extensive physical investigations on collodion mcmbranes, states specifically that dried collodion is non-sn-elling in u ater. Experimental proof for this statement, however, could not be found: it seems to be questionnble in Yiew of the well-knonn hygroscopicity of collodiun, nhich v a s investigated by Sorthrop (11) for the specific case of the dried collodion mernbinne. The questions nhich need an ansn er under theqe ciicumstance.; are: ( 1 ) Honmuch TI ater does a completely dried collodion membrane take up TI lien placed in atel.? (2) Does this TI a t w uptake cause any detectahle change in the volume oi the inembrane? ( 3 )Do diied collodion membranes sn-ell or shrink Then transferred from n ater to solutions of strong electrolytes? ($) Do 6sporou"' collodion membranes sn-ell or shridi i~ herl traixferred from TI nter to solution.. of strong electrolytes? (1) For thiq 11ork the conventional thin bag-shaped collodion membrane- ivere ucelesb; fairly thick flat sheets had to be used in order t o be able to make the necessary measurements. The membranes m r e prepared by pouring coilodion qolution on glass plates which floated on mercury. To obtain dried membranes the solvent I\ as allon ed to evaporate completely, to obtain porous membrane.; of 50 to 60 volume per cent water content, the drying u as interrupted after 90 min. The dried membranes, 0.10 to 0.15 mm. thick, after being cut to staix1.d -iae are kept in a desiccator; the porous membranes are kept in water. The volume of the membranes n as determined n ith a pvcnome'er filled 11 it11 7
I
178
KARL SOLLSER
mercury, the only liquid u e know that ~ o u l dnot 11-et the membranes. The weight determinations were carried out in n.eighing bottles. First, the volume and the weight of a dehydrated dried membrane were measured; next, the membrane was placed in water or the solution under inyestigation. ilfter a measured length of time it was removed froin the liquid, blotted free of surface liquid, and weight and volume determinations n-ere carried out. This process Tms repeated at intervals till equilibriuni was obtained. I t became apparent immediately that the behavior of dried collodion meiiibranes is less simple than is generally supposed. The membranes in equilibrium with water showed a volume increase of 5 to 11 per cent, depending on the brand of collodion used.
f
2 zoE
a
0
Baker Collodion
0
Mallinckrodt Parlcdion
Hours
FIG.4. Weight and volume increases of dried collodion membranes prepared collodio:i preparations on contact nit11 imter.
In ortier to visualize the quantitative nieaning of the ohwrved v eight x i r t 3:olume increases, they nere reduced to the standard unit of 1 cc. uf dry men:brane. The weight and \-olumc increases are evpressed in milligram. and cuhir millimeter., respectively. for 1 cc. of dry collodion. I n figure 4 are plotted the 11 eight 2nd volume increases against the i ~ i ~ l > e r \ l o , : Time obtained with mein1x:iiie+ prepared under x-ery similar condition- i'roc, se-;era! brands of collodion. l i e n i b r a n e ~prepared from the same colloclioii bu' under different conditiom hlio~vsmall though significant differences in n eight xn(: volume increav 1-noaidized and oxidized (activated) collodion (not ~h01~711 in the figure) oi the same origin +hen- identical 11ater uptake and swelling when p i p pared under identical conditions. Ii a dried collodion menihrane is transferred from water to a solution 01 A:. inorganic electrolyte, no TT eight or volume changes can be determined with certainty, 1101' c m any be r!eterniined irhen the membranes are transferreii hnek to witer.
PHYSICAL CHEMISTRY O F MEMBRAXES.
I1
179
“Porous” collodion membranes do not exist in an air-dry state; one can nieasure only the n eight and volume change on transfer from TI ater to electrolyte so!ution and back into n ater. There iq no change of volume under these conditions, and the observed c-eight changes are quantitatively accounted for by the differences in specific gravity of TI ater and solution filling the pores. The follon ing conclusions can be dran n froni these experiments: Contrary to the general assumption, dried collodion membranes >well on netting with I\ nter, the extent of the SITelling T arying TIith difierent brands of collodion. On wetting n-ith n ater, dehydrated dried collodion nienibranes undergo fundamental changes; the air-free and the ifater-n etted dried membranes are structurally different entities. Once netted with TI ater the dried collodion membrane determining t h e e:ect,rocheinical behai-ior !)f membranes of porous character. This general concept n-as actu:.tlly used some time ago by Teorell (19) and Xeyer and Severs (7, 8) in ts3:iie very interesting attempts to put, the theory of electrochemical mcmhrane hehn\-ior oil a quantitative basis. The charge density of a given surface is iclentical with its exchange capacity per unit area. The base-exchange capacity of collcdion in the fibrous state and of collodion membranes under various conditions \\-as studied, iii order t80see how far the general concept of \\-hat’one may call the “fixed-charge theory” is supported by the facts (18), ancl lion- far the Teorell, Xeyer-Sievers theoretical concepts hold true ( l i a ) . The iriherent acidity-the “acid. number”-of different collodion preparations, i.e., the equivalent \\-eight of these preparations, n-as also st,udied in this connection (I-&). Rase-exchange experiments are ususlly carried out, in the follon-ing manner: First the material to be investigated for its base-eschange properties is brought into a well-defined chemical state by saturating it’ n-ith cations of a single kind. For experimental reasons it is advisable to bring the material into the state of free acidity, either by treatment Tvith hydrochloric acid or by electrodialysis. Knon-n quantities of such matserial are brought into contact ryith a neutral salt solution. The hydrogen ions from the exchange body are replaced by the ot’her cations in the solution, and the latter becomes acid. The quantity of titratable hydrogen ions in the superiiatant, solut’ion represents the actual base exchange, if no complicating react’ions occur. If a relatively concentrated electrolyte solution is used for this exchange, it is complete within the limits of experimental error on the first t’reatment of the exchange body with the neutral salt solution. The free acid in solution can be determined either by titration or elect’rometrically. The difficulties and limitations of thi? method when it is applied to very small quantities, as in the case of collodion, have been discussed critically elsewhere (14, 18) and therefore need not be considereJ here in detail. The base-exchange values in table 1 are given in milliliters of 0.01 -V socliuni hydroxide per gram of dry collodion; t’he loirest of these values must be considered t o be fictitiously high (14, 18). The acid number! and therefore also the mean equivalent weight, of a collodion preparation can be determined by bringing i t into the state of free acidity, as cliscussed before, and titrating samples of this material dissolved in a suit’able organic solvent (alcohol-acetone) with alcoholic potassium hydroxide solution (14). The corresponding values are given in t’able 1 under the heading “acid number” and are expressed in milliliters of 0.01 N potassium hydroxide per grain of dry collodion. The last tn-o columns of table 1 give a measure of the electrochemical activity of the various collodion preparation?. l n e results of the esperinients listed in table 1 and the conclusions based on these experimental data can be sunimarized briefly as follows : High base-exchange capacity is aln--nys found n-itli preparat’ions of great “electrochemical activity”; medium and low base-exchange capacities occur n-ith electrochemically active as ire11 as with inactiT.e preparations. The acid numbers, espressed in “1
PHYSICAL CHEMISTRY OF M E U B R I S E S .
183
I1
milliliters of 0.01 S potassium 113-droside per gram of dry collodion, vary from 1.0 for a highly purified collodion preparation of very low electrochemical activity t o 3.3 for a highly oxidized sample of very high activity. Acid numbers of about 1.5 (corresponding to an equinileiit weight of about G7,OOO) are found both with inactive commercial and with fairly active oxidized preparations. The TABLE 1 Acid number, base exchange, and a c t i c i t y o j varioits collodion p r e p a r a t i o n s (5) I ~~~
BRAXD OF COLLODION A S D PRETREATMENT (ALL PREPAR~TIOSS WERE PRECIPITATED FROM ETHER-+LCOHOL SOLUTIOSS ~ K DDRIED)
SO.
~
~
~
BASE EXCHAXGE (48 HR.) XL. 0.01 X SaOH PER GRAY OF DRY 0.01 N KOH PER GRAM OF :COLLODION (CAL DRY COLLODIOS 1 CCLATED F R O Y ’ pH V - ~ L U E S ) ACID NUMRER (CORRECTED FOR ASH) m.
1
Osmotic rise a i t h 0.25 Y sucrose
Anomalous osmotic rise with A l l 5 1 2
mm.
inn1 .
110 120 127
36
0 go13
108 110 114
2s 1s 22
0 004
115 121 130
52 40 60
0.28
130 140 15s
145 135 I78
0 11
135 144 158
1so lG 172
0 0OGlj
122 138 11s
137 11.3 153
0 0066
135 140 150
125 134 120
i
,\Ialliiickrodt “Parlodion,” commercial preparation
(6)
ELECTROCHEYICAL ACTIVITS
’
1.2
0 0016
I
&so4
40
15
I
,\lrtllincl;rodt “Parlodion,” boiled 8 hr. in 90% alcohol ( t x o alcohol changes)
1.0
Baker Collodion U.S.P., coiiiniercial preparation
1.5
Oxidized collodion; (Baker Collodion Cotton. “Pyroxylin”) oxidized 45 hr. with 1 31 S a O R r and boiled several times v i t h water
3.3
Oxidized collodion ( S o . 4)] washed eight times n i t h
2.9
I
1
1
I
5
1
I
I I
Oxidized collodion ( S o . 4 ) , waslieJ sixteen tiiiies with 957c alcohol
2,s
Oxidized collodion ( S o . 4), boiled 3 hr. n-it11 9070 alcohol
1.6
I
base-exchange capacity of the same preparation. in the fibrous state (as measured after -16 hr. of exchange time) varies from 0.0013 nil. of 0.01 S sodium hydroxide per gram of dry collodion for the mo?t inactil-e preparation up t o 0.26 nil. of 0.01 S sodium hydroxide per gram for the most actiT-e preparations. Thus the acid numbers over the TI-holerange investigated differ only in the ratio of 1:3.3, ivhereaq the bare-exchange values differ in the range of 1:200.
184
KARL SOLLNER
I n the inactive preparation only one in 7'10 acid groups is available for base exchange, in the most active collodion one group in 13; values between these extremes are found TI ith commercial and alcohol-purified osidized preparations. The high base-exchange capacity of the electrochemically active preparations is not so much due to their higher acid number as to their more open structure. This difference in structure can be ascribed to the presence of a small fraction of low-molecular-weight material. which inhibits normal formation and arrangement of the collodion micelle?. So far we hal-e considered the babe-exchange data Ivhich correspond to the base exchange as obtained after prolonged contact of the electrolyte solution with the various collodion preparations. The base exchange which is obtained on short contact of the collodion with the electrolyte solution is much lower (14, 18), olving to the fact that fibrouq collodion has not only eaqily accessible outer surfaces but also inner surfaces \ \ hich are accesbible only 11ith difficulty; in addition, there may be also some do^ base eschange from the interior of the micelles. Since the electrochemical propertieq of the collodion membranes are usually established after short contact o i the membranes Ivith electrolyte solutions, we are actually more interested in the magnitude of the short-time base exchanges than in final, long-time base-exchange d u e s . Unfortunately, short-time baseexchangc studies yield rather poorly definecl and somewhat arbitrary baseexchange values. Seyertheless, this much can be said: short-time base-exchange experiments lvith fibrous collodion indicate that the number of acid groups IT hich in membranes n-ould be available for the typical electrochemical membrane functions may be estimated to be about 50 t o 1000 times less numerous than those found in the 4s-hr. base-exchange experiments discussed before. On the basis of these experiments Ivith collodion in the fibrous state i t can be estimated that even Tvith the most active collodion preparations not more than one in 500 acid groups may be available for the typical membrane functions; with the less active preparations, this ratio can be estimated to be only one in 1,000,000 or more. It may be added parenthetically that these figures can be taken as a measure of the discrepancy betn een the real situation and a fundamental postulate of the homogeneous-phase theory as applied to the dried collodion membrane,namely, equivalence of all critical groups. VIII. IS EXPERIMESTLL TEST O F SOVE ASPECTS O F T H E TEORELL, JIEYER-SIEVERS THEORY O F ELCCTROCHEMICIL MEMBRANE BEHAVIOR
The experimental results which have been discussed in the foregoing section were in many respects contrary t o the original expectation and, more important, seemed to be incompatible v-ith the Teorell and Meyer-Sievers theories. An attempt \\-as therefore made t o clarify the situation on a quantitative basis. The base exchange of membranez m s compared with what should be an identical or at least a similar value calculated according to the Teorell (19) and MeyerSievers (7, 8) theories on the basis of potentiometric measurements. A short exposition of these theories must be given. I n essence the tivo concepts are identical. The Teorell, Xeyer-Sievers theory is today the outstanding
PHYSICAL CHEMISTRY OF JIEMBRAXES.
11
185
example of an attempt to put on a rational basis the most important electrochemical function of membranes,-namely, their electromotive action. Since this theory is highly involved, it is preferable to give a short outline of it’ as presented by Neyer ( 7 ) in a condensed form stripped of eyerything that is not of primary importance here. , , , Consider a membrane consisting of an acid high-molecular substance, for instance of pectin chains, of nhich the carboxyl groups have been neutralized with metallic cations, e . g . , potassium ions. The membrane then possesses fixed anions and mobile cations. The cations may therefore be displaced if it supply of others is maintained from one side: the membrane is cation permeable. The concentration of the fixed anions, calculated in gram equivalents per litre of the imbibed liquid, is a quantity characteristic for each membrane which ive will call the ‘selectivity c o n s t a n t , ’ d . If now the membrane be immersed in a salt solution, both ions of the salt will penetrate into i t ; the equilibria then obtaining may be calculated from the Donnan equation: the actual membrane behaves like a solution bounded by two ideal Donnan membranes through which t,he fixed ions cannot pass. . . . ” ( l . , . If a current is passed across the membrane, the transport of the electricity will be divided b e h e e n the t n o kinds of mobile ions in accordance with the relative numbers of ions passing through the membrane. The ratio, ~ ~ y ’ n . 4between the numbers of cations and of anions traversing the membrane, vihich n-e will call the ratio of the transport or ‘traversal’ numbers, may be determined by the same methods as those used for the determination of transport numbers in a solution. iic/n.d depends on the rates of migration of the mobile ions and on their number; as mentioned above, the latter is dependent on the concentration of the ions in the surrounding liquid. We then obtain
n-here CCand C.4 are tlie rates of migration, c the molar concentration of the salt in the surrounding liquid, and A the selectivity constant . . . ” y is tlie concentration of mobile anions in the niembrane. The quotient C:C/CAand the selectivity constant of a nieiiibrane “can be determined by measuring the traversal numbers a t different concentrations. The potentiometric method is the best; the potential set up when the membrane separates two solutions of the same salt, but o i different concentrations is measured, the absolute concentrations being varied in such a ~ v a yt h a t their ratio is kept constant. When there is no ionic selectivity the potential is determined only by the quantity C:C/C.~, which is dependent on the absolute concentration; the greater the value of -1 as conipared with the external concentration, the more marlced will be the ionic selectivity. A , the selectivity constant, and C‘ciC.1 . . . can be quantitatively determined . . . nstructed plotting as ordinates the potential differences (“E”) measured between trvo solutions of the same binary electrolyte, the concentrations (c1, cz) of which A are always as 1 : 2 , and as abscissae the quantity log -’ For a n electrolyte, the two ions of C1
which have equal mobilities in tlie membrane (i.e., l : c / U ~= l ) ,a certain curve will then be obtained, while other values of U C / ~will A result in other curves lated curve for U C / U A = 1 is given belon- in figures 6 and 7 . ) . . . T o determine A and UclC’a for :in unknown membrane it will then only be necessary t o determine several values of E for different absolute vales of CI, the concentration cz being alvays kept equal t o 2cl. The observed values of E are then plotted against c1 (ordinates) using the same co-ordinates as before, and then the experimental curve is displaced sideivays (parallel to the abscissae) until i t has been successfully brought into co-
186
I i h R L SOLLNER
incidence with one of the curves already dravin; interpolation niay be necessary in this procedure. I n this way Cc/C.h is determined, the value depending only on the shape of the curve. The amount of the displacement as read off on the abscissa gives log A , and therefore $. . , ”
.
The quantity A , according to Neyer and Sievers, can thus be determined on the basis of potentiometric measurements. T’alues obtained in this manner are designated as A,. On the other hand. d by definition is the concentration of the fixed ions in the aqueous part (pore space) of the membrane, or more correctly the concentration calculated from the number of equivalents of anions (or anionic groups) fixed immovably t o the pore n-alls, divided by the pore space (in liters) of the same membrane. Electrically these anions are compensated for by an equivalent quantity of cations. These cations-potassium ions in the example given by Meyer-can be replaced by other cations if the membrane is brought into a suitable electrolyte solution. If one is able to determine the base-exchange capacity of all the pore \I alls in a membrane, Le., the base-exchange capacity of the membrane, and if one knows its pore space, one is able t o calculate the A value of the lleyer and Sievers theory, A being the base-eschange capacity in equivalents divided by the pore volume in liters. The A values vhich are calculated from base-exchange studies are denoted A b . If the theoretical assumptions on which the theory is based are correct, il, derived from potential measurements must be identical n-ith ~ 4 derived b from base-exchange studies. The main question 11-hich one can hope to decide by the comparison of the t\\o selectivity constants is n hether the Teorell, lleyer-Sievers theory is inherently a correct representation of the physical facts n-hich lie behind the observable potential-concentration relationships, or n hether it is only a formal n-ay of bringing the latter relations into a fictitious, fornially correct framework. Two kinds of measurements were performed n ith the same membranes : First, the potentiometric studies Tvere made; then the base exchange of the same membranes was determined, and in order to obtain the water content, the \vet and dry weights of the membranes were measured. From these data the selectivity constants, -4, and &,, u-ere obtained as explained in principle above. Since the de5 The water content of collodion membranes represents only a maxiniiini \ d u e for the available pore space. Some n a t e r , undoubtedly, is “bound” to the collodion and not of high n.ater available for the typical membrane functions. With L ‘ p o r o ~ smembranes ” content the fraction of “bound” water is negligibly sniall; TT ith dried membranes the “bound” v-ater may be a sizable fraction of the total water content (see C . \Y.Carr and IC. Sollner: J. Gen. Phgsiol. 27, 77 (1013)) In the calculations reported helon-, nhich involve the pore water content of the membranes, the assumption is made that the total \\ ater content is pore water. If we make the eyaggerated assumption t h a t 50 per rent of the water content of the “dried” nienibranes is “bound” Ti-ater, the figures for cases S o . 1-4 in column 3 of table 2 m-ould be twice as large as given, and the figures in colunin 6 would be half of the present values. This, however, would in no ~ a affect y the essence of the conclusions which Fill be dran n below from these figures
PHYSICAL CHEMISTRY OF MEUBRSNES.
187
I1
tailr of the esperiinental work involved and of the calculation niethods used will be published shortly elsewliere (17a), not many details need to be given here. Figures 6 and 7 sho\v in a graphical manner the results of experiments n-ith membranes prepared from yarious collodion preparations and also indicate how in a graphical manner the selectivity constant, -ID, is obtained. -1b is determined from the base-exchange capacity of the niembrnnes and their known water content.5 Tables 2 and 3 contain theqe magnitudes together nith the ratio A41,’.lb.
-2
-I
0
‘04
*P
+I
+2
+3
A FIG.6 . Potential-log 2 curves of “dried” and “porous” meiiibrniies prepared from various collodion c1
preparations.
According to the Teorell, lleyer-sievers theory, d p / - I b should be 1, or, considering the uncertainties of the theory, this ratio should not deviate too strongly from unity. -1glance at the last column of tables 2 and 3 shon-s that this ratio is spread over three orders of magnitude, varying from 0.08 to 107; in the majority of the caSes it is greater than 30. In evaluating the meaning of the -dD/-4b ratio, two facts must be considered. It i v m shon-n repeatedly (14, 18) that the long-time base-exchange yalueq, as they are used in tzbles 2 and 3 , are much higher than those obtainable after the
188
KARL SOLLNER
shorter periods 11-hich it takes the membranes to assume their final electrochemical properties. Therefore one may estimate that the Ab values which determine the functional behavior of the membranes are only 5 to 20 pec cent of those given in tables 2 and 3. Correspondingly, the A,/Ab values would be about ten times greater. I n addition there is another factor which tends to increase this discrepancy in the case of the membranes having only a very low base-exchange capacity. I n these cases the -1b values are fictitiously high; the
-2
0
-I
‘09
*P
+2
+3
c,
=1
FIG.7 . Potential-log 2’ curves of “oxidized membranes” of various porosities c1
true base exchange cannot be determined accurately 11-ith the available methods on account of difficulties discussed elsen-here (14,17a, 18). For this reaSon alone, values may easily be 10 or even 100 times lower than those given in table 2 . It is therefore a conservative estimate that the d,/d ,7 ratio in the case of the membranes having low base-eschange capacity is two, probably three, orders of magnitude smaller than given in the table; in the caqe of the membranes with high base-eschange capacity the A , / A b ratio is probably not much less than one
+4
189
It
PHYSIC \ L CHE.\IISTIZT OF RfEllBR1NES.
order of magnitude smaller than represented in the table. Considering this, we find that n reasonable agrcemeiit betn-ccn the :inti the .I b v d u e i exists only
(2) BRISD
O F COLLODION ASD TSIT OF MEMBRAhTS
Purified “Parlodion”; tlrietl membranes . . . . . . . . . . . . . . .1 Baker Collodion U.S.P.; dried membranes. . . . . . . . . . . . . . . i. Osidized collodion; drirtl inem-! branes. . . . . . . . . . . . . . . . . . . Oxidized collodion, purified; I dried membranes. . . . . . . . Purified “Parlodion”; porous I membranes . . . . . . . . . I Raker Collodion I-.S.P.;~ ) o r o u s inembrnnes . . . . . . . . . . . . Osidized collodion; porous iiicnibrnnes . . . . . . . . . . . . . . . 1 Osidized collodion, purilietl: porous membrniies. . . . . . ~
S.2“
1
62”
~
10.5”’
< 1 6:* ,
~
lti
I . .i
>7 3
~
,
3.3 ~.
.......
Coinpitre footnote 5 .
w e case, S o . 3 of table 3 . I n d l other cases t,he tlisagrcement varies bet\reen rilatively lon- figures :mcl figires as high as 10,000or more. I t does not seem t o be i\ithout sig1iific:mce that the disagreement i,e?!r.ecn tlic iii
190
KAItL SOLLXER
r l , and A b values is the smaller the greater the porosity of the membranes and the higher their base-exchange capacity. The attempt to verify the Teorell, JIeyer-Sies-crs theory as applied t o a variety of collodion membranes by the comparison of selectivity constants arrived at on the basis of two different, esperimentallp independent methods has failed completely. We shall discuss in a subsequent paper the probable significance of this disagreement. SUMMARY
1. Tarious new eqwimental approaches furnish proof that the characteristic behayior of “dried” collodion membranes with solutions of inorganic electrolytes and those non-electrolytes which are not strongly adsorbable must be explained on the basis of the porous, micellar-structural character of these membranes. The homogeneous-phase theory of membrane permeability cannot be applied to dried collodion membranes. 2. Air-dried “dried” collodion membranes s w l l slightly n hen immersed in water. S o specific sn-elling effect is observed with solutions of strong inorganic electrolytes and those organic conipounds TT hich are but weakly adsorba le; ithin these limits I\-atcr-wetted dried collodion membranes behave as rigid nonswelling structures. ,Idsorbable solutes in many instances cause pronounced specific slyelling. The behavior of highly porous collodion membranes towards the solutions of the various solutes i3 analogous t o that of the dried membranes. 3. The diqiociable groups located in the interstices of the membranes which determine the electrochemical behavior (“actil-ity”) of collodion membranes can be determined by base-exchange measr~emeiits. The base-eschange capacity of various collodion preparations in tlie fibrous state, as measured after 48 hr. of exchange time, varies f i om 0.0013 nil. of 0.01 S sodium h y d r o d e per gram of dry collodion for thc most inactive preparation up to 0 26 nil. of 0.01 ;L’sodium hydroxide per gram for the iiio3t active prepar::tion. High ba+,e-exhange capacity is aln ayq found 11ith preparations of great “clcctrochemicai actii*ity”; medium and 101v base-exchange capacities occur I\ ith electrochemically active as ell as nith inactive preparation.. -1. The inherent acidity of various collodion preparatio~:s, their “acid number.” TT a i Jetermiiicd by electrometric titration. Collodion in the acidic state u-aq titrated in an organic solvent mixture 11ith alcoholic p(,taAum hydroxide, using a c~uinhydroiieelcctrodc. The acid number-, c \ ; p ~ e ~ ‘ -in~ dmilliliters of 0.01 S potaisiurn hytli.o\ide per grain of dry collodion, x ary irom 1.0 for a highly purified collodion preparation of very 1011- clectroclieiiiical adivity to 3.3 for a highly ovidized sainplc of x-ery high activity. Tliu- the acid numbers over the nhole range inyeqtigated differ only in the ratio ot 1:3.3. uhereaa the baseexchange values differ in the range of 1:200. 6. The high base-eschange capacity of the electrochemicaiiy active preparations i-, clue not 50 murli t o their higher wid niimber as to their more open structurc.
PHTSIC.iL CHEXISTRY O F JlEJlBlIISES.
I1
101
6. Short-time base-eschange experiments indicate that in membranes prepared even from the most’active collodion not more than one in 500 acid groups may be available for the typical membrane functions; with the less acti\‘c preparations this ratio is estimated to be as high :is one in 1,000,000or more. 7. The Teorell, Meyer-Sierers theory characterizes the electiwhernical behavior of membranes by their selectirity constant -4,,, ivhich is derived conventionally from concentration potential measiiremeiits at varioiis concentration leyels. The selectivity constant may, lio\wver, be derived also from entirely independent, different, experimental data:---nnniely, base-exchange stndies. The constants arrived a t in this second way are designated as l i b . The selectivity constants derived by t’liese tn-o methods must be in reasonable, a t least semiquantitative agreement if t,he basic assumptions of the theory :ire correct. The selectivity constants -4, and Ad b n-ere determined for eleven different sets Of membranes of different electrochcmical activity and of diffcrmt (8.2 t o SO volume per cent) water content. The potentiometric seltctivitj- coii.,4nnts .dJ are in most cases s e r e i d ordew of magnitutle greatcr than the corresponding --Ib values. With membranes of great porosity m d high elect rochernical activity the ,4 b values approach at least in ordcr of niagiiituck t h e 1 vdiies. ’rhe cause of the unexpectedly large discrep:mcy h c t \ \ v n tlw .,t ;, ant1 ’1 h ~ ~ ~ 1 \vi11 1 1 be 1 ~ dis~ cussed i!i ?: subsequent papc’i’. ~
),
J
