357 The Microscopic Structure of Semip Membranes and the Part Played by Surface Forces Osmosis .
By Frank Tinker , M.Sc .
( Communicated by Sir Oliver Lodge , F.R.S. Received October 13 , 1915 .
) [ Plate 6 .
] In the following investigation an attempt has been made to arrive at some of the fundamental facts associated with the mechanism of osmosis .
Hitherto very few experimental observations have been made in this connection .
Raoulfc , Flusin , Kahlenburg , * and a few others have shown , however , that certain membranes such as parchment , gelatin , and rubber absorb the liquids to which they are permeable , and are impermeable to liquids which they do not absorb ; Bigelow and Bartellf have shown that , under certain conditions , the rate of flow of water through membranes such as copper ferrocyaniae obeys Poiseuille 's law for the rate of flow through capillary tubes ; whilst Beutner , Donnan , J and others have proved that certain precipitation membranes can act as electrodes , reversible with respect to various ions .
The further questions which are dealt with in the present communication are experimental ones : such as what is the size of the colloidal particles of which a semipermeable membrane is composed , and how is the membrane built up from those particles ; to what extent does a membrane show the properties of the gelatinous precipitates or gels as ordinarily prepared by bulk precipitation , and how does its structure differ from these ; how is the structure of the membrane altered by variations in the method of its formation , the nature of the solutions bathing it , and the treatment to which it is subjected ; and what is the size of its pores , the extent to which they are under the control of surface forces , and therefore of adsorption phenomena also .
* Raoult , ' Comptes Rendus , ' vol. 121 , p. 187 ( 1895 ) ; Flusin , ' Comptes Rendus , ' vol. 131 , p. 1110 ( 1901 ) ; Kahlenburg , ' Journ. Phys. Chem. , ' vol. 10 , p. 141 ( 1906 ) .
t Bigelow , 'Journ .
Amer .
Chem. Soc. , ' vol. 29 , p. 1675 ( 1907 ) ; Bartell , 'Journ .
Phys. Chem. , ' vol. 15 , p. 659 ( 1911 ) .
I Beutner , 'Journ .
Phys. Chem. , ' vol. 17 , p. 344 ( 1913 ) ; Donnan and Allmand , 1 Trans. Chem. Soc. , ' vol. 105 ( 2 ) , p. 1941 ( 1914 ) .
VOL. XCII.\#151 ; A. ' Mr. F. Tinker .
The Microscopic Structure 1 .
Microscopic Examination of Precipitation Membranes : Apparatus , Structure of Membrane , and Size of Grain .
In an investigation covering the foregoing ground it was thought desirable to limit the initial enquiry to the precipitation membranes first studied by M. Traube .
These membranes comprise the various silicates , tannates , and ferrocyanides of copper , iron , lead , etc. , and also tannate of glue , barium sulphate and silver chloride .
They show varying degrees of permeability to salts in solution .
Copper ferrocyanide is not permeable to either of the solutions from which it is formed , nor to barium chloride , calcium chloride , ammonium sulphate , barium nitrate , sugar , etc. , but it is permeable to potassium chloride .
Tannate of glue is permeable to most of the above substances , bub not to potassium ferrocyanide .
The metallic silicate and tannate membranes show an intermediate degree of semipermeability , in each case being practically impermeable to the salt solutions from which they are formed .
Barium sulphate , when used for clogging the pores of a tannate of glue membrane , renders the latter impermeable to ammonium sulphate and barium nitrate , in addition to potassium ferrocyanide ; whilst if silver chloride is precipitated in the pores of a copper ferrocyanide membrane , the latter will no longer allow potassium chloride to pass through it.* Fortunately , these membranes can be very easily prepared in thin transparent films suitable for microscopic examination .
In a preliminary set of experiments this was done by forming the semipermeable film in between two drops of the mutually precipitating liquids , e.g. , copper sulphate and potassium ferrocyanide , the one drop being placed on the microscope slide , and the other on the cover-glass .
On bringing the two drops together by gently lowering the cover-glass on to the slide , the film precipitated at the surface of contact of the drops was found to be quite thin and transparent , and quite stable enough to prevent further mixing and precipitation of the two solutions .
The films so prepared were first examined by the ordinary microscopic methods in which transmitted light is employed .
Several of them , notably the various silicates , showed a distinctly granular structure , very similar to a layer of sand particles placed side by side , the average size of the grain being different with different membranes .
The films of the ferrocyanides of copper and iron , however , seemed to be perfectly transparent and continuous .
" No structure could be detected with certainty , the particles forming the film * These facts are taken from M. Traube 's paper on these membranes ( ' Archiv fiir Anat .
Physiol , und Wissenschaft .
Medizin , ' pp. 87-165 ( 1867 ) ) .
of Semipermeable Membranes .
being apparently so small that the simple methods employed were not good enough to render them visible .
But the fact that the structure of some of the films had been resolved , rendered it likely that all might be resolved by improving the microscopic technique .
Two great improvements were made in the apparatus , in the lenses , and in the illuminating system respectively .
Through the generosity of Sir Oliver Lodge , Messrs. Carl Zeiss supplied me with a specially selected oil-immersion fluorite objective of great resolving and penetrating power .
Through the same agency I was also enabled to obtain a Zsigmondy-Seidentopf ultramicroscope* so that I might utilise its illuminating system , and attach it to the ordinary microscope .
In this illuminating system , which is the most perfect yet devised , the source of light is an intensely illuminated narrow slit which can be reduced to either point or line dimensions , thereby enabling the lenses in the optical system to exert their maximum resolving power .
In setting up this new and improved apparatus for the examination of the semipermeable films the ultramicroscopic outfit was made the basis .
To adapt it for the purpose in hand , the special microscope and separate Abb4 condenser supplied for the ultramicroscopic examination of colloids were removed from the optical bench , and an ordinary Zeiss microscope mounted instead .
A sketch of the complete apparatus so put together is given in Diagram 1 .
The beam of light from a powerful arc A is focussed by means of a lens B on to a graduated and adjustable slit C , which can be shortened and narrowed down to any desired extent .
The faint beam of light proceeding from the slit is then condensed twice ; first by means of the lens D , and subsequently by means of the Abbe condenser of the microscope E ; from which it passes on to and through the semipermeable film mounted on the microscope stage .
It was found advisable to reduce the length and width of the slit to as great an extent as possible , short of destroying the transmitted beam altogether .
The best results were obtained with a slit 1 mm. long and three-hundredths of a millimetre wide .
* A complete description of this instrument is given in the Zeiss catalogues , Micros .
227 and 228 .
Mr. F. Tinker .
The Microscopic Structure Much better films were obtained by using a Thoma 's hsemocytometer ( Diagram 2 ) instead of the simple microscopic slide and cover-glass used in COVER CLASS F/ LM Diagram 2 .
the preliminary experiments .
So prepared , the films were quite uniform and free from creases , and also quite stable when mounted vertically with the microscope , as in Diagram 1 .
Their thickness was about one-thousandth part of a millimetre ( 1 jj .
, ) .
Incidentally , it might be mentioned also that the solutions used for precipitating the films were usually decinormal .
More satisfactory films were obtained when the acetates of the metals were employed instead of the salts of the mineral acids .
The method of conducting the microscopic examination of the films prepared in the above way was as follows : The film was examined first by means of a No. 7 dry objective and a No. 5 eyepiece , in order to find a good portion free from furrows , and then more thoroughly by the selected Zeiss oil-objective already mentioned .
If the film was good enough , it was subsequently photographed by substituting a projection eyepiece for the No. 5 lens , and bringing the camera into position as shown in Diagram 1 .
Specimen photographs of the films are shown in Plate 6 , figs. 1 to 6 .
These photographs were usually taken about 15 minutes after the film had started to form .
An inspection of them shows that precipitation membranes are not as continuous and homogeneous in structure as commonly supposed .
They consist of collections of very small particles of precipitate , which are just large enough to be within the limit of microscopic vision ( 0T / jl ) .
These particles are of slightly different sizes , and are apparently aggregated together as closely as possible .
They approach the spherical in shape , but where they touch they seem to show a tendency to be flattened , and to cling together , just as if they were attracted to one another .
Moreover , the precipitate particles in the various films do not vary greatly in degree of fineness from film to film .
The average size of the particles in any one film is in all cases between 0T and 1*0 / jl .
The order of size , the colour of .
the film , and the approximate diameter of the particles , are given in the subjoined Table .
It will be seen from the Table that copper ferrocyanide and Prussian blue are the finest grained membranes , whilst tannate of peptone\#151 ; comparable to Traube'* tannate of glue\#151 ; is the coarsest .
The particles composing the former range from 0T to 0'4 ^ in diameter , whilst those in the latter lie of Semipermeable Membranes .
361 TaMe I.__Showing Order of Size of Particles in Precipitation Membranes .
Membrane .
Colour .
Size of particles .
Prussian blue Blue 0 T to 0 *4 fx .
Copper ferrocyanide Ferric si lien , the Chocolate red Yellowish white 0 T to 0 *4 fx .
Lead ferrocyanide White Copper silicate Greenish blue 0 2 to 0 *5 fx .
Tannate of iron Black Tannate of copper Dark brown Lead silicate White 0*6 to 0 *8 p. Lead tannate Dirtv w'hite White Peptone tannate 0*5 to 1-0/ u. Barium sulphate White 1 Coagulate into large particles , J 3 or 4 fx in diameter .
Silver chloride White 1 between 0*5 and 1-0 / x. Generally speaking , the iron membranes are slightly finer-grained than the copper , and copper membranes finer-grained than lead .
In the same way , the ferrocyanide membranes are not as coarse-grained as the silicates , and the silicates not as coarse as the tannates .
Although all the particles in the precipitation membranes are apparently visible by the microscopic method already outlined , it seemed advisable to ascertain whether there are any free unaggregated particles of ultramicroscopic dimensions ( i.e. less than 0T / x ) in such membranes .
In order to test this point , freshly precipitated copper ferrocyanide was shaken vigorously with excess of water for a few hours in a butter-churn driven by a small motor .
A microscopic examination of the emulsified mixture showed that the copper ferrocyanide in it was very difficult to break up .
Free particles of the same size as those in the membrane were present , however , and showed a vigorous Brownian movement .
In order to determine whether any uljramicroscopic particles had also been washed out of the aggregated precipitate , the emulsion was centrifuged until all particles visible by the microscopic method outlined above had been brought down .
When the centrifuged liquid was clear when examined by the microscope\#151 ; therefore containing no particles greater than 0T fi\#151 ; it was examined for the colour of copper ferrocyanide in the 18-incli cell of a Lovibond tintometer .
The tinge of colour , though slight , was too small to be measured .
It would hence appear that the small visible particles , of which precipitation membranes are made* , are totally fine suspensoid in character rather than colloid .
It would seem as if there are no free or unflocculated colloidal particles amongst their number at all .
But it must be remembered that the above experiment was performed in presence of traces of potassium sulphate , which would keep such ultramicroscopic copper ferrocyanide Mr. F. Tinker .
The Microscopic Structure particles precipitated .
But the fact that the membrane particles may themselves be flocculated ultramicroscopic particles ( see later ) .in no way detracts from the above observations as to membrane structure , for potassium sulphate is also present when the copper ferrocyanide membrane is formed from copper sulphate and potassium ferrocyanide .
The average size of the precipitate particles in semipermeable films is accordingly about the same as that of the metallic suspensions formed by the coagulation of colloidal metals , of the smallest dust particles in freshly distilled water , of the particles in gum gamboge and pigments in general , * and also the same as that of the particles of oil in kerosene emulsions and freshly shaken suspensions of water in kerosene.f In all probability this size is such as would be imposed by the combined action of the electrical charge on the particles and the tension at the boundary surface between the particles and the enveloping medium .
2 .
The Colloidal Properties of Precipitation Membranes : Relation to Sols andOrdinary Gels .
It is well known that the common precipitation membranes are colloidal in nature .
They are gels which are formed from colloidal solutions or sols by the segregation of the ultramicroscopic particles suspended in the latter .
But inasmuch as they are really very thin films of gel , and are formed by surface precipitation rather than bulk precipitation , it is to be expected that they will have properties peculiar to themselves , in addition to the properties of gels in general .
We will , therefore , consider the relation of the membrane to the sol on the one hand and the gel as ordinarily prepared on the other ; confining attention mainly to copper ferrocyanide .
Copper ferrocyanide sol was first described by Graham , J who showed that it is best prepared by dissolving the washed gelatinous precipitate dn just sufficient neutral ammonium oxalate solution , the latter being subsequently dialysed away .
In this way a saturated sol containing 3 or 4 per cent , of copper ferrocyanide can be obtained .
In size , the ultramicroscopic particles in the sol are not far below the microscopic limit of visibility .
The largest of them can be seen quite readily by means of the paraboloid condenser and dark ground illumination on the ordinary microscope .
When examined by the Zsigmondy-Seidentopf ultramicroscope , their size would appear to range from about 50 to 90 / x/ x. None * These facts are taken from Zsigmondy 's book on 'Colloids and the Ultramicroscope ' ( Alexander 's Trans. , pp. 132-136 ) .
t Lewis , ' Zeit .
Colloid .
Chemie , ' vol. 4 , p. 211 ( 1909 ) .
f ' Chemical and Physical Besearches , ' pp. 583 et seq. of Semipermeable Membranes .
of the copper ferrocyanide is in true solution\#151 ; as distinguished from colloidal solution\#151 ; since the characteristic ferrocyanide colour is removed completely by shaking the solution with a little Seitz absorbent asbestos .
This substance , which is much used in technical practice for clearing the haziness from wines , has little or no effect on the concentration of true solutions or even solutions of emulsion colloids .
It would appear therefore that the whole of the ultramicroscopic particles in copper ferrocyanide sol are in the submicron range .
It is also negative in character , * being precipitated by hydrion and the multivalent cations with the greatest ease .
The general properties of copper ferrocyanide gel are too well known to merit description .
To examine its structure microscopically , a thin film of it can be prepared in the way usually adopted for examining the structure of gels.f The structure seems to be that of a somewhat irregular network having a mesh of the order 05 to 1*0 / x. In this respect it seems to be similar to most of the other ordinary gels such as gelatin , silicic acid , etc. , which have been examined with great thoroughness by Biitschli , Hardy , van Bemmelen , von Weimarn , Quincke , Zsigmondy , Pauli and others .
The difficulties attendant on the preparation of gels , and the ease with which their real structure is altered in the handling of them have , however , caused their structure to be a subject of much dispute .
But there seems to be general agreement now that most of them consist of a lattice work system in which a semi-solid phase encloses a more liquid phase .
If we now consider the membrane , it will be seen that the small particles revealed by the refined microscopic examination are too large to be the individual ultramicroscopic particles which are suspended in the sol .
Each precipitate particle in the semipermeable film is rather an aggregate formed by the flocculation of the latter ; so that the membrane as a whole can be regarded as formed by the flocculation of sol particles into small precipitate particles just microscopically visible , whilst these small precipitate particles themselves aggregate together to form the continuous film .
The whole of this flocculation process by which the membrane is formed can be followed quite readily by the microscopic method already outlined .
The first stage in the production of a semipermeable film is the formation of a layer of ultramicroscopic particles at the boundary surface separating the two mutually precipitating liquids .
This is evidenced in the case of a copper ferrocyanide film by the appearance of a brick red tinge at the boundary surface some two or three minutes before any particles in the film * Duclaux , ' Comptes Rendus , ' vol. 143 , pp. 296 , 344 ( 1906 ) .
t An object glass is placed at the bottom of a shallow vessel containing the saturated sol , and a very thin layer of gel allowed to separate out on it .
364 Mr. F. Tinker .
The Microscopic Structure are microscopically visible .
After the first few minutes , precipitate particles produced by the flocculation of these colloidal particles begin to appear\#151 ; the process being fairly rapid at first , but going on at a gradually diminishing rate for some hours .
The gradual change which takes place in the appearance of the film with the lapse of time is shown in the two photographs of the same portion of a copper ferrocyanide film taken half-an-hour and eight hours respectively after formation ( figs. 1 and 2 , Plate 6 ) .
The flocculating effect shown in these photographs is a very marked and unmistakeable one .
But it is shown more strongly still by crystalloidal precipitates such as barium sulphate .
An inspection of fig. 6 ( Plate 6 ) indicates that after several hours the whole of a barium sulphate film has broken up completely and aggregated into large particles 2 or 3 microns in diameter .
But , as is now generally recognised , the difference between colloids and crystalloids in this respect is merely one of degree .
Furthermore , the rate at which the flocculation process takes place depends on the concentration of the precipitating solutions and their nature .
In the case of copper ferrocyanide the microscopically visible precipitate particles appear much sooner when strong precipitating solutions are used , and in presence of traces of acids or potassium sulphate .
They do not appear so readily when the acetate of copper is employed instead of the salts of the mineral acids ; neither do the particles grow to quite the same size .
The membrane is disintegrated by ammonium oxalate solution or on washing too thoroughly with water .
For it to remain stable traces of copper sulphate or potassium sulphate must be present in the solution bathing it .
Precipitation membranes accordingly show most of the properties of the corresponding gelatinous precipitates or gels which are produced when the mutually precipitating liquids are mixed in bulk .
Both the membrane and the gel proper are formed by the flocculation of colloidal particles of ultramicroscopic dimensions , and the behaviour of the aggregates so formed is the same in each case .
But although therefore we are justified in defining a precipitation membrane as a very thin layer of gel , it will be seen that the physical structure of the membrane is not the same as that of the gel produced by bulk precipitation .
There is a fundamental difference in the method of formation of the two , which seems to cause a fundamental difference in texture and geometrical form .
During the formation of the membrane , the interaction of the two liquids and the flocculation of the colloidal particles are limited to a plane ; whilst when the ordinary gel is produced by precipitation in bulk , these colloidal particles come together from all directions in the three dimensions of space .
This difference would seem to cause the texture of the membrane to be much finer than that of the gel proper .
Whereas the of Semipermeable Membranes .
ordinary gel has the structure of a somewhat open lattice-work , the precipitation membrane has the more irregular structure of a layer of sand grains laid close together side by side ; the nearest approach to the lattice-work structure being the tendency of the small particles to cling together and flatten at the point of contact .
It is probable that the lattice work structure is incompatible with the property of perfect semipermeability ; for if it were formed even temporarily the mutually precipitating liquids would leak into the mesh-work and precipitate one another still further , thus blocking up the cavities .
But in spite of these differences in physical structure between semipermeable films and gels proper , it is quite clear that there are the same binding forces operating in membranes as give rigidity to ordinary gels .
Copper ferrocyanide and Prussian blue films are possessed of great tensile strength\#151 ; a fact which accounts to a large extent for their success as membranes .
The silicate films are not so strong as the ferrocyanide films , and break up much more readily ; the tannate films fall to pieces more readily still ; whilst a barium sulphate film breaks up almost immediately after formation .
But it is important in this connection to remember that the stability of any given membrane film is determined almost wholly by the nature of the liquid surrounding it .
If , for example , the barium sulphate film is precipitated from solutions of sulphuric acid and barium acetate in alcohol diluted with a little water , the gelatinous film produced at the boundary surface remains quite stable and elastic , just as does the corresponding gel produced from similar solutions by bulk precipitation.* For the precipitation membrane to be stable , the substance of which it is composed must have a negligible true solubility in the surrounding medium .
The particles of the membrane will then undergo no change , since crystallisation is inhibited by the fact that molecules cannot be transferred from one particle to another by intermediate solution .
If on the other hand the substance of the membrane has a slight solubility in the medium , molecules can be transferred , and the large particles can grow at the expense of the smaller onesr the whole film meanwhile crystallising and breaking up .
Though these effects are known to occur in the case of colloids considered in extenso , it does not seem to have been generally recognised that they can take place in the case of membranes .
But there is no doubt that the nature of the liquid bathing the membrane can cause fundamental and irreversible changes in its colloidal character to take place , and also exercise a great influence on its capacity for showing osmotic properties.j- * Kato , 4 Journ. Chem. Soc. , ' A , vol. 11 , p. 850 ( 1910 ) .
+ In this connection we await wTith great interest the results of H. N. Morse , who announces that such changes in structure take place when a copper ferrocyanide Mr. F. Tinker .
The Microscopic Structure 3 .
Size of Pores in Membranes .
From what has been said above about the structure of semipermeable films it will be seen that there are two classes of pores perforating a precipitation membrane : ( 1 ) the larger pores consisting of the interspaces between the microscopically visible precipitate particles , ( 2 ) the smaller pores penetrating !
these visible particles and formed between the ultramicroscopic particles composing them .
There is , however , no sharp dividing line between these two classes , since the largest individual submicroscopic particles have a size ( 90/ / , / / , for copper ferrocyanide ) approaching that of the smallest precipitate particles ( 100/ / , / / , for copper ferrocyanide ) .
If we assume the precipitate particles and the submicroscopic particles from which they are formed to be approximately spherical , we can form an estimate of the diameter of both classes of capillaries .
To do this , consider three particles A , B and C ( Diagram 3 ) , which are of equal size and touching Diagram 3 .
one another .
The triangular interstice PQE formed between them is typical of all the pores in the membrane , whatever size they may be .
The centre 0 of the interstice is the point of greatest distance from the surfaces of the particles ; hence the perpendicular distance OP can be regarded as the radius of the interstice or pore , since it would be equal to the radius of a spherical membrane Is bathed by various aqueous salt solutions .
They also take place with zinc ferrocyanide and manganese ferrocyanide even in water ( ' Report on Osmotic Pressure of Aqueous Solutions/ Chaps .
IY and XI , Carnegie Institute of Washington , 1914 ) .
of Semipermeable Membranes .
particle which would just be able to go through the pore without being blocked mechanically .
Now it can be shown by simple geometry that the radius OP is approximately equal to 0'08 times the diameter of the particles A , B , and C , so that the pore diameter would be twice this , or about 016 times the size of the particles enclosing the interspace .
If , therefore , we know the size of the particles of a precipitation membrane , we can easily calculate the diameter of its pores .
We will apply the above result to find the approximate maximum , minimum , and average pore diameters in a copper ferrocyanide membrane .
The largest pores are formed in between the largest precipitate particles , and their maximum diameter is 0*16 x 0 4/ x , which is equal to 0*06 / x or 60/ x/ x. The smallest possible pores are enclosed by the smallest submicroscopic particles , their minimum diameter being 016 x 50 / x/ x or 8 / x/ x. In the same way the average diameter of the capillaries in between the submicroscopic particles is about 10 / x/ x , whilst that between the precipitate particles is about 48 / x/ x. The average for the membrane as a whole is much nearer the former .average , since the number of capillaries between the submicroscopic particles is much greater than the number between the precipitate particles .
Its value is probably between 15 and 20 / x/ x. If now we consider a series of precipitation membranes , it is evident that their average pore size increases as the average size of their component particles increases .
The order of the membranes with respect to pore diameter is therefore the same as that given for the size of the particles in Table I. Prussian blue and copper ferrocyanide have the smallest pores , whilst peptone tannate has the largest .
In the latter case the largest pores have a diameter approaching 160 / x/ x. If we proceed still further and compare the list of membranes placed in order of increasing size of particles and pore diameter with Traube 's list of their capacity for showing osmotic effects ( loc. cit. ) it will be seen that there is also a close relationship between the pore diameter of a membrane and its capacity for acting as a semipermeable film .
Copper ferrocyanide and Prussian blue have the smallest pores , and are also the most efficient of membranes .
The silicates and tannates have larger pores than the ferrocyanides , and are not such efficient differential septa ; whilst peptone tannate , with the largest pores of all , is the least efficient membrane of all .
This conclusion is in harmony with Bartell 's observations on the relation between pore diameter and the capacity for showing osmotic effects.* Using an indirect method of measuring the pore diameter based on Jurin 's Law , this investigator found that as the * k Journ. Phys. Chem. , ' vol. 16 , p. 318 ( 1912 ) .
368 Mr. F. Tinker .
The Microscopic Structure # # \#187 ; pore diameter of porcelain or porcelain clogged with various precipitates decreases , the osmotic effect simultaneously increases .
With copper ferrocyanide osmotic effects become noticeable at a pore diameter of about 900 / x/ x\#151 ; a figure which Bartell further showed is more or less the same for all membranes .
The osmotic effect increases very rapidly as the pore diameter gets less , and has become quite appreciable at 180 / x/ x , which was the least average diameter Bartell worked with .
It is quite understandable , therefore , that when the average pore diameter of a membrane has been reduced to 15 or 20 / x/ x\#151 ; which is the value for a well formed copper ferrocyanide film\#151 ; the membrane will have become almost truly semipermeable .
Hence pore size is one of the most important factors determining the osmotic properties of a precipitation membrane .
Although such a result would follow if the osmotic properties of a membrane were due to a selective mechanical blocking of the solute molecules , a moment 's consideration of the following approximate calculation will show that this cannot be the case , since the calculation indicates that the smallest pores in membranes have a diameter many times that of the average crystalloidal molecule .
Perrin and others have proved conclusively that there are very nearly 7 x 1023 molecules in the gramme-molecule\#151 ; a result which would give the diameter of the unassociated water molecule , for instance , as about 0*05 / x/ x. Since the smallest pore diameter in a copper ferrocyanide membrane is about 8 / x/ x , it follows that from 100 to 200 water molecules could be placed in a chain from one side of such a pore to the other , and that several thousands could be travelling across the cross section of the pore at the same time .
A selective mechanical blocking of even large hydrated crystallised molecules is hence out of the question , and this hypothesis for accounting for osmotic effects\#151 ; already largely rejected\#151 ; is no longer tenable .
Nevertheless , such a hypothesis might account for the blocking action exerted by the membrane on colloids , and on the suspension colloids more especially .
4 .
Range of Surf ace Forces inside Membrane Capillaries and Adsorption Effects .
Consider again the typical membrane interstice or pore formed between the three particles A , B , and C ( Diagram 4 ) .
Part at least of this interstice is always under the influence of the surface forces , whatever size the three particles may be ' ; and if the particles are sufficiently small , the whole of the interspace is within the surface force range .
Suppose first that the particles are large , and that the distance OP from the centre of the interspace to the surface of the particles is greater than the radius of molecular attraction PX .
There is then a central canal of Semipermeable Membranes .
( marked by the dotted line DEI ' in Diagram 4 ) , passing down the middle of the interspace , over which the surface forces have no control .
If the size Diagram 4 .
of the interstice or pore is reduced , the size of this central canal is diminished also ; and it is diminished very rapidly as compared with the rate of diminution of the pore itself , since the range of molecular attraction PX remains unchanged whatever size the pore may be .
When the distance OP from the middle of the interspace to the surfaces has become equal to the range PX of the surface forces , this central canal disappears , and the whole pore is under the control of the surface forces .
Many investigators have made determinations of the range of surface forces .
The most important for our present purpose are those based on surface condensation effects .
Such determinations are those of Parks , * who measured the thickness of the water film condensed on glass wool , and found it to be 13*4 x 10"6 cm .
or 130/ qa ; and of W. C. McLewis , f who determined the range of the adsorption-concentration effects at an oil-water interface and found it to be about 140 ( ip .
Other investigators , such as Quincke , Eeinold and Pucker , Plateau , Johannot , Vincent , etc. , working by different methods , have given values ranging from 100/ qa to 10/ qa .
The most likely interpretation of these different values is that the surface forces are very * 'Phil .
Mag. ' ( 6 ) , vol. 5 , pp. 517-523 ( 1903 ) .
t 'Phil .
Mag. ' ( 6 ) , vol. 17 , p. 428 ( 1909 ) . .
370 Mr. F. Tinker .
The Microscopic Structure strong up to a range of 10/ xya , and they become increasingly weaker from this point up to a range of 140/ zya , whilst beyond the latter point they are negligible . .
* We can now apply these results to determine the extent to which the pores of precipitation membranes are under the control of the surface forces .
In order to find out first which membranes have a central canal which is definitely out of range , we must take the upper limit of the above determinations ( 140fjbfju ) .
This figure would give a pore diameter of 280/ jl/ jl as the limit above which the central canal has a definite existence .
None of the precipitation membranes studied have pores so large as this , but it is noteworthy that the pores in gelatin , parchment , and porcelain capable of showing osmotic effects have values just above it ( 0*2 to lya).* On the other hand , if we wish to ascertain in which membranes the central canal has definitely vanished on account of the smallness of the pores , we must take the lower limit of the range of molecular attraction .
Taking this as 10/ Jbjjby the corresponding pore diameter for a completely and strongly controlled capillary is 20/ xya .
The bulk of the pores in copper ferrocyanide and Prussian blue membranes are at or below this limit , although some are possibly larger .
But the silicate and tannate membranes have pores with diameters just larger than this value , and the central spaces in their pores , if controlled by the surface forces at all , are controlled only very weakly .
The relation between the extent to which the capillaries of a membrane are controlled by the surface forces and the osmotic properties of the membrane is , however , an apparent and noteworthy one .
Copper ferrocyanide and Prussian blue\#151 ; the most perfect of membranes\#151 ; have their pores completely under control ; the tannate and silicate membranes , which are not so efficient , have a central space within their capillaries throughout which the surface forces are comparatively weak ; whilst membranes such as gelatin and parchment , which are permeable to all crystalloids , have pores which possess central canals outside the surface force range .
How , then , is this close relationship between the degree to which the capillaries are " controlled " and the osmotic properties accounted for ?
It is well known that any liquid in immediate contact with a surface is bound or adsorbed to the surface , the thickness of the adsorbed film being coincident with the surface force range .
All the liquid which gets into the pores of a precipitation membrane such as copper ferrocyanide is therefore adsorbed liquid .
Moreover , since adsorption is selective in nature , either the solvent or the solute is preferentially attracted to the surface and into the pores of the membrane .
Since it is evident also that those substances which can get * Of .
the work of Bartell already mentioned .
of Semipermeable Membranes .
into the pores of the membrane can also get through , it follows that a membrane such as copper ferrocyanide is relatively impermeable to solutes which it adsorbs negatively , but permeable to solutes which it adsorbs positively .
With the membrane bathed by cane sugar , for instance , the water would seem to be adsorbed into the capillaries rather than the sugar , and the water , being thus taken up into the pores , is able to get through , whilst the sugar , being unable to approach the surfaces of the particles , is not .
We must , however , qualify this generalisation according to the size of the pores .
If the membrane is one which has very fine capillaries , in which the central canal out of the surface force range has long vanished , the adsorption effects within the pores will be very strong .
In fact , they will be the same as the adsorption effects in the layers of surface film in immediate contact with a surface , which are not of necessity the same as those in the surface film as a whole or in its outer layers .
It is perhaps necessary to emphasise this point , since the modern measurements of adsorption effects with charcoal and other solids and liquids have referred to the complete surface film , and not to that portion in immediate contact with the surface .
If , on the other hand , we consider those membranes whose pores are large enough to possess a central canal out of range , it will be seen that the adsorption effects are not of such great importance , since free , unadsorbed solution of unchanged concentration can travel unhindered through the central canal .
But , though such membranes are permeable and leaky , they can show slight osmotic effects , since part of the liquid in their capillaries is adsorbed liquid .
Since the proportion of adsorbed liquid is increased by reducing the capillary diameter , osmotic effect is also increased in the same way ; but a perfect selective action is not attained until the pore size has diminished to such an extent that the central canal of unadsorbed liquid has vanished completely .
But I advance the above view of the cause of osmotic effects only tentatively .
To prove it more rigorously it will be necessary to show ( 1 ) that moisture is taken up by precipitation membranes from solutions bathing them , and ( 2 ) that such moisture is adsorbed on to the surfaces of the colloidal particles composing the membrane rather than absorbed into their substance by either hydration or solution proper.* Summary .
1 .
The common precipitation semipermeable membranes are composed of small precipitate particles , ranging from OT ^ to 1*0 these particles * Work now in progress tends to confirm these conclusions .
372 The Microscopic Structure of Semipermeable Membranes .
being packed closely together .
Each of these precipitate particles is , however , not simple in structure , but is itself an aggregate formed by the flocculation of submicroscopic colloidal particles .
The particles composing the membrane are smallest in the case of copper ferrocyanide and Prussian blue .
2 .
Precipitation membranes show most of the properties of gels , as ordinarily prepared , both in their method of formation and in the changes they undergo in various solutions .
Like ordinary gels , they are possessed of great tensile strength , which varies in membranes of different kinds .
Their stability in the colloidal condition also varies greatly .
But , although they show the physical properties of gels , they have not the same mechanical structure , the membrane being much more closely knit together than the gel proper .
3 .
The pores in a copper ferrocyanide membrane range from 8 to 60 yaya in diameter , the average diameter being from 15 to 20 ya/ a. The pore size is too great for the membrane to act osmotically by exerting a selective mechanical blocking action .
4 .
The order of a series of membranes in pore size is the same as that of their efficiency as semipermeable membranes .
Copper ferrocyanide and Prussian blue are the most efficient membranes , and they have also the smallest pores .
5 .
There is a very close connection between the osmotic properties of a membrane and the extent to which the membrane capillaries are under the control of surface forces .
Osmotic effects are probably the result of selective adsorption phenomena occurring at the surface of the membrane and in the capillaries , the membrane being relatively impermeable to solutes which are negatively adsorbed , but permeable to solutes which are positively adsorbed .
In conclusion , I desire to express my thanks to Sir Oliver Lodge , whose generous provision of instruments and materials has rendered the above work possible ; to Prof. A. J. Brown and Mr. T. H. Pope , B.Sc. , for much kindly advice and help ; and to Mr. Edward Robinson , Jr. , of Sir Oliver Lodge 's laboratory , for valuable assistance in taking the photographs .
i mer. Roy .
Soc. Proc. A , vol. 92 , PI .
6 .
Fig. 1 .
COPPER FERROCYANIDE x 550 , after 30 minutes .
Fig 3 .
PRUSSIAN BLUE x 550 , after 15 minutes .
Fig. 2 .
COPPER FERROCYANIDE x 550 , same film as Fig. 1 after 8 hours .
Fig. 4 .
COPPER SILICATE x 550 after 15 minutes Fig. 6 .
BARIUM SULPHATE x 550 .

