516
FRED HAZEL
T H E EFFECT OF SMALL CONCENTRATIONS OF HEXAMETAPHOSPHATE ON IRON OXIDE SURFACES FRED HAZEL Department of Chemistry and Chemical Engineering, University of Pennsyluania, Philadelphia, Pennsylvania Received January 16, 1048
The use of phosphates in water technology has been given impetus with the introduction of molecularly dehydrated compounds such aa sodium hexametaphosphate. Interest in the latter compound was stimulated by the observation of Rosenstein (17) that low concentrations inhibit the precipitation of calcium carbonate by ammonia from waters high in calcium bicarbonate. Subsequently it has been shown (6) that this inhibiting effect of hexametaphosphate manifests itself when the carbonate-ion concentration of the water is increased not only by the addition of ammonia but also by moderate treatment with caustic soda, lime, or sodium carbonate, or by heating. Since the concentration required to prevent the deposition of calcium carbonate is of the order of 1to 5 p.p.m., the term “threshold treatment” has been applied to the process. Buehrer and Reitemeier (2) have made a detailed study of the mechanism of calcium carbonate formation in the presence of hexametaphosphate and have reached the conclusion that the precipitation is retarded because of a “deranged crystallization” promoted by adsorption of hexametaphosphate on the crystal faces. Industrial application of threshold treatment has demonstrated that when it is used in a system which is incrusted with calcium carbonate scale, not only is further formation inhibited but also the old scale is slowly removed (6). Since a small amount of scale has a beneficial effect in retarding corrosion (l),removal of the protective coating might be expected to accelerate the formation of rust at the bare iron surface. However, under the conditions of the threshold treatment this corrosion problem is not encountered, and the experiments which have been conducted on the subject (7, 18) lead to the conclusion that low concentrations of hexametaphosphate strongly retard the rusting of iron. It has been proposed (7) that this is due to “the adsorption of hexametaphosphate, or a complex thereof, on the metal or metal oxide surface.” The present investigation was undertaken with a view of obtaining information, if possible, from the colloidal standpoint on the corrosion problem. The fact that colloidal hydrous oxides of iron are produced in the rusting of the metal served as a basis for the undertaking. Stability, flocculation and deflocculation, and mobility determinations were made with sodium hexametaphosphate on iron oxide sols at different concentrations arid a t different hydrogen-ion activities. The effect of calcium-ion concentration on the behaviors also was investigated. For purposes of comparison, parallel experiments were conducted with sodium pyrophosphate and sodium orthophosphate. In addition, this report includes some measurements on aluminum oxide suspensions which were made by Marlow F. Shute in tbe writer’s laboratory. Hexametaphosphate has an influence on numerous colloidal phenomena
EFFECT OF HEXAMETAPHOSPHATE ON IRON OXIDE SURFACES
517
(cf. reference 6 ) . It is especially effective as a stabilizer for colloidal suspensions. In this connection the property of the compound to peptize or disperse finely divided solids has been emphasized. However, it is well known that alkali phosphates, when used a t sufficiently low concentrations, are strong coagulators for positive colloidal systems, as, for example, hydrous iron oxide. Comparing different phosphates such as ortho and pyro, it is found that the coagulating power increases with the valence of the negative ion. The mechanism of the flocculation process involves adsorption of ions, which results in a decrease in mobility of the colloidal particles. With positive hydrous oxide sols, flocculated with strongly adsorbable negative ions, coagulation starts before the isoelectric point is reached and continues after the particles are recharged. At higher ionic concentrations the particles become strongly negative and stable. Thus, with regard to particle mobility, there are two zones on each side of the isoelectric point: viz., on the positive side a stable and instable zone and on the negative side an instable and stable zone. Addition of increasing concentrations of phosphate ions causes the mobility of the particla to change from the region of positive stability through the regions of positive instability and negative instability to the region of negative stability. EXPERIMENTAL
The ferric oxide sols were prepared by dropwise addition of a ferric chloride solution to boiling distilled water. Purification was effected by dialysis for 1 week a t room temperature with Visking casing. Sols of different concentration were prepared by dilution of the original with distilled water. Solutions of tetrasodium pyrophosphate and trisodium orthophcsphate were prepared from the reagent grade salts. Solutions of sodium hexametaphosphate were prepared from the commercial product, Calgon. The latter were used as soon as possible, since it is known that hexametaphosphate solutions possess a certain amount of instability with respect to reversion to the ortho form. Coagulation and mobility measurements were made by methods described previously (13,20). RESULTS
Table 1 gives stability data for the three phosphates with iron oxide sols of different concentrations and m e r e n t hydrogen-ion activities. The data show that coagulation occurred only over a limited range of electrolyte concentrations, Thus, with hexametaphosphate and system I, 0.005 millimole per liter or 3.1 p.p.m. of the salt were needed to coagulate the system, while 0.007 millimole per liter or 4.3 p.p.m. stabilized it.' 1 The flocculating value of sodium hesametaphosphate noted above, 0.005 millimole per liter, is for a positive sol. At a concentration of 0.007 millimole per liter the sol was recharged to a stable negative system. These effects were the result of adsorption of the polyvalent negative hexametaphosphate ion. At a much higher concentration of the electrolyte, e.g., 80 millimoles per liter (for the sol containing 1.0 g. of FelOJ per liter), the negative sol was coagulated, thus completing the irregular series. Coagulation of the negative so1 was induced, presumably, by adsorption of sodium ions and/or compression of the double layer surrounding the particles by the high electrolyte concentration.
518
FRED HAZEL
TABLE 1 Stability data for the three phosphates with iron ozide sols of different wneentrations and diffvent hydrogen-ion activities
System I: 0.2g. FerOa per liter; p H 4.6 Complete ...................... None ..........................
1 i I I 1 0.005
3.1 4.3
0.007
0.010 0.016
4.5 7.1
0.014 0.028
~
5.3 10.6
System 11: 0.2 g. FelOs per liter; pH 6.0 Complete.. .................... Xone . . . . . . . . . . . . . . . . . . . . . . . . . .
1 i:: ::; I i:;: 1 1 1 G:: 1 0.003
I0.007
0.008 0.014
~
~
~
0.012 0.024
~
4.6 9.1
System 111: 1.0 g. FepOs per liter; pH 4.4 Complete. ..................... None . . . . . . . . . . . . . . . . . . . . . . . . . .
0.036 0.076
j
13.7 28.9
26.8 0.072 62.4 10.17
1
27.4 64.6
0.032 0.056
1;:;
System IV: 2.75 g. FelO, per liter; pH 4.5 Complete.. .................... None . . . . . . . . . . . . . . . . . . . . . . . . . .
24.5 44.1
1
0.060
10.14
' ~
t +L
2 kl
f
0
+ J
-2
0
002
a04
MILLIMOLES ELECTROLYTE
am PER LITER
FIG.1
The flocculation values of all of the electrolytes were increased by an increase in concentration of the iron oxide particles and by an increase in hydrogen-ion activity of the systems. Insofar as polyvalent ions are concerned, these behaviors are entirely normal.
519
EFFECT OF HEXAMETAPHOSPHATE ON IRON OXIDE SURFACES
It may be seen from the data in table 1that hexametaphosphate had the greatest effect, while orthophosphate had the least effect on the stability of the systems. Figure 1 shows the mobility behavior of iron oxide particles with phosphates. The data are for a sol containing 0.2 g. of ferric oxide per liter and having a pH of 5.0. The order of effectiveness of the electrolytes in discharging TABLE 2 Effect of calcium chloride on the jlocculation and recharging values of phosphates for a ferric oxide sol
System I : 0.2 g. Fez08 per liter; pH 4.6 Complete.. . . . . . . . . . . . . . . . . Yone ...................... ~
0.005 0.007
1
t::
I :::;:1 "7; I
0.014 0.028
5.3 10.6
System IA: 0.2 g . Fez03 per liter; pH 4.6; 22.2 p.p.m. CaCL* Complete., . . . . . . . . . . . . . . . . None . . . . . . . . . . . . . . . . . . . . . .
~
0.006 0.025
~
3.7 0.008 3.6 15.31 0 . 2 189.2
0.011 14.0
I
4.2 15.2
System IB: 0.2 g. FenOa per liter; pH 4.6; 44.4 p.p.m. CaC1,* Complete . . . . . . . . . . . . . . ./ Kone.. . . . . . . . . . . . . . . . .
0.006 0.1
~
6;:;
1 ::? 1
17::;
I
o'olo
System IC: 0 . 2 g. Fez08 per liter; pH 4.6; 66.6 p.p.m. CaCL*
1
Complete . . . . . . . . . . . . . . . . . . None. .....................
I
0.006
~
o.25
3 . 1 ~0.007 153
~
1
3.1
IO.010
System ID: 0.2 g. Fez03 per liter; pH 4.6; 111.0 p.p.m. CaCl2* Complete.. . . . . . . . . . . . . . . . . S o n e . .....................
3.1
i
0.006
2.7
j
0.008
1
I 1
3'8
3.8
3.0
System IE: 0.2 g. FezOaper liter; pH 4.6; 222.0 p.p.m. CaCL'
I
Complete .................. 0.004 None.. .................... ca. 2 . 0
~
2.4 ea. 1224
1
0.005
1
2.2
1
0.007
2.7
*Added with the electrolyte.
the particles and recharging them with a negative sign is the same &s was observed in the coagulation experiments. The effect of calcium chloride on the flocculation and recharging values of phosphates for a ferric oxide sol is shown in table 2. It may be noted that calcium chloride decreased the amounts of orthophosphate and pyrophosphate required to flocculate the systems. On the other hand, a greatly increased amount
520
FRED HAZEL
of each of the phosphatea waa required to impart a sufficient negative charge to the particles for the systems to be stable under these conditions.* Mobility curves with hexametaphosphate are plotted in figure 2 for an iron oxide sol containing calcium chloride. The sol had a concentration of 0.2 g. of ferric oxide per liter and a pH of 5.0. The upward displacement of the curves
tl
+4
$ 0
r:
d
0
5
5
3
C"
-L
L
-4 MILLIMOLES bk6P'C$a FCR LITER
RPM CCI,
FIQ.3
FIG.2
TABLE 3. Flocculation behavior of aluminum ozide sols with phosphatea
.I
Complete.,. . . . . . . . . . . . . . . . . . Eone. . . . . . . . . . . . . . . . . . . . . .
0,010 ~
0.032
___.
! 1 I 6.1 20.8
0.018
0.014
0.038
178:;
~
0.058
1
8.0 25.9
2.1 g. AlzOs per liter; pH 6.0
.........I
Complete.. .................... Iione . . . . . . .
0.014 0.072
i 448:;
~
:!E
I 4's:: 1 ::;: ~
13.4 80.3
* D a t a by Marlow F. Shute.
toward the region of negative instability with increesing calcium chloride concentration correlates with the fact that under these conditions higher concentrations of hexametaphosphate were required to give the particles negative stability. 1 It was found impossible to investigate the recharging concentrations of pyrophosphate and orthophosphate a t the higher concentrations of calcium chloride, because of the insolubility of the calcium salts.
521
EFFECT OF HEXAMETAPHOSPHATE ON IRON OXIDE SURFACES
The effect of calcium chloride, alone, on the mobility of the same sol is shown in figure 3. Coagulation occurred at a concentration of 666 p.p.m. of calcium chloride and at a particle mobility of 3.0 p per second per volt per ent ti meter.^ The flocculation behavior of aluminum oxide sols with phosphates is shown in table 3 at two difTerent concentrations of the hydrous oxide. The sols were prepared by peptization of freshly precipitated hydrous aluminum oxide with dilute hydrochloric acid. They were purified by dialysis. The dilute system was prepared by dilution of the concentrated sol with distilled water. Solutions of sodium tetraphosphate were prepared from the commercial product, Quadrafos. Some indication of the stability of the hexametaphosphate solutions employed in this study is had from the fact that identical values were obtained with freshly prepared solutions and with solutions which had stood for 2 weeks. DISCUSSION
Increasing the pH of water by the addition of alkalies is a known method for controlling the corrosion of iron. This is of interest, since the property of alkalies of converting positive iron oxide surfaces to an electronegative condition is possessed also by alkali phosphates to a striking degree. With the alkali sodium hydroxide, the change in sign occurs at a pH of about 8.6 for a pure colloidal suspension of iron oxide (11). The isoelectric point is altered, however, by the presence of other ions in the system (12, 15). The effect of cations such as calcium ion is to shift the isoelectric point to higher pH values, while negative ions such as hexametaphosphate shift the isoelectric point to lower pH values. During the present study it was observed that in the presence of 2.7 p.p.m. of hexametaphosphate the iron oxide was isoelectric at pH 5.6.4 Pyrophosphate and orthophosphate were found to be less effective in this respect: 4.5 p.p.m. of pyrophosphate a t a pH of 6.0 and 7.0 p.p.m. of orthophosphate a t a pH of 6.7 were required to make the colloidal particles isoelectric. The amount of hexametaphosphate required to recharge iron oxide with a negative sign depends on three factors: first, the pH of the system; second, the presence of other ions and in particular multivalent cations such as calcium; third, the concentration of iron oxide. In regard to the first of these factors it was found that less hexametaphosphate was required at high pH values than at low ones; thus, while 2.7 p.p.m. were +requiredat pH 5.6, only 1.2 p.p.m. were needed at pH 6.8. Calcium ion has the revewe effect, increasing the amount of hexametaphosphate required to bring the particles to the isoelectric condition.
+
Critical mobilities of the phosphates employed in this study ranged between 1.5 and 1.1 per second per volt per centimeter for the same system. Monovalent ions usually have higher critical mobilities than polyvalent ions. This phenomenon has been discussed previously (8),as has been the effect of monovalent ions on the critical potentials of polyvalent ions (10). In the present investigation, the critical mobilities of the phosphates in 3.0 p per second per volt per centimeter. the presence of calcium chloride ranged up to I t has been stated (19) that charge reversal from positive to negative can be effected only with hydroxyl ions, i.e., in alkaline solution. From the above results and from those of previous atudies with potassium ferrocyanide (12), i t appears that this phenomenon is a general one associated with strongly adaorbable negative ions. a
+ 2.0 '
+
522
FRED HAZEL
This is illustrated by the following data obtained a t a pH of 5.9 with a sol containing 0.2 g. of iron oxide per liter. The system was isoelectric when the concentrations were as shown. WIM ctaopmr
HZXAYETAPXOSPEAIE
.p.p.m.
9.9.r. 2.4 8.4 12.6
0 22.2 66.6
While no measurements were conducted at concentrations of iron oxide lower than 0.2 g. per liter it is to be expected from the behavior of polyvalent ions in colloidal systems (3, 5, 9) that the amount of hexametaphosphate needed to recharge the surface of iron oxide at a given pH and calcium-ion concentration would decrease proportionally with the dilution of the metal oxide suspension. The property of alkalies and of hexametaphosphate to inhibit corrosion may be attributed to the property of these substances of converting the surface of iron oxide from an electropositive to an electronegative condition and thus rendering the surface of the metal more noble. It is possible that the mechanism of this process may involve the transformation of afilm of iron oxide, however thin, on the metal surface from a positive to a negative state. In practice it is found that when alkalies are added as corrosion inhibitors to bicarbonate waters there follows a precipitation of calcium carbonate due to the increase in pH. A scaling of pipes occurs (16),unless the process is carefully controlled (4,14),to a degree that exceeds the beneficial effects obtained from a protective coating of calcium carbonate. Hexametaphosphate not only inhibits the precipitation of calcium carbonate from alkaline solution but also retards the corrosion of iron at pH values below the neutral point, a fact which may be plausibly associated with its ability to recharge iron oxide under these conditions. THEORETICAL CONSIDERATIONS
Flocculation data with orthophosphate-calcium chloride and pyrophosphatecalcium chloride mixtures recorded in table 2 show that the amounts of phosphates needed to flocculate the systems decreased with increasing concentration of calcium chloride. I n the case of hexametaphosphate, however, there wasat first an increase and then a decrease in the amount required ~ t the s concentration of calcium chloride was increased. By considering calcium chloride as a coagulating electrolyte it is possible to tabulate the flocculation data on a percentage basis. This is done in table 4, and it may be seen that in no c a b are the flocculation values for the mixtures additive, i.e., the sums of the percentage flocculation values of calcium chloride and the percentage flocculation values of the phosphates do not equal 100 per cent. Thus, in the presence of 3.3 per cent (22.2 p.p.m.) of the flocculating concentration of calcium chloride, 78.6 per cent (0.011 millimole per liter) of the
523
EFFECT OF HEXAMETAPHOSPHATE ON IRON OXIDE SURFhCES
flocculating concentration of orthophosphate, 80 per cent (0.008 millimole per liter) of the flocculating concentration of pyrophosphate, and 120 per cent (0.006millimole per liter) of the flocculating concentration of hexametaphosphate were required, respectively, for coagulation. Similar effects with each of the electrolytes were observed at all concentrations of calcium chloride. The results with orthophosphate-calcium chloride and pyrophosphate-calcium chloride mixtures can be explained on the premise that calcium chloride compressed the double layer surrounding the particles. Attending the decrease in thickness of the double layer the critical potential was raised (10)and less phosphate ion was required for coagulation. The same factor operated in the case of hexametaphosphate-calcium chloride mixtures, as wm evidenced by the high critical mobilities a t which coagulation occurred. This factor, which tended to decrease the amount of hexametaphosphate needed for coagulation, was opposed by complex-ion formation between calcium and hexametaphosphate ions, resulting in a decrease in valence of the coagulating ion. In view of this it is probable that even greater concentrations of hexametaphosphate would have been required for TABLE 4 Percentage ficculation values
~0 3.3 6.7 10.0 16.7 33.3
N d 0 1
%%&
NslPlDi
?2&
100 78.6 75.7 75.7 57.1 50.0
100 81.9 82.4 85.7 73.8 83.3
100 80.0 80.0 70.0 60.0 50.0
100
CaClr
+
NaePlDta
NaQaoII
~
1
i
83.3 86.7 80.0 76.7 83.3
I j
100.0 120.0 120.0 120.0 100.0 80.0
1
100 123.3 126.7 130.0 116.7 113.3
coagulation in the presence of calcium chloride had it not been for the electrolyte concentration effect in decreasing the thicknesa of the double layer. With reference to table 2, it may be observed that the coagulation zone of hexametaphosphate was extended from 1.2p.p.m. with no calcium chloride present to over 1200 p.p.m. in the presence of 222 p.p.m. of calcium chloride. While the insolubility of the calcium salts prevented a complete investigation of the recharging effects of orthophosphate and pyrophosphate in the presence of calcium chloride, enough data are available to make clear that these phosphates were made much less active by calcium chloride than was hexametaphosphate. Since the former salts show a slight, if any, tendency to form complex ions with calcium, the explanation for their decreased effectiveness under these conditions may be attributed to the adsorption of calcium ions, which decreased the negative charge on the particles and thus opposed the recharging effect of the phosphate ions. REFERENCES (1) BAYLIS, J. R . : J. Am. Water Works Assoc. B, 408 (1922); Ind.Eng. Chem. 19,777 (1927). T. F., AND REITEMEIER, R. F . : J. Phys. Chem. 44,552 (1940). (2) BUEHRER,
524
COMMUNICATION TO THE EDITOR
(3) BURTON, E. F.,AND BISHOP,E.: J. Phys. Chem. U,701 (1920). (4) DEMARTINI,F. E.:J. Am. Water Works Aasoc. SO, 85 (1938). E., AND SORUM, C. H.: J. Phys. Chem. 44,62 (1940). (5) FISHER, (6) HATCH,G.B., AND RICE,0.: Ind. Eng. Chem. 81.51 (1939). (7) HATCH,G.B., AND RICE,0.: Ind. Eng. Chem. 91, 1572 (1940). (8) HAZEL,F.:J. Phya. Chem. 46, 731 (1941). (9) HAZEL,F.:J. Phys. Chem. 46, 738 (1941). (10) HAZEL,F.: J. Phya. Chem. 45, 747 (1941). (11) HAZEL,F.,AND AYRES,G. H.: J. Phys. Chem. I, 2930 (1931). 3148 (1931). (12) HAZEL,F., AND AYRES,G. H.: J. Phye. Chem. I, C. H . : J. Am. Chem. SOC.6% 49 (1931). (13) HAZEL,F., AND SORUM, (14) LANOELIER, W. F.:J. Am. Water Works Aeaoc. !U, 1500 (1936). (15) LARSON,T.E., AND BUSWELL,A. M.: Ind. Eng. Chem. 31, 132 (1940). E. P.: Ind. Eng. Chem. 91, 58 (1939). (16)RICE,O.,AND PARTRIDQE, (17)ROSENSTEIN, L.: U.s. patent 20,754(1938). EDW.C.: J. Am. Water Worka Assoc. 91. 1495 (1940). (18)TRAX, (19)WEISER,H.B.:Colloid Chemistry, p. 228. John Wiley and Sons, Inc., New York (1939). (20) WILLEY,A. R., AND HAZEL,F.: J. Phya. Chem. 41,699 (1937).
COMMUNICATION TO THE EDITOR LIQUID-VAF’OR COMPOSITION OF THE BOILING TERNARY SOLUTION ETHYL ALCOHOLGLYCEROLBENZENE Sir :
.
I am much indebted to Dr. Hugh J. McDonald for discussing with me a number of points in his paper, “Liquid-vapor composition of the boiling ternary solution ethyl alcohol-glycerol-benzene,’’ published in the Journal of Physical Chemistry 46,706 (1941). However, the following comments should be noted: (I) In figure 1, the residue curves are plotted with moles of each component upwards and total moles of residue out to the right; a point representing the change of residue composition during distillation (for any one component) then move8 from the upper right corner towards the origin; the motion is opposite on the distillate curves. (2) On page 711,the statement is made that the “slope of the curve a t any point is, of course, the mole fraction of each constituent in the vapor a t that point.” This is confusing. For the distillate curves, with moles of component plotted against total moles, the ratio of ordinate to abscissa is, of course, the mole fraction in the total distillate. The value of the slope has an additional term, as may be shown by simple calculus. For the residue curves, the ratio of ordinate to abscissa gives the mole fraction in the residue (as is correctly stated in the paper) ; the slope of a residue curve gives the mole fraction of that component in the “instantaneous” vapor just leaving the residue, not the mole fraction in the total distillate. (3) The curve for benzene (or either other component) in figure 1 must be