A New Method for Measuring Corrosion of Metals W. R.
I
VAN
WIJIC,Bataafsche Petroleum Maatschappij, Amsterdam, Holland
N STUDYING the action of corrosive agents it is usual
to apply some gravimetric method of measuring the material lost from a test piece, either by subjecting the test piece to the action of the corrosive agent for a certain length of time and weighing it before and after the test, or by determining analytically the amount of metal dissolved in the corrosive agent. To obtain anything like reliable results from either of these methods a fairly intensive corrosion has to be effected. As t,his in many instances involves too much time for practical purposes, accelerated c o r r o s i o n tests are often applied-for instance, by working a t a higher temperature or with a higher concentration of the corrosive agent than is usual in practice. It is well known, however, that such accelerated tests hardly ever give strictly comparable results, and in many cases the conditions and results of the test deviate so far from what takes place normally ‘T tr. that the conclusions drawn therefrom are entirely wrong. It is therefore of great imp o r t a n c e t o d e v i s e a method whereby even the slightest COITOsion can be measured quantitatively with great accuracy under normal conditions and within a reasonable time. Such a method has now been developed by subliming extremely thin layers of cm. thick) FIGURE1. CONTAINER met‘al (about 8 X FOR GLASSPLAQUES on glass plaques in high vacuum and measuring their thickness before and after subjecting them to corrosive action, the measurements being made by determining the coefficient of the light-transmissibility of the metal-coated plaques. Owing to the high absorption coefficient of metals, with an optical apparatus of great precision it is possible to measure a change in thickness of even one atomic diameter. In the author’s tests such a high degree of accuracy was not required, so that a simpler apparatus sufficed, with which he was able to measure differences of 5 atomic diameters. This method has been found to be very useful for measuring the degree of aggressiveness of a liquid or gas which is only slightly corrosive, as may be necessary in practice-for instance, in cases where a large quantity of liquid flows along metal parts where even the slightest wear may have undesirable consequences, even leading to a serious interruption of the work As an example may be mentioned the fuel pump of a solid-injection Diesel engine.
PREPARATION OF TESTPLATES The maximum thickness of a metal layer which can transmit light to a measurable degree is of the order of cm. or less. Such a layer can be made by subliming the metal in high vacuum on a base of transparent material (glass), and
if the necessary precautions are taken during the sublimation the metal layers adhere so well to the base that they can safely be heated in a liquid for a long time without risk of becoming detached. The glass plates (dimensions4 X 1 X 0.2 cm.) were first cleaned with nitric acid, washed with soapsuds, and rinsed first with tap water and then with distilled water; they can also be cleaned with chromic acid and rinsed with distilled water. Fourteen of these cleaned plates were laced u right in a glass container (Fjgure 1) and four more laizon the tottom, the container then being ready to be placed in an evaporation bell (Figure 2). The bell was 17 cm. high and 9 om. in diameter, closed a t the bottom with a ground-glass plaque 2 om. thick by means of a little vacuum grease. At the top a ground stopper was fitted in, through which a fine tungsten coil mounted on two nickel conducting wires could be passed down into the bell. The bell was connected by a tube at the side to a mercury diffusion pump, with a liquid nitrogen trap to prevent the mercury vapor from getting into the bell. The diffusion pump was connected up to a mercury-vapor jet pum for which the required low vacuum was drawn by a rotary oiy’pump. A weighed piece of cleaned metal was then put into the tungsten coil, the container with the glass plates laced in the bell, and the apparatus exhausted, (The tungsten &ament was lowered to a level just opposite the centers of the glass plates, at an average distance of about 3 cm. from the centers.) A few minutes after the pumps had been started, liquid nitrogen was supplied to the cooling vessel, and after waiting a few more minutes the current for the tungsten filament was switched on (from experience it was known that with the assembly used a high vacuum, 10-4 mm. of mercury, was obtained within a few minutes). The heating current was derived from a 36-volt storage battery, adjustment being made by means of a resistance so as to obtain a favorable rate of evaporation. The pro er working temperature, in general, is in the neighborhood of tEe melting point of the metal used. The coil should be heated slowly, allowing plenty of time for the gases liberated t o be pumped off; if heating takes place too quickly there is danger of particles of the metal flying off from the tungsten filament, or the gases may not be pumped off quickly enough, thus spoiling the vacuum, As soon as the piece of metal in the filament had entirely disa peared, the current was switched off. lfter cooling, air was admitted, the bell opened, and the container with the metal-coated glass plates removed. The plates were then marked and those not required for immediate use were k e p t immersed in kerosene previously treated with sodium. (It had been found that copper and zinc k e p t under kerosene not so treated showed the beginning of corrosion after a few weeks.) For making plaques wihh a layer of 2* iron of about 4 to 8 X 10-6 cm. thickRATION BELL FOR ness, a piece of f l o r i s t s ’ iron wire of TEMPERAabout 10-2 gram was used. Preferably, TURE Dure metals are used to insure good remohucibility, and in the author’; tests for the comparison of various oils it was also desirable to use pure metals rather than alloys, which naturally have t o differ in comgsition according to the purpose for which they are employed. evertheless, for the author’s own information he also examined some alloys and did not encounter any great difficulties.
IRONAND COPPER.The sublimation of iron and copper presented no difficulties, smooth and glossy metallic layers. being obtained on the glass plates.
January 15, 1935
ANALYTICAL EDITION
ZINC. In the case of this metal difficulties were encountered. At room temperature the zinc did not precipitate uniformly on the glass plates, because a t room temperature Zinc is so volatile that the atoms have very little chance of adhering to the glass. An attempt to remedy this by increasing the density of incidence of the zinc atoms (by more rapid heating) did not yield the desired result, but it was found that by cooling the plates smooth and glossy layers of
zinc be Obtained' This the use Of different type Of evaporation and holder for the glass plates (Figure 3). For these tests with zinc a nickel holder was used, being suspended from the ground joint in the neck of the bell, so that the Part of the glass wall of the bell that has to be cooled is nowhere in contact with metal or glass, which would involve the breaking; danger Of the done with liquid nitrogen. With zinc an interesting phenomenon was noticed which shows the mobility of the atoms on the glass surface: The zinc F I Q U R E3 . EVAPORATIONPrecipitated in floccular form on one Part BELLFOR Low of the plate, the rest of the plate remainTEMPERATURE ing free, this being a t t r i b u t a b l e t o a n C, container for accidental local conglomerat~on of conglass plaques
49
the infra-red ra s and thus preventing undesirable heating of the plaque), then tgrough a plate of Photographic red glass (the region of visible spectrum transmitted thereby being narrowed down from 5900 A. to infra-red, average 6200 A., so that the absorption coefficient and the refractive index of the metal could be considered as constants), and finally through a milk-glass plate (making the illumination more uniform). The plate was fastened in a holder with a spring and placed in front of the top half of a rectangular opening in diaphragm D. The image of this rectangular opening and the plaque was obtained through two lenses of the same focal distance (the two halves of a s ectacle-glass, 8 D, placed at a short distance one above the otEer) which could be brought sufficiently close together t o produce adjacent images on a focusing screen. Each lens was diaphragmed, leaving a rectangular opening 1.20 x 2 om., while the lower lens could be further diaDhraemed. leavinn a smaller, triangular opening adjustable acc'ordiig to a scde. From the position of a mark on the fixed diaphragm with respect to the scale, the free surface of the lens could be calculated by employing a conversion graph (a small lamp was mounted to illuminate the scale, and also a maenifvine - elass to make it easv to read the Position of the mark). Each of the images cast on the focusing screen consisted of a dark strip, corresponding to the part of the opening in the first diaphragm, D, covered by the plaque, which, being coated with metal, was only slightly translucent, and a light strip corresponding to the uncovered part of the first diaphragm opening. By adjusting the distance between the two lenses the images could be made to touch, so that four adjoining strips, dark-light-darklight, became visible on the screen. The two extreme strips were then diaphragmed off and the other two strips brought to equal strength by adjusting the triangular dia hragm opening until their line of demarcation disappeared. k i t h a uniform illumination the ratio of the lens surface remaining after diahragming to the total surface of the other lene is e ual to the rraction of incident light penetrating the metal-coatel plaquei. e., it is equal to the transmissibility of the pla ue. The position of the index mark on the fixed diaphragm b h n d the lenses with respect to the scale on the movable diaphragm was then noted and the corresponding transmissibility of the metal layer was read from the conversion graph mentioned above.
densation nuclei. After the Sublimation had ended, the zinc was found to be moving along the glass surface, ultimately covering the whole of the plate, also a t the zinc could be substanback. The spreading velocityof hially increased by heating the Plate by radiation from the bell (all being done in high vacuum) this being heated by means of hot air from a hair-drying machine. It is not, strictly speaking, the transmissibility of the metal INFLUENCE OF THE VACUUM. In preparing the metalplated test plaques care must be taken to use the right layer alone that is thus found, as the beam of light is also vacuum. To obtain smooth, glossy, and reflective metal layers it is essential that the metal atoms should not collide on their way; t h e y s h o u l d p a s s direct f r o m t h e m o l t e n .metal to the glass and stick ------- ----there. The mean free path for the atoms should therefore be of the order of the distance between the filam e n t coil and the glass plate, or g r e a t e r , w h i c h FIGURE 4. APPARATUS FOR TRANSMISSIBILITY MEASUREMENT corresponds to pressures of a few times 10-4 mm. of mercury or lower. If there is a considerable chance for slightly weakened by reflection a t the boundary between air collisions (pressures of the order of lo-* mm. of mercury), and glass on the noncoated side of the plaque. The same rethe metal particles will partially agglomerate on the way, flection also occurs, hovever, when the determination is reresulting in a layer that is not coherent and has a dull black peated with the same plaque after it has been subjected to appearance, as a consequence of diffuse reflection and accom- the action of the corrosive agent, so that this does not affect panying absorption of light. At higher gas pressures the the difference in transmissibility. Neither are the determetal and the tungsten coil oxidize in the air. minations affected by the fact that after corrosion the metal layer is less reflective than before, since the measurements are made in diffuse light. TRANSMISSIBILITY OF METALLAYERSTO LIGHT The question has been raised of the interference of protecThe transmissibility of the metal layers was measured op- tive films on light transmission where the film is opaque and ticab' by means of the apparatus Shown diaPammaticallY &mly attached to the metal and where Some of the passive in Figure 4, all the parts being mounted on a Zeiss rail to fa- films may be as thick as the test metal to be used. The cilitate centering. author has invariably met with corrosion products that could The source of light was an lam with milk-glass easily be removed and were practically transparent or a t most bulb fitted in a case with open front. The Eght first passed Pale Yellow in color. To x'educe as much as Possible the inthrough a tank fillsd with dilute copper sulfate solution (absorbing terference of films that might have been left behind, the ab-
50
INDUSTRIAL AND ENGINEERIKG CHEMISTRY
sorption of the layer of metal was determined in red light, where, as may be concluded from the pale yellow color of the corrosion products, their absorption was a minimum.
Vol. 7, No. I
ening by reflection against the metallic mirror. With thin layers this division is no longer possible, so that only the total transmissibility of the layer can be spoken of. The general relation between the intensity, I (d), of the light which has passed through an absorbing layer with a thickness of d (absorption coefficient k and refractive index n) and the original intensity, I (o), a t perpendicular incidence from air on metal and subsequently on glass (refractive index n’) is: I ( d ) / l ( o ) = 16n’(nS
+ n’W)/
d*4
e
’
[nW
+ (n + 1)*]
+ (n - n’)2] - 2[ (n?V + - nl(1 + n”) + - n*lc*(n’a+ 4n’ + 1)1cos + Cnk(n’ + 1) (na + n2kg - n’) sin (1) 4=d x 1
[nW
.‘)l
n’2
In this formula X is the wave length of the light used (4). If the layer is very thick (d great), the term % d
e
[nzk*
+ (n +
1 ~ 9 1[nak2
+ (n + n’111
dominates in the denominator, so that Z(d) should be written as 16n’ (n’
FIGURE5. TRANSMISSIBILITY OF LAYERS OF IRON AND STEEL AS FUNCTION OF THICKNESS
It must be recalled that to cause interference the coefficient of absorption of the corrosion products must be of the same order of magnitude as that of the metal, which will very rarely be the case. A substance so strongly absorbent that a 10-%m. (10 p ) layer is quite opaque (transparency less than O.OOOl), lets through 95 per cent of the light falling on a layer cm. (apart from reflection). of 5 X If a substance were found that could not be removed from the metal and that had an absorption of the same order of magnitude as the metal layer, the method could still be used if the coefficient of absorption differed from that of the metal. By allowing a thin layer of metal to corrode away entirely, the absorption of the equivalent layer of the substance affected may be determined-i. e., the layer, d*, of the corrosion product formed by complete corrosion of a layer of metal, d. By taking layers of different thicknesses, the connection between absorption and the equivalent thickness of a layer may be experimentally established. From the transmissibility of the plaque its thickness is also found graphically by means of the known absorption coefficient of the metal layer. The exact connection between the thickness of the layer and the optical constants of the metals should, however, be dealt with a t some length here. With the simple measuring apparatus described above, the thickness of the metal layers could be determined with an accuracy of lo-’ cm., which was more than sufficient for the purpose the author had in viewvie., the examination of gas oils, in order to investigate the reproducibility of corrosion by gas oils. OPTICALCONSTANTS OF THINMETALLAYERS The transmissibility of a metal layer of any thickness can be calculated by means of the electromagnetic theory of light. For a thick layer the weakening of the light due to the presence of the layer can be divided into two factors, one representing the weakening by absorption and the other the weak-
A
e
-4rhkd A
+ nW)
(2)
The intensity I ( d ) is now determined by two factors, the first of which does not contain the thickness of the plaque, (weakening by reflection), while in the second the thickness of the plaque occurs in the exponent (weakening by absorp tion). Equation 2 is so much simpler than 1 that there are certain advantages in using layers of such a thickness that 2 can be applied. For if we plot I(d)/I(o) logarithmically against the thickness of layer, the second formula is repre-
FIGURE 6. TRANSMISSIBILITY OF LAYERS OF COPPERAND ZINC AS A FUNCTION OF THICKNESS
January 15, 1935
A N A L Y T I C A. L E D I T I O N
sented by a straight line. It has been found that for the metals used by the author layers of about 3X10-6 cm. can be considered thick in the sense of Equation 2. In Figures 5 and 6 Equations 1 and 2 have been plotted graphically [I(o) = 1.00] for iron and copper with the values of n and k stated for these metals For steel (Fe 1 C) and for zinc only Equation 2 has been plotted. For the calculations the author has taken ?,?=‘n’ = 1.50 (the refractive index for glass). I \ or more, the curves For d = 3 X ‘\ I ; ‘\ for Equations 1 and 2 are practically parallel. For such thicknesses Equation 2 can therefore be used instead of Equation 1 to measure differences in thickness. When preparing the glass plaques the author invariably saw to it that the minimum thickness of 3 X cm. was exceeded, so that Equation 2 could be used. F1auRE OBThere is another reason why it is LIQUE GLASS p A Q u E F O R desirable always to use metal layers MEASUREMENTOFthicker than about 3 X 10+ cni. The n AND k o p t i c a l c o n s t a n t s of the metals (n and k ) are to a great e x t e n t d e p e n d ent on the crystal lattice in which the metal atoms are arranged. Now in a monoatomic metal layer the metal atoms are in conditions quite different from those in a crystal lattice. This lattice is gradually built up in proportion as the number of atom layers increases, so that in general the optical constants for very thin layers may be expected t o depend on the thickness of the layer, and to d 8 e r from the constants of a solid piece of metal. Only by measurement can it be established a t what thickness of layer these deviations become perceptible-i. e., become of the order of 1 per cent. Various investigators have occupied themselves with this dependence of n and k on the thickness of layer d. Leaving out of consideration the old, highly uncertain determinations from the time preceding the development of the high-vacuum practice, the author found that for none of the metals examined were any deviations noticed for layers thicker than 3 x 10- (4). With thinner layers n and k were sometimes found to vary very rapidly (Ag, Pt 1); in other cases layers of even 1 X cm. were found not to deviate in behavior (Sb). The atomic cm., and the distance between diameter of iron is 1.38 X the centers of two iron atoms in the crystal lattice about 1.5 X lo-* cm. A layer with a thickness of 3 X cm. is therefore built up out of 200 atoms. As a rule the author worked with layers of 5 to 6 X l o w 6c m . 4 . e., 300 to 400 atoms thick. It is to be expected that, as soon as the region is reached where the constants change, these changes will be great. (For silver, for instance, n varies from 0.2 to 0.8 for thicknesses below 3X cm.; in the same region k varies from 5.1 to 5.6,2.) For this reason the author tried to ascertain with very simple apparatus whether he could find a change in the constants of iron. To this end iron was evaporated on an oblique glass plaque, so that this was coated with a layer of iron of varying thickness (see Figure 7). The ratios between the thicknesses in given places can be calculated from the distance between these places
+
1I
\
*’
I
These values for n and k cannot be correct, as is seen on closer considera4
tion, for there is an inequality relation between n and k-viz., -(d -k n W
+
+
5
(a 1)* n W (3). Many of the values given in literature for n and IC are found not to comply with the inequality relation. This ha6 been pointed out by Veenemans (4). The author is very much indebted to C. F. Veenemana of the N. V . Philipa’ Gloeilampenfabrieken, Eindhoven, for hia valuable information on this point.
51
and the glowing filament, S, and the position of the glass plaque with respect t o the filament since, considering that the dimensions of the iron bar (length 0.6 cm.) are small in comparison with the distances from the filament to the plaque (minimum distance 5.5 cm., maximum 11.4 cm.), it may be assumed that the iron comes from a “point” source. The quantity of iron precipitated per square centimeter at a place P on the glass plaque is in this case directly pro ortional to the sine of the angle between the line PS and the surface of the glass, and inversely proportional to the s uare of its distance. (The absolute thickness of the layer cm%e calculated on the assumption that the iron evaporates uniformly in every direction. This su position is not quite correct, however, as the direction of the &ament and the direction perpendicular to it cannot be e uivalent. Presumably, all the directions perpendicular to %e filament are, however, equivalent; the author’s calculation was based on this supposition.) In this way the author has calculated the ratio of the thickness of the iron a t various points on the glass plaque and compared it graphically with the thickness found by the optical determination (n and k being taken to be independent of the thickness). As is shown by Figure 8, there were no systematic differences between the calculated and the meas-
0
0005
001
00/5
002
FIGURE 8. d AS A FUNCTION OF ljzr
OOe’5
SIN
c
First experiment, evaporated 8.5 mg. of iron Second experiment, evaporated 5 mg. of iron
ured ratios of thickness, from which it may be concluded that n and k are constants in this range. The thicknesses were calculated in accordance with Equation 1 (which is not an approximate one), n and k being assumed as constants. If this supposition is correct, then there is a linear relation 1 between the calculated thickness and - sin cp, in which T is fJ the distance from P to S and cp the angle between SP and the glass plaque.
APPLICATION TO CORROSION BY GASOILS The test method described here was applied to various gas oils. The fuel pump of a solid-injection Diesel engine is an instructive example of a case in which a large quantity of oil flows along places where even the slightest corrosion is inadmissible. In order t o have some basis to work on, the author calculated the order of magnitude of the corrosion permissible for a Bosch fuel pump with a plunger 6 mm. in diameter. Experience shows that such a pump no longer answers its purpose when the diameter of the plunger has worn down by about om. If this plunger has corroded over a distance of, say, 1 cm., the quantity of iron removed by corrosion is 0 . 6 ~X 1 X loba X 7.9 = 0.015 gram, this being dissolved in the gas oil. The gas oil is considered to be corrosive when the time in which this corrosion takes place is considerably shorter than the normal life of the pump (for instance, half). If, at an average load of 5 horsepower the normal life of the pump is assumed to be 30,000 horsepower-hours, then in half the time 3000 kg. of gas oil have
INDUSTRIAL AND ENGINEERING CHEMISTRY
52
been fed through the pump. About 5 per cent of this has leaked away along the plunger, so that 150 kg. of oil have caused the wear referred to. It follows that a corrosion of only 0.1 gram of iron per ton of gas oil is important. It is clear that a gravimetric determination on a laboratory scale of the quantity of corroded iron will be very difficult, even though during the short time of flow along the plunger the gas oil may not be nearly saturated with iron, so that in a laboratory test an appreciably higher content of iron per gram of oil might be found. In the case under discussion one is thus dealing with a corrosion represented by one part of iron dissolved by ten million parts of gas oil.
'7" t
1 :
//o
tice and those in the laboratory test are found to tally very well, for if the line of contact is drawn in the origin of the corrosion curve in Figure 9 it appears that a t the beginning of the process the corrosion is 5 X loW7 em. per hour. Assuming that a pump has an average life of 15,000 horsepowerhours when corrosive gas oil is used, the pump would have been in contact with the gas oil for about 3000 hours a t a load of 5 horsepower, which would point to a wear of 3 X los X 5 X cm. = 1.5 X om. This is quite consistent with the correct order of magnitude (see above). There is, therefore, a very good agreement between the figures, considering, in the first place, that the figures of cm. in the diameter to express the unserviceableness of the pump is based on an estimation, and secondly, that the temperature and t h e pressure of the oil in the pump differ from those in the laboratory test. Table I1 shows some gas oils examined by the author arranged in the order of decreasing corrosiveness (thickness of layer measured after 16 hours' heating with an iron plaque).
FIGURE9. DECREASE OF THICKNESS AS A FUNCTION OF TIMEOF CORROSION To compare the various gas oils, a 15-cc. oil sample was heated with an iron late in B test tube at 120' C. for 16 hours, after which the digrenee in the thickness of the plate was measured. It had been found that with a given corrosive gas oil (gas oil A) the thickness of layer removed by corrosion reaches a maximum after 16 hours. In this time the corrosive constituents in the oil have been entirely consumed, so that the corrosion after 16 hours is a gage of the total amount of corrosive constituents in 15 cc. of gas oil. (A fresh plate placed in oil that had been heated for 16 hours with an iron plate did not show corrosion.) As is always the case with corrosion phenomena, the duplicate determinations show divergences; t o reduce the influence of these divergences, the author invariably placed several plates separately in a test tube with 15 cc. of liquid. For the definite determinations he used 8 plates at a time. The accuracy of the final result is then about 1.5 X 1 0 1 cm.-i. e., 9 atomic diameters.
Several determinations with gas oil A to find the function of time in the process of corrosion are given in Table I, and have been plotted graphically in Figure 9. TABLEI. THICKNESS OF CORRODED LAYER AS TIMEOF ATTACK PERIOD OF ATTACK Hours 1
3 7.5 14 16 86
DECREASE IN THICKNESS OF IRONLAYER 10-1 cm. 4-7-9-4-3-7-4 5-10-7-9-7-3-6-11 10-16-23-11.5 13-15.5-15 4-10-15-16-7-16-17-8 13.5-10
A
FUNCTION OF
AVERAQE 10-7 cm. 5 7 15 14 13 12
From Figure 9 and Table I it follows that the corrosion approaches a final value after about 10 hours. The final value of the thickness of layer removed by corrosion is 13 X cm. In 15 cc. of this gas oil a maximum quantity of 13 X cc. = 52.10-' X 7.8 grams = X 4 X 1 cc. = 52 X 4.10-4 grams of iron might therefore dissolve-i. e., 3 grams of iron per ton of gas oil. Although this quantity is relatively extremely small, it is yet 30 times as large as the 0.10 gram of iron per ton of gas oil calculated as being inadmissible. The gas oil, when flowing along the metal, therefore absorbs only a small portion-a few per cent-of the maximum soluble amount of iron, As a matter of fact, the conditions for corrosion in the fuel pump differ fundamentally from those in the laboratory test, as in the former case fresh oil is always in contact with the iron for a short time, whereas in the second case a measured quantity of oil is brought into contact with the iron for a long time. It is therefore chiefly the corrosion brought about during the first period of contact between iron and oil which is decisive for the corrosion in the pump. As to the order of magnitude, the results obtained in prac-
Vol. 7, No. 1
TABLE11. CORROSIWN~SS OF VARIOUSGASOILS SAMPLE OF GASOIL
a
AVERAQB DXCR~ASE AVERAGE DECREASE IN TAICKNBSS SAMPL OF~ IN T H I C K N ~ S E OF LAYER" GASOIL OF LAYER" 10-1 em. lo-' om.
15 cc. of ga8 oil after 16 hours' heating.
The results obtained here tally, in so far as it was possible to verify this, with what was known from practice about the corrosion of the gas oil samples in the fuel pump. In Table I11 the acid values of some of the gas oils examined have been tabulated and compared with the maximum quantity of iron dissolved, according to the author's observations. In almost every case this quantity of iron is only a few tenths per cent or less of the amount calculated from the acid value, which shows that there is no correlation between the corrosion in the fuel pump and the acid value. It is not known what substances are responsible for the corrosion. At a decrease in thickness of a X lo-' cm., the oil has absorbed a X 10-7 X 4 x 7.88 grams of iron, whereas, a t an acid value b, the 15 cc. of gas oil used might have absorbed mol. wt. of Fe X l5
'P*
@*
2
x mol. wt. of KOH
TABLE111. LACKOF CORRELATION BETWEEN CORROSIVENESB AND ACIDVALUE DECR~ASE IN IRON SAMPLE THICKNEBS Absorbed Equivalent OF GAS OF ACID byeorro- t o acid OIL LAYER V A L U ~ Sion value 10-1 cm. Mo. Mg. 2 0.13 0.0063 0.80 H 0.022 8.0 7 1.3 F 0.00 29 0 4.7 L 1.5 2.07 0.0047 12.8 J 0.0063 2.5 2 0.4 I 0.86 8 0.14 0.025 C 0 0.25 0.000 1.55 M 10 1.75 0,032 10.8 B
IBON ABSORBED~
% 0.8 0.3 0.00 0.04 0.3 2.9 0.00 0.3
a Percentage of iron absorbed with respect to the quantity of iron equivalent to the acid value.
EXAMINATION OF ALLOYS The examination of alloys demands more caution than that of the pure metals, Small quantities of contaminations will evaporate with the rest and will be found in the layer on the glass. The author is unable to state how far the proper-
A N A L Y T I C A L EDIT1,ON
January 15, 1935
ties of the evaporated impure metal correspond with those of the solid metal, but believes that without further treatment of the metal layer they will generally differ. In fact, in many cases anticorrosive properties of an alloy are obtained by a special treatment, and, therefore, if its properties are to be studied, it is necessary to submit the test object to precisely the same treatment. Therefore it is necessary, in order to apply the method to alloys, to ascertain that the evaporated metal layer has not only the same chemical composition as the alloy to be examined, but also the same chemical propert i e s 4 e., the test plaque has to be treated in the same manner. The first condition-obtaining a metal layer of the same chemical composition as the alloy-may be satisfactorily fulfilled by heating the layer in a vacuum for a sufficient space of time, allowing the metals to diffuse. For instance, when evaporating brass the more volatile zinc is first evaporated and then the less volatile component copper, so that the metal layers on the plaques are white on the glass side, but red on the other. However, after heating for about 2 hours a t 100O C. no difference in color was visible between the front and the back of the coating, both sides showing the color of the brass. This experiment proves that in that time interval the metal atoms have completely diffused throughout the very thin layer. To accelerate the diffusion process, the metal might be sublimed in portions, so that in the case of brass alternating layers zinc and copper would be obtained and the distance over which the diffusion has taken place would be halved. This procedure has been applied to brass and a stainless steel without experiencing any difficulties. The possibility of also fulfilling the second condition, by giving the right treatment to the evaporated metal layer, is being studied at the moment, and the result of a preliminary experiment with a stainless steel was not unpromising. However, it is deemed expedient to defer the work until more experience has been gained. When applying the method to other cases, each case should be regarded separately and the method should be tested as to its utility by comparison with known data. ACCURACY OF
THE
METHOD
The accuracy of the measurement is determined by the divergences caused by the lack of uniformity of the corrosion. After the corrosion more or less sharply outlined dark and light spots can be observed on the plaques, so that even on each individual plaque considerable divergences would be found. It is, indeed, a generally known phenomenon that in corrosion there are very great local differences, without any apparent cause. This is sometimes accounted for by assuming the existence of “nuclei,” from which the corrosion is alleged to originate (in an aqueous medium these nuclei might be local elements). In this case, where the corrosion to be measured is extremely small, the local differences are relatively great. The divergences seem to be of an accidental nature, however, as appears from the observations on gas oil A. According to Gauss’ law of errors
-VzlV2
whereas -is found to be 0.23, in which v represents the dif-
w
ference between an observation and the average. The mean error in the average can therefore now be calculated, for it is equal to the mean deviation from the average value divided by the root of the number of observations minus 1. From this, cnlculations for the heavy corrosions (> 7 X 10-7) in Table I11 give a mean error in the average of about 1.5 X 10-7 for eight observations. The accuracy of the observa-
53
tions is therefore about 1.5 X 10 cm. It seems t o the author that this accuracy is for the present amply sufficient for a classification of products according to their corrosive properties, and for following a refining process quantitatively.
A CRNOWLEDGMENT The author wishes to thank the directors of the Bataafsche Petroleum Maatschappij, The Hague, for permission to publish this article, and his collaborators, J. Voss and J. Boelhouwer, of the laboratory of the Bataafsche Petroleum Maatschappij in Amsterdam. LITERATURE CITED (1) International Critical Tables, Vol. V, pp. 249-52. (2) Lorentr, H. A., Girtt. Nachr., 1911,94-6. (3) Murmann, Hans, 2.Physilc, 80, 176 (1933). (4) Veenemans, C.F.,dissertation, Utrecht, 1933. RECEIVED June 18, 1934.
Making Clean Liquid Sodium Potassium Alloy J. F. BIRMINGHAM, JR. Columbia University, New York, N. Y.
P
OTASSIUM alloys containing 25 to 50 per cent of sodium are liquid a t 15” C. The liquid surface makes them very useful where the maximum reactivity of an alkaline metal is desired, especially a t room temperatures, The ordinary preparation, where the elements are melted together under kerosene, gives small and dirty particles of variable composition. The author obtained bright, shiny globules up to 2 cm. in diameter-approximately 30 grams-by the following procedure: A gram of the liquid alloy (made under kerosene) is transferred into 75 cc. of benzene’ or other hydrocarbon contained in a 150-cc. beaker. A few cubic centimeters of a 10 er cent solutionof ethyl alcohol (95 per cent) in benzene are slow? added until the alloy reacts sufficiently to float. With a pair ottweezers a clean piece of potassium metal is pushed into the alloy until it is assimilated. Next a piece of sodium is added, and so on until the globule is of the desired size. A convenient method of transferring the alloy is to use a small porcelain spoon that holds 3 to 4 cc. Enough solution is scooped up at the same time to keep the surface shiny. Similarly, a very dilute solution of alcohol in benzene will convert the finelv divided residue from a reaction into shinv beads which may be stirred together into larger globules ana recovered for future use (sometimes the addition of some potassium metal is necessary to make the liquid more mobile); or they may be slowly and safely decomposed by successive additions of a few cubic centimeters of dilute alcohol solution in benzene. The alloy may be stored under a purified kerosene containing less than 0.5 per cent of ethyl alcohol. A hydrocarbon mixture such as the “Safety Solvent” of the Atlantic Refining Co. (boiling point range 150” to 200’ C.) distilled from sodium-potassium alloy, provides a water-white nonreactive medium. The alloy is put into the kerosene in glass-stoppered bottles which are loosely closed for several hours. Then the stopper is tightened and sealed in with sealing wax. Under these conditions the alloy may be kept in a bright state for 6 months. RECEIVED October 6, 1934. 1 With ether solution Bee Midgley, T.,Jr., and Henne, A. CHHIY., Anal. Ed., 1,76 (1929).
L., TND. Eno.
