INDUSTRIAL AATD ENGINEERING CHEMISTRY
January, 1927
I n run 63, 125 grams of glyceride were heated in a stream of pure nitrogen in a 500-cc. Pyrex flask. The glyceride was brought up to 293' C. in 15 minutes and held at this temperature, *2O, for the duration of the run. Table IV SAMPLE
Glyceride 1
2 3
AFTERHEATING MOL. Minutes WT.
Run 62
...
771
999 2687
(1
15 27
Solid Run 63
0
11 .-
Glyceride 1
2 3 4
... ... ...
176
34
162.3 162 162 160.3 158.9 151.6
2
1 ... ... ... ...
20
7 8 9 10
34.2
17 15
5
10 15 a
HEXABROMIDEIODINENo. No. OF OIL
31 40 50 60
132:o
Run 64 601 1095 1760 2618
... ... b
... ...
b
Solid
b
b b
95 a
10 20 23
145:2
... ...
70
...
... ... ...
126.9
b
_______ (I
b
Time required to heat up to 293" C. Duplication of run 62 with slightly purer linolenic glyceride.
65
The rate of molecular weight increase for runs 62, 63, and 64 is given in Table IV which also shows the rate of decrease of the hexabromide and iodine numbers in run 63.
Escape of HzO and other products of condensation is hindered by the slow stream of nitrogen in a closed flask and is facilitated when the heating is carried out in an open dish. Condensation reactions, with consequent large increase in molecular weight, are therefore favored by heating in the open dish. Conclusions
A study of the relative changes in hexabromide number, iodine number, and molecular weight indicates that in the first stages of heating linseed oil intramolecular change is the main reaction. When linseed oil was heated with sulfur, selenium, and tellurium in nitrogen, little or no condensation occurred. Organic sulfides, selenides, and tellurides were evolved and in the selenium run the molecular weight decreased. A linolenic glyceride was synthesized and studied. When heated it thickened and changed to a solid jelly. The rate of molecular weight increase was the most rapid observed so far for linseed oil and any of its derivatives. The thick products obtained by heating linolenic glyceride dry and form films similar to those derived from linseed oil.
Effect of Velocity on Corrosion of Steel under Water' By R. P. Russell, E. L. Chappell, and A. White DEPARTMENT OF CERhfICAL ENDINBRRING, MASSACHUSRTTS INSTITUTE
OF
TECHNOLOGY, CAMBRIDGE, MASS.
The present conception of the corrosion process' under water is that at the anode the reaction Fe 2 @ = Fe++ takes place. Simultaneously, a t the cathode, there is a reaction, which, under ordinary conditions, consumes dissolved oxygen. The anode and the cathode may be quite widely separated or may be very close together.
surface. The data in the literature on this subject are, however, decidedly contradictory. Heyn and Bauer3 report that if iron is corroded by water, the rate increases rapidly a t first with the rate of flow. They find that a maximum is reached, however, a t about 0.004 foot per second, and that subsequent increases in water velocity cause a decrease in corrosion. These tests were carried out by passing water over iron plates suspended in beakers. The velocities studied were all low
1 Received August 27. 1926. Presented before the Division of Industrial and Engineering Chemistry at the 72nd Meeting of the American Chemical Society, Philadelphia. Pa., September 5 to 11, 1926.
* Except under exceptional conditions, such as high acidity, high temperatures, etc. 8 Mitt. kgl. Maferiulprilfungsamt, 28, 62 (1910).
Theoretical Considerations and Previous Work
+
INDUXTUAL AND EN@,[NEERINQ CHEMISTRY
66
(maxiinurn of 0.07 foot per second) arid are somewhat uncertaiii on account of uiiknown coiivection currents in the beakers. Similar results, though with the niaximum corrosion corresponding to a much higher xater velocity, are reported by Friend..' He suspended strips oi Kahlbaum's electrolytic irori foil in glass tubes by means of platinum wire. Tap water was then run through the tubes at velocities as high as about 8 feet per second. IIis results are shown in Figure :3. Friend found that the corrosion rate increased up to a velocity of 0.15 feet. per second-i. e., approximately 30 times the velocity corresponding to maxiinurn corrosioii reported by Heyii arid Bauer. The corrosiori then decreased to a minimum a t about 2.5 feet per second, aiid subsequent velocity increases caused a slight increase in corrosion. At no point above 2.5 feet per second, however, did Friend find a corrosion rate as high as in quiet water. These results were advanced by Friend in support oi his colloidal theory of corrosion. Speller and Kendallbapproached the problem in a somewhat different way in an a t t e m p t t o d u p l i c a t e service conditions more closely. They determinod the corrosion rate by measuring the decrease in dissolved oxygen concentration of a known volume of water passed FIB-* I--Phofomirrobraph of Low-Carbon through commorcial Steel (Appmx. 0.10 Per cent Carbon) steel pipe of various X 150. Etched with nitai. Shows small grain .ize and reasonabie freedonl from sizes. Their results with 0.75-inch pipe at 27" C. are shown for comuarison in Fiieure 5. Sueller and Kendall found a regula; increase of-corrosion Awith velocity with no indication of a decrease in corrosiou over the entire velocity range covered by either of the previous investigations. The major differenees bet7reen the various sets of data would appear to be in the condition of tho metal suriaces employed, since Heyn and Bauer, and Friend used fresh, unrusted and presumably partly polished test pieces, while Speller and Kendall employed commercial pipe which was at least partly coated with rust. It seemed necessary, however, to obtain further experimental data beiore accepting this coriclnsion as final.
Val. 19, No. 1
before they were subjected to the different velocities. Five tubes of differing diameters were arranged in series; the water flowed througli them from a constant-head tank, the rate of flow being measured by moans of the pressure drop through a calibrated orifice. Four specimens were suspended along the centerlirie of each tube, as tar from the ends as possible in order to avoid undesirable (and unknown) effects of contraction and expansion disturbances. The oxygen content' of tlie water was essentially constant (6.0 to G.1 cc. oxygen per liter), and it was found by experiment that there was no appreciable decrease in oxygen conceiitratiori through tlie apparatus on account of the relatively small surface area of exposed metal as conipared with the large amount of water passing through. Tho tube diameters %-ere1.0, 1.42, 2.70, 5.87, and 11.6 em.?all approximately 1 meter long. The corrosion rates were determined by weighing the test pieces before and after each run, according to the equation: Averaee oenetration. centimeters ner "ear =
The temperature in all the runs was 21' C . * 1.0". The duration of runs varied from 48 hours to 2 weeks. Runs of 48hour duration were made with polished steel (brought down to 000 emery finish) and a,ith steel which had been roughened by a 3-minute pickliiig in G N hydrochloric acid a t 70" C. A 4-day run was made on specimens that had been rusted previously by subjecting them to a constant velocity (0.75 foot per second) until their entire surfaces were completely coated with a practically uniform rust film. A run of 2 weeks' duration %'as made on samples which had been rusted previously in the manner just described. The amount of corrosion which took place during the preliminary rnsting was determined by cleansing some of the test pieces and weighing. The cleaning in all eases was accomplished by W p i ing freefromrust with a wet towel and drying with alcohol. The metal used in all the runs was a lowcarbon steel a photomicrograph of which is shown in Figure 1. An x-ray diagram (by Experimental Procedure the pinhole method) Two procedures were outlined to obtain the necessary is shim in Figure 2. Note t i c sliyht % m w n t d residua1 &ai" and Flbering. Molybdenum target used data. The first was to study the effectof velocity by rotating test pieces were a vertical cylindrical test piece at surface speeds up to 8 of 20-gage metal, 6.3 by 0.8 em., drilled at each end so that feet per second in distilled water. This was done in a they could bo suspended in series in the glass tubes by means thesis investigation6 carried out under the direction of of cord. one of the present authors. The results, shown in Figure 4, Results are discussed later. The second procedure was to suspend steel test pieces in The results are given in the accompanying table and iu tubes through which was passed Cambridge (Mass.) tap water. Tube diameters were chosen to give essentially Figures 3,4, aud 5 . In Figure 3 is plotted the rate of corrothe velocity rarige covered by Friend and by Speller and sion us. water velocity for the 48-hour runs on polished and Kendall. Tesi-s by this method were planned to include rough steel. Friend's results are plotted in the same figure (1) polished specimens, (2) rough specimens, and ( 3 ) speci- for comparison. Figure 4 gives the results of the tests in mens which had been rusted under tlie same conditions wliich the vertical cylindrical test pieces were used. Figure 5 shows the results of the 4-day and 2-week runs in com~~
I ~ o nS l r d I n d . (London), Cnrnrgic Sihol. Mcm., 11, 128 (19221, 'rms J O ~ X W L , i s , 134 (1923i. 1 Rickeis, Jr.. M. S . Thesb, .M. I. T., 1923. 4
6
1 Determined arcording to standard Winkler procedure, Am. Publie 1 5 4 t h AEBOC., Standard Methods of Water Analyd.. 1928. p. 59.
INDUSTRIAL A N D ENGINEERING CHEMISTXY
January, 1927
parison with the results obtained by Speller and Kendall in 0.75-inch pipe at 27" C. Friend's results are also reproduced on a scale comparable with that used in Figure 3. 1240
1
I
67
and therefore affected by cvcn minute variations in conc1itio:is a t or on the metal surface, are almost always erratic. Unfortunately an absolute comparison could not be made between the penetrations observed by Friend and the prcawit authors, because c crtain essential data are missing f i om Friend's communication. Furthermore, no attempts berc made in the present investigation to determine the corrosiori at zero velocity, because under quiet conditions there is added t o the uncertainty of short t,csts doubt as to the oxygen concentration during the experiment. Effect of Velocity of Rate of Under-Water Corrosion
06
AVERAGE
VELOCITY OF WATER
0.044 0.173 0.824 3.03
02
Y
( A S RELATIVE CMIROSION)
Rough steel-48-hour
run
RATS"
0.036 0.054 0.100 0.056 0.062
6.45 VELOCITY-
YHNETKITION
CN./yenr
Ft./sec.
FTISEC
Figure 3
Discussion of Results
Figure 3, in which Friend's results are compared with the results of 48-hour runs on polished and rough steel, clearly demonstrates that, with uncorroded steel surfaces, the corrosion a t first increases with velocity up to a maximum. At higher velocities lower corrosion rates are obtained. It will be noted that the corrosion of the highly polished steel specimen is less a t the very high velocities than that observed a t very low velocities. The effect of roughening the steel surface is to shift the point of maximum corrosion to a higher water velocity and, also, to cause the rates observed a t water velocities above the maximum corrosion point to be greater than those obtained a t the low velocities. That is, there appears to be a gradual transition from the type of corrosion observed by Friend toward an over-all increase in corrosion with increasing velocity as the surface 06
051
4
I
j
Sfeel with uniform rust jlm-96-hour 0.048 0.187 0.890 3.27 6.95
Steel with uniform ~ u s frlm-Z-vetk t 0.040 0.156 0.745 2.74 5.81
run
0.070 0.065 0.105
0.151 0.103 run
0.048 0.047 0.101 0.107 0.064
Average of four specimens; average deviation less than 10 per cent; no attempt made to measure pitting.
In all the tests reported in this paperJ even at the highest velocities studied, rust appeared on the metal surfaces. This is contrary to the findings of Friend, who states that the action a t the highest velocities is one of erosion. It has been pointed out by W h i t n e ~ however, ,~ and later by Whitman,'o that under similar conditions there is no action in the absence of oxygen, and that the "erosion" postulated by Friend is merely corrosion at a velocity high enough to sweep away the products of corrosion from the relatively
rm
EFFECT OF VELOCITY ON CORROSION OF STEEL UNOLR WATER A S DETERMINED 0 Y I ROTATING CYLINDER METHOD (TEMP 2 P C )
-
40
W
-B2 5
32
" 24 01
I
I
I
I
P
5 +
W
OOO
I
2
3
4
VELOCITY-FT
/SEC
5 6 (SURFACE SPEED)
7
6
9
w 0
08
Figure 4
is progressively roughened. The explanation for this change may be one of differential oxygen concentration as suggested by Evans.8 That is, a t the higher velocities, with the greater turbulence which results, it may be more difficult to obtain differential oxygen concentration areas and corrosion may be more difficult to start, The results obtained in these two short tests demonstrate that the type of corrosionvelocity curve reported first by Heyn and Bauer, and later by Friend, is obtained when the surface is unrusted initially. The check is a remarkably good one when it is remembered that results obtained during the initial stages of corrosion, THISJOURNAL, 17, 363 (1925).
00
I
3
4
VELOCITY-
5 fT/SEC
6
9
Figure 5
smooth surface which he used. This conclusion is substantiated by the fact that rust appeared on the test pieces even a t the highest velocities used in the present investigation. Figure 4, which gives the results obtained with the rotating vertical cylinders, shows that results similar to those in Figure 3 are obtained when steel is corroded by distilled water (instead of tap water) at varying velocities. It is significant 10
THISJOURNAL, 17, 385 (1925). Chem. R e v , 2, 419 (1926).
68
INDUSTRIAL AND ENGINEERING CHEMISTRY
to note that the corrosion was lowest when the surface velocity was practically aero. When Figure 5 is compared with Figure 3 it is seen that over a wide velocity range, the effect of increasing the water velocity is to cause an increase in the corrosion rate of rusted steel. In other words, when test conditions approximate service exposure of uncoated metal under water, the results are similar to those reported by Speller and Kendall. This would be expected, as stated earlier, from the increased ease of diffusion of dissolved oxygen which results from an increase in water velocity. Thus by altering one surface condition of the metal it was possible to obtain results similar to those of Friend, and by suitable control of this same factor to pass to the other extreme of reported results and obtain data similar to Speller's. In the two longer runs on previously rusted metal shown in Figure 5, there is a range of maximum corrosion. This is not narrov, however, ,as is the case with short tests on bright metal but extends up to a velocity of nearly 4 feet
Vol. 19, No. 1
per second. The iilm of rust obtained at the highest velocities is particularly dense, hard, and tough. What effect more extended test periods would have on the corrosion at very high velocities is not known. This subject is being investigated, however, and the results will be reported later. Conclusions
1-The corrosion of steel under water increases with increasing water velocity over the range commonly encountered in practice providing the metal is coated with rust. 2-By suitably altering the surface condition of a steel it is possible to reproduce the apparently conflicting data which previously have appeared in the literature concerning the effect of water velocity on the corrosion of submerged steel. Acknowledgment
The photomicrograph of the steel used was made by V. 0. Homerburg and the x-ray diagram by E. W. Brugmann.
The Chemistry of Chinese Wood Oil',' By F. H. Rhodes and C. J. Welz' CORNELL UNIVERSITY, ITEACA, N. Y.
The effect of heat treatment upon the molecular HINESE wood oil conas the solvent, and that the weight, iodine number, and oxygen-absorbing power variations of the apparent sists essentially of the glyceride of eleomarof Chinese wood oil was determined. Gelatinized wood molecular weight have been garic acid. The raw oil conoil was extracted with petroleum ether, and the soluble due to experimental error. tains from 90 to 95 per cent and insoluble fractions thus obtained were examined. Both benzene and chloroform of this eleomargarin, the reThe heat of gelatinization of Chinese wood oil was are known to cause associamaining 5 to 10 per cent being measured. Upon the basis of the experimental retion of the dissolved oil, and olein, unsaponifiable matter, sults thus obtained, there is formulated a theory to Wolff considers that the apetc, Eleomargaric acid itself account for the observed changes which occur during parent variations in molecular the bodying and the gelatinization of Chinese wood oil. weight have been caused has been shown to be an unsaturated fatty acid of the by variations in the degree of formula :* association of the solute in the CHa(CH2)aCH:CH.CH: CH(CHZ)&OOH solvent. He does not, however, explain clearly why these The bodying of Chinese wood oil is accompanied by an experimental errors have been consistently in the direction increase in the viscosity, specific gravity, and apparent which indicate higher molecular weights for the heated oils. molecular weight of the oil, and by a decrease in the iodine Wolff himself made some cryoscopic measurements, using number and the amount of oxygen absorbed during drying. camphor as a solvent, and found that the apparent molecular There is little or no change in the acid-number or the saponi- weight of Chinese wood oil does not increase when the oil is fication number unless the heat treatment is effected a t un- bodied. In further support of his hypthesis, Wolff points out usually high temperatures. Most authorities agree that these that when Chinese wood oil is heated the increase in viscosity observed changes are due to the polymerization of the glyceryl and the decrease in iodine number do not take place concureleomargarate, although different investigators advance rently, the change in iodine number being almost completed different hypotheses as to the exact mechanism of the polym- before the most marked change in viscosity takes place. eriaation reaction. Wolff, however, holds that the thickenExperimental Work ing of the oils when heated is due solely to a change in the In the investigation of the effect of the heat treatment upon physical state of the 0il-i. e., to gelatinization-and that polymeriaation plays no essential part in this change. He the properties of Chinese wood oil here reported it was consays that the data for the determination of the average sidered desirable to develop a satisfactory method for followmolecular weights of the oil before and after heating have been ing the change in the molecular weight of the Oil with the obtained by cryoscopic methods, using benzene or chloroform temperature and time of heating. Hoffman6 and Long and Smull' have determined the molecular weights of heat1 Received June 29, 1926. treated Oils by the cryoscopic using benzene as a 2 This article is respectfully dedicated by the authors to Prof. I ,. M. solvent. This however, is open to criticism beDennis. I t will be reprinted as Article No. 4 in the Louis Munroe Dennis Quarter Century Volume, to be published in 1928 in commemoration of the cause of the known fact that benzene tends to cause some completion by Professor Dennis of 25 years of service a s head of the Departassociation of the solute and may therefore give incorrect ment of Chemistry a t Cornell University. The article is based upon the results. The use of camphor as a solvent in cryoscopic dethesis presented to the faculty of the graduate school of Cornell University termination has been suggested by Rast19 and by C. J. Welz in partial fulfilment of the requirements for the degree of doctor
C
of philosophy. I Du Pont Fellow in Chemistry at Cornell University. 4 Fokin, J , Russ. Phys.-Chcm. Soc., 46, 283 (1913). 5 2. angcw. Chcm., 81, 729 (1924).
THISJOURNAL, 17, 175 (1925). I b i d . , 11, 138 (1925). 8 Comfit. rend., 164, 1692 (1912). Bcr., 66, 1051,3727 (1922). e
@