urability of Its Aging in the Dark W. P. VAN OORT iV. Y . De Bataafsche Petrolerim 14aatschappij, Koninklijke-Shell Laboratorirwn, A m s t e r d a m , The 'Vetherlands
0
S E of the important factors deterniining the lifetime of constructions in which asphalt has been used is the influence exerted on the asphalt by the weather. Thc entire complcx of changes in the properties of asphalt by atmospheric influences, to the detriment of the construction concerned, is called aging. The degree to which asphalt resists these influences is called its durability. Literature (1, 3 ) on the aging phenomena of asphalt is extensive. However, most of the published studies describe only the phenomena observed. Frequently, correlations between aging phenomena and properties of the asphalt are sought by purely empirical mcthods. Many methods described aim a t obtaining direct information by short-time tests on asphalt behavior after long exposure. According to the literat'ure, the action of oxygen is one of t'he principal factors responsible for the occurrence of aging phenomena. When asphalt is exposed to atmospheric oxygen, a slow
autoxidation occui-2, the chemical nature of which depends to a very large extent upon the temperature. At temperatuiea above 100" C. dehydrogenation takes place, as is evident from the water produced. Some carbon dioxide is also formed (5). A t lower temperatures-e.g., 25" or 50" C.-the oxygen involved in the oxidation is quantitatively bound in the bitumen and no water or carbon dioxide is formed. The rate of the oxidation may be followed by means of oxygen absorption measurements. The over-all rate of oxygen absorption was found to be not only determined by the chemical nature of the asphalt, but also by the physical transport of the oxygen from the surrounding atmosphere to the interior of the material. Therefore, it is also a physical problem, one of diffusion in particular. A study of the time-absorption curve for oxygen is presented. Both experimenhl and t,heoretical investigations are included to acquire an understanding of phenomena involved by the transfer of oxygen and related factors determining the velocity of the entire process. In order not to complicate this fundamental study, aging in the absence of light was investigated. Such an investigation is very impor7 tant, as most of the asphalt is employed in road carpets and similar constructions, where the greater part of the asphalt is subjected t,o slon. oxidation in the dark, owing to the porous structure of the mixture that usually exists. Some practical data on the change in mechanical properties as a result of aging are given to substantiate this theoretical investigation. A microviscometer developed by Labout and van Oort ( 2 ) was used to collect these d a h .
I2
-.
:1$
sk
8
$
2 P
$ B
4
0
l0
23
40
30
50
RESULTS OF MEASURE>\IENTS
T i M f WEEKS
Figure 1.
Absorption of ox>-genby different asphalts in the dark In 7-micron layers at 22O C. and 1-atm. pressure Pen 25' C .
1. 2. 3. 4. 5. 6.
7.
Middle East Middle East Middle East Venezuelan Venezuelan Indonesian Residue of Dubbs -plant
Softening point Rand B C. 56 52
24 46
57 55 27
52
52 84 52
21 109
45
1196
Oxygen Absorption. The conditions under which the absorption of oxygen was measured 1%-erechosen so as to accord as closely as possible with conditions of actual service. Measurements were made on asphalt in thin films a t temperatures between 20" and 70' C. The procedure used for measuring absorption was based on the conventional
INDUSTRIAL AND ENGINEERING CHEMISTRY
July 1956
1195
2l
.o
PaNTS 0 DEERMNED AT 4OoC. THE OTHERS AT 2 5 % I
I
I
I
/O
20
30
40
I
I
50
60 T/M€ I W€€KS)
SOFTEN/NG POINT
Figure 2.
Viscosity-time curves of asphalts exposed on glass plates in layer thicknesses of 5 microns, in the dark in air at 22" C. Pen 25' C. Middle East Middle East Venezuelan a. Venezuelan 7. Residue of Dubbs plant 8. Indonesian
2. 3.
?.
volumetric method. In order to ensure reasonable accuracy of the measurements, several grams of asphalt were made into a thin film, of about 5 or 10 microns. Grains of sand as nearly as possible of uniform size were coated with asphalt by bringing together the required amounts of sand and asphalt in a heated mixer which could be provided with an inert atmosphere. An example of the curves obtained is shown in Figure 1, where the oxygen absorbed by seven different materials at 22" C. in the dark for 50 weeks is presented. In these tests the thickness of the film was invariably 7 microns. A common feature of the curves is that the rate of absorption decreases with time. These tests indicate that the absorption process is not complete a t the end of one year. Many other experiments under varied conditions have been performed (6) and the curves obtained are generally of the same type but vary in detail. Change in Viscosity. The change in viscosity after exposure of the asphalt in thin layers is measured with a specially developed micro technique (2). This gives an important extension to this theoretical work as changes in mechanical properties are most interesting from a practical point of view. Figure 2 shows the time-viscosity curves for six different materials during exposure in the dark in the air a t 22' C. in layer thicknesses of 5 microns. These curves show that the rate of hardening decreases with time. THEORETICAL DESCRIPTION OF OXYGEN ABSORPTION ( 4 )
Differential Equation for Aging Process. The mechanism of oxygen absorption in asphalt is visualized aa follows.
46 57 55 27 109 52
Softening point R and B C. 52
52 52 84 4.5
45
In the outer layer of asphalt, contiguous to the oxygen atmosphere, oxygen is dissolved a t a relatively rapid rate until a finite concentration is reached. This concentration of free, physically dissolved oxygen is dependent on the nature of the asphalt, on temperature and on the pressure of the oxygen in the gaseous phase. Because of the gradient in the concentration of free oxygen in the asphalt, some of the oxygen will diffuse to deeperlying layers, the rate of this diffusion depending on the gradient and the diffusion coefficient of oxygen in asphalt under the given conditions. In consequence, the asphalt a t various distances from the surface will contain different concentrations of free oxygen, varying with time. However, as soon as a concentration of free oxygen is present in a given layer of asphalt, the oxygen will react with the asphalt a t a rate determined by the concentration of free oxygen and by the reactivity of the asphalt. During the entire process each particle of asphalt is considered to be immobile in the layer. In drawing up a differential equation for this diffusion and reaction problem i t is assumed that the coat of asphalt on the grain of sand can be considered as a flat system. This is certainly permissible, since the ratio Thickness of layer 7 X 10-8 Radius of sand grain - 4 X 10-1
- about
1 50
The calculations are therefore carried out with one coordinate running from x = 0 a t the sand-asphalt interface to x = a a t the asphalt-gas interface. The differential equation based on the above conception then reads
INDUSTRIAL AND ENGINEERING CHEMISTRY
1198
Ga ( D g ) = z +a c5 T
ag
The right hand side of Equation 1 expresses the change with time of the total oxygen concentration (physically dissolved and chemically bound) present in a n infinitely thin layer dx at a distance x from the asphalt-sand interface. It is therefore a rate term. Consequently the left hand side is also a rate term, a fact which is discussed as follows. The diffusion coefficient D is described as the quantity of oxygen crossing unit cross-sectional area per second under the driving force of a unit oxygen concentration differential. Hence
bc
D-ax expresses the quantity of oxygen crossing unit cross-sectional
Vol. 48, No. 7
I n order now to arrive a t a calculation of the amount of chemically bound oxygen as a function of time and place in the layer, Equation 2 must be introduced. The total amount of absorbed oxygen can then be calculated by solving
M
=
lu(c+
g)dx, which leads to
The distribution of bound oxygen in the layer as a function of time is given by
area per second under the driving force of the oxygen concentra-
a3
tion differential experimentally present, and finally - D-
is
the change with depth (per unit depth) in the rate of oxygen passage per unit cross-sectional area. The general Equation 1 must be adapted to the specific problem under investigation, for which purpose use of the experimental results mentioned must be made. Some assumptions must be made, the justification of which must be sought in the final correlation of the derived equations with experiment.
z,
The first assumption deals with the reaction rate a9 which is taken to be of the first order. This assumption leads to
tanh (a Equation 7 contains the term co _ _ . This term can be P
omitted from the total amount of oxygen
M
-
absorbed per gram,
since c,, is the solubility of oxygen in asphalt under experimental conditions, a quantity of the order of some tenths of cubic centimeter per gram (1 cc./gram corresponds to about 1.4 kg. of tanh 'p oxygen per cubic meter of asphalt). Moreover, is always ~
9 at
=
P
kc(G, - g )
1, as will appear later from numerical
examples.
There seems to be no experimental evidence for assuming a higher reaction order. I n addition, this assumption simplifies mathematical treatment of the differential Equation 1. Equations 1 and 2 combined now produce
tanh p Without the term co _ _ Equation 7 may be written as :
(3) The second assumption made is based on the time-viscosity curves of Figure 2, which shows that viscosity increases with time-Le., with the amount of chemically bound oxygen g. Diffusion coefficient D will generally decrease with increasing viscosity, so t h a t the dependence of D on g will be a decreasing function. This may be written in a general form as
(4) Here Do is the diffusion coefficient for oxygen in unoxidized asphalt, 01 a positive constant, whereas p , . . .can be either positive or negative. This equation, as it stands, however, is insoluble. Therefore, two further simplifications were introduced, in hope that the solution of the modified equation would give representative results.
2
The first simplification consists in n-riting D 2aC
for
-
tx(
3
D -
the second amounts to assuming t = 0, which means that the concentration of free dissolved oxygen is stationary. As a nevt step, in the equation thus modified we put 01 = 1 and p = 0, which leads to
A simple method of trial and error allows Equation 9 to be adapted to the experiment, by first choosing the right value ot P, after which the values of k and D o can be evaluated unequivocally, provided Go and c o are known. I n Figure 3 the experimental time-absorption curve of a Venezuelan asphalt has been plotted. As expressed above, the values of Go and co must be known for further calculations with Equation 9. By extrapolation of absorption measurements after 100 hours a t different layer thicknesses to layer thickness zero an estimate of the value of Go could be obtained. A figure between 6 and 8 cc./gram (corresponding to about 8-12 kg. oxygen/crt meter asphalt) XTas indicated. The calculation of the expeiimental curve of Figure 3 was carried out with G o = 8 cc./gram (but Go = 6 cc./gram could have been chosen as well) and e , = 0.4 cc./gram. For a description of the experimental curve 'p had to be chosen a t 8. It followed that kc, = 2.4 X 10-5 sec. -I. I t then appears that k = 4.2 X 10-j cu. meter (asph.)/sec kg.(Oz).
As
(D
= a
42
-anda =
7 X 10-8 meters it is found
that D o= 3 6 X sq. meters/sec. To demonstrate clearly that the time-absorption curve can actually be described with the constants so calculated, the curve computed theoretically and the measured points are compared in Figure 3 . VERIFICATION OF THEORY
Assuming that the foregoing simplifications are justified, the solution of Equation 5 gives the distribution of the physically dissolved oxygen in the layer:
From the example given, Equation 9 provides a good description of the experimental time-absorption curve. Before examining the other experimental results and calculating the diatribution of chemically bound oxygen in the layer, it iE advisable to test the theory for its correctness. Some experiments mere carried out to ascertain this assumption
INDUSTRIAL AND ENGINEERING CHEMISTRY
July 1956
1199
\ D = Do ( 1 - g/G,). An endeavor was made to establish the relationship between D and g experimentally. A Venezuelan asphalt with the penetration of 55, thickness layer of 7 microns, was oxidized a t 50" C. for 50 hours under 734 mm. Hg oxygen pressure. From this curve and Equation 9, it is possible to calculate Do-i.e., the coefficient of diffusion in the fresh product-if g = 0 cc./gram. After 50 hours the asphaltsand mixture was taken from the absorption flask and again revolved for a short time in the mixer under carbon dioxide a t 150" C. This treatment was to homogenize the asphalt coating around the grains of sand, because oxidation has caused a certain distribution of oxidized product over the cross section of the coating. The material was returned to the absorption apparatus and another oxygen absorption test carried out for about 100 hours at 50" C. The product subjected to this test was an asphalt of which the bound oxygen content was precisely known. From the renewed absorption test the diffusion coefficient of the asphalt could be calculated, which value, combined with the known quantity of oxygen absorbed before homogenization, provided the second point in the Dlg graph. A second series of experiments along the same lines was carried out on another lot of fresh asphalt-sand mixture, allowing 115 hours for the first oxidation period. A third series produced data for oxidation of 268 hours. The calculations by Equation 9 were carried out with Go = 6 cc./gram for the fresh product, the amounts of oxygen already absorbed being subtracted from 6 cc./gram for the various homogenized products so as to obtain the new Govalue for the absorption test after homogenization. The result of plotting D against g is shown in Figure 4. An almost straight line can be drawn through the four points; this line, if extrapolated, intersects the axis a t about 6 cc./gram. In other words, the line complies with the formula D = D.(1
-
6). - I
The conclusion from these experiments is that D decreases linearly with an increase in g. These results are in agreement with the assumption that the decreasing function of D will be D D,(1 - g/G,). By ignoring &/at with respect b dg/dt in the differential equation the total amount of gas is assumed to be physically dissolved and also its distribution in the layer is constant during the entire absorption experiment. It has been possible to demonstrate experimentally that this supposition is reasonable. At two different moments the quantity of oxygen that can be pumped off is determined; this quantity represents the physically dissolved gas, amounting to co tanh p/p according to Equation 7. A direct determination of the amount of gas capable of being pumped off is not possible in the available apparatus, but can be found as fOflOWB.
0
1
2
3
4
5
6 crn.3r0,
q
I m
Figure 4. D as a function of g calcu1ated:from oxygen absorption measurements with Venezuelan asphalt at 50" C
An oxygen absorption experiment with a material of which the time-absorption curve was known was interrupted after a certain time by pumping off the gas in the absorption vessel. The pumping-off period was made sufficiently long to remove all physically dissolved gas. The oxygen absorption was then restarted. After some time the new time-absorption curve showed the same shape as that previously determined without interruption, if compared a t corresponding oxidation times (Figure 5 ) . The difference in the total amount of oxygen absorbed, read from between both time-absorption curves, gives the amount of physically dissolved gas a t the moment of interruption. It appears from Figure 5 that after oxidation of 101 hours in a '?-miiron layer at 50" C. of a Venezuelan asphalt having a penetration of 55, 0.11 cc./gram can be pumped off; after oxidation of 220 hours, 0.10 cc./gram. The difference remains within the error of measurement. If the time-absorption curve for the fresh product is calculated with Equation 9 and Go is taken to be 8 cc./gram it follows that p = 6 X co can then be calculated from 0.10 = co tanh p/p, and c , is found to be 0.6 cc./gram. This figure is in fair agreement with a value of 0.4 cc./yram that was found for the solubility of argon under the same conditions. These experiments show that the omission of dc/dt with respect to a g / d t is certainly permissible. In Equation 9 the constants c., Go, k , and Do occur. If timeabsorption curves are determined a t several oxygen pressures only the solubility co will acquire a different value. This solubility can be taken to be proportional to pressure. To ascertain whether the introduction of this dependence into the theory leads to a good description of the experiments, three time-absorption curves were determined on the same asphalt a t the same temperature a t 154, 500, and 734 mm. Hg oxygen pressure. The lower the pressure, the lower was the absorption of oxygen found. From Equation 9 the following figures for k and Do could be derived from the data so obtained. These results are in fair agreement.
INDUSTRIAL AND ENGINEERING CHEMISTRY
1200 c
voi. 48, NO. 'I
SOME APPLICATIONS OF THEORY
31,
I
I
io0
200
J
300 TIME INOURS)
Determination of the quantity C ,
Figure 5.
Tanh
(e
~
at
(e
50' C.
Go = 8 cc./gram Oxygen Pressure, Mm. Hg 154 500 734
1.7 X 1.3 X 1 . 9 x 10-5
9 . 8 X 10-'6 8.6 X 7 . 0 x 10-'6
A quantity that depends on experimental conditions is the thickness of layer a. If oxygen absorption measurements are carried out on an asphalt a t a given temperature and pressure but with a different thickness of layer, then by substituting new values in Equation 9 the various time-absorption curves can be described with the same k and D ovalues. To test this theory three measurements were carried out a t 50" C. on asphalt-sand mixtures having calculated thicknesses of 5, 7, and 9.5 microns, respectively. At smaller layer thicknesses, higher oxygen absorptions, expressed per gram of material, were found. The results indicate that a reasonable agreement is obtained. Layer Thickness, M
2 4 1.9 2 5
5 7 9.5
/0J6
x x x
5 5 x 10-16 7 0 x 10-16 7.2 X
10-5 10-6 10-6
-
k
864-
2-
NO. PEN, 25 OC.
som mmr R AND B OC.
56
8-
2 3 9
46 57 94
52 52 45
6IO
I n conclusion, the time-absorption curve for oxygen can be described theoretically. The velocity of the absorption process is characterized by the coefficient of diffusion of oxygen in asphalt and the constant for the reaction between oxygen and asphalt. The coefficient of diffusion decreases m-ith an increase in the amount of oxygen bound; the rate of the reaction decreases likewise. A detailed comparative investigation of different, asphalts by the oxygen absorption test and theoretical calculation of magnitudes D oand 12 is not a very reliable method, the difficulty being that Go is not known. Moreover, the amount of absorbed oxygen is not so important-from the practical point of view-as, for instance, the degree of hardening. However, the theory can be applied in the study of many problems related to the aging of asphalts. Though there is some uncertainty regarding the absolute values of D o and k as material constants, the variations of several experimental quantities occurring in the equations have little or no effect on the value of Do and k . Therefore, the theory can aid in ascertaining the relationship between viscosity of asphalt and the coefficient of diffusion. It is also possible to ascertain how the distribution of bound oxygen in t'he layer develops as a function of time and what changes it may undergo when k and D o change. Here the relationship of Iz to D0--i.e., the relationship of rate of consumption to rate of transport of oxygen-plays a role. The values of D ofor some >fiddle East asphalts, derived from absorption experiments a t 22" and 50" C., have been combined in one graph as log D oagainst log penetration (Figure 6). Over this long range of penetration the points lie on a straight line, as would be expected according t,o the theory that the coefficient of diffusion decreases xith increasing viscosit,y (decreasing penetration). By measuring the oxygen absorbed by one asphalt a t various temperatures it was ascertained how the 12 value to be derived from it depends on temperature. Then, the time-absorption curves a t 25', 50", and 75' C. for a Venezuelan asphalt having a penetration of 55 and 25" C., in a layer thickness of 7 microns and a t 734 mm. Hg oxygen pressure, n-ere measured. I n calculating Do,Go was taken as 8 cc./gram for all temperatures; the temperature dependence of solubility co was taken into account. The values for t.he reaction constant IC calculated from the three oxygen absorption lines are plotted in Figure 7 as log k against 1 / T . The activation energy that may be derived from this line is 24 kcal./mol. vhich is a normal value for a chemical reaction. The very slow penetration of oxygen into the asphalt and the slowness of the chemical reaction are the cause of the fact that the chemically bound oxygen is not evenly distributed in the layer, From Equation 8 t.he local concentration of chemically hound oxygen q can be calculated a,s a function of the distance to the asphalt-oxygen interface and of time. It is essential t o know IGo and 'p = a /AGO. I n this calculation the uncertainty in
/oo IO00 PENETRATION AT ABSORPTION TEMP.
Figure 6. Relationship between calculated diffusion coefficients and penetratrion for Middle East asphalts at 22" and 50" C.
t,he knodedge of G , makes it difficult to compare diffcrent asphalts. For the time-absorption curves of Figure 3 it was calculated that, with Go = 8 cc./gram, k = 4.2 X 10-5 cu. meter/kg. sec. and D o = 3.6 X lo-'* sq. meter/sec. The results of calculating g according to Equation 8 are shon-n in Figure 8. The slower increase in the amount of bound oxygen with time, as clearly evident in the various time-absorption curves, can also be seen in the distribution graph. At the end of oxidation of 300 hours hardly any bound oxygen is a t the bottom of the layer, a t the asphalt-sand interface; in other words, t'he depth to which the oxygen penetrates is very small. The action of oxygen forms a surface skin which is physically characterized by a decreasing coefficient of diffusion. The transport of oxygen through this
INDUSTRIAL AND ENGINEERING CHEMISTRY
July 1956
1201
8-
64-
Plo5,8-
64-
L
t
-
/OS
8-
64-
2
1
0:s SANDLASPHAL T BOUNDARY
Figure 7 .
Temperature dependence of calculated reaction constant
skin becomes increasingly difficult o n h g to the decrease in D and the growth of the skin. This oxidation skin therefore has a protective effect against further osidation of the material below it. The increase in g causes a rise in viscosity. Because of the distribution of g through the depth of the layer the viscosity of the asphalt will also vary with the situation of the latter in the layer. ,4n average viscosity is determined with a microviscometer, ( 2 )because the preparation technique applied involves homogenizing the aged product removed from the exposure glass or from the grains of sand. I n judging the practical consequences of aging this procedure must be considered in some cases, because a coating of asphalt aged in seqvice will have a harder surface, and therefore more brittle skin than that which follows from the measured viscosity. I n cases of small penetration of oxygen the measurement of aging phenomena, such as oxygen absorption and hardening, is only possible if very thin layers are used. The slight penetration accounts for t h e fact that modern road constructions, in which layers of 5 to 10 microns occur owing to the low content of asphalt, have a long life in spite of their porous structure. A similar factor is the additional retarding effect on aging exerted by the very narrow and elongated pores of the road carpet, which slow down the transport of oxygen. C O N C L ~ S I O Ns
The theory given accounts for the primary phenomena of aging, thus providing a satisfactory basis for gaining insight into the factors determining the rate of this process. A first conclusion is that experimental conditions correspond better t o reality if the aging properties of asphalt are examined by exposing thin layers of about 5-micron thickness t o atmospheric influences instead of thicker layers. An adequate tool for measuring changes in mechanical properties due t o the aging of these thin films is the microviscometer. The fundamental insight gained will be used in studying the effects on aging of varying the properties of an asphalt. Further, the development of a n accelerated aging test can now be undertaken with a good chance of arriving a t a procedure of which the significance can be fully understood. The use of high oxygen pressure without an increase in temperature is being investigated.
’
0.8
POSITION IN 7HE LAYER
I
p
I.
ASWALT’GAS BOUNDARY
Figure 8. Distribution of chemicallv bound oxveen in Vgnezueian asphalt, Pen (25’ C.) 55; softening point R and B 52” C. In a 7-micron layer after various periods of oxidation a t 50’ C. in the dark
NOMENCLATURE
a c
= layer thickness, meter = concentration of free oxygen in asphalt, kg. per cu. meter
co =
solubility in asphalt, kg. per cu. meter
D = diffusion coefficient, sq. meter per second
Do = diffusion coefficient of fresh material, sq. meter per second q = concentration of chemically bound gas, kg. per CLI. meter Go = maximum concentration of chemically bound gas, kg. per cu. meter
k
= reaction constant, cu. meter per kg.-sec.
t
= time, seconds
z
= coordinate, meter
CY
= constant
M = total quantity gas absorbed by whole layer, kg. per sq. meter
p = constant
ACKNOWLEDGMEXT
The work presented in this paper was carried out a t the Koninklijke/Shell-Laboratorium, Amsterdam. The author’s thanks are due to the management of these laboratories for their consent to publish this work. H e is indebted to R. N. J. S a d , J. M. Goppel, and J. J. van Deemter for their stimulating interest and for their valuable suggestions and t o J. P. Spaanderman and A. D. Langeveld for carrying out the experiments. LITERATURE CITED
(1) Highway Research Board, Washington, D. C., “Bibliography on
Resistance of Bituminous Materials to Deterioration Caused by Physical and Chemical Changes,” Bibliography No. 9, 1951. (2) Labout, J. W. A., Oort, W. P. van, Anal. Chem. 28, 1147 (1Q.56). \--_-,.
(3) Neppe, S. L., Trans. South Africa Inst. Civil Engrs., 1951, pp. 1 9 5 2 2 3 ; 1952, pp. 103-34. (4) Oort, W.P. van, “A Study of the Ageing of Asphaltic Bitumen,” thesis, Delft, Holland, 1954. ( 5 ) Pfeiffer, J. P., “The Properties of Asphaltic Bitumen,” pp. 111-13 Elsevier, Amsterdam, 1950. RECEIVED for review September 9, 1954. ACCEPTED February 20, 1956. Division of Petroleum Chemistry, 126th ACS Meeting, New York, September 1954.
