R. C. Hochel and E. A. Clark
Westinghouse Savannah River Company
Aiken, SC 29808

This report was prepared as an account of work sponsored by an agency of the United States Government. Neither the United States Government nor any agency thereof, nor any of their employees, makes any warranty, express or implied, or assumes any legal liability or responsibility for the accuracy, completeness, or usefulness of any information, apparatus, product or process disclosed, or represents that its use would not infringe privately owned rights. Reference herein to any specific commercial product, process or service by trade name, trademark, manufacturer, or otherwise does not necessarily constitute or imply its endorsement, recommendation, or favoring by the United States Government or any agency thereof. The views and opinions of authors expressed herein do not necessarily state or reflect those of the United States Government or any agency thereof.

This report has been reproduced directly from the best available copy.

Available for sale to the public, in paper, from:  U.S. Department of Commerce, National Technical Information Service, 5285 Port Royal Road, Springfield, VA 22161,  phone: (800) 553-6847,  fax: (703) 605-6900,  email:  orders@ntis.fedworld.gov   online ordering:  http://www.ntis.gov/support/ordering.htm

Available electronically at  http://www.osti.gov/bridge/

Available for a processing fee to U.S. Department of Energy and its contractors, in paper, from: U.S. Department of Energy, Office of Scientific and Technical Information, P.O. Box 62, Oak Ridge, TN 37831-0062,  phone: (865 ) 576-8401,  fax: (865) 576-5728,  email:  reports@adonis.osti.gov

This report is the second by the authors on characterizing the tritium content of cement and structural concrete. The first report reviewed the literature and used several new methods to characterize tritium on the surface and through the bulk of contaminated concrete at two facilities at the US Department of Energy Savannah River Site (SRS). In both cases, a relatively constant tritium concentration as a function of depth was observed, which has not been previously reported in the literature. These findings were explained as the relatively rapid transport of tritiated liquid through pores of the hydrated cement, and the exchange of tritium with hydrogen found primarily as free and bound water in the hydrated cement binder. The study reported here extended the measurement of surface and bulk tritium in concrete to three other locations at SRS.

The purpose of the current study was to characterize locations whose tritium exposure histories were well documented, and to characterize a location exposed exclusively to gaseous tritium, to confirm and possibly extend the knowledge gained from the earlier study. Results of the current study corroborate the earlier findings, in that the tritium concentration was constant through the bulk when exposed to aqueous tritium, even from a single aqueous tritium exposure. Exposure to gaseous tritium, on the other hand, lead to the well-known diffusion controlled variation of tritium concentration reported in the literature. Sufficient exposure history is available to enable a semi-quantitative explanation of the magnitude and depth dependence of the tritium in both the aqueous- and gas-exposed locations. The penetration of tritium from a liquid can be described by a hydraulic flow model, and gaseous tritium permeates in a diffusive manner. The general correlation of properly measured surface tritium activity to that in the underlying bulk found in the earlier study was confirmed. However, the surface and near surface tritium concentration is sensitive to changing humidity conditions and other surface and environmental effects, so the correlation may be only approximate in specific cases.

An earlier report by the authors [1] reviewed published data and introduced new tritium characterization methods for hydrated cement and concrete in contaminated facilities. That study took place in two separate Savannah River Site (SRS) facilities and led to the discovery of a previously unreported constant tritium concentration through the bulk of the concrete. The two facilities investigated were purposely selected because of what were assumed to be different contamination exposures and environmental histories. One was an indoor elemental tritium gas processing facility, and the other an outdoor facility that processed heavy water moderator used in the site’s nuclear materials production reactors. Neither of these locations were operating, and their contamination histories were largely unknown. However, it was expected that the two would typify differences in exposures from gaseous and liquid forms of tritium.

The goal of the previous work was to find better and faster methods to characterize tritium in cement and structural concrete. Small areas of contaminated floor in the two facilities were selected for tritium characterization using surface and bulk techniques. Results obtained from two different in situ ion chamber-type devices were compared with those obtained by liquid scintillation counting of surface smears to ascertain tritium surface activity. A handheld hammer drill was used instead of cumbersome core sections to collect samples to measure the tritium content as a function of depth through the bulk. The analysis of the hand drilled samples for tritium was simplified by a new rapid laboratory procedure, which enabled consistent and accurate results to be obtained in as little as one day. This procedure involves nitric acid leaching of tritium from the concrete powder sample, distillation, and finally liquid scintillation counting. The application of these new characterization methods proved very successful and superior to the traditional methods in cost, speed, contamination and personnel exposure, waste generation, and reliability of results. An additional bonus was what appeared to be strong correlation between tritium surface activity, as measured with the ion chamber devices, and that found in the underlying bulk.

After completing the earlier study, questions remained about the meaning of the relatively constant tritium depth profiles in the two facilities. The literature [2-6] strongly suggests that the concentration of tritium as a function of depth in concrete, as with many other materials, is established by a diffusion controlled process. The results from the earlier study seem to question this supposition. The depth profiles did not show the sharp tritium concentration decline with depth expected for a diffusion controlled process. Instead, the profiles were nearly constant with depth throughout the bulk, decreasing slightly at the boundary surfaces. This was interpreted in the earlier study to be the result of exposure of the concrete surfaces to tritiated liquid instead of tritiated vapor. The exposure at the heavy water facility to tritiated liquid is a historical certainty, but exposure to a liquid in the elemental tritium processing facility was unexpected. The nearly identical shapes of the tritium depth profiles in the concrete at the two facilities presented strong evidence that both locations were exposed to liquids and, in fact, anecdotal history of liquid spills in the elemental tritium processing area was substantiated by a former worker.

The initial report [1] considered a number of chemical and physical properties of concrete [6-12] and proposed a general model to explain this previously unreported constant bulk concentration. Briefly, hydrated cement binder in the concrete has the capacity to rapidly imbibe a considerable amount of water from a liquid-exposed surface through the many capillary voids which permeate the bulk. Once imbibed, much of the water or other tritiated liquid is incorporated into the bulk through various types of hydrogen exchange, radiolysis induced reactions, and diffusion. To further study the distribution of tritium in concrete and to corroborate this theory, several additional surface and bulk measurement studies recently completed at three locations in another SRS facility are reported. This facility has a well characterized exposure and environmental history. One of these locations is known to have been exposed solely to gaseous tritium (DTO), and the two other locations are known to only have been exposed to tritiated heavy water.

The 105-K main building of the SRS K-Reactor facility has ongoing missions to store spent fuel and plutonium, and is currently well staffed and well maintained for radiological work. It was the last SRS reactor to cease operating, and has the best documented history of contamination events.

All five former production reactors at SRS were moderated with non-pressurized heavy water (D₂O). They operated at above ambient temperature, just below the boiling point. In each, the reactor tank has an open top but is contained in a room with isolated ventilation. The K-reactor operated from 1953 to the end of 1991 during which time the average operational losses of moderator were approximately 45 kg per day, mainly by evaporation. The reactor moderator was transferred to a better storage tank in 1993, and since then losses have dropped to about 0.5 kg per day. Relatively small amounts of tritium were created in the heavy water upon neutron irradiation during reactor operation. The peak DTO concentration in the moderator reached about 10 Ci per liter in the past, but is now about 5 Ci/L. Small amounts of activated corrosion and fission products also exist in the reactor moderator.

In May 1999, an opportunity to use the tritium in concrete characterization tools and methods developed earlier arose. Because of seismic safety concerns for the 60 meter high 105-K building ventilation stack, a project to dismantle the top 20 meters was planned. Stack concrete samples at several different elevations were required by the project to properly characterize the radiological content of the rubble which would be generated. This provided the opportunity to study the stack concrete exposed exclusively to the gaseous tritium-containing heavy water vapor that evaporated from the moderator over the life of the reactor.

Although evaporation accounted for most moderator losses, from time to time moderator water spilled in various processing parts of the building. Two locations in K-reactor have had exposure to moderator water only (not gaseous tritium) and were well suited for a corroborative study of liquid tritium permeation in concrete. First, in February 1990 a line break in one of the heat exchanger header pipes, which unintentionally contained moderator water, resulted in a significant spill in the building, which was well documented. It is believed that this was a "one and only" exposure of the concrete floor to contaminated heavy water in the immediate area. Moderator of about 9 Ci/L stood on the floor for several days during cleanup operations. Second, in the so-called "makeup room", 208-liter moderator drums were routinely opened, sampled, weighed, and chemically adjusted, with occasional leaks or spills. A complete contamination history for this room does not exist, but enough is known to support or refute most measurement conclusions.

In this report, first results from the tritium concentration depth profile analyses at these three locations (K-reactor stack, heat exchanger, and makeup room) are presented and the profile results are discussed in light predictions from a diffusion model. Next, an alternative model, which incorporates bulk flow concepts with those of diffusion and better fits most of the profiles observed thus far at SRS, is proposed. Finally, tritium surface analyses results, as an indicator of bulk concentrations, at the same locations are compared and discussed.

The profile sampling and analysis methods for this work were developed in the previous study [1], with minor changes necessitated by work practices in the facility. Samples were collected by hammer drilling with either a 16-mm or 19-mm bit in 2.5-cm increments to a predetermined depth. The drill powder from each increment was collected in labeled glass vials and sent to an onsite laboratory to minimize analysis cost and reporting time. Tritium analysis consisted of nitric acid leaching of the concrete powder samples, distillation, and finally liquid scintillation counting to detect tritium. A part of the concrete powder from each sample was used to measure the calcium content by inductively coupled plasma-emission spectroscopy. As previously discussed [1], normalizing the tritium concentration to the calcium concentration reduces variations in the apparent tritium content due to the effects of drilling through locations having differing amounts of large aggregate, and so gives a more representative estimate of the true tritium concentration variation with depth despite the small sample size. The sequence of grinding and vacuuming the surface of the measurement location and making a surface activity measurements prior to the profile drilling had to be altered in some cases because of operational constraints. These changes are described separately for each location discussed in this section.

Characterization of the 105-K building ventilation stack had to be specially designed because a minimum stack air flow was maintained at all times for proper building ventilation. Drilling from the outside inward was rejected because it would have required twice as many samples to reach the interior surface region where the exposure took place, and because the thickness of the stack varied with elevation which risked "punch through" and loss of sample. A team of four persons (driller, collector, backup, and monitor) was lowered in a safety cage into the stack to remove samples at four different elevations, by drilling from the inside out. Despite these very difficult working conditions, 24 samples (six 2.5-cm depth increments at each elevation) were safely obtained in only a few hours. Surface activity measurements would have been very difficult in the stack as it stood, and so measuring surface activity was deferred until suitable pieces were available after dismantling (discussed below).

The depth profile data, to a depth of 15 cm, for the four elevations taken in the stack are shown in Fig. 1. The tritium concentrations have been normalize to the calcium content in each sample. The legend indicates the approximate elevation down from the top of the stack, and the approximate total thickness of the stack at that elevation. The "plenum" is a concrete splitter wall about two meters high at the bottom of the stack, and it separates the flow of reactor process area air from that of the rest of the building. Below the plenum, one side of the wall was exposed only to clean air while the other was exposed to air containing moderator vapor. Above the splitter, the two air streams begin to merge as they move up the stack. The stack is 40 meters high, and sits on top of the 20-meter high reactor building.

The lines connecting the sample points are only a visual aid in Fig. 1. Because of the challenging conditions during drilling and sample collection, individual increments may have been cross-contaminated with the previous sample, or even accidentally placed in the wrong sample vial. Only the general shape and approximate relative concentrations of the curves should be viewed as significant. The tritium profiles are peaked and relatively high for the locations near the center and upper two-thirds of the stack, but less peaked and lower at the top and bottom locations.

The depth profile data from the floor under one of the 105-K heat exchangers are shown in Fig. 2. Holes 1 and 2 were drilled about 10 cm apart to a depth of 23 cm and the tritium concentrations with depth are seen to be nearly the same, as expected for such close proximity samples. The measured concentrations are approximate because the drill holes could not be vacuumed between increments, and cross contamination is more likely than in our previous study (facility rules require special approved HEPA-filtered vacuum cleaners, which were not available at the time of the work). The tritium concentrations in this case are about two orders of magnitude higher than found in the 105-K stack (Fig. 1), and the concentration profiles are relatively flat and constant over a greater depth. The concrete slab at this location is 90 cm thick, yet the general shape of the profile over the top 25% of the total bulk changes little except, possibly, to increase slightly in the first 5 cm. The painted surface coating at this location was removed by grinding [1] but no vacuuming was done. Surface activity measurements, which require much more time than profile drilling, were delayed several days because of schedule and equipment constraints.

The steeper profile slopes in the makeup room compared to the heat exchanger can be understood, when the room’s function is considered. Nitric acid was often added to SRS reactor moderator in the makeup room, to maintain required moderator chemistry. Acid spills were not uncommon. All strong acids react quickly with the hydrated calcium silicate gel in hydrated cements and concrete to neutralize Ca(OH)₂. This liberates water and calcium ions. Much of this water, including any tritium it contains, would then be lost by evaporation. The calcium ions left in solution in the capillaries are then free to react with dissolved CO₂ to form insoluble CaCO₃. The net result is that carbonate replaces tritium and water near the surface of acid-exposed concrete, and this is evident in Fig. 3. This same process also occurs naturally in concrete exposed to only CO₂, but is slow and self-limiting.

Comparing the concentration depth profiles of the stack (Fig. 1) to those in the floor of the heat exchanger (Figs. 2) and the makeup room (Fig. 3), convincingly demonstrates that exposures of concrete surfaces to tritiated liquids produce a concentration profile distinctly different from exposures to tritiated vapors. Liquids deposit large amounts of tritium through most of the bulk, whereas vapors deposit significantly less and in much more varying concentration profiles. These data confirm most of the observations in the previous paper [1].

The approximate tritium concentration and exposure duration at each of the three sampled locations in the building allow the measured profiles to be compared to diffusion models, at least on a semi-quantitative basis. Fig. 4 shows the ideal diffusion profiles for the spreading in one dimension of an amount of substance M deposited at time t = 0 in the plane x = 0 according to Eq. (1) by Crank [13],

C = M/(2(pDt)^1/2) exp(-x²/4Dt)       (1)

in which the concentration C at any point into the infinite medium varies with M, depth x, diffusivity D, and time t. Rigorous treatment of the actual problem yields a similar but a more computationally difficult solution. For the sake of simplicity, Eq. (1) is a useful approximation for showing the progress of diffusion with time; however, an estimate of the tritium surface activity initially deposited must first be made. Assuming a value of 25 mCi (5 l of 5 mCi/mL moderator) seems to produce profiles within about an order of magnitude of the ones measured above.

The values plotted in the upper two curves of Fig. 4 are for the heat exchanger floor spill of tritiated moderator (a 10-year exposure) but with grossly different values of D (5x10^-8 cm²/s vs. 5x10^-5 cm²/s). For the stack exposure to DTO vapor depicted by the bottom curve in Fig. 4, the average surface tritium was reduced by a factor of about 800 to account for the lower density of vapor compared to liquid. The concentrations on a mass basis are approximated by dividing the results of Eq. (1) by of 2.3 g/cc, a nominal density of concrete. The profiles are plotted only down to the drill depth at each location, which is clearly adequate to depict the curve shape. The diffusion coefficients are assumed to be a constants independent of the tritium concentration, whose magnitudes depends upon the identities and physical properties of the diffusing material and the medium. Krasznai [5] has reported a diffusivity for DTO vapor of about 5 x 10^-8 cm²/s for normal density concrete as measured at a heavy water reactor, and this is the assumed D for the top and bottom curves. The diffusion profile for the stack is for a 40-year exposure duration.

From the monitored loss rates of moderator (45 kg/day at an average concentration of about 5 Ci/kg diluted by a 3110 m³/m flow), the stack can be estimated to have been exposed to a nearly continuous flow of air having about 50 mCi/m³ of DTO during the 40 years of reactor operation (1953-1993). From 1993 to the present, the air flow has continued, but contains only about 1% of the previous exposure activity. Six years ago, the tritium depth profile presumably would have looked very much like the curve in Fig. 4: a slow decrease in tritium concentration with depth into the concrete. This profile shape is also very similar to that reported by Numata et al. in concrete walls in an operational heavy water reactor [4]. The profile for the 105-K stack, however, has been changed by a subsequent diffusion of tritium out of the concrete and back into the relatively clean air that now flows across the inside surface of the stack. This has resulted in the two peaked curves of Fig. 1, which actually approximate the reverse of that in Fig. 4. The depth location of the peak depends on the thickness of the stack, the relative tritium concentration of air inside and outside the stack, and the length of time the back diffusion has been in progress. If these actual boundary conditions of the stack are known, it is possible to calculate a more specific profile using finite difference or other numerical methods. Even without such a calculation, the measured curves are very similar in shape and magnitude to what is expected qualitatively from the mathematics of diffusion.

All the profiles in Fig. 1 suggest a concave down shape and have maxima near the center of the stack wall. This is most obvious for the minus 12- and 20-meter elevations. The stack has a tapered thickness which decreases from bottom to top. At minus 12-meters, the stack is about 4 cm thinner than at minus 20 meters. In each case, the profiles seem to peak near the middle of the stack, but without directly sampling the outer half of the stack this is unproven. However, the data definitely indicate deep penetration of tritium into the stack during the exposure to higher activity air flow, which is now diffusing out both to the lower activity air flow inside the stack and to the outside of the stack. The profiles at the top and bottom plenum have a less pronounced peak shape; the peaks are visible if the scale in Fig. 1 is expanded.

The difference in concentrations at the upper and lower stack elevations compared to the middle warrants some explanation. The bottom plenum splitter has the thinnest cross-section sampled (15 cm). Because of the functional dependence of Eq. (1), concentrations with depth into the thin plenum splitter change faster in response to a similar change in the tritium source term than those in thicker regions. Also, one side of the splitter was always bathed in a flow of uncontaminated air from the non-process areas of the building, so it is logical for the splitter to have a lower tritium concentration. The lower tritium activity at the top is unexpected. This might be due to more rapid weathering and back diffusion at the high elevation of the top, when compared to the splitter at the bottom that is largely unaffected by weather. Complicated flow effects of winds at the top would also play a role. Given detailed information about initial and boundary conditions, diffusion theory presumably could predict the shapes and magnitudes of tritium concentration with depth at different stack elevations in a much more quantitative manner.

The two upper curves of Fig. 4 show the dramatic effects of a large change in D (or t) on the shape of the hypothetical diffusion profile. The small D value assumed for the top curve in no way resembles the concentration dependence with depth that was actually observed for the liquid DTO moderator spills in the 105-K building (Figs. 2 and 3). However, several studies have shown that diffusivity and permeability of dry-cured cement, after being resaturated with water just once, can increase by several orders of magnitude due to microcracking [14], which could lead to a much higher diffusivity of perhaps 5 x 10^-5 cm²/s instead of 5 x10^-8 cm²/s. It is interesting to note that this higher value is comparable to those generally found for liquid-in-liquid diffusivities (~ 10^-5 cm²/s) [15]. It may be that the huge, moist surface area of the hydrated cement solid actually looks like a liquid for diffusion purposes.

As mentioned in the introduction, the capillaries and pores of concrete are normally filled or partially filled with free or loosely bound liquid water. Accordingly, the transport of tritiated water through concrete could be viewed as a liquid diffusing through the water filled capillaries. In fact, an aqueous tritium diffusion coefficient of about 5 x 10^-5 cm²/s was measured by Eichholz et al. [8] in experiments on large prepared samples of water saturated concrete. Using this higher diffusivity, again for a ten-year exposure time, the very flat curve of Fig. 4 is obtained. The shape of the profile matches what is seen in Fig. 2, but the magnitude is low by about a factor twenty. Rapid diffusion both into and out of the bulk results in the loss of much of the initial 25 mCi surface exposure over a ten-year exposure duration, so a much higher initial surface exposure (500 mCi) would be required to produce a profile matching those in Fig. 2.

It is interesting to further compare these profile shapes with those in our previous paper [1], as shown in Fig. 5. The five highest profiles are all SRS facilities. An average of the multiple sample drillings at each location is used for each data point. The profile for the SRS 105-K stack is the average of the minus 12- and 20-meter data in Fig. 1. The lowest profile in Fig. 5 is plotted from a wall profile at the heavy water Japan Research Reactor (JRR-3) reported in Ref. [4]. The approximate concentration of tritium in the exposing fluid, either liquid or vapor, is shown adjacent to each plotted profile. For the SRS 234 H Tritium Facility, the exposure medium may have been water or oil having an activity of 5 Ci/L or more (previous report [1]). The value at the surface of each profile is calculated by subtracting the value drop between the first two measured points from the value of the first point. Such a value for the 105-K stack turns out to be negative and is not plotted in Fig.5. Simple extrapolation of the two points back to zero depth would give a higher approximation of the expected bulk concentrations near the surface. The true value probably is somewhere in between. Also, in contrast to Fig. 1, 2, and 3, the data points are connected by a smooth curve to give a better picture of what the expected continuous profile curve would look like. The slope of each curve between the calculated surface value and the first bulk increment value correlates well, in a qualitative sense, with what we know of the environment conditions at each facility.

The two facilities with the greatest slopes are both subject to some or all of the outdoor weather elements, and profiles of the outdoor sump (previous report [1]) and 105-K stack (current work) are quite similar despite the fact that the sump was exposed to liquid while the stack was exposed to vapor. The long 40-year active exposure of the stack no doubt reached a zero-gradient steady state where rates of inward and outward diffusion were about equal. The subsequent 8-10 year period of air flow containing only a trace of the former levels has been rapidly removing bulk tritium, especially in the surface region. Bulk levels are only slightly higher than the JRR-3 wall which was exposed for only half the time and at 1% the concentration of the stack. Presumably this is so because the thick reactor wall was sampled while the reactor was operating and nearly all the tritium it was exposed to remains. The wall also shows a profile expected for vapor diffusion into a thick slab (Fig. 4) when not exposed to harsh outdoor weathering effects.

Because 105-K stack flow has been continuous, the inside surface suffered little from the weather but was bathed in relatively dry air which led to rapid evaporation of the near surface moisture. In contrast, the outdoor rework sump was exposed numerous times to moderator of about 5 Ci/L, and between spills was significantly affected by the weather. In addition in this outdoor location, the concrete likely reached full hydration before it was ever exposed to moderator and much tritium would have remained unbound in the capillaries relatively free to move. Cycles of partial evaporation of tritiated water from the pores and capillaries followed by the influx of tritium-free rainwater has further reduced the present tritium content to a fraction of its original exposure values; the last known exposure to moderator was about a year (around 130 cm of rainfall) prior to the sampling. Wet and dry weather cycles are so frequent that the near surface is probably striped of tritium faster than it can be replenished by that in the bulk.

The JRR-3 wall, the SRS 234-H Tritium Facility floor, and SRS K-Reactor heat exchanger building floor (HX) are all indoor environments protected from most weather changes and water intrusion. Only the heat exchanger floor shows a slight downturn in the near surface profiles. The floor was exposed to tritiated liquid only once ten years ago; still back diffusion and evaporation appear to have resulted in only a minor perturbation to the bulk concentration. The tritium facility floor exposure history is mostly unknown, but it is in a climate-controlled location where surface losses would be expected to be small and slow.

The 105-K Reactor building floor (MR) continues to subjected to occasional spills even now as drummed moderator is stored or transported. As an indoor location, surface evaporation losses should be fairly low. But nitric acid spills, as explained above, have depleted the near surface tritium and the concentration depth profile more resembles that of a weathered outdoor facility than that expected for an indoor facility.

According to Crank [13], valid solutions of the problem of diffusion into a semi-infinite medium having zero initial concentration and a surface maintained at a constant value involve but a single parametric expression: x/2Ö (Dt). It follows then that depth of penetration and the amount of material diffusing through a unit surface area is proportional to the square root of time, and the time for any point to reach a given concentration varies directly with the square of the depth and inversely with the diffusion coefficient. These properties only hold provided the initial concentration is uniform and the surface concentration remains constant. Clearly, the application of such idealistic diffusion solutions is rather limited.

An alternate treatment can be used to model the permeation of tritium into concrete, based on the flow of an incompressible fluid in a nondeformable porous medium. Assuming a steady state laminar flow, Darcy’s law relates water flow to the hydraulic pressure gradient. The hydraulic pressure gradient arises almost entirely from large suction forces within the tiny gel pores and capillaries of the concrete bulk. In one dimension, the advance of a sharp wetting front with time is given by [14a].

where B is the penetration coefficient, which depends on the capillary forces and the chemical interaction of the liquid and solid. Thus Eq. (1) is replaced by one of much simpler form, where the depth of penetration from a surface exposure to tritiated liquid is just proportional to the square root of time. (Note the functional dependence on time is the same as that of a diffusion process.) B can be determined experimentally by so-called "Infiltration and Capillary Absorption" measurements on real concrete specimens. Infiltration introduces the liquid from the top of the specimen while capillary absorption allows the liquid to enter from a slight immersion of the bottom. Measured values, in both cases, are the volume or mass of liquid absorbed and the depth of the wetting front. The wetting front is measured visually after splitting the exposed specimen axially.

Some results from the literature are shown in Table I [14 b]. Included in the table are the type of test, the water to cement ratio, the moist curing time, age at the time of the test, the duration of the test, the measured values, and the calculated penetration coefficients. The data are not extensive but the trends are clear. Penetration increases with the water to cement ratio and is aided even more so if the specimen has been dried. Most interior concrete in buildings is not moist cured for more than a few days and will have dried somewhat with age. Since a water to cement ratio of 0.54 is common for general-use concrete, the last two penetration coefficient values in Table I (average 15.9 mm/d^1/2) are likely to be conservative if assumed for further calculations.

Eq. (2) predicts the progression of the wetting front into the bulk of concrete with time using a single parameter, which is obtained experimentally from controlled measurements on real materials. It is also applicable to liquid spills because it includes the observed effects of imbibition. The advantage of this hydraulic model compared to diffusion models is that the measured penetration coefficient accounts for all the unknown physical and chemical effects that complicate the diffusion model.

Two assumptions must be made in applying the hydraulic model to the floor spills in the 105-K building. First, DTO is assumed to behave like normal water. The more massive DTO molecule presumably will move much slower than normal water, but radiolytic effects probably speed isotopic exchange and may compensate for the slowness due to mass difference. Second, it is assumed that the same square root dependence measured for the short times in Table I is valid for much longer times. This is not intuitive, but there is some support for this again from the work of Eichholz et al. [8]. It reports that "the rate of wetting (i.e., the velocity of any wetting front) did not show any difference between thin and thick samples". Also from an empirical perspective, perhaps we can expect the same dependence for much longer times and down to much deeper depths. The table data indicate that after three days, a wetting front in 0.54 w/c concrete will have progressed to about 27 mm (15.9 mm/d^1/2 x (3d)^1/2). The physical and chemical conditions in the capillaries beyond this depth should be relatively constant as only the upper most portions are subject to large changes in moisture and porosity that affect the capillary flow. If so, one would predict the penetration to advance to 42 mm in seven days. The 7-day infiltration test of a 0.57 w/c sample gave a 50-mm penetration in reasonable agreement. Only 1.66 L/m² was absorbed, which is likely a reflection of the 44-day sample age and wetter capillaries than those at 90 days.

In the case of the spill onto the floor beneath the heat exchanger, the surface exposure duration was, at most, a few days. Table I suggests that in a few days each square meter of floor absorbed perhaps 2.5 L of tritiated moderator, or on average about 0.25 mL/cm². This is equivalent to 1.25 mCi/cm² of tritium taken into the bulk of the concrete slab. The slab is about 90 cm thick. Assuming the same initial rate of penetration has continued over the past ten years (3650 days), the wetting front could have progressed to 96 cm or all the way through the slab. Distributing this evenly from top to bottom in one dimension of the 2.3 g/cc concrete slab would result in about 6 mCi/g. Allowing for penetration in all three dimensions, the tritium concentration would be about 2 mCi/g of concrete. The actual value measured is about 0.5 mCi/g. Since none of the surface losses have been accounted for, this agreement must be considered quite good. Using the same approach for the 30 cm floor slab in the makeup room, full penetration would occur in about a year to give a concentration about three times higher or 6 mCi/g. This also compares well to the actual measured value of about 2 mCi/g. Tritium concentrations in the makeup room floor are among the highest encountered thus far at SRS. Presumably this is the result of sustained multiple spills reaching back nearly 50 years. Tritium concentrations probably cannot go much higher because of some continual loss from evaporation and diffusion out of this relatively thin slab.

It is, however, not necessary to assume the hydraulic model will hold over these long times. Once a spill allows imbibition to progress for a week or so, the wetting front will have moved far enough into the bulk to reach capillaries that are mostly water-filled, regardless of most ambient drying conditions. At this point, the more rapid liquid-in-liquid diffusion illustrated in Fig. 4 is sufficient to produce a nearly constant depth profile to 25-30 cm in about a year. This is a short time for nearly any facility, so any past spill of a tritiated liquid onto a concrete surface will likely produce the profiles observed in these SRS studies.

As mentioned, facility constraints at the105-K building forced the surface activity measurements to be made under conditions less ideal and controlled than in our earlier study [1]. In the case of the stack, surface activities measurements were not attempted until six months after the depth profile work. The earlier profile data showed that maximum bulk tritium concentration occurred at about mid-stack height. Stack dismantlement started at the top and worked down, so the sought pieces were the last to be generated. A piece roughly 15 cm across and 5 cm thick, coming from the inside surface of the stack, was taken to the area’s radiological control laboratory for measurements using E-Permä electrets [1]. Electrets are polymer disks that are initially charged to a fixed voltage. Any tritium or other weak beta surface activity produces ionization of the air in the narrow gap between the and the contaminated surface that discharges the disk. The amount of surface activity can be related to the reduction of the electret voltage over time. The flat surface was ground smooth and vacuumed prior to placing two electrets on the surface: one open to measure the surface activity of the tritiated concrete and the other capped to measure background. The electrets were checked over a two week period. Voltage changes were at the sensitivity limit of the devices, and the surface activity was estimated at about 1x10^-3 nCi/cm².

The same procedure of placing two electrets side by side was used at the heat exchanger and makeup room locations; however, measurements times of a few days or less were sufficient to measure the tritium surface activity. Although the surfaces were again ground relatively smooth, they could not be vacuumed prior to the measurements, only lightly brushed. The surface activity in the makeup room measured 0.41 nCi/cm² and nearly the same as the 0.45 reading obtained on the unprepared surface some four months earlier while scouting for suitable study locations.

Scheduling difficulties required a change in this procedure at the heat exchanger location. After grinding the surface, the sampling holes were drilled and the excess drill powder was swept away from the holes with a brush. Then the electrets were placed for four days. The surface activity measured 0.36 nCi/cm², but the electrets had been errantly placed over the drill holes. The measurements were repeated for two days with the electrets placed properly, and the surface activity was found to be 0.43 nCi/cm². The increase compensated almost exactly for the hole area missed in the first measurement, and matched the 0.42 reading obtained some months earlier.

In our earlier paper [1], a correlation between surface and bulk activity was postulated and tentatively shown. The data used to establish this correlation included only two measurement sites and the surface activity at the one of lower bulk activity, an outdoor location, could only be estimated because electret readings suffered from weather effects. Measurements at the higher activity indoor location were much more certain. Bulk activity there was about 1 m Ci/g and the surface activity measured about 15 nCi/cm², suggesting a correlation factor of around 0.015 g/cm².

Conversely if the measured surface activity is divided by this factor, an estimate of the underlying bulk activity is obtained. (It was shown in Ref. [1] that this empirical relationship has a theoretical basis and is related to the large apparent surface areas of hydrated cements.) Doing so for the three measured surface activities in this study would give the following approximate bulk values: 0.07 nCi/g for the stack, 27 nCi/g for the makeup room, and 29 nCi/g under the heat exchange. From Fig. 5, the agreement is seen to be quite poor, with predicted bulk values being well below the measured ones.

The shapes of two of the curves near the surface (stack and makeup room) suggest the problem in applying this analysis. The correlation factor is based on data from the SRS tritium facility floor spill, which has a very flat profile in contrast to the other four SRS measurements. Three of these show steep drops in the calculated surface activity and a moderate drop in the fourth. Recall that the surface activities in Fig. 5 are calculated from the drop between the first and second profile samples. These represent an estimate of what the bulk concentration at the near surface might be if it could be measured. A simple extrapolation of first two data points back to the surface provides another estimate. The calculated values are all smaller than the extrapolated ones, and the difference tends to be larger for the steeper sloping profiles. Although the term "surface activity" is frequently used and applied, such as in limits for radiological classification and for free release criteria, most attempts (except smears) to measure the apparent surface activity (that read from any direct measurement device) also respond to some of the subsurface tritium in porous materials. Because of this it is expected that the bulk concentration very near the surface should be roughly reflected by apparent surface activity measurements.

Table II shows that this is true. The surface activities for all five SRS locations in Fig. 5 were measured using electrets which, as discussed in Ref. [1], actually can detect tritium betas within about 70 microns of the concrete surface. From the measured surface activities in the second column and a single correlation factor (0.015 g/cm²), the near surface bulk activity is calculated in the third column. The best estimate of the actual near surface bulk activity is the average of the estimated minimum (calculated near surface activity) and the estimated maximum activities (extrapolated near surface activity). Surprisingly, agreement (within a factor of three) between columns three and four is best for the sump and stack, which have the sharpest profile slopes. The lesser sloping profiles of the makeup room and the heat exchanger floors both are low by a factor of 12 or so. This is unexpected, but has a plausible explanation. The floors in both locations were painted relatively recently. Prior to this waxes and sealers were applied to the makeup room floor to maintain appearance.

In sampling, the floors were ground to remove the paint layer, but there was no deep abrasion or other attempt to remove any clear subsurface materials that might be partially sealing the surface. This might be the cause the observed low surface activity measurements. Wax may be clogging the pores of the concrete in the makeup room, and the low reading at the heat exchanger may have a similar, but as yet unapparent, explanation.

The relatively good correlation of surface measurements to near surface bulk is, as stated above, expected when using a device such as an electret, which detects some tritium below the top surface. However, the ultimate goal of establishing such a correlation is to eliminate the need for subsurface sampling. A glance at Fig. 5 or the last column of Table II, which shows the tritium bulk concentrations below the first few centimeters, suggests that some limited subsurface sampling is prudent, given the uncertainty in accounting for the detailed exposure history and present-day near surface conditions of concrete. For surfaces that have been weathered or altered by chemicals, bulk activities from surface measurement correlations could be in error by factors approaching 100 in some cases. In facilities that are still active or have only recently been deactivated, personal communications and process knowledge can greatly reduce profile uncertainties and only a few subsurface samples may be needed. However, the older the facility or the more time between characterization studies and deactivation, the greater will be the need to corroborate or establish profiles. In either case, the liberal use of electrets or other surface/subsurface measurement techniques should significantly reduce the chance of missing contaminated areas. Smears should be used only for the most rudimentary types of characterization or for radiological control tasks.

The observations made during this study strongly support the concept that tritium concentration depth profiles in concrete produced from liquid exposure are fundamentally different from those produced from vapor exposure. This opportunity to do further study in an operational facility with a relatively well documented history clarifies questions from the original work [1]. Relatively flat tritium depth profiles invariably arise from any exposure of concrete to tritiated liquids, and additionally this constant concentration profile is established much faster than diffusion theory predicts. A hydraulic flow model can explain the distribution of tritium in liquid exposed concrete, where capillary suction of the concrete simulates a hydraulic head causing the liquid to move quickly into the bulk in all directions without the aid of gravity. Within a few days the wetting front moves far enough into the bulk to encounter water-filled capillaries and rapid liquid-in-liquid diffusion augments the permeation process.

Depending on the tritium concentrations in the exposing liquid, bulk tritium concentrations in the concrete can far exceed those encountered from gaseous exposures. Regardless of the actual mode of exposure (liquid or gas), time, weather, and environmental conditions further alter the original profile. Tritium is present in concrete as relatively free pore and capillary water and as bound water in hydroxides and hydrates. Cycles of drying and wetting of the surface from evaporation and rain, or from large changes in relative humidity, invariably change the near surface moisture content quickly, which then slowly affects the bulk. Because of concrete’s affinity for water, in the absence of hydraulic flow, tritium once taken into the bulk is removed only very slowly, primarily in response to humidity changes at air-exposed surfaces. As a result, the interior bulk of concrete may hold much more tritium than surface or near surface measurements indicate. This must be anticipated in characterization work.

Direct surface activity measurements with devices able to respond to both surface and near surface tritium are superior to surface smears in characterization. Such measures correlate well with the underlying bulk, and contaminated areas are less likely to escape detection. Still, some limited bulk sampling to confirm the tritium profile down to about 15 centimeters is recommended. Using an inexpensive hammer drill, such bulk sampling can be accomplished in minutes and, with an at- or on-site laboratory, results can be available in a day or two if necessary. Data quality are better and costs are but a fraction of traditional D&D methods.

1. R. C. Hochel and E. A. Clark, Tritium Characterization in Cement and Concrete, Report WSRC-TR-99-00081, Rev.0, Westinghouse Savannah River Company, (1999).
2. R. S. Dickson, Tritium Interactions with Steel and Construction Materials in Fusion Devices – A Literature Review, Report AECL-10208, CFFTP-G-9039, Chalk River Laboratories, November, 1990.
3. S. Numata et al, Diffusion of Tritiated Water in Cement Material, Journal of Nuclear Materials 171, pp. 373-380, (1990).
4. S. Numata et al., Diffusion of Tritiated Water Vapor into Concrete, Fusion Technology, Vol. 19, pp.140-145, (1991).
5. J. P. Krasznai, The Radiochemical Characterization of Regular- and High-Density Concrete from a Decommissioned Reactor, Waste Management, Vol. 13, pp. 131-140, (1993).
6. H. Masuzuru et al., Leaching Behavior of Tritium from a Hardened Cement Paste, Annals of Nuclear Energy, Vol. 6, pp. 417-423, (1979).
7. E. Zamorani. Water Corrosion and Release Mechanism of Cement Matrix Incorporating Simulated Medium-Level Waste, Nuclear Technology, Vol. 77, pp. 313-319, (1987).
8. G. G. Eichholz et al. Tritium Permeation Through Concrete, Waste Management, V9, pp. 27-36, (1989).
9. N. S. Martys and C. F. Ferraris. Capillary Transport in Mortars and Concrete, Cement and Concrete Research, Vol. 27, No. 5, pp. 747-760, (1997).
10. S. D. Beyea et al. Magnetic Resonance Imaging and Moisture Content Profiles of Drying Concrete, Cement and Concrete Research, Vol. 28, No. 3, pp. 453-463, (1998).
11. F. M. Lea, The Chemistry of Cement and Concrete, Third Edition, Edward Arnold (Publisher) Ltd., (1970).
12. A. M. Neville, Properties of Concrete, Second Edition, John Wiley and Sons, New York, pp. 1-36, (1973).
13. J. Crank, The Mathematics of Diffusion, Second Edition, Clarendon Press, Oxford England, pp. 9-37, (1975).
14. Materials Science of Concrete III, Editor Jan Skalny, The Am. Ceramic Soc., Westerville, Ohio, p. 51, (1992).
• 14a. Ibid, p. 215.
• 14b. Ibid, pp. 224-225.
15. Encyclopedia of Fluid Mechanics, Volume 1, N. P. Cheremisinoff, Editor, Gulf Publishing Company, Houston, Texas, pp. 84-85, (1986).