Abstract

Displacement of groundwater by relatively hydrogen&endash;poor solvents affects the response of thermal neutron devices by changing the scattering and thermal neutron capture properties of a contaminated aquifer. Since chlorine has a large capture cross&endash;section compared to other elements commonly found in aquifers, the capture phenomenon significantly influences thermal neutron count rates in the presence of chlorinated solvent contamination.

Thermal neutron logging was performed during a controlled release of perchloroethylene (PCE) with tools having short (8 cm) and long (41 cm) source&endash;detector spacings. These results demonstrated that thermal neutron logging successfully detected and monitored immiscible phase chlorinated solvents in the subsurface at residual concentrations. Comparison of the data from the two different tools indicates that the predominate mechanism affecting tool response is the change in macroscopic capture cross&endash;section associated with the presence of solvent&endash;based chlorine.

Laboratory tests were conducted to determine the response of the short spacing neutron moisture probe used in the field experiments (i) to changes in chlorine concentration in a homogeneous environment and (ii) to the spatial distribution of chlorine&endash;rich layers. These tests demonstrated that this device is sensitive to the presence of chlorine in a homogeneous medium to concentrations below 0.1 moles Cl / l and that chlorine&endash;rich layers as thin as 0.2 cm (1/16 in) caused an appreciable tool response. These laboratory results were used to convert the field data into apparent PCE saturation measurements that compare favorably with PCE saturations obtained from core samples.

The results of this combined field and laboratory study conclusively show that this simple, readily available technology can be effectively used to evaluate and monitor chlorinated solvent contamination at residual concentrations under suitable geological conditions.

The contamination of aquifers by immiscible organic liquids is widely recognized as a serious environmental problem. It is common to separate these liquids into two categories according to their densities. Those with specific gravities less than water are referred to as LNAPLs (light non&endash;aqueous phase liquids); DNAPLs (dense non&endash;aqueous phase liquids) have specific gravities greater than water. LNAPL contamination commonly occurs from leaking storage tanks and pipelines; these lighter contaminants are largely confined above the water table due to their buoyancy.

Conversely, DNAPLs will move downward into the saturated zone, making the location and remediation of these contaminants more difficult. Their migration through a typical heterogeneous soil is downward and laterally in response to variations in capillary pressure conditions, displacing water as a non&endash;wetting phase. After a DNAPL has moved through a region, some portion of the contaminant remains as residual and pools. In residual accumulations, the DNAPL is no longer a connected phase and is unable to drain. Residual saturations of DNAPL in sands commonly range between 1% and 15% (Kueper et al., 1993). The DNAPL in pools exists as a connected fluid phase and is capable of movement. A dissolved phase plume is advected away from the pooled and residual DNAPL accumulations by the movement of the groundwater. Chlorinated solvents, such as perchloroethylene (PCE) and trichloroethylene (TCE), form an important group of DNAPLs. PCE (C2Cl4), referred to as tetrachloroethylene by the United States Environmental Protection Agency, is 85.5% chlorine by weight. This solvent has a very low solubility (1100 ppb at 20°C); however, the World Health Organization's acceptable drinking water standard is 30 ppb. Hence, one liter of solvent has the potential to contaminate more than 10⁷ liters of groundwater for a period of many decades.

Since the physical properties of these DNAPLs are generally very different from those of the displaced freshwater, geophysical methods can be useful in their detection and monitoring in the subsurface. To evaluate the usefulness of various geophysical techniques, a controlled release of PCE was performed and monitored over time by means of surface and borehole ground penetrating radar (GPR), surface and surface to borehole resistivity measurements, in situ time domain reflectometry (TDR) and resistivity probes and logging of boreholes with gamma&endash;gamma density, thermal neutron and electromagnetic induction devices (Brewster and Annan, 1994; Redman and Annan, 1992; Sander et al., 1992; Schneider and Greenhouse, 1992).

The thermal neutron logging data obtained during this release are presented in this paper. This field experiment demonstrated that thermal neutron logging can be successfully used in the detection and monitoring of free and residual phase chlorinated DNAPLs in the subsurface; however, the results are qualitative in nature. To quantify these findings, laboratory experiments, also presented in this paper, were conducted to assess the response of a neutron moisture gauge used in the field program to changes in chlorine concentration in a homogeneous environment and to the spatial distribution of chlorine&endash;rich layers. These experimental results showed that the gauge is very sensitive to the presence of chlorine and provide a basis for expressing the borehole data in terms of an apparent PCE saturation. This transformation of the thermal neutron data compares favorably with core information.

The underlying physics governing neutron transport in porous rocks and soils is rather complicated; the reader is referred to Ellis (1987) and Hearst and Nelson (1985) for a more extensive discussion on the subject. The response of thermal neutron devices is chiefly affected by two mechanisms in the presence of chlorinated solvents contamination: moderation of neutrons by hydrogen atoms and thermal neutron capture by the chlorine atoms. Moderation (i.e., the ability of materials to slow down fast neutrons emitted from a source to the lower energy thermal regime) is caused by the interactions between the neutrons and the nuclei in the medium; the primary mechanism is elastic scattering. The ability of a medium to moderate neutrons is characterized by its slowing&endash;down length, which is approximately the root&endash;mean&endash;square (rms) distance that a neutron travels from the source until it reaches the thermal energy level. The average kinetic energy loss by a neutron in an elastic collision with a given nucleus decreases as its atomic weight increases. Hydrogen is, by far, the most efficient element for neutron moderation. Hence, the ability of porous rocks and soils to moderate neutrons is commonly used as an estimate of its hydrogen content. Since hydrogen is present primarily as a component of water in rocks and soils, thermal neutron measurements are usually regarded as an assessment of water content. By replacing hydrogen&endash;rich water with hydrogen&endash;poor PCE, the slowing&endash;down length of the medium increases. This change in slowing&endash;down length influences the count rate (i.e., the number of thermal neutrons detected over a given time period) of a thermal neutron device; the manifestation of this effect for a given device is determined by its source center&endash; detector center separation (Hearst and Carlson, 1994).

In addition, thermal neutrons are removed from the system through the radiative capture mechanism, a process that decreases the count rate measured by all thermal neutron logging devices. The ability of a nucleus of a given element to capture thermal neutrons is expressed in terms of its microscopic capture cross&endash;section, a. To describe the capture of thermal neutrons in bulk materials, the macroscopic capture cross&endash;section, S, is used. This quantity gives the probability that a thermal neutron will be captured per unit length of its travel path. For a medium composed of M elements, the macroscopic capture cross&endash;section is given by where n_m is the number of nuclei of the mth element per cm3 and sm is its microscopic capture cross&endash;section.

The chlorine nucleus has a capture cross section that is large in comparison to other elements commonly found in aquifers. If the predominant chlorine source in an aquifer is a chlorinated solvent, observed changes in the response of thermal neutron devices can be related to the amount of solvent present. For example, consider a silica sand with a given porosity containing varying amounts of water and PCE in the pore space. Using microscopic capture cross&endash;sections for thermal neutrons at 0.025 eV given by Lamarsh (1966), S can be determined as a function of the volume fraction of the pore space filled with PCE (i.e., PCE saturation, S_PCE) for a sand having various porosity values; Figure 1 shows this linear relationship. It can be seen that the macroscopic cross&endash;section of the aquifer grows significantly with increasing PCE saturation.

A controlled injection of 770 liters of industrial grade PCE into a sandy aquifer was geophysically monitored for about 1200 hrs in the summer of 1991. Details of this experiment at CFB Borden (located approximately 100 km north of Toronto, Ontario) are given in Greenhouse et al. (1993) and Brewster et al. (1994) and are briefly summarized below. A portion of a sandy aquifer was isolated in a 9 m by 9 m cell constructed of steel sheet piling with sealable joints driven into a clay aquitard about 3.3 m below ground level. The aquifer is a medium to fine&endash; grained Pleistocene beach sand (Bohla, 1986) having a porosity of approximately 36%. Natural gamma ray logging performed prior to the injection found no clay layers in the aquifer within the cell. Throughout the experiment, the watertable within the cell was maintained at a depth of 15 cm below surface.

Figure 2 shows the cell in plan view. The central injection well is surrounded, at 1 m distance, by four in situ probes. Two of these probes measured electrical resistivity; the two TDR probes sampled the dielectric permittivity. Eight boreholes for subsurface geophysical measurements were uniformly situated on a circle of 3 m radius around the injection well (AT&endash;1 through AT&endash;8); a ninth borehole was placed 1 m from the injection well (AT&endash;9). Thermal neutron logging was performed in all nine boreholes over the duration of the experiment. The boreholes were cased with Schedule 40, 6.4 cm (2.5 in) polyvinyl chloride (PVC) pipe which was sealed at the base. While the casing material contains chlorine, this did not seriously impair the ability of the thermal neutron devices to detect chlorinated solvents in the surrounding aquifer.

Thermal neutron logging was initially performed with a Mineral Logging Systems (MLS) tool with a 3 Ci Am&endash;Be source and a 41 cm (17 in) source center&endash;detector center separation. This tool uses a 3He proportional detector. Continuous logging of the boreholes was performed at a rate of 2.5 m / min. The sampling interval along the borehole was 2.5 cm (1 in); the count rate is expressed in terms of thermal neutrons per second. The theoretical standard deviation in count rate due to source variation is estimated to be less than 5.1% for all samples. Logistical problems and operator safety considerations led to the use of a soil

moisture gauge (model 503 DR Hydroprobe~ manufactured by the Campbell Pacific Nuclear (CPN) Corporation) for thermal neutron logging after 215 hrs. This device has a 50 mCi AmBe source at the base of the probe with a source center&endash;detector center separation of approximately 7.6 cm (3 in). This device also possesses a 3He proportional detector. Given its portability and the relatively low intensity of its neutron source, this instrument is well suited for logging at shallow depths. The sampling interval chosen for logging with this device was 10 cm. A four second count time was used; the count rate is expressed as the equivalent rate for a 16 second count time (the standard output format for this instrument).

The theoretical standard deviation in count rate due to source variation is (4 x count rate)^1/2; this is less than 3.1 % for all samples. The depth reference for each tool is approximately the midpoint of the detector assembly.

Figure 3a and 3b show the thermal neutron measurements taken at various times during the experiment in borehole AT&endash;2 with the MLS and CPN systems, respectively. It can be seen that zones of reduced count rate, which are indicated by arrows on Figures 3a and 3b, develop over time. The depths of a given zone differ by approximately 10 cm between the MLS and CPN systems; this is probably due to the difference in detector&endash;source spacing and the effective sampling depth for the two devices. Comparison with a core (CP&endash;3) taken near borehole AT&endash;2 reveals that the location of PCE in the core correlate with the zones of lower count rate in AT&endash;2; this comparison will be discussed further below. Since both devices exhibit reduced count rates in the presence of PCE, it can be inferred that neutron capture by the chlorine nuclei is the principal mechanism governing the response of these thermal neutron devices to the chlorinated solvent.

The appearance of the deeper PCE contaminated layers with time reflects the progressive downward migration of PCE. In addition, it can be seen that these zones exhibit a general decrease in the amplitude over time. This temporal variation corresponds with the accumulation and drainage of mobile PCE phase in a horizontal pool as it migrates, leaving behind a residual phase. Changes in event reflectivity in the GPR data (Brewster and Annan, 1994) and the response of in situ TDR and resistivity probes (Redman and Annan, 1992; Schneider and Greenhouse, 1992) also exhibit this type of behavior.

Spatial and temporal information about PCE migration in the cell is obtained from the cylindrical cross section through the eight boreholes along the 3 m radius circle around the injection point. Logs from borehole AT&endash;6 showed no zones of reduced count rate throughout the experiment, indicating that this borehole was not intersected by migrating PCE; this interpretation is supported by the GPR data (Brewster and Annan, 1994). Assuming that the thermal neutron transport properties of the uncontaminated aquifer do not significantly vary laterally, the thermal neutron log from borehole AT&endash;6 is used as a reference for determining the location and magnitude of reduced count rate zones on the logs from the other boreholes. Figure 4 shows the cylindrical cross section, unwrapped at AT&endash;6, for 217.5 hrs and 988.5 hrs after the start of the spill; the thermal neutron measurements are expressed in terms of the count rate reduction relative to AT&endash;6.

At 217.5 hrs, it can be seen that a laterally extensive zone of PCE contamination has formed at approximately 1.0 m depth; this is the initial pool that formed during the PCE injection and subsequently drained. Downward migration along the cylindrical cross&endash;section has primarily occurred in the vicinity of AT&endash;2; in addition, a minor zone of PCE contamination is detected in AT&endash;5 at about 1.5 m depth. By 988.5 hrs, significant downward migration of PCE has taken place in the neighborhood of AT&endash;1 and AT&endash;8 on the cylindrical cross&endash;section. Further, the anomaly at AT&endash;5 has grown in magnitude, indicating increased PCE contamination. A possible zone of contamination is detected near the aquifer&endash;aquitard interface at 3.4 m depth in AT&endash;2. The location of the PCE contaminated zones detected along boreholes within the aquifer are consistent with the interpretation of the GPR data by Brewster and Annan (1994) and Brewster et al. (1994)

The field experiment clearly demonstrates the usefulness of thermal neutron logging in detecting the presence of chlorinated solvents; however, these results are qualitative in nature. To obtain further information from this data set, it is necessary to quantify the effects of chlorinated solvent contamination on the behavior of the neutron moisture probe. Analytical expressions for the behavior of neutron moisture probes have only been derived by employing fairly broad assumptions (Gemmell et al., 1966; Ølgaard and Haahr, 1967); further, essential parameters required by these theoretical descriptions are often difficult to determine (Kreft, 1974; Czubek et al., 1983).

To overcome these difficulties, laboratory measurements were undertaken to empirically quantify the response of the CPN moisture gauge used in the field experiment. Silvestri et al. (1991) demonstrated that the response of this gauge can be studied within a small test tank. Our test tanks were fabricated out of cylindrical metal barrels having a diameter of 57 cm and a height of 87 cm. Schedule 40, 6.4 cm (2.5 in) PVC pipe, sealed at the bottom to provide a dry interior, was placed along the vertical axis of each tank to simulate the borehole environment of the field experiment. The outer surfaces of the test tanks were lined with paraffin for added safety. Baseline measurements indicated that boundary effects due to the finite extent of the tanks could be neglected when the probe is located in the central region of the test tank.

The effects of varying chlorine concentration in a homogeneous medium and spatial distribution in chlorine&endash;rich layers were investigated; the interior configuration of the test tanks varied with the experiment type. For the chlorine concentration experiment, the test tank contained a medium having a uniform chlorine concentration. Due to the safety hazards in using chlorinated solvents, the effects of chlorine concentration were determined by using a calcium chloride (CaCl2 ) solution. Initially, concentration data were obtained by filling the test tank with only the CaCl2 solution. The influence of a solid matrix on neutron transport was observed by packing the test tank with a crushed silica sand and repeating the concentration experiment. The sand was packed in a water saturated state to minimize the air trapped in the pore space. From three samples, the porosity of the sand pack was estimated to be39+ 1%.

Thirty measurements, each having a 16 second count time, were performed for each CaCl2 concentration; the average of these data is taken to be the count rate for that chlorine concentration. The estimated standard deviation of an average count rate is >/average value/30; the error in the (average count rate) is less than 0.6% in all cases. The CaCl2 concentration was determined by a mass balance technique, density measurements and/or quantitative chemical analysis; the value used is the average of the concentrations obtained from these methods. The average estimated error in CaCl2 concentration is less than 8%; the estimated error is less than 15% in all cases. The data acquired for the response of the neutron moisture probe to varying chlorine concentration in the solution and solution saturated sand are shown in Figure 5a. The range of total chlorine concentrations obtained in the solution saturated sand is equivalent to PCE saturations from 0.0% to 7.2%. It can be seen that the count rate significantly decreases with increasing chlorine concentration and that relatively low chlorine concentrations produce a significant count rate reduction. The rate of change is greater at lower chlorine concentrations, indicating that the technique is more sensitive to variations in chlorine content in this range.

The relationship between total chlorine concentration and count rate for these two cases are not simple multiples of one another. However, it was discovered that similarities can be found between the two data sets if the following quantity is considered. The normalized count rate reduction at a given total chlorine concentration, Q(Cl), is defined as

where n(Cl) and n(O) are the count rates for the given chlorine concentration and for the chlorine free system, respectively. Figure 5b shows Q(Cl) for the solution and solution saturated sand plotted with respect to the logarithm of total chlorine concentration. At higher concentration levels, a linear trend can be seen in both data sets; a least squares fit to the linear trends found that the slopes are almost equal (i.e., 0.174 for the sand mixture and 0.179 for the solution). Since the range of equivalent PCE saturation for the sand mixture is restricted to relatively low values, we propose to use this linear trend to extrapolate to higher total chlorine concentrations not obtained in the laboratory work.

The effects of spatial chlorine variation were observed by suspending assemblages of chlorine&endash;rich plates in a water&endash;filled tank. PVC plates were used because of their high chlorine concentration; using a general composition for rigid PVC (Gobstein, 1990), the chlorine concentration was estimated to be 20 moles / liter. The probe was kept stationary within the central region of the tank while the plate assemblage was moved, thus simulating the motion of the probe passing chlorine&endash;rich layers in the subsurface. The plate assemblages were moved in 1 cm increments. The probe position is the location of the detector center relative to the plate assemblage center; positive values indicate that the probe is located above the assemblage center. A single measurement having a 16 second count time was obtained at each plate assemblage position. The estimated standard deviation of a measurement is (sample value)^1/2; this is less than 1.7% for all measurements. The effects of spatially varying chlorine concentration can be divided into two distinctive types of problems: the detection of a single layer and the resolution of two closely spaced layers. Therefore, two different types of plate assemblages were used. For the detection problem, PVC plates were stacked to form a single layer; for the resolution problem, two thin plates were separated by varying distances.

The response of the CPN moisture gauge to a single chlorine&endash;rich layer in a water background was determined for layer thicknesses between 0.2 cm (1/16 in) to 16.5 cm (6 1/2 in). Figure 6 shows the CPN signature for selected layer thicknesses. The CPN signature for a given layer thickness is characterized by its maximum count rate reduction from the background level of 9SOO counts / 16 s, to be called the signature amplitude, and the signature width at one half maximum amplitude.

The moisture gauge is very sensitive to the presence of a thin chlorine&endash;rich layer; the signature amplitude for a 0.2 cm thick layer is 827 counts / 16 s, a count rate reduction of approximately 9% from the background level. As layer thickness initially increases, the signature amplitude increases rapidly, while half width expands slowly. This behavior suggests the device is primarily responding to an apparent increase in chlorine concentration within some elemental sampling volume. Further, it can be estimated that the impulse response, the limiting case of an infinitely thin layer with a unit mass of chlorine, has a half width of approximately 9.5 cm.

As layer thickness continues to increase, it will eventually exceed the effective sampling volume of the CPN gauge. In this case, the gauge will predominately sample the layer's uniform chlorine&endash;rich environment over some interval of probe positions. Hence, signature amplitude should asymptotically approach some limiting value with increasing layer thickness. From the experimental data, the signature amplitude flattens appreciably for layer thicknesses exceeding the length of the source&endash;detector assembly (12.7 cm). Further, it can be observed that the signature for the 16.5 cm layer in Figure 6 is developing a region of relatively uniform readings about its apex (Unfortunately, the test tank design made it very difficult to use layers thicker than 16.5 cm; thus, it can only be conjectured that the lower limits of the asymptotic behavior have been encountered.). In addition, it was found that half width is a good estimate of layer thickness plus 3 cm when the layer thickness surpasses 13 cm.

The ability to resolve closely spaced chlorine&endash;rich layers was examined by using an assemblage consisting of two 0.6 cm (1/4 in) thick PVC plates that are separated by varying distances; layer separation is specified by the distance between the inner surfaces of the two plates. Figure 7 shows the observed tool response as the plate separation increases from 2.5 cm (1 in) to 12.7 cm (5 in). For layer separations of less than 5.1 cm, the response of the two layer system is not easily distinguished from the response of a single layer of 1.3 cm (0.5 in ) thickness (i.e., the zero separation case). This indicates the device is primarily responding to an apparent constant chlorine concentration within an elemental sampling volume, which is consistent with the behavior observed in the layer thickness test. For layer separations of 12.7 cm and greater, the signature of the individual layers are visually discernible. At intermediate separation distances, the response of the two layers interact in a rather complex manner; a simple convolutional model cannot be used to explain the behavior.

Re&endash;examination of the Field Results

The field results will now be reconsidered in light of the laboratory work. It is apparent that the spatial sampling of the CPN field data at a 10 cm interval makes it difficult to perform a detailed analysis in terms of layer thicknesses and signature amplitudes or to resolve closely spaced thin layers. However, given the estimated 9.5 cm half width of the impulse response, the CPN logging of the boreholes should have detected all zones of PCE contamination. To translate the field data into a more understandable format, the chlorine concentration data for the sand&endash; solution mixture are used to convert the field readings into an apparent PCE saturation for an equivalent homogenous environment.

Figure 8 shows the relationship between the normalized count rate reduction Q, as defined by equation (2), and the equivalent PCE saturation for a sand having 39% porosity corresponding to the chlorine concentrations in the sand&endash;solution laboratory data. Curve A is the third order polynomial

S_PCE =0.784Q³ -0.0169Q² +.133Q+.00019 (3)

determined by a least squares analysis of the data. The curve B is the counterpart to the linear extrapolation to higher chlorine concentration in Figure 5b and is given by

S_PCE = 0.0106e^{5. 74Q}. (4)

Relationship A is used to convert Q>0.3 to an equivalent S_PCE; relationship B is used to convert Q > 0.3

Let us first examine the CPN measurements from borehole AT&endash;2 at 988.5 hrs. At about 1000 hrs, core CP&endash;3 was taken from a location within 0.5 meters of AT&endash;2. PCE saturations in 5 cm long samples along the core were determined by spectrophotometer and gas chromatograph methods; the resulting S_PCE values are given in Figure 9. Considering the subtle lateral variations in aquifer properties that can affect PCE migration, the zones of PCE contamination in the core correlate well with the zones of reduced count rate at AT&endash;2.

Using the data from AT&endash;6 for the count rate in the uncontaminated aquifer, T was determined as function of depth for the thermal neutron log from AT&endash;2. An apparent S_PCE log was generated using equations 3 and 4; this log is shown in Figure 9. Apparent S_PCE less than about 1% appear to be within the uncertainty introduced by measurement error and variation in the physical properties of the aquifer between AT&endash;2 and AT&endash;6. In an attempt tomake the core data more compatible with the CPN measurements, the core data at a given depth were averaged with its two adjacent values to obtain an effective S_PCE over a 15 cm interval, a more appropriate length scale compared to the effective sampling volume of the CPN device. These values are superimposed on the apparent S_PCE log in Figure 9. Given that an analog system employing CaCl2 and a sand with a different porosity (39% versus 36% for the Borden aquifer) was used to estimate the relationship between the response of the CPN device and S_PCE, the apparent S_PCE log and effective S_PCE compare favorably; this suggests that useful information is contained in an apparent S_PCE obtained in this manner.

We will now reconsider the spatial distribution of PCE at 217.5 hrs and 988.5 hrs original shown in Figure 4. The apparent S_PCE was determined from these data sets assuming that AT&endash;6 is the profile through the uncontaminated aquifer; the results are shown in Figure 10. Since the CPN device is more sensitive to changes in chlorine content at low concentrations, the conversion to apparent S_PCE surpresses the disproportionately higher count rate reduction due to relative low levels of contamination. The spatial changes in the PCE distribution that were previously noted are again apparent; the drainage from earlier pools is somewhat better illustrated in terms of apparent S_PCE.

Discussion and Conclusions

Results from a controlled PCE spill clearly demonstrate the usefulness of thermal neutron logging in the detection and monitoring of chlorinated solvents in aquifers. Zones of reduced count rate in thermal neutron logs obtained with long and short source&endash;detector spacing tools correspond with PCE contamination found in a nearby core. Spatial and temporal information show the progressive downward PCE migration in a water saturated aquifer, as well as the accumulation and drainage to residual levels of the PCE phase in horizontal pools.

Laboratory testing of the CPN moisture gauge was performed in small test tanks to quantify its response to the presence of chlorine. The relationship between count rate and chlorine concentration in a uniform environment was obtained by using CaCl2 solutions of varying strength. The CPN device responds to relatively small amounts of chlorine; however, its sensitivity to chlorine content variations decreases as concentration increases. The effects of spatial variations in chlorine content were determined by using assemblages of chlorinerich PVC plates in a water background. A layer as thin as 0.2 cm (1/16 in) was found to be detectable. For plate assemblage less than 5 cm wide, the CPN device responds primarily to the apparent chlorine content within an elemental sampling volume. As layer thickness increases, the signature amplitude asymptotically approaches the value for a uniform environment composed of the layer material and the signature half width can be used to determine layer thickness. Closely spaced thin layers cannot be visually resolved until their separation is on the order of the source&endash;detector separation.

Using the concentration information from the laboratory experiment, a relationship between count rate reduction and apparent PCE saturation was established. When applied to the field data, the results compared favorably with core data. While a preferable method for quantifying field measurements would use mathematical formulations based on the underlying physics, it appears that expressions based on reasonable analog systems used in laboratory calibration can render useful information when such formulations are not easily obtained. The major drawback of this approach is the difficulty in generalizing such laboratory results; conditions in the field can vary significantly from the chosen analog systems. This consideration includes not only the geological media, but also the borehole environment (Keller et al., 1990). Casing materials can contain elements with large thermal neutron capture cross sections, such as chlorine in PVC and boron in certain glasses, that can alter the response of a thermal neutron device.

Results from the plate assemblage experiments raise the question of the proper spatial sampling when using the CPN device to detect zones of chlorinated solvent contamination. The 10 cm interval used in the field experiment seems too large in light of the laboratory results. We suggest that a minimum 5 cm sampling interval is required to obtain aufficient structure in the thermal neutron depth profile; however, a 2.5 cm interval would be more desirable.

The response of the long and short spacing tools indicates that radiative capture of the thermal neutrons is the predominate mechanism in detecting the presence of chlorinated solvents. While either device used in this experiment is clearly very sensitive to chlorine, their response can be ambiguous in the absence of other data. A possible alternative approach is to use a dual detector thermal neutron device so that the macroscopic capture cross&endash;section of the geological material can be estimated directly. However, when used in conjunction with each other or with other geophysical borehole measurements, thermal neutron logging can be a valuable aid in detecting and monitoring chlorinated solvents in the subsurface. It should be noted that the effectiveness of thermal neutron logging would be seriously imparted when groundwater contains chlorine from other sources since the contrast in chlorine content between contaminated and clean zones is decreased.

In the Borden field experiment, we found that the resistivity and TDR in situ probes were also able to detect the presence of PCE in the aquifer. Neutron moisture probes, such as the CPN 503DR Hydroprobe, offer several advantages over these methods. The gauges are available off the shelf, as opposed to the resistivity and TDR equipment which is custom built. The resistivity and TDR probes require complicated installation; the neutron method needs only a standard borehole. The thermal neutron method responds directly to the presence of a chlorinated solvent, whereas electrical and electromagnetic method detect changes in water content. The neutron method has a sensitivity to chlorinated solvents comparable to other geophysical methods The main impediments to the thermal neutron approach are the regulations governing the use of nuclear materials in many jurisdictions and the perceived hazards (real and imaginary) associated with nuclear devices.

Acknowledgments

This project was funded by the Solvents in Groundwater Project at the University of Waterloo and the Waterloo Centre for Groundwater Research. A. L. Endres was supported by a postdoctoral fellowship from the Natural Sciences and Engineering Research Council (NSERC) of Canada during this work.

References

Bohla, J., 1986, MSc. Thesis, University of Waterloo.

Brewster, M. L., Annan, A. P., Greenhouse, J. P., Kueper, B. H., Olhoeft, G. R., Redman, J. D., and Sander, K., 1995, Ground Water, in press.

Brewster, M. L., and Annan, A. P., 1994, Geophysics, 59, 1211.

Czubek, J. A., Drozdowicz, K., Krynicka&endash;Drozdowicz, E., Igielski, A., and Woznicka, U., 1983, International Journal of Applied Radiation and Isotopes, 34, 143.

Ellis, D. V., 1987, Well Logging for Earth Scientist, Elsevier, New York, N. Y., 227&endash;280.

Gemmell, W., McGregor, B., and Moss, G. F., 1966, International Journal of Applied Radiation and Isotopes, 17, 615.

Gobstein, S., 1990, Handbook of Plastic Materials and Technology, I. I. Rubins (ed.), WileyInterscience, New York, N. Y., 549&endash;566.

Greenhouse, J., Brewster, M., Schneider, G., Redman, D., Annan, P., Olhoeft, G., Lucius, J., Sander, K., and Mazzella, A., 1993, The Leading Edge, 12, n. 4, 261.

Hearst, J. R., and Carlson, R. C., 1994, Nuclear Geophysics, 8, 165.

Hearst, J. R. and Nelson, P. H., 1985, Well Logging for Physical Properties, McGraw&endash;Hill, New York, 240&endash;263.

Keller, B. R., Everett, L. G., and Marks, R. J., 1990, Ground Water Monitoring Review, Winter 1990, 96.

Kreft, A., 1974, Nukleonika, 19, 145.

Kueper, B. H., Redman, J. D., Starr, R. C., Reitsma, S., and Mah, M., 1993, Ground Water, 31, 756.

Lamarsh, J. R., 1966, Introduction to Nuclear Reactor Theory, Addison&endash;Wesley, Reading Mass., 558&endash;561.

01gaard, P. L. and Haahr, V., 1967, Nuclear Engineering and Design, 5, 311.

Redman, J. D., and Annan, A. P., 1992, Proceedings of the Fourth International Conference on Ground Penetrating Radar, P. Hanninen and S. Autio (eds.), 191

Sanders, K. A., Olhoeft, G. R., and Lucius, J. E., 1992, Proceeding of the Symposium on the Application of Geophysics to Engineering and Environmental Problems, 455.

Schneider, G. W., and Greenhouse, J. P., 1992, Proceeding of the Symposium on the Application of Geophysics to Engineering and Environmental Problems, 619.

Silvestri, V., Sarkis, G., Bekkouche, N., Soulie, M., and Tabib, C., 1991, Geotechnical Testing Journal, 14, 64.

Figures

Figure 1: Macroscopic capture cross&endash;section as a function of PCE saturation for sandwater&endash;PCE system. Porosity values are denoted for each model sand&endash;water&endash;PCE system.

Figure 2: Plan view of 9 m by 9 m cell used in field experiment. Boreholes in which neutron logging was performed&endash; . Other features in cell: Injection point&endash;Of Resistivity probes&endash;n; TDR probe&endash;A and GPR profiles&endash;dashed lines.

Figure 3: Thermal neutron logs taken at various times in borehole AT&endash;2 during field experiment using (a) MLS system and (b) CPN moisture gauge. Arrows indicate zones of reduced count rate. Time given for log is in hours after start of PCE release.

Figure 4: Cylindrical cross&endash;sections along 3 m circle of differenced thermal neutron logs for 217.5 hrs and 988.5 hrs. after start of PCE release.

Figure 5: Response of CPN moisture gauge as a function of chlorine concentration in a uniform environment: (a) observed count rate and (b) normalized count rate reduction T(Cl) Lines give least squares best fits to the higher concentration data.

Figure 6: CPN moisture gauge response to a single chlorine&endash;rich layer in a water background as a function of probe location relative to the layer center. Layer thicknesses are denoted above each graph.

Figure 8: Relationship between PCE saturation and normalized count rate reduction T for a sand with 39% porosity inferred from CACl2 experimental data. Curves A and B correspond to equations 3 and 4 in text, respectively.

Figure 9: Comparison of PCE saturation determined from core CP&endash;3 and the apparent PCE saturation computed from thermal neutron log taken at 988.5 hrs in borehole AT&endash;2. CP&endash;3 data superimposed on AT&endash;2 apparent PCE saturation has been averaged over a 15 cm interval for comparison purposes.

Figure 10: Cylindrical cross&endash;sections along 3 m circle of apparent PCE saturation computed from thermal neutron logs for 217.5 hrs and 988.5 hrs. after start of PCE release.