Dense non-aqueous phase liquid (DNAPL) source zones are one of the most complex and difficult types of contaminated sites to remediate. Sites characterized by low permeability and high subsurface heterogeneity are particularly challenging. To overcome these limitations, thermal conduction heating (TCH) is gaining attention for usage even in heterogeneous subsurface settings with layers of low permeability and for contaminants with moderate to high boiling points. Thermal wells contain electrical heating elements operating at 400-700°C and can heat both high- and low-permeability media by a combination of TCH and thermally induced convection processes. The vaporized contaminants are mainly extracted by soil vapor extraction (SVE). The combination of thermal wells and SVE is also known as in situ thermal desorption.

The objectives of this project were to (1) determine the relative significance of the various contaminant removal mechanisms below the water table (e.g., steam generation, steam stripping, volatilization), (2) assess the DNAPL source removal efficiency and accompanying change in water saturation at various treatment temperatures/durations through boiling, and (3) evaluate the potential for DNAPL mobilization, either through volatilization and recondensation, and/or pool mobilization outside of the target treatment zone (TTZ) during heating.

Technical Approach

Bench- and larger-scale remediation experiments were conducted. In parallel, MK Tech Solutions’ simulation modeling was optimized based on the experiments, enabling numerical simulations to be used to design the 3-D large-scale experiments. Several quasi 2-D experiments were conducted, examining heat propagation in various soil types. In addition, during an initial round of 3-D heat-transfer experiments in large containers (up to 150 m³), the progression of heating to 100°C and accompanying desaturation were monitored using 300 temperature sensors and 35 time-domain reflectometry probes, respectively, allowing comparisons of the physical experiments and accompanying numerical simulations. Remediation experiments were then conducted in the 2-D and 3-D large-scale containers, with controlled release of DNAPL into lower permeability layers simulating aquitards beneath the water table.


During this research, dominating processes, those of more minor importance, and even small-scale effects could be identified. The process understanding of the application of TCH in the saturated zone was significantly improved. As occurs with application of TCH for remediation of volatile organic compounds (VOCs) in the unsaturated zone, steam distillation is the major dominating process. It is helpful that the process of steam distillation (co-boiling of water and NAPL) first occurs at temperatures below 100°C, e.g., below the boiling point of either pure water or pure tetrachloroethene (PCE). Consequently, contaminants present as a separate phase have already vaporized by the time that steam production due to boiling of water in the initially water-saturated aquitard layer begins. Nevertheless, an efficient NAPL recovery from the saturated zone by an overlying SVE system requires not only vaporization of the NAPL but also development of one or more contiguous flow paths via which the gaseous phase can travel towards and reach the unsaturated zone and the SVE recovery wells. It follows that steaming of the water within the aquitard in the given large 3-D physical models, which occurs at about 100°C (depending on pressure), is important in enabling the desired migration of gaseous NAPL constituents into and through the unsaturated zone.

The generation of steam at a given location during the experiments was significantly affected by the energy intensity (a function of heater power and spacing), soil permeability, depth below the groundwater level, and the cooling effect caused by the groundwater flux. Furthermore, the extent of lateral versus vertical steam zone propagation is mainly affected by: (a) site-specific conditions such as anisotropy, presence of preferential flow paths, and capillary pressure - saturation relations; (b) heater system design factors such as aspect ratio (i.e., the ratio of heater length to heater spacing), the number of heaters, and their lateral and vertical positioning relative to the TTZ, and (c) vapor recovery system design factors such as the position of SVE wells and specifically their filter screens, negative pressure exerted, and recovery rate (especially for filter screens in the former saturated zone). The recovery of contaminant vapors emanating from a previously saturated aquitard layer through placement of SVE wells only in the unsaturated zone is possible. For field applications, extension of the SVE recovery system into the lower permeability layer (LPL) (aquitard) is often recommended to accelerate the remediation process, even though energy losses from the LPL will be higher.

Condensation of vaporized contaminants at the steam front results in longer remediation times. To minimize formation of this condensate, the heater array must fully encompass and surround the TTZ. This requires adequate site characterization prior to design. A certain amount of condensate is transported by the so-called heat-pipe effect as liquid phase back into the steamed zone.

For a given level of heater power, the higher the soil permeability, the greater will be the cooling effect of groundwater flux, and thus achieved temperatures will be lower at steady state conditions. For the given power and spacing of the heater wells relative to the position of the NAPL, a high groundwater flux limits the lower expansion of the steamed zone, indicating that NAPL located close to the boundary between water-saturated aquitard and aquifer may not be recovered without implementation of combined remedies.


The experimental results indicate that advanced numerical simulations would be helpful to predict steam front propagation, temperature distribution, and DNAPL recovery. The accuracy of input data such as geological and hydrogeological information, soil parameters, and contaminant distribution, however, affect simulation results. In some respects, predictions gleaned from numerical simulations proved of value to process understanding; in other respects, they differed from the experimental results.

Advanced simulations are needed to answer very specific questions in detail. Lack of basic knowledge, such as the impact of temperature on three-phase parameters, prevents numerical results from being of sufficient reliability. Further research is needed to improve the results by application of some (small-scale) effects of interest. It was found that the model design itself can affect the overall results of the simulations.

To facilitate future simulations, it is useful to bear in mind that many flow, transport, and remediation effects can be anticipated with no more than an accurate prediction of the temperature distribution. Future research should focus on the quantification of those factors that affect heat transport in order to improve the reliability of numerical simulations.