Summary Electrically conductive concrete is usually produced by mixing conductive materials into the wet concrete to decrease its resistivity. A voltage applied across the concrete causes the temperature of the material to increase by Joule heating. Although many applications of such a material are possible, this study focused on its use in highway anti-icing. A conductive concrete overlay could be used to heat the roadway surface, thus preventing an ice-road surface bond without the use of chemical deicers. This technique is particularly valuable for highway sections in which the surrounding environment is sensitive to an increase in salt. In this study, the effect of welded wire fabric on the heating ability of conductive concrete was examined. The height of the fabric and the placement of electrodes were varied to find the optimal arrangement for surface heating. Introduction To fully remove snow and ice from highways, both mechanical and chemical methods are generally required. Although snow plows can move the bulk of the mass off the road surface, a bond exists between the pavement and an underlying layer of ice that is difficult to remove by plow alone. Chemical deicers or salts such as NaCl (most common), CaCl, and MgCl break this bond by lowering the temperature at which water freezes, melting some of the ice and allowing plows and traffic to move the remaining slush off the road. These salts are effective and inexpensive, which are very appealing qualities to departments of transportation. However, excessive use of salt can have devastating effects on the adjacent environment, particularly lakes and ponds (Figure 1). In the Upper and Lower Cascade Lakes in Keene, NY for example, chloride levels are 100 to 150 times greater than that of a comparable lake (Langen et al, 2006). Figure 1. Section of Route 73 adjacent to Upper Cascade Lake in Keene, NY. Road salt damages plants by disturbing the water balance in the soil, leading to dehydration. The excess sodium cations introduced to the soil by heavy use of NaCl reduce soil fertility by forcing out other essential ions such as calcium and magnesium and can promote erosion. Chloride can bind with toxic heavy metals, forming complexes that are more easily absorbed by plants. Sufficient levels of chloride concentration in lakes and ponds can lead to meromixis, a state in which the body of water remains stratified and does not “turn over” in the spring and fall. This condition leads to a concentration of nutrients and an absence of oxygen in the bottom portion of the lake. (Langen et al, 2006) One possible solution to this problem is the use of electrically conductive concrete instead of deicing chemicals on roads near chemically sensitive areas. This material is produced by mixing conductive aggregate or fibers into concrete. The relatively small amount of conductive material renders the entire sample conductive by the connections between each piece, which form a current-carrying network. When a voltage is applied across the concrete, the resistivity of the sample causes it to warm through Joule heating. The heat emitted by the concrete would fill the same role as the salt, breaking the bond between ice and pavement that prevents a clear, safe highway. The concrete can also be used as an anti-icer if it begins to heat before the storm event, thus melting all precipitation as it hits the road. Previous Research A substantial body of research has been developed in recent years in electrically conductive concrete aimed particularly at its possible deicing and anti-icing applications (Yehia, Tuan, et al, 2000; Tuan, 2004; Green and Janoyan, 2005; Vincent and Janoyan, 2005). The optimal proportions of steel shavings and steel reinforcing fibers has been studied extensively and found to be 15 – 20% shavings and 1.5% fiber by volume, optimized for electrical conductivity (heating performance) and strength (Yehia, Tuan, et al, 2000). These proportions created a concrete that sufficiently lowered the resistivity of the material (less than 10 Ω∙m) while maintaining workability, surface finish, and necessary mechanical properties with a heating efficiency of approximately 88%. This study also found that the use of shavings or fibers alone did not create sufficiently conductive concrete. The Roca Spur Bridge in Nebraska was completed in 2002 with a conductive concrete overlay (Tuan, 2004b). The bridge demonstrated the effectiveness of conductive concrete in preventing snow and ice accumulation on a bridge surface (Figure 2). Figure 2. Roca Spur Bridge in use during February 2004. Image credit: http://www.cement.org/tech/cct_con_design_conductive.asp Further research has been done to investigate the use of materials in conductive concrete that are less expensive and easier to implement. A comparison of steel fiber reinforced concrete, wire mesh, and nickel wires found that mesh cut in the middle, suspended in fiber reinforced concrete, warmed the surface best (Vincent and Janoyan, 2005). It was also found that the samples because hottest where the separate sides of mesh were closest. This study aimed to continue investigating methods of creating electrically conductive concrete using materials and methods that departments of transportation already use, namely welded wire fabric and steel fibers. Although steel shavings can be used to create conductive concrete with excellent mechanical and electrical properties, this material is difficult to use in bulk quantities due to the cleaning necessary to remove contaminants that might compromise the properties of the final product (Green and Janoyan, 2005). Welded wire fabric is commonly used by many departments of transportation to reinforce roadway surfaces. Thus, understanding its effect on the heating ability of conductive concrete is vital if this technology is to be accepted by departments of transportation for use on highways. Materials Four 15x15x2” resistivity/heating samples were tested, incorporating two welded wire fabric heights and two concrete mixes. In one set of samples, the 12x15” sections of 4x4W4xW4 mesh were placed above the thinnest layer of concrete possible, about ½” above the base of the mold. The higher one was placed approximately in the center of the mold, at about 1”. The mesh was originally covered in a layer of rust, which was removed using wire brushes. In both sets of samples, the mesh was positioned so it was not in direct electrical contact with the electrodes (Figure 3). One concrete mix contained 0.25% intact Dramix 80 steel fibers by volume, while the other contained 0.5% fibers cut in half (Appendix A). These low percentages were used due to the poor workability and finish of concrete incorporating large amounts of steel fiber. Figure 3. Mesh placement (gray) in relation to electrodes (black). Three 6x12” compression cylinders were poured for each mix. Concrete molds were constructed from ¾” plywood, connected with 1 ¼” drywall screws. Steel strips (3/16” x 3/4”) were used to create the electrodes. They were drilled with ¼” holes every inch to facilitate bonding with the concrete. On each sample, four 10” and four 4” electrodes were attached to the concrete by connecting the steel to the side of the mold with ¼ x 5 ½” bolts while curing (Figure 4). The bolts were used as electrical contacts during testing. Multiple electrodes were used to allow for investigating the effects of electrode placement. Thermistors were installed using ¼” wooden dowels to record the temperature of the concrete at the surface and at 1” depth at nine locations, as well as ambient temperature (Figure 4). Figure 4. Concrete mold with electrodes and thermistor temperature probes. The steel bars bolted to the sides of the mold are electrodes. Inside the mold, each dowel is attached to one or two thermistors. Procedure Sample Constuction Concrete samples were poured in the structures lab at Clarkson University on June 29, 2006. Both mixes were prepared by first wetting the 1/3 yd3 drum and adding the dry components, then adding the premixed water and admixture. The drum was not sufficiently dampened before mixing the full-length fiber sample, which resulted in a large amount of dry material adhering to the sides of the drum. A scoop and rod were used to dislodge the material during a ten minute period of short bursts of mixing and resting. The more evenly distributed sample was then mixed for a final three minutes. The slump was initially found to be 8”, which was very different from the expected 1-3” range. Upon closer inspection, the concrete appeared to be not yet uniformly mixed. The concrete was therefore mixed further by hand for approximately five minutes and retested with a slump of 3”. To avoid the problems of the previous mix, the half-length fiber sample was mixed in two standard cycles of three minutes mixing, three minutes rest, and three minutes final mixing. When poured out for a final hand mixing, the concrete appeared to have a much higher slump than expected. The concrete was allowed to sit, covered, for fifteen minutes to allow it to set. After this time the concrete still did not appear sufficiently stiff, so the true water/cement ratio of the mix was examined and found to be .39, higher than the design .35. More cement was added to bring this ratio to its proper value. The slump was tested and initially found to be 8”. After waiting five minutes for further setting of the recently introduced cement, the slump was found to be 5.5”. The compression cylinders were poured in accordance with AASHTO T 126-93, in three layers that were rodded 25 times each. It was observed that the half-length fibers allowed for a better initial surface finish than the full-length fibers. Forms in which the half-length fibers were used were much easier to strike off. In both mixes the heating and resistance molds were difficult to strike off due to the large number of probes. After pouring, all samples were placed in the curing room, where they were kept damp by an automatic watering system. After seven days the samples were removed from the molds and replaced in the curing room. One 4” electrode was found to have not bonded to the concrete on the sample using the half length fiber mix with the lower mesh height. The following day the samples were relocated to the testing location. In the process, a 4” and a 10” electrode was removed from the sample using the half length mix with the upper mesh height. Later, another 4” electrode was observed to be loose on the same sample. The protruding ends of the thermistor supports were removed with a hacksaw, which provoked a 4” electrode on the full length upper mesh location sample to detach. The samples were covered with wet burlap. However, after three days the burlap was removed and silver-filled conductive epoxy used to reattach the electrodes to the samples. Sample Testing All thermistors were tested prior to concrete pouring and found to be functioning properly. Initial tests showed the thermistors to be severely damaged by the curing process, as they gave disparate and unreasonable temperature readings whether or not there was a voltage across the sample (Appendix B, Figures 1, 2), although they were still able to respond to extreme external heating (Appendix B, Figure 3). New surface temperature probes were constructed and attached to samples using duct tape followed by a layer of caulk (Figure 5). The duct tape served as a waterproof layer, while the caulk provided thermal insulation. These samples were tested by using a BK Precision 1651A power supply to apply 12 VDC across the sample and measuring temperatures at the surface as recorded by the thermistors. Data for the heating tests was collected automatically using an HP (Agilent) 3852A data acquisition system. Figure 5. Sample with partially attached new thermistors. The plywood visible on the sides of the sample is being used as a clamp for the electrodes as the epoxy cures. The resististance values of the samples were found using the four wire method. The power supply was used to apply 12 V across the sample, verified by a multimeter (true voltage for all tests was 12.3 VDC). A multimeter was also used to measure the current passing through the sample. All samples were tested for resistivity and heating ability multiple times, using different electrode arrangements. The arrangements tested were using all electrodes, using all top electrodes, using all bottom electrodes, using half top and half bottom electrodes, using the diagonal of 4” electrodes on top, and using the diagonal on the bottom (Figure 6). In the halflength fiber samples, not all tests were able to be performed since many of the detached electrodes did not sufficiently bond to the concrete through epoxy. All Electrodes Half Upper, Half Lower Upper Electrodes Lower Electrodes Upper Diagonal Lower Diagonal Figure 6. Electrode arrangements used during testing. Results The compressive strengths of the concrete are suitable for highway applications (Table 1). Departments of transportation generally require a minimum strength of 3000 psi for concrete used in road overlays (Green and Janoyan, 2005). As expected, the mix containing full-length fibers had a higher strength than the half-length mix. Without considering the effect of the steel fibers, the projected strength for these mixes was 6000 psi from the water/cement ratio of 0.35. Higher values as were observed were expected due to the reinforcing effect of the fibers. Full-length mix Half-length mix Cylinder 1 8751 7310 Cylinder 2 9154 7640 Cylinder 3 8238 7431 Mean 8714 7460 Table 1. Compressive strengths of concrete mixes in psi. The resistivity data suggest that none of the samples would be able to function as highway anti-icers or deicers (Table 2). In all samples and in all trials the resistivity was lowest when all electrodes received the applied voltage and highest when only one electrode on each side received the applied voltage (Appendix C). This observation is not surprising, since more electrodes imply a greater number of current paths. While low, these values were still not low enough to fall below 10 Ω∙m, the value determined by Yehia and Tuan (2000) to be the maximum resistivity for functional deicing or anti-icing conductive concrete. The median resistivity of each sample also increased with time in almost all cases (Table 3). This suggests that the heating ability of the concrete would become still worse over time. An increase in the resistivity of conductive concrete over time was also observed by Yehia and Tuan, which they explained as resulting from the progression of hydration and predicted would eventually reach a constant value (2000). It was observed that during resistivity tests the ammeter readings began at a high current and eventually decreased to a steady value. This behavior suggests that the effective resistivity of the concrete increases as current flows through it, although it eventually stabilizes. Full length fibers Upper mesh Half length fibers Upper mesh Lower mesh Ω∙m 27.3 Ω∙m 27.9 Ω∙m 16.8 Ω∙m 32.5 Lower mesh Table 2. Lowest resistivity measurements of all samples. The resistivity values given are the mean of all resistivity measurements using the all electrode arrangement. 13 days 17 days 22 days Upper mesh 65.9 46.6 72.9 Full length fibers Lower mesh 52.2 104.6 114.9 Upper mesh 25.2 34.1 51.0 Half length fibers Lower mesh 19.4 29.2 45.9 Table 3. Median resistivity values, in Ω∙m, of all possible arrangements of electrodes for each sample at different times after curing. Heating tests indicate that applying a voltage across the samples does not heat them considerably (Figure 7). The surface thermistors appear to register a delayed, muted version of the temperature variations recorded by the ambient probe. In no trial did the temperature in one location vary by more than 1˚ F (Appendix B, Figures 4 - 24). These observations suggest that the samples cannot function as anti-icers or deicers. Figure 7. Temperature variations of surface and ambient probes during heating test. Although all loose rust was removed from the welded wire reinforcement before placement in the samples, a fine layer that could not be removed by wire brush remained. Although very thin, this layer was found to effectively electrically insulate the mesh. When electrodes were connected to the slightly rusted area, generally no measurable current passed through the wire fabric, although in certain contact locations the resistance could be found to be 2.04 MΩ. However, when electrodes contacted the unrusted areas where the mesh had been cut, the resistance was approximately 1 Ω. These observations suggest that any effect the mesh has on the resistivity of the sample is only through contact with the cut ends of the fabric, a very small surface area. Since it is unlikely that departments of transportation use welded wire fabric completely devoid of rust, similar resistance values are likely to be observed in a large-scale application. Despite the small surface area of usable contacts on the mesh, the results suggest that the fabric did decrease the resistivity of the concrete. The electric resistivity of concrete containing 2% steel fibers has been found to be 0.54 M Ω∙m (Yehia, Tuan, et al, 2000). Although resistivity measurements of the fiber reinforced concrete alone were not taken in this study, it is reasonable to assume that the concrete mixes used in this study would have higher resistivity values due to the lower concentration of steel fibers. However, all observed resistivity values observed were much lower than this value, suggesting that the addition of welded wire fabric substantially decreased the resistivity of the sample (Appendix C). It is important to note that previous research has shown samples with higher resistivity values exhibit greater heating. In one trial a conductive concrete sample having a resistivity of approximately 20 Ω∙m experienced a temperature increase of 21.3˚ F over 5000 seconds in one location (Green and Janoyan, 2005). In another experiment, a conductive concrete sample containing steel fibers was found to have a resistivity of 41.0 Ω∙m but increased in temperature by 2.5˚ F over 6000 seconds (Vincent and Janoyan, 2005). These previous results suggest that resistivity is not the only important factor in determining the heating ability of conductive concrete. The latter experiment particularly suggests that the absence of heating of the samples in this study is not due simply to the high resistivity values of the concrete, since similar resistivity values were found for these samples. Although the embedded mesh decreases the overall resistivity, the extremely low resistivity paths that it produces do not contribute to Joule heating. The lowered resistivity is a false encouragement, since the increased current cannot noticeably contribute to heating. To attempt to determine the cause of the mass thermistor failure, a probe was constructed and placed in a cylinder of dirt and water for a week. The thermistor’s room temperature resistance was then measured using the data acquisition system and found to have decreased from 53 kΩ to 45 kΩ, suggesting that the thermistors failed due to water damage. This thermistor also exhibited the same trend as observed in the embedded probes of steadily decreasing in resistance as measurements were made. Conclusions These results suggest that electrically conductive concrete composed of 0.25 - 0.5% steel fiber and welded wire mesh isolated from electrodes is not effective for deicing or anti-icing applications. Using more area on the sides of the concrete as electrodes decreases the resistivity, although this does not increase its heating capability. The data suggest the inability of the samples to sufficiently increase in temperature is due to the low resistivity of the embedded mesh, which lowers the resistivity of the sample without allowing the current passing through it to perform Joule heating. If reinforcing mesh is to be used with electrically conductive concrete, the mesh may require complete electrical isolation for the concrete to heat properly.