Laboratory Investigation of Performance and Impacts of Snow and Ice Control Chemicals for Winter Road Service

For this research, the deicers and reagent-grade chemicals used in various laboratory tests are listed in Table 1. The reagent-grade NaCl (99%, solid) was purchased from ScienceLab.com (Houston). The noninhibited NaCl-based deicer (IceSlicer) and salt-sand mixture (with 10–25% NaCl) were both provided by the Colorado Department of Transportation (CDOT) out of its solid stockpile. IceSlicer consists of naturally occurring complex chlorides (mostly NaCl) and more than 40 trace minerals. The two ${\mathrm{MgCl}}_2$-based deicers were provided by the CDOT out of its liquid tank and by the vendor, respectively, with active ingredient concentration of 27–29%. The three agro-based deicers were provided by CDOT out of its liquid tank and by the vendors, respectively, containing high concentrations of either ${\mathrm{MgCl}}_2$ or ${\mathrm{CaCl}}_2$. The NaAc/NaFm deicer was a blend made from NAAC (solid, 97% anhydrous NaAc, from Cryotech, Fort Madison, IA) and Peak SF (98% granulated NaFm, from the Blackfoot Company, Toledo, OH) at $50:50$ weight ratio. The liquid deicer CF7 was obtained from Cryotech with 50% KAc. The solid CMA deicer was obtained from Cryotech with 96% of hydrated calcium magnesium acetate. The test solution of CMA, CF7, NAAC, and Peak SF deicers contained 18.4, 0.2, 6.6, and 11.4 mM chloride, respectively, based on chloride sensor measurements. The reagent-grade potassium formate, KFm (99%, solid), was purchased from Alfa Aesar (Ward Hill, MA).

Laboratory measurements of ice melting capacity of various deicers were conducted following the SHRP H205.1 and H205.2 test methods (Chappelow et al. 1992). The SHRP H205.1 test measures the ice melting capacity of solid deicer pellets spread randomly across an ice surface of uniform thickness. The results of the test provide a measurement of the ice melting capacity of the deicer relative to the generated brine, or melted ice. The test utilized 25 mL of deionized water to form a sheet of ice of uniform thickness in a 3.5 cm (radius) Petrie dish. Once frozen, ice extrusions from the ice surface were melted. The sample was then refrozen for 24 h to equilibrate. After equilibration at the desired temperature, 1 g of solid was broadcast over the ice specimen. At 10, 20, 30, 45, and 60 min after application of deicer, the generated brines were removed from the specimen dish and weighed. The generated brine was then reintroduced to the same specimen dish. The process of removal of brine, weight, and reintroduction was completed within 1 min for each sample. The process was completed at 0, $-5$, $-18°\mathrm{C}$ (32, 23, $-0.4°\mathrm{F}$). Special consideration was taken to use separate weighing dishes for each deicer to avoid cross-contamination. For liquid deicing solutions (SHRP H205.2), similar procedures were followed, except that 0.9 g of liquid deicer was pipetted onto the ice surface.

Laboratory measurements of ice penetration of various deicers were conducted following the SHRP H205.3 and H205.4 test methods. The SHRP H205.3 test measures the penetration behavior of solid deicer pellets on a cylinder of ice. The results of this test provide a measure of the penetration rate of the selected deicer over time.

The test requires a custom-built Plexiglas apparatus with holes drilled to form cavities (Fig. 1). A 0.50 in (1.3 cm)-thick Plexiglas with dimension $2\times 8$ in ($5\times 20\mathrm{cm}$) had 10 cavities drilled vertically along the margin using a $5/32$ in (0.4 cm) drill bit to a depth of 3.5 cm with 1.7 cm spacing between cavities. The cavities were then enlarged at the top with a countersink bit to create a surface cone. These cavities were filled with deionized water and then frozen for 12–24 h. The solid deicers were placed in the same freezer to ensure equilibration to the test temperature. Once completely frozen, any extrusions of ice were melted. The top of the ice surface was then wiped with an absorbent tissue to remove all liquid water and the apparatus was replaced in the freezer for an additional 1–2 h to re-equilibrate. A pellet of solid deicer was weighed and placed on the surface of the ice specimen along with a tracer dye solution (McCormick, red) and was placed in the freezer. The penetration depth of each pellet was measured using a ruler at 10, 20, 30, 45, and 60 min. The process was carried out at 0, $-12$, and $-20°\mathrm{C}$ (32, 10.4, $-4°\mathrm{F}$) in triplicate.

For liquid deicing solutions (SHRP H205.4), similar procedures were followed, except that 25 mLl of liquid deicer were dyed with 2 drops of (McCormick, red) food coloring. The dyed liquid deicer solution was mixed and allowed to equilibrate to the given test temperature during the freezing of the ice specimen. Three drops of dyed liquid deicer solution were pipetted onto the ice specimen. Often the penetration depth of dyed solutions was difficult to visually assess. When this occurred, a small metal probe was inserted in the cavity until it contacted the ice interface to determine penetration depth. The probe was maintained at the test temperature and did not contribute to any penetration or melting.

Laboratory measurements of ice undercutting of various deicers were conducted following the SHRP H205.5 and H205.6 test methods. The SHRP H205.5 test measures the ice undercutting ability of a solid chemical deicer. The results of the test provide an area where the ice-concrete bond has been broken by the deicer as a function of time.

Portland cement concrete (PCC) blocks were made $6\times 9\times 2$ in ($15.2\times 22.8\times 5.1\mathrm{cm}$) and allowed to cure in the mold for 48 h in a wet chamber. The concrete blocks were then demolded and placed in a saturated lime solution for 28 days to cure. Once removed from the lime solution, the substrate was washed and allowed to dry. A 1 in (2.54 cm) wide rubber strip was affixed to the exterior of the concrete block, forming a wall around the top side of the substrate. The interface between the substrate and the rubber wall was sealed using silicon sealant.

Subsequently, 150 mL of deionized water were allowed to freeze on top of a concrete block for 12–24 h. The solid deicers were placed in the freezer to equilibration to the test temperature. Two drops of (McCormick, red) food coloring were deposited at three equally spaced locations across the ice surface and were allowed to freeze for 3–4 h. Two weighed deicer pellets were placed at the center of each frozen red dye stain. Photographs were taken at 10, 20, 30, 45, and 60 min. The process was carried out at 0, $-6$, $-10$, and $-16°\mathrm{C}$ (32, 21.2, 14, $3.2°\mathrm{F}$) in triplicate.

The digital pictures taken of the undercut area were imported into Adobe Photoshop CS2. The magic wand, polygon selection, lasso selection, and pixel count tools were used to select the extent of undercut area. A pixel count was obtained of the given selection and an area of undercut substrate was calculated for each photo.

For liquid deicing solutions (SHRP H205.6), similar procedures were followed with the exception that, when the ice specimen was completely frozen, a 0.5 in metal rod was heated and pressed vertically at three evenly spaced locations across the specimen. The resulting liquid was wiped clean, leaving a shallow cylinder free of ice terminating at the concrete surface. Three drops of dyed deicer liquid solution were pipetted into the center of the three shallow holes.

Laboratory measurements of changes to PCC through freeze-thaw testing in the presence of deicers were conducted following the SHRP H205.8 test method entitled “Test Method for Rapid Evaluation of Effects of Deicing Chemicals on Concrete” (1992) with minor modifications. The SHRP H205.8 test evaluates the effects of chemical deicing formulations and freeze-thaw cycling on the structural integrity of small test specimens of non-air-entrained concrete. The method quantitatively evaluates degradation of the specimen through weight loss measurements. This test method is not intended to be used in determining the durability of aggregates or other ingredients of the concrete.

The concrete mix design was specified following the ASTM C 672-91 standard (2003). An ASTM specification C150-07 (2007) Type I/II low-alkali portland cement (Ash Grove Cement Company, Clancy, MT) was used in this study. The chemical composition and physical properties of the cement are available from the writers’ previous paper (Shi et al. 2009e). The fine aggregates used were clean, natural silica sand sifted to allow a maximum aggregate size of 1.18 mm before proportioning and admixing. The coarse aggregates used were crushed limestone with a maximum size of $3/8$ in (0.95 cm). The concrete mix design had a water-to-cement ratio (w/c) of 0.51, a coarse-aggregate-to-cement ratio of 2.36, and a fine-aggregate-to-cement ratio of 1.75. The coarse aggregates and fine aggregates were prepared to saturated-surface-dry (SSD) condition before mixing, featuring a moisture content of 0.55% and 3.6%, respectively. The fine and coarse aggregates were added to the 3-cubic-feet (86-L) mixer and mixed until a homogeneous mixture was obtained. Then the cement was added and mixed again until a homogeneous mixture was obtained. Next, water was added from a graduated cylinder and mixed until the concrete is homogeneous and of the desired consistency. The batch was remixed periodically during the casting of the test specimens and the mix container was covered to prevent evaporation. The fresh concrete featured a slump of nearly zero [by the ASTM C143 method (2005)] and air content of 3% after compaction [by the ASTM C173 method (2010b)]. Concrete specimens were made in $11/2$ in diameter $\times 17/8$ in height ($3.8\times 4.8\mathrm{cm}$) polyvinyl chloride piping with a volume of $54{\mathrm{cm}}^3$. At least four specimens were made for each deicer solution or control solution. To facilitate demolding, one saw cut was made through each specimen mold along the $17/8\times (4.8\mathrm{cm})$ length axis. The cut was then covered with duct tape. In the first 24 h of molding, the concrete specimens were placed on a rigid surface under ambient temperature and at a relative humidity of approximately 50% and covered to present excessive evaporation of water. Next, the specimens were demolded and cured in a moist cure room with relative humidity of 98% for 27 days (which deviates from the curing regime specified by the SHRP method). The 28-day compressive strength of test cylinders [by the ASTM C873 method (2010a)] was 6,619 psi (45.6 MPa), well above the recommended 4,000 psi (27.6 MPa).

Once fully cured the concrete specimens were allowed to dry overnight and weighed. For each deicer, four concrete specimens were placed on a cellulose sponge inside a dish containing 310 mL of deicer solution and then covered with plastic wrap to avoid water evaporation and to slightly compress each test specimen into the sponge. The deicer solutions were made at a 3% by volume for liquids and 3% by weight solution for solids using deionized water. This 3-to-100 ratio was chosen to simulate the field dilution of liquid or solid deicers once applied onto the pavement. One end of each test specimen was in full contact with the sponge. The test specimens (along with the deicer sponge and dish) were placed in the freezer for 16 to 18 h at $-17.8\pm 2.7°\mathrm{C}$ ($-0.04°\mathrm{F}$), and every diluted deicer tested froze at this temperature once it was placed in the freezer for some time. Subsequently, the specimens (along with the deicer sponge and dish) were placed in the laboratory environment at $23\pm 1.7°\mathrm{C}$ (73.4 °F) and with a relative humidity ranging from 45 to 55% for 6 to 8 h, at which temperature every diluted deicer tested thawed once it was taken out of the freezer for some time. This cycle was repeated 10 times. The average heating rate and cooling rate was observed to be $0.4°\mathrm{C}/\min$ and $1.2°\mathrm{C}/\min$, respectively. After complete thawing following the tenth cycle, test specimens were carefully removed from the dish, individually rinsed under running tap water, and hand-crumbled to remove any material loosened during the freeze-thaw cycling. The largest intact part of each test specimen was then placed in open air to dry for 24 h at $23\pm 1.7°\mathrm{C}$ (73.4 °F) and a relative humidity ranging of 45–55%. After drying, test specimens were weighed and the final weights recorded.

Concrete samples were made (3 cm diameter $\times 1\mathrm{cm}$ thick) with a w/c of 0.46 and aggregate and sand sizes no greater than $1/6\mathrm{cm}$. Samples were cured in the moist cure room for 28 days at approximately 100% humidity. The compressive strength of the concrete at seven days was tested to be 4,929 psi (34.0 MPa), by the ASTM C873 method.

Concrete samples and the deicers were brought to the CSM Instruments Laboratory (Needham, MA) where 1.7 mL of deionized water (approximately 2 mm thick) were allowed to freeze (approximately 30 min) on the concrete surface. The freezer temperature was $-16.2\pm 1.6°\mathrm{C}$ ($2.8\pm 2.8°\mathrm{F}$) and the freezer humidity was $50.6\pm 11.6$ percent, and the lab temperature was maintained at $23.7°\mathrm{C}$ ($74.8°\mathrm{F}$) and 30% humidity. Deicers were then applied to ice surface and allowed to sit for 5 min in the freezer. Samples were then loaded onto the tribometer and run for 100 laps at $3\mathrm{cm}/\mathrm{s}$ with a 10 mm diameter and 5 Newtons of force applied. The pin was cleaned between each sample using water and isopropyl alcohol. The liquid deicers were diluted to 3% by volume and two drops were pipetted onto the surface. Solid deicer were weighted and applied so that by weight the deicer did not exceed 3% concentration if all the ice on the sample were melted. Each deicer was run at least in triplicate.

Laboratory testing was conducted using a differential scanning calorimetry (DSC) to quantify the thermal properties of deicers. The DSC (TA Instruments Q200) was set to run from 25 to $-60°\mathrm{C}$ (77 to $-76°\mathrm{F}$) with cooling/heating rates at 2°C per minute. All samples were run as liquids. Solid samples were made into a saturated liquid solution with deionized water. Samples were run at $1.5:1$ dilution. Ten $\mu \mathrm{L}$ of each sample were pipetted into the aluminum test chamber and sealed, and then weighed. All samples were run in triplicate. Data were used from the warming cycle, which were less prone to the supercooling effect and featured better reproducibility than those from the cooling cycle.

Corrosion to mild steel [ASTM A36 (2008)] and galvanized guardrail steel (Trinity Highway Products) was measured using a Gamry Instruments Potentiostat with an eight-channel electrochemical multiplexer (ECMB). Deicer solutions were 3% by weight for solid and by volume for liquid samples, made with deionized water. Before testing, the metal samples were cleaned with acetone and deionized water, and then dried. Four samples of each metal type were placed in the deicer solution and the open circuit potential (OCP) was monitored for 48 h.

Electrochemical techniques provide an attractive alternative to the gravimetric method [e.g., Pacific Northwest Snowfighters Association (PNS)/National Association of Corrosion Engineers (NACE) test] in terms of allowing for rapid determination of corrosion rates of metals and revealing information pertinent to the corrosion mechanism and kinetics. As such, at 48 h of immersion in the deicer solution, the weak polarization curve of each metal sample was taken to rapidly measure the corrosivity of the deicers. Weak polarization is an experimental technique that measures the current-potential relationship of a metal in the electrolyte when an external potential signal (perturbation) is applied within $\pm 120\mathrm{mV}$ range of its corrosion potential ($E_{\mathrm{corr}}$) at a given sweeping rate. Such a current-potential plot, known as a potentiodynamic polarization curve, provides information on the corrosion mechanisms occurring at the metal-electrolyte interface and the instantaneous corrosion rate of the metal, based on which the corrosivity of various deicers can be evaluated.

The ice melting capacity of the selected deicers varied as a function of product type and test temperature. At 0°C, the differences in the 60-min ice melting capacity were relatively small for all liquid and solid deicer products (Fig. 2). At $-5°\mathrm{C}$, the solid deicer products (one NaCl-based, one NaFm-based, and one NaAc-based) performed better than the liquid deicer products (one ${\mathrm{MgCl}}_2$-based and one agro-based). However, at $-18°\mathrm{C}$, liquid deicers outperformed the solid deicers, and the solid deicers based on NaAc or NaFm failed to melt any ice. Overall, the reagent-grade salt [NaCl ($r$, $s$)] outperformed all the other products by performing well at all three temperatures. Nixon et al. (2007) reported that at 0°F ($\sim -18°\mathrm{C}$), ${\mathrm{MgCl}}_2$ slightly outperformed NaCl, and at 23°F ($\sim -5°\mathrm{C}$) NaCl performed the best, followed by ${\mathrm{MgCl}}_2$ and the agro-based one, which coincides with this study’s results.

Although results from the SHRP ice melting capacity test method provide performance data that can be easily understood and used by field practitioners, reproducibility issues have been reported and the method has been modified in many studies to generate more consistent results (Goyal et al. 1989; Chappelow et al. 1993; Nixon et al. 2005). The rate of dissolution of solid deicers may have been a factor affecting reproducibility, which is dependent on the particle size and the amount of brine needed for reasonably accurate measurements. This study reduced the surface area of the ice based on recommendations by Chappelow et al. (1993) to limit the errors resulting from absorption, but this greatly limited the amount of brine generated, especially at colder temperatures.

The ice penetration capability of the selected deicers also varied as a function of product type and test temperature. Overall, the liquid deicers outperformed the solids at all temperatures (Fig. 3). At 0°C, the two liquid ${\mathrm{MgCl}}_2$-based deicers, the KAc-based deicer, and the agro-based deicer all penetrated to the bottom within 30 min of the 60-min test. Penetration performance for all deicers gradually diminished as the temperature got colder.

As a group, the solid deicers performed similarly well at 0°C, with the exception of the solid NaAc-based deicer, which failed to penetrate into ice at the three test temperatures. The solid NaFm-based deicer failed to penetrate into ice once the temperature dropped to $-12°\mathrm{C}$. The solid NaCl-based deicer showed penetration at 0°C, but not at $-12°\mathrm{C}$. At $-18°\mathrm{C}$, the solid NaCl-based deicer pellet became lodged, but the generated brine became the agent behind the penetration. Nixon et al. (2007) reported much lower ice penetration rates at 30°F ($\sim 0°\mathrm{C}$), less than 6 mm for all deicers tested, compared with 30 mm of ice penetration depth shown in Fig. 3. Yet, the relative deicer performance in these two studies was consistent, where ${\mathrm{MgCl}}_2$ and KAc showed good ice penetration and NaCl showed poor ice penetration.

The writers do not recommend this test method for evaluating solid deicers because the solid pellets often became lodged in the ice column and led to reproducibility issues. This is consistent with the study by Nixon et al. (2007), which indicated that the ice penetration test was not a useful specification test.

The ice undercutting abilities of solid deicers were very difficult to measure, as shown from the lack of data in Fig. 4. At temperature ranges between $-16°\mathrm{C}$ and $-10°\mathrm{C}$, the solid NaCl-based deicer was the only deicer to reach the substrate; all other solid deicers either demonstrated no notable effect to the ice surface or only partially melted down to the substrate, with some refreezing by 60 min. The solid NaAc-based deicer showed some undercutting at $-10°\mathrm{C}$, but was ultimately refrozen by 60 min. At 0°C, all test deicers generally reached the substrate, however, few generated sufficient brine to begin the undercutting process, i.e., the two liquid ${\mathrm{MgCl}}_2$-based deicers, the KAc-based deicer, and the agro-based deicer. This is consistent with the outstanding ice penetration performance of these four liquid deicers discussed earlier.

Overall, the liquid deicer products worked better at undercutting and breaking the ice-concrete bond than solids, but no one liquid deicer significantly outperformed others in the group. Generally, the deicers performed better at warmer temperatures, with the exception of the solid NaAc-based deicer showing its peak performance at $-6°\mathrm{C}$. Salt brine (23% liquid, reagent grade), the NaFm-based deicer, and the sodium acetate/formate (NaAc/NaFm, 50% blend of the two)-based deicer were tested, but all showed no ice undercutting at the three test temperatures.

Although results from this test method provide insight and performance data that can be used to guide field applications, the method itself is plagued by reproducibility issues that are difficult to address, especially for solid deicers. In addition, many deicers initially appeared to be undercutting and breaking the ice-concrete bond, but in fact the dye was moving across the surface of the ice.

The presence of deicers generally degraded the freeze-thaw resistance of PCC. In this test program of various deicers, the controls were deionized water ($\mathrm{DI}\mathrm{\text{-}}{\mathrm{H}}_2\mathrm{O}$) and NaCl (reagent grade, solid). Using the PCC samples with air content of greater than 3%, $\mathrm{DI}\mathrm{\text{-}}{\mathrm{H}}_2\mathrm{O}$ had little effect on the structural integrity of the concrete, whereas NaCl caused approximately 43% weight loss after the SHRP freeze-thaw test. Among the deicers tested, the solid NaCl-based deicer and the liquid KAc-based deicer caused the most weight loss of the concrete sample (approximately 50%), followed by the solid NaFm-based deicer and the solid NaAc/NaFm-based deicer (approximately 24%), whereas the liquid ${\mathrm{MgCl}}_2$-based deicer and the solid CMA-based deicer caused the least weight loss (as shown in Fig. 5).

The significant damage of PCC in the presence of NaCl or NaCl-based deicers can be explained by the role of NaCl in significantly accelerating the freeze-thaw cycles and thus exacerbating the physical and chemical damages. ${\mathrm{MgCl}}_2$ did not seem to have as much impact on the freeze/-thaw resistance of PCC, at least after 10 days of testing. On the other hand, multiple studies have suggested ${\mathrm{MgCl}}_2$-based deicers to be more destructive to PCC than NaCl-based deicers (Cody et al. 1996; Mussato et al. 2004). The difference in the findings can be attributed to the differences in the deicer concentrations (concentrated versus diluted), the PCC samples used (old cored concrete versus new concrete that feature different cement hydration products and concrete microstructure), the test procedures (wet-dry cycles and freeze-thaw cycles), and the test duration. Although previous studies used mostly concentrated deicer solutions, this study tested diluted deicers (assuming a 3-to-100 dilution ratio for all liquid and solid deicers), because the deicer concentrations experienced by the PCC structures and pavements in the field are generally low in the long term. Laboratory findings from either the SHRP H205.8 or other test methods do not necessarily accurately predict the deicer/concrete interactions in the field and research is still needed to establish the correlation between the laboratory data and the field data. Additional work at the writers’ laboratory examined the chemical changes in the cement paste after the SHRP freeze-thaw test and found that ${\mathrm{MgCl}}_2$ did react with the cement hydrates to form fibrous or needlelike crystals characteristic of magnesium chloride hydroxide hydrate (Shi et al. 2009c), which is consistent with the findings by Sutter et al. (2008), Cody et al. (1994, 1996), Deja et al. (1999), and Lee et al. (2000). Given longer exposure time, the ${\mathrm{MgCl}}_2$-based deicer might cause considerable damage to the concrete as well.

The alternative deicers tested all caused significant damage to the PCC, with the only exception being the CMA-based deicer. Because this study did not use aggregates susceptible to alkali silica reactivity (ASR), the exacerbated deterioration of PCC in the presence of KAc, NaFm, and NaAc merits further investigation since it cannot be explained by the mechanism proposed by Rangaraju et al. (2005, 2006).

The friction coefficient on the ice was lower than the friction coefficient on the wet concrete surface (Fig. 6), showing that the ice is more slippery than wet concrete. Partly attributable to the immature nature of this test method, the friction test results had great variance and did not show any significant difference between liquid and solid deicers or between chloride-based deicers and the acetate-or formate-based deicers. Nonetheless, the tribometer test revealed that the agro-based deicer led to the lowest friction coefficients on both the ice and the deiced concrete, whereas the NaCl-based solid deicer had the greatest variance of friction coefficients on the ice. The latter may be attributed to the clay or aggregate material naturally embedded in the NaCl-based solid deicer.

The current practice for DOTs is to measure the friction coefficient of wet and icy roads with a friction wheel that is attached to the back of a vehicle on a trailer and towed. Al-Qadi et al. (2002) found that the use of friction measurements can improve winter maintenance operations and mobility, but using a device with an extra wheel may not be practical in the field. There have been recent advances in friction measurement devices, such that they can provide real-time data in order to facilitate the determination of deicer application rates based on road surface conditions. Bergström et al. (2003) also reported the use of a portable friction tester to measure friction on cycleways that was able to differentiate various winter road conditions and various pavement materials.

The quantification of friction coefficients of deiced pavement is a new application of the tribometer. Although the data in Fig. 6 demonstrate the feasibility of this technology, the tribometer and test procedures need to be modified to provide laboratory data that are reliable and representative of field experience. The tribometer holds promise for a potential standard test protocol to assess the friction implications of deicer products, if more research is conducted to establish its test results with field friction test data.

The DSC was used to quantify the thermal properties of deicers. Fig. 7 shows the heating-cycle thermogram for six general types of deicers. The peak at the warmer temperature range for each thermogram defines the characteristic peak, which corresponds well with the effective temperature for the deicer. If the temperature got colder than the characteristic peak, ice crystals would begin to form and lead to icy pavement in the field environment. The DSC thermograms were very reproducible for each deicer at a given dilution rate and heating rate, and thus may serve as a fingerprint tool for quality assurance of deicers. For instance, the NaCl-based deicers were the only deicers consistently featuring two equally strong peaks in the warming cycle.

The two peaks observed in the NaCl-based deicer warming cycle represent two endothermic phase transitions, where the peak on the left represents the separation of the ice from the subcooled NaCl or the pseudoeutectic formation (Table 2) and the peak on the right represents the warmest temperature at which ice crystals begin forming (Huidt and Borch 1991; Koefod 2008). This may also help explain the two peaks observed in the NaFm and NaAc/NaFm-based deicer warming cycles.

Table 2 shows the characteristic temperature and heat flow of the tested deicers obtained from the DSC thermograms. The table shows that KAc-based deicer had the coldest effective temperature, followed by the ${\mathrm{MgCl}}_2$-based deicer. Table 3 shows the ice melting capacity of the tested deicers from the SHRP test. The table shows that the solid NaCl and solid NaCl-based deicers had the highest ice melting capacity. In general, the SHRP ice melting data had more variance than the DSC data, as indicated by the higher numbers in the coefficients of variance (COV).

In Table 2, the change in heat flow ($\Delta H$) was calculated by subtracting the total heat of fusion for pure water ($334\mathrm{J}/\mathrm{g}$) from the measured heat flow of the characteristic peak. Statistical analysis revealed the following correlation between the DSC data and the SHRP data at 0°C, as shown in Eq. (1):

The positive coefficient associated with ${\mathrm{Log}}_{10}(\Delta H)$, i.e., 8.58, is consistent with the notion that the more powerful a deicer is, the less external heat it needs to melt the ice (and thus the higher value in $\Delta H$). The high $R$-square value confirms that there is a strong correlation between the change in heat flow ($\Delta H$) at the characteristic temperature measured with the DSC and the SHRP ice melting capacity at 0°C. The less-than-one $R$-square can be attributed to experimental error especially in the SHRP test or the different behavior between solid and liquid deicers in the SHRP test. As such, DSC may hold the promise for a reliable standard test protocol to assess the deicer performance under certain road weather conditions. More details related to the development of a DSC-based test protocol for quantifying the performance of liquid deicers can be found in a recent report (Akin and Shi 2010).

For deicers diluted at 3% by weight or volume (for solid and liquid deicers, respectively), electrochemical testing of their corrosion to mild steel and galvanized steel showed that acetate-based deicers were much less corrosive to mild steel than chloride-based deicers and the agro-based deicer. The corrosion rates [in milli-inches per year (MPY)] of mild steel for $\mathrm{NaCl}-$, ${\mathrm{MgCl}}_2-$, and acetate-based deicers from our electrochemical test (weak polarization) were similar to those reported by Levelton Consultants (2007) who used another electrochemical test method, linear polarization resistance (LPR), to assess the deicer corrosivity to mild steel (ASTM A36).

In general, steel is considered to be passive when its corrosion current density $i_{\mathrm{corr}}<0.1\mu \mathrm{A}/{\mathrm{cm}}^2$, and active corrosion occurs when $i_{\mathrm{corr}}>1.0\mu \mathrm{A}/{\mathrm{cm}}^2$. As such, it can be concluded that the acetate-based deicers were noncorrosive to mild steel, whereas the chloride-based deicers and the agro-based deicer were very corrosive (Table 4). Nonetheless, the galvanized steel in the acetate-based deicers was found to be corroding at comparably high rates to the other deicers. For both metal types, the specific agro-based deicer product showed high corrosivity, showing little benefits of the agro-based additive and contradicting the claim by a previous study (Kahl 2004).

Data in Table 4 add to the existing knowledge base concerning the corrosive effects of deicers to metals and illustrate that the additives beneficial in reducing the corrosivity of deicers to one metal may not work for a different type of metal. Kennelley (1986) conducted electrochemical and weight loss tests of 14–17 month duration, which indicated that bridge structural metals, including steel, cast iron, aluminum, and galvanized steel corroded considerably less in CMA solutions than in NaCl solution. Callahan (1989) demonstrated that pure CMA could effectively inhibit chloride-induced corrosion of reinforcing steel, whereas CMA as an additive to NaCl did not inhibit the rebar corrosion in concrete. Ushirode et al. (1992) reported that NaAc, urea, and CMA were only marginally effective as corrosion inhibitors for reinforced concrete.

The writers intend to lay the groundwork for a framework that allows an agency to assess the various deicer products on the market or to formulate its own deicer product based on laboratory testing and multicriteria decision-making (MCDM). The notions of alternatives, multiple attributes, conflicts among criteria, decision weights and decision matrix from the deterministic, stochastic, or fuzzy MCDM methods (Triantaphyllou 2000) can be adopted for the selection or formulation of snow and ice control materials. The crux is to strike the right balance in meeting multiple goals of the highway agency, including safety, mobility, environmental stewardship, infrastructure preservation, and economics (Shi 2005).

The following paragraphs describe a deicer composite index that would allow winter maintenance managers to numerically evaluate deicers based on their agency priorities or local needs and constraints. As an example,the CDOT user priorities are applied to the composite index, which would allow the most applicable deicer to be selected based on its relative ranking. Each deicer attribute category or subcategory was assigned a decision weight based on the CDOT users’ input, and more details are available in the project final report by Shi et al. (2009a). As shown in Table 5, the deicer composite index for each deicer product is calculated by multiplying the relevant decision weights with the attribute values indicating where the product’s cost, performance or impacts fall in the specific category or subcategory.

The decision weights are scaled between 1 and 10, with 1 being least important and 10 being most important. The attribute values are normalized between 1 and 10, which in this work were semiquantitatively estimated based on the overall examination of existing literature, and experimental data and were provided for concept demonstration purpose only. One could also quantitatively calculate these attribute values based on the experimental data alone, assuming that the laboratory test protocols used can effectively capture the relative performance or relative impact of the deicer products. For performance attributes, the attribute values ranged from 1 to 10, with 1 being the worst and 10 being the best. For impact attributes, the attribute values ranged from 1 to 10, with 1 being the most deleterious and 10 being the least.

As shown in Table 5, the deicer composite index was calculated to be 46.6, 57.1, and 46.5 for noninhibited NaCl, inhibited liquid ${\mathrm{MgCl}}_2$, and KAc-or NaAc/NaFm-based deicers, respectively. This illustrates the challenges still faced by the road maintenance agencies, given that none of the deicers evaluated is close to being perfect (which would have a deicer composite index of 100). With the CDOT user priorities, the inhibited liquid ${\mathrm{MgCl}}_2$ deicer products present a better alternative than either the noninhibited NaCl or the KAc-or NaAc/NaFm-based deicers. Nonetheless, such comparisons may no longer hold true when applied to a different agency that has user priorities different from those of CDOT.

This is a preliminary attempt to illustrate how user priorities and experimental data can be integrated into a defensible decision-making process for selecting or formulating snow and ice control materials. There are many caveats in this simplified MCDM framework, such as the use of laboratory data to predict field performance and impacts as well as the lack of an appropriate index to quantify environmental impacts, given that environmental concerns are often site-specific. At this point in time, there is no laboratory test method available for deicing or anti-icing performance and friction coefficient that directly correlates with the performance and friction of deicers in the field. As such, the existing laboratory tests can only provide a baseline to contrast various products under well-controlled conditions, and the findings derived from such tests need to used with caution. Additional research is needed before such a framework can be robust and reliable enough to be adopted by the DOT managers to guide their operations.