Reverse osmosis (RO) membrane desalination has been increasingly used to produce drinking water from salt water. Most of the United States has relied on fresh water resources to produce drinking water, but even water-rich regions, such as states in the northeast, are experiencing a strain on traditional resources. Furthermore, many regions lack sufficient fresh water resources and are turning to alternate resources, such as seawater or brackish water, to sustain water needs. In particular, a growing number of inland communities have both insufficient fresh water and unused brackish water (500 – 10,000 mg/L total dissolved solids) resources. A key financial and technical limitation to inland RO desalination is disposal of the waste stream (concentrate); typically, 10 – 25% of the influent volume becomes the RO concentrate stream. This waste volume is large when compared to the waste volume produced by traditional fresh water treatment (less than 1%). To improve the feasibility of RO desalination, RO system recovery (volume of product water per volume of feed water) must be increased to decrease the concentrate volume.
Brackish water RO treatment recovery is limited by sparingly soluble salt (CaCO3, CaSO4, BaSO4, SrSO4, silica) precipitation. Specifically, calcium carbonate (CaCO3) is known to be a key, omnipresent precipitate. Salt precipitation can be limited or prevented through a combination of chemical addition, pH control, and decreased recovery. Chemicals called antiscalants are dosed prior to membrane treatment and delay precipitation through association with and modification of the salt crystals in solution. Antiscalants are typically synthetic organic phosphonates, acrylic polymers, or polymer blends. Precipitation prevention through antiscalants can be achieved within a limited range of specific salt concentrations. As recovery is increased, antiscalant control is overcome and precipitation occurs. An alternate approach is thus required for further recovery augmentation.
This research focuses on the development of a novel process to treat brackish water RO concentrate and increase overall recovery. This research focuses on a novel process and includes three stages: antiscalant degradation, salt precipitation, and solid/liquid separation.
This study focused on the second stage, salt precipitation, using model synthetic concentrates and several antiscalant types to evaluate the precipitation of calcium carbonate. The model concentrate is based on data from a brackish water source in Arizona, assuming an 80% recovery and 100% ion retention. First, the precipitation of calcium carbonate as a function of pH was determined for solutions with a phosphonate antiscalant and with no antiscalant. For the model concentrate used (Saturation Index, SI, for CaCO3 = 2.3 at pH 8), samples without antiscalant resulted in 87% calcium precipitation, while, as expected, samples with antiscalant showed 0% calcium precipitation. This result corresponds to the recommended SI limit for successful precipitation prevention by an antiscalant (2.5 for CaCO3) . As pH increased, samples with antiscalant showed increased calcium precipitation (92 – 93% at pH 10.5). However, the same sample without antiscalant always resulted in a higher calcium precipitation (99.7% at pH 10.5). Subsequent experiments with varied water composition showed a decrease in calcium precipitation with the addition of magnesium and sulfate.
The effect of antiscalants on CaCO3 precipitation was also evaluated through particle size and particle number distributions, using a laser granulometer Mastersizer S (Malvern Instruments) and a laser particle counter (Met One). Several antiscalants, including organic phosphonates and one acrylic polymer blend, were tested. Most samples showed particle size distributions similar to samples without antiscalant. However, two antiscalants (amino tri(methylene phosphonic acid), or AMPA) and diethylenetriamine penta(methylene phosphonic acid), or DTPMP), showed markedly different results at the highest antiscalant concentration tested (85 mg/L and 100 mg/L, respectively, i.e. 17 mg/L and 20 mg/L in the hypothetical RO influent). Particle size measures taken directly after precipitation in the presence of 85 mg/L AMPA and 100 mg/L DTPMP showed a bimodal particle size distribution. Both peaks represented smaller sizes than peaks obtained for samples without antiscalant. Particle size measures taken 3 days after precipitation showed a significant increase in particle size, and results also varied as a function of pH. Photos taken with a light microscope (10x and 25x) confirm the results obtained from the laser granulometer. Particle counter results showed the majority of particles are within a particle size range of 2 – 50 micrometers; the number of larger particles (50 – 300 micrometers) increased with time for samples with 85 mg/L AMPA.
Light microscope photos also showed differences in particle shape and size with different antiscalants, as well as for solutions with and without antiscalant present. These photos, along with flux data manipulation, were used to provide explanations for the type of membrane fouling during dead end filtration.
Dead end filtration (MWCO = 0.1 micrometers, Millipore nitrocellulose membranes) was performed to compare the fouling potential of the precipitated solutions as a function of water composition, antiscalant type, and antiscalant concentration. A simplified concentrate, containing sodium chloride (NaCl), sodium bicarbonate (NaHCO3), and calcium chloride (CaCl2*2H2O) was first tested. The water composition was then changed by separately adding magnesium chloride (MgCl2*6H2O) and sodium sulfate (Na2SO4). Finally, the full water data set was tested. Results showed differences between antiscalants and water compositions. AMPA and DTPMP showed flux decreases of approximately 30% over a period of 7 minutes for the simplified concentrate, while the same two antiscalants showed only a 19% flux decrease when magnesium was added to the simplified concentrate. For the four antiscalants tested, there was no consistent relationship between antiscalant concentration and membrane flux. For all antiscalants tested, while a greater antiscalant concentration resulted in less calcium precipitated, a greater antiscalant concentration did not result in lower permeate flux decline. Subsequent tests on solutions containing only antiscalant and deionized water indicated membrane fouling occurred through antiscalant adsorption onto the membrane surface.
Future experiments will focus on tests with real water samples to evaluate the effect of natural organic matter.
 Chemical Pretreatment for RO and NF. Hydranautics, Technical Application Bulletin No. 111, Rev. B, 2003.