Long-running comparison of feed-water scaling in membrane distillation

Abstract

Membrane distillation (MD) has shown promise for concentrating a wide variety of brines, but the knowledge is limited on how different brines impact salt scaling, flux decline, and subsequent wetting. Furthermore, past studies have lacked critical details and analysis to enable a physical understanding, including the length of experiments, the inclusion of salt kinetics, impact of antiscalants, and variability between feed-water types. To address this gap, we examined the system performance, water recovery, scale formation, and saturation index of a lab-scale vacuum membrane distillation (VMD) in long-running test runs approaching 200 h. The tests provided a comparison of a variety of relevant feed solutions, including a synthetic seawater reverse osmosis brine with a salinity of 8.0 g/L, tap water, and NaCl, and included an antiscalant. Saturation modeling indicated that calcite and aragonite were the main foulants contributing to permeate flux reduction. The longer operation times than typical studies revealed several insights. First, scaling could reduce permeate flux dramatically, seen here as 49% for the synthetic brine, when reaching a high recovery ratio of 91%. Second, salt crystallization on the membrane surface could have a long-delayed but subsequently significant impact, as the permeate flux experienced a precipitous decline only after 72 h of continuous operation. Several scaling-resistant impacts were observed as well. Although use of an antiscalant did not reduce the decrease in flux, it extended membrane operational time before surface foulants caused membrane wetting. Additionally, numerous calcium, magnesium, and carbonate salts, as well as silica, reached very high saturation indices (>1). Despite this, scaling without wetting was often observed, and scaling was consistently reversible and easily washed. Under heavy scaling conditions, many areas lacked deposits, which enabled continued operation; existing MD performance models lack this effect by assuming uniform layers. This work implies that longer times are needed for MD fouling experiments, and provides further scaling-resistant evidence for MD.