Environmental engineers are faced with handling large volumes of water containing very little solids. Filtration depends on area, so the problem is to create a large area of filtration surface at a low cost. One solution is to use beds of sand, or more correctly, beds of granular material. Usually a layer of coarser material supports a layer of sand. The sand may be precoated; this means a layer of excellent granular material such as diatomaceous earth on top of the sand.
The mechanisms are straining that gives a cake on top of the sand and interception on sand grains deeper in the bed (depth filtration). Both occur with larger particles predominating in the cake and almost exclusively tiny particles down deep in the bed. The cake itself is a porous medium in which tiny particles may impact on a larger particle to be captured.
A "slow sand filter" is 1 to 1.5 meters thick. The filtration comes from sand particles that are about 0.35 mm in diameter and fairly uniform. The solids in the water collect mostly on top of the filter and are scraped away at intervals of 1 to 6 months. At this time the sand is cleaned by vigorous flushing with water flowing upward. This type of filter is popular in Europe because a previous step for coagulation can be omitted. Few such filters have been used in the U.S. because the water sources tend to have fine clay that penetrates and clogs them.
Sand adsorbs organic compounds on which microorganisms can feed. The highest nutrient concentration is right on the sand granules, so that is where the organisms grow. Microbial growth clogs filters and shortens the interval until cleaning. Too much head loss is the signal that cleaning is required. Chlorination kills the microorganisms and keeps the filter going longer.
A "rapid sand filter" is less deep; half a meter to one meter is a typical thickness. Time until cleaning is a matter of hours, not days. Smaller particles would pass though, so coagulation prior to filtration is mandatory.
Filtration does not always perform satisfactorily. There is a crisis in the U.S. because of outbreaks of giardasis and cryptosporosis. The former is popularly called "beaver fever" and resembles a really bad case of intestinal flu. The latter is called "crypto" for short and is much more often deadly. The spores that cause crypto are fairly large and should not be difficult to remove by filtration, but some cities have encountered periods when their water supplies are unsafe.
The coagulation step can be operated poorly to impair the subsequent filtration step. When coagulation depends on establishing insolubility, precipitates form. However, precipitation is not instantaneous, and rates depend on available surface on which to deposit. Seed crystals can be added on which the precipitate grows, or natural supersaturation can start the precipiation process. The problem stems from not providing sufficient time for complete precipitation. If the water from the coagulation step goes too quickly to the filtration step, precipitation may cement sand grains together. This makes cleaning difficult and gives uneven particle size with different amounts of cementing at different depths. Particles of sand stuck together and trapping junk are called "mudballs".
A common question on our quizzes is "Why don't sand particles overflow during backwashing ?" The answer is easy if you remember the classes of sedimentation. A compacted bed of sand has the grains in a settling category with very slow rates. When water flows up faster that sand settles down, the bed of sand expands. As it expands, the particles have less interference with each other and settle faster. There is sufficient room above the original bed that the particles get into Class 1 settling with excellent rates that are well above the upflow rate of water. The net result is a bed expanded to the degree where its settling rate matches the upflow rate of water. Typical conditions are 3 to 10 minutes of backwashing with the bed expanded 15 to 30 %. The head loss can be equated to the weight of the suspended sand.
Backwashing is fairly effective for removing the junk that was filtered, but some may adhere strongly to the sand. There is some banging of one sand particle against another to dislodge the junk, but not much. Mechanical agitation would help and may be used for sand that is very dirty. An easier way to create turbulence is to blast some air or to direct jets of water into the sand.
Poor cleaning of sand impairs performance. A non-homogeneous bed can have cracks that offer regions of less flow resistance and thus poor distribution of the liquid. Channels of different flow rates can form. Chemicals can discourage growth of microorganisms in the sand bed. Air trapped in the sand bed is bad because channels must form. Non-uniform flow leads to an interesting phenomenon know as negative head.
A sudden change in head or surge in flow is bad because particles in the bed may be dislodged. If not captured deeper in the bed, there will be junk in the product water.
Operation of a sand filter aims at avoiding cake compression. Filter aids combat cake compaction, and intelligent application of pressure can optimize rates by using low driving force when the cake could be crushed and more head later when you might as well get what you can.