ABSTRACT Cold-applied, asphalt-modified, elastomeric polyurethane waterproofing membranes (polyurethane membranes) have been popular for roofing and waterproofing applications in the Pacific Northwest and British Columbia for at least the past 15 years. Relative low cost and easy application resulted in their widespread use in inverted roofing and waterproofing membrane assemblies (IRMA) applied to concrete decks. However, water-filled blisters under these membranes have been discovered on numerous buildings in the Pacific Northwest in recent years. In some cases, the blisters were so large that replacement of the membrane was required. Water leakage to the interior can result when the blister expands to a crack or joint in the concrete slab. The local building science and roofing industry is aware of the problem but lacks a complete understanding of causal effect or of the physics of moisture transfer. Water vapor diffusion and capillary flow do not adequately explain the pressures or volumes of water contained within these discrete water blisters. Moisture transfer via osmosis can result in blisters under significant pressure and potentially explains the observed conditions. Osmosis is the physical transfer of water through a semipermeable membrane when separating solutions of different dissolved ion (salt) concentrations. Under osmotic pressures, water will flow through a membrane from the less salty side to the more salty side, in order to reach equilibrium. A series of laboratory experiments was performed to demonstrate that the required conditions for osmosis to occur exist in the field. Laboratory testing of several of these membranes confirmed they are semipermeable to water (in order of 60 to 420 ng/Pa·s·m2 for typical thicknesses). Re – searchers also confirmed a significant dissolved- salt-ion concentration in the water collected in the field from beneath the membranes. Finally, osmotic flow was measured through several of the membranes using a controlled laboratory apparatus. The measured flow through these membranes in the laboratory is in the correct order of magnitude to explain the large water-filled blisters and pressures observed in the field. This article demonstrates osmotic flow through polyurethane membranes and attempts to create an industry awareness of the issue. Ongoing research is under way to refine polyurethane waterproofing membranes to reduce their susceptibility to osmosis and prevent future occurrences of water-filled membrane blistering. INTRODUCTION Over the past several years, RDH Build – ing Engineering, Ltd. has reviewed dozens of asphalt-modified polyurethane membrane applications ranging from five to 15 years of age. The membranes are applied to sloped concrete slabs in both insulated and uninsulated inverted roof and waterproofing membrane assemblies (IRMAs). Typically, water-filled blisters have formed between the membrane and the concrete deck and are often under considerable pressure. These self-contained, pressurized water blisters have no identifiable leakage path through or around the membrane. Blisters range in size from a penny to entire roof deck areas and can contain significant quantities of water. In some cases, large blisters (>50 mm deep) have displaced concrete pavers, creating hazardous walking conditions. As blisters expand over cracks or joints in the concrete, water can leak to the interior. In the Pacific Northwest, the blisters described above have been observed with asphalt-modified polyurethane membranes used in IRMA construction and not in other conventional roofing systems such as hot rubberized asphalt or sheet-applied SBS modified bitumen. In an IRMA, the membrane is installed directly on the structural concrete beneath insulation (if separating heat space) and ballast or wear course. In the wet Pacific Northwest climate, moisture remains in contact with the membrane for much of the year. Hygrothermal analysis shows that vapor diffusion can transport water through polyurethane membranes due to their relatively high vapor permeance compared to other roofing and waterproofing membranes (>1 U.S. perm vs. <0.01 perm). However, the quantity of water transported by vapor dif- DE C E M B E R 2010 I N T E R FA C E • 1 5 This article is reprinted from the Proceedings of the RCI Building Envelope Technology Symposium, held Oct. 26-27, 2009, in San Diego, CA. fusion is not of the magnitude required to ex – plain the blistering, nor are water vapor pressures sufficient to ex – plain the high hydrostatic pressures that exist within the blisters. To explain the large volumes of water and high pressures within the blisters, we hypothesize that osmosis is acting to transport water through the membrane. In the roofing industry, the concept of water flow by osmosis is uncommon. The process of osmotic flow has been reported to cause failure of flooring and traffic membranes and is a consideration in the design of bridge decks and water tanks where exposed to groundwater, saltwater, or road de-icing salts. It is also a reported problem and design consideration with glass fiber-reinforced boat hulls, where osmosis can form blisters within the fiberglass. Objectives The objectives of this research study are as follows: • Confirm that osmotic flow is a significant contributor to the observed in-situ blistering of polyurethane membranes directly applied to concrete in IRMA construction. • Develop a test method to determine the susceptibility of a membrane to osmotic flow, and measure the rate of osmotic flow under various salt concentrations. • Determine a possible relationship between water vapor permeance and osmotic flow. • Create awareness and highlight the need of osmosis control for the inclusion in current Canadian and U.S. standards for liquid-applied waterproofing membranes. Background Asphalt-modified polyurethane membranes have been used in hundreds of buildings constructed over the past 15 years in the Pacific Northwest and Lower Main land of British Columbia. Their relatively low cost and easy application have led to widespread use in IRMA construction, both insulated and uninsulated, as well as for planters, fountains, and foundation walls. RDH and other local consultants have reviewed dozens of buildings in which the polyurethane membranes have blistered. Blisters are also a frequent occurrence in fluid-applied polyurethane membranes in Japan (Tanaka et al. 2007 and 2008). Water-filled blisters range from penny sized Figures 1 and 2 – Typical blistered roof membranes. Blisters range from penny sized to areas several square feet in a 5- to 10-yearold membrane. Figure 4 – Water beneath blistered membrane over entire deck. 16 • I N T E R FA C E DE C E M B E R 2010 Figure 3 – Large “waterbed”-type blister lifting pavers. 3” 32” to those that encompass entire decks (Figures 1 and 2). Unlike vapor blisters filled with air and caused by other mechanisms, these are filled with water under considerable pressure. Larger blisters can lift ballast and pavers, creating a hazardous “wa – terbed” effect when one walks on the surface (Figures 3 and 4). Small blisters typically do not result in direct leaks to the interior; however, larger blisters that encompass a crack or joint in the concrete tend to manifest into leaks to the interior. Investigating IRMA construction is challenging because ballast and insulation need to be removed to expose the membrane. As a result, only small, random areas of the membrane are typically reviewed. Larger areas of membrane typically get exposed only when leaks are reported, when blisters have become so large that they lift up the ballast or pavers, or when the membrane is replaced. Blisters typically occur on horizontal surfaces, but they have also been observed on vertical surfaces of water features, planters, and green roofs. In IRMA construction located in the Pacific Northwest, it is not uncommon for water to remain at the membrane surface year round. When inverted roofing and waterproofing are investigated in the summer, even after many weeks of dry weather, water exists at the membrane surface, held by capillary forces between the membrane and insulation/ drainage mat layer and prevented from evaporating by the dimpled polyethylene drain mat or insulation installed above. In RDH’s experience, the following factors appear to increase the severity and size of the blisters: • Blisters are often more severe at low points and at areas of poor slope, i.e., ponding. However, blisters still do occur where the slab is well sloped to drains. • Blisters are typically larger and more prevalent at areas where the membrane is thinner. • The size or severity of the blisters does not appear to be affected by the use of either drainage mat or extruded polystyrene insulation applied directly over the membrane. • Blisters are almost always larger and more frequent in older membranes. In review of available literature, the authors found little information regarding the blistering of polyurethane membranes in IRMA construction, suggesting that the mechanisms causing the failure are not well understood by the roofing and waterproofing industry. MECHANISM OF MEMBRANE BLISTERING Osmosis Process Osmosis is a naturally occurring phenomenon, wherein water (or other solvent) flows through a semipermeable membrane from a solution of low-salt (solute) concentration (hypotonic) to a solution of high-salt concentration (hypertonic), without the input of energy (Oxtoby et al., 1999). Osmosis can be countered by increasing the pressure of the hypertonic solution with respect to the hypotonic side. Osmotic pressure is the pressure required to maintain equilibrium between the two sides with no net movement of solvent. Osmotic pressure depends only on the molar concentration of the solute, not the type of solute present. Therefore, if any difference in solute is present across a membrane, osmosis will occur. Essentially, if a semipermeable membrane separates a tank of fresh water and saltwater, the fresh water will flow through AIA and CSI members: earn Sustainable Design credit online with “Cool Roofing: A Solution to National Energy and Environmental Challenges” COOL ROOF Do your part to reduce the urban heat island effect. Learn more at WWW.VINYLROOFS.ORG WHAT MAKES ONE ROOFING MATERIAL COOLER THAN ANOTHER? High solar reflectance + High thermal emittance = Lower energy demand DE C E M B E R 2010 I N T E R FA C E • 1 7 the membrane to the salty side until equilibrium (equal concentration) is achieved. If left unrestrained, fresh water would essentially fill up the salty side until the water head pressure was equal to the osmotic pressure. The semipermeable membrane must be permeable to the solvent (i.e., water) but not to the majority of solutes (salt, metal, and contaminant ions); otherwise, equilibrium will be achieved by dissolution through the membrane. Depending on the molecular structure and pore size of the membrane, certain salt ions may pass through freely while other larger and heavier metal ions may not. In this case, osmosis will still occur. Reverse osmosis used in water filtration systems essentially applies a high water pressure (>50 psi) to counteract the osmotic pressure and force water ions through a specially developed semipermeable membrane to create fresh water. Reverse osmosis membranes have been developed with these properties in mind, to allow only H2O ions to pass through, rejecting other, larger salt ions, which provides greater than 99% effectiveness. The processes of osmosis and reverse osmosis are demonstrated in Figure 5. The two laws governing the osmotic pressure of a dilute solution were discovered by the German botanist W.F.P. Pfeffer and the Dutch chemist J.H. van ’t Hoff (Oxtoby et al., 1999). The laws state that the osmotic pressure of a dilute solution at a constant temperature is directly proportional to its concentration and that the osmotic pressure of a solution is directly proportional to its absolute temperature. Osmotic pressure is analogous to Boyle’s law and Charles’s Law for gases. The ideal gas law, PV = nRT, has an analog for ideal solutions in the form of πV = nRTi. This is rearranged in Equation 1 in terms of molar concentration and solving for osmotic pressure, π. Equation 1 Where: π = osmotic pressure; i = the number of ions produced during dissociation of the solute; M = the molar concentration of all solutes, moles/L; R = 8.3145 J/K·mol (0.083145 L·bar/moles·K), the molar gas constant; and T is absolute temperature, Kelvin. In solutions containing multiple types of dissolved salts, the partial osmotic pressure for each is added to determine the overall osmotic pressure across the membrane. Essentially, it is the difference in total dissolved solids (TDS) that causes the pressure. Reverse osmosis membrane manufacturers have simplified this formula to the following in terms of pressure in psi to sizereverse osmosis filtration systems (Lenntech 2008): Equation 2 Where: Σmj = the sum of molality concentration of all constituents in a solution (moles of solute/kg of solvent), and T is the absolute temperature in Kelvin. For example, saltwater from the ocean will have a total dissolved solids concentration of approximately 36,000 mg/L or ppm (Lenntech, 2008). This concentration has a total osmotic pressure difference of almost 26 bar (2.6 MPa) at 20°C (68°F). Brackish water (i.e., well water contaminated with ground salts) may have a total dissolved solids concentration of 500 mg/L, which results in a pressure of 0.25 bar (25 kPa) at 20°C (68°F). These osmosis pressures give an indication of the pressure required in a reverse osmosis filtration system. Figure 5 – Osmosis, equilibrium, and reverse-osmosis flow through a membrane. 18 • I N T E R FA C E DE C E M B E R 2010 Blistering Process We hypothesize that the formation of a blister occurs in two stages. In stage 1, a film of liquid water forms at the concrete-tomembrane interface, likely at a surface void in the concrete beneath the membrane. Various sources can form the initial film of water, including vapor diffusion and capillary flow downward through the polyurethane waterproofing membrane or water initially in the concrete slab from construction or rainwater. Once a film of water forms, vapor pressures on both sides of the membrane are equal and vapor diffusion ceases. In stage 2, the water film dissolves minerals from the concrete, increasing the salinity. Once this osmotic cell starts, osmotic pressures draw water through the membrane, creating water-filled blisters. As the pressure in the blister increases, it debonds from the concrete at the perimeter, enlarging the blister and allowing the process to continue and the blister to grow in size over time. A typical IRMA is depicted in Figure 6. Rainwater flows down through the ballast, insulation, and drainage layers to the waterproofing membrane. Throughout the wetting process, water is adsorbed on the surfaces and into the pores of the materials within the assembly, remaining there even after the bulk of the rainwater is drained away. This creates a layer of water and vapor at 100% RH at the membrane surface. Extruded polystyrene has a vapor permeance of approximately 15 ng/Pa·s·m2 (per 100 mm); and drainage mat, made of dimpled, high-density polyethylene, has a vapor permeance of less than 1 ng/Pa·s·m2. The ability of an IRMA to dry upward by vapor diffusion through these materials is limited. The concrete also has a relatively low vapor permeance, with a wet permeability estimated from 0.5 to 5 ng/s·m·Pa (WUFI, 2009; ASHRAE, 2005; Onmura et al., 2009). Thus, a 150-mm-thick slab has a vapor permeance ranging from 3 to 30 ng/Pa·s·m2. Interior ceiling finishes or coatings will reduce this further. The vapor pressure differences during wet periods of the year are shown graphically in Figure 7. Even during the warmer summer months and under solar heating, the insulation above the membrane maintains the concrete temperature relatively constant, and, as a result, minimal heat is provided to dry out the excess or absorbed water. Field experience and hygrothermal modeling show that even when bulk surface water dries out, the relative humidity above the membrane re mains above 90% year round, maintaining a constant vapor pressure drive from the exterior to the interior. Therefore, whether liquid wa – ter or water vapor is present at the roof membrane sur face, there will be an almost constant vapor pressure drive in – wards, which will effectively prevent significant drying from occurring up – wards through the Roofing contractors & specifiers choose Durapax coal tar roofing systems. Coal Tar: First Choice for Flat Roofs Durapax: First Choice in Coal Tar 610.579.9075 Durapax.com • Coal tar roofing provides low cost and long life (25+ years) • Many coal tar roofs last more than 50 years • Coal tar’s cold flow properties provide self-healing • Superior technical & customer support • Delivery you can depend on • Comprehensive warranties • UL & FM approved systems Specify your next flat roof with a Durapax coal tar roofing system. Figure 7 – Vapor drive from saturated roof membrane interface to the interior and exterior. DE C E M B E R 2010 I N T E R FA C E • 1 9 Figure 6 – Typical IRMA waterproofed with a polyurethane membrane. membrane. As a result, indoor humidity and temperature do not have a significant effect on the degree or severity of blistering. This is supported by the authors’ field observations of blistered membranes over all types of occupied and unoccupied spaces, including occupied residential suites, mechanical rooms, pools, parking garages, and exterior spaces. Figure 8 depicts the concrete moisture content output from a seven-year WUFI 4.1 hygrothermal simulation of the inverted roof assembly discussed above, with two different levels of membrane vapor permeance. The plot compares a 30-mil polyurethane membrane (vapor permeance = 400 ng/Pa·s·m2) and a two-ply SBS membrane (vapor permeance < 1 ng/Pa·s·m2). WUFI does not account for moisture flow by osmosis, but it does model vapor and capillary transport through the membrane and concrete. The impact of rainwater wetting and sitting on the membrane between rain events was accounted for in the model; however, the waterproofing membrane was assumed to have no capillary suction, consistent with other membrane material properties listed in the WUFI database. The simulation demonstrates that the polyurethane membrane will allow a net wetting of the concrete from vapor diffusion. During a typical year, more moisture is transported through the membrane into the concrete than can dry out of the concrete itself. This analysis is sufficient to show that over time the concrete under the membrane will become saturated. However, once the top surface of the concrete becomes saturated, the vapor pressure will be equal on both sides of the membrane, diffusion will stop, and there will be no driving pressure to create a blister. Once the concrete is wet, salt ions in the concrete aggregate, cement, and admixtures dissolve into available water at the surface. This dissolution of ions from the concrete creates the required salt concentration difference between the top and bottom of the membrane for osmosis to begin. Once started, osmosis continues until the salt concentration gradient is removed or water is removed from the freshwater side of the membrane. The process is slow to start but accelerates rapidly once small quantities of liquid water are present beneath the membrane. The osmosis mechanism is summarized in Figure 9. HYPOTHESIS OF OSMOTIC FLOW For osmotic flow to occur across a membrane, two requirements must be satisfied: 1. The membrane must be semipermeable to water molecules and not salt molecules, and 2. Liquid water of different salt concentrations must be present on both sides of the membrane. To prove the hy – pothesis that osmotic flow can occur across polyurethane membranes, samples of membrane and water from the blisters were collected from several build ings. The water vapor permeability of the blistered mem branes was mea sured and the sampled water analyzed to determine the salt ion concen- Figure 8 – Modeled moisture content of concrete slab with SBS and polyurethane waterproofing. Figure 9 – Blister formation mechanism by osmosis. 20 • I N T E R FA C E DE C E M B E R 2010 tration of both the blister water and the water collected from the top surface of the membrane. Water Vapor Permeance Testing of Polyurethane Roof Membranes Water vapor permeance of the sampled polyurethane membranes was measured under dry-cup, wet-cup, and inverted-wetcup conditions in general conformance with ASTM E96. Vapor permeance of new polyurethane membrane samples provided by several manufacturers and control samples of other waterproofing membranes were also measured for comparison. Vapor permeance test results correlated with published and unpublished data provided by the membrane manufacturers. Measured water vapor dry-, wet- and inverted-wet-cup permeance data for three different polyurethane membranes are provided in Figure 10. Results for three lowpermeance roofing membranes are also shown for comparison purposes. Invertedwet- cup testing (wet-cup sample inverted so that liquid water is in direct contact with sample) is not always performed by most roofing membrane manufacturers but provides the best indication of vapor permeance under realistic exposure conditions for a membrane in an IRMA system. Two of the results are for aged polyurethane membrane samples removed from their sites: Membrane 1 and 2, both approximately ten years old. Membrane 3 is a new sample of a similar polyurethane membrane from the same manufacturer as Membrane 1. The installed membrane thick nesses of 30 and 60 mils for Mem – branes 1 and 2 are considered thin by today’s standards but were common in the Vancouver market 5 to 15 years ago. The results show that the polyurethane membranes are relatively permeable compared to other roofing and waterproofing membranes, especially inverted-wet-cup measurements. The relationship between water vapor permeance and osmotic flow through a membrane is not fully understood at the microporous level; however, we believe they are intrinsically related. If the membrane has a measureable water vapor transmission rate, then it makes sense that it will have an “osmotic flow” transmission rate as well. If the membrane pore structure allows the passage of water ions but not all dissolved salt and metal ions, then osmotic pressures can be developed. Outperforms every other system on the market when it comes to maximizing energy and cost savings. Watertight within 30 seconds of being applied to a dry, clean substrate. Exceptionally reflective to abate urban heat islands. Fully adhered, with no penetrating fasteners, therefore attaching easily to most types of decks and substrates. Environmentally friendly, using no harmful ozone-depleting chemicals, leading the Environmental Protection Agency to award NCFI a Stratospheric Ozone Protection Award in 2002. Because of its durability and the fact that it can be recoated multiple times, there are buildings that still have coated foam roofs that were installed over 30 years ago. Adheres to any creative feature or curve. We know roofs. In fact, we’ve applied more than 8 million square feet of foam and we use EnduraTech™ Premier Roofing Systems by NCFI. EnduraTech™ is watertight, seamless and can be applied, multiple times, over existing roofs. EnduraTech™ lowers energy costs, keeps tenants more comfortable, and requires less of their time in repairs and maintenance. It’s good for our customers and good for our business.“ The Science of Comfort. To learn more, visit www.enduratechsystems.com Matt Segura and Aaron Lewis Southwest Spray Foam Santa Fe, NM Healthy Schools Healthy Children lthy Sc hy Chi “ The Science of Comfort DE C E M B E R 2010 I N T E R FA C E • 2 1 Water Absorption of Polyurethane Membranes Water uptake testing of the poly ure – thane membrane samples was performed to determine if water absorption is related to osmotic flow. If a material will absorb water into its pore structure, it follows that water may also be able to flow through it. Membrane samples were submerged in water for over three months and weighed periodically until the mass remained constant for longer than two weeks. Both aged membrane samples absorbed approximately 16% to 17% moisture content by mass, and the new membrane sample absorbed approximately 4%. Moisture absorption in all samples generally stopped after two to three weeks. The higher absorption in the aged samples may be due to the poly ure – thane filler materials and reinforcing mesh fabric used by both manufacturers at the time. These water absorption tests appear to indicate that water is able to pass through the pore structure of these membranes. Dissolved Salt Concentration of Blister Water Samples of water and membrane at large blisters from two different buildings were collected for laboratory testing. The water was analyzed to confirm different salt concentrations between samples collected from above and below the membrane. Samples of water were extracted from the membrane blister with a syringe and small hand pump and collected from above the membrane (rainwater). Both samples were sent to an independent laboratory for analysis of dissolved metals concentration. At a large blister in Membrane 1, approximately 12 in of hydrostatic water head (~3 kPa) was measured prior to the extraction of the sample water. The wa – ter inside the blister had a dark brownish tint, similar to all of the blisters reviewed, and is likely to be the result of continual contact with the bitumen in the asphalt-modified polyurethane membrane. See Figures 11 and 12. A water-quality laboratory analysis determined the dissolved solid concentration for 30 of the most common dissolvedmetal ions. Table 1 includes the results provided by an independent water-testing laboratory of the membrane blister water and rainwater taken from one of the buildings where Membrane 1 was used. As suspected, the blister water had high levels of several dissolved metal ions, the majority of which were sodium and potassium ions. The mix of dissolved metal ions at the blister water is likely from minerals within the aggregates, cement, admixtures, and polyurethane membrane itself. Silicon, present in cement (calcium silicates CaO·SiO2), indicates dissolution from the concrete. The level of dissolved ions within the blister water was considerably higher than rainwater, and for reference, is more saline than brackish water but much less than seawater. Figures 11 and 12 – Cutting of membrane blister and expulsion of water during extraction process. Figure 10 – ASTM E96 – Vapor permeability and permeance laboratory results for membrane samples. 22 • I N T E R FA C E DE C E M B E R 2010 Initial calculations predict that an osmotic pressure of approximately 326 kPa was present at this blister. While this pressure could not have been physically contained within this elastic membrane, it is better visualized as a pressure potential that causes suction of liquid water through the membrane. In reality, the pressure within the blister is moderated by failure of the membrane-to-concrete bond at the sides of the blister and stretching of the membrane under tensile stresses. VERIFICATION OF OSMOTIC FLOW Apparatus To measure osmotic flow through the membranes, an apparatus was developed and tested. A piece of the polyurethane roof membrane is used to separate distilled and saltwater. If osmosis flow occurs across the membrane, water will flow from the distilled water reservoir through the membrane and into the saltwater until equilibrium occurs or the pressure developed in the container is equal to the osmotic pressure (Oxtoby et al., 1999). By measuring the mass and volumetric change of the container holding the saltwater and membrane at regular intervals, the osmotic flow can be measured. Distilled water was used to represent rainwater, and water removed from several membrane blisters was used as the saltwater. Polyurethane and other roofing membranes, ranging in thickness from 30 to 150 mils, were tested, as was a commercial reverse osmosis membrane, as a “proof-ofconcept” exercise. Figure 13 demonstrates a concept of the apparatus, and Figure 14 shows a schematic of the container and photograph of one of the proof-of-concept test specimens. As a proof-of-concept experiment, a commercial reverse osmosis water filter was disassembled and a sample of the osmotic membrane put into the apparatus with the blister water removed from site. Other salt solutions using common table salt were also tested. Flow through the membrane initially was in the order of 15 L/m2/day before immense pressures developed in the containers. This order of flow is comparable to what is advertised by the osmosis membrane in a water filtering application. As a result of the high pressures developed, several of the container lids burst during the test procedure. In Figure 14, a total of 58 mL (25 L/m2) of water was transported through the reverse osmosis membrane, as was measured in the container at the end of 31 days. Procedure Samples of polyurethane membranes discussed in the previous sections were tested using the osmotic-flow test apparatus. Samples of saltwater removed from the blisters were used in conjunction with the membrane samples to demonstrate osmotic flow under the specific field conditions. Initial testing focused on two of the aged polyurethane membranes (1 and 2). Samples of the membranes and saltwater from the blisters were put into the test apparatus and tested as follows: 1. Samples of membrane were cut to precisely fit into the screw-top “open-top” lid of the glass jar. Each sample was initially weighed, and the thickness was measured. 2. A known mass of saltwater was placed into the glass jars. 3. Membrane samples were bedded into waterproof epoxy within the lid flange of the glass jar. Epoxy was also used between the membrane and the glass. The screw-top lid provided compression and a watertight fit of the membrane at the edges. The membrane acted as a gasket; Table 1 – Dissolved metals concentration – Building 1 (30-mil membrane sample). Figure 13 – Osmotic flow testing apparatus schematic. 24 • I N T E R FA C E DE C E M B E R 2010 WATER FROM WATER FROM ABOVE BENEATH MEMBRANE #1 MEMBRANE (BLISTER WATER) (RAINWATER) Soluble Metal Ions in Dissolved Solid Dissolved Solid Solution Concentration, mg/L, ppm Concentration, mg/L, ppm Sodium 2960 1.89 Potassium 574 0.47 Sulphate 75.3 <1.0 ppm Magnesium 1.83 0.35 Phosphorous 1.82 <0.2 ppm Silicon 29.9 <1.0 ppm Calcium 3.4 4.0 Other dissolved metals Trace amounts of None present several, <1ppm Total dissolved solids ~3650 ppm ~7 ppm Hardness, CaCO3 equivalent 16.0 11.5 how ever, waterproof epoxy was used to aid with sealing the membrane in place and to seal the container. 4. After the epoxy had cured and the container was leak tested, the initial mass of the container, membrane, and saltwater to – gether were measured. 5. Membrane coupons (blank samples) were produced, measured, weighed, and submerged in the freshwater bath. The water level of the freshwater was kept equal to that within the container to eliminate the effect of hydrostatic pressure head. 6. At regular intervals, the containers and blank samples were removed from the freshwater bath, dried thoroughly, and weighed. This process was repeated approximately twice a week for several months. For samples with significant osmotic flow (i.e., through an osmosis membrane), the volumetric increase can also be measured using graduations on the sample container. 7. The flow of water through the membrane was measured by subtracting the incremental mass from the initial container, water, and sample mass. The glass container, lid, and epoxy did not absorb water (confirmed by producing empty container blanks and submerging them in water), so any change in mass was Figures 14 a (above) and b (right) – Osmotic flow testing con – tainer schematic and photograph of increase in water volume. Building Envelope Technology Symposium 2011 The Westin Charlotte, NC October 10 – 11, 2011 RCI, Inc. 800-828-1902 www.rci-online.org/symposium.html Technical Education for Building Envelope Professionals The Symposium on Building Envelope Technology is a two-day educational program - ing design, construction, maintenance and - lopes. The event’s open, inclusive environment encourages attendees to ask questions and - Call For Abstracts The Building Envelope Committee is solicit- www.rci-online.org/downloads/ B ildi E l P Educatio f i l on a inclus nology edu Buildin sive cational ng TechProfessionals Co mmittee down nloads/ DE C E M B E R 2010 I N T E R FA C E • 2 5 the result of absorption into the membrane and flow into the salty water within the container. 8. To isolate osmotic flow, the absorption rates of the membrane samples must be known. For the two membranes tested, each gained approximately 17% moisture by mass within three weeks and remained constant for the remainder of the test. Blank samples were used to determine the required time and mass of water for absorption to saturation to occur. Following this initial uptake, additional mass/volume gain of the containers was by osmosis flow through the membrane. This can also be confirmed by measuring the volume of water within the container and the subsequent pressure developed. Results Measured osmotic flow rates through 30- and 60-mil samples of Membranes 1 and 2 are presented in Figure 15. Several additional samples of these membranes are currently being tested and show similar flow rates. Small deviations result from precision of weighing the samples (to nearest ±0.01g of container mass), but otherwise a constant flow rate was measured. With this setup, it appears to take a few weeks for the membrane to become fully saturated before osmotic flow rates can be determined. On average, an osmotic flow rate of between 8 and 13 g/m2/day was measured. As shown, both membrane samples have similar order-of-magnitude osmoticflow rates. Over time, the osmotic pressure should decrease with a lower salt concentration; however, the volume of water needed to reach equilibrium is many times the size of the jars. Thus, we wouldn’t expect to see a significant change in slope on the graph measuring osmotic flow in the 150- day test period. However, as pressure develops within the blisters in the field and the membrane pore structure stretches and opens up, the osmotic flow rate may be affected. The blister water contains many different salt ions at varying concentrations. Over time, some of the low concentration salts will reach their equilibrium pressure, marginally reducing the flow rate. Further experiments are ongoing to determine the effects of the osmotic pressure and osmotic flow. Figure 16 shows initial results for 0.1 and 1.0 molar NaCl salt solutions compared to distilled water (blank sample) and the blister water. Comparing the osmotic flow rates of the three salt solutions to the distilled water confirms that the salt concentration has a direct effect on the rate of osmosis through the polyurethane membrane. The lack of flow through the con trol distilled wa ter sample also confirms that the process of osmosis is pulling water through the membrane, not capillary suction or another mechanism. Based on continued research under way with new unaged polyurethane membranes with and without various primer coatings, preliminary osmotic flow rates of 0.5 g/m2/day to 7 g/m2/day have been measured in the laboratory. The intent of this further research is to reduce the osmotic flow rate through new membranes to as close to zero as possible to prevent blisters from forming. Discussion The measured osmotic flow rate through the aged polyurethane membranes (with the saltwater removed from the blisters) is, on average, between 8 and 13 g/m2/day. For a membrane in an IRMA application continually exposed to water, in one year, this flow rate equates to between 3 and 5 L/m2 of water (3 to 5 mm deep) transported by Figure 16 – Comparison of osmotic flow through Membrane 1 with different osmotic suction pressures. 26 • I N T E R FA C E DE C E M B E R 2010 Figure 15 – Polyurethane Membranes 1 and 2 – measured osmotic flow through membrane. osmo sis. In ten years, this is on the order of 30 to 50 L/m2 or water (30 to 50 mm deep), which corresponds to the volume contained beneath the membrane in blisters we have observed at several buildings. Blisters are often observed to be 3 mm to 25 mm tall, and in extreme cases, where the pavers are floated, greater than 50 mm. Continuous exposure to liquid water will affect blister formation, and it is likely that some blisters may grow and shrink seasonally when water is present on top of the membrane. In addition, both membrane adhesion and tensile strength of the membrane will affect how large the blisters grow in the field. Blisters will also stop growing if the pressure becomes equal to the osmotic suction pressure or if the blister does not continue to grow by adhesion failure or stretching, which may occur at low salt concentrations and osmotic pressures. In the IRMA application, wetting of the concrete surface by osmosis is shown to be up to an order of magnitude higher than the drying capability of the assembly. Above the membrane, RH remains high, between 90% and 100% year round. As a result, drying outward is slow through the insulation and drainage mat: at the rate of less than 0.1 g/m2/day. On the bottom side, drying through the concrete is also a very slow process, and vapor flow inward through the concrete is estimated to be on the order of 1 g/m2/day (dependent on concrete properties). These mechanisms for drying are slower than the wetting process, and moisture accumulates beneath the membrane. The osmotic cell develops immense suction pressures that cause blisters to form and expand, acting to delaminate and stretch the membrane. As the process progresses, the blisters expand into each other until very large water-filled blisters develop. CONCLUSIONS Severe water-filled blistering of coldapplied, asphalt-modified elastomeric poly – urethane waterproofing membranes is a frequent problem for inverted roof membrane assemblies in the Pacific Northwest. RDH’s testing and research demonstrate that the water-filled blisters can be explained by the fluid transfer mechanism of osmosis. The research confirms that osmotic flow does occur through these membranes, the conditions for osmosis to occur exist in the field, and our test results replicate the same order of magnitude of moisture transfer observed in the field. The rate of osmotic flow is a function of the vapor permeance of the membrane. Therefore, lowering the vapor permeability of the polyurethane membrane will likely reduce the potential for osmosis to start by decreasing the potential for the top surface of the concrete to become saturated and will likely result in a lower rate of flow under osmotic pressures. The aged polyurethane membranes that were removed from blistered roofs and tested were found to be semipermeable and have a vapor permeance ranging from 60 to 420 ng/Pa·s·m2, depending on application thickness and chemical composition. Some new polyurethane membranes that have also been tested have similar order-ofmagnitude vapor permeance values (up to 120 ng/Pa·s·m2), even when tested with certain concrete primers. Osmotic flow rates measured through aged polyurethane membranes that were removed from blistered locations are on the order of 8 to 13 g/m2/day. Preliminary testing has also been performed on new primed and unprimed polyurethane membranes that are currently available on the market, with measured flow rates of between 0.5 to 7 g/m2/day, depending on membrane chemistry, thickness, and primer application. These lower flow rates are still in excess of most other waterproofing and roofing membrane systems, and at this time, it is not known if these flow rates are low enough to prevent blisters from occurring within the expected service life of the membrane. Further research is needed to develop an acceptable solution. The two most relevant standards that cover the manufacture and installation of asphalt-modified polyurethane membranes are ASTM C836-00, Standard Specification for High-Solids-Content, Cold, Liquid-Applied Elastomeric Waterproofing Membranes for Use With Separate Wearing Course; and CAN/CGSB – 37.58-M86, Membrane, Elas – tomeric, Cold-Applied Liquid for Nonexposed Use in Roofing and Waterproofing. These standards do not contain maximum values for vapor permeance, requirements for reporting inverted wet-cup permeance numbers, or osmosis-testing requirements. Based on the field observations and the testing performed in this study, the existing standards do not have adequate test requirements to prevent premature blistering of polyurethane membranes. Soil to Soil Contact Means Shared: Moisture • Nutrients • Beneficial Organisms First and Only Hybrid Green Roof System Monolithic Surface • Subterranean Modules • Instant Green LiveRoof.com (800) 875-1392 LiveRoof® a built-in-place system with the convenience of a modular system DE C E M B E R 2010 I N T E R FA C E • 2 9 RECOMMENDATIONS Awareness and understanding of the physical process of osmosis and its potential impact on building materials will help the industry come up with solutions and avoid the problem in the future. We recommend that maximum allowable values for membrane vapor permeance, tested under inverted-wet-cup conditions, and osmosis testing requirements be included in current industry standards referenced by polyurethane membrane manufacturers, specifically ASTM C836-00 and CAN/CGSB–37.58-M86. We also recommend that an industryaccepted standard be developed to test new IRMA roofing and waterproofing membranes for susceptibility to osmotic flow. Additional research is needed to determine allowable osmotic- and vapor-flow rates that can be safely accommodated by moisture flow through concrete slabs. The effects of aging and exposure to wet and alkaline conditions on the material properties of polyurethane membranes in the field should also be further researched in this context. Research should also be performed to examine the effects of concrete primers and sealers to prevent the passage of salts to the membrane interface. Based on these findings, new poly ure – thane membranes should be modified to be sufficiently impermeable to vapor and osmotic flow to prevent blistering within their expected service lives, while still maintaining their other desirable physical properties for waterproofing. ACKNOWLEDGEMENTS We thank Mark Lawton, PEng, of Morrison Hershfield, Ltd.; Bob Matich of BASF Canada; and Cor Claus at Pacific Polymers for their assistance and support during this research project. REFERENCES ASTM Standard E96/E96M-05, Stan – dard Test Methods for Water Vapor Transmission of Materials, ASTM International, 2005. ASHRAE Handbook of Fundamentals, American Society of Heating, Re frig – erating and Air-Conditioning Engi – neers, Inc., Atlanta, GA, 2005. Lenntech, Osmotic Pressure Calculator, available at www.lenntech.com, 2008. S. Onmura, S. Hokoi, T. Matsushita, D. Ogura, K. Kominami, Y. Yasui, “A Measurement of Concrete Hygro – thermal Properties and the Influence of Its Scattering on Hygrothermal Behaviour in Concrete Walls,” ASTM Second Symposium on Heat-Air- Moisture Transport: Measurement and Implications. Vancouver, BC, 2009. D. Oxtoby, H. Gillis, N. Nachtrieb, Prin – ciples of Modern Chemistry, 4th Edition, Saunders College Publishing, 1999. Kyoji Tanaka, Structural Engineering Research Center, Tokyo Institute of Technology, e-mail discussion re – garding blistering of fluid-applied polyurethane membranes, 2008. Kyoji Tanaka and Hiroyuki Miyauchi, “Waterproofing Systems in Japan,” Interface, RCI, Inc., October 2007. Wikipedia, osmosis definition at www.en.wikipedia.org, 2009. WUFI Pro 4.2 IBP Material Database, Fraunhofer Institut Bauphysik, 2009. Brian Hubbs, PEng, a senior building science specialist with RDH Building Engineering Ltd., has over 18 years of experience as a consultant focused exclusively in the field of building science. This work has included the design of new building enclosures as well as forensic investigation, rehabilitation, maintenance, and litigation support on existing buildings. Hubbs’s experience includes steep- and low-slope roofing systems, wall cladding of all types, windows, glazing, glass/metal curtain walls and skylights, and below-grade and plaza waterproofing systems. He is a Building Envelope Professional (BEP) through the Architectural Institute of British Columbia Professional Engineers and Geoscientists (IBC/APEGBC) program. Brian Hubbs, PEng, BEP Graham Finch is a building science research engineer with RDH in Vancouver, BC. He holds a master’s degree specializing in building science from the University of Waterloo and is actively involved with numerous research projects at RDH, leading to several papers. Finch has several years of experience as an engineering consultant on building enclosure issues across North America. This work has included new construction design, hygrothermal modeling, material evaluation and testing, forensic investigation, whole building monitoring, and several industry research studies. Graham Finch Robert Bombino, PE, senior science specialist and principal with RDH, has extensive experience with a variety of building enclosure systems, components, and materials used across the U.S. and Canada. Prior to joining RDH in Seattle in 2004, Robert spent five years practicing in Boston, MA. He holds a bachelor’s degree in civil engineering from the University of Waterloo and a master’s in architectural engineering from Penn State University. He is a registered professional engineer in Massachussets and Washington and a member of ASTM Committees C16, D08, and E06, as well as a member of ASHRAE Technical Committee 4.4. Robert Bombino, PE 30 • I N T E R FA C E DE C E M B E R 2010
