- Email: [email protected]
Tough, void-free, flame retardant phenolic matrix materials
Construction and Building Materials 13 Ž1999. 343]353
Tough, void-free, flame retardant phenolic matrix materials C.S. Tyberg a , M. Sankarapandian a , K. Bears a , P. Shiha , A.C. Loosa , D. Dillard a , J.E. McGrath a , J.S Riffle a,U , U. Sorathiab a
NSF Science and Technology Center for High Performance Polymeric Adhesi¨ es and Composites, Virginia Polytechnic Institute and State Uni¨ ersity, Blacksburg, VA 24060, USA b Na¨ al Surface Warfare Center, 9500 MacArthur Bl¨ d., West Bethesda, MD 20817, USA Received 25 April 1998; received in revised form 26 April 1999; accepted 2 May 1999
Abstract Polymer matrix composites may be effective alternatives to steel and concrete in many civil structures due to their anticipated superior environmental stability. However, one major obstacle limiting the use of organic materials in construction is their high combustibility. Phenolic resins are high volume materials widely utilized for many purposes, including structural adhesives, composites, coatings, etc. They are also well known to provide fire resistance in many applications. Normally, their curing process involves reagents which produce volatile by-products such as water, formaldehyde, or ammonia. This feature has severely limited processability and their utilization in void-free structural applications. This paper is focused on a catalyzed epoxy-phenolic reaction utilizing commercially available phenolic resins and liquid epoxies to promote a predominately Ž50]90% wrw. phenolic network which produces little or no volatiles. Both ductility and flame resistance are enhanced in compositions containing these high phenolic concentrations. The materials appear to have significant potential for adhesives and structural composites, wherein the absence of voids could permit significant new products to be prepared for infrastructure and other applications. Relationships between the chemical and network structures of these materials and their properties are described. Q 1999 Elsevier Science Ltd. All rights reserved. Keywords: Composites; Flame retardant; Phenolic matrix materials
1. Introduction Lightweight, stiff polymer matrix composites may be effective alternatives to steel and concrete in many civil structures. It is anticipated that polymer matrix composites’ superior oxidative resistance relative to steel and better freeze]thaw durability relative to concrete could lead to structural components with the needed improved environmental resistance. However, one major obstacle limiting the use of organic materials in construction is their high combustibility. U
Corresponding author. Department of Chemistry, Virginia Polytechnic Institute, 3107 Hahn Hall, West Campus Drive, Blacksburg, VA 24061-5976, USA. Tel.: q1-540-231-8214; fax: q1-540-231-8517.
Phenolic resins, both Novolacs and Resols, are widely used commercially due to their excellent flame retardance and low cost w1]10x. There has also been significant academic interest in understanding the mechanisms of curing and decomposition of the phenolics in order to understand the reasons for their excellent flame retardance w2x. As a result of the low smoke generation of phenolic resins they are being used in the interiors of planes such as the DC-10 w3x. Phenolics have a broad range of applications varying from construction to electronics and aerospace w11,12x. Commercial Novolac resins are cured with reagents such as hexamethylenetetramine ŽHMTA. which yields networks with relatively high crosslink densities w12,13x. Curing with HMTA produces volatile by-products such
0950-0618r99r$ - see front matter Q 1999 Elsevier Science Ltd. All rights reserved. PII: S 0 9 5 0 - 0 6 1 8 Ž 9 9 . 0 0 0 2 2 - 7
344
C.S. Tyberg et al. r Construction and Building Materials 13 (1999) 343]353
as water, formaldehyde, and ammonia, which lead to voids in the materials w14]20x. This has been a negative feature in phenolic matrix composites because the high void content leads to brittle components. Therefore, current commercial Novolacs are limited to applications where high strength is not a requirement. On the other hand, epoxy resins are used for applications where high strength and toughness are needed, although they are highly combustible materials. The flame retardance of epoxy resins can be improved by incorporating bromine into the structure, but bromine also increases the toxic smoke emission w21x. Phenolics have been used as curing agents for epoxy resins where the epoxy is the major component w22,23x. In many cases, these systems make up the base resins for semiconductor packaging due to their improved hydrophobicity relative to amine cured epoxies and their low cost. Effects of crosslink densities on the physical properties of the networks have been studied w24x. Unfortunately, these systems do not have good flame retardance because of the high content of combustible epoxy resin. This paper focuses on curing phenolic resins with epoxies, as opposed to HMTA ŽFig. 1. w25x. The cure reaction produces no volatiles and therefore produces tough, void-free networks. Network density can be at least somewhat controlled by the stoichiometric offset between phenol and epoxy groups. The approach is to utilize networks high in phenolic content Žup to ; 80 wt.%. so that the flame retardant properties of the phenolic material are retained while tailoring mechanical properties via crosslink densities and molecular structure. In addition, these systems should be much tougher than the phenolic Novolac resins cured with
HMTA as a result of the lack of voids in the networks. The intent is to prepare materials as strong as epoxy networks which retains the flame retardance of phenolic resins w26,27x.
2. Experimental section 2.1. Materials The commercial phenolic resins were provided by Georgia]Pacific ŽProduct no. GP-2073.. The phenolic resin used had a molecular weight of ; 741 grmol. Epon 828 epoxy resin Ž1. was obtained from Shell Chemical. XUR-1544-55284-35 epoxy Ž2. and DER 354 epoxy Ž3. were obtained from Dow Chemical Co. ŽFig. 2.. Triphenylphosphine ŽTPP. was obtained from Aldrich and used as received. 1,1,3,3-Tetramethyldisiloxane was obtained from United Chemical Technologies, Inc. and used as received. Allyl glycidyl ether and hydrogen hexachloroplatinate ŽIV. hydrate were obtained from Aldrich and used as received. 2.2. Synthesis of 1,3-bis(3-glycidoxypropyl)tetramethyldisiloxane [28] To a 250-ml three-neck flask equipped with a dropping funnel, nitrogen inlet, thermocouple, and magnetic stirrer, was added allyl glycidyl ether Ž87.56 g. and hydrogen hexachloroplatinate ŽIV. hydrate Ž0.0083 g.. This solution was stirred at 258C for 45 min to activate the catalyst. Tetramethyldisiloxane Ž39.58 g. was added to the dropping funnel. After 45 min a few drops were added to the allyl glycidyl etherrcatalyst solution. This
Fig. 1. Curing of phenolic novolac with epoxy.
C.S. Tyberg et al. r Construction and Building Materials 13 (1999) 343]353
345
Fig. 2. Epoxy resin structures Žbisphenol-A based epoxy Ž1., brominated bisphenol-A based epoxy Ž2.; bisphenol-F based epoxy Ž3.; and siloxane epoxy Ž4..
was stirred at 25]308C until a temperature increase of ; 58C due to the reaction exotherm was observed. At this point the tetramethyldisiloxane was added dropwise in order to control the exotherm to less than 458C. After all the siloxane was added, the solution was allowed to cool to room temperature. The remaining allyl glycidyl ether was vacuum stripped at 608C and 2 torr pressure. The epoxy was then vacuum distilled at 2008C and 2 torr pressure. 2.3. Solution mixing of the epoxy and phenolic components with catalyst The epoxy component Ž1, 2, 3, or 4., Novolac, and catalyst were weighed into a round bottom flask, then stirred in acetone Ž1r10 wrw polymerrsolvent. until dissolved ŽFig. 2.. The solvent was evaporated under nitrogen, and the sample was placed in a vacuum oven at room temperature for 24 h. The sample was then heated above Tg at 608C for 2]3 h. Samples prepared in this manner were used for melt viscosity measurements. 2.4. Preparation of fire test panels and fracture toughness bars with epoxies 1, 3, and 4 To a three-neck round bottom flask equipped with a vacuum tight mechanical stirrer and a vacuum adapter
was added phenolic Novolac and ; 75% of the required epoxy. The flask was heated in an oil bath to 1708C. When the Novolac began to melt at 1708C, mechanical stirring was begun. The vacuum was applied incrementally to prevent the material from swelling into the vacuum line. Once final vacuum was achieved Ž2]5 torr. the solution was stirred for approximately 20 min to degas the blend. During this time the remaining epoxy, with the catalyst Ž0.1]0.5 wt.% based on the total weight of epoxy. dissolved in it, was degassed in a vacuum oven at ; 808C. The stirring solution was cooled to between 130 and 1508C, depending on the viscosity of the blend being prepared, and the vacuum was temporarily released to add the remaining epoxy with catalyst. This was stirred for approximately 4 min to fully degas the samples. The melt was then poured into a mold Žwhich had been preheated to the initial cure temperature of the material. in a preheated oven and covered with mold release cloth and a heavy metal plate. The samples with epoxy 1 and 0.1 wt.% triphenylphosphine based on the weight of the epoxy were cured at 1808C for 1 h and then 2008C for 1 h. The samples with epoxy 3 and 0.1 wt.% TPP Žbased on the weight of epoxy. were cured at 1808C for 1 h, 2008C for 1 h, and 2208C for 1.5 h. The samples with epoxy 4 and 0.25 wt.% TPP Žbased on the weight of epoxy. were cured at 2008C for 1.5 h and 2208C for 2 h. The cure
346
C.S. Tyberg et al. r Construction and Building Materials 13 (1999) 343]353
schedules were determined by isothermal differential scanning calorimetry ŽDSC. measurements of uncured samples as well as dynamic mechanical analysis ŽDMA. to determine Tg values of the cured samples. The samples cured by the indicated cure cycles showed no change in Tg with additional heating, indicating that the samples were fully cured. Each phenolicrepoxy system required different cure cycles based on the cure rate of the material. 2.5. Preparation of fire test panels and fracture toughness bars with epoxy 2 After samples were mixed by the solution method and dried in a vacuum oven, the solid blends were crushed to powder. The powders of each composition were placed in molds and heated to ; 1108C for approximately 10 min in a vacuum oven to melt and degas the samples. The molds consisted of the appropriate shape grooves of 3.18= 6.35= 38.1 mm for fracture toughness bars and 10 = 10 cm = 6.3 mm for fire test panels; in addition, an extra 1-cm border was added around the perimeter of each mold to prevent the samples from bubbling out of the molds during the degassing process. The molds were then removed from the vacuum oven and placed in a preheated oven to cure the samples. The samples were heated at 1808C for 1 h and 2008C for 1 h. This method produced void free samples for fracture toughness and fire testing. 2.6. Hot-melt prepregging and composite fabrication A lab scale Model 30 prepregger manufactured by Research Tools Corporation, Ovid, MI, USA was used in composite preparation ŽFig. 3.. In this apparatus, a 36K sized AS4 carbon fiber tow was passed through a wedge-slit die at the bottom of a heated resin pot containing the matrix resin. The wetted tow was then passed between a pair of flattening pins and around a guide roller before being wound on a drum. The flattening pins and the guide rollers were independently heated. Unidirectional carbon fiber prepregs were prepared with various uncatalyzed phenolicrepoxy 1 resin systems using the hot-melt technique. The set-point temperature of the resin pot, flattening pin, and roller were determined by viscosity data. The set-point temperature was 1208C for the 50r50 wrw phenolicrepoxy 1 and 1408C for the 70r30 wrw and 80r20 wrw phenolicrepoxy 1 resins. These high temperatures were necessary to give a low enough melt viscosity to process the materials. Low melt viscosity of the resin systems is critical to permit good wet-out of the reinforcing fiber tows and yield uniform resin content. The prepregs are then cut and placed in a metal mold and cured with a thermal cycle of 2308C for 4 h to make composite panels. The volume fraction of fibers in the composites,
Fig. 3. The hot]melt prepregging process.
calculated based on densities of the fiber and resin, was 59]63% fiber. 2.7. Measurements Differential scanning calorimetry measurements were carried out using a Perkin]Elmer differential scanning calorimeter, model DSC-7. Samples of approximately 8]12 mg were placed in aluminum pans and cured using a dynamic temperature scan from 30 to 2608C with a heating rate of 58Crmin. The peak in the reaction exotherm was used to determine cure temperature and rate. The temperature calibration was performed using indium and zinc as standards. Thermogravimetric analyses were carried out using a Perkin]Elmer Thermogravimetric Analyzer model TGA-7. Samples of approximately 5 mg were placed in an aluminum pan, and heated at 108Crmin from 30 to 6008C under nitrogen. Dynamic mechanical analyses were used to determine the glass transition temperature and the relative trends in rubbery moduli. These measurements were performed using a Perkin]Elmer dynamic mechanical analyzer, model DMA-7. The Tg values were calculated from the peaks in the Tan delta curves. The analyses were conducted using a three-point bend setup with rectangular specimens. The test was run with tension control set at 110% using amplitude control adjusted to fix the room temperature modulus to 1 = 10 9 Pa. Fully cured samples were heated at 58Crmin from 25 to 2008C. Soxhlet extractions with methyl ethyl ketone ŽMEK. as the solvent were used to determine the sol fractions
C.S. Tyberg et al. r Construction and Building Materials 13 (1999) 343]353
of the networks. Cellulose thimbles were soaked in MEK for 24 h then dried in the vacuum oven at room temperature for 24 h. They were subsequently heated to 1408C for approximately 12 h. After the samples were allowed to cool to room temperature in the vacuum oven, they were weighed within 30 s of releasing the vacuum and the weight was recorded. Samples were extracted for 72 h, placed in the vacuum oven at room temperature for 24 h, then heated to 1408C under vacuum for approximately 12 h and allowed to cool to room temperature under vacuum. After being removed from the vacuum oven, the thimbles containing sample were weighed within 30 s to prevent any significant weight gain due to moisture absorption. Flame retardance was directly measured using cone calorimetry on void-free neat panels Žno fiber. in a horizontal orientation. Panels with a surface area of ; 0.01 m2 and 6.35 mm thick were used with a radiant heat flux of 50.0 kWrm2 . Three-point bend tests were utilized to characterize the toughness of the phenolic-epoxy networks in terms of the critical stress intensity factor K 1c , according to ASTM standard D 5045-91 w29x. The specimens had thickness b Ž3.12 mm. and width w Ž6.28 mm.. The single edge notch bending method was used. First, a sharp notch was created in the sample by sawing. A natural crack was initiated by inserting a cold razor blade Žwhich had been immersed in liquid nitrogen. into the notch and tapping with a rubber hammer. The depth of the crack Ž a. was between 40 and 60% of the width Ž w .. The pre-cracked notched specimen was loaded crack down, into a three-point bend fixture and tested using an Instron model 4204 instrument. The single edged notched bending rig had moving rollers to avoid excess plastic indentation. The three-point bend fixture was set up so that the line of action of applied load passed midway between the support roll centers within 1% of the distance between these centers. The crosshead speed was 1.27 mmrmin, and the testing was conducted at room temperature. Elastic moduli were determined using a Dynastat calibrated using digital calipers accurate to 0.01 mm. The Dynastat was designed for materials testing: performing creep, relaxation, and dynamic tests. It is capable of loading a small specimen in tension or compression up to 10 kg. Displacement control is also supported, as in a relaxation test, and the full range is 10 mm. The oven temperature is controlled automatically, with a range of y150]2508C. The test specimens, with dimensions of height h Ž3.18 mm., width w Ž6.35 mm., and length l Ž38.1 mm., were placed on two flat supports with a span of 2.54 cm, in the three point bend setup. The samples were heated to 608C above Tg to measure the elastic moduli. Once the sample reached the set temperature a small load Ž0.01 kg. was placed on the sample and the displacement at equilibrium was
347
measured. Equilibrium was obtained in - 3 s, indicating that the material was in the rubbery plateau region at the testing temperature of Tg q 608C. The load was increased by increments of 0.01 kg up to approximately 0.08 kg and at each load the equilibrium displacement was recorded. From this data the load]displacement curves were generated and linear regression analysis was performed to determine the slopes of the lines. These slopes were used to determine the moduli according to Eq. Ž1.. The displacement values measured from the Dynastat are accurate to within 0.05 mm. Es g Ž PrD .Ž L3r48 I . PrD g L I
s s s s
Ž1.
slope of load]displacement data gravitational constant s 9.81 mrs 2 length between supports s 2.54 cm Ž1r12. wb 3
Room temperature density measurements were performed using a Mettler]Toledo AG204 balance adapted with a Mettler]Toledo density determination kit for ATrAG and PGrPR balances. Rectangular samples with dimensions of approximately 19 = 6.35= 3.18 mm were sanded, then polished, to prevent any trapping of air bubbles. Deionized water was degassed in the vacuum oven at 238C for approximately 30 min prior to use. Using this setup, the weight of the solid in air ŽA. and the weight of the solid in water ŽB. were measured. The temperature of the water was recorded to within 0.18C and the density of distilled water at that temperature was determined using a density table. Room temperature density was calculated according to Eq. Ž2..
r s Ar Ž A y B .4 r Ž H 2 O .
Ž2.
The density at 608C above Tg was calculated using the coefficients of thermal expansion below Tg and above Tg , measured by Thermal Mechanical Analysis using the large quartz parallel plate measuring system in TMA mode. Using these densities and the elastic moduli Ždetermined from Dynastat measurements . the molecular weights between crosslinks Ž Mc . were calculated by Eq. Ž3.. Mc s 3 RTrrE9
Ž3.
Room temperature water absorption was determined for phenolicrepoxy Ž1. networks cured with 0.1 wt.% TPP based on the weight of epoxy and compared to an epoxy control cured with diaminodiphenylsulfone at 1808C for 2 h and 2208C for 2 h. Rectangular samples with dimensions of approximately 19.5= 6.4= 1 mm were sanded and polished to remove any flaws which might affect the results. Two samples of each compo-
C.S. Tyberg et al. r Construction and Building Materials 13 (1999) 343]353
348
sition were accurately weighed then placed in vials filled with water at room temperature. The samples were periodically removed from the water, dried with a Kimwipe, then weighed in order to measure the moisture uptake. The sample weight and the amount of time elapsed from the start of the experiment were recorded. The time was normalized for the varied thickness by dividing by the sample thickness. For each composition, the average moisture uptake of the three samples was recorded and plotted vs. the square root of the time in hours divided by the thickness. Data was collected until the curves leveled off. Rheology experiments were carried out on a Bohlin VOR Rheometer in continuous oscillation mode with a frequency of 1 Hz. Temperature control was accomplished with a Bohlin HTC. The auto-strain was set to control the torque to 25% of the maximum torque allowed. The maximum strain for the instrument was 0.25. Samples of approximately 0.7 g of dry powder were pressed into a pellet then placed between the preheated 25-mm diameter parallel plates of the rheometer. The gap was closed to ; 1 mm and the sides were scraped to remove excess sample before the run was started. The samples used in these test were prepared by solvent mixing. 2.8. Composite panel characterization The composite panels were C-scanned for overall quality using a Sonix model HF 1000 by a pulse-echo arrangement using a 15-MHz transducer with a focal length of 38.1 mm. Scanning electron photomicrographs, obtained from an ISI SX-40 SEM, were used to establish the void-free nature of the composites and to confirm good resin distribution and fiber wet-out. Transverse and longitudinal flexural tests following ASTM standard D-790 were used to evaluate a systematically varied series of phenolicrepoxy 1 compositions.
3. Results and discussion A phenolic Novolac resin with an average functionality of 7.3 phenolic hydroxy groups was cured with four different epoxy resins, 1 Žbisphenol-A based epoxy., 2 Žbrominated bisphenol-A based epoxy., 3 Žbisphenol-F based epoxy., and 4 Ž1,3-bisŽ3-glycidoxypropyl.tetramethyldisiloxane. to investigate the effects of epoxy structure on network properties ŽFig. 2.. For each chemical structure of the epoxy component, systematically varied compositions were studied ranging from 50 to 80 wt.% phenolic Novolac. Structural and compositional effects on network toughness, transition temperatures, and flame retardance were investigated. In general, network toughness would be expected to increase with the average molecular weight between crosslinks Žincrease in ratio of phenols to epoxies. up to some point where the amount of unconnected phenolic chains begins to dominate properties. Likewise, glass transition temperatures of fully cured networks should decrease as the distance between crosslinks increases. Thus, it was of interest to predict and measure network densities and correlate these chemical structures to physical properties of the networks. ‘Theoretical’ average distances between crosslinks were predicted by considering the stoichiometries and assuming a linear phenolic chain ŽTable 1.. In considering the chemical structures of these networks, it was predicted that the crosslink densities should decrease as the stoichiometric offset Žexcess of phenols. was increased. These predicted values were then compared to experimental measurements of crosslink densities derived from rubbery moduli and elasticity theories. The following assumptions were made in order to calculate the predicted molecular weights between crosslinks: Ž1. all chains were of equal length Žthe effect of molecular weight distribution was not considered.; Ž2. a segment was defined by any chain connected by two junction points; and Ž3. a junction point was defined as a point were at least three segments
Table 1 Physical properties of phenolicrepoxy networks Epoxy
Phenolicr epoxy Žwrw.
Phenolicr epoxy Žeq.req..
Tg Ž8C.
Modulus at Tg q 608C ŽPa.
Mc Žmeasured.
Mc Žtheoretical .
1 1 1 2 2 3 3 4 4
50r50 65r35 80r20 50r50 65r35 65r35 80r20 65r35 80r20
1.8r1 3.3r1 7.2r1 3.1r1 5.8r1 3r1 6.4r1 3.3r1 7.2r1
151 127 114 148 130 118 109 87 96
2.09= 107 9.35= 106 2.87= 106 1.10= 107 4.43= 106 1.53= 107 4.99= 106 1.12= 107 2.97= 106
643 1413 4539 1554 3511 872 2622 1030 4051
492 3564 ` 2179 ` 1768 ` 3518 `
C.S. Tyberg et al. r Construction and Building Materials 13 (1999) 343]353
349
Fig. 4. Dynamic mechanical analysis ŽDMA. of phenolicrepoxy 1 networks.
intersect. The number of segments for each system was calculated based on the stoichiometric ratio of epoxy to phenol and the average functionality of the phenolic chains Ž f s 7.31.. The calculation for Mc s involves dividing the total molecular weights of all chains considered by the total number of segments. This method should correlate well to measured values for cases where very few ‘dangling ends’ Ži.e. elastically inactive segments. exist. The number of so-called ‘dangling ends’ would be minimized when the phenolic chains were either very long or for cases where the crosslink density was very high. On the other hand, since this method for calculating Mc includes the total weights of all the chains considered Žincluding the weights of the ‘dangling ends’. and it did not count those chain ends as segments, then the theoretical values should be higher than those measured from moduli data. According to rubber elasticity theory, the rubbery moduli of these networks should be proportional to the crosslink densities. Absolute values for the moduli in the rubbery plateau regions were determined from creep tests at 608C above Tg using a Dynastat. The slopes of the load]displacement curves were used to calculate moduli according to Eq. Ž1., and the molecular weights between crosslinks ŽTable 1. were calculated from the elastic moduli and the densities at 608C above Tg wEq. Ž3.x. Eq. Ž3. was derived to apply to elastically active network chains only Žnot dangling ends or unconnected chains. w30x and depends on the following four basic assumptions: Ž1. the chains have gaussian statistics; Ž2. the material deforms affinely; Ž3. the internal energy of the system is independent of the conformations of the individual chains; and Ž4. the chains are treated as phantom networks Žthere is no excluded volume. w30x. The experimental trend in Mc values follows the expected increases with stoichiometric offset Žhigher ratios of phenols to epoxy groups. in all cases. The experimental Mc values, in most cases, are lower than those predicted from stoichiometry ŽTable 1.. This difference can be partially attributed to the
distribution of molecular weights of the phenolic oligomers, which was not considered in the theoretical predictions, and the fact that as the ratio of phenols to epoxies is increased, the number of chains which do not contribute to networking increases Ži.e. the number of dangling ends increases .. Therefore as the offset increases, the molecular weight between crosslinks increases significantly due to the large increase in the number of phenolic chains unconnected Žin the network. and the increased fraction of dangling ends. Dynamic mechanical analyses were used to determine both the glass transition temperatures ŽTg s. and to compare the relative elastic moduli. The Tg s were calculated from the peaks in the Tan delta curves ŽTable 1.. Results of these DMA analyses show that as the stoichiometric offset Žexcess of phenols. was increased, both the Tg s and the elastic moduli decreased ŽFig. 4.. This was the expected trend due to the expected decrease in network density in going from 2r1 Žphenolrepoxy. to 7r1 Žphenolrepoxy.. Fracture toughness of the networks was measured by Table 2 Sol fraction and fracture toughness of phenolicrepoxy networks Epoxy
Phenolicr epoxy Žeq.req..
1
Control Žepoxy 1 cured wrDDS. 1r1 2r1 3r1 7r1 3.1r1 5.8r1 3r1 6.4r1 3.3r1 7.2r1
1 1 1 1 2 2 3 3 4 4
Phenolicr epoxy Žwrw.
Sol fraction
K1c ŽMPa]m1r2 .
]
]
0.62
36r64 53r47 63r37 80r20 50r50 65r35 65r35 80r20 65r35 80r20
2.6 3.4 12.5 24.5 ] ] ] ] ] ]
0.57 0.64 0.85 0.70 0.84 0.74 0.87 0.41 0.77 0.62
C.S. Tyberg et al. r Construction and Building Materials 13 (1999) 343]353
350
Table 3 Flame retardance of phenolicrepoxy networks Epoxy
Phenolicr epoxy Žwrw.
Peak heat release rate ŽkWrm2 .
1
Control epoxy ŽEpon 828rDDS. 35r65 50r50 65r35 80r20 50r50 65r35 65r35 80r20 65r35 80r20
1230
5
44
477 382 357 260 158 165 397 266 325 226
17 23 29 33 9 8 17 26 24 15
45 36 34 27 175 189 76 53 24 15
1 1 1 1 2 2 3 3 4 4
determining the plane-strain stress intensity factors, K 1c . Higher values of K 1c indicate improved resistance to crack propagation, i.e. increased toughness. Toughness increases with the increased offset in stoichiometry to a maximum of 0.85 Mpa]m1r2 for the 3r1 phenolrepoxy 1 networks. This is consistent with the drop in moduli and decrease in network density. The toughness decreases slightly when the stoichiometric offset becomes too large Ž7r1 phenolrepoxy.. This is undoubtedly related to the increase in dangling ends and unconnected phenolic chains at these very high phenol to epoxy ratios. Most of the materials with a phenolic content of 50 wt.% or greater have significantly higher toughness than the epoxy control Žepoxy 1 stoichiometrically cured with p, p9-diaminodiphenylsulfone.. This is considered a breakthrough because
Char yield Žwt.%.
Smoke toxicity ŽCO yieldrCO2 yield. Ž=10y3 .
the commercially cured phenolic resins are much more brittle than epoxies. The K 1c values for the 65r35 Žwrw. networks using epoxies 2, 3, and 4 are also relatively high ŽTable 2.. Thus, it can be reasoned that phenolic-epoxy mechanical properties using this particular phenolic oligomer can be optimized by controlling stoichiometry at approximately 3r1 phenol to epoxy. Lower ratios lead to increased brittleness associated with higher network densities, and higher ratios result in too many dangling ends and unconnected chains for optimum mechanical properties. It would be interesting to compare these same values using a phenolic oligomeric precursor with a higher average functionality since this would reduce the inactive chain segments as stoichiometric offset is increased. The flame retardance of neat panels of the
Fig. 5. Heat release rate curves of Ža. epoxy 1 cured wrDDS and Žb. 80r20 phenolicrepoxy 1 network.
C.S. Tyberg et al. r Construction and Building Materials 13 (1999) 343]353
phenolicrepoxy networks was measured using a cone calorimeter with a heat flux of 50 kWrm2 and 20.9% O 2 Žatmospheric oxygen.. The heat release rate curves from the phenolicrepoxy networks show a much lower peak heat release rate and lower total heat released than the epoxy control ŽFig. 5.. The heat release rate of the 80r20 wrw phenolicrepoxy 1 network never reaches 300 kWrm2 ; whereas, the epoxy control has a maximum heat release rate of approximately 1200 kWrm2 ŽTable 3.. In addition to the different maximum heat release rate values another notable difference is evident from the heat release rate curves ŽFig. 5.. It can be speculated that the initial spike in the heat release rate curve of the phenolicrepoxy networks may be a result of the buildup of volatiles from thermal degradation prior to ignition. The consumption of these volatiles may cause the immediate spike as the sample ignites, followed by a reduction in the heat release rate before the gradual increase due to the burning of the material. It is possible that this spike is not seen in the control epoxy because the initial burning is much more rapid, overshadowing this effect. Networks with the brominated epoxy 2 have the lowest heat release rates, but the smoke produced has a much higher toxicity. On the other hand, networks with the siloxane epoxy 4 have a low heat release rate as well as very low smoke toxicity ŽTable 3.. From the cone calorimetry data we know that the networks have a high degree of flame retardance, but we do not fully understand what the factors involved in the flame retardance are. One possibility is that the free phenolic groups on the Novolac may act as free radical traps to retard degradation at the burning front. It was reasoned that such behavior might result in the feed of volatiles to the flame being reduced. Another possibility is that the high aromatic content of the Phenolic resin leads to a high char yield at the burning surface. The char would not burn readily, and might also shield heat transfer from the flame to the sub-
351
strate. High char yields for the networks high in phenolic content were observed by thermogravimetric analyses. These thermograms show that materials with a higher content of phenolic material in the network have higher char yields. Moisture uptake of a series of materials with varied stoichiometries was measured for the phenolicrepoxy 1 networks and compared to an epoxy control. These samples were immersed in water at room temperature for several months until equilibrium was achieved ŽFig. 6.. The phenolicrepoxy showed significantly increased moisture uptake when compared with the control epoxy. In addition, the moisture uptake increased significantly as the phenolic content increased, as a result of the higher concentration of hydroxyl groups. These results can be explained by increased hydrogen bonding of phenolicrepoxy networks as the phenolic concentration is increased. The phenolicrepoxy material with the highest network density and correspondingly the lowest concentration of hydroxyl groups Žincluding both aliphatic hydroxyls and phenols., 50r50 wrw phenolicr epoxy 1, was closest to the epoxy control as was expected. Future work in this area will include investigations into methods to reduce the moisture uptake of these systems. 3.1. Composite properties Composites made from the phenolicrepoxy 1 resins exhibited superior transverse and longitudinal flexural strengths and moduli compared to the epoxy 1 control ŽFig. 7.. The panels prepared from the 50r50 and 70r30 wrw phenolicrepoxy compositions exhibited the highest values, consistent with the superior toughness of these materials which was measured for the neat networks. The excellent transverse flexural properties of the phenolicrepoxy compositions may also reflect improved fiberrmatrix adhesion. This might be expected due to the unusually high potential for hydro-
Fig. 6. Water absorption of novolacrepoxy 1 networks.
352
C.S. Tyberg et al. r Construction and Building Materials 13 (1999) 343]353
Fig. 7. Transverse and longitudinal flexural test results of composite panels Žepoxy 1 control cured with DDS, 50r50 wrw phenolicrepoxy 1, 65r35 wrw phenolicrepoxy 1, and 80r20 wrw phenolicrepoxy 1.
gen bonding which is inherent in the chemical backbone structure for the phenolic material. These issues will be subjects for further investigation.
Novolac and 20 wt.% siloxane epoxy yielded both very low heat release rates and the lowest smoke toxicities. Thus, by increasing the Novolac content up to a phenol to epoxy ratio of approximately 3, both mechanical properties and flame retardance could be maximized.
4. Conclusions Relationships between the chemical network structures and the physical and flame performance properties of new phenolic-epoxy materials have been elucidated. Network densities have been explored by measuring the moduli in the rubbery regions and these experimental values were compared with those predicted from stoichiometry. At least trends in experimental results followed predictions. It was shown that the Tg ’s decreased, and toughness increased, as the phenolic Novolac content in the network was increased. Both results could be correlated to the decrease in network densities along this series. Toughness values exceed those of typical untoughened epoxy networks and far exceed existing commercial phenolic resins. In addition, an increase in Novolac content improves the flame retardance rather dramatically. The maximum flame retardance was achieved for networks with 80 wt.% of the phenolic component, and these networks have significantly better flame retardance than the epoxy control. Networks containing 80 wt.% phenolic
5. Future work One aspect of future research in this area will be focused on increasing the processing window of the phenolicrepoxy systems for rapid ‘on-line’ composite fabrication techniques such as pultrusion or resin infusion. Other directions for pursuit will be in identifying epoxy components containing silicon and phosphorus in order to improve flame properties even further.
Acknowledgements The authors are grateful for the financial support of the NSF Science and Technology Center for High Performance Polymeric Adhesives and Composites at Virginia Tech ŽDMR 9120004., to the Dow Chemical Co. and to the GenCorp Foundation. They are also grateful to the Dow Chemical Co. and to Georgia] Pacific Corp. for materials donations and for close technical cooperation.
C.S. Tyberg et al. r Construction and Building Materials 13 (1999) 343]353
References w1x w2x w3x w4x w5x w6x w7x w8x w9x
w10x w11x
w12x w13x w14x w15x
Hshieh F, Beeson HD. Fire Mater 1997;21:41]49. Antony R, Pillai CKS. J Appl Polym Sci 1994;54:429]438. Kunz HE. Macromol Chem Macromol Symp 1993;74:155]164. Dailey TH Jr. Plast Eng, 1989. Sorathia U, Lyon R, Ohlemiller T, Grenier A. A review of fire test methods and criteria for composites. Sampe J 1997; 33Ž4.:23]31. Scudamore MJ. Fire performance studies on glass-reinforced plastic laminates. Fire Mater 1994;18Ž5.:313]325. US Pat 3 352 744 to Elmer C, Mestdagh JJ, Monsanto Co. Self-extinguishing phenolic resin compositions and laminates prepared therefrom. 11r14r67. US Pat 3 369 056 to Schwarzer CG, Shell Oil Co. Flame retardant phenolic resin compositions. 2r13r65. Fire and smoke resistant interior materials for commercial transport aircraft. NMAB -477]1. National Materials Advisory Board Commission on Engineering and Technical Systems. Washington, DC: National Research Council, National Academy Press, 1995. Brown JR, St. John NA. Fire-retardant low-temperature-cured phenolic resins and composites. Trend Polym Sci 1996; 4Ž12.:416]420. Knopp A, Bohmer V, Pilato LA. Comprehensive polymer science, the synthesis, characterization, reactions and applications of polymers, chap 35, 5. New York: Pergamon Press, 1989. p. 35. Mark, Bakales, Overberger, Menges. Encyclopedia of polymer science and engineering, vol. 11, Chap Phenolic Resins. New York: John Wiley and Sons, 1988. Zhang X, Looney MG, Solomon DH, Whittaker AK. Polymer 1997;38Ž23.:5835]5848. Klett MW, Dailey TH Jr, Allison R. New advancements in phenolic resin pultrusion. Proc. 25th Internat’l. SAMPE Tech. Conf., 1993. Branco CM, Ferreira JM, Fael P, Richardson MOW. A comparative study of the fatigue behaviour of GRP hand lay-up and pultruded phenolic composites. Int J Fatigue 1995; 18Ž4.:255]263.
353
w16x US Pat 4 403 066 to Brode GL, Chow S-W, Hale WF, Union Carbide Corp. Composites made from liquid phenol formaldehyde resins. 9r6r83. w17x US Pat 4 419 400 to Hindersinn RR, of Occidental Chemical Corp. Pultruded reinforced phenolic resin products. 12r6r83. w18x Wu H-D, Wu Y-D, Su Y-F, Ma C-CM. Pultruded fiber-reinforced polyurethane-toughened phenolic resin. II. Mechanical properties, thermal properties, and flame resistance. J Appl Polym Sci 1996;62:227]234. w19x Thorning H. Phenolic matrices used for pultruded profiles. Proc. of the 17th Reinforced Plastics Congress, 1990. p. 284]287. w20x Walton G. Manufacturers tackle phenolic processing challenges. High Perf Comp, 1998:34]38. w21x Hshieh F, Beeson HD. ASTM Spec Tech Publ 1995;STP 1267:152]167. w22x Hale A, Macosko CW. J Appl Polym Sci 1989;38:1253]1269. w23x Bair HE. Proceedings of the 21st North American thermal analysis society conference. Atlanta, GA, 1992. p. 60]71. w24x Ogata M, Kinjo N, Kawata T. J Appl Polym Sci 1993; 48:583]601. w25x Biernath RW, Soane DS. Contemp Topic Polym Sci 1992; 7:103]159. w26x Ghassemi H, Shobha H.K., Sandarapandian M, et al. Volatilefree phenolic networks for infrastructure. In: Saadatmanesh H, Ehsani M, editors. Fiber composites in infrastructure, vol. 1. U. of Arizona: Engineering Professional Development, 1997. p. 46]56. w27x Sankarapandian M, Shobha HK, Sensenich C et al. Void-free phenolic networks. Proc. 43rd International SAMPE Symposium and Exhibition. Anaheim, CA, 1998. w28x US Pat 4 689 383 to Riffle JS, Yilgor I, Thoratec Laboratories Corp. Hydroxyl-functional disiloxanes and polysiloxane oligomers. 8r25r87. w29x Standard Test Methods for Plane-Strain Fracture Toughness and Strain Energy Release Rate of Plastic Materials Designation: D 5045-91. w30x Aklonis JJ, MacKnight WJ. Introduction to polymer viscoelasticity, 2nd ed. New York: John Wiley and Sons, 1983. p. 111.
Tough, void-free, flame retardant phenolic matrix materials C.S. Tyberg a , M. Sankarapandian a , K. Bears a , P. Shiha , A.C. Loosa , D. Dillard a , J.E. McGrath a , J.S Riffle a,U , U. Sorathiab a
NSF Science and Technology Center for High Performance Polymeric Adhesi¨ es and Composites, Virginia Polytechnic Institute and State Uni¨ ersity, Blacksburg, VA 24060, USA b Na¨ al Surface Warfare Center, 9500 MacArthur Bl¨ d., West Bethesda, MD 20817, USA Received 25 April 1998; received in revised form 26 April 1999; accepted 2 May 1999
Abstract Polymer matrix composites may be effective alternatives to steel and concrete in many civil structures due to their anticipated superior environmental stability. However, one major obstacle limiting the use of organic materials in construction is their high combustibility. Phenolic resins are high volume materials widely utilized for many purposes, including structural adhesives, composites, coatings, etc. They are also well known to provide fire resistance in many applications. Normally, their curing process involves reagents which produce volatile by-products such as water, formaldehyde, or ammonia. This feature has severely limited processability and their utilization in void-free structural applications. This paper is focused on a catalyzed epoxy-phenolic reaction utilizing commercially available phenolic resins and liquid epoxies to promote a predominately Ž50]90% wrw. phenolic network which produces little or no volatiles. Both ductility and flame resistance are enhanced in compositions containing these high phenolic concentrations. The materials appear to have significant potential for adhesives and structural composites, wherein the absence of voids could permit significant new products to be prepared for infrastructure and other applications. Relationships between the chemical and network structures of these materials and their properties are described. Q 1999 Elsevier Science Ltd. All rights reserved. Keywords: Composites; Flame retardant; Phenolic matrix materials
1. Introduction Lightweight, stiff polymer matrix composites may be effective alternatives to steel and concrete in many civil structures. It is anticipated that polymer matrix composites’ superior oxidative resistance relative to steel and better freeze]thaw durability relative to concrete could lead to structural components with the needed improved environmental resistance. However, one major obstacle limiting the use of organic materials in construction is their high combustibility. U
Corresponding author. Department of Chemistry, Virginia Polytechnic Institute, 3107 Hahn Hall, West Campus Drive, Blacksburg, VA 24061-5976, USA. Tel.: q1-540-231-8214; fax: q1-540-231-8517.
Phenolic resins, both Novolacs and Resols, are widely used commercially due to their excellent flame retardance and low cost w1]10x. There has also been significant academic interest in understanding the mechanisms of curing and decomposition of the phenolics in order to understand the reasons for their excellent flame retardance w2x. As a result of the low smoke generation of phenolic resins they are being used in the interiors of planes such as the DC-10 w3x. Phenolics have a broad range of applications varying from construction to electronics and aerospace w11,12x. Commercial Novolac resins are cured with reagents such as hexamethylenetetramine ŽHMTA. which yields networks with relatively high crosslink densities w12,13x. Curing with HMTA produces volatile by-products such
0950-0618r99r$ - see front matter Q 1999 Elsevier Science Ltd. All rights reserved. PII: S 0 9 5 0 - 0 6 1 8 Ž 9 9 . 0 0 0 2 2 - 7
344
C.S. Tyberg et al. r Construction and Building Materials 13 (1999) 343]353
as water, formaldehyde, and ammonia, which lead to voids in the materials w14]20x. This has been a negative feature in phenolic matrix composites because the high void content leads to brittle components. Therefore, current commercial Novolacs are limited to applications where high strength is not a requirement. On the other hand, epoxy resins are used for applications where high strength and toughness are needed, although they are highly combustible materials. The flame retardance of epoxy resins can be improved by incorporating bromine into the structure, but bromine also increases the toxic smoke emission w21x. Phenolics have been used as curing agents for epoxy resins where the epoxy is the major component w22,23x. In many cases, these systems make up the base resins for semiconductor packaging due to their improved hydrophobicity relative to amine cured epoxies and their low cost. Effects of crosslink densities on the physical properties of the networks have been studied w24x. Unfortunately, these systems do not have good flame retardance because of the high content of combustible epoxy resin. This paper focuses on curing phenolic resins with epoxies, as opposed to HMTA ŽFig. 1. w25x. The cure reaction produces no volatiles and therefore produces tough, void-free networks. Network density can be at least somewhat controlled by the stoichiometric offset between phenol and epoxy groups. The approach is to utilize networks high in phenolic content Žup to ; 80 wt.%. so that the flame retardant properties of the phenolic material are retained while tailoring mechanical properties via crosslink densities and molecular structure. In addition, these systems should be much tougher than the phenolic Novolac resins cured with
HMTA as a result of the lack of voids in the networks. The intent is to prepare materials as strong as epoxy networks which retains the flame retardance of phenolic resins w26,27x.
2. Experimental section 2.1. Materials The commercial phenolic resins were provided by Georgia]Pacific ŽProduct no. GP-2073.. The phenolic resin used had a molecular weight of ; 741 grmol. Epon 828 epoxy resin Ž1. was obtained from Shell Chemical. XUR-1544-55284-35 epoxy Ž2. and DER 354 epoxy Ž3. were obtained from Dow Chemical Co. ŽFig. 2.. Triphenylphosphine ŽTPP. was obtained from Aldrich and used as received. 1,1,3,3-Tetramethyldisiloxane was obtained from United Chemical Technologies, Inc. and used as received. Allyl glycidyl ether and hydrogen hexachloroplatinate ŽIV. hydrate were obtained from Aldrich and used as received. 2.2. Synthesis of 1,3-bis(3-glycidoxypropyl)tetramethyldisiloxane [28] To a 250-ml three-neck flask equipped with a dropping funnel, nitrogen inlet, thermocouple, and magnetic stirrer, was added allyl glycidyl ether Ž87.56 g. and hydrogen hexachloroplatinate ŽIV. hydrate Ž0.0083 g.. This solution was stirred at 258C for 45 min to activate the catalyst. Tetramethyldisiloxane Ž39.58 g. was added to the dropping funnel. After 45 min a few drops were added to the allyl glycidyl etherrcatalyst solution. This
Fig. 1. Curing of phenolic novolac with epoxy.
C.S. Tyberg et al. r Construction and Building Materials 13 (1999) 343]353
345
Fig. 2. Epoxy resin structures Žbisphenol-A based epoxy Ž1., brominated bisphenol-A based epoxy Ž2.; bisphenol-F based epoxy Ž3.; and siloxane epoxy Ž4..
was stirred at 25]308C until a temperature increase of ; 58C due to the reaction exotherm was observed. At this point the tetramethyldisiloxane was added dropwise in order to control the exotherm to less than 458C. After all the siloxane was added, the solution was allowed to cool to room temperature. The remaining allyl glycidyl ether was vacuum stripped at 608C and 2 torr pressure. The epoxy was then vacuum distilled at 2008C and 2 torr pressure. 2.3. Solution mixing of the epoxy and phenolic components with catalyst The epoxy component Ž1, 2, 3, or 4., Novolac, and catalyst were weighed into a round bottom flask, then stirred in acetone Ž1r10 wrw polymerrsolvent. until dissolved ŽFig. 2.. The solvent was evaporated under nitrogen, and the sample was placed in a vacuum oven at room temperature for 24 h. The sample was then heated above Tg at 608C for 2]3 h. Samples prepared in this manner were used for melt viscosity measurements. 2.4. Preparation of fire test panels and fracture toughness bars with epoxies 1, 3, and 4 To a three-neck round bottom flask equipped with a vacuum tight mechanical stirrer and a vacuum adapter
was added phenolic Novolac and ; 75% of the required epoxy. The flask was heated in an oil bath to 1708C. When the Novolac began to melt at 1708C, mechanical stirring was begun. The vacuum was applied incrementally to prevent the material from swelling into the vacuum line. Once final vacuum was achieved Ž2]5 torr. the solution was stirred for approximately 20 min to degas the blend. During this time the remaining epoxy, with the catalyst Ž0.1]0.5 wt.% based on the total weight of epoxy. dissolved in it, was degassed in a vacuum oven at ; 808C. The stirring solution was cooled to between 130 and 1508C, depending on the viscosity of the blend being prepared, and the vacuum was temporarily released to add the remaining epoxy with catalyst. This was stirred for approximately 4 min to fully degas the samples. The melt was then poured into a mold Žwhich had been preheated to the initial cure temperature of the material. in a preheated oven and covered with mold release cloth and a heavy metal plate. The samples with epoxy 1 and 0.1 wt.% triphenylphosphine based on the weight of the epoxy were cured at 1808C for 1 h and then 2008C for 1 h. The samples with epoxy 3 and 0.1 wt.% TPP Žbased on the weight of epoxy. were cured at 1808C for 1 h, 2008C for 1 h, and 2208C for 1.5 h. The samples with epoxy 4 and 0.25 wt.% TPP Žbased on the weight of epoxy. were cured at 2008C for 1.5 h and 2208C for 2 h. The cure
346
C.S. Tyberg et al. r Construction and Building Materials 13 (1999) 343]353
schedules were determined by isothermal differential scanning calorimetry ŽDSC. measurements of uncured samples as well as dynamic mechanical analysis ŽDMA. to determine Tg values of the cured samples. The samples cured by the indicated cure cycles showed no change in Tg with additional heating, indicating that the samples were fully cured. Each phenolicrepoxy system required different cure cycles based on the cure rate of the material. 2.5. Preparation of fire test panels and fracture toughness bars with epoxy 2 After samples were mixed by the solution method and dried in a vacuum oven, the solid blends were crushed to powder. The powders of each composition were placed in molds and heated to ; 1108C for approximately 10 min in a vacuum oven to melt and degas the samples. The molds consisted of the appropriate shape grooves of 3.18= 6.35= 38.1 mm for fracture toughness bars and 10 = 10 cm = 6.3 mm for fire test panels; in addition, an extra 1-cm border was added around the perimeter of each mold to prevent the samples from bubbling out of the molds during the degassing process. The molds were then removed from the vacuum oven and placed in a preheated oven to cure the samples. The samples were heated at 1808C for 1 h and 2008C for 1 h. This method produced void free samples for fracture toughness and fire testing. 2.6. Hot-melt prepregging and composite fabrication A lab scale Model 30 prepregger manufactured by Research Tools Corporation, Ovid, MI, USA was used in composite preparation ŽFig. 3.. In this apparatus, a 36K sized AS4 carbon fiber tow was passed through a wedge-slit die at the bottom of a heated resin pot containing the matrix resin. The wetted tow was then passed between a pair of flattening pins and around a guide roller before being wound on a drum. The flattening pins and the guide rollers were independently heated. Unidirectional carbon fiber prepregs were prepared with various uncatalyzed phenolicrepoxy 1 resin systems using the hot-melt technique. The set-point temperature of the resin pot, flattening pin, and roller were determined by viscosity data. The set-point temperature was 1208C for the 50r50 wrw phenolicrepoxy 1 and 1408C for the 70r30 wrw and 80r20 wrw phenolicrepoxy 1 resins. These high temperatures were necessary to give a low enough melt viscosity to process the materials. Low melt viscosity of the resin systems is critical to permit good wet-out of the reinforcing fiber tows and yield uniform resin content. The prepregs are then cut and placed in a metal mold and cured with a thermal cycle of 2308C for 4 h to make composite panels. The volume fraction of fibers in the composites,
Fig. 3. The hot]melt prepregging process.
calculated based on densities of the fiber and resin, was 59]63% fiber. 2.7. Measurements Differential scanning calorimetry measurements were carried out using a Perkin]Elmer differential scanning calorimeter, model DSC-7. Samples of approximately 8]12 mg were placed in aluminum pans and cured using a dynamic temperature scan from 30 to 2608C with a heating rate of 58Crmin. The peak in the reaction exotherm was used to determine cure temperature and rate. The temperature calibration was performed using indium and zinc as standards. Thermogravimetric analyses were carried out using a Perkin]Elmer Thermogravimetric Analyzer model TGA-7. Samples of approximately 5 mg were placed in an aluminum pan, and heated at 108Crmin from 30 to 6008C under nitrogen. Dynamic mechanical analyses were used to determine the glass transition temperature and the relative trends in rubbery moduli. These measurements were performed using a Perkin]Elmer dynamic mechanical analyzer, model DMA-7. The Tg values were calculated from the peaks in the Tan delta curves. The analyses were conducted using a three-point bend setup with rectangular specimens. The test was run with tension control set at 110% using amplitude control adjusted to fix the room temperature modulus to 1 = 10 9 Pa. Fully cured samples were heated at 58Crmin from 25 to 2008C. Soxhlet extractions with methyl ethyl ketone ŽMEK. as the solvent were used to determine the sol fractions
C.S. Tyberg et al. r Construction and Building Materials 13 (1999) 343]353
of the networks. Cellulose thimbles were soaked in MEK for 24 h then dried in the vacuum oven at room temperature for 24 h. They were subsequently heated to 1408C for approximately 12 h. After the samples were allowed to cool to room temperature in the vacuum oven, they were weighed within 30 s of releasing the vacuum and the weight was recorded. Samples were extracted for 72 h, placed in the vacuum oven at room temperature for 24 h, then heated to 1408C under vacuum for approximately 12 h and allowed to cool to room temperature under vacuum. After being removed from the vacuum oven, the thimbles containing sample were weighed within 30 s to prevent any significant weight gain due to moisture absorption. Flame retardance was directly measured using cone calorimetry on void-free neat panels Žno fiber. in a horizontal orientation. Panels with a surface area of ; 0.01 m2 and 6.35 mm thick were used with a radiant heat flux of 50.0 kWrm2 . Three-point bend tests were utilized to characterize the toughness of the phenolic-epoxy networks in terms of the critical stress intensity factor K 1c , according to ASTM standard D 5045-91 w29x. The specimens had thickness b Ž3.12 mm. and width w Ž6.28 mm.. The single edge notch bending method was used. First, a sharp notch was created in the sample by sawing. A natural crack was initiated by inserting a cold razor blade Žwhich had been immersed in liquid nitrogen. into the notch and tapping with a rubber hammer. The depth of the crack Ž a. was between 40 and 60% of the width Ž w .. The pre-cracked notched specimen was loaded crack down, into a three-point bend fixture and tested using an Instron model 4204 instrument. The single edged notched bending rig had moving rollers to avoid excess plastic indentation. The three-point bend fixture was set up so that the line of action of applied load passed midway between the support roll centers within 1% of the distance between these centers. The crosshead speed was 1.27 mmrmin, and the testing was conducted at room temperature. Elastic moduli were determined using a Dynastat calibrated using digital calipers accurate to 0.01 mm. The Dynastat was designed for materials testing: performing creep, relaxation, and dynamic tests. It is capable of loading a small specimen in tension or compression up to 10 kg. Displacement control is also supported, as in a relaxation test, and the full range is 10 mm. The oven temperature is controlled automatically, with a range of y150]2508C. The test specimens, with dimensions of height h Ž3.18 mm., width w Ž6.35 mm., and length l Ž38.1 mm., were placed on two flat supports with a span of 2.54 cm, in the three point bend setup. The samples were heated to 608C above Tg to measure the elastic moduli. Once the sample reached the set temperature a small load Ž0.01 kg. was placed on the sample and the displacement at equilibrium was
347
measured. Equilibrium was obtained in - 3 s, indicating that the material was in the rubbery plateau region at the testing temperature of Tg q 608C. The load was increased by increments of 0.01 kg up to approximately 0.08 kg and at each load the equilibrium displacement was recorded. From this data the load]displacement curves were generated and linear regression analysis was performed to determine the slopes of the lines. These slopes were used to determine the moduli according to Eq. Ž1.. The displacement values measured from the Dynastat are accurate to within 0.05 mm. Es g Ž PrD .Ž L3r48 I . PrD g L I
s s s s
Ž1.
slope of load]displacement data gravitational constant s 9.81 mrs 2 length between supports s 2.54 cm Ž1r12. wb 3
Room temperature density measurements were performed using a Mettler]Toledo AG204 balance adapted with a Mettler]Toledo density determination kit for ATrAG and PGrPR balances. Rectangular samples with dimensions of approximately 19 = 6.35= 3.18 mm were sanded, then polished, to prevent any trapping of air bubbles. Deionized water was degassed in the vacuum oven at 238C for approximately 30 min prior to use. Using this setup, the weight of the solid in air ŽA. and the weight of the solid in water ŽB. were measured. The temperature of the water was recorded to within 0.18C and the density of distilled water at that temperature was determined using a density table. Room temperature density was calculated according to Eq. Ž2..
r s Ar Ž A y B .4 r Ž H 2 O .
Ž2.
The density at 608C above Tg was calculated using the coefficients of thermal expansion below Tg and above Tg , measured by Thermal Mechanical Analysis using the large quartz parallel plate measuring system in TMA mode. Using these densities and the elastic moduli Ždetermined from Dynastat measurements . the molecular weights between crosslinks Ž Mc . were calculated by Eq. Ž3.. Mc s 3 RTrrE9
Ž3.
Room temperature water absorption was determined for phenolicrepoxy Ž1. networks cured with 0.1 wt.% TPP based on the weight of epoxy and compared to an epoxy control cured with diaminodiphenylsulfone at 1808C for 2 h and 2208C for 2 h. Rectangular samples with dimensions of approximately 19.5= 6.4= 1 mm were sanded and polished to remove any flaws which might affect the results. Two samples of each compo-
C.S. Tyberg et al. r Construction and Building Materials 13 (1999) 343]353
348
sition were accurately weighed then placed in vials filled with water at room temperature. The samples were periodically removed from the water, dried with a Kimwipe, then weighed in order to measure the moisture uptake. The sample weight and the amount of time elapsed from the start of the experiment were recorded. The time was normalized for the varied thickness by dividing by the sample thickness. For each composition, the average moisture uptake of the three samples was recorded and plotted vs. the square root of the time in hours divided by the thickness. Data was collected until the curves leveled off. Rheology experiments were carried out on a Bohlin VOR Rheometer in continuous oscillation mode with a frequency of 1 Hz. Temperature control was accomplished with a Bohlin HTC. The auto-strain was set to control the torque to 25% of the maximum torque allowed. The maximum strain for the instrument was 0.25. Samples of approximately 0.7 g of dry powder were pressed into a pellet then placed between the preheated 25-mm diameter parallel plates of the rheometer. The gap was closed to ; 1 mm and the sides were scraped to remove excess sample before the run was started. The samples used in these test were prepared by solvent mixing. 2.8. Composite panel characterization The composite panels were C-scanned for overall quality using a Sonix model HF 1000 by a pulse-echo arrangement using a 15-MHz transducer with a focal length of 38.1 mm. Scanning electron photomicrographs, obtained from an ISI SX-40 SEM, were used to establish the void-free nature of the composites and to confirm good resin distribution and fiber wet-out. Transverse and longitudinal flexural tests following ASTM standard D-790 were used to evaluate a systematically varied series of phenolicrepoxy 1 compositions.
3. Results and discussion A phenolic Novolac resin with an average functionality of 7.3 phenolic hydroxy groups was cured with four different epoxy resins, 1 Žbisphenol-A based epoxy., 2 Žbrominated bisphenol-A based epoxy., 3 Žbisphenol-F based epoxy., and 4 Ž1,3-bisŽ3-glycidoxypropyl.tetramethyldisiloxane. to investigate the effects of epoxy structure on network properties ŽFig. 2.. For each chemical structure of the epoxy component, systematically varied compositions were studied ranging from 50 to 80 wt.% phenolic Novolac. Structural and compositional effects on network toughness, transition temperatures, and flame retardance were investigated. In general, network toughness would be expected to increase with the average molecular weight between crosslinks Žincrease in ratio of phenols to epoxies. up to some point where the amount of unconnected phenolic chains begins to dominate properties. Likewise, glass transition temperatures of fully cured networks should decrease as the distance between crosslinks increases. Thus, it was of interest to predict and measure network densities and correlate these chemical structures to physical properties of the networks. ‘Theoretical’ average distances between crosslinks were predicted by considering the stoichiometries and assuming a linear phenolic chain ŽTable 1.. In considering the chemical structures of these networks, it was predicted that the crosslink densities should decrease as the stoichiometric offset Žexcess of phenols. was increased. These predicted values were then compared to experimental measurements of crosslink densities derived from rubbery moduli and elasticity theories. The following assumptions were made in order to calculate the predicted molecular weights between crosslinks: Ž1. all chains were of equal length Žthe effect of molecular weight distribution was not considered.; Ž2. a segment was defined by any chain connected by two junction points; and Ž3. a junction point was defined as a point were at least three segments
Table 1 Physical properties of phenolicrepoxy networks Epoxy
Phenolicr epoxy Žwrw.
Phenolicr epoxy Žeq.req..
Tg Ž8C.
Modulus at Tg q 608C ŽPa.
Mc Žmeasured.
Mc Žtheoretical .
1 1 1 2 2 3 3 4 4
50r50 65r35 80r20 50r50 65r35 65r35 80r20 65r35 80r20
1.8r1 3.3r1 7.2r1 3.1r1 5.8r1 3r1 6.4r1 3.3r1 7.2r1
151 127 114 148 130 118 109 87 96
2.09= 107 9.35= 106 2.87= 106 1.10= 107 4.43= 106 1.53= 107 4.99= 106 1.12= 107 2.97= 106
643 1413 4539 1554 3511 872 2622 1030 4051
492 3564 ` 2179 ` 1768 ` 3518 `
C.S. Tyberg et al. r Construction and Building Materials 13 (1999) 343]353
349
Fig. 4. Dynamic mechanical analysis ŽDMA. of phenolicrepoxy 1 networks.
intersect. The number of segments for each system was calculated based on the stoichiometric ratio of epoxy to phenol and the average functionality of the phenolic chains Ž f s 7.31.. The calculation for Mc s involves dividing the total molecular weights of all chains considered by the total number of segments. This method should correlate well to measured values for cases where very few ‘dangling ends’ Ži.e. elastically inactive segments. exist. The number of so-called ‘dangling ends’ would be minimized when the phenolic chains were either very long or for cases where the crosslink density was very high. On the other hand, since this method for calculating Mc includes the total weights of all the chains considered Žincluding the weights of the ‘dangling ends’. and it did not count those chain ends as segments, then the theoretical values should be higher than those measured from moduli data. According to rubber elasticity theory, the rubbery moduli of these networks should be proportional to the crosslink densities. Absolute values for the moduli in the rubbery plateau regions were determined from creep tests at 608C above Tg using a Dynastat. The slopes of the load]displacement curves were used to calculate moduli according to Eq. Ž1., and the molecular weights between crosslinks ŽTable 1. were calculated from the elastic moduli and the densities at 608C above Tg wEq. Ž3.x. Eq. Ž3. was derived to apply to elastically active network chains only Žnot dangling ends or unconnected chains. w30x and depends on the following four basic assumptions: Ž1. the chains have gaussian statistics; Ž2. the material deforms affinely; Ž3. the internal energy of the system is independent of the conformations of the individual chains; and Ž4. the chains are treated as phantom networks Žthere is no excluded volume. w30x. The experimental trend in Mc values follows the expected increases with stoichiometric offset Žhigher ratios of phenols to epoxy groups. in all cases. The experimental Mc values, in most cases, are lower than those predicted from stoichiometry ŽTable 1.. This difference can be partially attributed to the
distribution of molecular weights of the phenolic oligomers, which was not considered in the theoretical predictions, and the fact that as the ratio of phenols to epoxies is increased, the number of chains which do not contribute to networking increases Ži.e. the number of dangling ends increases .. Therefore as the offset increases, the molecular weight between crosslinks increases significantly due to the large increase in the number of phenolic chains unconnected Žin the network. and the increased fraction of dangling ends. Dynamic mechanical analyses were used to determine both the glass transition temperatures ŽTg s. and to compare the relative elastic moduli. The Tg s were calculated from the peaks in the Tan delta curves ŽTable 1.. Results of these DMA analyses show that as the stoichiometric offset Žexcess of phenols. was increased, both the Tg s and the elastic moduli decreased ŽFig. 4.. This was the expected trend due to the expected decrease in network density in going from 2r1 Žphenolrepoxy. to 7r1 Žphenolrepoxy.. Fracture toughness of the networks was measured by Table 2 Sol fraction and fracture toughness of phenolicrepoxy networks Epoxy
Phenolicr epoxy Žeq.req..
1
Control Žepoxy 1 cured wrDDS. 1r1 2r1 3r1 7r1 3.1r1 5.8r1 3r1 6.4r1 3.3r1 7.2r1
1 1 1 1 2 2 3 3 4 4
Phenolicr epoxy Žwrw.
Sol fraction
K1c ŽMPa]m1r2 .
]
]
0.62
36r64 53r47 63r37 80r20 50r50 65r35 65r35 80r20 65r35 80r20
2.6 3.4 12.5 24.5 ] ] ] ] ] ]
0.57 0.64 0.85 0.70 0.84 0.74 0.87 0.41 0.77 0.62
C.S. Tyberg et al. r Construction and Building Materials 13 (1999) 343]353
350
Table 3 Flame retardance of phenolicrepoxy networks Epoxy
Phenolicr epoxy Žwrw.
Peak heat release rate ŽkWrm2 .
1
Control epoxy ŽEpon 828rDDS. 35r65 50r50 65r35 80r20 50r50 65r35 65r35 80r20 65r35 80r20
1230
5
44
477 382 357 260 158 165 397 266 325 226
17 23 29 33 9 8 17 26 24 15
45 36 34 27 175 189 76 53 24 15
1 1 1 1 2 2 3 3 4 4
determining the plane-strain stress intensity factors, K 1c . Higher values of K 1c indicate improved resistance to crack propagation, i.e. increased toughness. Toughness increases with the increased offset in stoichiometry to a maximum of 0.85 Mpa]m1r2 for the 3r1 phenolrepoxy 1 networks. This is consistent with the drop in moduli and decrease in network density. The toughness decreases slightly when the stoichiometric offset becomes too large Ž7r1 phenolrepoxy.. This is undoubtedly related to the increase in dangling ends and unconnected phenolic chains at these very high phenol to epoxy ratios. Most of the materials with a phenolic content of 50 wt.% or greater have significantly higher toughness than the epoxy control Žepoxy 1 stoichiometrically cured with p, p9-diaminodiphenylsulfone.. This is considered a breakthrough because
Char yield Žwt.%.
Smoke toxicity ŽCO yieldrCO2 yield. Ž=10y3 .
the commercially cured phenolic resins are much more brittle than epoxies. The K 1c values for the 65r35 Žwrw. networks using epoxies 2, 3, and 4 are also relatively high ŽTable 2.. Thus, it can be reasoned that phenolic-epoxy mechanical properties using this particular phenolic oligomer can be optimized by controlling stoichiometry at approximately 3r1 phenol to epoxy. Lower ratios lead to increased brittleness associated with higher network densities, and higher ratios result in too many dangling ends and unconnected chains for optimum mechanical properties. It would be interesting to compare these same values using a phenolic oligomeric precursor with a higher average functionality since this would reduce the inactive chain segments as stoichiometric offset is increased. The flame retardance of neat panels of the
Fig. 5. Heat release rate curves of Ža. epoxy 1 cured wrDDS and Žb. 80r20 phenolicrepoxy 1 network.
C.S. Tyberg et al. r Construction and Building Materials 13 (1999) 343]353
phenolicrepoxy networks was measured using a cone calorimeter with a heat flux of 50 kWrm2 and 20.9% O 2 Žatmospheric oxygen.. The heat release rate curves from the phenolicrepoxy networks show a much lower peak heat release rate and lower total heat released than the epoxy control ŽFig. 5.. The heat release rate of the 80r20 wrw phenolicrepoxy 1 network never reaches 300 kWrm2 ; whereas, the epoxy control has a maximum heat release rate of approximately 1200 kWrm2 ŽTable 3.. In addition to the different maximum heat release rate values another notable difference is evident from the heat release rate curves ŽFig. 5.. It can be speculated that the initial spike in the heat release rate curve of the phenolicrepoxy networks may be a result of the buildup of volatiles from thermal degradation prior to ignition. The consumption of these volatiles may cause the immediate spike as the sample ignites, followed by a reduction in the heat release rate before the gradual increase due to the burning of the material. It is possible that this spike is not seen in the control epoxy because the initial burning is much more rapid, overshadowing this effect. Networks with the brominated epoxy 2 have the lowest heat release rates, but the smoke produced has a much higher toxicity. On the other hand, networks with the siloxane epoxy 4 have a low heat release rate as well as very low smoke toxicity ŽTable 3.. From the cone calorimetry data we know that the networks have a high degree of flame retardance, but we do not fully understand what the factors involved in the flame retardance are. One possibility is that the free phenolic groups on the Novolac may act as free radical traps to retard degradation at the burning front. It was reasoned that such behavior might result in the feed of volatiles to the flame being reduced. Another possibility is that the high aromatic content of the Phenolic resin leads to a high char yield at the burning surface. The char would not burn readily, and might also shield heat transfer from the flame to the sub-
351
strate. High char yields for the networks high in phenolic content were observed by thermogravimetric analyses. These thermograms show that materials with a higher content of phenolic material in the network have higher char yields. Moisture uptake of a series of materials with varied stoichiometries was measured for the phenolicrepoxy 1 networks and compared to an epoxy control. These samples were immersed in water at room temperature for several months until equilibrium was achieved ŽFig. 6.. The phenolicrepoxy showed significantly increased moisture uptake when compared with the control epoxy. In addition, the moisture uptake increased significantly as the phenolic content increased, as a result of the higher concentration of hydroxyl groups. These results can be explained by increased hydrogen bonding of phenolicrepoxy networks as the phenolic concentration is increased. The phenolicrepoxy material with the highest network density and correspondingly the lowest concentration of hydroxyl groups Žincluding both aliphatic hydroxyls and phenols., 50r50 wrw phenolicr epoxy 1, was closest to the epoxy control as was expected. Future work in this area will include investigations into methods to reduce the moisture uptake of these systems. 3.1. Composite properties Composites made from the phenolicrepoxy 1 resins exhibited superior transverse and longitudinal flexural strengths and moduli compared to the epoxy 1 control ŽFig. 7.. The panels prepared from the 50r50 and 70r30 wrw phenolicrepoxy compositions exhibited the highest values, consistent with the superior toughness of these materials which was measured for the neat networks. The excellent transverse flexural properties of the phenolicrepoxy compositions may also reflect improved fiberrmatrix adhesion. This might be expected due to the unusually high potential for hydro-
Fig. 6. Water absorption of novolacrepoxy 1 networks.
352
C.S. Tyberg et al. r Construction and Building Materials 13 (1999) 343]353
Fig. 7. Transverse and longitudinal flexural test results of composite panels Žepoxy 1 control cured with DDS, 50r50 wrw phenolicrepoxy 1, 65r35 wrw phenolicrepoxy 1, and 80r20 wrw phenolicrepoxy 1.
gen bonding which is inherent in the chemical backbone structure for the phenolic material. These issues will be subjects for further investigation.
Novolac and 20 wt.% siloxane epoxy yielded both very low heat release rates and the lowest smoke toxicities. Thus, by increasing the Novolac content up to a phenol to epoxy ratio of approximately 3, both mechanical properties and flame retardance could be maximized.
4. Conclusions Relationships between the chemical network structures and the physical and flame performance properties of new phenolic-epoxy materials have been elucidated. Network densities have been explored by measuring the moduli in the rubbery regions and these experimental values were compared with those predicted from stoichiometry. At least trends in experimental results followed predictions. It was shown that the Tg ’s decreased, and toughness increased, as the phenolic Novolac content in the network was increased. Both results could be correlated to the decrease in network densities along this series. Toughness values exceed those of typical untoughened epoxy networks and far exceed existing commercial phenolic resins. In addition, an increase in Novolac content improves the flame retardance rather dramatically. The maximum flame retardance was achieved for networks with 80 wt.% of the phenolic component, and these networks have significantly better flame retardance than the epoxy control. Networks containing 80 wt.% phenolic
5. Future work One aspect of future research in this area will be focused on increasing the processing window of the phenolicrepoxy systems for rapid ‘on-line’ composite fabrication techniques such as pultrusion or resin infusion. Other directions for pursuit will be in identifying epoxy components containing silicon and phosphorus in order to improve flame properties even further.
Acknowledgements The authors are grateful for the financial support of the NSF Science and Technology Center for High Performance Polymeric Adhesives and Composites at Virginia Tech ŽDMR 9120004., to the Dow Chemical Co. and to the GenCorp Foundation. They are also grateful to the Dow Chemical Co. and to Georgia] Pacific Corp. for materials donations and for close technical cooperation.
C.S. Tyberg et al. r Construction and Building Materials 13 (1999) 343]353
References w1x w2x w3x w4x w5x w6x w7x w8x w9x
w10x w11x
w12x w13x w14x w15x
Hshieh F, Beeson HD. Fire Mater 1997;21:41]49. Antony R, Pillai CKS. J Appl Polym Sci 1994;54:429]438. Kunz HE. Macromol Chem Macromol Symp 1993;74:155]164. Dailey TH Jr. Plast Eng, 1989. Sorathia U, Lyon R, Ohlemiller T, Grenier A. A review of fire test methods and criteria for composites. Sampe J 1997; 33Ž4.:23]31. Scudamore MJ. Fire performance studies on glass-reinforced plastic laminates. Fire Mater 1994;18Ž5.:313]325. US Pat 3 352 744 to Elmer C, Mestdagh JJ, Monsanto Co. Self-extinguishing phenolic resin compositions and laminates prepared therefrom. 11r14r67. US Pat 3 369 056 to Schwarzer CG, Shell Oil Co. Flame retardant phenolic resin compositions. 2r13r65. Fire and smoke resistant interior materials for commercial transport aircraft. NMAB -477]1. National Materials Advisory Board Commission on Engineering and Technical Systems. Washington, DC: National Research Council, National Academy Press, 1995. Brown JR, St. John NA. Fire-retardant low-temperature-cured phenolic resins and composites. Trend Polym Sci 1996; 4Ž12.:416]420. Knopp A, Bohmer V, Pilato LA. Comprehensive polymer science, the synthesis, characterization, reactions and applications of polymers, chap 35, 5. New York: Pergamon Press, 1989. p. 35. Mark, Bakales, Overberger, Menges. Encyclopedia of polymer science and engineering, vol. 11, Chap Phenolic Resins. New York: John Wiley and Sons, 1988. Zhang X, Looney MG, Solomon DH, Whittaker AK. Polymer 1997;38Ž23.:5835]5848. Klett MW, Dailey TH Jr, Allison R. New advancements in phenolic resin pultrusion. Proc. 25th Internat’l. SAMPE Tech. Conf., 1993. Branco CM, Ferreira JM, Fael P, Richardson MOW. A comparative study of the fatigue behaviour of GRP hand lay-up and pultruded phenolic composites. Int J Fatigue 1995; 18Ž4.:255]263.
353
w16x US Pat 4 403 066 to Brode GL, Chow S-W, Hale WF, Union Carbide Corp. Composites made from liquid phenol formaldehyde resins. 9r6r83. w17x US Pat 4 419 400 to Hindersinn RR, of Occidental Chemical Corp. Pultruded reinforced phenolic resin products. 12r6r83. w18x Wu H-D, Wu Y-D, Su Y-F, Ma C-CM. Pultruded fiber-reinforced polyurethane-toughened phenolic resin. II. Mechanical properties, thermal properties, and flame resistance. J Appl Polym Sci 1996;62:227]234. w19x Thorning H. Phenolic matrices used for pultruded profiles. Proc. of the 17th Reinforced Plastics Congress, 1990. p. 284]287. w20x Walton G. Manufacturers tackle phenolic processing challenges. High Perf Comp, 1998:34]38. w21x Hshieh F, Beeson HD. ASTM Spec Tech Publ 1995;STP 1267:152]167. w22x Hale A, Macosko CW. J Appl Polym Sci 1989;38:1253]1269. w23x Bair HE. Proceedings of the 21st North American thermal analysis society conference. Atlanta, GA, 1992. p. 60]71. w24x Ogata M, Kinjo N, Kawata T. J Appl Polym Sci 1993; 48:583]601. w25x Biernath RW, Soane DS. Contemp Topic Polym Sci 1992; 7:103]159. w26x Ghassemi H, Shobha H.K., Sandarapandian M, et al. Volatilefree phenolic networks for infrastructure. In: Saadatmanesh H, Ehsani M, editors. Fiber composites in infrastructure, vol. 1. U. of Arizona: Engineering Professional Development, 1997. p. 46]56. w27x Sankarapandian M, Shobha HK, Sensenich C et al. Void-free phenolic networks. Proc. 43rd International SAMPE Symposium and Exhibition. Anaheim, CA, 1998. w28x US Pat 4 689 383 to Riffle JS, Yilgor I, Thoratec Laboratories Corp. Hydroxyl-functional disiloxanes and polysiloxane oligomers. 8r25r87. w29x Standard Test Methods for Plane-Strain Fracture Toughness and Strain Energy Release Rate of Plastic Materials Designation: D 5045-91. w30x Aklonis JJ, MacKnight WJ. Introduction to polymer viscoelasticity, 2nd ed. New York: John Wiley and Sons, 1983. p. 111.