Flame Retardant Thermosettable Resin Compositions

Resin

Thermosettable resins that can be used as component (a) in the composition of the invention are epoxy, N,N'-bismaleimide, cyanate ester monomers and prepolymers and mixtures thereof that can preferably be cured by conventional methods such as thermally or via radiation such as UV or electron beam, with or without the use of a catalyst, at room temperatures or above.

1. Epoxy

Epoxy monomers and prepolymers that can be used in the composition of the invention include those that are well known in the art. They are compounds in which there is present at least one linear or branched aliphatic, cycloaliphatic or aromatic structure and one or more epoxy groups, e.g., ##STR1## The most preferred epoxy compounds are the aromatic polyepoxides, including the polyglycidyl ethers of polyhydric phenols, glycidyl esters of aromatic carboxylic acids, N-glycidylaminobenzenes, and glycidylamino-glycidyloxy-benzenes.

Examples of epoxies useful in the thermosettable resin composition of the invention are described in Chemistry and Technology, "Epoxy Resins", Marcel Dekker Inc., 2nd edition, New York (May, 1988) and in U.S. Pat. No. 4,882,370. Examples of curatives, hardeners and catalysts for epoxy resins may also be found in these same references.

2. N,N'-bismaleimide

N,N'-bismaleimide monomers and prepolymers that can be used in the composition of the invention are also well known, many of which can be purchased commercially. They are compounds which comprise more than one maleimido group. Preferably, the compound has two maleimido groups. The compound has the following general formula: ##STR2## wherein: Y represents a divalent group containing at least 2 carbon atoms, preferably 2 to 6 carbon atoms, and a carbon-carbon double bond; and Z is a divalent aliphatic, cycloaliphatic, aromatic or heterocyclic group containing at least 2 carbon atoms and generally no more than 20 carbon atoms. Examples of N,N'-bismaleimides useful in the present invention and processes for their preparation can be found in U.S. Pat. Nos. 3,562,223; 3,627,780; 4,468,497; 4,946,908; 4,100,140 and 3,839,358. Examples of curatives, hardeners and catalysts for N,N'-bismaleimides can also be found in these same references.

3. Cyanate Ester

Cyanate ester monomers and prepolymers that can be used in the composition of the invention include those that are well known. Dicyanate esters are compounds having the general formula:

wherein R is a divalent aromatic hydrocarbon residue that comprises at least one aromatic moiety such as benzene, naphthalene, anthracene, phenanthrene and the like, and contains a total of up to 40 carbon atoms, including the aromatic moiety. For example, the dicyanate ester of bisphenol A, wherein the cyanate ester groups are attached in the para positions of the benzene rings, is one useful compound for the present invention.

Useful cyanate ester materials, and examples of curatives, hardeners and catalysts for these materials may be found in U.S. Pat. No. 4,608,434. Useful bismaleimide/cyanate ester combinations are described in U.S. Pat. No. 4,769,440 and U.S. Pat. No. 4,820,797.

Commercially available cyanate esters that are useful in the present invention include: AroCy.TM. B10, AroCy.TM. L10 and RTX.TM.-366 cyanate esters available from Ciba-Geigy; XU71787, a polycyanate ester of a polyphenol adduct of dicyclopentadiene, available from Dow Chemical Company; and BT 2160 resin, a prepolymer of bisphenol A dicyanate ester containing 10% by weight of 4,4'-methylenedianiline bismaleimide, available from Mitsubishi Gas Chemical Company.

Other combinations of resins, such as bismaleimide/epoxy, cyanate ester/epoxy, and bismaleimide/epoxy/cyanate esters are also of utility in the present invention.

Non-Functional Cyclophosphazene

The cyclophosphazene of the composition of the present invention does not covalently bond with the resin network. In addition, it is substituted with aryloxy groups such as phenoxy or substituted aryloxy groups. The cyclophosphazene is believed to provide an effective amount of phosphorous and nitrogen to render the thermosettable resin composition flame retardant. The quantity of cyclophosphazene employed in the resin may be varied, depending on the flame retardant and the resin system employed. In general, it is desirable that the amount of cyclophosphazene present in the resin be at least 3 parts by weight of cyclophosphazene per 100 parts by weight of resin. The amount of cyclophosphazene may be less than 3 parts by weight when additional flame retardant is employed. An upper limit for the amount of cyclophosphazene present in the resin is not critical to the invention; however, at high cyclophosphazene loadings in some resin networks, other parameters (such as glass transition temperature, T.sub.g) may be negatively impacted. A preferred upper limit for the amount of cyclophosphazene is about 50 parts by weight per 100 parts resin, more preferably 40 parts by weight.

The amount of aryloxy substitution present in the cyclophosphazene of the present invention will impact the thermal stability and moisture uptake of the resin. It is preferred that the amount of aryloxy substitution be greater than 50%, more preferably 100%, for performance applications such as resins used or processed at high temperatures or in aggressive environments. For some applications, however, acceptable performance may be achieved at aryloxy substitution levels below 50%, particularly if the remaining substituents are fluorinated alkoxy of the type OCH.sub.2 R.sub.f.

The cyclophosphazene may be a trimer, tetramer or higher cyclic. Preferably, the cyclophosphazene is a trimer or tetramer having the general formulae, respectively: ##STR3## wherein R is selected from the group consisting of R.sup.1, XR.sup.1, and OCH.sub.2 R.sub.f. R.sup.1 is selected from the group consisting of linear or branched alkyl having one to twelve carbon atoms; cycloalkyl having five to six carbon atoms; and substituted and unsubstituted aryl. X is selected from the group consisting of oxygen, sulfur, and NR.sup.2, wherein N is nitrogen and R.sup.2 is an alkyl having one to four carbon atoms. When X is oxygen, two R.sup.1 groups may be covalently bonded to form a divalent group that is selected from the group consisting of 1,8-naphthalenediyl, 4,5-naphthalenediyl and 2,2'-biphenylene. R.sub.f is a polyfluoroaliphatic group having 2 to 20 carbon atoms and up to 5 ether oxygen atoms. Preferably R.sub.f is a polyfluoroalkoxide having the structure OCH.sub.2 CF.sub.3 or OCH.sub.2 CF.sub.2 OC.sub.4 F.sub.9. It is also required that no more than three R groups comprise R.sup.1 as defined above or XR.sup.1 where X is oxygen and R.sup.1 is linear or branched alkyl.

The cyclophosphazenes are prepared as taught in the example section and references indicated therein. The physical characteristics of the aryloxy cyclophosphazenes may be easily controlled. In general, single substituent aryloxy cyclophosphazenes with ortho or para substitution can typically be isolated as solids with melting points that vary with the nature of the aryloxy side group. For instance, hexaphenoxycyclotriphosphazene has a melting point of 105.degree. C., whereas hexakis(4-phenylphenoxy)cyclotriphosphazene has a melting point of 201.degree. C. When aryloxy groups are substituted in the meta position, the melting point is significantly lower and often times the meta-substituted phenoxy cyclotriphosphazene may be a liquid at room temperature. Alternatively, the cyclophosphazenes can be synthesized to incorporate a combination of aryloxy side groups, such as tris(phenoxy)-tris(m-trifluoromethylphenoxy)cyclotriphosphazene. This is generally referred to as a co-substituent or mixed substituent aryloxy group substitution. When the synthesis of such compounds involves the sequential addition of the nucleophiles, the final structure is actually a distribution of different substituted products, represented, for example, by N.sub.3 P.sub.3 (OC.sub.6 H.sub.5).sub.6-x (OC.sub.6 H.sub.4 --CF.sub.3 -m).sub.x wherein m represents the meta isomer position and x is an integer from 0 to 6. The result is a product that tends to resist solidification and has compatibility characteristics that are effectively the average of the two different single substituent flame retardants. Thus, the processing properties and compatibility or solubility characteristics of the flame retardant may be tailored by controlling the substituted nature of the cyclophosphazene. For example, the cyclophosphazene may be designed to be melt compatible with the resin, yet phase separate upon curing. This could facilitate preservation of the glass transition temperature of the cured resin. Moreover, the nature of the non-functional side group does not seem to significantly influence the efficacy of the flame retardant. It is also possible to physically blend two or more cyclophosphazenes to tailor the process properties and compatibility of the resulting flame retardant. The blend may comprise, for example, mixtures of same or different trimer, tetramer and/or higher cyclic cyclophosphazenes.

Statement of Use

The thermosettable compositions of the invention are suitable for use as impregnating, laminating or molding resins. They can be used as sealants, insulating materials and adhesives especially as a sealant for semiconductors and other electronic circuit parts. The compositions may be used in the electronics industry to bond electronic components to circuit boards. The compositions may also be used to make the circuit boards themselves. They are particularly suitable for use in prepregs for the production of space age reinforced composite structures.

The thermosettable resin composition of the invention can be used to impregnate woven or non-woven webs, filaments, rovings or the like in the preparation of prepregs. Fibers that can be used in such prepregs include organic and inorganic fibers such as glass fibers, carbon or graphite fibers, ceramic fibers, boron fibers, silicon carbide fibers, polyimide fibers and the like, and mixtures thereof. The thermosettable composition of the invention can also be used to coat various substrate including woven and non-woven webs, films and foils of organic and inorganic materials including, for example, such organic materials as polyolefins, polyesters, polyimides, and the like and inorganic materials such as ceramics, copper, nickel, and gold.

The composition of the invention can also contain additives to modify characteristics of the cured composition. Additives that can be used include: inert fillers, pigments, and dyes. Specific examples of additives include: chopped fibers comprising ceramic, glass, boron or carbon; inorganic powders comprising kaolin, chalk, silica, antimony oxides, titanium oxide, or carbon; and solid microspheres or hollow microballoons comprising glass, ceramic, or metal. The composition of the invention may also be modified with toughening agents and other additives to render the composition suitable for use as a structural adhesive.

The thermosettable compositions of the invention are prepared by mixing the curatives, hardeners, and optionally catalysts, and modifying additives into the thermosettable resin at a temperature at which the thermosettable resin is fluid, generally from about 30.degree. C. to 150.degree. C. The mixing is generally accomplished by conventional high shear mixing devices, for example, a planetary mixer, a kneader, a vacuum mixer, a ball mill, paint mill or high speed stirrer. The resin is then typically poured into a mold and allowed to cure.

The present invention also includes the incorporation of other additional flame retardants to the thermosettable resin comprising non-functional cyclophosphazene. The additional flame retardant may work in combination with the non-functional cyclophosphazene to improve the flame retardancy of the resin in an additive or synergistic fashion. One benefit of employing additional flame retardant may be to reduce quantity requirements of the flame retardant material. This may result in improved physical resin properties at a reduced cost. For example, when the thermosettable resin is based on an epoxy material, it is possible to include nitrogen-containing heterocyclic materials, which by themselves, may have flame retardant properties. Heterocyclic materials include nitrogen-containing epoxies such as Araldite.TM. PT 810 and Aralcast.TM. XU AY 238 epoxies, both available from Ciba-Geigy. Other heterocyclic materials include triazine-containing materials such as trichlorotriazine, tri-isocyanurate materials such as melamine isocyanurate and materials derived from cyanate esters.

Objects and advantages of this invention are further illustrated by the following examples, but the particular materials and amounts thereof recited in these examples, as well as other conditions and details, should not be construed to unduly limit this invention. In the examples all parts and percentages are by weight and temperatures are in degrees centigrade unless otherwise noted.

Cyclophosphazene Synthesis:

The cyclophosphazenes described in these examples were prepared by conventional methods familiar to one skilled in the art of chemical synthesis. In general, the non-functional, aryloxy cyclophosphazenes were prepared by the reaction of unsubstituted or substituted phenol with hexachlorocyclotriphosphazene and potassium hydroxide in toluene. At temperatures between 80.degree. and 110.degree. C., water was removed from the reaction mixture by azeotropic distillation. The progress of the reaction was easily monitored by various analytical methods including HPLC, GC and .sup.31 P NMR spectroscopy.

Mixed substituent aryloxy cyclophosphazenes were prepared in a similar manner; however, sequential addition of the desired number of equivalents of the phenols was made to produce a partially substituted cyclophosphazene which was then reacted with additional unsubstituted or substituted phenol until all remaining P-Cl bonds were replaced. The progress of the reaction was easily monitored by .sup.31 P NMR spectroscopy. The reaction was considered complete when no further changes were detected in the .sup.31 P NMR spectrum of the reaction mixture. Typical chemical shifts of the aryloxy phosphazene trimers were approximately .delta.+6.5 ppm relative to phosphoric acid. The reaction mixture was filtered to remove, potassium chloride (KCl) followed by water washing and base extraction to remove residual phenols and KCl. The solvent was then removed under vacuum. Upon solvent removal, the single substituent phosphazenes typically crystallized from the oily product mixture, whereas mixed substituent cyclophosphazenes remained as viscous oils or crystallized only very slowly. Additional experimental details are provided in Example 1. Specific structures of the phosphazenes are listed in FIGS. 1-6.

Other alternative methods of preparation of these and related compounds are described by Allcock, H. R. et al in the Journal of Chemical Society, Chem Comm., Volume C4, page 149 (1970); Allcock, H. R., et al, in Inorganic Chemistry, Volume 14, page 2433 (1975); and Jeneau, et al in U.S. Pat. No. 4,576,806. Tris(2,2'-oxybiphenyl) and 1,1-dichloro3,5-bis(2,2'-oxybiphenyl) cyclotriphosphazene were prepared according to the method of Allcock, H. R., et. al. as described in Inorganic Chemistry, Volume 5, page 1016 (1966). Amine functional cyclophosphazenes, 1,1-N.sub.3 P.sub.3 (NH.sub.2).sub.2 (OR).sub.4 were prepared by the methods of Takahashi and Ishikawa described in Kobunshi Ronbunshu, Volume 47, page 727 (1990); Feister and Moellar, Journal of Inorganic Nuclear Chemistry, Volume 29, page 2731 (1967) or Allcock, H. R. et al, Organometallics, Volume 21, page 323 (1988). Anilino and phenol functional aryloxy cyclophosphazenes, N.sub.3 P.sub.3 (OC.sub.6 H.sub.5).sub.x (OC.sub.6 H.sub.4 --Y).sub.6-x where Y=NH.sub.2 or OH and x can be 0 to 6; were prepared according to the method of Allcock and Chang described in Macromolecules, Volume 24, page 993 (1991), and Gleria et al, Macromolecules, Volume 25 page 2569 (1992), respectively. Alkoxy cyclophosphazenes [NP(OC.sub.2 H.sub.5).sub.2 ].sub.3 ; [NP(OC.sub.3 H.sub.7).sub.2 ].sub.3 or [NP(iso-OC.sub.3 H.sub.7).sub.2 ].sub.3 were prepared according to the method of Fitzsimmons and Shaw described in the Journal of Chemical Society, Volume 41, page 1735 (1964).

A mixture of 700 ml of toluene, 429 g of phenol (4.56 moles), 270 g potassium hydroxide (4.82 moles), and 73.5 g of tetrabutyl ammonium bromide (as a phase transfer catalyst) were placed into a flask equipped with an addition funnel and Dean Stark trap. The mixture was heated to reflux and a toluene solution of 203 g of hexachlorocyclotriphosphazene (0.58 moles, available from Nippon Soda Co., Ltd. Japan) in 300 ml of toluene was added over a period of about 15 minutes. The mixture was heated at reflux (about 110.degree. C.) for about 12 hours with water being removed by azeotropic distillation.

The reaction mixture was analyzed using .sup.31 P NMR spectroscopy. When the .sup.31 P NMR spectrum showed a singlet at .delta.+6.5 ppm, complete Cl atom substitution had been achieved. The reaction mixture was filtered to removed potassium chloride, followed by water washing and base extraction to remove residual potassium chloride and phenol. The toluene solvent was removed by roto-evaporation to produce a light brown, viscous oil. Upon further drying, the oil solidified to produce 355 g (85% yield) of hexaphenoxy cyclotriphosphazene as a brown solid having a melting point of 98.degree.-103.degree. C. If desired, it is possible to purify the crude product further by recrystallization from a hot solution of methanol and tetrahydrofuran (9 to 1 by volume) followed by cooling the solution to 4.degree. C. for 16 hours. The purified solid is a free flowing powder with a light tan color having a melting point of 104.degree.-106.degree. C.

Examples 2-9 which follow describe the methods used to prepare epoxy, cyanate ester and bismaleimide resin systems with and without flame retardant materials. The glass transition temperature for each resin system is also provided in these examples. The sample numbers used in Examples 2-9 are referenced throughout the rest of the example section where various properties of the resin systems are evaluated.

Glass transition temperatures (T.sub.g) for cured resins were determined by thermomechanical analysis (TMA) using a TA 2940.TM. Instrument with an expansion probe with an applied force of 0.05 Newtons. Data were collected on a second heat using a temperature ramp of 5.degree. C./min from 50.degree. to 250.degree. C. The samples were protected from thermal oxidation by maintaining a positive flow of nitrogen gas through the furnace cell for the duration of the experiment. The glass transition temperatures were calculated from the inflection point of the graph of thermal expansion (ppm) vs temperature (.degree. C.).

Epoxy resin samples 1-15 were prepared by melting the desired amount of PR 500 molding resin (available from 3M Company) as listed in Table 1 with the desired flame retardant at a temperature between 75.degree. C. and 125.degree. C. After stirring and degassing, a homogeneous, viscous syrup was obtained (unless otherwise noted). The resin/flame retardant mixture was transferred to a mold and cured at 177.degree. C. for 4 hours. The mold was selected from the following list: (1) a 57 mm diameter aluminum pan; (2) a 4".times.4".times.0.125" (102 mm.times.102 mm.times.3 mm) premade glass mold or (3) a 8".times.4".times.0.125" (204 mm.times.102 mm.times.3 mm) premade glass mold. The glass molds were lightly coated with a mold release agent such as RAM.TM. 225 Mold Release available from Ram Chemical Company. The cured resin was allowed to cool to room temperature before removal from the mold.

The flame retardants employed in the resin samples are defined by reference to the structures given in FIGS. 1-6, except that Samples 14-16 comprised commercially available flame retardants as indicated in Table 1. Sample 14 comprised Hostaflam.TM. TPAP-750 flame retardant, a particulate poly(ammonium phosphate) flame retardant available from Hoechst-Celanese Corporation. Sample 15 comprised Kronitex.TM.-3600 an alkyl, aryl phosphate ester available from FMC Corporation.

Sample 16 comprised a bromine-containing epoxy that is often used in conventional epoxy systems to impart flame retardant properties. Sample 16 was prepared by mixing 35.0 g Quatrex.TM. 6410 (a brominated bisphenol A epoxy available from Dow Chemical, 49% w/w bromine) with 65.0 g EPON HPT.TM. 1079 (fluorene based epoxy available from Shell Chemical, hereinafter referred to as "FEP" or HPT.TM. 1079) at approximately 175.degree. C. to produce a clear viscous syrup. The epoxy mixture was allowed to cool to about 150.degree. C. and was mixed with 48.8 g of CAF curative, 9,9-bis(3-chloro-4-aminophenyl)fluorene, prepared according to the method of Schultz et al, U.S. Pat. No. 4,684,678. The mixture was stirred and degassed at approximately 150.degree. C. until a clear, viscous syrup was obtained. The warm, viscous syrup was poured into a mold and cured for 4 hours at 177.degree. C. followed by a one hour post cure at 200.degree. C. The cured sample was allowed to cool to room temperature before removal from the mold.

The specific formulations for Samples 1-16 are summarized in Table 1 below, along with the corresponding T.sub.g for each sample.

FEP/CAF epoxy resin systems (samples 17-19) were prepared by allowing the desired flame retardant (see Formulae 1-6) to melt into FEP at approximately 150.degree. C. with mechanical stirring. After complete mixing, the mixture was allowed to cool to about 130.degree. C. The FEP/flame retardant mixture was then mixed and degassed at this temperature with CAF to produce a homogeneous viscous syrup. While still warm and pourable, the mixture was then transferred to a mold and cured for 4 hours at 177.degree. C. followed by a 1 hour post cure at 225.degree. C. The cured specimens were then allowed to cool to room temperature before removal from the mold. Specific formulations are given in Table 2 below, along with the T.sub.g of each specimen, determined as in Example 2.

For samples 20-25, a master batch of FEP/DGEBA was prepared by mixing equal parts by weight FEP with DGEBA (diglycidylether bisphenol-A epoxy, available from Shell Chemical as EPON.TM. 828, hereinafter referred to as either DGEBA or EPON.TM. 828). The epoxy mixture was stirred at about 100.degree. C. until a clear viscous syrup was obtained. The desired amount of this epoxy mixture was mixed at about 100.degree. C. with the desired flame retardant (see Formulae 1-6) until a homogeneous viscous mixture was obtained. See Table 3 for quantities used. After the epoxy and flame retardant were mixed fully, CAF was added, and the resin was stirred and degassed at 100.degree. C. for about 5 minutes. While still warm and pourable, the mixture was then transferred to a mold and cured for 4 hours at 177.degree. C. followed by a 1 hour post cure at 225.degree. C. The cured specimens were then allowed to cool to room temperature before removal from the mold. The specific formulations are given in Table 3, along with the T.sub.g values determined as in Example 2.

The epoxy resin systems (samples 26-27) were prepared by allowing the desired flame retardant (see Formulae 1-6) to dissolve into Araldite.TM. MY-720 (N,N,N',N'-tetraglycidyl-4,4'-diaminodiphenylmethane, available from Ciba-Geigy, at 110.degree. C. to 120.degree. C. while being stirred. After complete mixing of these components, the curative DDS (4,4'-diaminodiphenylsulfone, available from Ciba-Geigy as Hardener HT-976) was added and the mixture was stirred further while being degassed. After about five minutes, the warm viscous syrup was poured into a mold and cured at 177.degree. C. for 4 hours followed by a 1 hour post cure at 225.degree. C. The specific formulations are listed in the Table 4, along with the T.sub.g values determined as in Example 2.

Fiber composites were prepared by dissolving the desired flame retardant into PR 500 resin with mechanical stirring and degassing as described for Example 2. After thorough mixing, the warm resin was transferred to a one gallon pail equipped with a resin transfer molding (RTM) pump attachment. A mold was prepared with either AS4-PW carbon fabric or 7781 glass fabric; both fabrics are available from Hercules Company. The carbon fiber composites had 14 ply and [0/90].sub.7s orientation to provide a thickness of 10.2 mm (0.4"). The glass fiber composites had 6 ply and a [0/90].sub.3s orientation to provide a thickness of 10.2 mm (0.4"). The mold was closed and evacuated. The modified PR 500 resin was warmed to 70.degree. C. and injected into the mold and cured at 177.degree. C. (350.degree. F.) for 2.2 hours.

Unless otherwise noted, the cyclophosphazene flame retardants were easily processed into the resins and were fully compatible after curing to yield clear, homogeneous thermosets. The specific composite formulations are given in Table 5, along with the T.sub.g values, determined as in Example 2.

Samples 32-35 comprised epoxy resins DEN.TM.438, an epoxy novolac resin, and DER.TM.332, a DGEBA material, (both from Dow Chemical Co.) blended at a 1:1 weight ratio. Samples 34 and 35 further comprised nitrogen-containing epoxies, Araldite.TM. PT810 and Aralcast.TM. XU AY 238 hydantoin epoxy, respectively, both epoxies available from Ciba-Geigy. PT 810 comprises 1,3,5-tris(oxiranylmethyl)-1,3,5-triazine-2,4,6 (1H,3H,5H)trione. XU AY 238 comprises 5-ethyl-5-methyl-1,3-bis(oxiranylmethyl)-2,4-imidazolidinedione. The epoxy materials were first blended together, and then the epoxy blend was mixed with CAF curative and flame retardant (if any). The specific formulations of each sample are given in Table 6. The resulting mixture was heated to about 90.degree. C. and thoroughly mixed using a mechanical stirrer for about 15 minutes. The pasty mixture was then de-gassed in a heated vacuum oven at 100.degree. C. for about 5 minutes. It was then poured into a pre-heated glass mold and cured at 177.degree. C. for 4 hours.

Samples 36-37 comprised DEN.TM.438 (49.1% wt), DER.TM.332 (43.0% wt.), EPON.TM. 828 (5.9% wt.), polysulfone (2.0% wt., available as Udel.TM. P3500 from Amoco Corp.) and dicyandiamide catalyst ("DICY", available as Amicure.TM. CG-1400 from Air Products and Chemicals, Inc., 0.5% wt). OTBAF [bis 9,9-(3-methyl-4-aminophenyl)fluorene, prepared according to the method of Schultz et al, described in U.S. Pat. No. 4,684,678] was used as the curative. Sample 37 further comprised a flame retardant given in Formulae 1, Structure 1a and Aralcast.TM. XU AY 238, a nitrogen-containing hydantoin epoxy described above. The DEN.TM.438, DER.TM.332, EPON.TM.828, polysulfone, hydantoin epoxy (if any) and flame retardant (if any) were first blended together. This blend was then mixed with OTBAF and DICY in a Brabender plasti-corder at 57.degree. C. for 15 minutes to give a pasty mixture. The specific formulations are given in Table 6. Typically, a small amount of a given formulation was placed in an aluminum pan and cured in a 177.degree. C. oven for 4 hours. Alternatively, the various components were mixed by means of a mechanical stirrer at about 70.degree. C. The mixture was then de-gassed in a vacuum oven at 100.degree. C. for about 5 minutes. It was then poured into a pre-heated glass mold and cured at 177.degree. C. for 4 hours. Table 6 shows the amount of the various components used in the preparation of samples 32-37, along with the T.sub.g for each sample determined using differential scanning calorimetry (DSC) using a TA 912.TM. Instrument at a heating rate of 15.degree. C./min from 50.degree. C. to 250.degree. C.

A measured amount of the neat cyanate ester resin (AroCy.TM. B30, AroCy.TM. L10 or RTX.TM.-366 available from Ceiba-Geigy) was melted and blended with the desired flame retardant. The resulting mixture was heated in a vacuum oven at 100.degree. C. for about 5 minutes. It was then poured into a pre-heated glass mold and cured at about 180.degree. C. for 15 hours, followed by a post-cure at 250.degree. C. for 1 hour. The specific formulations and the T.sub.g values (determined using either TMA as described in Example 2 or DSC as described in Example 7) are given in Table 7.

The base bismaleimide resin was prepared by mixing 80 parts of Shell Compimide.TM. 358 resin and 20 parts of Shell Toughening Modifier TM-121. 30 g of this base resin was mixed with the flame retardant. The resulting mixture was degassed at about 150.degree. C. for about 5 minutes, and then poured into a warm glass mold and cured according to the following schedule: 2 hours at 190.5.degree. C. and 5 hours at 248.8.degree. C. The control specimen was prepared similarly without the flame retardant. The specific formulations and T.sub.g values for each sample are given in Table 8.

The thermal stability of eight resins were determined by thermogravimetric analysis (TGA) to demonstrate that the incorporation of non-functional aryloxy cyclophosphazene flame retardants into epoxy resins does not negatively affect the thermal stability of the resins.

A 15 to 30 mg sample of the resin (listed in Table 9 by reference to a sample described in previous examples) was heated in a TA 2950.TM. TGA. The TGA was programmed so that the specimen were exposed to sequential, 60 minute isotherms of 220.degree. C., 260.degree. C. and 300.degree. C. A positive flow (approximately 60 cc/min) of air was maintained through the furnace cell during the experiment. The weight losses for each sample observed after each of the 60 minute isotherms are given in Table 9.

From this experiment it can be concluded that the non-functional aryloxy phosphazenes (samples 2, 4 and 13) do not negatively impact the thermal stability of the resin. In contrast, the samples comprising alkoxy phosphazenes (Samples 7 and 8), the functional phosphazenes (Samples 10 and 11) and the brominated epoxy flame retardant (Sample 16) exhibit higher weight loss and hence possess lower thermal stability.

In this example, neat epoxy resin samples corresponding to samples 1, 2, 4, 6, 12, 17, 18, 19, 20, 21, 23-27 were burned using a vertical burn test to evaluate the ability of the phosphazenes to impart flame retardant properties to the thermoset network. The resin samples were prepared as described previously in the examples and cut into dimensions of 5".times.0.5".times.0.125" (127 mm.times.12.7 mm.times.3.18 mm). A flame source was procured according to the method described in IPC [Institute for Interconnecting and Packaging Electronic Circuits] Test Method 2.3.10. Each sample was exposed to a first flame application (1st FA, in Table 10) for 10 seconds. The flame source was removed, and the time to self-extinguish was reported. If the sample did not burn, a second flame application was applied to the sample (2nd FA, in Table 10) for another 10 seconds. The flame source was removed and the time to self-extinguish was reported. If after these two flame exposures the sample did not support combustion, the flame was applied a third time (3rd FA, Table 10) until sustained ignition occurred and the time to ignition was reported. In all cases, the burn performance was compared to a control sample that contained no flame retardant. Burn time values reported in Table 10 represent the average of at least three samples. The structures of the cyclophosphazene flame retardant additives are shown in Table 10 by reference to Formulae 1-6.

From these data it can be concluded that the resin samples which contained the non-functional phosphazenes (samples 2, 4, 9, 12, 18, 21, 27) were effective in imparting flame retardant properties to the epoxy network. In all cases, the thermoset systems that contained the non-functional cyclophosphazenes were more flame retardant then thermoset systems that did not contain any flame retardant additive or contained a ring functional phosphazene of the type shown in Formulae 3 or 4 (Samples 19 and 25).

In this example, a modification of the burn test of Example 11 was utilized. The samples were exposed to a flame(procured as in Example 11) until sustained ignition was achieved. The average time to ignite was recorded as T.sub.ign. Once the sample started burning, the average time to self extinguish was recorded as T.sub.ext. Results for various samples previously described are listed in Table 11. The values reported represent the average of at least three samples.

From these data it can be concluded that the non-functional cyclophosphazenes (Samples 29, 30, 31, 33, 34, 35, 46 and 48) are effective flame retardants in epoxy/fiber composites and in cyanate ester resins. The samples containing the non-functional cyclophosphazenes had longer T.sub.ign times and shorter T.sub.ext times than the control resins. The data also indicate that the resin transfer molding (RTM) composite systems that contain the non-functional cyclophosphazenes (Samples 29-31) were more flame retardant than composites that did not contain any flame retardant (Sample 28).

Moisture uptake data were obtained by exposing neat resin or fiber composite samples to 100% relative humidity at 82.degree. C. until equilibrium moisture uptake was obtained. At least two replicate samples were measured, and average values are given in Table 12. After equilibrium moisture pickup was obtained, each sample was redried to a constant weight to determine if any of the phosphazene or other components had been extracted during the experiment. Typically, the redried weight was within .+-.0.05% of the initial weight.

From these data it can be concluded that the non-functional cyclophosphazenes (Samples 3, 5, 9, 18, 29, 31, 34, 35 and 40) do not have a negative impact on the moisture pick up of the neat resin or composite samples. In most cases, the non-functional cyclophosphazene flame retardants reduced the moisture pickup of the resin. Moreover, the data show that the ring functional amino cyclophosphazenes such as structures 3a, 3b, 4a and 4b are not beneficial in the resin network since these structures promote moisture pickup. The observation that sample 19 did not reach equilibrium after 30 days is indicative that chemical hydrolysis of the flame retardant may be occurring.

In order for a resin comprising flame retardant to be useful in processing techniques such as resin transfer molding (RTM), the viscosity of the resin must be essentially unaffected by the incorporation of the flame retardant. In addition, the flame retardant preferably has a minimal effect on the cure kinetics of the resin and does not cause the resin to cure prematurely. For example, it would be undesirable for the flame retardant to initiate the resin cure while it is being injected into the mold.

One means of assessing resin processibility is obtained from measurement of the change in the resin viscosity as a function of time at a fixed temperature. In this experiment, samples were ramped from 25.degree. C. to 104.degree. C. at 5.degree. C./min. The samples were then held isothermally at 104.degree. C. for 12 hours. The viscosity was measured throughout the experiment. The samples tested were:

(A) Sample 1, PR 500 unmodified, containing no flame retardant;

(B) Sample 2, PR 500 with 15% hexaphenoxy cyclotriphosphazene (Formulae 1, Structure 1a);

(C) Sample 10, PR 500 containing 15% of tris(phenoxy), tris(4-aminophenoxy)cyclotriphosphazene (Formaule 5, Structure 5a); and

(D) Sample 11, PR 500 containing 15% tris(phenoxy) tris(4-hydroxyphenoxy)cyclotriphosphazene (Formulae 5, Structure 5c).

Viscosity data were collected using a Rheometrics RDA-II.TM.. The neat, degassed uncured resin samples were cast between two 25 mm parallel plates. A strain amplitude of 2% and a frequency of 100 rad/sec were used. Viscosity data are provided in tabular form in Table 13 and plotted in FIG. 1.

In this experiment, the properties of melt processability, compatibility, moisture pickup (% water), glass transition temperature (T.sub.g) and thermal stability were determined for the following samples:

Sample 1: PR 500 with no flame retardant;

Sample 2: PR 500 modified with 15% hexaphenoxycyclotriphosphazene (Formula 1, Structure 1;a);

Sample 4: PR 500 modified with 15% hexa(2-chlorophenoxy)cyclotriphosphazene (Formula 1, Structure 1c);

Sample 13: PR 500 modified with 15% 1,1 bis(3-methoxyphenoxy)-3,5 bis(2,2'-oxybiphenyl) cyclotriphosphazene (Formula 6, Structure 6a, Sample 13)

Sample 14: PR 500 modified with 15% Hostaflam.TM. TPAP-750.TM., polyammonium phosphate available from Hoechst-Celanese (Sample 14)

Sample 15: PR 500 modified with 15% Kronitex.TM.-3600, alkyl-aryl phosphate ester, available from FMC Corp; and

Sample 16: Brominated epoxy sample.

Compatibility was determined by visually inspecting the cured samples for phase separation. Melt processability was determined by the ability of the flame retardant to dissolve into the resin prior to cure. T.sub.g was determined by differential scanning colorimetry (DSC) using a TA 912.TM. at a heating rate of 15.degree. C./min from 50.degree. C. to 250.degree. C. Thermal stability and moisture pickup were determined as described previously in Examples 10 and 13 respectively.

As can be seen from these comparative examples, the flame retardant in samples 14, 15, 16 had deleterious effects on several critical properties of the thermoset. For example, the particulate flame retardant (Sample 14) was not melt processable and was not compatible in the resin. Moreover, the moisture pick up of this resin was exceedingly high, not reaching equilibrium after 20 days at 82.degree. C. and 100% RH. Additionally, it was found that Sample 14 lost 7% of its initial weight after 46 days at 82.degree. C. and 100% R.H. This suggests that the ammonium phosphate undergoes hydrolysis and degrades the network.

The alkyl-aryl phosphate, Sample 15, reduced the T.sub.g of the resin to such an extent that such a network would not be useful at high temperatures. Additionally, Sample 15 exhibited poor thermal stability losing significant weight at 260.degree. and 300.degree. C.

The brominated epoxy, Sample 16, suffered catastrophic thermal degradation at 300.degree. C. losing 40% of its weight and exhibiting significant off-gassing. Moreover, this material contains bromine which is considered to be environmentally unfriendly.

Flame retardant analyses were conducted as described in Example 12, and results are reported in Table 15.

As can be seen from the Table 15, only the brominated flame retardant, Sample 16, was able to impart similar flame retardant performance as the non-functional cyclophosphazenes. The data in both Tables 14 and 15 show that the non-functional cyclophosphazenes imparts flame retardancy to thermoset networks while not detracting from other critical processing or performance properties.

The use of non-functional alkoxy cyclophosphazenes as flame retardant additives for thermoset systems is described in Japanese Kokai Patent Application No. SHO 61[1986]-120850. As will be demonstrated in this example, the alkoxy cyclophosphazenes, as compared to the aryloxy cyclophosphazenes, are not suitable flame retardants for thermosettable resins. In this example, the following samples were tested:

Sample 2: PR 500 modified with 15% of hexaphenoxycyclotriphosphazene (Formula 1, Structure 1a);

Sample 6: PR 500 modified with 15% hexa(ethoxy)cyclotriphosphazene (Formula 2, Structure 2a);

Sample 7: PR 500 modified with 15% hexa(n-propoxy)cyclotriphosphazene (Formula 2, Structure 2b); and

Sample 8: PR 500 modified with 15% hexa(isopropoxycyclotriphosphazene) (Formula 2, Structure 2c).

The resins were prepared as described in Example 2 and were cured at 177.degree. C. for 4 hours. After curing, it was noted that samples 6 and 8 were porous. Porosity in cured thermoset systems suggests a significant amount of volatile species were evolved during the curing of the resin. Such a resin is unsuitable for any thermoset application as the voids in the resin weaken the structure. The resins were analyzed for T.sub.g (using the procedure described in Example 15), Moisture pick up (using the method described in Example 13), and weight loss (using the method described in Example 10). Values are reported in Table 16.

From these data it can be concluded that the alkoxy phosphazenes (Samples 6, 7, and 8) do not have the thermal or moisture stability to be considered useful as flame retardants in thermosets. The moisture pickup studies showed that the alkoxy cyclotriphosphazenes (Samples 6, 7, and 8) did not reach equilibrium moisture pick up after 30 days at 82.degree. C. and 100% relative humidity. This suggests that the alkoxy cyclotriphosphazenes do not exhibit adequate hydrolytic stability to be useful in thermoset materials.

To further substantiate the thermal stability differences of the non-functional aryloxy cyclophosphazenes compared to non-functional alkoxy cyclophosphazenes, neat samples (approximately 1.5 g) of ([NP(OC.sub.6 H.sub.5).sub.2 ].sub.3 (Formula 1, Structure 1a), ([NP(OC.sub.2 H.sub.5).sub.2 ].sub.3, (Formula 2, Structure 2a) ([NP(O.sup.n C.sub.3 H.sub.7).sub.2 ].sub.3 (Formula 2, Structure 2b), ([NP(O.sup.i C.sub.3 H.sub.7).sub.2 ].sub.3, (Formula 2, Structure 2c) and N.sub.3 P.sub.3 (OC.sub.6 H.sub.5).sub.2 (OCH.sub.2 CF.sub.3).sub.4, (Formula 2, Structure 2e) were placed in heavy walled glass tubes and evacuated to 0.25 mm Hg for 30 minutes. The tubes were sealed under vacuum and place in an oven at 200.degree. C. (.+-.5.degree. C.) for 8 hours. After this time, the tubes were allowed to cool to room temperature and opened. The contents of each tube were evaluated by .sup.31 P and .sup.1 H NMR spectroscopy. By NMR analyses it was found that hexaphenoxycyclotriphosphazene ([NP(OC.sub.6 H.sub.5).sub.2 ].sub.3 (Formula. 1, Structure 1a), and N.sub.3 P.sub.3 (OC.sub.6 H.sub.5).sub.2 (OCH.sub.2 CF.sub.3).sub.4, (Formula 2, Structure 2c) did not undergo any chemical equilibration and were unchanged by this thermal process. Hexa(ethoxy) and Hexa(propoxy) cyclotriphosphazene, ([NP(OC.sub.2 H.sub.5).sub.2 ].sub.3, (Formula 2, Structure 2a) and ([NP(O.sup.n C.sub.3 H.sub.7).sub.2 ].sub.3, (Formula 2, Structure 2b), underwent significant chemical equilibration and decomposition to higher cyclic species or other rearranged products. Moreover, the tube containing hexa(isopropoxy)cyclotriphosphazene (See Formula 2, Structure 2c) detonated during this thermal treatment suggesting a rapid thermal decomposition to volatile species that exceeded the pressure limits of the sealed tube.

From these experiments, it was concluded that the non-functional alkoxy cyclophosphazenes do not have the thermal stability to be useful as flame retardants in these thermosets.

In this analysis, neat samples of the cyclophosphazenes 1a, 2a, 2b, 2c, and 2e were analyzed by thermogravimetric analysis (TGA). The samples were held for two hours at 200.degree. C., during which time the weight loss of each sample was measured at intervals of 10, 20, 30, 60 and 120 minutes. From these experiments it could be determined if volatility and/or decomposition of the cyclophosphazenes at a typical processing temperature of 200.degree. C. were likely to be problematic. The weight loss data are listed in below in Table 17.

From these data, it was concluded that the alkoxy cyclophosphazenes (Structures 2a, 2b, 2c and 2e) would not be beneficial in thermoset resins that are used or processed at elevated temperatures since the decomposition and/or volatility of the flame retardant at these temperatures would be problematic. These data also show that the alkoxy containing cyclophosphazenes possess inherent drawbacks that would limit their utility as flame retardant additives in thermosettable materials.

Relative permittivity (dielectric constant) and dissipation factor (loss tangent) were determined for thermoset resins using ASTM D150 Micrometer Electrodes at frequencies of 1 KHz and 1 MHz. Rolled-on foil was used for the measurements of dielectric constant, while no foil was used for measurements of loss tangent. Data are listed in Table 18.

From these data it is apparent that the non-functional cyclophosphazene flame retardant additives do not negatively impact the dielectric performance of the epoxy or cyanate ester thermoset resins.

The incorporation of the non-functional aryloxy cyclophosphazene flame retardant of the type shown in FIG. 1, Structure 1a, results in about a 20% decrease in the T.sub.g of the PR 500 resin. In most cases, this reduction in T.sub.g is still within acceptable limits. However, it was found that other cyclophosphazene structures would prevent this type of T.sub.g reduction. The T.sub.g of the PR 500 resin can be maintained by incorporation of either an incompatible aryloxy cyclophosphazene of the types shown in FIG. 1 as Structures 1b or 1e, or aryloxy bicyclophosphazenes of the types shown in Formula 6 as Structures 6a or 6b, or by blending aryloxy cyclophosphazenes of the types shown in FIG. 1 as Structures 1a and 1e. These affects on T.sub.g are summarized in Table 19. T.sub.g was measured as in Example 2. The term "compatibility" used in Table 19 refers to the visual detection of non-homogeneity after the sample was cured.

From these data it can be seen that the T.sub.g of the resin can be maintained by selecting a cyclophosphazene structure that can phase separate after the cure. Alternatively, a rigid non-functional cyclophosphazene such as bicyclophosphazenes of the type given in FIG. 6 may be used to preserve the T.sub.g of the resin.

A two-component epoxy structural adhesive was prepared with and without a non-functional cyclophosphazene flame retardant. Adhesives consisted of a base component and a curative component as shown in Table 20 in parts by weight.

The adhesive base portion of the composition was prepared by mixing the indicated amount of the crystalline solid cyclophosphazene with the liquid EPON.TM. 828 epoxy resin. The mixed materials were then heated to approximately 120.degree. C. in a glass jar until the cyclophosphazene was completely liquefied. The hot mixture was agitated by swirling the jar until the cyclophosphazene and epoxy resin mixture became homogeneous. Other components of the adhesive base, if any, were added to the cyclophosphazene-epoxy blend after it had cooled to room temperature. Similarly, the adhesive curative was made by melt-mixing the cyclophosphazene and mercaptan curative components listed in Table 20 at 120.degree. C., then allowing the mixture to cool to room temperature.

The base and curative compositions were mixed 1:1 by volume and dispensed simultaneously using a 3M Duo-Pak.TM. dual syringe and static mixer combination. Three rectangular plaques having dimensions 5.0"L.times.1.0"W.times.0.125" (12.5 cm.times.2.5 cm.times.0.31 cm) of each adhesive formulation (Samples 53 and 54) were obtained by dispensing the mixed adhesive into cavity molds milled from polytetrafluoroethylene (PTFE) and allowing a 7-day cure in ambient laboratory conditions before testing. Burn testing was done by suspending the plaques in a horizontal position within a steel frame on three evenly spaced 0.020" (0.508 mm) thick steel wires. Burn tests were conducted in a laboratory hood with the sash closed and fan off to minimize stray drafts. The burner flame height was set by adjusting the natural gas flame with burner air inlets closed to get a 1.0" (2.5 cm) height yellow flame, then opening the air inlets until a blue flame of the same height was obtained. The burner flame was applied to one end of the test specimen for 10 seconds, then withdrawn. The time of burning after removal of the flame, the qualitative intensity of the flame from the burning adhesive, and the time required for the flame front to travel 1.0" (2.5 cm) from the end of the plaque were recorded. Samples were allowed to burn for a maximum of 90 seconds, if they did not self-extinguish. Horizontal burn tests were run in triplicate for each sample, and the results are reported in Table 21.

The horizontal burn test results given in Table 21 illustrate that the addition of a non-functional cyclophosphazene imparts flame retardancy to a two-component epoxy structural adhesive.

The same formulations as given in Table 20, except for the replacement of 1 part of glycidoxypropyltrimethoxysilane with 1 part of EPON.TM. 828 epoxy resin were used for lap shear and peel testing. These formulations are identified in Table 22 as Samples 55 and 56. Lap shear specimens were prepared by spreading each mixed epoxy composition (prepared using the same procedure described above) onto an acetone-wiped, 0.60" (1.5 cm) thick, 1".times.4".times.(2.5.times.10.2 cm) aluminum alloy coupon (Alcoa 6111). Bond thickness was controlled by mixing approximately 1 wt % glass beads of nominal 0.010" (0.025 cm) diameter with the adhesive before application. After approximately 18 hours at room temperature, the bonds received a 30 minute post-cure at 250.degree. F. (120.degree. C.). After cooling to room temperature, the bonds were tested for Lap Shear Strength according to ASTM D1002-72 on an Instron tensile tester at a crosshead speed of 0.5" per minute (0.125 cm/min). Results are given in Table 22 as an average of the number of specimens indicated.

The lap shear results of Sample 55 (no flame retardant) and Sample 56 indicate that addition of the cyclophosphazene flame retardant has essentially no effect on the initial adhesive shear strength.

T-peel specimens were prepared using 8.0"L.times.1.0"W .times.0.042"T (20.3 cm.times.2.5 cm.times.0.11 cm) aluminum alloy (Alcoa 2008-T4) strips which were acetone-wiped before use. The adhesive was mixed and applied to approximately 6" (15.2 cm) of the coupon. Bond thickness was controlled by mixing approximately 1 wt % glass beads of nominal 0.010" (0.25 mm) diameter with the adhesive before application. The adhesive sides of two adhesive-bearing strips were then clamped together with binder clips for approximately 18 hours at room temperature. After this, the bonds received a 30 minute post-cure at 250.degree. F. (120.degree. C.) and were tested after cooling to room temperature. Bonds were tested according to ASTM D1876-72 on an Instron tensile tester at a crosshead speed of 10.0" per minute (25.4 cm/min). The recorded value in Table 23 is the average sustained peel load for the number of specimens indicated in the table.

Brittle failure means that the adhesive exhibited sudden catastrophic fractures within the resin during the peel test. Ductile failure means that the adhesive exhibited a continuous, yielding fracture of the resin during the peel test. In general, ductile failure of structural adhesives in the peel or cleavage mode is the preferred failure mode, and may be interpreted as a measure of the fracture toughness of the adhesive. As may be seen from the results in Table 23, the peel strength and failure mode of the adhesive are both improved by the addition of the cyclophosphazene flame retardant. Furthermore, there was no visible evidence on the specimens of migration of the non-functional flame retardant out of the epoxy matrix.

Heat-curable epoxy compositions 57, 58, 59, 60 and 61 were prepared to evaluate additive or synergistic effects of a non-functional cyclophosphazene flame retardant combined with other nonhalogenated flame retardants and halogen-free curative. Compositions were prepared with and without a cyclophosphazene-synergist combination and evaluated for flame retardant behavior.

Composition 57 comprised an epoxy resin (EPON.TM. 828) having cyclophosphazene flame retardant (Formula 1, structure 1a) with no other additional flame retardants. Composition 58 comprised the same components as Composition 57 with the further addition of Araldite.TM. PT 810 epoxy resin described previously in Example 7. Composition 59 comprised the same components as Composition 57 except for the addition of particulate melamine isocyanurate (available from Aldrich Chemical) flame retardant and the omission of fumed silica to counter the viscosity increase imparted by the particulate melamine isocyanurate. Compositions 60 and 61 comprised the same components as Composition 57 except for the further addition of AroCy.TM. L10 cyanate ester and the omission of fumed silica. The formulations employed are listed in Table 24 with quantities given as parts by weight.

Composition 57 was prepared by mixing the indicated amount of the crystalline solid cyclophosphazene (Formula 1, Structure 1a) with the liquid EPON.TM. 828 epoxy resin. The mixture was heated to 120.degree. C. and agitated until all the cyclophosphazene was completely liquified and the mixture became homogeneous. Particulate components [OTBAF (described in Example 7) and fumed silica] were then added to the hot resin-flame retardant mixture and dispersed with stirring.

Composition 58 was prepared by grinding the solid epoxy resin Araldite.TM. PT 810 into a fine powder and melt-mixing it with the EPON.TM. 828 resin at about 120.degree. C. Crystalline solid cyclophosphazene (Formula 1, Structure 1a) was then added to the melt. The mixture was held at 120.degree. C. with agitation until the cyclophosphazene was completely liquified and the mixture became homogeneous. Particulate components, OTBAF and fumed silica, were added to the hot resin-flame retardant mixture and dispersed with stirring.

Composition 59 was prepared using the same procedure described above for Composition 57 except that (1) finely ground melamine isocyanurate particulate was added to the hot resin-cyclophosphazene blend with the OTBAF and dispersed with stirring and (2) fumed silica was omitted from the formulation.

Composition 60 was prepared using the same procedure described above for Composition 57 except that liquid AroCy.TM. L10 cyanate ester was added to the EPON.TM. 828 epoxy resin at room temperature prior to the addition of the cyclophosphazene. Fumed silica was again omitted from this formulation.

Composition 61 was prepared according to the same procedure described above for Composition 57 except that cured cyanate ester particles were added to the hot resin-cyclophosphazene blend with the OTBAF and dispersed with stirring and (2) fumed silica was omitted from the formulation. The cyanate ester particles were prepared by pre-curing uncatalyzed AroCy.TM. L10 cyanate ester at 190.degree. C. for 72 hours. The solidified cyanate ester resin was then hammer-milled to an average particle size of 10 microns.

Rectangular plaques of dimensions 12.5 cm L.times.2.5cm W.times.0.31 cm T were obtained by curing each of the compositions in cavity molds milled from PTFE. The cure cycle consisted of a 30 minute hold at 70.degree. C. followed by ramp to 180.degree. C. at 1.degree. C. /minute, followed by a two hour hold at 180.degree. C. Burn testing was then done with the de-molded plaques clamped in a vertical position. Burn tests were conducted in a laboratory hood with sash closed and fan off to minimize stray drafts.

The burner flame height was set by adjusting the natural gas flame with the burner air inlets closed to get a 2.5 cm height yellow flame, then opening the air inlets until a blue flame of the same height was obtained. The burner flame was applied to one end of the test specimen for 10 seconds, then withdrawn. The time of burning after removal of the flame was noted. A second 10 second application was made after the sample self-extinguished, and burn time after removal of the flame was again noted. Results of the vertical burn tests of these formulations are summarized in Table 25; two tests were run per formulation.

The results in Table 25 show that the use of either Araldite.TM. PT810, melamine cyanurate or cyanate ester with cyclophosphazene flame retardant improves the flame retardancy of the epoxy composition. This is especially apparent in the data reported for the second flame application where markedly improved results are given for the formulations comprising both cyclophosphazene with either Araldite.TM. PT810, melamine cyanurate, or cyanate ester.

The results also show that cyanate ester may be added to the epoxy resin system as a cured resin or as an uncured resin, and in either case, functions as a synergist with the cyclophosphazene flame retardant in the cured thermoset system.