Dimethyl secondary amine chain extenders: a conceptual approach to in situ generation of advanced epoxy resins for rapid cure, low VOC coatings.
Keywords Poly(N-methylazetidine), Dimethyl-meta-xylylenediamine. Pseudo-first-order kinetic. Amine reactivity, Epoxy curing agent, Epoxy resin, Epoxy coating, Flexibility. Crosslink density, Miller-Macosko calculation. Chain extender, VOC, Low temperature cure. Environmental movement, Electrochemical impedance, Amine epoxy reaction mechanism, Mannich base
It is a great honor to have been awarded the Joseph Mattiello Lecture. Dr. Mattiello was a war hero--he gave one of his legs in service to his country in World War I. and served as a civilian chemical expert for the U.S. Army Quartermaster Corp in World War II--a distinguished contributor to the development of the science of paints and coatings, and rose to the level of Vice President and Technical Director of the Hilo Paint and Varnish Co. Over the years that I have been active in coatings research, it has been my pleasure to listen to many Mattiello lectures. Every one, without exception, was intellectually stimulating, and a great pleasure to hear.
While the conservation movement is rooted in events of the 19th century, the birth of the modern environmental movement is often traced to two events that both occurred in my formative years. In 1962, Rachel Carson courageously published Silent Spring, (1) in which she documented the decline in bird populations and other environmental damage associated with the widespread use of pesticides, most particularly DDT. Despite vigorous attacks on the work and threats of lawsuits, all anticipated by Carson and others involved in the publication of the work. Carson's conclusions were generally backed by the academic community, (2) and over time her arguments prevailed in the court of public opinion. The Environmental Defense Fund, established in 1967, initiated lawsuits that by 1972 secured a phase-out of the use of DDT in the United States. The establishment of the Environmental Protection Agency in 1970 may also be attributed in part to the criticism, voiced by Carson, of the inherent conflict of interest in the USDA, which was responsible for regulating pesticides, while also being responsible for promoting the concerns of the agricultural industry. (3) It is worth pointing out that despite protestations to the contrary, Carson never called for the outright ban on the use of pesticides, but rather argued for responsible and carefully managed use with an awareness of their effect on the entire ecosystem. (4)
In September 1969, Senator Gaylord Nelson of Wisconsin announced there would be a nationwide grassroots demonstration for the environment in the spring of 1970, and the first United States* Earth Day was held on April 22. The event has been held annually ever since. On that first Earth Day, Woodbury High School in New Jersey held an event in the school's auditorium, where I was one among several speakers. In my naivete, I warned my classmates of the environmental damage caused by the internal combustion engine, and urged them to abandon use of the automobile. To my surprise, the speech received less than universal approbation. To the best of my knowledge (and great relief) no text of this speech has survived. Like my classmates, in the following year I was unable to resist the allure of the freedom offered by the automobile, and I obtained my New Jersey driver's license at the earliest permissible date.
Upon receiving the Ph.D. degree from Yale University in 1982, I joined the Pioneering Research department at Rohm and Haas, and immediately began working on the development of nonaqueous dispersions targeted for use in the coatings market. Their purpose was to provide coatings with lower solvent content than could be afforded by traditional high molecular weight solvent-borne lacquers, while addressing some of the practical limitations associated with aqueous dispersions (emulsion polymers) in certain applications.
In the intervening years between the start of that first research project and the present, nearly every project that I have worked on--whether it be the general areas of high solids, waterborne, or UV cure technology--has included reducing the solvents emitted to the atmosphere as well as simultaneously meeting numerous other performance and cost criteria as a primary objective of the project. When working on such projects, progress usually seems slow. Technology advances by one careful, time-consuming experiment at a time: a small insight here, a tiny advance there.
While it has certainly not been easy, and has involved the efforts of thousands of dedicated scientists and engineers, over the span of decades our industry has made enormous progress. Table 1 shows the U.S. market estimates of the quantities of solvent sold into the paint and coatings market at various times. Also shown are the estimates for the volume of paint sold into the U.S. market. It is clear that our industry has made great strides in reducing the amount of solvent released to the environment in order to protect and beautify a given surface. All the while, this has been accomplished while generally maintaining--and in some cases markedly improving--the performance properties of our products. I take pride in having been a part of this industry, and having made contributions, however minor, to the progress that we as a whole have managed to accomplish.
Table 1: Solvent content of U.S. coatings since the advent of the environmental movement Year U.S. coatings volume Solvent volume in coatings Average pounds (MM gal) (a) (MM lb) (b) per gallon 1970 827 5414 6.55 1975 890 4424 4.97 1980 890 5636 6.33 1985 976 4054 4.15 1990 1102 4086 3.71 1995 1192 4408 3.70 2000 1286 3498 2.72 2002 1291 3698 2.86 (a) Source: E. Linak and A. Kishi, Chemical Economics Handbook, SRI Consulting, 2005 (b) Source: EPA's National Emissions Inventory (http://www.epa.gov/ttn/chief/net/index.html). This data is based on EPA's National Emissions Inventory. It does not include exempt solvent volumes.
Unfortunately, Joe Mattiello passed away in 1948, before the advent of the modern environmental movement. I believe Joe would likewise have taken great pride in our progress: pride based, in part, on the contributions he made to our science which laid the foundation for later advances, as well as his efforts toward the establishment and success of professional societies like the FSCT which through their mission of professional development and education have also contributed to our progress.
Issues in high solids coatings
The most straightforward approach to reducing a coating's solvent content is the high solids approach, which in its most basic manifestation simply involves removing some of the solvent in a formulation. Naturally, this creates a significant and obvious problem. Paints are always formulated to a relatively narrow window of viscosity that results in successful application using the application method suitable for the job at hand. While some benefit can be gained by manipulating solvent blends or application parameters (such as heating the liquid coating or increasing pressure capabilities of spray equipment), in reality the benefits to be gained in this way are quite limited, and result in only modest reductions in VOC by themselves.
The result of the need to address this viscosity problem has been that nearly all high solids paints are based on lower molecular weight resins and curatives than the conventional solids paints based on the same basic polymer chemistry. This has resulted in particularly difficult problems in the area of solvent-borne lacquer technology, where there is only limited ability to reduce molecular weight before the performance properties of the resulting film are so reduced as to be beyond usable limits. Thus, nearly all modern high solids technology is based on reactive chemistries where the low molecular weight resin components react to form crosslinked networks after application. The majority of these chemistries are two-component (2 K) technology where separate reactive components are mixed before application, with the exception of some high solids oxidative cure technology (which has its own distinct set of technological issues).
In general, it is found with 2 K technology that reducing the molecular weight of reactive components has three very negative consequences: the "pot life" or the amount of time that the liquid coating remains below the maximum viscosity for successful application is decreased, the amount of time it takes for the coating to reach an acceptable degree of cure (dry-to-handle lime, dry-to-stack time, walk-on time, etc.) is increased, and the flexibility of the coating is decreased.
Chemistry of traditional and high solids solvent-borne epoxy coatings
Before the advent of VOC regulations, typical conventional solids solvent-borne epoxy paint utilized a solid epoxy resin with an epoxy equivalent weight (EEW) of about 500 or so. For corrosion-resistant coating applications, it was most commonly formulated with a polyamide curing agent with a viscosity in the range of 400,000 mPa s at room temperature. To reduce viscosity and increase solids, high solids formulations are usually based on "liquid" epoxy resin (LER) with an EEW of about 190 and a polyamide curing agent with a much lower viscosity, typically in the range of 15,000-50,000 mPa s.
Both standard liquid and solid epoxy resin are oligomers derived -from bisphenol-A, with chemical structure 1. They differ in the average value of n. For LER [M.sub.n] is about 380, the viscosity is about 12,000 mPa s at room temperature, and n [congruent to] 0.15. There are several grades of solid epoxy resin, but the one most commonly used in coatings has an [M.sub.n]of about 1050, with n [congruent to] 2.5. Because of its physical state, the viscosity of solid epoxy resin cannot be measured directly at room temperature, but if viscosity vs % solids is plotted, the extrapolated value of this resin's viscosity is in the millions of mPa s.
Using epoxy resin chemistry as a model, we quantitatively elucidated the source of the problems of short pot life and slow cure speed, (5) and due to the relevance to the results described in this article, the information is reviewed here. As shown in Table 2, a conventional epoxy resin formulation based on a 525-epoxy equivalent weight (EEW) formulated at a VOC of 600 g/L has an epoxide group concentration of 0.54 eq/ L. An EEW of 525 is the equivalent weight of the least viscous, lowest molecular weight conventional solid epoxy resin. Simply eliminating solvent to get to 420 g/L increases the epoxide concentration by 72% to 0.93 eq/L. This is probably about the practical limit that can be achieved in practice formulating solid epoxy resin. At this point, replacing solid epoxy resin with LER (EEW = 190) at the 420 g/L VOC level results in an even larger (115%) increase in concentration to 2.01 eq/L. It is important to note that this is the effect on the concentration of reactive end groups caused by decreasing resin molecular weight, not by a change in solvent content. Eventually, totally eliminating solvent increases the epoxide concentration to 3.97 eq/L.
Table 2: Epoxide concentration vs VOC for model epoxy coating formulation VOC g/L (lb/gal) Resin EEW %Solids [Epoxide] (eq/L) 600 (5.0) 525 36.0 0.541 420 (3.5) 525 58.1 0.931 420 (3.5) 190 58.1 2.01 335 (2.8) 190 67.4 2.40 180 (1.5) 190 83.4 3.13 0 (0) 190 100 3.97 Calculated %solids and [epoxide] for clearcoats of varying VOC content, with epoxide equivalent weights of typical solid and LERs. Calculations assume that resin and curing agent both have a density of 1.15 g/mL and are present at 1:1 stoichiometry. The curing agent has an equivalent weight of 100 g/N-H, and the solvent density is 0.85 g/mL
The kinetics of the reaction of epoxy resins with an amine-curing agent are somewhat complex (as will be reinforced by results presented later in this article), because there are mixtures of 1[degrees] and 2[degrees] amines present, and because the reaction is strongly catalyzed by hydroxyl groups that are formed during the reaction. (6) However, in the presence of an alcohol solvent, the concentration of OH groups is more or less constant, and the kinetics of the reaction might then be expected to be at least approximately of second order.* In other words, the rate of consumption of epoxide is nearly proportional to the concentration of amine times the concentration of epoxide, with a proportionality constant equal to some rate constant, k:
[[-d[E]]/[dt]] = k [E] [A]]
where [E] is the concentration of epoxide groups and [A] the concentration of amines.
By inserting the epoxide concentration into the integrated form of a second-order rate equation, (7) it is possible to calculate the time to reach any degree of conversion. Data for 50% (half-life) and 25% (quarter-life) conversions are shown in Fig. 1. A rate constant of 0.0001 L [mol.sup.-1] [s.sup.-1] was assumed, which is within the range of reported second-order rate constants for amines with aromatic glycidyl ethers at room temperature. (8) At what degree of conversion pot life will end depends on the particular system studied, and the absolute values obtained are not as important as the trends. The results are shown in Fig. 1.
[FIGURE 1 OMITTED]
With the highest VOC formulation of solid epoxy resin, the half-life is calculated to be 9.5 h. By simply reducing VOC to 3.5 lb/gal, the half-life decreases to 3.21 h, whereas decreasing equivalent weight from 525 to 190 (switching from solid to liquid resin) further reduces the half-life to 0.69 h. At 100% solids, the half-life is only 11 minutes! These trends are in good qualitative agreement with the enormous decreases in pot life that are experienced in actual practice as VOC has been continuously reduced. Formulators, who used to be able to deliver products with pot lives that would last an entire shift, may struggle to achieve a half-hour pot life in a 100% solids formulation. Thus, the decrease in pot life in high solids formulations is largely explained by the increased concentration of reactive groups, and straightforward chemical kinetics.
A useful concept for understanding dry speed is that a coating will appear to be dry when the film does not appreciably flow under the force to which it is subjected in the test. For example, a coating is dry-to-touch when it does not flow under light finger pressure. Since viscosity is by definition resistance to flow, it follows then that the state of dryness will also be related to viscosity. It has been reported. (9) for example, that a coating with a viscosity greater than 1,000,000 mPa s will pass "dry-to-touch." A much higher viscosity is required to pass a blocking test.
Conventional epoxy coatings generally pass dry-to-touch and dust-free tests in a matter of minutes. This is probably too short a time for chemical reactions to play a significant role in the drying process. Since they are based on a solid epoxy resin (with a viscosity in the millions of mPa s) and a polyamide resin with a viscosity of about 400,000 mPa s. it is quite reasonable that once most of the solvent has evaporated and perhaps a small amount of chemical reaction has occurred, the film will become dry-to-touch. Drying that is solely dependent on solvent evaporation is referred to as "lacquer dry."
To understand the differences in the drying behavior of high solids formulations, computer modeling of epoxy cure was conducted using Miller-Macosko calculations. For curing agents, a polyamide resin was modeled with idealized triethylenetetramine (TETA)--dimer acid oligomers (2) chosen with values of n = 2 for a high solids, and n = 8 for a low solids system.
The data required for the calculations are shown in Table 3, along with the calculated gel point. The weight average molecular weight as a function of conversion is given in Fig. 2, and the crosslink density, calculated as the concentration of effective strands, is given in Fig. 3.
[FIGURE 2 OMITTED]
[FIGURE 3 OMITTED]
Table 3: Data used for Miller-Macosko model calculations Property High solids Low solids Epoxide resin mol. wt. 380 1050 Epoxide resin functionality 2 2 Curing agent mol. wt. 1490 5522 Curing agent functionality 10 22 Stoichiometry 1:1 1:1 Calculated result Gel point 33.3% 21.8% Crosslink density at 100% conversion (mol/g) 0.00118 0.000586
The viscosity of concentrated polymer solutions is proportional to [M.sub.w] to the first power below the critical entanglement chain length. (10) Since the low solids gives lacquer dry, we can assume that the viscosity of the blend in the absence of solvent is probably in excess of 1,000.000 mPa s. The high solids composition, on the other hand, must achieve roughly 20% conversion to reach a comparable [M.sub.w] and viscosity, depending on the accuracy of this modeling. The low solids system also gels at much lower conversion. However, at about 52% conversion, the crosslink density of the high solids composition begins to exceed that of the low solids composition, and the final crosslink density of the high solids system is about twice that of the low solids system.
As crosslink density increases, modulus increases for otherwise identical materials. Although there is some change in the ratio of epoxy resin to hardener in going from low to high solids, it is reasonable to assume that this increase in crosslink density would lead to higher modulus in this case. This would then be expected to yield higher hardness, as is normally observed when these changes are made. Unfortunately, the decrease in distance between crosslinks (increased concentration of effective strands) would be expected to lead to increased shrinkage, reduced flexibility, and increased internal stress. This is indeed also usually observed when these formulation changes are made, since high solids epoxy coatings often perform poorly in impact, mandrel bend, and hygrothermal stress tests. Finally, solvent resistance is usually a function of crosslink density within families of similar materials. These calculations would suggest that solvent resistance at high degrees of conversion should be superior in a high solids formulation, which is also commonly observed.
In situ chain extension concept
Solid epoxy resin (1, n ~ 2.5) can be formulated to yield long pot life, fast dry speed and adequate flexibility, but cannot be formulated to meet aggressive VOC targets. LER can be used to meet VOC requirements, but often yields short pot life, slow cure, and brittle films. How can this dilemma be addressed?
We developed the following conceptual framework for investigation. Could a difunctional diamine 3 be developed:
such that when combined with other multifunctional amines necessary for crosslinking it would react with epoxy resin so that in situ, oligomer 4 (or remnants thereof in the crosslinked network) would result? Would the properties of 4 be similar enough to solid epoxy resin that upon cure, films with solid epoxy resin cure speed and flexibility properties would result?
To meet the multitude of requirements of modern high solids epoxy coatings, what characteristics would difunctional diamine 3 need to possess? First and foremost, it would have to be based on readily available and inexpensive raw materials. The process for preparing 3 should preferably be amenable to existing capital equipment. Optionally, the costs should be so low as to make investment in such equipment attractive, though in my experience business decision makers are often unwilling (for good reason) to undertake the risks inherent in such investments for technology unproven by the marketplace. The diamine would need to be of reasonably low molecular weight to meet viscosity requirements. We hypothesized that it would also preferably be an amine with a relatively fast kinetic rate constant for reaction with epoxy resins. A fast rate of reaction should be beneficial for cure speed. Fast reaction might also reduce phase separation of curing agents and epoxy resins during cure. This leads to numerous problems, referred to in the industry as blush and exudate. These defects are caused by the development of a layer of less than fully cured amines on the surface, which may be in the form of semi-crystalline carbonates formed by reaction of primary amines with carbon dioxide (blush), or simply a greasy layer on the surface (exudate). In either event, surface appearance is negatively impacted, and intercoat adhesion of a topcoat is often compromised.
Naturally, fast reaction rate would be expected to negatively impact pot life. However, there is reason to believe that a difunctional chain extender reacting with a difunctional epoxy might be less problematic in this regard than would a multifunctional amine with comparable reactivity. Reaction of 3 with epoxy resin results in a linear polymerization, whereas a multifunctional amine results in a branched structure, and eventually gelation and crosslinking. Viscosity is related to the weight average molecular weight of a polymer, (11) and weight average molecular weight increases more rapidly with branching than it does with linear polymerization as a function of conversion. (12)
All solvents, diglyme internal standard, phenyl glycidyl ether (PGE), N,N'-dimethyl-l,3-propane diamine (DMPDA). N,N'-dimethyl-l,6-hexamethylene diamine (DMHMDA). and salicylic acid were obtained from Aldrich Chemical Co. and were used without purification. Meta-xy1ylenediamine(MXDA) was obtained from Mitsubishi Gas and Chemical Co. Mixed polycycloaliphatic amines (MPCA) were obtained from Air Products and Chemicals. Inc., as Ancamine[R] 2168, and had an amine hydrogen equivalent weight (AHEW) of 57.
Poly(N-methylazetidine) (p-NMAz) and N,N'-di-methyl-meta-xylylene diamine (DMMXDA) were prepared by processes discussed in general terms later in the text. Process details are proprietary.
Competitive pseudo-first-order rate constant determinations
In this example, the competitive pseudo-first-order (PFO) rate constants for the reaction of DMMXDA and MXDA with PGE were determined. Other rate constants were measured by an analogous procedure. Into a disposable tri-pour beaker was placed 22.5460 g (3.0025 M) of PGE, along with a small amount of isopropyl alcohol (IPA). Into a dry 50-mL volumetric flask were placed 0.6709 g of diglyme (internal standard), 1.3673 g of MXDA (0.2009 M), and 1.9135 g of a sample that was analyzed by GC to contain 86.3% DMMXDA and 13.7% N-methyl-meta-xylylenedi-amine (MMMXDA), and a small amount of IPA. Thus, the amount of DMMXDA in the mixture was 1.65 g (0.201 M). Working rapidly, the contents of the tri-pour were poured into the volumetric flask, and washed carefully several times with small amounts of IPA to ensure complete transfer. The volumetric flask was then filled to make up 50 mL with IPA. and timing started. After mixing the contents, a 1-g sample was immediately removed and diluted 10:1 with IPA. The volumetric sample was placed in a water bath held at 25[degrees]C and continuously swirled for 2 min to rapidly achieve thermal equilibrium. The diluted sample was then immediately analyzed by GC. Additional samples were withdrawn every 15 min for GC analysis and likewise diluted with IPA. The GC analysis was conducted on an HP 5973 GC with an auto-injection system, using a 30 m x 0.32 mm x 0.25 [micro]m film thickness HP-5 column operated at an initial temperature of 80[degrees]C ramped to 280[degrees]C at a rate of 20[degrees]C/min, with a 2 min hold at the final temperature. The data were analyzed by the internal standard method.
Miller-Macosko calculations (13-15) were performed on an Excel[R] spreadsheet program developed by the author, which was adapted from a program published by Bauer. (16) The output of the program used in this work is the weighted concentration of effective strands. (17) The required inputs for the program are the stoichiometric imbalance r of amine-hydrogen to epoxy groups (set at 1.00 for these calculations), the extent of reaction [alpha](set at 1.00 for these calculations), and the mol%, molecular weight, and functionality of every component for both the epoxy and amine sides of the formulation. For the epoxy resin, a composition of 81 mol% diglycidyl ether of bisphenol-A. and 19 mol% DGEBA n = 1 oligomer was the input, corresponding approximately with the composition of Epon[R] 828 LER (Hexion Specialty Chemicals, Inc.). Crosslink densities were calculated for curatives that were part of a clearcoat mixture experimental design for which other performance results are reported in this study. The curatives used in the design were three-component mixtures of DMMXDA, MXDA. and MPCA. The composition of MPCA utilized was based on the averages from multiple commercial batches as analyzed by GC.
Dynamic mechanical analysis
The above calculated results were compared to glass transition temperatures and crosslink densities obtained by dynamic mechanical analysis (DMA) run on free films of these same compositions. The DMA's were run at isothermal steps spaced 6[degrees]C apart at temperatures between -97 and 200[degrees]C at an applied oscillation frequency of 1 Hz in tensile mode on an RSA II-controlled strain rheometer manufactured by Rheometric Scientific (now TA Instruments, New Castle, DE). To ensure maximum reaction of amine and epoxide, the sample was cooled and the temperature ramp repeated. The rescan values were used for comparison with the calculations. The [T.sub.g] was taken as the inflection point of the rapid drop in modulus between the glassy state and the rubbery plateau. The molecular weight between crosslinks ([M.sub.c]) is determined from the expression [M.sub.c] = [[3[rho]RT]/E'] using the temperature where the modulus value was at its minimum in the rubbery plateau. (18)
Clear coatings for Electrochemical Impedance Spectroscopy (EIS) measurements were applied to cold-rolled steel test panels, ground one side (size 0.8 x 76 x 152 mm, Q Panel Lab Products). Coatings were applied using a 75-[micro]m wire bar to result film thicknesses between 60 and 70 [micro]m following 7-day cure at 23[degrees]C and 60% RH. EIS was conducted using an FAS1 potentiostat with CMS 100 Electrochemical Measurement System (Gamry Instruments, Inc.). The coating surface was exposed to an aqueous solution of NaCl (3 wt%) which served as the electrolyte. Using a graphite rod electrode, resistance and capacitance data of the coatings were collected against a saturated calomel electrode (SCE) during 100 s by applying a 10-mV AC excitation voltage.
Resistance was plotted as a function of frequency and impedance data was analyzed and fit with the system's software to obtain the pore resistance. Water uptake (V in wt%) was calculated from the change in capacitance between 1- and 24-h exposure, using the following equation:
V = 100 * ([[log[R.sub.C,1h]/[R.sub.C,24h]/[log 80]])
Preparation and testing of coatings and mixture experimental designs
Coating properties were measured using the test methods described in Table 4.
Table 4: Coating evaluation test methods Property Response Test method Drying time: BK Thin film set times, ASTM D5895 recorder phases 2 & 3 (h) Drying time: Set-to-touch and ASTM D1640 thumb twist method thumb-twist time (h) Specular gloss Gloss at 20[degrees] and ISO 2813, ASTM D523 60[degrees] Persoz pendulum Persoz hardness (s) ISO 1522, ASTM D4366 hardness Impact resistance-- Direct and reverse impact ISO 6272, ASTM D2794 tubular impact tester (kg cm) Mandrel bend test: Elongation (%) ISO 1519, ASTM D1737 cylindrical bend Mandrel bend test: Elongation (%) ISO 6860, ASTM D522 conical bend
For the mixture design experiments, hardener mixtures were prepared by combining and mixing the components given in Tables 7 and 8. They were then thoroughly mixed with DGEBA at the use level (parts per hundred resin. PHR) indicated. Clear coatings for drying time by BK recorder and hardness development by Persoz pendulum hardness were applied to standard glass panels. Clear coatings for measuring drying time by thumb twist method, specular gloss, and resistance to carbamate formation were applied to uncoated. matte paper charts (AG5350, Byk). Coatings were applied at 75-[micro]m WFT (wet-film thickness) using a Bird applicator resulting in dry-film thicknesses from 60 to 70 [micro]m. Films were cured at 5[degrees]C and 60% RH (relative humidity) or 25[degrees]C and 60% RH in a Weiss climate chamber (type WEKK0057), and Persoz Hardness was measured at the times indicated.
Table 7: Coating performance data for DMMXDA/MXDA/MPCA design space X1 X2 X3 Y1 Y3 MXDA DMMXDA MPCA BZA AHEW MIX VIS TFST25 0.300 0.700 0.000 0.428 80 641 4.2 0.000 0.570 0.430 0.428 100 1500 3.3 0.500 0.000 0.500 0.428 63 3200 4.1 0.325 0.350 0.325 0.428 74 1620 3.8 0.325 0.350 0.325 0.428 74 1640 3.8 0.325 0.350 0.325 0.428 74 1700 3.8 0.000 0.500 0.500 0.428 99 1870 4.3 0.000 0.700 0.300 0.428 103 980 3.5 0.220 0.280 0.500 0.428 78 2240 3.8 0.500 0.480 0.020 0.428 67 750 3.5 [R.sup.2] 0.996 0.45 LoF RMSE 73 0.3 Y6 Y4 Y2 Y7 GLSS 25[degrees] PEND 25[degrees] TFST 5[degrees] GLSS 5[degrees] 59 339 10.6 17 100 350 6.9 8 95 352 14.6 15 50 341 12.3 17 40 319 12.6 10 38 334 13.9 24 99 337 8.8 91 105 327 7.4 97 94 338 9.9 96 4 323 9.3 3 [R.sup.2] 0.980 0.450 0.850 0.43 LoF RMSE 6 10 1.4 34 Y6 Y5 Y'1 Y'2 GLSS 25[degrees] PEND 5[degrees] [T.sub.g]_FINAL [M.sub.c]_FINAL 59 188 85 1247 100 138 92 1423 95 163 130 641 50 141 112 750 40 170 112 750 38 155 111 700 99 101 102 1030 105 85 82 1577 94 109 122 694 4 164 101 774 [R.sup.2] 0.980 0.620 0.974 0.970 RMSE 6 23 3 191 Table 8: Coating performance data for p-NMAz[1:1]/MXDA/MPCA design space X1 X2 X3 Y2 Y3 MXDA p-NMAz MPCA BZA AHEW MIX VIS TFST25 0.000 0.620 0.380 0.428 116 650 3.1 0.000 0.740 0.260 0.428 123 420 3.4 0.140 0.420 0.440 0.428 93 990 3.5 0.170 0.510 0.320 0.428 94 710 3.4 0.250 0.750 0.000 0.428 96 340 3.1 0.300 0.390 0.310 0.428 81 830 3.5 0.300 0.390 0.310 0.428 81 830 3.5 0.300 0.390 0.310 0.428 81 830 3.5 0.500 0.000 0.500 0.428 63 1620 3.8 0.430 0.550 0.020 0.428 77 440 3.3 1.000 0.000 0.000 0.428 49 750 3.6 [R.sup.2] 0.999 0.868 RMSE 10 0.1 Y11 Y7 Y5 Y13 GLSS 25[degrees] PEND 25[degrees] TFST 5[degrees] GLSS 5[degrees] 103 303 8.4 78 100 265 8.0 50 84 294 8.9 64 40 318 9.2 55 2 298 9.3 30 64 341 9.1 43 65 331 9.0 42 64 334 9.4 40 2 331 11.2 16 1 326 8.5 1 27 348 10.1 1 [R.sup.2] 0.866 LoF 0.814 0.940 0.950 LoF 16 RMSE 12 0.3 6 Y11 Y9 Y'2 Y'3 GLSS 25[degrees] PEND 5[degrees] [T.sub.g]_FINAL [M.sub.c]_FINAL 103 86 60 4557 100 54 50 7405 84 102 82 1577 40 108 74 1821 2 89 57 2668 64 108 89 1157 65 110 89 1160 64 105 89 1200 2 81 135 582 1 139 77 916 27 120 127 476 [R.sup.2] 0.866 LoF 0.959 0.999 0.977 LoF 16 RMSE 6 0.3 437
Mix viscosity was determined using a Rheolab MC20 apparatus (Physica) equipped with a Visco-therm VT10 water bath and MC20 temperature control unit. The equipment is set up with the TEK 150 cone-plate and connected to a desk-top computer. After the apparatus has equilibrated at 25.0[degrees]C, the gap between the cone (MK22) and plate is set to 50 urn. Samples were equilibrated at 25[degrees]C 24 h before use. After mixing as indicated, excess product running out of the gap was removed and rotational viscosity of the mixed product was recorded at 200 [s.sup.-1] shear rate after 30 s.
Clear coatings for impact resistance and mandrel bend testing were applied to cold-rolled steel test panels, ground one side (size 0.8 x 76 x 152 mm) and cold-rolled steel, smooth finish (0.5 x 76 x 152 mm), using a 75-[micro]m wet-film thickness (WFT) wire bar. Metal test panels were obtained from Q Panel Lab Products.
Results and discussion
Identification of target chain extenders
Having identified general characteristics of desired chain extenders, the next task was to pick target molecules that might meet our needs. Using pseudo-first-order competitive kinetic analysis. Marsella and Starner (19) showed that the rate of reaction of a methyl-substituted 2[degrees] amine with phenyl glycidyl ether was about 1.2-2.1 times faster that than the rate of the corresponding primary amine. Select data from their study is presented in Table 5. This was remarkable, in that alkyl substitutions other than methyl always lead to lower rate of reaction, and the effect is attributed to a fortuitous combination of the electron-donating effect of the methyl substitution, and a methyl group's small steric effect.
Table 5: Pseudo-first-order competitive rate data for primary and corresponding secondary amines (19) Primary amine [k.sub.2[degrees]Me/1[degrees]] Solvent n-Butylamine 2.1 Ethanol t-Butylamine 1.2 Ethanol Cyclododecylamine 1.2 Ethanol Cyclohexylamine 1.4 t-Amyl alcohol Ethylene diamine 1.6 t-Amyl alcohol
Based on this insight, we decided to target two types of dimethyl substituted diamines: one with a linear chain separating the methylamine groups, and one with a xylylene moiety separating them.
Commercially, MXDA is prepared by reductive hydrogenation of isophthalonitrile. The methyl substituted analog of MXDA. DMMXDA, was prepared by reduction of isophthalonitrile with hydrogen in the presence of excess methylamine over a noble metal catalyst (Scheme 1). After distillation DMMXDA was obtained as 88% of a mixture with 12% of the monomethyl substituted product as shown in Scheme 1.
For use in model studies, it was planned to prepare N,N'-dimethyl-1,3-propanediamine (DMPDA) by the Michael addition reaction of excess methylamine to acrylonitrile, to yield 3-(N-methylamino)propionitrile, followed by hydrogenation of the resulting product in the presence of methylamine. (Due to its high volatility, 1,3-PDA was not expected to be of practical value as a curing agent.) Surprisingly, it was found that the yield of 1,3-PDA was very low, ranging from a few percent to about 13%, depending on conditions.
Further investigation of the reaction products indicated that most of the product was an oligomer stream based on the repeating unit -[CH.sub.2][CH.sub.2][CH.sub.2]N([CH.sub.3])-as shown in Scheme 2. Poly(N-methylazetidine) (p-NMAz) is the common name that was assigned to this oligomer stream (the nomenclature is explained below). After process optimization, the following general procedure was developed. First, acrylonitrile is reacted with methylamine to yield a mixture of the 1:1 and 2:1 adducts. The resulting mixture is hydrogenated over a noble metal catalyst in a methylamine atmosphere. By changing the ratio of acrylonitrile to methylamine in the first step, and by variation of the methylamine concentration in the second step (as well as varying the temperature, hydrogen pressure, etc.) the distribution of products in the oligomer stream can be varied to some degree. Note that GC may underestimate the concentration of higher molecular weight species in the distribution.
The name poly(N-methylazetidine) refers to the macromolecule that in principle could be prepared by polymerization of N-methylazetidine with methylamine (or methylamide anion under strongly basic conditions) as the initiator (Scheme 3). This is analogous to the name poly(ethylene oxide) for the polymer prepared by hydroxide initiated polymerization of ethylene oxide. In practice, it might prove difficult if not impossible to effect such a polymerization, as azetidines (4-member ring compounds) are much less prone to polymerization than are aziridines (3-member rings). Furthermore, azetidines tend to be more difficult to prepare than aziridines, and the author is unaware of any economical route for the preparation of this monomer.
A proposed mechanism for the polymerization is shown in Scheme 4. Reduction of the nitrile proceeds step-wise to yield the unsubstituted imine. which is in rapid equilibrium with the methyl-substituted imine through an aminal intermediate. Hydrogenation of the methyl-substituted imine yields 1,3-PDA, which can likewise participate in a further addition to another imine intermediate to yield the oligomerized aminal. Reductive hydrogenolysis yields p-NMAz with n = 1. Of course, it is also possible that 1,3-PDA is likewise derived directly via reductive hydrogenolysis of the aminal intermediate, rather than through the methyl-substituted imine as shown.
In solution, the reaction of amines with epoxides probably occurs by both catalyzed and non-catalyzed pathways, though the catalyzed pathway is significantly lower in energy and therefore dominant when hydrogen-bonding species are present. A proposed mechanism for the catalyzed reaction is shown in Scheme 5. Though other published mechanisms (20) have depicted hydrogen-bonded species, the six-member ring transition state shown here appears not to have been published, though the author believes it is consistent with the known facts concerning this reaction. The process begins with activation of the epoxide ring by hydrogen-bonding with a hydrogen bond donor present in the system, most commonly a hydroxyl group, or some other H-bond donor added as a catalyst. Attack of a primary amine then affords ring opening of the epoxide through a six-member ring transition state whereby protons are transferred among heteroatoms. The resulting 2[degrees] amine reacts with a second-activated epoxide to yield a tertiary amine. It is generally found that the rate constant for reaction of primary amines with aromatic glycidyl ether-type epoxides is approximately twice that of the secondary amine adduct with a second epoxide.
Since hydroxyl groups are catalysts for the reaction, their formation during the reaction accounts for the auto-accelerating kinetics observed when epoxy-amine reactions are conducted in the absence of hydrogen-bonding solvents.
Horie has proposed the following equation to describe the complexities of the kinetics of amine-epoxy chemistry (21):
[[dx]/[dt]] = [k.sub.1] [a.sub.1]ex + [k'.sub.1][a.sub.1] [ec.sub.0] + [k.sub.2] [a.sub.2]ex + [k'.sub.2][a.sub.2][ec.sub.0]
which includes terms for both 1[degrees] and 2[degrees] amine reactions by both catalyzed and uncatalyzed processes. Unfortunately, determining kinetic rate constants through the use of such a complex equation is exceedingly complex. (22)
A more practical approach, and one much more suited to answering the question "what are the relative reactivities of this series of amines" was developed by Marsella and Starrier. (19) Relative rate constants in this approach are estimated using pseudo-first-order (PFO) conditions, and the compounds are all measured in competition experiments vs a reference amine. In pseudo-first-order experiments one of the reagents, in this case PGE, is employed in large (greater than 10-fold) excess. Thus, for practical purposes, the concentration of PGE can be treated as constant, reducing the mathematical analysis from a second-order problem to a much simpler first-order problem. The rate constant is the slope of the line obtained from a plot of In[conc.] vs time as long as good linear behavior has been obtained. Since the rate of first-order reactions is independent of concentration, (23) small errors in purity or weighing have little or no effect on the final results. By performing competition experiments pegged to a common reference amine, small changes in temperature, which can have large effects on absolute rate constants, generally have little effect on the relative rates [[[K.sub.target]]/[[k.sub.ref.]]], since one is only comparing the slope of the first-order rates of the two reactions, which are both occurring at the same temperature. (These errors are inherent, however, in the calculated PFO rate constants also given in this study.) Finally, by conducting the reactions in an alcohol solvent such as IPA the auto-accelerating effect is eliminated, since for practical purposes the concentration of catalytic OH can be treated as a constant.
Figure 4 shows the results of a PFO competition experiment for DMMXDA vs MXDA as reference amine conducted at 25[degrees]C in isopropyl alcohol (IPA). Excellent straight line behavior was observed, and [[[K.sub.DMMXDA]]/[[k.sub.MXDA]]] is 2.9. as shown in Table 6.
[FIGURE 4 OMITTED]
Table 6: Chain extender competitive pseudo-first-order rate data Target amine Ref. Temp. Solvent PFO rate amine ([degrees]C) constant ([min.sup.-1]) DMMXDA MXDA 25 IPA 0.0587 DMMXDA MXDA 0 IPA 0.0071 DMMXDA MXDA 25 Xyl na DMMXDA MXDA 25 Xyl/0.5 M IPA 0.0063 DMMXDA MXDA 25 Xyl/1.0 M IPA 0.0101 DMMXDA MXDA 25 Xyl/2.0 M IPA 0.0187 DMMXDA MXDA 25 Xyl/4.0 M IPA 0.0332 DMMXDA MXDA 25 Xyl/1.0 M IPA/0.1 na M SA DMMXDA MXDA 25 Xyl/2.0 M IPA/0.1 0.0239 M SA DMMXDA MXDA 25 Xyl/4.0 M IPA/0.1 0.0331 M SA DMMXDA MXDA 25 Xyl/1.0 M BzA 0.0158 DMMXDA MXDA 25 Xyl/2.0 M BzA 0.036 DMMXDA MXDA 25 Xyl/4.0 M BzA 0.0785 DMPDA MXDA 25 IPA 0.0546 DMPDA DMMXDA 25 IPA 0.0543 p-NMAZ(n = 1) MXDA 25 IPA 0.0558 p-NMAZ (n = 2) MXDA 25 IPA 0.0534 Target amine [k.sub.target]/ Data fit [k.sub.ref]. DMMXDA 2.9 Good DMMXDA 3.2 Good DMMXDA na V Poor, auto-acceleration DMMXDA 3.7 Fair, some noise DMMXDA 3.1 Fair, some noise DMMXDA 3.2 Slight curvature possible DMMXDA 3.2 Slight curvature possible DMMXDA na Noisy; degree of curvature hard to judge DMMXDA 1.7 Good DMMXDA 1.8 Good DMMXDA 3.7 Slight curvature DMMXDA 3.5 Slight curvature DMMXDA 3.4 Good, but only 3 data points DMPDA 3.4 Good DMPDA 1.1 Slight curvature p-NMAZ(n = 1) 3.6 Good p-NMAZ (n = 2) 3.5 Good
Figure 5 shows the results for this same set of amines at 0[degrees]C. Again good behavior for analysis by PFO kinetics is observed. In this case DMMXDA is reacting at 3.2 times the rate of MXDA. Comparison of the actual PFO rate constants shown in Table 6 indicates that the reaction of DMMXDA at 25[degrees]C is occurring at about eight times the rate at 0[degrees]C, while the rate constant differences for MXDA (not shown in the table) differ by about 9. This magnitude of change in rate with temperature is not uncommon, and goes a long way toward explaining the very large changes in cure speed observed for epoxy coatings applied under ambient conditions.
[FIGURE 5 OMITTED]
Results of a competition experiment between DMMXDA and MXDA in xylene are shown in Fig. 6. This is an excellent representation of the auto-accelerating effect of the formation of hydroxyl groups (20) on the rate of epoxy-amine reactions observed in non-protic media. Because of the total lack of fit (LoF) to a straight line, a PFO rate constant would be highly misleading if calculated. The slope of the line connecting t = 0 with the first measurement at 15 min of reaction is 0.0011 [min.sup.-1], whereas for 105-120 min the slope is 0.0054 [min.sup.-1]. approximately a fivefold increase in rate. Furthermore, this admittedly crude estimate of the change in rate probably underestimates the effect of hydroxyl ups for at least three reasons: at time 0 there was undoubtedly a small amount of water present; the phenyl glycidyl, which was not purified before use. probably contained some hydroxylic impurities; and calculating the average rate from 0 to 15 min exaggerates the instantaneous rate at t = 0. The initial rate of reaction of MXDA cannot be estimated in a similar manner, since the concentration at 15 min actually appears to be greater than at t = 0. due to experimental error. However, observation of the slope of the lines through the first hour of reaction clearly indicates the rate is extremely slow, even in the presence of the alcohol formed throughout that time by the faster reaction of DMMXDA with PGE.
[FIGURE 6 OMITTED]
To get a more quantitative understanding of the catalytic effect of alcohols on the reaction rate, experiments were conducted in xylene/alcohol mixtures with varying concentrations of IPA (0.5, 1, 2, and 4 M) and benzyl alcohol (BzA, 1, 2, and 4 M). BzA was chosen for this study because it is widely used in coating and civil engineering epoxy formulations as both a plasticizer and cure accelerator. Plasticizers are commonly used in epoxy formulations to prevent vitrification before adequate cure has been achieved, since upon vitrification the rates of chemical reactions decrease by several orders of magnitude. (24) Normally, a molecule with the volatility of BzA would not be chosen as a plasticizer, since it often has a greater contribution to VOC when tested by the ASTM method than do other less volatile plasticizers. Its widespread use provides testament to its utility in improving the rate of cure. To give the reader a sense of the quality of the data for these experiments, the results in 1 and 2 M IPA are shown in Figs. 7 and 8. In 1 M IPA there is significantly more noise in the data than was present when the reaction was run in pure IPA. In 2 M IPA. though the data are less noisy, the data as plotted suggests the possibility of slight curvature, suggesting that some auto-acceleration may be occurring. Thus, in both cases the exact value of the PFO rate constants should be interpreted conservatively. However, there are clear trends from the data, which bear discussion. In going from 1 to 4 M IPA. there is about a 3.4-fold increase in rate for DMMXDA. Over the same range, the MXDA rate increases by a factor of about 3.1. Over the same range in BzA. DMMXDA increases fivefold, and MXDA by 5.5-fold.
[FIGURE 7 OMITTED]
[FIGURE 8 OMITTED]
Figure 9 shows the effect of increasing alcohol concentration for DMMXDA, and Fig. 10 shows the same for MXDA. If the catalytic effect is first order in alcohol, one would anticipate a linear response, crossing the origin at the uncatalyzed rate constant. In this plot, we have set that rate constant at zero, since as discussed above, the uncatalyzed rate is clearly quite low (less than 0.0011 [min.sup.-1] for DMMXDA. and even lower for MXDA).
[FIGURE 9 OMITTED]
[FIGURE 10 OMITTED]
Comparing the slopes obtained by linear least squares fit, BzA has 2.7 times the catalytic activity of IPA with DMMXDA, and 2.6 times with MXDA. This difference may be attributed to the difference in acidity of the alcohols. The p[K.sub.a] of IPA is estimated to be 17.1, whereas that of BzA is 15.4. (25) Others have noted a similar dependence of rate on the acidity of catalysts. (26)
Another commonly used and potent accelerator for epoxy formulations is salicylic acid (SA). Experiments were also run in xylene and 1, 2. and 4 M IPA mixtures with the addition of 0.1 M SA, with very interesting results. In the case of DMMXDA, there was no increase in reaction rate when IPA was present in 40 times the concentration (4 M) of SA. and a 1.3-fold increase in 2 M IPA. Unfortunately, the data with 1 M IPA were deemed too poor to calculate a rate constant. With MXDA. on the other hand. SA had an even stronger catalytic effect. For example, in 4 M IPA/ 0.1 M SA, DMMXDA was only 1.8 times as reactive as MXDA, and for both the 2 and 4 M alcohol levels, the rate was actually higher than the corresponding rate in an equal concentration of BzA without added SA. The fact that there is no difference in rate on going from 1 to 2 M IPA would seem to indicate that the SA-catalyzed pathway is so much lower in energy, that almost all the reactions utilize that pathway even when the alcohol is present at 20 times the concentration of the SA.
A possible explanation of the data is as follows. The p[K.sub.a] of salicylic acid is about 3. so that in the presence of amine, it is essentially completely ionized. Thus, the actual catalyst is not SA, but rather an alkyl ammonium salt. In general, 2[degrees] amines are stronger bases than 1[degrees] amines. In the present case, therefore, it is likely that DMMXDA is protonated to a greater degree than is MXDA. In these experiments, the amines were approximately 0.2 M at the start of the reaction, and SA was 0.1 M, which is a fairly high catalyst concentration. It is reasonable to assume that diamine in its protonated form is much less reactive than free amine. Obviously, the amine group that is actually protonated has no lone pair for reaction. However, the other amine is probably also significantly less nucleophilic due to the inductive effect of the positive charge.
If the difference in basicity for DMMXDA vs MXDA is large enough, there is a significant difference in amine available for reaction with PGE. In the case of DMMXDA in 4 M IPA, it appears that the 40-fold higher concentration of the alcohol relative to SA is sufficient such that SA does not further enhance the rate. On the other hand, the concentration of MXDA available for reaction is less affected, and its rate of reaction increases significantly in the presence of SA even in 4 M IPA/xylene.
Experiments were conducted on a sample of pure N,N'-dimethyl-l,3-propanediamine (DMPDA). The ratio [[[k.sub.DMPDA]]/[[k.sub.MXDA]]] was slightly greater than the ratio [[[k.sub.DMMXDA]]/[[K.sub.MXDA]]] (Table 6). However, the values were so close that DMPDA was run in direct competition vs DMMXDA (Fig. 11) to be sure of the result. The rate of DMPDA was indeed about 1.1 times faster.
[FIGURE 11 OMITTED]
Data for competitive rate experiments with p-NMAz are shown in Fig. 12. In this experiment, the initial concentrations of total amine hydrogen from p-NMAz and MXDA were approximately equal, but of course the concentration of any given oligomer in the polymeric mixture of p-NMAz is much lower than the MXDA concentration. Fortunately, this does not affect the results for [[[k.subn.target]]/[[k.sub.ref]]]. The rates of disappearance of both the n = 1 and n = 2 oligomers were followed. Other oligomers were also present, but their relatively low concentrations led to signal-to-noise difficulties in quantifying their rate of reaction. Both oligomers reacted at about the same rate, which was also about the same as the rate of reaction of DMPDA (see above), which is identical to the n = 0 oligomer.
[FIGURE 12 OMITTED]
Both DMMXDA and p-NMAz are difunctional secondary amines, and hence incapable of crosslinking. They are therefore designed to be utilized as components of mixtures also containing multifunctional amines. Because it has low viscosity, aromatic character which improves compatibility with aromatic Bisphenol-A type epoxy resins, and a record of use in marine coatings applications, MXDA was one of the crosslinking amines chosen for study as a crosslinker. Another crosslinker was mixed polycycloaliphatic polyamines (MPCA). The latter is a complex mixture of cycloaliphatic amines with an AHEW of 57. Representative components of this mixture are shown as follows. Though considerably more viscous than MXDA. MPCA generally imparts even better compatibility with epoxy resins, especially at low temperatures, and can impart excellent solvent resistance and corrosion resistance to epoxy coating formulations.
Using DMMXDA as the chain extender, these components were evaluated as components of mixture design experiments with 10 experimental points across the design space (Table 7) which included all the corners in the design space, and three repeats of the center point. The design was constructed to evaluate both linear and quadratic terms. Each curing agent formulation contained 42.86 parts of BzA to 100 parts of amine mixture (30% of total curing agent) by weight. All the formulations in the mixture designs were tested at calculated 1:1 stoichiometric ratios of epoxy groups to amine-hydrogen, and the use ratios of the mixtures by weight varied from 33 to 53 parts per hundred resin (PHR). This in turn meant that the level of BzA in the total formulation after mixing with epoxy resin varied from 7.5% to 10.5%. The results of this study are shown through the series of three-component contour plots of the design space that follow.
The mix viscosity of the amine curing agent and epoxy resins (Fig. 13) was measured at 25[degrees]C on samples pre-equilibrated for 24 h using a cone and plate viscometer operated at 200 [s.sup.-1]. The samples were thoroughly mixed for 1 min before the measurement was taken. Viscosity varied from about 600 to over 3000 mPa.S. Viscosity increased with increasing MPCA content, and DMMXDA was slightly more effective at reducing viscosity than was MXDA, indicating that DMMXDA is well suited for the formulation of low VOC products.
[FIGURE 13 OMITTED]
Dry speeds were measured under constant temperature and humidity (50-60% RH) conditions as phase 2 thin-film set times (TFST) using a Beck-Koller drying recorder. At 5[degrees]C (Fig. 14), TFST ranged from about 7.5 to over 14 h. Note that fast dry speeds were predicted along the axis connecting 100% DMMXDA and 100% MPCA. This probably reflects the combination of two factors: the fast kinetic rate of reaction of DMMXDA, and the higher weight average molecular weight afforded by MPCA. At 25[degrees]C (Fig. 15) the correlation coefficient was low ([R.sup.2] = 0.45) and the analysis of variance (ANOVA) indicate a LoF. However, the response surface was in reality quite flat, as the data ranged only from 3.4 to 4.2 h, so the LoF is not of significant concern a practically speaking. Although not the very fastest region, the 100% DMMXDA-100% MPCA axis again represented a region of fast cure speed.
[FIGURE 14 OMITTED]
[FIGURE 15 OMITTED]
Persoz pendulum hardness development was measured after 3 days at constant temperature and 50% RH. At 5[degrees]C (Fig. 16), it was shown to increase strongly with MXDA content, with a slight bias toward higher hardness with DMMXDA as opposed to MPCA. Once again at 25[degrees]C, the fit was poor, but this is the result of what is now an extremely flat surface (Fig. 17). Thus, hardness varied only during the time from 320 to 350 s, and all the systems were excellent in this regard.
[FIGURE 16 OMITTED]
[FIGURE 17 OMITTED]
The 20[degrees] gloss obtained with 5[degrees]C cure is shown in Fig. 18. In addition to measuring an appearance property of the coating, in epoxy-amine coating systems low angle gloss measurements are often good indicators of the compatibility of the system, and low gloss is indicative of the presence of surface defects such as exudate and carbamation, all of which negatively impact gloss. At such a low temperature, it can be seen that MXDA had a negative impact, and coatings with high levels of MXDA exhibited very poor compatibility. The best gloss is also exhibited on the 100% MPCA-100% DMMXDA axis. This result is not surprising for MPCA. which is frequently incorporated in curing agents for its excellent low temperature compatibility with epoxy resin. However, the sharp difference imparted by DMMXDA and MXDA would be somewhat surprising if one were to consider chemical compatibility alone, since these amines differ only by the presence of two methyl groups. A likely cause of the improved compatibility is fast reaction of DMMXDA with epoxy resin, which will tend to compatibilize the system and reduce phase separation. It should be noted that these results over the entire mixture design must be interpreted cautiously due to the poor LoF as evidenced by a correlation coefficient of only 0.43. However, the gloss was indeed at its highest on the 100% MPCA-100% DMMXDA axis as indicated in the data in Table 7.
[FIGURE 18 OMITTED]
At 25[degrees]C, the gloss data (Fig. 19) yields an excellent value of [R.sup.2]. The data trend at this temperature is quite similar to that at low temperature, though the best model to fit the data was nonlinear (whereas as a linear model it gave the best fit at 5[degrees]C). This does tend to support the conclusions drawn from the 5[degrees]C data, and again the best gloss and compatibility is shown along the 100% MPCA-100% DMMXDA axis.
[FIGURE 19 OMITTED]
Dynamic mechanical analysis was performed on the free films prepared from the same mixtures employed to measure the coating properties described above, with scans performed from -97 to 200[degrees]C at an oscillation frequency of 1 Hz. Due to experiment scheduling issues, there was some variability in the time the films were analyzed, although since all the films were cured a minimum of 14 days at room temperature, this would be expected to have little effect on the results. The [T.sub.g] was measured as the inflection point in the rapid drop in modulus between the glassy state and the rubbery plateau. Each sample was subjected to two temperature scans. Figure 20 shows the [T.sub.g] obtained during the first temperature scan. The transitions varied from 65 to 77[degrees]C, and the surface shows significant curvature. Since epoxy-amine systems rarely go to complete cure at room temperature, all the systems were then immediately re-scanned through the same temperature range to determine the system's ultimate [T.sub.g] The data, shown in Fig. 21, indicate that there was indeed a significant degree of further cure that occurred during the first DMA scan, and [T.sub.g]'S now vary from 80 to 130[degrees]C. The data is also clearly far simpler to interpret, as the model shows much less curvature, and the correlation coefficient (0.994) is extremely high. This figure clearly indicates that [T.sub.g] shows a strong and almost linear decrease as DMMXDA content is increased. Since [T.sub.g] is strongly dependent on crosslink density. (24) this data confirms the expected effect of DMMXDA as a linear chain extender.
[FIGURE 20 OMITTED]
[FIGURE 21 OMITTED]
This conclusion is further supported through determinations of molecular weight between crosslinks calculated from the minimum modulus in the rubbery plateau. (18) As for the [T.sub.g] measurements, determinations were made both for initial (Fig. 22) and rescan (Fig. 23) of the DMA. In the first scan, weight between crosslinks varied from 726 to 3479, and the plot showed a significant degree of curvature. In the rescan. there is a significant reduction in molecular weight between crosslinks to a range of 581 to 1617 (i.e.. the crosslink density has significantly increased). The rescan data are also simpler to interpret, and display much less curvature than the original scans. As was the case for rescan [T.sub.g] there is a clear and almost linear dependence on the concentration of DMMXDA. as one would expect for a linear chain extender.
[FIGURE 22 OMITTED]
[FIGURE 23 OMITTED]
Certain properties of thermosetting networks can be calculated by the method of Miller and Macosko. (13-15) This approach calculates probabilities, based on the molar compositions and functionality of the components and the extent of reaction, that "looking into" or "looking out from" a particular bond, leads to a finite chain, or instead leads back to the cross-linked network.
Figure 24 displays the calculated crosslink density profile for the experimental design space. The data presented are the inverse of the concentration of effective strands. Effective strands (16) are essentially crosslinking sites, and the units are moles per g. From the DMA, we obtained the average molecular weight between crosslinks. [M.sub.c], which is in units of g per mole. Thus, to compare results, the inverse of [effective strands] is actually presented in Fig. 24. Since [/[[effective strands]]] and [M.sub.c] are not truly identical measures of crosslink density, what is of importance here is not the absolute values shown in the graphs, but the shapes of the curves. In the Miller-Macosko calculations, crosslink density decreases (weight between crosslinks increases) as a function of the concentration of DMMXDA. The crosslink density is essentially independent of whether the rest of the curative is MXDA or MPCA, although there is a small but increasing amount of curvature as the concentration of DMMXDA is diminished.
[FIGURE 24 OMITTED]
When compared to the actual measured values shown in Fig. 23, the similarity is remarkable, especially in light of the assumptions of Miller-Macosko theory such as lack of any side reactions. The principal difference in the actual data is that MPCA led to lower crosslink densities than did MXDA when utilized in the same amount with DMMXDA. For example, the [M.sub.c] of 60 MPCA/40 DMMXDA is predicted by the statistical model derived from the DMA results to be about 920, whereas that of 60 MXDA/40 DMMXDA is predicted to be about 1300.
In general, the mixture design response surfaces obtained from the study of p-NMAz/MPCA/MXDA mixtures (Table 8 and Figs. 25-33) were quite similar to those from the DMMXDA design. In these experiments, the sample of p-NMAz had an AHEW of 100. Mix viscosity (Fig. 25) generated a surface very similar to that from the DMMXDA design, though the viscosity range was significantly lower (340-1620 mPa.s). This is not surprising given the linear structure of p-NMAz and the aromatic structure of DMMXDA and the higher AHEW of p-NMAz relative to DMMXDA. resulting in a coating with a higher weight fraction of chain extender. Thus, p-NMAz appears to be particularly well suited for the development of low VOC formulations.
[FIGURE 25 OMITTED]
[FIGURE 26 OMITTED]
[FIGURE 27 OMITTED]
[FIGURE 28 OMITTED]
[FIGURE 29 OMITTED]
[FIGURE 30 OMITTED]
[FIGURE 31 OMITTED]
[FIGURE 32 OMITTED]
[FIGURE 33 OMITTED]
The best TFST's at 5[degrees]C (Fig. 26) were nearly an hour longer with p-NMAz than with DMMXDA, but the optimum composition was again on the 100% p-NMAz-100% MPCA axis. At 25[degrees]C, cure was slightly faster with p-NMAz, and in this case the correlation was much improved (Fig. 27).
When cured at 5[degrees]C, pendulum hardness (Fig. 28) was generally lower for p-NMAz. This may again be the result of its linear structure. The surface was also more complex than the DMMXDA surface. The hardness obtained with a room temperature cure (Fig. 29), however, was also somewhat lower compared to DMMXDA in formulations containing high levels of chain extender, though this may partly be due to measurement at only one day of cure.
Gloss obtained at 5[degrees]C for p-NMAz (Fig. 30) also led to a far better correlation coefficient than for DMMXDA, though the model showed LoF. Gloss ranges were similar, and MXDA in this case also generally detracted from gloss, with the best gloss values obtained on the 100% p-NMAz-100% MPCA axis, or near to it. At 25[degrees]C. MXDA also reduced gloss and compatibility (Fig. 31).
The rescan [T.sub.g] measurements for p-NMAz yielded a response surface (Fig. 32) very similar in shape to the DMMXDA surface, though [T.sub.g]'s with high levels of chain extender were almost 30[degrees]C lower. The molecular weight between crosslinks (Fig. 33), however, did not follow the rather straightforward increase with chain extender observed with DMMXDA. It is possible that this effect may be attributed to the presence of homopolymerization in p-NMAz systems, as will be shown below.
Unfortunately, flexibility as estimated by direct and reverse impact resistance did not model well for either chain extender. This is perhaps not surprising considering the large number of variables that can affect impact resistance, including not only the inherent material properties of the polymer, but also film thickness, adhesion, temperature, film defects, and various substrate effects. Data from the various mixtures of the DMMXDA design are shown in Fig. 34. Data are shown for films cured for 10 days ambient, as well as 10 days ambient followed by 2 h at 80[degrees]C. All reverse impact values after 10 days ambient cure had very poor impact resistance (about 4 kg cm) and are not shown. Direct impact resistances of coatings cured at ambient temperature showed values around 40-50 kg cm at 60-[micro]m dry-film thickness for most coating compositions. When these panels were submitted to 2 h post-cure at 80[degrees]C. the typical level of direct impact increased to 70-90 kg cm. At the same time, DFT decreased to 50-55 [micro]m. most likely as a result of partial evaporation of BzA from the coating. However, particular coating compositions, especially those without MXDA and where relatively high levels of DMMXDA were used, showed significantly improved impact resistance. Similar trends were found for reverse impact resistance. Likewise, all the samples that showed > 120 kg cm direct impact after the 80[degrees]C cure also passed a 3-mm cylindrical bend test (not shown), whereas all the other samples only passed 16-25 mm mandrel tests. The only sample with a high level (50% or more) DMMXDA that did not show excellent flexibility (first data point) contained a high level of MXDA, and had a low gloss value. Perhaps compatibility-driven adhesion issues may account for the poor flexibility.
[FIGURE 34 OMITTED]
The impact results for p-NMAz with an AHEW of 100 (Fig. 35) display very similar trends, with a high temperature post cure significantly improving results, and with samples prepared from high levels of p-NMAz combined with MPCA showing the best results. However, there were two mixtures in this case that yielded direct impact of over 160 kg cm, suggesting that p-NMAz can lead to even higher flexibility than DMMXDA, though it should be noted that these samples received a 14-day cure vs the 10-day cure for DMMXDA. The data in Fig. 36 show that by increasing the molecular weight of the p-NMAz, even further improvements in impact resistance can be obtained, with values of about 200 kg cm reverse impact obtained even with room temperature cure.
[FIGURE 35 OMITTED]
[FIGURE 36 OMITTED]
The primary purpose of experimental design work (a point that is widely misunderstood) is not to directly come upon a formulation that meets one's needs, but rather to use the models generated to identify an area of design space where the multiple needs normally required for commercial product success may be met. Table 9 shows the results of two DMMXDA formulations selected to have optimum properties (paperchamps), and one selected largely to test the validity of the models in regions of generally poor properties (paperchump). The properties of the clear-coat paperchamps were indeed generally very good. At 25[degrees]C, the properties were very close to those predicted, the biggest difference being in TFST. which had the poorest predictive model. Fortunately the cure times were actually less than predicted. At 5[degrees]C, both gloss and cure speed results were significantly better than predicted. This is not surprising for the gloss value, since that model exhibited LoF. but is somewhat unexpected for TFST. Fortunately, this result is also significantly better than predicted. The [T.sub.g]'s were also very close to predictions. The largest discrepancy in the data is for the molecular weight between crosslinks for the second champ formulation, though the fact that [T.sub.g] was so close to prediction brings into doubt the validity of that particular determination. The paperchump results tended to also validate the models.
Table 9: Model predictions vs actual data of coatings based on DMMXDA Paperchamp 1 Paperchamp 2 MXDA 0 0 DMMXDA 65 75 MPCA 35 25 BZA 42.8 42.8 AHEW 101 103 Predict Actual Predict Actual Mix viscosity 25[degrees]C 1197 1170 908 897 (mPa-s) TFST. phase 2 (h) 25[degrees]C 3.6 3.2 3.6 3.0 Gloss. 20" 25[degrees]C 99 106 108 108 Persoz hardness, 7 25[degrees]C 335 344 337 332 days (s) TFST. phase 2 (h) 5[degrees]C 7.8 7.0 8.0 6.5 Gloss, 20[degrees] 5[degrees]C 66 89 61 84 Persoz hardness, 7 5[degrees]C 113 114 118 103 days (s) [T.sub.g], after 86 87 75 79 rescan ([degrees] C) [M.sub.c], after 1475 1424 1473 1902 rescan (g/mol) Paperchump 1 MXDA 40 DMMXDA 20 MPCA 40 BZA 42.8 AHEW 68 Predict Actual Mix viscosity (mPa-s) 25[degrees]C 2187 2060 TFST. phase 2 (h) 25[degrees]C 3.8 3.8 Gloss. 20" 25[degrees]C 53 45 Persoz hardness, 7 days (s) 25[degrees]C 337 339 TFST. phase 2 (h) 5[degrees]C 13.2 12.1 Gloss, 20[degrees] 5[degrees]C 24 29 Persoz hardness, 7 days (s) 5[degrees]C 155 129 [T.sub.g], after rescan ([degrees] C) 121 122 [M.sub.c], after rescan (g/mol) 620 757
The p-NMAz model predictions and results were likewise in generally good agreement. The largest differences for the champs were in pendulum hardness at 5[degrees]C (both samples), and 20[degrees] gloss (for one of the samples). The chump formulation also showed good agreement with predictions (Table 10).
Table 10: Model predictions vs actual data of coatings based on p-NMAz Paperchamp 1 Paperchamp 2 MXDA 0 0 p-NMAz[1:1] 72 80 MPCA 28 20 BZA 42.8 42.8 AHEW 122 127 Predict Actual Predict Actual Mix viscosity (mPa-s) 25[degrees]C 451 480 331 370 TFST, phase 2 (h) 25[degrees]C 3.3 3.2 3.3 3.0 Gloss, 20 [degrees] 25[degrees]C 90 95 72 100 Persoz hardness, 25[degrees]C 283 290 272 262 1 day (s) TFST, phase 2 (h) 5[degrees]C 8.2 8.7 8.4 8.6 Gloss, 20 [degrees] 5[degrees]C 60 65 50 71 Persoz hardness, 5[degrees]C 61 91 42 74 2 days (s) [T.sub.g], after 52 45 rescan ([degrees]C) [M.sub.c], after 5650 7695 rescan (g/mol) Paperchump 1 MXDA 45 p-NMAz[1:1] 7 MPCA 48 BZA 42.8 AHEW 66 Predict Actual Mix viscosity (mPa-s) 25[degrees]C 1492 1520 TFST, phase 2 (h) 25[degrees]C 3.7 3.6 Gloss, 20 [degrees] 25[degrees]C 18 21 Persoz hardness, 1 day (s) 25[degrees]C 332 331 TFST, phase 2 (h) 5[degrees]C 10.8 10.2 Gloss, 20 [degrees] 5[degrees]C 25 18 Persoz hardness, 2 days (s) 5[degrees]C 92 110 [T.sub.g], after rescan ([degrees]C) 126 [M.sub.c], after rescan (g/mol) 168
Tertiary amine catalyzed epoxy homopolymerization
In the course of this study, stoichiometric ladders of amine-hydrogen to epoxy resin were run (Table 11). The evaluations were performed using a curing agent consisting of a mixture of 52.5 parts p-NMAz, 17.5 parts MPCA. and 30 parts of BzA. Significantly increasing the ratio of epoxy to amine resulted in longer drying times but also increased hardness. These results show that it may be beneficial to deviate from stoichiometric epoxy to amine ratios when optimizing coating performance. For example, improvement in Persoz hardness is beneficial assuming that cure speed can be retained sufficiently.
Table 11: Clearcoat performance as a function of stoichiometric ratio of p-NMAz amine-hydrogen to epoxy resin p-NMAz[1:1] 52.5 MPCA 17.5 BzA 30.0 Epoxy:amine groups 1:0.7 1:0.8 1:0.9 1:1.0 Use Level with LER (PHR) 44 50 56 62 Thin film set time 9.9 8.9 8.0 8.0 at 5[degrees]C (h) Persoz hardness (s) Day 1/day 7 at 5[degrees]C 40/185 55/185 50/175 50/125 Day 1 at 25[degrees]C 320 290 275 205
It is well known that tertiary amines can catalyze the homopolymerization reaction of epoxy resins as illustrated in Scheme 6.
Since the p-NMAz repeat unit contains a methyl-substituted tertiary amine, we investigated whether a significant amount of epoxy homopolymerization might be occurring where it was used as a substantial portion of a curing agent formulation. Figure 37 shows plots of absorption vs time for epoxide (2280 nm) and 2[degrees] amine (1540 nm) absorption bands from the near 1R spectra of mixtures of p-NMAz and excess LER. To aid the reader, the absorbance at 1540 nm as shown has been multiplied 10-fold, since this is a much weaker absorbance than the epoxy band. Two ratios of p-NMAz to epoxy resin were studied: 10% and 30% by weight, which correspond to 16% and 49% 2[degrees] amine-hydrogen to epoxy stoichiometric ratios, respectively. The mixtures were cured at 40[degrees]C, and spectra were obtained every 10 min. The figure also shows percent conversions at 100 and 350 min, obtained by dividing absorbance at time t by absorbance at [t.sub.0].
[FIGURE 37 OMITTED]
The 2[degrees] amine is consumed rapidly, and by 100 min, conversions are 86.8% and 96.2% for the 49% and 16% stoichiometric ratios, respectively, and nearly quantitative conversion is achieved after 350 min. At the 100 min mark, some 22% and 54% of the epoxide groups have been consumed. However, based on the initial stoichiometric ratios and amine conversions at this point, reaction with amine could account for only 15% and 43% loss of epoxide. Despite the fact that most of the amine has been consumed within 100 min. significant amounts of epoxide groups continue to be consumed well after this point. For the mixture with 49% 2[degrees] amine, ultimately 79% of the epoxide groups are consumed, or 30% more than can be accounted for by the amine hydrogen available. For the mixture with a 16% ratio of amine-hydrogen to epoxy. 56% of the epoxy groups are consumed. 40% of which cannot be accounted for because of reaction with amine. These data provide strong evidence that p-NMAz is a potent catalyst for epoxide homopolymerization, and may explain why better hardness is obtained in clearcoats formulated with a large excess of epoxy resin, though the high level of flexibility and relatively low [T.sub.g] imparted by p-NMAz may also be contributing factors.
A high level of protection from corrosion is often a requirement for epoxy-based coatings used in marine, industrial maintenance, and other applications areas. Electrochemical impedance spectroscopy has been proven to be an effective tool to rapidly screen coatings for their corrosion resistance properties. Coatings with good corrosion resistance properties show linear or near-linear behavior in a plot of log Modulus (ohm) vs log frequency even at frequencies as low as 0.1 [s.sup.-1] (Bode plot). (27), (28)
Figure 38 shows impedance spectra of one of the p-NMAz paperchamp compositions at different times of testing with 3 wt% NaCl solution. Even the initial measurement is showing poor corrosion resistance properties, and by 1 h, there is clear evidence of corrosion processes occurring at the metal surface. Shown for reference is a clearcoat based on MPCA and BzA after 24 h of testing. The cause for the poor corrosion resistance imparted by p-NMAz is most likely the result of the hydrophobic nature of the tertiary amines in the repeat unit structure. Reduction of [T.sub.g] and crosslink density, which will increase diffusion rates of water and ionic species in the film, may also play a role. Epoxy homopolymerization as noted above may also have played a role, though attempts to markedly improve the EIS properties by changing the epoxy/amine stoichiometry of the formulation have so far been unsuccessful.
[FIGURE 38 OMITTED]
Impedance experiments were performed on a 75 DMMXDA/25 MPCA/42.8 BA composition as well as a clearcoat prepared from DMMXDA and a phenalkamine curing agent Sunmide[R] CX-105 (Air Products and Chemicals, Inc.). Like other phenalkamines, CX-105 is a Mannich base derived from cardanol, a meta-substituted phenol containing a [C.sub.15] alkyl chain with one to three double bonds separated by a methylene group (the same arrangement found in naturally occurring unsaturated fatty acids like oleic, linoleic, and linolenic acids). Phenalkamines are utilized extensively in the marine coatings field due to their excellent corrosion resistance and reasonably good low-temperature cure properties, though there is a need to further improve their low temperature cure capabilities.
The impedance spectrum of a 75/25 mixture of DMMXDA/CX-105 cured with epoxy resin is compared to the spectrum of a 100% CX-105--epoxy clearcoat in Fig. 39. After 24 h immersion, the pore resistance of the DMMXDA coating is significantly better than that of the CX-105 by itself, even at this high level of modification.
[FIGURE 39 OMITTED]
Impedance results were also measured for a 75 DMMXDA/25 MPCA/42.8 BA composition (spectrum not shown). The pore resistances were 2.5 x [10.sup.9] and 2.35 x [10.sup.8] after 1 and 24 h immersion, respectively. While not quite as high as for the DMMXDA/CX-105 composition shown in Fig. 39, this clearcoat demonstrates far better corrosion-resistance properties than any of the formulations based on p-NMAz, which were tested. It is likely that the lower pore resistance compared to DMMXDA/CX-105 was the result of the BzA plasticizer in the formulation.
Mannich base chemistry of dimethyl secondary diamines
Mannich base curing agents are utilized extensively in commercial epoxy-resin formulations, especially where fast cure speed is required at low ambient temperatures. There are two general routes for the preparation of Mannich bases. In the direct route, phenol (or some type of substituted phenol) is condensed with formaldehyde and polyamine. A wide variety of polyamines, including the polyethyleneamines (ethylenediamine, diethylenetetramine. etc.) and various cycloaliphatic amines (e.g., isophorone diamine) are employed in the reaction.
The phenol can be substituted as usual at the ortho and para positions with up to three substituents per phenol, and a primary amine can react with one or both of its available amine hydrogens. Thus, when polyamines are employed it is quite possible to gel such a reaction mixture. In practice, this means that such Mannich bases are always prepared using starting ingredient ratios and reaction conditions that result in substantial amounts of residual phenol in the final product. This is an undesirable result since phenol is highly toxic, and Mannich bases derived from phenol always require the use of extensive personal protective equipment, toxic labeling, etc., all of which reduce the market for their use.
The exchange route to Mannich bases employs tertiary amine Mannich bases derived from volatile secondary amines such as dimethylamine. Fortunately, such a Mannich base--tris-(dimethylamino-methyl)phenol (TDMAMP)--is commercially available. The reaction is conducted by heating TDMAMP in the presence of a polyamine, with loss of dimethylamine. With this route, it is possible to prepare phenol-free Mannich base curing agents, but it results in a much more expensive product than does the direct route.
We first explored the preparation of Mannich bases derived from p-NMAz by the exchange route. As expected, the reaction proceeded smoothly to yield products with the structure shown below.
Although interesting, it would be more likely to be of commercial significance were it to prove possible to prepare such Mannich bases directly from p-NMAz, phenol and formaldehyde, but under conditions that would result in free phenol content of less than 1% (the limit currently requiring mandatory labeling as toxic in the United States). Interestingly, this did prove to be possible, and by optimization of starting point formulations and reaction conditions, Mannich bases could routinely be prepared by the direct route that contained less than 0.1% residual phenol, less than one tenth of the target maximum. Though the exact composition of starting formulas and process conditions is proprietary, the reaction times were well within necessary limits for practical commercial production. The particular p-NMAz Mannich base utilized in the formulations described below had a viscosity of 750-1250 mPa s, and an AHEW of 175.
Through experimental design optimization, a paper-champ formulation consisting of 50 parts CX-105, 37.5 parts of a p-NMAz Mannich base, 12.5 parts of MPCA, and 5 parts of Epodil [R] L hydrocarbon resin diluent (Air Products and Chemicals, Inc.) was developed. In Table 11, its clearcoat properties are compared with clearcoat formulas based on 100% CX-105, and a blend of CX-105 plus 5 parts of Ancamine [R] K54 tertiary amine accelerator. The latter formulations were employed at 1:1 stoichiometry as would normally be recommended, whereas the p-NMAz MB formulation utilized a considerable stoichiometric excess of epoxy to account for the homopolymerization tendencies of p-NMAz discussed earlier.
Table 12 shows the results obtained. In this case, the far more aggressive dry-to-handle time, as measured by the thumb-twist method, was utilized to evaluate cure speed. By itself, CX-105 took over 30 h to pass this test. When modified with 5% tertiary amine accelerator, cure time was reduced to 16 h, but the p-NMAz MB came close to matching this result (2 h). Unfortunately, as is usually the case, addition of the accelerator embrittled the film, resulting in only 50 kg cm direct, and <5 kg cm reverse impact, respectively. The best impact was obtained from the p-NMAz MB formulation with 100/20 kg cm direct/reverse impact, respectively. The pore resistance for all the formulations was the same, whereas water uptake for the p-NMAz MB formulation was slightly higher than the other formulations. Thus it proved possible to incorporate p-NMAz in a Mannich base, and by formulating it with a hydrophobic phenalkamine curative, take advantage of the cure speed and flexibility offered by the p-NMAz structure while maintaining promising corrosion resistance properties as indicated by EIS studies.
Table 12: Clearcoat results for phenalkamine curing agent formulations Condition Standard Catalyzed Std + p-NMAz phenalkamine phenalkamine phenalkamine Sunmide CX-105 100 100 50 Ancamine K54 5 p-NMAz 37.5 phenalkamine MPCA 12.5 Epodil L 5.0 Total 100 105 105 PHR with LER 76 76 Stoichiometry 100 100 75 (% of theoretical) Cure speed (h) Dry-to-handle 5[degrees]C >30 16 22 Impact flexibility Direct (cm kg) 14 days 80 <50 100 Reverse (cm kg) 14 days 10 <5 20 Pore resistance ([OMEGA]) 24 h 7 days 2.0 x 2.0 x 2.0 x immersion [10.sup.+09] [10.sup.+09] [10.sup.+09] Water uptake (wt%) 24 h immersion 7 days 4 5 8
In this work the utility of dimethyl secondary diamine chain extenders to modify the properties of epoxy thermoset coatings has been explored. The potential range of properties that can be obtained were evaluated through two new raw materials, both of which can be prepared from reasonably inexpensive and readily available starting materials. Dimethyl-meta-xylylenediamine was obtained by reductive hydrogenation of isophthalonitrile in the presence of methylamine. A new oligomerization reaction which yielded p-NMAz was discovered when it was attempted to similarly prepare 1,3-PDA from hydrogenation of the Michael adduct of acrylonitrile and methylamine. With PFO kinetic competition experiments, it was demonstrated that methyl substitution increased the rate of reaction of these amines with aromatic glycidyl ethers by a factor of about 3, when compared to MXDA.
Clearcoat performance was optimized through the use of a combination of statistical design experimental strategies. Dynamic mechanical analysis of these clear-coats showed that as expected, the dimethyl secondary diamines reduced crosslink density (increased molecular weight between crosslinks) and [T.sub.g]. By modeling the network properties of the coatings from this design space using Miller-Macosko calculations, it was shown that the response surface for l/[effeetive strands] was qualitatively very similar to the surface obtained for molecular weight between crosslinks, thereby confirming that our theoretical understanding of the effect of chain extenders on network structure was at least qualitatively verified. Clearcoats with an excellent balance of fast cure speed and gloss could be obtained even at temperatures of 5[degrees]C. While DMMXDA could significantly improve the flexibility of coatings as measured by direct and reverse impact resistance, p-NMAz, because of its linear structure, gave even higher levels of impact resistance, as well as lower viscosity. On the other hand, incorporation of p-NMAz in the coatings generally led to poor corrosion resistance as measured by EIS. whereas coatings with excellent properties in EIS were obtained from DMMXDA.
It proved possible to prepare Mannich base curing agents with very low levels of residual phenol from p-NMAz by the economical direct route. The p-NMAz Mannich base was also evaluated in blends with a phenalkamine curing agent to prepare epoxy clear-coats. Through experimental design optimization, it proved possible to prepare clearcoats that exhibit an excellent balance of low temperature cure and flexibility, while maintaining the same pore resistance as the unmodified phenalkamine.
Thus, in situ chain extension with dimethyl secondary diamines would appear to be a promising concept for the development of high solids coatings that possess an intriguing combination of performance properties.
"No man is an island, entire of itself.
Everyone is a piece of the continent, a part of the main. ...
Therefore, never send to know for whom the bell tolls;
It tolls for thee." John Donne, 1623
There is only a single Mattiello awardee each year, but in this case the award lecture is based upon the work of many. Gamini Vedage led the synthesis efforts with the help of Gene Lutz. Applications work was directed by Rob Rasing with the able assistance of Edwin Lijffijt. 1 had the pleasure of working with two very skilled bench chemists in this project. Steve Boyce and Renee Keller. However, the results described here were truly a group effort, with all these people plus Michael Cook, a very skillful manager in Air Products' epoxy R&D effort, participating in regular sessions of brainstorming and data evaluation, where insights and ideas were shared with great generosity. The results would not have been possible without all of their contributions. John Marsella and Bill Starner (now with CVC Specialty Chemicals) made earlier contributions to the concept of methyl-substituted amines as useful curing agents, and were the first to note their effect on cure speed. We were also aided, as usual, by the fine department of analytical chemists at Air Products. Most notably, Chris Walsh conducted the DMA experiments, Gary Johnson conducted the near IR evaluations, and Anne Kotz ran and interpreted NMR experiments, which though not discussed in this article, were absolutely essential to the work.
Throughout my career I have been honored to work with, and more importantly learn from, many excellent colleagues. At Air Products. I wish especially to thank Charlie Hegedus, Ernie Galgoci, Dave Dubowik. Dilip Shah, Ellen O'Connell, Pete Lucas. Frank Pepe, John Dickenson, and Herb Klotz. I've also worked with many line business people, including Bob Thomas, Zay Risinger, Kurt Junge, Bruce Thoet, and Neil Hunt. I am well aware that without their successful leadership of our business from a commercial perspective, the financial backing to conduct this work would not have been available.
I began working in the coatings industry at the Rohm and Haas Corp., and am deeply indebted to the members of that line research organization for teaching me how to conduct industrial research. Bill Emmons (now deceased) was a true intellectual giant in the field, and a source of great advice, who took an almost fatherly interest in my career, as he did for many others. I got wonderful guidance from other managers there, including Andy Mercurio. Dave Clemens, and Jerry Levy, and benefited from a host of great colleagues, including Alvin Lavoie, Dan Bors. Willie Lau, Joe Tanzer. Gary Larson. CJ Chang, and many others. The success these folks have had in their careers is well deserved, and has been a pleasure to watch. At Akzo Coatings in Troy, Michigan. I reported to the VP of R&D. John Gardon, who was a great friend as well as a continual source of technical inspiration, who never failed to pick my spirits up when I needed it most with his wonderful sense of humor. I was also lucky to have been trained as a scientist by a remarkable group of professors. I spent several years "borrowing" Josef Michl's equipment at the University of Utah. His remarkable intelligence was inspiring, and not a little intimidating. Ken Wiberg, my graduate school advisor at Yale University, combined remarkable patience with an encyclopedic knowledge of all the aspects of organic chemistry, and quite simply taught me to become a chemist. At Bloomfield College, Ron Trost, a professor of experimental psychology--my undergraduate major--took an interest in me. and encouraged my interest in science. He was also responsible for hiring Mel Winokur, who was quite simply the best teacher I've ever known. It was Mel's ability to bring alive the fascinating process by which chemists unravel nature's secrets as they unfold in the world of organic chemistry that inspired me to pursue a career in that science.
I sincerely thank all of you.
(1.) Carson, R, Silent Spring. Houghton Mifflin. New York (1962)
(2.) Lear, L, Rachel Carson: Witness for Nature. Henry Holt, New York (1997)
(3.) Hynes, HP, The Recurring Silent Spring. Pergamon Press. New York (1989)
(4.) Murphy, PC. What a Book Can Do: The Publication and Reception of Silent Spring. University of Massachusetts Press. Amherst (2005)
(5.) Dubowik, DA, Walker. FH. Starner, WE. "Novel Curing Agent Technology for High Solids 6. Epoxy Coatings." Proc. Waterborne, High Solids and Powder Coating Symp., Vol. 26, pp. 13-25 (1999)
(6.) Rozenberg, BA, "Kinetics. Thermodynamics and Mechanism of Reactions of Epoxy Olilgomers with Amines." Adv. Polym. Sci., 75 113-165 (1986). doi:10.1007/BFb0017916
(7.) Moore, WJ, Physical Chemistry, 4th ed., pp. 334-336. Prentice-Hall. Englewood Cliffs. NJ (1972)
(8.) Tanaka, Y, Bauer. RS. "Curing Reactions." In: May. CA (ed.) Epoxy Resins Chemistry and Technology, 2nd ed., pp. 293-304. Marcel Dekker, New York (1988), and references therein
(9.) Burrell, H, "Glass Transition in Coating Formulations." Official Digest, 34 (445) 131-161 (1962)
(10.) Bird. RB. Armstrong, RC. Hassager, O, Dynamics of Polymeric Liquids, pp. 190-194. Wiley. New York (1977)
(11.) Walker, FH, Introduction to Polymers and Resins. 2nd ed., pp.13-14. Federation of Societies for Coatings Technology (1999)
(12.) Odian, G, Principles of Polymerization, 2nd ed., pp. 113 126. John Wiley and Sons. New York (1981)
(13.) Macosko, CW, Miller. DR. "A New Derivation of Average Molecular Weights of Nonlinear Polymers." Macromolecules. 9 199-206 (1976)
(14.) Miller, DR, Macosko. CW, "A New Derivation of Post Gel Properties of Network Polymers." Macromolecules. 9 206-211 (1976). doi:10.1021/ma60050a004
(15.) Miller, DR, Valles, EM. Macosko. CW, "Calculation of Molecular Parameters for Stepwise Polyfunctional Polymerization." Polym. Eng. Sci., 19 (4) 272-283 (1979). doi:10.1002/pen.760190409
(16.) Bauer, DR, "How to Calculate Crosslink Structure in Coatings." J. Coat. Technol., 60 (758) 53 66 (1988)
(17.) Miller, DR, Macosko. CW. "A New Derivation of Post Gel Properties of Network Polymers." Macromolecules. 9 206-211 (1976). doi:10.1021/ma60050a004
(18.) Hill, LW, "Structure/Property Relationships of Thermoset Coatings." J. Coat. Technol., 64 (808) 28 42 (1992)
(19.) Marsella, JA, Starner. WE. "Acceleration of Amine/Epoxy Reactions with N-Methyl Secondary Amines." J. Polym. Sci. A: Polym. Chem.. 38 921-930 (2000). doi:10.1002/(SICI)1099-0518(20000301)38:5<921::AID-POLA17>3.0.CO:2-P
(20.) Xu, L, Fu. JH. Schlup. JR. "In situ Near-Infrared Spectroscopic Investigation of Epoxy Resin-Aromatic Amine Cure Mechanisms."J. Am. Chem. Soc., 116 2821-2826 (1994). doi:10.1021/ja00086a015
(21.) Horie, K, et al.. "Calorimetric Investigation of Polymerization Reactions. III. Curing Reaction of Epoxides with Amines." J. Polym. Sci. A-l, 8 1357-1372 (1970)
(22.) See for example Paz-Abuin. S. et al., "Autoacceleration and Inhibition: Free Volume. Epoxy-amine Kinetics." J. Polym. Sci A Polym. Chem., 36 1001-1016 (1998). doi:10.1002/(SICI) 1099-0518(19980430)36:6<1001::AID-POLA15>3.0.CO;2-9
(23.) Moore, WJ, Physical Chemistry, 4th ed., pp. 333-334. Prentice-Hall, New Jersey (1972)
(24.) Wisanrakkit, G, Gillham, JK. "The Glass Transition Temperature (Tg) as an Index of Chemical Conversion for a High-Tg Amine/Epoxy System: Chemical and Diffusion Controlled Reaction Kinetics." J. Coat. Technol., 62 (783) 35 50 (1990)
(25.) Buckingham, J, MacDonald, F (eds.), Dictionary of Organic Compounds, 6th ed. Chapman and Hall. London (1996)
(26.) Bowen, DO, Whiteside. RC. Jr. In: Lee, H (ed.). Epoxy Resins, ACS Advances in Chemistry Series, pp. 48-59. Washington, DC (1970)
(27.) Monetta, T, Bellucci, F, Nicodemo. L, Nicolais, L, "Protective Properties of Epoxy-based Organic Coatings on Mild Steel." Prog. Org. Coat., 21 353-369 (1993). doi:l0.10l6/0033-0655(93)80050-K
(28.) Rammett, V, Reinhard, G, "Application of Electrochemical Impedance Spectroscopy (EIS) for Characterizing the Corrosion-Protective Performance of Organic Coatings on Metals." Prog. Org. Coat., 21 205-226 (1992). doi:10.1016/0033-0655(92)87005-U
This paper was presented as the Mattiello Memorial Lecture, presented at 2008 FutureCoat! Conference, sponsored by Federation of Societies for Coatings Technology. October 15-16, 2008. in Chicago, IL.
F. H. Walker ([??]), M. Cook, G. Vedage, R. Rasing
Air Products and Chemicals, Inc., 7201 Hamilton Blvd., Allentown, PA 18195, USA
* The United Nations celebrates Earth Day on the March equinox. This tradition was started by peace activist John McConnell in 1969.
* The complication of primary amine converting to secondary amine is ignored in this analysis.
|Printer friendly Cite/link Email Feedback|
|Author:||Walker, Frederick H.; Cook, Michael; Vedage, Gamini; Rasing, Rob|
|Date:||Sep 1, 2009|
|Previous Article:||Study of corrosion and friction reduction of electroless Ni-P coating with molybdenum disulfide nanoparticles.|
|Next Article:||Confocal Raman microscopy study of the melamine distribution in polyester-melamine coil coating.|