SEARCH

SEARCH BY CITATION

Keywords:

  • ageing;
  • calorimetry;
  • glass transition;
  • nanocomposites;
  • reinforcement

Abstract

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. EXPERIMENTAL
  5. RESULTS AND DISCUSSION
  6. CONCLUSIONS
  7. Acknowledgements
  8. REFERENCES AND NOTES

Although the effects of filler nanoparticles size and surface treatment on the glass transition temperature of the matrix phase have received well-deserved attention, nanofiller effects on other physical parameters associated with the glass transition have received less interest. To better understand how the incorporation of nanofillers affects the enthalpic relaxations associated with the glass transition, differential scanning calorimeter measurements were carried out on silica–polyvinyl acetate nanocomposites with respect to filler content, annealing temperature, and annealing period. As expected, longer annealing periods below the glass transition temperature resulted in an increase of the subsequent enthalpic relaxations. However, the presence of filler substantially reduces the enthalpic relaxation relative to that of the neat polymer at longer annealing periods only. The underlying enthalpic relaxations and the effects suppressed by the fillers are specific to the annealing temperature. These results suggest a significant alteration in the physical state of the matrix because of the presence of the filler particles. However, this does not imply the existence of a glassy layer or layers with a glass transition gradient near the filler surface. © 2008 Wiley Periodicals, Inc. J Polym Sci Part B: Polym Phys 46: 2733–2740, 2008


INTRODUCTION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. EXPERIMENTAL
  5. RESULTS AND DISCUSSION
  6. CONCLUSIONS
  7. Acknowledgements
  8. REFERENCES AND NOTES

The mechanisms responsible for reinforcement in filled systems and the origin of the Payne effect, namely, the reduction in reinforcement with strain amplitude, have been topics of great debate. Reinforcement effects are more pronounced in nanocomposites where the fillers posses relatively large surface area per unit volume. Several models have been proposed to explain these phenomena. These models can be categorized based on the root origin of the phenomena, that is, some of them are based on the spatial distribution of the fillers, such as agglomeration, deagglomeration, filler percolation threshold, or the alteration of chain dynamics near the filler surface, with both short- or long-range effects and with finite bond lifetime or permanently adhered bonds that form glassy layers near the filler particles.1–4

The presence of a glassy layer at the filler surface as the underlying mechanism of reinforcement has been proposed by several authors.5–8 The model states that a glassy layer or layers with a gradient of glass transition temperature persists near the walls of the filler even if the temperature is above the glass transition temperature of the neat or composite material. Struik, for example, argues that the presence of these layers is evident from the physical ageing behavior portrayed by the filled system at temperatures above the glass transition temperature of the neat system, thereby leading him to believe that there is a distribution of the glass transition temperature.5 However, he argues a shift in glass transition temperature can be observed only if the glassy layer extends over the whole amorphous region. Similarly, he draws a conclusion on the ageing of semicrystalline polymer above the glass transition due the same glassy layer around the crystalline domains. More recently, NMR measurements indicate a decrease in chain mobility and some ordering of chains near filler surfaces.6–8 Of major importance is the issue of whether or not a decrease in mobility necessarily implies the formation of a glassy layer, and if it does, whether or not the glassy layer is of sufficient extent and stiffness to account for the very high reinforcement values observed above the glass transition temperature. In this regard, there are several observations that have been made in filled elastomers well above the glass transition temperature that suggest alternate explanations to the formation of a glassy layer.3, 4 These explanations are based on the labile bonding of the polymer to the filler surfaces and the resultant effects this bonding has on the reduction of conformational entropy of the matrix chains.4 Thus, entropic hardening is the result of enthalpic bonding of the polymer segments to the filler surfaces, but this bonding need not produce a glassy layer to be effective at producing high reinforcement through far-field interactions transmitted by chain entanglements.

Of major concern with the concept of a glassy layer as a means of reinforcement is the fact that the glass transition in itself is not well understood. It has been shown that the glass transition temperature depends on the method employed to determine it.9 For example, glass transition temperatures measured from calorimetric measurements may not necessarily be equal to the temperatures obtained from mechanical measurements at a given frequency. Furthermore, the glass transition temperatures of polymers depend on the physical dimensions of the test specimens10, 11 as well as their state of physical confinement.12 In comparison with the glass transition temperature of bulk polymers, the glass transitions of physically restricted polymers have been observed to increase, decrease, remain the same, or go undetected. A topical review article is presented in ref.13. In general, however, it seems that a strong interaction between the polymer and the surface tends to increase the glass transition temperature. Likewise, for polymer nanocomposites their glass transition temperatures has been observed to increase,14 decrease,15 disappear,14 or remain unaltered.16

Although the effects of physical size, confinement, and incorporation of fillers on glass transition temperature have received major attention, physical ageing of these materials has not received adequate investigation. All polymers exhibit significant physical ageing in their glassy state.5, 17 For instance, dynamic dielectric measurements of thin poly(methyl methacrylate) films show that the relaxation due to ageing decreases with decreasing film thickness.18 Fictive temperature analysis suggests that the rate of physical ageing in confined ortho-terphenyl, o-TP, is faster than in the bulk (unconfined) o-TP.19 Fluorescence measurements of alumina poly(2-vinyl pyridine) nanocomposites, however, show significant reductions in the physical ageing rates in comparison with neat poly(2-vinyl pyridine).20

In this article, we present our enthalpic relaxation measurements of fumed silica filled poly(vinyl acetate) nanocomposites with different filler concentrations after annealing at different temperatures and for several annealing periods. This work is motivated by the following key issues: (a) if there is a glassy layer or layers around the filler particles, then these layers should influence the rate of the physical ageing at a given annealing temperature, (b) if there are glassy layers with a distribution of glass transitions, then their physical ageing should manifest distinguishably for different annealing temperatures, and (c) if there is a glassy layer around a filler, then the observed effect on physical ageing should systematically increase with filler content.

EXPERIMENTAL

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. EXPERIMENTAL
  5. RESULTS AND DISCUSSION
  6. CONCLUSIONS
  7. Acknowledgements
  8. REFERENCES AND NOTES

Polyvinyl acetate (PVAc) with a number-average molecular weight of 83 kDa was purchased from Aldrich and fumed silica particles with average base-particle diameter of 15 nm were obtained from Cabot. The materials used and details of composite sample preparation are the same as presented by Sternstein and Zhu.4 However, a brief outline is given here.

Filler concentrations in the composites are reported as parts per hundred by weight of the matrix resin, or phr. The volume concentration can be approximated from the phr value by dividing by 2, thus, 25 phr is approximately 12.5 vol %. To appropriately disperse these fillers, the desired amount of filler was dispersed in acetone and mechanically stirred with about 17% of the required matrix weight. The solution was then ultra-sonicated in a temperature-controlled bath. The remaining amount of matrix was added with additional acetone and mechanically stirred until the polymer was dissolved. This mixture was again ultrasonicated and then poured in Teflon Petri dishes for air drying. The air-dried composites were transferred into a vacuum oven and dried for at least 12 h at 110 °C. The dried samples were ground to produce powders, from which ∼ 1 mm thick films were compression molded at 110 °C under slight pressure.

Power compensated Differential Scanning Calorimeter, Perkin–Elmer Diamond DSC, was used to evaluate the thermal properties of the materials. Experimental heating and cooling protocols followed in the DSC under nitrogen atmosphere are depicted in Figure 1. Any residual thermal history was first erased by heating the specimen to 110 °C at 10 °C per minute followed by isothermal annealing at 110 °C for 20 min. The samples were then cooled to the annealing temperatures at 10 °C per minute and held for various annealing periods. This was followed by additional cooling to 10 °C at 10 °C per minute. Subsequent heating at 5 °C per minute is considered second heating and provides the data used for analysis. The final cooling to 10 °C was performed to allow the fully developed steady-state controlled heating rate at 5 °C per minute to be achieved before the sample reached the 30–45 °C region in which the hysteresis occurs.

thumbnail image

Figure 1. DSC protocols used to erase thermal history and the subsequent cooling to the annealing temperature followed by the annealing period and second heating.

Download figure to PowerPoint

RESULTS AND DISCUSSION

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. EXPERIMENTAL
  5. RESULTS AND DISCUSSION
  6. CONCLUSIONS
  7. Acknowledgements
  8. REFERENCES AND NOTES

Figure 2(A) shows the complete heating and cooling curves of the neat polyvinyl acetate, and Figure 2(B) gives an enlarged view of the enthalpic hysteresis curves from the second heating obtained by repetitive annealing cycles with increasing annealing period times on the same DSC specimen without removal from the apparatus. The glass transition temperature, as determined from the half Cp extrapolation or inflection point, is about 40 °C and does not change significantly with the annealing period time. The large jump in heat flow at the beginning of the first heating, the large drop in heat flow at the end of the heating cycle, and the spike in heat flow at the end of the cooling cycle are due to instrumental instabilities associated with a startup, shutdown, or change in direction of temperature programming and have no physical significance. Note must be made that the annealing step is masked between the two cooling steps, that is, cooling from 110 °C to 30 °C and cooling from 30 °C to 10 °C. The necessity to cool the samples further to 10 °C before the second heating is to avoid this overwhelming stabilization signal in the range of temperatures of interest. The time required to cool the samples from the annealing temperature to 10 °C and then initiate the second heating is substantially smaller than the annealing period time, and hence is considered to have minimal effect for all practical purposes on the reported annealing times or resultant hysteresis values.

thumbnail image

Figure 2. A: DSC heating and cooling traces of neat PVAc during several cycles with various annealing periods (10, 20, 30, 60, 120, and 240 min) at 30 °C conducted on the same DSC specimen without removal from the calorimeter. B: Expanded view of the data shown in (A) showing the hysteresis curves from the 2nd heating run obtained by multiple annealing cycles on the same DSC specimen of neat PVAc. The sample was annealed at 30 °C for 10, 20, 30, 60, 120, and 240 min before heating.

Download figure to PowerPoint

From the second heating cycle, Figure 2(A), it can be seen that the system has stabilized before it reaches the glass transition temperature range. The second heating, which terminates at 110 °C and is immediately followed by annealing at 110 °C for 20 min, is used as a means of rejuvenating and erasing the thermal history of the sample. Cycling the sample in this way, but of course changing the annealing period at 30 °C, is thus used to study the ageing process at 30 °C with minimal introduction of systematic and random errors. This is evident in Figure 2(A), where there are five indistinguishable cooling traces. This suggests that there is no significant change in the baseline of the machine, there is no significant change in the thermal contact of the sample with the sensors, and the technique used to erase the thermal history of the sample is effective enough for the purposes intended in the analysis to follow. Furthermore, repeated cycles under identical cooling, annealing, and heating conditions of a given sample were reproducible to within 5% of the area analysis to be presented.

Overlaid in Figure 2(A), there are six second heating curves, with the region of interest around the glass transition temperature expanded in Figure 2(B). These are distinguishable only because of the different degrees of overshoot (i.e., hysteresis) around the glass transition corresponding to the different annealing periods at 30 °C. Apparently, there is no difference among the curves in the heat flow at temperatures below and above the glass transition, granted the initial heat flow is dictated by the instrument configuration. Physical ageing did not result in any significant change of the glass transition temperature of the materials, as determined from half Cp and inflection point. No change in the glass transition temperature of the composite materials due to incorporation of the silica particles was observed, the results being similar to what is shown in Figure 2(B) for the neat polymer. A typical comparison is presented in Figure 3. In this case, all of the samples have been annealed at 30 °C for 240 min before each heating curve. These curves have been shifted vertically for presentation purposes. Clearly, there are no significant differences among the curves other than the magnitudes of the normalized overshoot areas.

thumbnail image

Figure 3. Heating curves of neat PVAc and PVAc silica nanocomposites after annealing at 30 °C for 240 min. The curves have been shifted vertically for presentation purposes.

Download figure to PowerPoint

To quantify the relationship of the overshoot with respect to the annealing period, we have adopted a technique similar to that used by Petrie and coworker,21, 22 but using a different algorithm for evaluating the area under the curve. Petrie and coworker21, 22 compute the areas by using a sigmoidal baseline. However, the choice of sigmoidal baseline is not unique. It depends on the choice of the left and right temperature limits used to compute the area. This is problematic if the baseline of the system changes from sample to sample. Although, in our case, the baseline does not change at least within a given cycling of a sample, Figure 2(A), it is conceivable if not usual to expect change in the baseline due to difference in thermal contact. A typical example can be seen in Petrie's work.21 Comparison of Figure 3 with Figure 4 of ref.21 indicates the overall baselines had positive and negative slopes, respectively. It is for similar reasons that annealing the specimen in the DSC without removing it and cycling it repeatedly is found to be a more attractive procedure for obtaining data on the quantitative effects of annealing conditions (temperature and time) on hysteresis at the very least within one sample.

thumbnail image

Figure 4. Techniques employed to compute enthalpic relaxation due to the overshoot in the second heating. The enthalpic relaxation is computed from the areas bounded by the tangent lines, vertical line, and the heat flow traces. This is the heating curve of neat PVAc after being annealed for 240 min at 30 °C.

Download figure to PowerPoint

The area of the overshoot (or amount of hysteresis) is calculated by extending a tangent line from a point above the glass transition until it intersects the heating curve as shown in Figure 4. The intersection point is then vertically dropped to intersect a tangent line that started below the glass transition temperature thereby defining two areas between the heating curve and the dotted lines. Both areas are computed and added to obtain the total hysteresis value. In effect, this assumes that the glass transition should have followed a step function at a temperature where the vertical line intersects the tangent lines, as shown in Figure 4. This does not imply that the glass transition proceeds in a step, but rather expresses the fact that the heating rate is much faster than the rate at which the sample can equilibrate with the changing temperature. Thus, our analysis has some similarity with the computation of fictive temperature.9, 23, 24 Fictive temperatures are computed by comparing the areas above and under the heating curves and then finding the temperature at which these areas are equal. This procedure inadvertently includes the overshoot area. As a result, fictive temperatures are influenced by the overshoot areas and how rapidly the curves rise. Large overshoot areas result in lower fictive temperatures, likewise smaller overshoot result in higher fictive temperature. Any physical process or procedure that reduces the overshoot thus results in increasing the fictive temperature, and vice versa. In our case, the fictive temperatures plots versus either annealing time or temperature portray somewhat a vertical mirror image of the trends depicted in our results, with the vertical axes in temperature scale rather than energy per unit mass. We did not adopt this technique, because it would be difficult to account for the filler weight, which does not contribute to the overshoot but does contribute to the normalization with respect to weight. In addition, the magnitudes of the hysteresis values as related to annealing conditions and filler content are of major interest here, and not the exact manner in which one computes or defines the “value” of the glass transition temperature itself.

Figure 5 shows enthalpic overshoot for the neat polymer and nanocomposites prepared with 15 and 25 parts per hundred by weight of silica particles. These materials were annealed at 30 °C for different periods. The values presented in Figure 5 and all other figures have been corrected to account for the mass of the silica in the sample. Thus in Figure 5, the ordinate represents change in enthalpy per unit mass of polymer in the sample. Clearly, the enthalpic relaxations increase in all cases with respect to annealing time. This is not very surprising because the annealing temperature is below the glass transition temperature of the neat polymer. However, there are differences in the rate at which the increases occur. At shorter annealing periods, that is, annealing periods less than 30 min, there are no significant differences between the neat polymer and the composites. The physical ageing manifested in these enthalpic relaxations seems to be the response of the matrix only without alteration by the presence of filler. It is as if the fillers did not exist in these annealing periods. This seems in contradiction with the thesis of existence of glassy layers around the filler.5, 7, 8 If there were glassy layers around the filler, their effects on the ageing of the material should have been immediately felt. If one argues that such effects are beyond the detection limit of the instrument or the technique employed, then the same argument should hold at the intermediate and longer annealing periods. At these annealing periods, the effects of the fillers are to reduce the magnitude of the relaxation strength. It appears that the fillers have an effect at longer timescales. In fact, for the composites with 25 phr of silica the relaxation appears to plateau. Simon et al.19 showed that the fictive temperature of confined o-TP reaches a terminal value. Arguing that the terminal value has been achieved earlier in the confined o-TP faster than the bulk o-TP, they concluded that confinement induces physical ageing. On the other hand, it is possible that the confinement had reduced the mobility sufficiently to give an observed value that changes so slowly with time as to be mistaken for a terminal equilibrium, as suggested by Rittigstein and Torkelson.20 Thus, the apparent disagreement or controversy is rooted in what is considered to be an equilibrium or terminal value. There is no disagreement that the rate at which equilibrium is approached at longer annealing or ageing periods for both the composites or confined systems is slowed either by the presence of the filler or by confinement. The particle to particle distances in the cases presented here and computed using a simple cubic unit cell analysis with one filler particle per unit cell are 13.7 nm and 9.2 nm for the 15 phr and 25 phr nanocomposites, respectively. These are within the range of confinement studies presented for o-TP. However, unlike the confinement effect there is no change in the glass transition temperature in these cases whether they were annealed for short periods of time or for longer annealing periods.

thumbnail image

Figure 5. Enthalpic relaxation of poly(vinyl acetate) and poly(vinyl acetate)–silica nanocomposites annealed at 30 °C for different annealing periods. Diamond represents the relaxation of the neat poly(vinyl acetate), whereas square and triangle represent the relaxations of the nanocomposites with 15 phr and 25 phr silica, respectively. The purpose of the connecting lines is for presentation purposes only.

Download figure to PowerPoint

If the existence of glassy layers around the fillers were assumed then it follows that a distribution of glass transition temperatures would be present. According to this conjecture, annealing the materials at different temperatures should thus promote specific relaxations depending on the annealing temperature. Any structural or enthalpic relaxation is thus to manifest a skewed increase in its time evolution at the annealing temperature. Our analysis of the second heating curves, however, did not indicate any change in the glass transition temperature of either the neat material or the nanocomposites as a function of annealing temperature. What has changed as a function of the annealing temperature is the amount of hysteresis. Figure 6 shows the effect of annealing temperature on the enthalpic relaxation of the neat PVAc and its nanocomposites. There is no surprise that the enthalpic relaxation goes through a maximum with annealing temperature for a given annealing time period. The physical picture is similar to that used to explain the maximum in crystallization rate with degree of super cooling below the melting point for crystallizing materials. However, we are not claiming the glass transition is the same as crystallization. At annealing temperatures far above the glass transition temperature, there is no evidence of vitrification. Accordingly, the area of the overshoot is minimal. As the annealing temperatures are lowered, the vitrification is evidenced and likewise the increase in the overshoot of the area. This is in agreement with what Sasabe and Moynihan25 have showed, where the mean equilibrium relaxation times for the structural relaxation of PVAc above its glass transition increases from seconds to hours within a few degrees of temperature decrease. At an annealing temperature well below the glass transition temperature, the vitrification is inhibited by kinetics. There is insufficient mobility to allow whatever molecular motions are needed to evolve the hysteresis producing structure of the glassy state within the allowed 1-h timescale. These results are in agreement with the fictive temperature analysis of neat o-TP by Simon et al.19 According to their results presented in their Figure 3,19 for a given annealing period, fictive temperatures of o-TP annealed below the glass transition temperature increase with decreasing annealing temperature. As fictive temperature and the overshoot areas are inversely related, what we have observed in PVAc is similar to the observation in o-TP.

thumbnail image

Figure 6. Enthalpic relaxation of poly(vinyl acetate) and poly(vinyl acetate)–silica nanocomposites annealed for 1 h at different annealing temperatures. Diamond represents the relaxation of the neat poly(vinyl acetate), whereas square and triangle represent the relaxations of the nanocomposites with 15 phr and 25 phr silica, respectively. The purpose of the connecting lines is for presentation purposes only.

Download figure to PowerPoint

Interestingly though, at annealing temperatures greater or equal to the glass transition temperature, the materials exhibited minimal ageing compared with annealing temperatures below the glass transition temperature. This is contrary to what Struik presents as evidence for the existence of the glassy layer. Furthermore, had the presence of fillers promoted the formation of glassy layers then a higher filler concentration should increase the ageing of the material. In Figure 6, however, the opposite scenario is depicted with more filler producing less ageing. Clearly, the measurements presented here do not support the formation of a glassy layer or layers around the fillers. However, the presence of fillers appears to reduce the extent of ageing relative to that of the neat polymer.

The experimental results reported here are not consistent with the hypothesis that glassy layer formation is responsible for the abnormally high reinforcement values known to exist in nanofilled polymers and elastomers above the glass transition temperature. The reinforcement value is known to be obtained quickly above Tg and does not require any significant annealing time to develop. Also, the stress levels that are required to reduce the reinforcement value (so-called Payne effect) are far lower than expected based on the dependence of Tg on hydrostatic stresses. It has also been shown that the Payne effect is related to the dynamic, and not static, component of the strain history and that the Payne effect would vanish at very low strain rates.3 If a glassy layer is responsible for the high reinforcement then why does reinforcement vanish if observed at a slow strain rate? Additionally, a static strain has no effect on the observed Payne effect.3 Why would a glassy layer disappear when subjected to a static stress but not a dynamic stress? In summary, the glassy layer as a root cause of the reinforcement mechanism raises more questions than it answers.

However, there is little question but that fillers do result in a lowering of polymer segmental mobility at or near a filler interface. But this reduced mobility does not require that the near-surface layer is glassy. There are reinforcement mechanisms based on the concept that labile attachments of polymer chains to a filler surface govern the Payne effect.4 This labile attachment produces entropic hardening of the polymer chain by restricting the conformational entropy states that the chain may inhabit.4 Relatively few chain attachments to one or a few filler particles is quite adequate to result in a chain stiffness increase of several decades through the well-known Langevin chain statistics that properly considers the situations at high relative chain elongations for which Gaussian statistics are not valid. The entropic hardening concept is far more efficient in terms of filler contacts needed to produce high mechanical reinforcement values than the glassy layer concept for which there are many more filler contacts required, but as shown here evidently too few to observe. Finally, recent work has shown that detailed examination of the rate of reinforcement with filler concentration is not consistent with either agglomeration or percolation as an alternate mechanism for high reinforcement at low filler concentrations.26

CONCLUSIONS

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. EXPERIMENTAL
  5. RESULTS AND DISCUSSION
  6. CONCLUSIONS
  7. Acknowledgements
  8. REFERENCES AND NOTES

This work is based on the conjecture that physical ageing is a phenomenon exhibited by glassy polymers. Accordingly, if the origin of reinforcement in filled system is due to the formation of glassy layers around the fillers, physical ageing of the composite materials should have been promoted by the presence of the fillers. However, contrary to this expectation the ageing of the filled systems was observed to be suppressed in comparison with the neat polymer. Furthermore, there was no evidence of alteration of the glass transition temperature or broadening of the glass transition temperature due to the incorporation of the fillers. Physical ageing of the neat polymer obtained at different temperatures was replicated by the filled materials but with smaller magnitudes for the hysteresis even after the correction for the weight of the fillers. The effects of the fillers at short ageing periods seem minimal but may serve to diminish or retard the ageing significantly at longer ageing times. Thus, insofar as enthalpic hysteresis associated with annealing in the glassy state, the presence of filler is manifested primarily at long timescales and appears to retard the ageing process. This is consistent with findings from studies on the mechanical reinforcement of the same materials used in this study, among others, namely, the effect of filler concentration on the rate of reinforcement that indicates a far-field entropic-hardening effect that is related to the extent or strength of the filler–polymer interactions (the enthalpic surface interactions).4, 26 Thus, the importance of a bound polymer layer (perhaps of 2 mer units thickness at most) at the filler surface is of major importance, but not because it is or is not glassy. Additionally, it is also found that filler addition has a major effect on the relaxation time spectrum of the matrix polymer, with the longest relaxation times being dramatically enhanced or extended to longer times (lower frequencies).26 Finally, the abnormally high reinforcement is found to increase with temperature26 above Tg and this is inconsistent with the concept of a glassy layer as the primary reason for high reinforcement.

Acknowledgements

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. EXPERIMENTAL
  5. RESULTS AND DISCUSSION
  6. CONCLUSIONS
  7. Acknowledgements
  8. REFERENCES AND NOTES

The authors would like to acknowledge the Union College Faculty Research Fund for financial support and Cabot Corp. for supplying the fillers. We would like to thank C. T. Moynihan for his valuable insights and discussions.

REFERENCES AND NOTES

  1. Top of page
  2. Abstract
  3. INTRODUCTION
  4. EXPERIMENTAL
  5. RESULTS AND DISCUSSION
  6. CONCLUSIONS
  7. Acknowledgements
  8. REFERENCES AND NOTES