lable at ScienceDirect

Polymer 52 (2011) 2910e2919

Contents lists avai

Polymer

journal homepage: www.elsevier .com/locate/polymer

Microcellular and nanocellular solid-state polyetherimide (PEI) foams usingsub-critical carbon dioxide II. Tensile and impact properties

Dustin Miller*, Vipin KumarDepartment of Mechanical Engineering, University of Washington, Seattle, WA 98195, USA

a r t i c l e i n f o

Article history:Received 25 January 2011Received in revised form22 April 2011Accepted 26 April 2011Available online 4 May 2011

Keywords:MicrocellularNanocellularProperties

* Corresponding author. Tel.: þ1 206 543 7504; faxE-mail address: mille35@uw.edu (D. Miller).

0032-3861/$ e see front matter � 2011 Elsevier Ltd.doi:10.1016/j.polymer.2011.04.049

a b s t r a c t

Microcellular foams, closed-cell polymer foams with cells of order 10 mm, have been studied for over twodecades. These foams have shown significant improvements in mechanical properties, such as strength-to-weight ratio, over conventional foams with cells on the order of 1 mm. Will an additional 100-foldreduction in cell size yield further improvement in properties? Here we answer this question in the solid-state PEIeCO2 system, a unique gasepolymer system in which cellular structures can be createdthroughout the polymer volume at either micro or nano scales. The tensile and impact behaviors ofmicrocellular (cells in 2e5 mm range) and nanocellular (cells in the 50e100 nm range) structures areexperimentally compared for foams with relative densities, to that of the starting solid, in the range of0.75e0.90. We found that nanofoams show a significantly higher strain to failure, resulting in animprovement in the modulus of toughness by up to 350% compared to microcellular foams. Fallingweight impact tests show evidence of a brittle-to-ductile transition in nanofoams resulting in impactenergies that are up to 600% higher compared to microcellular foams. These results point to the promiseof nanofoams as an important new class of materials.

� 2011 Elsevier Ltd. All rights reserved.

1. Introduction

Microcellular foams refer to thermoplastic foams with cells ofthe order of 10 mm in size. Typically these foams are rigid, closed-cell structures. The idea to incorporate such small bubbles inthermoplastics goes back to early work at MIT [1,2] where theadvent of microcellular polystyrene was described using nitrogenas the blowing agent. The invention was in response to a challengeby food and photographic film industries to reduce the amount ofpolymer used for packaging. Thus was born the idea to createmicrocellular foam, where we could have, for example, 100 bubblesacross a 1 mm thickness, and expect to have a reasonable strengthfor the intended applications. The basic solid-state batch processused for polystyrene [1,3,4] has been used to create microcellularfoams from a number of amorphous and semi-crystalline polymers,such as polyvinyl chloride (PVC) [5], polycarbonate (PC) [6,7],acrylonitrileebutadieneestyrene (ABS) [8], polyethylene tere-phthalate (PET) [9], crystallizable polyethylene terephthalate(CPET) [10], and polylactic acid (PLA) [11], to name a few.

There are two basic steps in the solid-state foaming of thermo-plastic polymers. The first step consists of saturation of the polymer

: þ1 206 685 8047.

All rights reserved.

with gas under high pressure. This step is normally carried out atroom temperature. Given sufficient time for diffusion of gas into thepolymer, the gas attains an equilibrium concentration that isconsistent with the solubility of gas in the polymer and the gaspressure. In the second step, bubbles are nucleated in the gasepol-ymer system by creating a thermodynamic instability. This is ach-ieved by either a sudden drop in pressure [12] or sudden increase intemperature [5,6]. Both strategies suddenly reduce the solubility ofthe gas, driving the gas out of the polymermatrix and into nucleatedbubbles. One consequence of dissolving gas in the polymer is plas-ticization, reducing the polymer’s glass transition temperature [13].After saturation, the temperature of the gas-saturated polymer onlyneeds to be raised to the effective glass transition temperature of thegasepolymer system to nucleate bubbles. Hence the phrase “solid-state foam” is used to describe such foams, as opposed to theconventional foams produced from a polymer melt. Microcellularfoams can, however, also be produced in extrusion and injectionmolding during molten state [14e17].

Since the invention of microcellular foams many researchershave investigated the mechanical properties of these materials.Results have shown that unlike conventional foams, with typicallylarge variations inproperties,microcellular foams exhibit consistentand improved tensile properties in systems such as PC [18], PET [9],ABS [19], and PVC [20]. For instance, microcellular PC tensilestrength reduces proportionally with the relative density of the

D. Miller, V. Kumar / Polymer 52 (2011) 2910e2919 2911

foam [18,21], where relative density refers to the ratio of the densityof the foam relative to that of the solid. It has been hypothesized thatthis occurs due to the reductionof cell size aswell as the uniformandhomogeneous cell structure ofmicrocellular foams.However, due tothe large variation in mechanical behavior of conventional foamsand the difficulty in obtaining foams of a given density but differentcell sizes, determining the effect of cell size on the mechanicalproperties of polymer foams remains a challenge. Previous attemptsto study the effect of cell size on the tensile properties of micro-cellular foams determined that within the range of 3e37 mm, cellsize did not affect tensile strength [22].

The influence of foam morphology on impact strength ofmicrocellular foams has been studied in several polymer systems.Results for the influence of cell size on impact properties have beenmixed. In PC, impact toughness is improved within the micro-cellular range by increasing the average cell size [23]. Othersstudying cell size effects in PS [24] found no change in impactstrength by varying cell size within the microcellular range. Ingeneral, high relative density microcellular foams have been foundto provide increased impact resistance when compared to the neatpolymer. This increasewas found across a range of polymer systemssuch as PET [25], PC [7], and PVC [26].

Nanocellular foams, or nanofoams, refer to a recent extension ofmicrocellular technology where foamed polymers have pore sizesin the range of nanometers. The idea of creating nano-scale cells inpolymers is exciting but only a handful of polymer systems withuniform nano-scale morphologies have so far been discovered. Theprogression to nanocellular foams first started with ultra-microcellular or supermicrocellular foams with sub-micrometersized cells [27,28]. Nanocellular foams have been created in thesolid-state process primarily in polyimides [29e32], thoughnanofoam processing has also been extended to HDPE [33] andpolymer blends using a block copolymer method [34,35]. One ofthe limitations to mechanical property evaluation of nanofoams isthe production of sufficiently large samples that can be subjected tostandard test protocols.

It has been speculated that nanofoams will offer many uniqueproperties that are superior to existing materials, such as a higherstrength-to-weight ratio, improved toughness and significantlylower thermal conductivity. One popular hypothesis suggests that ifnanocells are smaller than the mean free path for atmosphericmolecular collisions, roughly 70 nm at standard temperature andpressure, then the cells will essentially have a vacuum inside them,reducing the thermal conductivity. However, thus far, studies onnanofoam thermal properties are not available.

The current processing techniques available to produce poly-meric nanofoams are also limited. Nanofoams have been producedfor microelectronic applications using a block copolymer method[29,30], where the copolymers consist of thermally stable andthermally labile blocks. Upon heating, the thermally labile blockundergoes thermolysis, leaving nanopores behind. Nanofoamswithhigh thermal stability and relatively low dielectric constant havebeenproduced thisway.However, void fractions achievedwere onlyin the 15e25% range. In a low-temperature process developed byHanda and Zhang [27] for PMMAeCO2 system, cells in 200e400 nmrangewere achieved. In a recent publication [36],wehave describedthe PEIeCO2 system where we can create either microcellular ornanocellular structures at like densities. This processing advance-ment enables the study of the effect of cell morphology on materialproperties over a range of cell sizes spanning several orders ofmagnitude.

Our interest in PEI foams was initially motivated by the need forhigh strength lightweight materials for the protection and insu-lation of electronic components at high service temperatures. Dueto its relatively high glass transition temperature, we chose to

investigate PEI foams. PEI meets the Federal Aviation Administra-tion requirements for flame, smoke and toxicity. We discoveredthat the PEIeCO2 system provides us, for the first time, the capa-bility to create cellular structures at a given density where the cellsize can be varied by a factor of 100 or more [36]. This gives us theopportunity to investigate the effect of varying cell size over themicro-to-nano range on mechanical properties, the subject of thisstudy.

2. Experimental

2.1. Materials

We used 1.5 mm thick sheets of Ultem 1000 PEI, made by SABIC.It is an amorphous thermoplastic with a measured density of1.28 g/cm3 and a glass transition temperature of 216.4 �C. The rawmaterial was stored at room temperature in the original shippingpackaging until sample specimens were cut for testing.

2.2. Foaming procedure

The PEI specimens were foamed by the solid-state batchfoaming process. From previous processing investigations [36] itwas determined that a foam density range between 75 and 90%relative density would provide testing samples where structurescould be produced with either microcellular or nanocellular cells atlike densities. This density range was chosen to reduce variationfrom sample to sample and densities could be produced accuratelywithin a �1% relative density variation. High relative densitysamples also offer the smallest cell sizes in the nanocellular range.Three density values were chosen, 75, 82.5 and 90% relative density,to evenly span the range of desired densities.

Microcellular structures were created by saturating the 1.5 mmthick specimens at 1 MPa for 500 h, and foaming in the 165e185 �Crange. Nanocellular structures were created by saturation the PEIspecimens at 5MPa for 300 h, and foaming in the 110e145 �C range.The foaming time was 2.5 min. Details of the processing conditionscan be found in our previous work [36].

2.3. Tensile testing

Tensile specimens were prepared according to ASTM D638“Standard Test Method for Tensile Properties of Plastics” and TypeIV sample specifications. A milling operation cut samples into TypeIV dimensions. After the foaming procedure was performed toachieve the desired density reduction, tensile testing samples werethen desorbed of the remaining CO2 for a minimum of 800 h beforetesting. Tensile testing was performed on an Instron 5585H.A constant crosshead rate of 10 mm/min was used while testingsamples. Ten samples were tested at each processing condition andresults are given as averages with standard deviations.

2.4. Impact testing

Impact samples were created according to ASTM D5420 “Stan-dard Test Method for Impact Resistance (Gardner Impact)”. Samplesizes of 50 mm square were produced for impact testing. Gardnerimpact data applies to awide variety of polymer applications due tothe impact mechanism of the test (falling dart). The test dataproduced by this method is useful in comparing impact resistanceof materials on a relative basis, but does not allow for correlationwith data obtained by other test protocols. A total of 26 sampleswere tested at each processing condition. Six are used to roughlydetermine the mean failure height, while the other 20 samples areused in the final test according to the ASTM standard. Results are

D. Miller, V. Kumar / Polymer 52 (2011) 2910e29192912

given as mean fracture toughness per unit thickness. Fracturetoughness per unit thickness was calculated to account for theincrease in sample thickness due to foaming and representsa standardized reporting technique within the ASTM standard.

2.5. Differential scanning calorimetry

Differential scanning calorimetry (DSC) measurements wereperformed on a TA Instruments Q20 setup. The DSC procedurefollowed the ASTM D7426-08 technique using a heat rate of 10 �C/min and the glass transition temperature (Tg) was determined froma half-height method. A minimum of 3 samples was run at eachfoaming condition in order to validate results. Foamed sampleswere tested after totally desorbing samples of all remaining CO2 gasresulting from foaming. The mean value of Tg is reported with errorbars representing �2 standard deviations.

3. Results and discussion

3.1. Microstructure

Scanning electron microscope images of the various micro-cellular and nanocellular structures representing the range testedin this study are shown in Fig. 1. Microcellular average cell sizes

Fig. 1. Microcellular PEI structures created at 1 MPa saturation pressure. (A) 91.5% relative d56.0% relative density and avg. cell size ¼ 4 mm. Nanocellular PEI structures created at 5 Mrelative density and avg. cell size ¼ 50 nm, (F) 47.6% relative density and avg. cell size ¼ 120 nnanocellular foams.

range from 2 to 4 mm between the densities produced. Micro-cellular sample A, with a relative density of 91.5%, exhibits a struc-ture where individual cells are typically distributed with clear areasof solid polymer between each cell. In contrast, microcellularsamples B and C show cells that are all directly adjacent to oneanother and are characterized as sharing cellular walls. Celldensities for microcellular samples A, B and C are on the order of1010 cells/cm3.

Nanocellular samples shown in Fig. 1(D, E, F) have a range ofaverage cell sizes between 30 and 120 nm. Cell densities in nano-cellular samples are of order 1014 cells/cm3. At higher densities,nanocellular morphologies appear to be similar to that of micro-cellular samples with cells uniformly distributed throughout thevolume. As the density is reduced, cells begin to appear adjacent toothers often sharing cellular walls, although micrographs in Fig. 1still show significant areas of solid polymer. Many solid regionsare of the order of the average cell size or larger, as in Fig. 1E.

3.2. Tensile results

Fig. 2 presents the stressestrain data obtained in our tensileexperiments. The key tensile parameters, namely the averagevalues of elastic modulus, stress and strain at break, and themodulus of toughness are summarized in Table 1. Also given in

ensity and avg. cell size ¼ 2 mm, (B) 70.2% relative density and avg. cell size ¼ 3 mm, (C)Pa saturation pressure. (D) 69.9% relative density and avg. cell size ¼ 30 nm, (E) 59.8%m. Cell densities range from 3.8 to 12.9 � 1010 for microcellular and 1.3 to 2.0 � 1014 for

Fig. 2. (A) Unprocessed PEI averaged tensile stressestrain curves. (B), (C), and (D) show the stressestrain plots for microcellular and nanocellular foams of nominal relative densityof 90%, 82.5%, and 75%, respectively. The open symbols represent the strain at which an individual sample fractured.

D. Miller, V. Kumar / Polymer 52 (2011) 2910e2919 2913

Table 1 are standard deviations for the stress and strain at break,which give us a measure of the variability in these important foamproperties. Finally, the data in Table 1 is plotted in Fig. 3 to bring outthe trends and the important differences as a result of changing thecell size from micro to nano.

First, some overall observations. The macroscopic foam prop-erties are primarily determined by the density, or void fraction, ofthe foam. Thus, the tensile stress and strain at break, as well as themoduli of elasticity and toughness, all decrease as foam densitydecreases; see Fig. 3. Models for this influence of foam density onfoam properties are well established [37] and typically referred toas the “Rule of Mixtures”.

We now focus on the differences we observe, at a given foamdensity, when the cell size changes from micro-to-nano scale.Fig. 2A shows the stressestrain behavior of solid, unprocessed PEIas a reference. The open symbol (>) shows the strain at which thespecimen broke. We note that the strain to break for the 10 spec-imens tested ranges from 1.0 to 1.35, with a mean of 1.1, and

Table 1Summary of tensile test measurement data.

Yield (MPa) Mean stress atbreak (MPa)

Standard deviatiostress at break

Unprocessed 108.3 99.6 5.490% Relative density Micro 87.7 78.1 6.7

Nano 82.21 93.26 3.6282.5% Relative density Micro 78.1 73.1 2.99

Nano 71.5 84.7 4.9275% Relative density Micro 68.3 66.7 3.32

Nano 64.1 71.5 2.95

a standard deviation of 0.11. Table 1 and Fig. 2BeD show the resultsof tensile tests for micro- and nano-foams at 90%, 82.5%, and 75%relative density. Perhaps the most striking feature we see in thisdata is the impact of cell size on strain to failure, range, mean, anddistribution. At 90% relative density, for example, the 10 micro-cellular specimens failed at strains ranging from 0.20 to 1.08, witha mean strain of 0.71. The nanocellular specimens, on the otherhand, failed over a range of 1.12e1.54, with a mean strain of 1.34.The corresponding standard deviations are 0.29 for micro foamsand 0.13 for nanofoams. This shows that nanofoam ductility, asrepresented by average strain at failure, has nearly doubledcompared to micro foams, while at the same time the variability inmeasurements has reduced by a factor of 2 as well.

Microcellular PEI foams get increasingly brittle as voids growand the foam density reduces. Referring to Table 1, the strain atbreak reduces from 0.71 at 90% relative density, to 0.36 at 82.5%relative density, to 0.29 at a relative density of 75%. Thus, ata relative density of 75%, the average strain to break has dropped

n Mean strain at break(mm/mm)

Standard deviationstrain at break

Modulus ofelasticity (MPa)

Toughness(J/cm3)

1.1 0.11 3454 98.70.71 0.29 2597 51.11.34 0.13 2628 103.10.36 0.28 2265 24.71.23 0.18 2226 86.50.29 0.23 1978 17.61.05 0.12 1929 65.1

Fig. 3. Results from tensile tests at different foam densities. (A) strain at break, (B) stress at break, (C) modulus of elasticity, (D) modulus of toughness.

D. Miller, V. Kumar / Polymer 52 (2011) 2910e29192914

from 1.1 for the solid PEI to 0.29; a reduction of 74%. By contrast, thenanofoams maintain the ductility at a level that is comparable tosolid PEI, as the data in Table 1 shows.

Directly resulting fromthe increased strain at break, themodulusof toughness of nanocellular PEI is significantly improved overmicrocellular PEI. The area under the stressestrain curve, whichresults in the amount of energy required to fracture the sample,calculates toughness. The largest toughness difference betweennanocellular PEI and microcellular is seen at 75% relative density,where the nanocellular material was 3.8 times tougher than themicrocellular material. Fig. 3 compares the material propertiesmeasured in the tensile tests.

In a previous study, Kumar andWeller [22] investigated the effectof cell size on tensile properties of microcellular PC foams. Theyfound that over a cell size range of 2.8e37 mm, there was no signifi-cant change in tensile behavior. However, as our results show in PEI,the ductility (strain to break) increased by 150%, and the modulus oftoughness increased by nearly 300%, when, at a relative density of75%, the cell size reduced from approximately 3 mme30 nm, a factorof 100.

When we compare these tensile results for microcellular andnanocellular foams in Fig. 3 we see that cell size plays an importantrole in the tensile properties of these foams. In part (A) of the figurewe clearly see the differences between microcellular and nano-cellular with that of unprocessed PEI in the strain at break. Theamount of strain at break in the nanocellular foams remains rela-tively unchanged throughout the density range. Microcellularfoams, shown in the same plot, display a decrease and greatervariation in the strain at break. In part (B) the tensile stress at breakis detailed. Due to the increased strain at break for nanocellularfoams, these materials undergo increased strain hardening andthus have greater values of stress at break. In part (D) of the figurethe tensile toughness of the materials studied are shown. Also,primarily resulting from the increased strain at break in the

nanocellular foams, the resulting tensile toughness is greatlyimproved over that of the microcellular foams. In Fig. 3 we see thatall the tensile material properties studied decrease with decreasingdensity for both microcellular and nanocellular foams, except forstrain at break for nanocellular foams.

3.3. Impact results

The PEI impact testing data contained in this paper presentsthree densities where foam samples had either a microcellular ornanocellular structure. Foam processing conditions repeated thoseused to create tensile specimens. The impact data collected is usedto decouple the influence of cell size from sample density. Theseresults give a greater understanding of how these two variables caninfluence the impact behavior of PEI foams and provide a furtherunderstanding of the properties of nano-structuredmaterials. Fig. 4plots the impact energies in solid and cellular PEI specimens asa function of density. Impact energy is displayed in the units ofJoules and Joules per millimeter to account for thickness variationsbetween the various sample densities.

Unprocessed PEI samples were first tested to benchmark thebehavior of PEI. The mean energy for breakage was calculatedaccording to ASTM D5420 and for solid PEI it was found to be12.45 J/mm, see Fig. 4 and Table 2 for tabulated values of impactenergy. Microcellular and nanocellular foamed samples withdensities of 75, 82.5, and 90% relative densities are also shown inTable 2. In Fig. 4 we see that there are significant differencesbetween unprocessed and foamed samples as well as variationsbetween the two cellular structures. The impact strength of 90%relative density microcellular PEI is 3.16 J/mm. A 90% relativedensity foam represents a material that contains gas bubbles in 10%of the samples total volume. Microcellular foams with only a 10%reduction in density exhibit a 75% reduction in impact strengthcompared to the unprocessed material. Further density reduction

Fig. 4. Impact strengths of unprocessed, microcellular and nanocellular samples asa function of relative density in Gardner impact. Nanocellular foams have higherimpact strengths than the unprocessed PEI across the entire density range wheremicrocellular foams have lower impact strengths than the solid.

D. Miller, V. Kumar / Polymer 52 (2011) 2910e2919 2915

with microcellular voids shows an increase in strength to a point atwhich the 75% relative density microcellular PEI impact strength,8.79 J/mm, is slightly below that of the unprocessed PEI.

Impact data for nanocellular PEI shows the opposite trend thanthat observed in microcellular PEI. Nanocellular PEI showsa decreasing impact strength trend as a function of densityreduction. The nanocellular 90% relative density samples had animpact energy of 20.54 J/mm, which is a 65% improvement inimpact strength over the unprocessed material. Comparing bothcellular structures at 90% relative density, nanocellular PEI exhibitsa 600% increase from that of the microcellular material. As nano-cellular foam density is further reduced to 82.5% and 75% relativedensity, the impact strength reduced to 16.35 J/mm and 13.70 J/mmrespectively. At 75% relative density, nanocellular PEI still showsa 10% improvement over the unprocessed material and a 55%increase over that of the microcellular material at the same density.

All nanocellular PEI samples across the density range hadimpact strengths greater than that of unprocessed PEI. The trend forboth cellular structures extrapolates a convergence at 70% relativedensity to the impact strength observed for the starting unpro-cessed material. This is significant result in that PEI can achievea 75% reduction in density and still provide the impact strengthcomparable to that of the starting unprocessed material. Similartrends of decreasing impact strength as a function of reduceddensity have been shown in PC, PVC and CPET microcellular foams[7,23,25,26,38]. Previous research conducted by Barlow et al.

Table 2Impact strength of unprocessed, microcellular and nanocellular samples in Gardnerimpact.

Relative density Impactenergy (J)

Average samplethickness (mm)

Impact energy(J/mm)

Unprocessed PEI1 20.79 1.67 12.45

Microcellular0.75 17.40 1.98 8.790.825 10.84 1.95 5.560.9 5.81 1.84 3.16

Nanocellular0.75 25.20 1.84 13.700.825 30.73 1.88 16.350.9 36.15 1.76 20.54

[23,25] on the PCeCO2 system shows that there is a strong need todecouple the contributions of cell size and relative density on theimpact behavior of solid-state microcellular foams. In 1994, a studyby Collias et al. [7] conducted Charpy impact tests on polycarbonatesamples with cell sizes ranging from 5 to 45 mm and 97 to 72%relative density. This study found that impact strengths wereimproved over unprocessed PC and it was hypothesized that theimprovement in impact strength came from the cellular structurechanging the mode of fracture from brittle to ductile.

Barlow et al. also investigated extending the bubble size range inthe PCeCO2 system by controlling the solid-state processing vari-ables resulting in foams of 71% relative density with four differentcell sizes ranging from 7.56 to 18.05 mm. By keeping the relativedensity somewhat constant, Barlow was able to study the effect ofcell size on impact properties while removing the contributions ofdensity. Impact strengths were relatively constant throughout thecell size range except for the sample with an average cell size of18.05 mm which had a 50% increase in impact strength. The studywas inconclusive in determining the contribution of cell size onimpact properties, but did demonstrate a clear relationshipbetween relative density and impact strength.

In Fig. 5 we visually compare the failure mechanisms ofunprocessed PEI (A) to that of microcellular and nanocellularsamples of 82.5% relative density, (B) and (C) respectively. Unpro-cessed PEI exhibits little ductility, and upon failure, breaks witha brittle fracture. For the unprocessed samples, the falling dartstriker would cause failure around the circumference of theimpacted region and cause the entire region to shear from the restof the sample. Microcellular foamed samples (B) would break ina brittle fashion, much like that of the starting unprocessed mate-rial, though cracks initiated in the impact regionwould continue topropagate outside this region traveling through the material untilreaching the edge of the sample. Fracturing was much more brittleand violent for microcellular foamed samples compared to theunprocessed material. The microcellular sample (B), of 82.5% rela-tive density, shown in the figure, had less than 50% the impactresistance as compared to unprocessed PEI.

In contrast, nanocellular foamed samples (C) of all densitiestested had impact energies greater than that of the microcellularand standard unprocessed PEI. Nanocellular samples tested belowthe reported failure energy show a transition from the brittle-to-ductile stretching. Fractured nanocellular samples also exhibiteda ductile failure, showing evidence of a brittle-to-ductile transition.Upon fracture, cracks initiated in the impacted region propagatetoward the edge of the impacted region and arrest at the edge ofthis transitional region. Nanocellular ductility was observed athigher impact energies than that of the solid, and because thefalling weight mass was fixed, the velocity of the striker was greateras well. These failure mechanisms were characteristic throughoutthe various nanocellular density conditions studied, although theactual impact energies changed due to foam density. The intro-duction of nanocellular voids in PEI provides for increased ductilityat high impact strain rates, resulting in increased impact resistanceover the unprocessed material.

Comparing the results of impact with that found in the tensiletests we find that the introduction of nanocellular voids in PEIprovides both advantages and disadvantages in material propertiescompared to unprocessed PEI. For the exception of a slight reduc-tion in tensile yield strength, nanocellular structures provide foreither comparable or increased material properties in both tensileand impact compared to microcellular samples. This data showsthat nanocellular PEI foams have superior properties to micro-cellular PEI foams. The most significant improvement in nano-cellular PEI foams is the increased ductility under tensile andimpact loads when compared to microcellular foams. The rate of

Fig. 5. Images of impacted samples showing mode of failure in increased ductility. Sample (A) ¼ unprocessed PEI, sample (B) ¼ microcellular 82.5% relative density, sample(C) ¼ nanocellular 82.5% relative density.

D. Miller, V. Kumar / Polymer 52 (2011) 2910e29192916

strain in tensile testing, 10 mm/min, and high strain rates in impactplayed a role in the nanocellular foam ductility when comparing tothe unprocessed material. For lower strain rates in tensile tests,nanocellular ductility, or total amount of strain to break, wascomparable to that of the unprocessed material. Impact testing,where strain rates are much greater than tensile, nanocellular foamductility was improved compared to the unprocessed material andresulted in improved impact properties across the entire densityrange studied.

3.4. Discussion of material properties

Tensile and impact testing reveals that nanocellular foamsexhibit increased strain to failure compared to microcellularsamples. Unique to impact, where strain rates are much greaterthan that in tensile tests, nanocellular samples were found to havegreater strains than both microcellular and unprocessed PEI.Nanocellular samples in both tests were observed to have greaterelongation as can be seen from the tensile data as well as impactsamples shown in Fig. 5. The elongation observed highlights themolecular mobility of the solid before crack propagation and frac-ture. The introduction of voids in a polymer substrate will alter themacroscopic properties by the physical ordering of these voids.Such an effect is highlighted well in the tensile strength of polymerfoams where the area fraction of voids in the cross section limit theengineering stress of the material and can be estimated throughthe classical approach of a two-phase material or more commonly,the “Rule of Mixtures”. The mixture being that of both solid poly-mer in the cell walls and air within the cell. However, the intro-duction of voids in a polymer greatly increases the free surface.Voids structure the solid into a network of cell walls. Depending onthe cellular morphology, cell wall thickness can vary with averagecell size and density. Cell wall thickness typically decreases as cellsize and density are reduced as can be seen by the micrographs inFig. 1. Molecular mobility is of particular interest when length scalesare below w200 nm in dimension, where it has been found tochange such properties as Tg [39,40]. Thus, the reduction of lengthscale in these discrete regions of solid polymer has the effect ofaltering the fundamental material properties of the polymer itself.

Studying the micrographs of microcellular and nanocellularfoams and the fundamental differences between these two struc-tures highlights the importance of the discussion on molecular

mobility. In Fig. 1A, the introduction of microcellular sized voids inthe polymer matrix reduces the thickness of the polymer regions,i.e. cell wall, to approximately 2e5 mm. As the density is reduced bythe addition of more cells, the cell wall thickness reduces until twoneighboring cells share a single cell wall. The cell wall thickness isreduced at lower densities to roughly 1 mm. In contrast, nano-cellular voids for high relative density foams, like that shown inFig. 1D, have cell walls on the order of 200e300 nm in thickness.Nanofoam density reduction reduces the cell wall thickness to100 nm on average. Cell wall thicknesses differ by one order ofmagnitude between microcellular and nanocellular structures.Further, 100 nm cell wall thickness in a nanocellular foam are wellwithin the w200 nm range where molecular motilities have beenshown to be affected in thin films.

In order to characterize the effect of cell size and cell density onthe material properties of these foams, DSC measurements weretaken on a range of foam densities of both the microcellular andnanocellular foams. Local changes of molecular mobility in the cellwall can be measured through bulk DSC measurements due to theuniform cellular structure in which the cell walls constitute thebulk morphology of the foam.

Glass transition measurements taken on the unprocessed PEImaterial and the two cellular structures across awide density rangeare plotted in Fig. 6 as a function of density. First, we will describethe overall observations from this data and then apply thoseobservations to the cellular structures and densities tested intensile and impact. From Fig. 6 we see that the Tg of microcellularPEI foams remains fairly flat over the density range and equal to theunprocessed material. Microcellular samples tested with relativedensities ranging from 0.65 to 0.9 have mean values greater thanthat of the solid. Accounting for �2 standard deviations, only themicrocellular sample with a density of 0.82 came out statisticallydifferent from the solid.

Tg values collected for nanocellular foams diverge from themicrocellular trend by showing an increase in temperature asdensity is reduced. The increase in Tg becomes statistically signifi-cant for density reductions below 0.82. The lowest density nano-cellular sample had an average Tg slightly above 221 �C, or roughly5 �C above the unprocessed material. Relating the cellular wallthicknesses created in the nanocellular foams, Fig. 1DeF, to the Tgdata in Fig. 6, we see that as the cell wall thickness reduces fromroughly 300 nm to within the range of 100e200 nm for the

Fig. 6. Glass transition temperature of unprocessed PEI, microcellular and nanocellularfoamed samples. The plot shows a wider density range than that of the tensile andimpact samples range in order to determine trends. Points represent the mean value of3 samples. Error bars show �2 standard deviations from the mean.

D. Miller, V. Kumar / Polymer 52 (2011) 2910e2919 2917

densities shown, the Tg of the foam raises by 5 �C. An increase of thismagnitude (5 �C) over the length scales between 100 and 200 nmwas also found in spin-cast polystyrene thin films [39]. Over thedensity range tested for tensile and impact, 0.75e0.9, nanocellularfoams have an increase in Tg of 3 �C at most, as compared tomicrocellular foams with a negligible increase.

The Tg data for nanocellular foams shows an increase intemperature over the density range tested for tensile and impact. Anincrease in Tg indicates a decrease in the molecular mobility of themolecules within the foamed samples. This effect appears small, butnotable. It does not seem that the introduction of nano-sized voids

Fig. 7. Schematic of bubble stretching stages during tensile testing with corresponding micro(B) structures appear spherical before beginning loading. Stage #3 shows the moment at fra(C) and microcellular (D). Note the appearance of the nanocellular material in image C wh

into the polymer matrix can explain the increased elasticity, ormolecular mobility, of the nanocellular tensile and impact samples.In fact, the data suggests the opposite. We can then conclude thatthe presence of nanocellular voids in the polymer do not contributeto the increased mobility seen in tensile and impact on the strictbasis of altering the fundamental base properties of the polymer.

Tensile and impact properties of microcellular and nanocellularPEI foams are likely best explained by a discussion on the fracturemechanics of polymer foams. Investigations into fracturemechanics of microcellular foams for tensile and impact considerthe presence of voids as stress risers. However, during crackpropagation, cellular material ahead of the crack tip will yield andthus relieve the triaxial stress state ahead of the crack tip[7,21,23,24]. These studies call attention to the reduction of stressconcentration when cell sizes are reduced, but all found noconclusive relationship between toughness and cell size for thesmall cell ranges studied. The PEI system has the unique ability toreduce cell sizes into the nano range, and thus the stress concen-tration reduction would be greatly enhanced. There are othercharacteristics of nanocellular foams that can further contribute tothe fracture mechanics in this system. These potential advantagesinclude cell sizes on the order of individual molecules and densi-fication of nanocellular voids during testing.

Nanocellular foams with bubble sizes ranging between 50 and100 nm have cellular morphologies on the same order of magni-tude as the length of PEI molecules, typically referred to as contourlength. Due to the similar length scales of molecules and voids,a single molecule could stretch from one side of the void to theother around the circumference. This would allow for macroscopicloads to span across the circumference of a single cell, loading theconnection of primary bonds. A microcellular void requiresthe connection of multiple molecules to span the distance aroundthe circumference of the cell, thus relying primarily on secondarybonding forces and entanglement for loading. It remains unclear

graphs of starting and fractured samples. In stage #1 nanocellular (A) and microcellularcture and the resultant micrographs of the elongated cellular structure for nanocellularere no cells are visible.

D. Miller, V. Kumar / Polymer 52 (2011) 2910e29192918

what the contribution of this factor has on the measured tensileand impact properties presented here.

Perhaps the most evident mechanism for the strength proper-ties of microcellular and nanocellular foams is the deformation ofthe cellular matrix under loading. A schematic of cellular defor-mation with corresponding micrographs of tensile samples isshown in Fig. 7. As tensile dog-bones are loaded, microcellular andnanocellular bubbles are stretched, yielding then collapsing thecellular structure. Failure occurs form the propagation of a crackthrough the sample causing fracture. As can be seen from Fig. 7, thetwo cellular structures differ in the amount of densification fora given strain. Just as reducing overall cell size has the effect ofreducing stress concentrations, the densification of cellular struc-tures, would result in the reduction of stress concentrations in thematerial. This is particularly the case for nanocellular structures,where a complete collapse can been seen. Microcellular structure atthe point of fracture can still be clearly seen in the tensile speci-mens. As the cell size is reduced in nanocellular foams to the pointshown in the schematic for stage #3, opposing cell walls begin totouch. A nanocellular material’s unique ability to undergo plasticdeformation to a point of cellular collapse would significantlycontribute to a reduction of stress concentrations. Tensile andimpact tests would both benefit from cellular collapse, thus givingthe material greater ability to strain. After full collapse additionalmolecular movement would allow for molecules to redevelopintermolecular bonds and perhaps reintroduce entanglementbetween molecules on opposite cell walls. The rate dependence ofdensification, redevelopment of intermolecular bonding, andentanglement remains an area for future investigation.

4. Conclusions

We have reported the effect of microcellular and nanocellularmorphologies on tensile and impact properties of polyetherimide(PEI) foams. Tensile and impact samples were prepared using thesolid-state batch foaming process using sub-critical CO2 asa blowing agent. A range of foam morphologies were testedincluding both microcellular structures with cell sizes around 5 mmand nanocellular structures with cell sizes on the order of 40 nm.Three relative densities were chosen to determine the effect of cellsize over a density range of 75e90% relative density. Based on thetwo cellular structures and three relative densities, diagrams for allconditions were presented showing the effect of cell size anddensity on material properties.

Tensile results reveal that yield strength is primarily a functionof part density and is not influenced by cell size. The simple “Rule ofMixtures” can accurately estimate yield strength. PEI exhibitsa reduction in tensile yield strength for higher saturation pressures,like that also found in other polymermaterials. Cell size did also notaffect the modulus of elasticity in tension. Strain to failure and thustensile toughness was the primary difference between foamingsamples and was increased by a factor of 2e3 by reducing cellmorphologies from the micro- to the nano-scale. It is thought thatthe smaller size of the nanocellular structure enables increasedmolecular mobility, thus allowing chain sliding under tensile stress.This might be due to the similar scale of both nanocellular voidsand individual polymer molecules. As density was reduced,microcellular samples displayed increased variability in the strainat break. However, the strain at break for nanocellular samplesremained unchanged from that of the unprocessed material acrossthe entire density range tested, even though the ultimate stress wasreduced.

Impact properties for foamed samples varied widely betweenthe two cellular structures. Microcellular sample impact strengthwas immediately reduced upon foaming at 90% relative density but

had a trend of increasing impact resistance as density was reducedfurther. Only PS has shown this same trend while other polymersystems usually show the opposite of a reduction in impact strengthas density is reduced. Nanocellular samples showed the exactopposite trend from that of the microcellular samples. Nanocellularsamples exhibited an initial increase in impact strength that linearlydecreased as density was reduced. The two impact strength trendsfor microcellular and nanocellular morphologies eventuallyconverged at 75% relative density also corresponding to the strengthof the original unprocessed material.

Further investigation into the material properties of the twocellular structures reveals that the introduction of nanocellularvoids reduces the cellular wall dimension to between 100 and200 nmwhich in turn increases the glass transition of the polymermatrix by 5 �C. This increase in the glass transition temperatureshows a reduction in the molecular mobility of the polymer matrix,thus does not explain the resulting strength properties in tensileand impact. Further investigation into the fracture mechanisms ofthe polymer samples shows that nanocellular structures reduce thestress concentrations within the material during testing by densi-fying the foam to a point at which the cellular structure are nolonger visible. This densification reduces the stress concentrationsof the voids and allows the nanocellular foam to reach strainequivalent to that of the starting material.

This study reveals, for thefirst time, a direct comparisonbetweenthe material properties of microcellular and nanocellular foams intensile and impact. The results of this research also have wideapplicability in the field of nanomaterials by offering experimentalresults of how material properties change as morphologies arereduced from the micrometer to nanometer length scales. Since theinvention of microcellular polymers at MIT, many have wonderedwhat role cell size plays in the material properties of these foams.The results found in this work clearly show that nanocellular foamsoffer superior properties compared to microcellular foams. Theseresults are encouraging and should motivate the next decade ofnanocellular foams research to continually explore how propertieschange in nano-structured polymers. In order to fully answer thisquestion, additional research will be required in extending thedensity range in the solid-state PEI process, as well as incorporatingdata from additional nanocellular polymer systems.

References

[1] Martini J, Suh NP, Waldman FA. Microcellular closed cell foams and theirmethod of manufacture. USA: Massachusetts Institute of Technology; 1984.

[2] Martini J, Waldman FA, Suh NP. The production and analysis of microcellularthermoplastic foam, in SPE ANTEC; 1982. San Francisco, CA. p. 674.

[3] Otsuka T, Taki K, Ohshima M. Nanocellular foams of PS/PMMA polymerblends. Macromol Mater Eng 2008;293(1):78e82.

[4] Taki K, Waratani Y, Ohshima M. Preparation of nanowells on a PS-PMMAcopolymer thin film by CO2 treatment. Macromol Mater Eng 2008;293(7):589e97.

[5] Kumar V, Weller JE. A process to produce microcellular PVC. Int Polym Proc1993;VIII(I):73e80.

[6] Kumar V, Weller JE. Production of microcellular polycarbonate using carbondioxide for bubble nucleation. J Eng Ind 1994;116:413e20.

[7] Collias DI, Baird DG, Borggreve RJM. Impact toughening of polycarbonate bymicrocellular foaming. Polymer 1994;35:3978e83.

[8] Nawaby V, Handa P. Fundamental understanding of the ABS-CO2 interactions,its retrograde behavior and development of nanocellular structures, in ANTEC;2004. Chicago, IL. pp. 2532e2536.

[9] Shimbo M, Higashitani I, Miyano Y. Mechanism of strength improvement offoamed plastics having fine cell. J Cell Plast 2007;43:157e67.

[10] Handa YP, Wong B, Zhang Z, Kumar V, Eddy S, Khemani K. Some thermody-namic and kinetic properties of the system PETG-CO2, and morphologicalcharacteristics of the CO2-blown PETG foams. Polym Eng Sci 1999;39(1):55e9.

[11] Richards E, Rizvi R, Chow A, Naguib H. Biodegradable composite foams of PLAand PHBV using subcritical CO2. J Polym Environ 2008;16:258e66.

[12] Goel SK, Beckman EJ. Generation of microcellular polymers using supercriticalCO2. Cell Polym 1993;12(4):251.

D. Miller, V. Kumar / Polymer 52 (2011) 2910e2919 2919

[13] Chow TS. Molecular interpretation of the glass transition temperature ofpolymer-diluent systems. Macromol 1980;13:362e4.

[14] Baldwin DF, Park CB, Suh NP. An extrusion system for the processing ofmicrocellular polymer sheets: shaping and cell growth control. Polym Eng Sci1996;36(10):1425e35.

[15] Han X, Koelling KW, Tomasko DL, Lee LJ. Continuous microcellular poly-styrene foam extrusion with supercritical CO2. Polym Eng Sci 2002;42(11):2094e106.

[16] Park CB, Behravesh AH, Venter RD. Low density microcellular foam processingin extrusion using CO2. Polym Eng Sci 1998;38(11):1812e23.

[17] Shimbo M, Nishida K, Nishikawa S, Sueda T, Eriguti M. Foams extrusiontechnology of microcellular plastics. Porous Cellular and Microcellular Mate-rials: ASME 1998;82:93e8.

[18] Kumar V, VanderWel M, Weller J, Seeler K. Experimental characterization ofthe tensile behavior of microcellular polycarbonate foams. J Eng MaterTechnol 1994;116:439e45.

[19] Nadella K, Kumar V. Tensile and flexural properties of solid-state microcellularABS panels. In: Experimental analysis of nano and engineering materials andstructures. Netherlands: Springer; 2007. p. 765e6.

[20] Kumar V, Weller J, Ma M, Montecillo R, Kwapisz R. The effect of additives onmicrocellularPVC foams:part II. Tensilebehavior. Cell Polym1998;17(5):350e61.

[21] Bureau MN, Kumar V. Fracture toughness of high density polycarbonatemicrocellular foams. J Cell Plast 2006;42(3):229e40.

[22] Kumar V, Weller JE. The effect of cell size on the tensile behavior of micro-cellular polycarbonate. Cell Microcell Mater ASME 1996;76:17e26.

[23] Barlow C, Kumar V, Flinn B, Bordia R, Weller J. Impact strength of high densitysolid-state microcelluar polycarbonate foams. J Eng Mater Technol 2001;123:229e33.

[24] Doroudiani S, Kortschot MT. Polystyrene foams. II. structureeimpact proper-ties relationships. J Appl Polym Sci 2003;90(5):1421e6.

[25] Kumar V, Juntunen RP, Barlow C. Impact strength of high relative density solidstate carbon dioxide blown crystallizable poly(ethylene terephthalate)microcellular foams. Cell Polym 2000;19(1):25e37.

[26] Juntunen RP, Kumar V, Weller J, Bezubic W. Impact strength of high densitymicrocellular poly(vinyl chloride) foams. J Vinyl Add Techy 2000;6(2):93e9.

[27] Handa YP, Zhang Z. Manufacturing ultramicrocellular polymer foams at lowpressure; 1999. USA.

[28] Baldwin DF, Suh NP, Park CB, Cha SW. Supermicrocellular foamed materials.USA: Massachusetts Institute of Technology; 1994.

[29] Hedrick JL, Carter KR, Cha HJ, Hawker CJ, DiPietro RA, Labadie JW, et al. High-temperature polyimide nanofoams for microelectronics applications. ReactFunct Polym 1996;30:43e53.

[30] McGrath JE, Jayaraman SK, Lakshmanan P, Abed JC, Afchar-Taromi, F. Poly-imide nanofoams: materials design and development. In: Proceedings of theACS New Orleans Meeting. 1996: New Orleans, pp. 136e137.

[31] Krause B, Sijbesma HJP, Munuklu P, Van Der Vegt NFA, Wessling M. Bicon-tinuous nanoporous polymers by carbon dioxide foaming. Macromolecules2001;34:8792e801.

[32] Krause B, Diekmann K, Der Vegt Van NFA, Wessling M. Open nanoporousmorphologies from polymeric blends by carbon dioxide foaming. Macro-molecules 2002;35:1738e45.

[33] Miller D, Kumar V. Microcellular foaming of high-density polyethylene usingsub-critical carbon dioxide, in SPE Polyolefins; 2008. Houston, TX.

[34] Nemoto T, Takagi J, Ohshima M. Control of bubble size and location in nano-/microscale cellular poly(propylene)/rubber blend foams. Macromol Mater Eng2008;293(7):574e80.

[35] Nemoto T, Takagi J, Ohshima M. Nanoscale cellular foams from a poly(-propylene)-rubber blend. Macromol Mater Eng 2008;293(12):991e8.

[36] Miller D, Chatchaisucha P, Kumar V. Microcellular and nanocellular solid-statepolyetherimide (PEI) foams using sub-critical carbon dioxide I. Processing andstructure. Polymer 2009;50(23):5576e84.

[37] Gibson LJ, Ashby MF. Cellular solids: structure & properties. In: . New York:Pergamon Press; 1988.

[38] Kumar V. Microcellular plastics: does microcellular structure always lead toan improvement in impact properties? in SPE ANTEC; 2002. San Francisco, CA.

[39] Sills S, Overney RM, Chau W, Lee VY, Miller RD, Frommer J. Interfacial glasstransition profiles in ultrathin, spin cast polymer films. J Chem Phys 2004;120(11):5334e8.

[40] Frank B, Gast AP, Russell TP, Brown HR, Hawker C. Polymer mobility in thinfilms. Macromolecules 1996;29(20):6531e4.