Electrospun mullite fibers from the sol–gel precursor - Springer Link

J Sol-Gel Sci Technol (2015) 74:208–219 DOI 10.1007/s10971-014-3599-7

ORIGINAL PAPER

Electrospun mullite fibers from the sol–gel precursor Zhaoxi Chen • Zhao Zhang • Chen-Chih Tsai Konstantin Kornev • Igor Luzinov • Minghao Fang • Fei Peng

•

Received: 27 August 2014 / Accepted: 11 December 2014 / Published online: 23 December 2014  Springer Science+Business Media New York 2014

Abstract Mullite fibers with diameters from 400 nm to 10 lm were fabricated from the sol–gel precursors using the electrospinning method. During the precursor synthesis, the hydrolysis was controlled to obtain highly viscous mullite sols. The viscous mullite sols were then diluted and mixed with a small amount of polyethylene oxide. Controlling the precursor rheology and spinning conditions, we obtained mullite fibers with the relatively uniform microstructure and narrow diameter distributions for each e-spinning condition. We carried out the mechanical tests for the electrospun mullite fibers since the mechanical performances of e-spin ceramic fibers have not been often reported. The tensile strengths of electrospun mullite fibers were determined using the single filament tensile test. The average tensile strength was 1.46 GPa for 5 mm gauge length, and 1.25 GPa for 10 mm gauge length. The Weibull modulus was estimated to be 3–4, which is comparable to commercial ceramic fibers. The fiber exhibited an average elastic modulus of 100 GPa. In this study, we show that controlling the hydrolysis can reduce the polymer additive amount required for electrospinning. Thus the electrospun mullite fiber has the similar mechanical properties to the dry spun counterparts. Keywords

Sol–gel  Mullite fiber  Weibull modulus

Z. Chen  Z. Zhang  C.-C. Tsai  K. Kornev  I. Luzinov  F. Peng (&) Department of Materials Science and Engineering, COMSET, Clemson University, Clemson, SC 29634, USA e-mail: [email protected] M. Fang School of Materials Sciences and Technology, China University of Geosciences, Beijing 100083, People’s Republic of China

123

1 Introduction Mullite has excellent high temperature strength, creep resistance, and good chemical stability [1–4]. Thus mullite fibers have been widely used as the reinforcement in ceramic matrix composites [5, 6]. These fibers are also used as the high temperature or electrical insulating materials [6, 7]. Mullite fibers have been produced for decades using the dry-spinning technology [8–16]. The diameters of the dry-spun fibers are generally about tens of micrometers. The precursors were often derived from aluminum alkoxide and tetraethyl orthosilicate (TEOS) [9– 11]. The polymerization reaction was controlled via hydrolysis and condensation until the appropriate viscoelastic behavior was reached. In many applications, it is desirable to have fibers with small diameters and high surface areas. For instance, thin fibers are desired in sensing and catalyst applications. Owning to their high surface areas, the sensitivity and catalytic activity have been significantly improved [17–19]. In structural ceramic composites, nanosized ceramic fibers in a ceramic matrix can lead to an improvement in the mechanical properties [20, 21]. Hence nanofibers become especially important. The dry spinning technology is difficult to achieve sub-micrometer diameters. Electrospinning is a versatile processing tool for ceramic fibers, especially nanofibers [22–24]. Recently the mullite fibers have been fabricated using electrospinning of aluminum isopropoxide (AIP), aluminum nitrate (AN) and TEOS directly mixed with polymer additives, such as polyvinylpyrrolidone (PVP), polyvinyl butyral (PVB) or polyvinyl alcohol (PVA) [25– 27]. A high concentration of PVP, normally 5–8 % PVP in the precursor was required for electrospinning [25–27]. The reports on mechanical properties of the electrospun mullite fibers can rarely be found. In this work, we present

J Sol-Gel Sci Technol (2015) 74:208–219

209

an approach for electrospinning of mullite fibers which only requires a small concentration of polymer spinning aid of \0.3 %. We show that high molecular weight polyethylene oxide (PEO) can help to improve the formation of e-spun fiber. The diameters of produced fibers had narrow distributions. The thermal decomposition of the gel fibers were studied using DTA and TGA. The crystalline phases and microstructure of calcined fibers were characterized using XRD and SEM. The fibers microstructure and mechanical properties were comprehensively studied.

2 Experimental 2.1 The mullite sols Aluminum isopropoxide (AIP, Al(C3H7O)3, 98 %,), aluminum nitrate (AN, Al(NO3)39H2O, 98 %, Alfa Aesar, MA, USA) and tetraethyl orthosilicate (TEOS, Si(OC2 H5)4, 98 %, Acros Organics, NJ, USA) were used to synthesize the precursor. The precursor compositions are given in Table 1. Water was used as the solvent. For each batch, the total concentration of AIP and AN were kept at 0.6 mol. AN was dissolved in deionized water at room temperature by vigorously stirring it for 30 min. Then AIP and TEOS were added into the solution and stirred for 20 h. AIP and TEOS were dissolved completely, and clear solutions were obtained. Each solution was then refluxed at 80 C for 5 h. Approximately 2/3 part of the solvent was removed using a rotary evaporator (IKA RV 10 digital, IKA, China). The obtained solutions were then set in an oven at 80 C until viscous sols were formed. The spinnability of these sols was determined using hand-drawing with a glass rod. 2.2 The e-spinning precursors A PEO (MW 1,000,000, Aldrich, MO, USA) solution of 2 wt% in H2O was prepared separately as the spinning aid Table 1 The initial precusor compositions Sol code

AIP (mol)

AN (mol)

TEOS (mol)

H2O (mol)

1

0.3

0.3

0.2

10

2

0.35

0.25

0.2

10

3.1

0.36

0.24

0.2

10

3.2

0.38

0.22

0.2

10

3

0.4

0.2

0.2

10

3.3

0.42

0.18

0.2

10

3.4

0.44

0.16

0.2

10

4

0.45

0.15

0.2

10

5

0.5

0.1

0.2

10

solutions. Sol 3.4 from Table 1 was diluted in ethanol, and then mixed with small amount of PEO solution. These solutions were ready for electrospinning, and are called ‘Esol’ in Table 2. The volume ratios between the initial mullite precursor, PEO solution, and ethanol are given in Table 2. The calculated mullite yields and PEO concentrations with respect to the E-sol volume are also given in Table 2. 2.3 Electrospinning and firing The fibers were electrospun under an applied electrical field generated using a high voltage supply (Model PS/ FC60P02.0-11, Glassman High Voltage Inc., NJ, USA). A positive voltage of 10 kV was applied to the needle of the syringe containing e-spun solutions driven by a syringe pump (Model NE-300, New Era Pump System Inc., NY, USA). The flow rate was set to the vicinity of 0.5 ml/h. The needle was placed 20 cm apart from the collector. The fibers were produced at 25–35 % ambient relative humidity and collected using a rotating collector. The rotating collector has four grounded stainless steel bars, with a gap of *10 cm between each pair of adjacent bars. The fibers were then cut from the collector and collected in the form of mats. The obtained fibers were dried at 60 C for 24 h before firing. The heating rate was set at 1 C/min below 500 C and 10 C/min above 500 C. The fibers were fired at 1,000, 1,200, or 1,400 C for 2 h. 2.4 Characterizations The viscosity of sols was measured using a viscometer (Viscolead ADV, Fungilab Inc., NY, USA) at room temperature. The thermal characteristics were studied at different heating rate under flowing air condition using DTA (DTA7, Perkin Elmer, Waltham, MA, USA), and TGA (TGA7, Perkin Elmer, Waltham, MA, USA). The fiber microstructure was characterized using X-ray diffraction (XRD, Rigaku Co., Ltd., Tokyo, Japan) and scanning electron microscopy (SEM, Hitachi S4800, Hitachi, Ltd., Tokyo, Japan). The average size of fiber diameters of the as-spun and fired fibers was calculated from more than 100 randomly selected fibers taken from SEM micrographs. 2.5 Mechanical test The single filament tensile tests were carried out using a single filament tensile testing machine (Instron 5582, Instron Ltd., High Wycombe, Buckinghamshire, UK). During each test, a single mullite fiber was mounted and fixed using superglue or tape onto a C-card. After mounted on the test machine, the C-card’s neck was cut open. The strain rate is set as 1 mm/min. Two gauge lengths (5 and

123

210

J Sol-Gel Sci Technol (2015) 74:208–219

10 mm) were used. The fiber diameters for each test were measured using optical microscope (Olympus BX51, Olympus Optical Co. Ltd., Tokyo, Japan).

3 Results 3.1 The precursor characteristics for electrospinning Table 3 summarizes the appearance mullite precursor after mixing and condensation. The pH value and spinnability after condensation are also given in Table 3. A viscous sol was obtained from sol 3, 3.3, 3.4, 4 and 5 respectively without apparent phase separation. The viscosity changes with time of the mullite precursors are shown in Fig. 1. In general, sols 3 and 3.4 had relative mild slopes in the late stage of polymerization, especially when the viscosities

Table 2 The electrospinning sol compositions E-sol code

M:P:E (volume ratio)

Calculated mullite yield (grams per 100 ml E-sol)

PEO concentration (grams per 100 ml E-sol)

E-1

4:1:0.5

32

0.27

E-2

4:1:2

28

0.23

E-3 E-4

4:1:4 4:1:8

24 18

0.20 0.15

E-5

4:1:12

14

0.12

E-6

4:1:16

12

0.10

E-7

4:1:20

10

E-8

4:2:22

9.4

0.16

E-9

4:2:26

8.3

0.14

E-10

4:2:30

7.5

0.13

E-11

4:2:34

6.8

0.11

E-12

4:2:40

6.0

0.10

0.09

‘M:P:E’ means volume ratios between initial mullite sol, PEO solution, and ethanol

Table 3 The appearances of mullite initial precursor before and after condensation

The pH value and spinnability after condensation are also given in this table. The spinnability is determined using the hand drawing method

123

Sol code

were greater than 100 poises. Sols 4 and 5 showed steep increase of viscosity with time. In the sol–gel process, the sol viscosity can be greatly affected by a small change in the processing conditions, such as initial concentration, pH value, and setting temperature. Sometimes the ‘relative time’ is used to compare the polymerization behaviors [14]. Relative time is defined as t/tg, where t is the real time and tg is the total gelation time when the sol becomes the solid gel. Figure 1b shows the viscosity change versus relative time when the viscosity was above 100 poises. At this stage, the viscosity increased drastically. Sols 4 and 5 showed narrow viscosity-time windows for processing at the final stage of gelation compared with sols 3 and 3.4. Thus sol 3.4 was selected for fabrication of electrospinning precursors. All the mullite fibers obtained in this study were from sol 3.4. The initial precursors were too viscous for e-spinning. Therefore, mullite sols were diluted with ethanol and PEO solutions. Figure 2 shows the viscosity of modified precursors (E-sols) at different shear rate. All E-sols exhibited shear-thinning behavior. With increasing ethanol content, the viscosity decreased gradually. 3.2 Thermal analysis and phase identification In the DTA traces, shown in Fig. 3a, the endothermic peaks at 100–200 C were attributed to the evaporation of absorbed solvent and low molecular weight organics [13]. The endothermic peak at 300–400 C was due to the decomposition of polymer chains [13]. The exothermic peak at around 1,000 C corresponded to the crystallization of mullite phase. These peaks are consistent with other reports [12, 13]. With increasing heating rate, the exothermic peaks shifted to the higher temperatures. The TGA result (Fig. 3b) was consistent with the DTA results. With a heating rate of 5 C/min, the significant weight loss occurred at 100–400 C. There was no significant weight loss at above 500 C, indicating the completion of organic decomposition.

Sol appearance Before condensation

pH value

Spinnable

After condensation

1

White precipitation, opaque

Opaque and precipitation

1.28

No

2

Silver–gray and clear

Opaque and precipitation

1.62

No

3.1 3.2

Colorless and clear Colorless and clear

Opaque and precipitation Opaque and precipitation

1.69 1.90

No No

3

Colorless and clear

Colorless and clear

2.32

Yes

3.3

Colorless and clear

Colorless and clear

2.77

Yes

3.4

Colorless and clear

Colorless and clear

2.88

Yes

4

Lightly gray and clear

Silver–gray and clear

2.89

Yes

5

Lightly gray and clear

Silver–gray and clear

3.13

Yes

J Sol-Gel Sci Technol (2015) 74:208–219

211

3.3 Fiber microstructure

Fig. 1 Viscosities of sol 3, 3.4, 4, and 5 versus polymerization time. a Versus real time and b versus relative time

The precursor spinnability and obtained mullite fiber microstructure are summarized in Table 4. Except E-12, the rest of the E-sols were spinnable. Only beads were obtained without fiber from E-12. The SEM micrographs of electrospun fibers are shown in Fig. 5. Mullite fibers with diameters from 500 nm to 15 lm were obtained from the solutions with the compositions described in Table 2. Figure 5 shows some of the electrospun fibers. The fibers in Fig. 5a–e showed uniform microstructure. The surfaces of the fibers were smooth. With the increasing ethanol content up to 81 vol%, the diameter of the fiber gradually decreased to about 500 nm. Further increasing the ethanol content, we were unable to obtain uniform fibers. The beads appeared and the fiber diameters became uneven, as shown in Fig. 5f, g. When the viscosity was too low, only beads were collected (Fig. 5h). The mullite fibers fired at 1,200 C for 2 h are shown in Fig. 6. The calcined fibers did not fuse with each other. There were no cracks or pores observed on the fiber surface and cross-sections. The surfaces of the heat-treated fibers exhibited tiny grains and thus were rougher than the asspun fiber surfaces. This feature was caused by the mullite grain growth. The fiber diameters can be achieved from above 10 lm to about 400 nm. Figure 6a shows the largest diameter, which is about 12 lm. Figure 6b–e show the mullite fibers with different diameters obtained from E-sols in Table 2. The thinnest mullite fiber obtained in our study was about 400 nm in diameter. In general, the obtained mullite fibers have narrow diameter distribution for each batch, not including the fibers obtained from E-10 and E-11. Those fibers have uneven fiber diameters as spun. Figure 7 shows the diameter distribution of mullite microfibers obtained from E-3. These fibers were selected for mechanical test, because the fiber diameter of *3 lm is about the limit for our tensile test machine. We were not able to test smaller diameter fibers. 3.4 Mechanical properties 3.4.1 Tensile strength

Fig. 2 Viscosities of diluted precursors versus shear rate. The bottom pictures show the effect of the dilution

The XRD results on the mullite sol after heat-treatment at 800, 1,000 and 1,200 C was shown in Fig. 4. The labeled peaks indicate the well-defined pure mullite phase without spinel phase. In the sol–gel processing, elimination of the spinel formation is challenging and important [28–30].

There were two mounting methods used to fix the fibers on the C-cards. We use superglue or tapes to fix the fibers on the C-cards. When superglue was used to fix the two ends, the fiber was not able to slide during the test. This is because after superglue was cured, it completely became solid. When the tape was used, the fiber can slip during the test. Figure 8 shows the difference of the stress–strain curves of these two mounting methods. If the fiber did not slide during stressing, the modulus measurement was accurate. The tape mounting did not generate any

123

212

J Sol-Gel Sci Technol (2015) 74:208–219

Fig. 3 DTA (a) and TGA (b) traces of mullite sol

4 Discussion 4.1 Controlling the electrospinning precursors

Fig. 4 XRD trace of mullite powder after calcination at 800, 1,000 and 1,200 C for 2 h (mullite filled circle)

meaningful modulus values, but the measured tensile strengths were accurate. The success rate for the 5 mm gauge length measurement was extremely low (\10 %) if superglue was used. Most of the fibers were broken at the superglue contact spots rather than in the middle of the fibers. The mechanical properties of the mullite fibers were summarized in Table 5. The result values were obtained based on at least 20 successful measurements. The 5 mm gauge length test was done using the tape to mount the fibers. Thus no modulus data were obtained with this group of testing. The average tensile strengths of the mullite fibers were about 1.46 GPa for 5 mm gauge length, and 1.25 GPa for 1 cm gauge length. An average elastic modulus of about 100.02 GPa was determined when superglue was used to mount the fiber.

123

One of the purposes of this study is to form the mullite fiber using the polymer spinning aid of as less amount as possible. The low polymer content is desired to achieve high density of the fibers. High polymer contents generally result in porous fiber microstructure consequently with poor mechanical performance. The sol–gel precursors in this study differ from previously reported electrospinning precursors in that it was achieved with controlled sol–gel reactions, including hydrolysis, polymerization, and condensation. In the other studies, mullite fibers were derived from the precursors without controlling the sol–gel reactions. The polymer spinning aid was used to form the fiber backbone during electrospinning. We controlled the hydrolysis process in the beginning, so that the mullite sol itself was spinnable. Thus the content of the required polymer spin aid can be minimized. We can even e-spin diluted sol without adding any PEO. However adding PEO did improve the continuity and uniformity of the fibers. Our polymer contents in the E-sols are substantially lower compared with the other studies. Approximately 0.1–0.3 % PEO was used, which is dramatically less than the amount in other reports (e.g. 5–8 % in refs. [25–27]). Mullite fibers are usually used as the reinforcement for ceramic matrix composites (CMCs). The toughening mechanism of CMCs is caused by the crack deflection at the fiber/matrix interfaces. The thinner fibers will result in more interfaces per unit volume. Currently there is no study on using submicrometer mullite fibers as the CMC reinforcement. Thus it is desirable to have a method to control the fiber diameters from micrometers to submicrometers for systematic research. We find that the major factor determining the fiber diameter is the total solid concentration in the precursor. From Table 2 and Figs. 5, 6

J Sol-Gel Sci Technol (2015) 74:208–219 Table 4 Summary of microstructural characteristics of electrospun mullite fibers

213

Sol code

Average diameter of as-spun fibers (lm)

E-1

*17

Average diameter of mullite fibers after firing (lm)

Coefficient of variation

Fiber appearance

12.24

0.167

Uniform

E-2

*6.0

4.65

0.079

Uniform

E-3

*4.5

3.53

0.141

Uniform

E-4

*3.0

1.50

0.123

Uniform

E-5

*1.5

0.95

0.147

Uniform

E-6

*0.9

0.77

0.115

Uniform

E-7 E-8

*0.5 *1.2

0.32 0.87

0.130 0.144

Uniform Uniform

E-9

*0.8

0.68

0.121

Uniform

E-10

*0.6

0.41

0.262

Uniform with a few beads

E-11

*0.5

*0.4

NA

Nonuniform with beads

E-12

–

–

and 7, we can see that the lower solid concentration, the smaller diameter of the fiber will be. Other factors such as the feeding rate and the magnitude of the electric field will also play important roles in controlling the fiber diameter. Those effects will be discussed in Sect. 4.2. However we found that controlling the solid concentration is the most convenient way for us to have the well-controlled diameters. The fiber diameter increases from *315 nm in sol E-7 to *12 lm in sol E-1, which is due to higher viscosity of the solution at higher AIP–AN–TEOS–PEO concentration. The fiber diameter obtained from sol E-8 was greater than that from E-7. This is probably due the viscosity difference in the two precursors, since higher viscosity generally results in a larger fiber diameter. The viscosities of E-7 and E-8 are not presented in here, because of the limitation of our viscosity testing instrument. The viscosities of E-7 and E-8 were too low to be accurately determined, both were below 0.1 poise. The concentration of PEO nearly doubled in sol E-8 than in E-7; and the sol concentration for both, in the term of mullite yield, was similar. In sols E-10 to 11, surface tension was dominant and beaded fibers were formed during the electrospinning process [31, 32]. The conventional e-spin method of fabricating mullite fibers shows difficulties in making thick fibers with diameters above 1 lm. Merely increase the polymer content cannot achieve thick fibers. This is because of the formation of wide and flat ribbon shaped fibers from the high polymer content precursors [25–27]. In our process, only low PEO concentrations are required, we did not observe any ribbonlike fibers. The fibers obtained from the sol–gel/e-spin precursor showed round cross-sections with controlled diameters from nano- to micrometers, which has not been reported in the other studies related to electrospinning of mullite fibers.

Only beads

4.2 The decomposition, phases, and microstructure The mixing level between TEOS and aluminum alkoxide during hydrolysis is important because sometimes these two chemicals can phase separate during hydrolysis, resulting in the so-called diphasic gel [8]. The diphasic gel decreases the fiber strength and creep resistance at high temperatures [12]. To avoid phase separation, monophasic gel can be achieved using an aqueous mixture of AN, AIP, and TEOS [8, 12–14]. The obtained mullite fibers had low crystallization temperature at below 1,000 C. Thus the grain growth can be effectively prohibited [12–14]. In the DTA experiment, the activation energy for mullite phase formation (also known as mullization) can be calculated from the DTA exothermic peaks using the Kissinger’s equation [33],  2   TP Ea Ea ln  ln v ð1Þ ¼ ln þ Pr R RTp where Tp is the exothermic peak temperature; Pr is the heating rate; Ea is the activation energy for crystallization; R is the gas constant; and v is the frequency factor constant. This method has been reported to determine the activation energy for mullization by many authors [34–36]. The activation energy calculated from Eq. (1) was Ea = 1,411 kJ/mol. This value is in good agreement with many of the previous reports [34–36]. Okada showed that in monophasic gels, the Ea values for mullization range from 800 to 1,400 kJ/mol [36]. The corresponding crystallization temperature has a maximum of about 1,000 C [36]. The monophasic gel is desired for two reasons. First, the complete phase formation without significant grain growth at low temperatures is important for mechanical strength. Second, the spinel phase can be avoided to ensure the high temperature creep resistance. Many studies have shown

123

214

J Sol-Gel Sci Technol (2015) 74:208–219

Fig. 5 Microstructure of electrospun green fibers obtained from. a E-3, b E-4, c E-6, d E-7, e E-9, f E-10, g E-11, and h E-12

that if mullite was synthesized using AIP–AN–TEOS as the precursors, mullite was the only phase crystallized at temperatures below 1,000 C [34–40]. The single mullite phase formation was caused by the controlled hydrolysis reaction in moderate acidic condition (e.g. pH = 3–4.5). This yields Al–O–Si bonding in atomic level [39]. The simple unit of aluminosilicate complex then polymerizes in a preferentially linear arrangement, which facilitates the

123

fiber formation [41]. Phase separation does not occur in the above mentioned precursors. They are often referred as the monophasic precursors. In contrast, diphasic precursor will have phase separation, and as a result, the spinel phase was observed during crystallization [8]. In order to enhance the mechanical performance of the obtained fibers, we need to not only limit the grain sizes, but also increase the final density. In our work, grain

J Sol-Gel Sci Technol (2015) 74:208–219

215

Fig. 6 SEM images of mullite fibers after calcination at 1,200 C for 2 h. a Thick fibers with diameter (d) [10 lm, b, c microfibers with 1 \ d \ 5 lm, d sub-micro fibers with d *800 nm, and e nanofiber with d *300 nm

growth was inhibited in all fibers with average grain size below 10 nm after firing at 1,000 C, because the fibers only need to be heated up to 1,000 C to obtain the mullite phase. In the di-phasic system, the grain size is about 100 nm due to a higher phase formation temperature at 1,200 C, where co-existence of c-AlO3 phase was also observed [25]. The XRD results show that only single mullite phase was observed at between 800 and 1,200 C without any spinel phase. Both the XRD results and the calculated activation energy of crystallization indicate the monophasic character of the precursor. In order to ensure the final density of the fibers, we need to have high ceramic yield from the precursor, or low weight loss. The average shrinkage of the diameter of the as-spun fibers to the calcined fibers was about 30 % in our study, which is less than the value of about 50 % mentioned in other reports [26, 27]. The ceramic yield was examined using the TGA results. In our study, the weight loss was\60 % after the

gel was dried. On the other hand, about 70 % weight loss was mentioned in the other study [26]. According to the TGA analysis, over 95 % of the weight loss happened at the temperatures below 500 C. During the annealing process, slow heating rate (1 C/min) was employed at T\500 C, in order to prevent the rapid weight loss and cracks formation. Similar heat-treatment method was used for mullite fibers obtained by the dry-spinning method [15]. In other previous studies, mullite fibers with diameters around 100 nm were achieved using electrospinning [26, 27]. The feeding rate used in their studies was 0.2 ml/min, and in this study it was 0.5 ml/min. The similar electric field was used. In other previous studies, the collection distances (*13–15 cm) and applied voltages (10–20 kV) are similar to this study (*20 cm and 10 kV respectively). We did not achieve the thin fibers as the previous studies. This is because the purpose of this study is to significantly reduce the polymer content, so that the

123

216

J Sol-Gel Sci Technol (2015) 74:208–219

Fig. 7 Statistical diameter distribution of mullite fibers in Fig. 6a–e. The distribution was fitted by normal function (dashed line)

ceramic yield can be greatly improved. Experimentally, thinner fibers could be obtained with the low feeding rate combined with the high electric field. When the applied electric field was high, the electric force tended to break the jets and form short fibers, resulting in non-uniform microstructure. However, relatively high voltages with low feeding rates were applied in those studies [26, 27]. This is because in those studies, high polymer contents (e.g. 5–8 %) were used to facilitate the stretching process. In our study \0.3 % polymer aid was used. In our study, the resistance to the breakdown of the jets came from the wellcontrolled polymerization process, or the sol–gel process, instead of merely using polymer aid to form the fibers. Fig. 8 Stress strain curve of mullite single fiber

123

J Sol-Gel Sci Technol (2015) 74:208–219

217

Table 5 Average tensile strength, elastic modulus measured at different gauge length Gauge length (mm)

Average tensile strength (GPa)

Elastic modulus (GPa)

10

1.25

96–104

5

1.46

–

The result values were obtained based on at least 20 successful measurements

4.3 Fiber strength and elastic modulus To our best knowledge, there are very few works have been done on measuring the mechanical properties of electrospun mullite fibers. However some works can be found for measuring the mechanical properties of small diameter (e.g. 3–5 lm) dry-spun mullite fibers other than the Nextel and Altex fibers [42, 43]. Li et al. [42] reported a tensile strength of 1.1–1.4 GPa for the phase pure mullite fiber of diameter of 3–5 lm. A tensile strength of 1.3–1.6 GPa was reported for alumina rich electrospun mullite fibers of diameter 3–12 lm [43]. These two results cannot be directly compared because the phases in ref. [43] were a mixture of c-Al2O3 and mullite. Both above studies did not report the fibers’ elastic moduli. The tensile strength of the mullite fibers obtained in our study is similar or slightly better than the studies of dryspinning mullite fiber without any polymer spinning aid [42, 43]. This shows the importance of controlling the spinning aid contents. The polymer spinning aid usually leaves porous microstructure after burned off. Controlling the hydrolysis process can significantly reduce the needed spinning aid content, and thus resulting in improved mechanical performance. The grain size to diameter ratio has a crucial effect on the mechanical properties of the fibers. For instance, significant and exaggerated grain growth was found to deteriorate the failure strength of mullite microfibers [44]. In nanofiber systems, grain growth determines the nanocrystalline structure and mechanical performance [19]. Sometimes good mechanical properties of nanofibers may take advantage of the small grain size and high volume fraction of grain boundary of the nanocrystalline structures. On the other hand, grain boundaries may act as extrinsic defects that are important in dictating the mechanical properties [19]. The detailed mechanism remains unclear and more studies need to be done. 4.4 Fiber strength distribution and Weibull modulus The mullite fiber strength statistics is an important indication of flaw density [45–51]. The statistical variation of the tensile strength of fibers is now commonly reported in

terms of Weibull modulus [45–47]. This is a model for the failure probability of a fiber under an applied stress with a surface distribution of defects. This method has been applied to the commercial mullite fibers (Nextel 720 and 550) manufactured by 3 M Company, and proved to be important [47–49]. According to Weibull distribution method, the probability of failure, P, is known as [45, 47]:     V r m P ¼ 1  exp  ð2Þ V0 r 0 where m is the Weibull modulus, V is the test volume, r is the stress and V0, r0 are scaling constants [45]. Assuming the fibers have the same diameter, then Eq. (2) can be derived into [47] ln lnð1=ð1  PÞÞ ¼ m ln r þ k

ð3Þ

where k is a constant. Figure 9 shows the fracture probability of mullite fibers from the tensile test in this work. The Weibull modulus obtained was between 3.1 and 3.2 for both 1 and 0.5 cm gauge lengths. This Weibull modulus value range is comparable to many commercial ceramic fibers. For example the Weibull moduli for Nicalon fiber was 3–4, and for carbon fibers were 3–8 [50, 51]. However the reported Nextel 720 fiber had a Weibull modulus of 7–8 [47]. Actually the ‘true’ Weibull modulus of the mullite fibers obtained in this study should be well above 3, since the diameters of the electrospun fibers are typically less uniform than the diameters of dry-spun fibers. Electrospun fibers are formed during liquid ejection under applied electric field where whipping and bending instability acts within a very short time to stretch the thick liquid jets into very thin fibers. The distribution of fiber diameters is usually broader than the dry-spun counterparts because the dry spinning process uses the mechanical force to stretch the precursor. Currently the dry spinning method is so mature that the stretching velocity can be precisely controlled to ensure highly uniform diameters. However controlling the stretching force of electrospinning is difficult. The fluid is stretched into much smaller diameters within a much shorter period of time, compared to dry spinning. The fiber diameters are sensitive to small changes to local environment, e.g. the electric field, the surface charge, and the solvent vaporization. The deviation of fiber diameter affects the calculated modulus using the Weibull plot [47]. The measured diameter had an absolute deviation of about 15 %. The variation in gauge length was about 12 %. These deviations may cause up to 48 % variation in volume. However, the volume is assumed to be constant if Eq. (3) is used. The scattering effect on the calculated Weibull modulus caused by the volume deviation is shown in Fig. 10. If all the

123

218

J Sol-Gel Sci Technol (2015) 74:208–219

m ¼

Dy þ lnðV=V0 Þ m Dy

ð4Þ

where Dy is the difference between the maximum and minimum values of ln ln(1/(1 - P)). The estimated true Weibull modulus m* is 3.9.

5 Conclusion

Fig. 9 The plot of Eq. (3) for Weibull modulus calculation

Mullite fibers have been processed from monophasic sol– gel precursor using electrospinning. The fiber diameter can be controlled with the solute concentration in the electrospinning precursors. Fiber diameters can be controlled from 400 nm up to *10 lm. The obtained fibers showed narrow diameter distributions. The fiber had an average strength of 1.25 GPa for 1 cm gauge length, and 1.46 GPa for 0.5 cm gauge length. These mechanical properties show that electrospinning can be used to fabricate small diameter mullite fibers with good mechanical properties, which is comparable to the dry-spun counterparts. Acknowledgments This Project was funded by the Air Force Office of Scientific Research, Contract FA9550-12-1-0459. The authors would like to express their appreciation for the helpful suggestions and support of their contract monitor, Dr. Ali Sayir.

References

Fig. 10 The effect of volume deviation in Weibull probability plot

fibers had the exact same volume, then the Weibull plot would follow the dashed lines in Fig. 10. This is because in a typical ln lnð1=ð1  PÞÞ versus ln r plot, one assumes constant testing volume V, which yields a linear function with slope m and intercept k, as can be derived from Eq. (2). The testing volume information was reflected in the value of k. However, certain volume deviation always exists in our study. This is because of the distribution of fiber diameter. Consequently, the ln lnð1=ð1  PÞÞ versus ln r plot then reflects a group of testing specimen with testing volume V1, V2, V3, …, Vn. Each test volume corresponds to the same slope (true modulus m* here) and different intercept ki (i = 1, 2, 3, …, n). Each testing volume Vi corresponds to a dashed line in Fig. 10. Assume that the Weibull modulus is not affected by this V, then different volumes correspond to a series of parallel dashed lines. The overall Weibull modulus plot from a group of fibers with volume deviation, will intersect with all dashed lines, as shown in Fig. 10. The true Weibull modulus, m*, can be derived as:

123

1. Aksay IA, Dabbs DM, Sarikaya M (1991) Mullite for structural, electronic, and optical applications. J Am Ceram Soc 74(10):2343–2358 2. Kriven WM, Palko JW, Sinogeikin S, Bass JD, Sayir A, Brunauer G, Schneider J (1999) High temperature single crystal properties of mullite. J Eur Ceram Soc 19(13):2529–2541 3. Dokko PC, Pask JA, Mazdiyasni KS (1977) High-temperature mechanical properties of mullite under compression. J Am Ceram Soc 60(3–4):150–155 4. Kanzaki S et al (1985) Sintering and mechanical properties of stoichiometric mullite. J Am Ceram Soc 68(1):C-6 5. Chawla KK, Xu ZR, Ha J-S (1996) Processing, structure, and properties of mullite fiber/mullite matrix composites. J Eur Ceram Soc 16(2):293–299 6. Schneider H, Schreuer J, Hildmann B (2008) Structure and properties of mullite—a review. J Eur Ceram Soc 28(2):329–344 7. Merrill GB, Morrison JA (2000) High temperature insulation for ceramic matrix composites. U.S. Patent No. 6,013,592, 11 Jan 2000 8. Cividanes LS, Campos TM, Rodrigues LA, Brunelli DD, Thim GP (2010) Review of mullite synthesis routes by sol–gel method. J Sol-Gel Sci Technol 55(1):111–125 9. Kim GD, Lee DA, Lee HI, Yoon SJ (1993) A study on the development of mullitent fibers using the sol–gel process. Mater Sci Eng A 167(1):171–178 10. Yogo T, Aksay IA (1994) Synthesis of mullite fibre from an aluminosiloxane precursor. J Mater Chem 4(2):353–359 11. Al-Assafi S, Cruse T, Simmons JH, Brennan AB, Sacks MD (1994) Sol–gel processing of continuous mullite fibers. In: Proceedings of the 18th annual conference on composites and

J Sol-Gel Sci Technol (2015) 74:208–219

12.

13.

14. 15.

16.

17.

18. 19.

20. 21.

22. 23.

24.

25.

26.

27.

28.

29.

30.

advanced ceramic materials-b: ceramic engineering and science proceedings, vol 15, Issue 5, Wiley, pp 1060–1067 Song KC (1998) Preparation of mullite fibers from aluminum isopropoxide–aluminum nitrate–tetraethylorthosilicate solutions by sol–gel method. Mater Lett 35(5):290–296 Okada K, Yasohama S, Hayashi S, Yasumori A (1998) Sol–gel synthesis of mullite long fibres from water solvent systems. J Eur Ceram Soc 18(13):1879–1884 Chen X, Gu L (2009) Sol–gel dry spinning of mullite fibers from AN/TEOS/AIP system. Mater Res Bull 44(4):865–873 Zhang Y, Ding Y, Gao J, Yang J (2009) Mullite fibres prepared by sol–gel method using polyvinyl butyral. J Eur Ceram Soc 29(6):1101–1107 Tan H, Ding Y, Yang J (2010) Mullite fibres preparation by aqueous sol–gel process and activation energy of mullitization. J Alloys Compd 492(1):396–401 Formo E, Lee E, Campbell D, Xia Y (2008) Functionalization of electrospun TiO2 nanofibers with Pt nanoparticles and nanowires for catalytic applications. Nano Lett 8(2):668–672 De Jong KP, Geus JW (2000) Carbon nanofibers: catalytic synthesis and applications. Catal Rev 42(4):481–510 Ramaseshan R, Sundarrajan S, Jose R, Ramakrishna S (2007) Nanostructured ceramics by electrospinning. J Appl Phys 102(11):111101 Sternitzke M (1997) Structural ceramic nanocomposites. J Eur Ceram Soc 17(9):1061–1082 Chronakis IS (2005) Novel nanocomposites and nanoceramics based on polymer nanofibers using electrospinning process—a review. J Mater Process Technol 167(2):283–293 Li Dan, Xia Younan (2004) Electrospinning of nanofibers: reinventing the wheel? Adv Mater 16(14):1151–1170 Dai Y et al (2011) Ceramic nanofibers fabricated by electrospinning and their applications in catalysis, environmental science, and energy technology. Polym Adv Technol 22(3):326–338 Li D, Wang Y, Xia Y (2003) Electrospinning of polymeric and ceramic nanofibers as uniaxially aligned arrays. Nano Lett 3(8):1167–1171 Wu J, Lin H, Li JB, Zhan XB, Li JF (2009) Fabrication and characterization of electrospun mullite nanofibers. Mater Lett 63(27):2309–2312 Mohammad Ali Zadeh M, Keyanpour-Rad M, Ebadzadeh T (2013) Synthesis of mullite nanofibres by electrospinning of solutions containing different proportions of polyvinyl butyral. Ceram Int 39(8):9079–9084 Mohammad Ali Zadeh M, Keyanpour-Rad M, Ebadzadeh T (2014) Effect of viscosity of polyvinyl alcohol solution on morphology of the electrospun mullite nanofibres. Ceram Int 40(4):5461–5466 Selvaraj U, Komarneni S, Roy R (1993) Structural differences in mullite xerogels from different precursors characterized by 27Al and 29Si MASNMR. J Solid State Chem 106(1):73–82 Cassidy DJ, Woolfrey JL, Bartlett JR, Ben-Nissan B (1997) The effect of precursor chemistry on the crystallisation and densification of sol–gel derived mullite gels and powders. J Sol-Gel Sci Technol 10(1):19–30 Ban T, Hayashi S, Yasumori A, Okada K (1996) Characterization of low temperature mullitization. J Eur Ceram Soc 16(2):127–132

219 31. Sigmund W, Yuh J, Park H, Maneeratana V, Pyrgiotakis G, Daga A, Nino JC (2006) Processing and structure relationships in electrospinning of ceramic fiber systems. J Am Ceram Soc 89(2):395–407 32. Dzenis YA (2004) Spinning continuous fibers for nanotechnology. Science 304:1917–1919 33. Kissinger HE (1957) Reaction kinetics in differential thermal analysis. Anal Chem 29(11):1702–1706 34. Okada K, Kaneda JI, Kameshima Y, Yasumori A, Takei T (2003) Crystallization kinetics of mullite from polymeric Al2O3–SiO2 xerogels. Mater Lett 57(21):3155–3159 35. Douy A (2006) Crystallisation of amorphous spray-dried precursors in the Al2O3–SiO2 system. J Eur Ceram Soc 26(8):1447–1454 36. Okada K (2008) Activation energy of mullitization from various starting materials. J Eur Ceram Soc 28(2):377–382 37. Lee JS, Yu SC (1992) Mullite formation kinetics of coprecipitated Al2O3SiO2 gels. Mater Res Bull 27(4):405–416 38. Takei T, Kameshima Y, Yasumori A, Okada K (2001) Crystallization kinetics of mullite from Al2O3–SiO2 glasses under nonisothermal conditions. J Eur Ceram Soc 21(14):2487–2493 39. Chakravorty AK (1994) Effect of pH on 980 C spinel phasemullite formation of Al2O3–SiO2 gels. J Mater Sci 29(6):1558–1568 40. Li DX, Thomson WJ (1990) Mullite formation kinetics of a single-phase gel. J Am Ceram Soc 73(4):964–969 41. Hench LL, West JK (1990) The sol–gel process. Chem Rev 90(1):33–72 42. Li CS, Zhang YJ, Zhang JD (2009) Polycrystalline mullite fibers prepared by sol–gel method. J Inorg Mater 4:044 43. Chatterjee M, Naskar MK, Chakrabarty PK, Ganguli D (2002) Mullite fibre mats by a sol–gel spinning technique. J Sol-Gel Sci Technol 25(2):169–174 44. Schmu¨cker M, Flucht F, Schneider H (1996) High temperature behaviour of polycrystalline aluminosilicate fibres with mullite bulk composition. I. Microstructure and strength properties. J Eur Ceram Soc 16(2):281–285 45. van der Zwaag S (1989) The concept of filament strength and the Weibull modulus. ASTM J Test Eval 17(5):292–298 46. Goda K, Fukunaga H (1986) The evaluation of the strength distribution of silicon carbide and alumina fibres by a multi-modal Weibull distribution. J Mater Sci 21(12):4475–4480 47. Wilson DM (1997) Statistical tensile strength of NextelTM 610 and NextelTM 720 fibres. J Mater Sci 32(10):2535–2542 48. Fernando JA, Chawla KK, Ferber MK, Coffey D (1992) Effect of boron nitride coating on the tensile strength of Nextel 480TM fiber. Mater Sci Eng A 154(1):103–108 49. Deleglise F, Berger MH, Jeulin D, Bunsell AR (2001) Microstructural stability and room temperature mechanical properties of the Nextel 720 fibre. J Eur Ceram Soc 21(5):569–580 50. Wu HF, Netrwavali AN (1992) Weibull analysis of strength– length relationships in single Nicalon SiC fibres. J Mater Sci 27(12):3318–3324 51. Simon G, Bunsell AR (1984) Mechanical and structural characterization of the Nicalon silicon carbide fibre. J Mater Sci 19(11):3649–3657

123