Organic Liquid Impregnation Behavior into Nanofibrous Membranes: Quantitative Analysis of the Effects of Structural Parameters.

This paper reports the effects of structural parameters on organic liquid impregnation behavior into nanofibrous (NF) polymer membranes. The NF membranes were prepared from organic liquidphilic polymers, poly(amide-imide)s (PAIs), by electrospinning. The impregnation velocity of the organic liquid, ethylmethylcarbonate, into the as-spun PAI NF membranes with diameters ranging from 400 to 900 nm was approximately 10-20 times higher than that into commercial cellulose nonwoven membranes. Our theoretical analyses based on the Kozeny-Carman equation and multivariate statistics clearly indicate that in addition to the porosity of the membranes, the variation in fiber diameter as well as the average fiber diameter is a crucial factor for controlling the liquid impregnation behavior.

Introduction

Electrospinning is a straightforward and versatile method for the formation of continuous thin fibers based on an electrohydrodynamic process.[1−3] This method has the following advantages: (i) it is applicable to a broad spectrum of molecules, such as synthetic polymers, biological polymers (e.g., proteins and DNAs), and inorganic molecules; (ii) it has the ability to produce thin fibers with diameters in the micrometer and nanometer ranges; and (iii) it enables one-step formation of nanofibrous (NF) membranes (or nonwoven webs) with random network structures. The membrane thickness can also be controlled by the spinning duration. Electrospun NF membranes with high porosity, an interconnected pore structure, and large surface-to-volume ratios have recently attracted much attention in applications such as high-performance air and liquid filter media, battery separators, electrode materials, protective clothes, composites, drug delivery systems, and biomaterial scaffolds for tissue engineering.[1−9]

Some researchers, including our group, reported unique wetting behavior on the surface of electrospun NF membranes,[10−12] such as excellent water repellency,[12] excellent oil repellency,[13] and a metastable Cassie–Baxter wetting state.[14,15] Their wetting behaviors could be controlled by optimizing both the chemical composition and the surface structure of the electrospun NF membranes (i.e., roughness and porosity).[12−16] For the practical applications of NF membranes as battery separators and liquid filters, the liquid impregnation behavior as well as the wetting behavior is important. For example, the initial liquid impregnation into the membranes influences the production process: poor liquid impregnation could delay the production time (i.e., takt time) or reduce the production yields of cells or stacks due to insufficient liquid filling.

To improve the liquid impregnation behavior into NF membranes, elucidation of the crucial factors for liquid impregnation behaviors (e.g., topology of membrane structure, properties of the membrane material and liquid) is strongly required. However, studies on organic liquid impregnation behavior into NF membranes are very limited. This lack of understanding prevents the establishment of a protocol for rationally designing the NF membranes. Thus, in the present study, the influence of NF membrane structures on the liquid impregnation behavior was examined. We prepared NF polymer membranes using electrospinning. In this study, commonly used organic liquidphilic polymers, poly(amide-imide)s (PAIs), were used as the nanofiber material. In addition, we evaluated the impregnation velocities of the organic liquid, ethylmethylcarbonate (EMC), into the as-spun NF membranes with diameters ranging from 400 to 900 nm using time-resolved contact angle measurements.[17] Herein, we attempted a first-approximation theoretical analysis of liquid impregnation kinetics into nanofibrous porous media based on the Kozeny–Carman equation[18] for fibrous nonwovens and a practical theoretical analysis based on multivariate statistics[19] to investigate the influence of the structural parameters of the membranes on the liquid impregnation behavior.

Results and Discussion

Preparation and Characterization of NF Membranes

Figure shows surface scanning electron microscopy (SEM) images of the prepared NF membranes. The average fiber diameters (δ0) ± standard deviation (σ) of the prepared NF membranes were 410 ± 60 nm for PAI-1, 750 ± 270 nm for PAI-2, 710 ± 112 nm for PAI-3, and 912 ± 338 nm for PAI-4, whereas the δ0 ± σ of 570 ± 438 nm for the reference cellulose nonwoven membrane. The structural parameters (fiber diameter and porosity) and wetting and impregnation properties (contact angle and impregnation time and velocity) of the prepared PAI NF membranes and the cellulose nonwoven membrane are summarized in Table . The impregnation velocity (v) for the prepared PAI NF membranes was larger than that for the reference cellulose nonwoven membrane; specifically, the v values for the PAI NF membranes were 13–19 times larger than that of the cellulose nonwoven membrane. This result is partly because the porosity of the as-spun NF membranes (86–92%) is larger than that of the reference cellulose nonwoven membrane (65%) (see Figure S3a). However, approximately 20–25% porosity change does not allow a quantitative explanation of the significant improvement in the EMC impregnation velocity in the range of 1300–1900%. The v–δ0 relationship for the NF membranes is shown in Figure S3b. The v value increased with an increase in the δ0 value for the highly porous PAI NF membranes.

Figure 1

Surface SEM images of (a) the reference cellulose membrane and the prepared (b) PAI-1, (c) PAI-2, (d) PAI-3, and (e) PAI-4 NF membranes. The scale bar is 10 μm.

Table 1

Structural Parameters and Wetting Properties of the Cellulose Nonwoven Membrane and the Prepared PAI NF Membranes

samples	cellulose	PAI-1	PAI-2	PAI-3	PAI-4
average fiber diameter, δ0 [nm]	570 ± 438	410 ± 64	750 ± 270	710 ± 112	912 ± 338
porosity, ε [%]	65 ± 5	86 ± 2	92 ± 2	92 ± 2	90 ± 2
initial contact angle, θa [deg]	15	14	15	15	15
impregnation time, t [s]	2.70 ± 0.04	0.15 ± 0.05	0.12 ± 0.05	0.10 ± 0.05	0.11 ± 0.05
impregnation velocity, v [mm/s]	1.0 ± 2.0	13 ± 5	16 ± 10	19 ± 10	18 ± 15
fiber density, ρfb [g/cm3]	1.5	1.4	1.4	1.4	1.4

Measured 65 ms after the deposition of a 0.2 μL EMC droplet.

Used in eq for the determination of membrane porosity.

Theoretical Analysis of Impregnation Behavior

To investigate the influence of three parameters (i.e., membrane porosity, fiber diameter, and intrinsic contact angle of fiber material) on the liquid impregnation velocity (v) into electrospun NF membranes, we used the Kozeny–Carman equation, which describes the capillary rise (h) of the fluid flowing through the porous media,[18] as followswhere ΔP is the pressure difference across the porous medium, k is the Carman constant, ε is the porosity, Sp is the real specific surface area per unit volume of a porous medium, and η is the fluid viscosity. The model NF membrane is shown in Figure S4, and a detailed description of the Kozeny–Carman equation for NF membranes is included in the Supporting Information.

The impregnation velocities (vcalc) for the commercial cellulose separator and NF membranes were calculated by the Kozeny–Carman equation for NF membranes (eq S11). For calculation, the experimental values of the porosity and fiber diameter (see Table ), the intrinsic contact angle of the fiber materials (see the Supporting Information), and the solution properties of EMC (density, ρL: 1015 [kg/m3]; viscosity, η: 0.65 [mPa s]; and surface tension, γL: 27.1 [mN/m])[20] were used. Figure shows a comparison of the experimental and calculated values of the EMC impregnation velocity into NF membranes. A distinct difference between the experimental results and calculated results appears for all membranes except for the PAI-1 NF membrane.

Figure 2

Experimental and calculated values of the impregnation velocity (v) into NF membranes. Calculations were carried out using the minimum, average, and maximum fiber diameters.

To investigate the difference between the experimental and calculated values of the EMC impregnation velocity into NF membranes in detail, the ratio (R) of the experimental impregnation velocity (v) to the calculated one (vcalc) was plotted as a function of the standard deviation (σ) of the fiber diameter, as shown in Figure . The R values showed a negative correlation with the σ values: the R value increased toward unity with a decrease in the σ value. This observation suggests that the variation in fiber diameter influences the actual liquid impregnation behavior. Subsequently, we recalculated the impregnation velocities using the minimum and maximum fiber diameters instead of the average diameter. The minimum and maximum fiber diameters are the minimum and maximum values obtained from the image analysis of the SEM images of the electrospun fibers, respectively. The results are shown in Figure . The experimental value roughly agrees with the value recalculated using the minimum fiber diameter. In contrast, the difference between the experimental and calculated values became larger when the maximum fiber diameter was used for calculation. This result clearly indicates that compared to the maximum fiber diameter, the minimum fiber diameter is more suitable for describing the liquid impregnation velocity into NF membranes.

Figure 3

Ratio (R) of the experimental impregnation velocity (v) to the calculated one (vcalc) for the NF membranes as a function of the standard deviation (σ) of the fiber diameter.

To examine the reason why the value calculated using the minimum fiber diameter agreed with the experimental value, we reconsidered the Kozeny–Carman equation. In general, for the electrospun NF membranes, the average fiber diameter, δ0, positively correlates with the interfiber spacing, δ−δ0 (δ is the center–center distance between the closet fibers; see Figure S4): a thinner NF membrane has a smaller interfiber spacing.[4] Therefore, we attempt to derive the relationship between δ0 and δ.

The total fiber length (L) is given as follows.where Wf is the weight of the NF membrane with a unit area, La × Lb is the area size of the NF membrane, and ρf is the fiber density.

The porosity of the NF membrane (ε) is obtained by removing the total fiber volume from the total volume of the NF membranewhere Lc is the membrane thickness.

Equation can be rewritten as followsFrom the expression for the total volume of NF membrane δ2L = La × Lb × Lc (see Figure S4), the following equation can be obtained.Finally, the relationship between δ0 and δ is obtained from eqs , 4, and 5.The liquid impregnation behavior into NF membranes is significantly influenced by the interfiber spacing (pore size), δ−δ0 = δ0{1 – (π/4(1 – ε))1/2}. Figure shows the schematic of the liquid impregnation behavior into NF membranes prepared from liquidphilic materials with different interfiber spacings. The impregnation velocities, v1 and v2, have the following relationship (a detailed description for the deviation of eq is included in the Supporting Information).When the spacing δ1–δ1,0 is smaller than δ2–δ2,0, the velocity v1 is smaller than v2. In other words, the impregnation velocity slows down with a smaller interfiber spacing (pore size). Therefore, when there is spacing distribution (i.e., fiber diameter distribution) in the NF membranes, the small spacing formed locally in the membranes substantially prevents fast liquid impregnation and consequently decreases the apparent impregnation velocity into the NF membrane by increasing the actual tortuosity. However, this mechanism for the contribution of the minimum fiber diameter in an NF membrane to the impregnation velocity is plausible. Since considering the effect of the variation in fiber diameter based on the Kozeny–Carman equation is difficult, we attempted a practical multivariate analysis to examine the influence of the variation in fiber diameter on the impregnation velocity.

Figure 4

Schematic of the liquid impregnation behavior into NF membranes with different interfiber spacings The spacing, δ1–δ1,0, is smaller than δ2–δ2,0.

Multivariate Statistical Analysis

To discuss the influence of the variation in fiber diameter in detail, the relationship between the structural parameters, i.e., average fiber diameter (δ0), porosity (ε), standard deviation of fiber diameter (σ), and the impregnation velocity (v) was quantitatively investigated by multivariate statistical analysis[19,21,22] using software (JMP version 11.0.0, SAS Institute Inc.). In this study, to improve the accuracy of our multivariate statistical analysis, the data for an additional NF membrane was used in addition to the data shown in Table (more detailed information on our calculation is included in the Supporting Information).

Figure shows the relationship between structural parameters (ε, δ0, σ) and the impregnation velocity (v) obtained from our multivariate analysis. The correlation coefficient R2 was 0.94, which was a relatively good value. Our analysis revealed positive correlations between v and ε and between v and δ0. The latter correlation indicates that the impregnation velocity slows down with a smaller interfiber spacing because a thinner NF membrane has a smaller interfiber spacing.[4] This reflection will not contradict the plausible mechanism for the contribution of the minimum fiber diameter based on eq . More interestingly, there is a negative correlation between v and σ.

Figure 5

Relationship between the structural parameters, porosity (ε), average fiber diameter (δ0), and standard deviation of fiber diameter (σ), and the impregnation velocity (v) obtained from multivariate statistical analysis.

In other words, the impregnation velocity (v) increases with a decrease in the variation in the fiber diameter (σ). The relationships among the impregnation velocity (v), average fiber diameter (δ0), and variation in fiber (σ) diameter obtained from the multivariate statistical analysis under a fixed porosity (ε) of 92% (the maximum value for our calculation) are shown in Figure . The NF membrane with very small diameters of less than 100 nm usually has inferior liquid impregnation properties. However, this estimation clearly indicates the possibility that liquid impregnation into NF membranes with very thin diameters can be improved by the precisely controlling the fiber diameter during spinning. In fact, we succeeded in preparing poly(vinylidene fluoride) (PVDF) NF membranes with a narrow distribution of fiber diameter (δ0 ± σ = 240 ± 32 nm, ε = 90%) by precisely controlling the charge density of the electrified liquid jet during electrospinning based on the analysis of the direct observation using a high-speed camera.[3] The EMC impregnation velocity was substantially improved to 9.7 from 6.0 mm/s for the PVDF NF membrane with a broad fiber diameter distribution (δ0 ± σ = 230 ± 92 nm, ε = 90%).

Figure 6

Relationships among the impregnation velocity (v), average fiber diameter (δ0), and variation in fiber diameter (σ) obtained from the multivariate statistical analysis at a porosity of 92%.

Conclusions

In the present study, organic liquid impregnation behaviors into electrospun NF membranes were characterized and analyzed based on the Kozeny–Carman equation and multivariate statistics. It was revealed that the variation in the fiber diameter, the membrane porosity, and the average fiber diameter are the determining factors that control the actual liquid impregnation behavior. These results provide the fundamental information for the rational design of electrospun NF membranes and coatings utilized in application fields such as liquid filters, diaphragms, battery separators, functional coatings, and sensors.

Experimental Section

Materials

Poly(amide-imides)s with average molecular weights of 30 000 (PAI-30k for PAI-1 NF), 26 000 (PAI-26k for PAI-2 NF), and 60 000 (PAI-60k for PAI-3 and PAI-4 NFs) were purchased from Toyobo, Japan (PAI-30k and PAI-60k), and Hitachi Chemical, Japan (PAI-26k), respectively. Carboxymethylcellulose (CMC, CMC Daicel Grade 1390) was obtained from Daicel FineChem Ltd., Japan. N,N-Dimethylacetamide (DMAc, semiconductor grade) and ethylmethylcarbonate (EMC, semiconductor grade) were purchased from Wako Pure Chemical Industries, Japan. These reagents were used without further purification. A commercial cellulose nonwoven membrane for a battery separator (Nanobase 2, Mitsubishi Paper Mills, Ltd., Japan) was used as the reference porous membrane for characterization.

Electrospinning

PAI-30k/DMAc (11.25 wt %) and PAI-26k/DMAc (11.25 wt %) solutions and 29 and 30 wt % PAI-60k/DMAc solutions were electrospun using a commercial device (NF-103, MECC, Japan). A stainless steel nozzle (0.3 mm internal diameter, Unicontrols, Japan), connected to a high-voltage regulated DC power supply, was used as the spinneret. The grounded collector used as the counter electrode was an aluminum plate (250 × 200 mm2). The applied voltage was 40 kV, the distance between the nozzle tip and the collector was 135 mm, and the flow rate was 16.6 μL/min for PAIs. All electrospinning procedures were carried out at 28–33 °C and at 20–30% relative humidity. The thicknesses of all the prepared NF membranes are approximately 20 ± 1 μm.

Characterization

The surface morphologies of the prepared NF membranes were observed using a scanning electron microscope (SEM, S-2000, Hitachi High Technologies, Japan) operated at 0.7–5 kV. The average fiber diameter was determined by SEM image analysis. The weight and thickness of the membranes were determined using an analytical balance (HR-202i, A&D, Japan) and a height gauge (digimatic indicator ID-H, Mitutoyo, Japan), respectively. The apparent membrane porosity (ε) was determined as followswhere Wf is the basis weight of membrane, the weight per the unit area (1 × 1 cm2); ρf is fiber density; and Lc is the membrane thickness. The measurements were carried out 3 times for each membrane, and the mean value (±standard deviation) is indicated.

The time courses of the contact angle measurements were also measured using a DropMaster 500 (Kyowa Interface Science Co., Japan) at approximately 25 °C (Figure S1). An organic liquid (EMC) droplet with a volume of 0.2 μL was used as a probe liquid. The impregnation velocity (v) was determined by the measured impregnation time (t) required for the contact angle to reach zero degrees after the deposition of the droplet. For the thin-film NF membranes, at t, the cylindrical liquid permeation (wetting) region is always formed because Lc is much smaller than the apparent flow distance (see Figure S2, more detailed information is given in the Supporting Information). Here, we determined the apparent impregnation velocity based on these experimental results. At t, the apparent permeation volume reaches the droplet volume divided by the apparent membrane porosity (ε). Therefore, the apparent permeation volume can be expressed as (v × t)2 × π × Lc. Consequently, v is given as followswhere the droplet volume is 0.2 μL and the membrane thickness (Lc) is 20 μm. The measurements were carried out 5 times for each membrane, and the mean value (±standard deviation) is indicated. In this study, we used fiber materials with the low intrinsic contact angles of 3.3–6.6° (a detailed description is included in the Supporting Information). Therefore, the chemical effects of fiber materials on the liquid impregnation behavior are negligible. In addition, we used EMC, which is a commonly used solvent for electrolyte of lithium-ion batteries, as a model organic solvent. The surface tension of EMC is 27.1 mN/m at 20 °C[20] and is in the range of representative organic solvents (e.g., ethanol, acetone, cyclohexane, and benzene), which is 20–30 mN/m at room temperature.