Supplementary Figure S1 / DSC charts of a polystyrene (PS) solution during rapid cooling. a, DSC thermogram

obtained from a 20 wt% DMF solution. b, Changes in temperature and cooling rate with time. The cooling rate decreased

with decreasing temperature, but remained above 10 K/min even at −150 °C.

Exo

Endo

Tg

Cooling (quenching)

b a

2

Supplementary Figure S2 / SEM images of nanofiber networks. a, PSF (NF1); b, PS-40 (NF7); c, PC (NF3); d, PES (NF4); e,

PAN (NF5); f, PVC (NF6). Insets show enlarged images.

a

c

e

b

d

f

100 nm

250 nm

100 nm

250 nm

100 nm

250 nm

100 nm

250 nm 250 nm

100 nm

100 nm

250 nm

PSF (NF1)

PC (NF3)

PAN (NF5)

PS-40 (NF7)

PES (NF4)

PVC (NF6)

3

Supplementary Figure S3 / Nanofiber diameter distributions. a, SEM image of PS nanofiber network (PS-20); b−g, diameter

histograms of PSF (NF1), PS-20 (NF2), PC (NF3), PES (NF4), PAN (NF5) and PVC (NF6) nanofibers. The nanofiber diameter

distributions were determined by statistically analyzing more than 1,000 data points from over 50 SEM images for each sample.

a

b

e f

c d

g

PS-20 (NF2)

PSF (NF1)

PES (NF4) PAN (NF5)

PS-20 (NF2) PC (NF3)

PVC (NF6)

4

Supplementary Figure S4 / N2 adsorption/desorption isotherms and pore size distributions. a, PSF (NF1); b, PS40 (NF7); c,

PC (NF3); d, PES (NF4); e, PAN (NF5); f, PVC (NF6). The upper plot for each polymer shows N2 adsorption/desorption isotherms

at 77 K. The lower plot presents the pore size distributions calculated from these isotherms using Barrett–Joyner–Halenda (BJH)

analysis. Red and blue circles indicate data points for adsorption and desorption processes, respectively.

b a c

e d f

PS-40 (NF7) PSF (NF1) PC (NF3)

PAN (NF5) PES (NF4) PVC (NF6)

5

Supplementary Figure S5 / TEM tomography image of a PSF nanofiber network.

Rotational movie is available in mpg format.

Size: 160 × 175 × 65 nm

x

y

z

6

Supplementary Figure S6 / Reversing plots from temperature-modulated DSC. a, PSF (NF1); b, PS-20 (NF2); c, PC

(NF3); d, PES (NF4); e, PVC (NF6). Blue and red curves indicate the first and second scans, respectively. Glass transition

temperatures are indicated by arrows.

a

1st

2nd

Tg

TgL

TgH

TgL

TgH

Tg 1st

2nd

TgL

TgH

Tg 1st

2nd

TgL

TgH

Tg 1st

2nd

TgL

TgH

Tg

1st

2nd

c b

e d

a

PC (NF3) PS-20 (NF2)

PVC (NF6) PES (NF4)

PSF (NF1)

7

Supplementary Figure S7 / Stress-strain curve of a mesoporous PSF sheet. A sheet

sample 10 mm in width, 40 mm in length and 65 μm thick was examined at an initial gap size

of 20 mm.

8

Supplementary Figure S8 / DSC charts of polystyrene (PS) solutions in dimethylformamide (DMF). a, DSC thermograms of

pure DMF and 5–30 wt% PS solutions during rapid cooling; b, DSC thermograms of pure DMF and 30 wt% PS solution at elevated

temperatures (heating rate: 10 K/min). During the rapid cooling process, 5–15 wt% PS solutions exhibited exothermic peaks due to the crystallization of the solvent. The

peak intensity gradually decreased with increasing concentration of PS and the peak apex shifted to lower temperatures. The

exothermic peak disappeared when the PS concentrations were approximately 20 wt% (see Supplementary Figure S1a) and a

baseline shift due to glassification of polystyrene was observed near −135 °C.

During the heating of a rapidly cooled 30 wt% PS solution, cold crystallization of DMF was observed within the range of −85 to

−65 °C. This range was about 30 °C higher than the exothermic peak seen for the 20 wt% PS solution at −105 °C (see Figure 1c of

our paper). These results indicate that cold crystallization of DMF is inhibited when the PS concentration is high.

b a

9

Supplementary Figure S9 / Pore size distributions of polyetherimide (PEI) nanofiber

network. Red and blue plots were obtained from N2 adsorption and desorption isotherms,

respectively.

PEI nanofiber network was prepared from the 1:1 (w/w) mixture of dimethylsulfoxide (DMSO) and

dimethylacetamide (DMAc). PEI was obtained from PolyScience (Tg: 217 °C, Mw: 30,000). A 20

wt% PEI solution was spun into cold methanol at −100 °C and solvent extraction was followed in a

manner described in Supplementary Methods. The resultant nanofiber network had extremely

small mesopores of 1.5−4 nm and macropores of about 30 nm in radius.

10

Supplementary Figure S10 / Correction effects in high-pressure CO2 sorption experiments. The amount of CO2 sorption was

corrected by considering four factors associated with high pressure conditions. The raw data (P0) was obtained for a PSF nanofiber

network at 318 K. Deviation from an ideal gas was significant at pressures above 10 atm. Since non-ideal behavior of CO2 gas was observed, particularly at high pressures, the following four correction factors were

considered for the sorption experiments. The curve P0 is a plot of the raw sorption data without any corrections, in which the sorption

amount appears to decrease with increasing gas pressures above 10 atm. Using the virial coefficient of CO2 at 297 K obtained from

the NIST Chemistry Webbook24

, the curve P0 was corrected to curve P1. However, in the present experiments, CO2 gas introduced to

the sample cell was heated to the measurement temperature of 318 K and thus, when the virial coefficient at 318 K was separately

used for the CO2 in the sample cell, the curve P1 was corrected to curve P2. Furthermore, curve P3 was obtained after subtracting

the volume of the polymer nanofiber network from the volume of the empty cell. The nearly ideal curve P4 was finally obtained after

baseline correction of the empty cell for CO2 sorption. These corrections significantly improved the accuracy of the CO2 sorption

experiments, especially at high pressures above 10 atm.

P0

P1

P2

P3

P4

Correction Factors P0 P1 P2 P3 P4

Virial coefficient at room

temperature – + + + +

Virial coefficient at each

sorption temperature – – + + +

Sample density – – – + +

Sorption amount for empty

vessel – – – – +

–: not corrected, +: corrected.

11

Supplementary Figure S11 / CO2 sorption properties of nanofiber network. a, CO2 sorption isotherms of PSF

nanofiber network at different temperatures; b, CO2 sorption/desorption kinetics monitored by the gravimetric

method.

273 K 263 K 283 K

293 K

303 K

313 K

a

Vacuum

1st 2nd 3rd

CO2 10 atm b

12

Supplementary Figure S12 / Condensation of organic vapor. a, Vapor sorption of PSF nanofiber

network at relative pressures (p/p0) of 0.3, 0.6, and 0.9 at 298 K. Contact angle (CA) on a bulk PSF

film is shown beside the histogram of each solvent; b, Sorption isotherms of methanol vapor at

different temperatures. Mesoporous polymer: PSF (NF1 in Supplementary Table S2). Due to the large mesopore volumes and narrow pore-size distributions, the nanofiber networks can

capture a large amount of organic solvents from the vapor phase. PSF nanofibers, however, have

hydrophobic surfaces and therefore less than 1 mmol/g of water vapor was adsorbed, even at high

partial pressures. In sharp contrast, the sorption of methanol and hexane increased rapidly at

relative pressures (p/p0) above 0.6, indicating capillary condensation in the mesopores. In five

sorption isotherms of methanol, large variations in the saturation vapor pressure (5.5–44 kPa) are

observed over the temperature range 278–318 K. These data indicate that the nanofiber network

can capture a large amount of methanol vapor at low temperatures and will subsequently release the

captured methanol in response to mild heating.

278 K 288 K

298 K

308 K

318 K

Vtotal: 0.89 cm3/g

a

b

13

Supplementary Figure S13 / Sorption isotherm of tetrahydrofuran from aqueous solution.

Blue and red curves were obtained at 20 °C and 40 °C, respectively. Mesoporous polymer: PS

(NF7 in Supplementary Table S2).

14

Supplementary Figure S14 / Adsorption isotherms of m-cresol dissolved in pure water. Blue:

adsorption isotherms on commercially-available activated carbon (provided from a Japanese

leading company), red: commercially-available polymer adsorbent (XAD4, DOW Chemical). Closed

and open symbols represent the adsorption isotherms at 20˚C and 80˚C, respectively.

15

Supplementary Table S1. Basic information for polymers used in this study.

Polymer Abbreviation CAS

number Supplier Mw

ρ

(g/cm3)

Tg (°C) Tm (°C)

Polystyrene PS 9003-53-6 Wako Chemical Inc. 200,000 1.05 100 N.D.

Poly(bisphenol-A-carbonate) PC 103598-77-2 Scientific Polymer Inc. 60,000 1.20 149 N.D.

Polysulfone PSF 25154-01-2 Scientific Polymer Inc. 60,000 1.24 190 230

Poly(p-phenylene ether-sulfone) PES 25608-63-3 Scientific Polymer Inc. N.A. 1.37 N.A. N.A.

Polyacrylonitrile PAN 25014-41-9 Scientific Polymer Inc. 150,000 1.18 125 314

Poly(vinyl chloride) PVC 9002-86-2 Scientific Polymer Inc. 275,000 1.40 85 285

Mw: weight-averaged molecular weight, ρ: density, Tg: glass transition temperature, Tm: melting point. These values are from suppliers’ catalogs.

16

Supplementary Table S2. Fabrication parameters for mesoporous polymer nanofiber networks.

Sample ID Polymer Concentration Good solvent Freezing temperature

NF1 PSF 20 wt% DMF −196 °C

NF2 PS 20 wt% DMF −196 °C

NF3 PC 20 wt% Cyclohexanone −196 °C

NF4 PES 20 wt% DMF −196 °C

NF5 PAN 10 wt% DMF/DMAc (2/1) −196 °C

NF6 PVC 10 wt% DMF/DMAc (2/1) −196 °C

NF7 PS 40 wt% DMF −196 °C

DMF: dimethylformamide, DMAc: dimethylacetamide.

17

Supplementary Table S3. Thermal properties of nanofiber networks.

ID Polymer First scan

TgL(°C) Tg

H(°C)

Second scan

Tg (°C)

NF1 PSF 161.3±13.9 184.0±8.8 187.2±5.9

NF2 PS 87.3±4.9 106.1±5.2 105.3±5.1

NF3 PC 123.6±14.6 152.6±5.7 151.3±5.4

NF4 PES 189.9±7.4 223.8±9.7 225.1±5.6

NF6 PVC 58.7±8.3 87.0±7.7 87.4±7.2

TgL and Tg

H are the two glass transition temperatures observed at low and high temperatures

during the first heating operation, respectively. The error range shows the width of the transition

range. Tg is the glass transition temperature observed in the second heating run.

18

Supplementary Table S4. Water fluxes and separation properties of mesoporous PSF sheets.

ID Polymer solution

(wt% in NMP)

Spin-coating

condition

Freezing

temperature

Thickness

(μm)

Flux

(L/m2h)

Rejection

(%)

PSF-1 20 2000 rpm −100 °C 25±1 92 93

PSF-2 20 3000 rpm −100 °C 18±1 62 92

PSF-3 25 2000 rpm −100 °C 38±2 11 96

PSF-4 25 3000 rpm −100 °C 26±1 23 95

NMP: N-methyl-2-pyrrolidone. Flux was measured under a reduced pressure of 80 kPa. Rejection was calculated from the

decrease in the absorbance of 5 nm Au nanoparticles between the feed and the permeate.

19

Supplementary Table S5. Dual sorption model parameters evaluated for CO2 sorption of PSF nanofiber networks.

T

(K)

kD

(cm3STP/cm

3·atm)

CH

(cm3STP/cm

3)

b

(atm–1

)

C

(cm3STP/cm

3)

CD/C

10 atm

CD/C

30 atm

263 6.34 27.6 1.53 89.3 0.71 0.88

268 5.41 26.7 1.34 78.9 0.69 0.86

273 4.59 25.8 1.12 69.6 0.66 0.85

278 3.85 25.7 0.90 61.6 0.62 0.82

283 3.51 25.4 0.72 57.4 0.61 0.81

288 2.80 28.4 0.52 51.8 0.54 0.76

293 2.54 27.1 0.46 47.7 0.53 0.75

298 2.04 30.0 0.32 43.2 0.47 0.69

303 1.63 31.3 0.26 39.1 0.42 0.64

308 1.53 28.2 0.26 35.6 0.43 0.65

313 1.31 27.2 0.24 32.2 0.41 0.62

318 1.07 29.2 0.19 29.7 0.36 0.56

kD: Henry’s dissolution parameter, CH: Langmuir’s capacity parameter, b: Langmuir’s affinity parameter, C: total sorption amount of CO2 at

10 atm, CD/C: contribution of Henry-type sorption to total CO2 sorption at 10 and 30 atm.

20

Supplementary Table S6. Adsorption equilibrium of m-cresol from aqueous solutions.

Polymer Temp. Initial conc. (wt%)

ce (wt%) / qe (mg/g)

PS

20 °C 0.51 0.76 1.03 1.54 2.04

0.392 26.0 0.575 40.2 0.719 68.4 0.978 126 1.298 164

80 °C 0.52 0.75 1.00 1.53 2.00

0.440 18.0 0.664 18.9 0.901 22.9 1.333 44.6 1.789 45.7

PSF

20 °C 0.51 0.76 1.03 1.54 2.04

0.164 76.7 0.269 108 0.397 140 0.645 200 0.951 241

80 °C 0.52 0.75 1.00 1.53 2.00

0.266 56.7 0.452 66.0 0.621 85.1 1.022 114 1.352 143

PES

20 °C 0.51 0.76 1.03 1.54 2.04

0.140 82.0 0.246 113 0.353 150 0.590 212 0.858 262

80 °C 0.52 0.75 1.00 1.53 2.00

0.236 63.3 0.399 77.8 0.562 98.2 0.929 134 1.237 168

Ce: equilibrium concentration, qe: equilibrium adsorption capacity.

21

Supplementary Note 1

Production costs of mesoporous polymer nanofiber networks

As of 2013, the market prices for polystyrene (PS) and polycarbonate (PC) are approximately 2.0 and 3.5 $US/kg,

respectively, while those for polysulfone (PSF) and poly(p-phenylene ether sulfone) (PES) (engineering plastics with

high Tg) are relatively high at about 20 and 25 $US/kg. The raw material cost of the polymer, however, will be very

low as compared with that of the separation unit. In addition, the solvents used for the production of mesoporous

polymer nanofiber networks can be repeatedly recycled. The major cost associated with the production of these

materials will be the electricity required for cooling, although this will not be an obstacle in terms of the economics

of mass-production.

Macroscopic fibers of mesoporous polymer nanofiber networks were prepared by directly injecting the polymer

solution into excess methanol precooled to −100 °C, using a metal syringe with a needle with a 0.5 mm inner

diameter. The resultant fibers had diameters in the range of 0.5–1.5 mm, depending on the viscosity of the polymer

solution. A video of the continuous production of macroscopic fibers of mesoporous polymer nanofiber network is

available in mpg format (Supplementary Movie 2). These fibers were cut into pellets 4 mm in length using a

mechanical crusher.

Supplementary Methods

Material information and preparation procedures

Supplementary Table S1 provides basic information (supplier, molecular weight, density, glass transition

temperature, melting point, etc.) for the polymers used in this study. Mesoporous polymer nanofiber networks were

fabricated by the following four-step process. (I) A concentrated polymer solution (20 wt%) was prepared by

dissolving polymer pellets (2.0 g) in a good solvent (8.0 g) via overnight magnetic stirring. In cases where

incomplete dissolution was evident at room temperature, the solution was gradually heated to 120 °C. (II) The

polymer solution was quickly frozen by immersing it in liquid nitrogen (or immersing in excess methanol precooled

to −100 °C) and subsequently aged for one to several days at a temperature below the melting point of the solvent.

(III) Solvent extraction was performed by immersing the frozen polymer solution in excess methanol (more than 10

times the volume of the frozen solution) precooled to −80 °C. The extraction was allowed to proceed for one day

(sheet and fiber forms) to over five days (disk-shaped sheet form) in a cryogenic freezer (−80 °C). (IV) Residual

22

methanol in the solidified sample was exchanged with tert-butanol at room temperature and the sample was then

freeze dried. Residual solvent was completely removed by subsequent vacuum drying at temperatures up to 50 °C for

one to five days. We confirmed that the thermal and physicochemical properties of the resulting mesoporous polymer

nanofiber networks were unchanged after one day of aging. The freeze drying in procedure IV was not essential for

the formation of the mesoporous structure, but was performed so as to obtain a high level of reproducibility by

minimizing undesirable shrinkage of small pores during drying. Suitable solvents and optimal polymer

concentrations differed for different polymers. Supplementary Table S2 summarizes the typical fabrication

conditions. The bulk density (or apparent density) of each specimen was determined from the weight and geometric

volume of disk-shaped nanofiber sheet. A digital caliper was used to determine the thickness and diameter of the

sheet.

Polymer solution freezing process

The freezing and melting behaviors of each polymer solution were characterized by DSC (TA Instruments,

Q2000). Supplementary Figure S1a shows a DSC curve obtained during rapid cooling. The heat flow was recorded

for a 20 wt% polystyrene (PS) solution in dimethylformamide (DMF). A baseline shift is clearly seen near −135 °C

and there is no exothermic peak down to −150 °C, demonstrating that the polymer solution was frozen in a glassy

state without solvent crystallization31

. The solidified solution was also observed to remain transparent. These

characteristics are very different from those of a slowly frozen solution in which both liquid–liquid phase separation

of the polymer solution and/or solvent crystallization occur with decreasing temperature32–34

. Supplementary Figure

S1b shows changes in the temperature (blue line) and the cooling rate (red line) during rapid cooling. The cooling

rate was initially greater than 70 K/min, but gradually reduced with decreasing temperature. The rate was higher than

10 K/min even at −150 °C.

SEM observations

Since polymer nanofiber networks are damaged by external mechanical stress, the samples should not be cut with

a razor blade, torn at room temperature nor freeze fractured without first filling the mesopores with a poor solvent.

Cutting with a razor blade destroys the nanofiber morphology, while the other two procedures do not preserve the

original (as-prepared) nanofiber structure. In fact, nanofibers were often elongated into fibril structures in the torn or

fractured direction, similar to the well-known crazing phenomenon in polymers35

. Supplementary Figure S2 shows

23

SEM images of nanofiber networks prepared from various polymers. All the images were obtained at a magnification

of 50,000× (the insets were obtained at a magnification of 300,000×). The same morphology was observed in all

parts of the sample, as confirmed by low-magnification observations (1,000×). The uniform mesoporous structure

with an absence of macrovoids is very different from the structure of porous polymer sheets prepared by

conventional phase inversion36,37

.

Analysis of nanofiber diameters

The SEM images were analyzed to determine the nanofiber diameters. As shown in Supplementary Figure S3a,

the middle points between the nodes of the nanofiber network were selected in an SEM image (obtained at a

magnification of 300,000×) as a means of calculating the diameter. Over 1,000 data points were selected from more

than 50 images. The diameter was determined using a scaling function of the instrument software. The resulting

histograms and Gaussian fitting curves are also shown in Supplementary Figure S3. The thickness of the platinum

layer was not considered in these calculations, and the geometric specific surface area, Scalc, was calculated from the

actual polymer density (ρ), the nanofiber diameter (d) and the probability function of the diameter histogram, f (d), as

Scalc = ∑f×4/(ρ×d)33,38

.

Analysis of pore size distributions

The pore size distributions of polymer nanofiber networks were analyzed using a volumetric gas adsorption

apparatus (BEL Japan, Belsorp-max). A glass sample tube was filled with small pieces of dried sample (40–70 mg

for N2 adsorption) and thoroughly dried overnight under vacuum at 70 °C (except for PVC, which was dried at

50 °C). All measurements were performed after conducting a 3 min leak check to ensure sufficient sealing and

complete drying of the specimens. Brunauer–Emmett–Teller (BET) and Barrett–Joyner–Halenda (BJH) analyses

were systematically carried out using commercially available software (BEL Japan, BEL MasterTM

). The saturated

vapor pressure of liquid nitrogen, p0, was experimentally corrected for in-situ monitoring.

Nitrogen adsorption isotherms obtained at 77 K are shown in Supplementary Figure S4. In all the isotherms, the

amount of N2 adsorbed increased drastically at a relative pressure (p/p0) of about 0.8, reaching values as high as

400–700 cm3STP/g at p/p0 = 0.99. The units of measurements used in these isotherms (cm

3STP/g) describe the

volume of adsorbed gas molecules at standard temperature and pressure per gram of sample. Some hysteresis was

observed between adsorption and desorption processes, which is often the case with capillary condensation in

24

mesoporous materials39

. The pore size distributions of the nanofiber networks are shown below the corresponding

nitrogen adsorption isotherms. These pore size distributions were quantitatively obtained from BJH analyses of the

higher pressure regions (p/p0 > 0.3) of the adsorption/desorption isotherms using the software associated with the

instrumentation (BEL Japan, BEL MasterTM

) and the relevant parameters for N2 (molecular weight: 28.013,

cross-sectional area: 0.1620 nm2)

40. The average and standard deviation of each pore radius were determined by

fitting a Gaussian curve to the pore size distribution obtained separately for the adsorption and desorption processes.

The relative pressures of 0.385, 0.963 and 0.990 correspond to pore radii of 1.2, 25.8 and 98.9 nm, respectively.

Under the present experimental conditions, N2 cannot fill macropores that have radii larger than 100 nm39

. Based on

the IUPAC definition, the pore volumes of mesopores (1–25 nm in radius) and macropores (25–100 nm in radius) are

listed separately in Table 1 of our paper.

The actual polymer density values in Table 1 of our paper were calculated from the weight and volume of

samples after excluding the void volume, as determined using a helium pycnometer (BEL Japan, Belsorp-max). To

ensure the accuracy of the volume estimation, the same experiment was repeated at least 10 times, and the averaged

value was used to calculate the actual polymer density.

TEM tomography

A polymerizable resin containing 5 nm (in diameter) gold nanoparticles was embedded in the mesopores of a

nanofiber network sample (NF1 in Supplementary Table S2). These nanoparticles functioned as markers for digital

reconstruction of a tomographic image. The embedded resin was further stained with osmium and ultrathin slices of

the resulting specimen were prepared by frozen sectioning using an ultramicrotome and subsequently used for

transmission electron microscopy (TEM; JEOL, JEM2100F) observations41

. The images were obtained at an

acceleration voltage of 200 kV and a magnification of 400,000× and the tile angle was varied over the range of ±62°.

Images were recorded at every degree using commercial software (Gatan, DigitalMicrographTM

) at a resolution of

0.268 nm/pixel and were then digitally compiled into a tomographic image. As shown in Supplementary Figure S5,

the negatively stained sample gave a clear tomographic image. The black region indicates the presence of

interconnected pores of 10–20 nm in the nanofiber network. It is clear that the PSF nanofibers (white regions) have a

three-dimensional network structure. The rotational movie is available in mpg format (Supplementary Movie 1).

25

Analysis of glass transition temperatures

Supplementary Figure S6 presents temperature-modulated differential scanning calorimetry (MDSC) results for

the polymer nanofiber networks made of polysulfone (PSF), polystyrene (PS), poly(bisphenol-A-carbonate) (PC),

poly(p-phenylene ether-sulfone) (PES) and poly(vinyl chloride) (PVC). The measurements were performed using

DSC (TA Instruments, Q2000) with a temperature amplitude of ±1 K and a 60 s cycle at a constant heating rate of 2

K/min. In the case of the PSF nanofiber network, the first heating scan showed a minor change in specific heat

capacity at 161 °C (TgL), followed by a major glass transition at 184 °C (Tg

H). In contrast, the second heating scan

exhibited a single glass transition temperature (Tg) at 187 °C. The values of TgH and Tg obtained in this manner were

almost the same as the glass transition temperature reported for bulk PSF42

. All other polymers shown in

Supplementary Figure S6 had TgL and Tg

H in the first scan, and the latter values were almost consistent with the

glass transition temperatures of bulk samples. The nanofiber networks made of polyacrylonitrile (PAN) were an

exception, since they could not be characterized due to self-cross-linking during heating. These data are summarized

in Supplementary Table S3.

The weak thermal transition (TgL) 20 to 30 °C lower than the major glass transition always disappeared when the

nanofiber structure was destroyed by heating to 250 °C. Therefore, the thermal transition at the lower temperature

might be related to the nanofiber structure. The mesopores were usually collapsed near bulk Tg (or TgH). Judging

from these observations, TgL might be explained as a surface effect of polymer nanofibers. When the polymer

contains 5 wt% of good solvent, the glass transition temperature, in general, decreases about 20 °C. The thermal

stability of the mesopores of polymer nanofiber networks also decreases with increasing amount of adsorbed organic

solvents. This should be considered when they are applied as oil adsorbents. In our MDSC measurements, all the

specimens were dried for one to five days at 50 °C in vacuum in order to minimize the plasticization effect caused by

the residual solvent.

Mechanical properties of a mesoporous PSF sheet

Supplementary Figure S7 shows the stress-strain curve of the mesoporous PSF sheet shown in Figure 3a of our

paper. The measurements were conducted using a TA Instruments RSA-G2. The sheet was prepared according to the

method described in this Supplementary Methods. The Young’s modulus of the material, as averaged from four

measurements, was 0.23±0.02 GPa and the average maximum stress at break and the elongation were 7.4±0.6 MPa

and 26.5±9.0 %, respectively. It is known that bulk PSF typically has an elastic modulus of 2.5–2.6 GPa, a stress at

26

break of about 70 MPa and an elongation at break of about 50–100 %. The elastic modulus of the mesoporous PSF

sheet was therefore about 10 times smaller than the value of non-porous PSF.

Water permeance through a mesoporous PSF sheet

Supplementary Table S4 summarizes water flux through a mesoporous PSF sheet with different thicknesses

prepared according to the method described in this Supplementary Methods. In spite of their significant thicknesses

(of a few tens of micrometers), the sheets allowed high levels of water flux (11−92 L/m2h) under a reduced pressure

of 80 kPa. It is therefore apparent that mesopores penetrate from one side of the sheet to the other side. These sheets

also showed rejection values of 92−96 % for 5 nm Au nanoparticles.

CO2 sorption experiments

CO2 adsorption measurements were conducted up to 30 atm using a volumetric gas sorption system (BEL Japan,

Belsorp-HP). A standard cell with a known dead volume (measured at ambient temperature) was maintained at a

temperature of 24.00±0.01 °C. The ambient temperature was also maintained at 24 °C to prevent temperature

variations during experiments. A vacuum-dried sample (0.4–0.7 g) was sealed in a 9.1 cm3 stainless-steel sample

tube. Prior to each measurement, gas leakage was confirmed to be less than 1 Pa/min, which ensured complete

sealing and dryness of the sample. The CO2 sorption temperature was carefully controlled in the range of 263–318 K

with an accuracy of 0.1 °C using a circulating temperature bath (Taitec, Coolnit CL-80R) equipped with an external

temperature sensor. The PID coefficients were optimized at each temperature. More than 25 data points were

recorded over the pressure range of 1–30 atm and the amount of CO2 sorption was determined as the equilibrium

value after a pressure variation of less than 0.03 atm (0.1% of the full scale of the pressure gauge) was recorded over

600 s.

Supplementary Figure S11a shows CO2 sorption isotherms of the PSF nanofiber network over the temperature

range 263–318K. The amount of adsorbed CO2 increased with increasing partial pressure and decreasing sorption

temperature and CO2 sorption at 263 K reached 89.3 cm3STP/cm

3 (14.0 wt%) at 10 atm. The sorption isotherms were

analyzed using the DMS model43

. This model describes gas sorption as the sum of Henry and Langmuir sorption

terms:

bp

bpCpkC H

D

1

, (S1)

27

where C is the amount adsorbed, kD is the Henry’s law constant, CH is the Langmuir capacity and b is the Langmuir

affinity constant. These parameters were derived by curve fitting the sorption isotherms. Supplementary Table S5

lists the values of each parameter as well as the total sorption amount of CO2 at 10 atm (C) and the contribution of

Henry sorption (CD/C) at 10 and 30 atm. The CH values were essentially constant (28±3 cm3STP/cm

3) while the kD

values increased significantly from 1.07 cm3STP/cm

3·atm (318 K) to 6.34 cm

3STP/cm

3·atm (263 K). The Henry’s

law constant (kD) at 293 K was about three times larger than that of bulk PSF films27,44

. Interestingly, the value at 298

K was very similar to that of liquid toluene (2.1 cm3STP/cm

3·atm)

45.

In general, glassy polymers have a limited gas sorption capacity due to their internal free volume. The PSF

nanofiber network, however, demonstrated a much higher gas sorption capacity. When the free volume of bulk PSF

(0.116 cm3/g)

46 is completely occupied with liquid-like CO2, the adsorbed amount is calculated to be 67 cm

3STP/cm

3.

The PSF nanofibers should have almost the same free volume, since the actual polymer density of the nanofibers

(1.25 g/cm3) is very similar to that of bulk PSF (1.24 g/cm

3). Nevertheless, the sorption amount of PSF nanofibers is

significantly in excess of 100 cm3STP/cm

3 under conditions of low temperature and high pressure. Judging from

these results, the nanofibers appear to be swollen by contact with CO2 at increasingly high gas pressures. Swelling or

plasticization phenomena are frequently observed for glassy polymer/CO2 systems28

.

Supplementary Figure S11b shows CO2 sorption/desorption kinetics of a PSF nanofiber network. These

measurements were conducted using an in-house gravimetric gas sorption apparatus equipped with a pneumatic

pressure controller (GE Sensing, Pace5000-7MPa module). CO2 was introduced at a pressure of 10 atm for 5 min

with a high-pressure controller (GE Sensing, The PACE5000) and then evacuated using a rotary-vacuum pump

(Ulvac Kiko, GLD-051). The sorption and desorption behaviors are shown with red and blue curves. Within 1 min,

the nanofibers adsorbed up to 90 % of the saturation level of CO2 and the adsorbed gas was rapidly and completely

desorbed by vacuum evacuation. The amount adsorbed at 10 atm was 9.6 wt% at 298 K, which is six times higher

than the N2 adsorption.

Vapor sorption experiments

Vapor sorption experiments were performed using a volumetric gas adsorption apparatus (BEL Japan,

Belsorp-max). A glass sample tube was filled with small pieces of dried sample (50–100 mg) and thoroughly dried

under vacuum at 70 °C overnight. All measurements were performed after conducting a 3 min leak check to ensure

sufficient sealing and complete drying of the specimens. Vapor sources (such as methanol, hexane and water) were

28

completely degassed by at least three cycles of freezing and evacuating. All the transfer tubes of the gas adsorption

apparatus were heated to at least 15 K above the sorption temperature. The sorption temperature was carefully

controlled in the range of 278–318 K with an accuracy of 0.1 °C using a circulating temperature bath (Taitec, Coolnit

CL-80R) equipped with an external temperature sensor. The PID coefficients were also optimized at each

temperature. The saturation vapor pressures at each sorption temperature were calculated from the Antoine equation

using the coefficients compiled in the NIST Chemistry Webbook24

.

Removal of organic solvents in water

The polymer nanofiber networks can effectively and rapidly adsorb organic solvents in water, due to their

continuous mesoporous structure and very high free surface area. We confirmed that these materials are able to

remove toluene from aqueous solutions containing concentrations of a few tens of ppm. Aromatic alcohols such as

m-cresol were also effectively removed from aqueous solutions with concentrations of 1–2 wt%. The nanofiber

networks can quantitatively release these adsorbed organic solvents on subsequent treatment with methanol. During

our experimental trials, these adsorption/desorption procedures did not inflict any damage on the nanofiber network

structures. Supplementary Figure S13 shows sorption isotherms of aqueous tetrahydrofuran (THF) at 20 and 40 °C.

These isotherms were obtained using 2 g of PS nanofibers (NF7 in Supplementary Table S2) to remove THF in the

concentration range of 1.0–10.0 wt%. During these trials, the amounts of equilibrium adsorption at 20 °C were

higher than those at 40 °C. The adsorbed amount also increased significantly when the liquid phase concentration

approached 5 wt%, indicating that the polymer nanofibers were swollen by the THF.

Adsorption isotherms of m-cresol on PES and PSF nanofibers are presented in Figure 3e of our paper. Those

equilibrium data obtained are summarized in Supplementary Table S6 with the data obtained for a PS nanofiber

network. The adsorption amount (qe, in mg/g) was calculated using the equation described in literature47

. Compared

to the PS nanofiber network, networks of both PSF and PES strongly adsorbed m-cresol and the measured adsorption

amounts were routinely higher at 20 °C than at 80 °C. The difference in the adsorption amount between the two

temperatures was 130 mg/g when the m-cresol concentration in water was 1.0 wt%. When three leading

commercially-available activated carbons were used for comparison experiments, the differences in adsorption

amounts between 20 °C and 80 °C were about 50 mg/g at a m-cresol concentration of 1.0 wt%, as shown in

Supplementary Figure S14. The values of three leading commercially-available cross-linked polymer adsorbents

were about 60 mg/g under the same experimental conditions. We also observed that the adsorption rate of m-cresol

29

on mesoporous polymer nanofiber networks is very fast, as compared with the rates obtained using

commercially-available cross-linked polymer adsorbents.

30

Supplementary References

31. Francois, J., Gan, J. Y. S. & Guenet, J. –M. Sol-gel transition and phase diagram of the system atactic

polystyrene-carbon disulfide. Macromolecules 19, 2755–2760 (1986).

32. Tsai, F. –J. & Torkelson, J. M. Roles of phase separation mechanism and coarsening in the formation of

ploy(methyl methacrylate) asymmetric membranes. Macromolecules 23, 775–784 (1990).

33. Aubert, J. H. Structural coarsening of demixed polymer solutions. Macromolecules 23, 1446–1452 (1990).

34. Song, S. –W. & Torkelson, J. M. Coarsening effects on microstructure formation in isopycnic polymer solutions

and membranes produced via thermally induced phase separation. Macromolecules 27, 6389–6397 (1994).

35. Horiuchi, S., Nakagawa, A., Liao, Y. & Ougizawa, T. Interfacial entanglements between glassy polymers

investigated by nanofractography with high-resolution scanning electron microscopy. Macromolecules 41,

8063–8071 (2008).

36. Strathmann, H. & Kock, K. The formation mechanism of phase inversion membranes. Desalination 21, 241–255

(1977).

37. Smolders, C. A., Reuvers, A. J., Boom, R. M. & Wienk, I. M. Microstructures in phase-inversion membranes.

Part 1. Formation of macrovoids. J. Membr. Sci. 73, 259–275 (1992).

38. Ma, P. X. & Zhang, R. Synthetic nano-scale fibrous extracellular matrix. J. Biomed. Mater. Res. 46, 60–72

(1999).

39. Kaneko, K. Determination of pore size and pore size distribution 1. Adsorbents and catalysts. J. Membr. Sci. 96,

59–89 (1994).

40. Barrett, E. P., Joyner, L. G. & Halenda, P. P. The determination of pore volume and area distributions in porous

substances. I. Computations from nitrogen isotherms. J. Am. Chem. Soc. 73, 373–380 (1951).

41. Radzilowski, L. H., Carragher, B. O. & Stupp, S. I. Three-dimensional self-assembly of rodcoil copolymer

nanostructures. Macromolecules 30, 2110–2119 (1997).

42. Matsuoka, S. Thermodynamic theory of viscoelasticity. J. Thermal Anal. 46, 985–1010 (1996).

43. Vieth, W. R., Howell, J. M. & Hsieh, J. H. Dual sorption theory. J. Membr. Sci. 1, 177–220 (1976).

44. Maeda, Y. & Paul, D. R. Effect of antiplasticization on gas sorption and transport. I. Polysulfone. J. Polym. Sci.,

Part B: Polym. Phys. 25, 957–980 (1987).

45. Wilhelm, E. & Battino, R. Thermodynamic functions of the solubilities of gases in liquids at 25 °C. Chem. Rev.

73, 1–9 (1973).

46. Huang, Y., Wang, X. & Paul, D. R. Physical aging of thin glassy polymer films: free volume interpretation. J.

Membr. Sci. 277, 219–229 (2006).

47. Strivastava, V. C., Swamy, M. M., Mall, I. D., Prasad, B. & Mishra, I. M. Adsorptive removal of phenol by

bagasse fly ash and activated carbon: Equilibrium, kinetics and thermodynamics. Colloids Surf. A 272, 89–104

(2006).