Author's Accepted Manuscript

Permeability thickness dependence of poly-

dimethylsiloxane (PDMS) Membranes
G. Firpo, E. Angeli, L. Repetto, U. Valbusa

www.elsevier.com/locate/memsci

PII: S0376-7388(14)00946-6
DOI: http://dx.doi.org/10.1016/j.memsci.2014.12.043
Reference: MEMSCI13390

To appear in: Journal of Membrane Science

Received date: 19 September 2014

Revised date: 6 November 2014
Accepted date: 17 December 2014

Cite this article as: G. Firpo, E. Angeli, L. Repetto, U. Valbusa, Permeability

thickness dependence of polydimethylsiloxane (PDMS) Membranes, Journal of
Membrane Science, http://dx.doi.org/10.1016/j.memsci.2014.12.043

This is a PDF file of an unedited manuscript that has been accepted for
publication. As a service to our customers we are providing this early version of
the manuscript. The manuscript will undergo copyediting, typesetting, and
review of the resulting galley proof before it is published in its final citable form.
Please note that during the production process errors may be discovered which
could affect the content, and all legal disclaimers that apply to the journal
pertain.
 Permeability Thickness Dependence of PolyDiMethylSiloxane (PDMS)
Membranes
G. Firpo, E. Angeli, L. Repetto, U. Valbusa

Nanomed Labs, Physics Department, University of Genova

Via Dodecaneso, 33 16146 Genova I

Corresponding author G. Firpo. E-mail: giuseppe.firpo@unige.it

Abstract

The paper describes a reliable technique to prepare PDMS membrane used to determine the permeability to

CO2 and He. SEM controls morphology of surface membrane and SEM/FIB measures the membrane

thickness. The data show that permeability becomes thickness dependent below some tens of micrometers.

The Islam’s model, based on non-equilibrium sorption-desorption process at the interface, fits quite well the

experimental data. The comparison shows that flux models based on interface reaction could apply to

describe quantitatively the thickness-dependence of the permeability. The model allows to determine the

surface reactions rate constants for CO2 and He on PDMS. The introduction of a specific characteristic

thickness Lc provides a functional form, which describes very accurately the results of the present paper and

other experiments on polymers. By reducing the membrane thickness below 200 nm, both permeability and

selectivity change, indicating Knudsen diffusion as flux mechanisms responsible for permeation. In this

range of thickness, SEM images reveal membrane defects.

Keywords: PDMS permeability, membrane thickness, sorption rate, ultra-thin membrane.

1. Introduction

Permselectivity polymer membrane technology is a dynamic and rapidly growing field. Among the main

areas of application [1, 2, 3, 4] there are air separation for gas pure production, hydrogen separation in

cryogenic and petrochemical applications, CO2 capture from coal-fired power plant emissions and removal

of acid gas from natural gas. Polymer membranes are commonly used in different types of sensors for biogas

monitoring [5], CO2 detection [6], environmental control, pressure detection [7, 8] and, because of their

biocompatibility, as substrates for cell adhesion in tissue engineering, drug delivery and biochemical sensing.

For these specific purposes, ultra-thin hyperelastic and selectively permeable membranes [9] are necessary.

1
In vacuum technology, for instance, thin polimeric membranes are used in micro pumps [10, 11] and also in

gas leakage measurements [7]. Smart single-chip gas sensor microsystems, based on sensitive polymeric

layers are used for detecting airborne volatile organic compounds [12, 13]. Very thin (600 nm)

nanotube/polymer membranes [14] are proposed for high flux gas transport.

In the last decade, several papers as reference [15] described procedures to fabricate freestanding polymer

membranes. The references [16, 17, 18] estensively investigated electromechanical as well as mechanical

and structural properties as function of thickness. The references [19, 20] studied gas transport

characteristics of thin polymeric membranes, in different conditions of membrane preparation like nano

particles mixing [21], in mixed-matrix membranes [22] and by plasma treatments [23, 24]. Dependence of

transport properties on working temperature, pressure [25], and physical aging [26] was also investigated.

However, in spite of fact that very thin polymeric membranes are employed in several miniaturizated

devices, the dependence of the measured permeability from membrane thickness L is little known. Studies on

vulcanized rubber [27] and glassy polymers [28] concluded that it is independent on L, others concerning

Oxygen transport in Kapton polyimide [29] and, more recently, in nano-constrained glassy polymers [30],

showed instead dependence from L. To explain the behavior ref. [28, 29, 30] proposed a morphological

change of polymer membrane with thickness accountable for fractional free volume alterations, which

modifies gas transport mechanism. An alternative explanation was proposed by Islam and Buschatz [31, 32],

who considered non-equilibrium reactions at the interface instead of morphological changes. Under this

hypothesis they noted that, for thickness such that the reactions at the interface can’t be more considered

‘near’ equilibrium, the measured permeability becomes an “apparent” permeability. This quantity deviates

from the “true” permeability that is, instead, a material property independent of thickness, and is generally

defined as product of solubility and diffusivity. The model showed good agreement with experiment and

allowed to evaluate surface reaction rates which are generally not available in literature. Their individual

knowledge, important in determining the rate of capture from the surface [33], does not require sorption

experiments to determine the equilibrium constant. The question if the model is still valid on smaller scales

where the membrane permeability seems to be more dependent from morphology than from gas transport

mechanism is still matter of investigation [9].

2
This paper studies the permeability of PDMS membrane to CO2 and He for the thickness smaller than 55 µm.

The reasons of the choice of PDMS are in its wide use: from gas separation membrane field, to sensors

technology, and, because its biocompatibility, in life science [7, 8, 9, 15]. CO2 is extensively studied in gas

capture and storage technology because its involvement in environmental science [2, 5, 6]. He, mostly

employed in leak detection, is inert and much lighter than CO2: a comparison with CO2, allows an easy

discrimination among different permeation mechanisms. A Scanning Electron Microscope (SEM) monitors

possible morphological changes. The results show that the permeability P follows the Islam and Buschatz’s

model for L > 200 nm. The data allow determination of rates of sorption and desorption for CO2, He on

PDMS not available in literature, and show a CO2 permeability almost ten times greater than He, thus

confirming that the main mechanism of permselectivity is the solution-diffusion described in [34].

For L < 200 nm the SEM images show the appearance of pinhole defects and the selectivity and permeability

strongly change. In this case He permeates better than CO2 indicating the activation of different gas transport

mechanism [35] and previous cited model is not more relevant. The results are in agreement with the studies

of Ryoo et al. [9] on PDMS nano membranes that below 200 nm become sufficiently permeable to air and

methanol.

2. Experimental

2.1 Membrane support preparation

Fig. 1 reports description of the support of the polymer membrane. The set up warranties mechanical

stability and adequate area exposed to gas flow. A non-permeable silicon chip coated with a 200 nm Silicon

Nitride (Si3N4) low stress film, anisotropically etched, has a freestanding area (102 µm2) at the center of the

chip as shown in Fig. 1(a). For supporting membranes with L larger than 5µm the Silicon Nitride film is

completely removed. In this case the effective membrane area is 102 µm2.

For supporting membranes with L smaller than 5µm a Focused Ion Beams (FIB) (Fig. 1(b) and (c)) drills

holes in the Si3N4 freestanding area of the chip, varying size and number of the apertures depending on the

thickness of the membrane.

3
Fig. 1. Membrane support. (a) Silicon chip coated with 200 nm of Si3N4 with a freestanding square area of

102 µm2, (b) FIB milling of holes, (c) SEM image of porous support, (d) cross section of the support for

membrane with L smaller than 5µm.

FIB is a part of a Dual Beam instrumentation by Zeiss, CrossBeam® 1540XB, that combines a Focused Ion

Beam with an Ultra High Resolution Field Emission Scanning Electron Microscopy (UHR-FE-SEM). The

instrument allows versatile and fast fabrication of the support structure.

2.2 PDMS Membranes Preparation

Fig. 2 reports detailed scheme of the membrane fabrication process. A glass cover slip supports a thick

PDMS layer (Fig. 2(a)). The base and curing agent, mixed (10:1) (Sylgard 184 Silicone Elastomer Kit,

product by Dow Corning), are degassed in a vacuum desiccator for 15 minutes, spun on a glass cover slip (22

× 22 mm2) at 500 rpm for 60 s and finally baked at 100 °C for 1 hour. The layer is next treated with an

oxygen plasma (model Tucano purchase from Gambetti) for 60 s (power 30 W) and then silanized by

exposing it to an anti-stiction agent of 1H,1H,2H,2H-per-FluoroOctylTrichloroSilane (FOTS, purchase from

Sigma-Aldrich) in a vacuum chamber for 30 minutes (Fig. 2(b)). After, the PDMS film (the membrane) is

spin coated on the anti-stiction layer and baked at 60 °C (Fig. 2(c)). The PDMS membranes are fabricated

4
using different procedures depending on desired thickness. For membranes thicker than 1 µm, base and

curing agent are mixed (10:1), spun on the substrate previously described varying the rotational speed and

time to tune the thickness, and finally baked in oven at 60 °C for 1 hour. For thicknesses in the range 1 µm -

100 nm, base and curing agent are mixed (10:1) and diluted in hexane 1:25 (w/w). After treatment in oxygen

plasma for 60s (power 30 W), the PDMS membrane is brought into conformal contact with the support of

Fig. 1, gently pressed with tweezers and baked in oven at 60 °C for 15 min (Fig. 2(d)). Finally, the PDMS

membrane is transferred on the support previously described by peeling off the substrate (glass and thick

PDMS) (Fig. 2(e)). For PDMS membrane with L larger than 5µm as stated before the effective membrane

area is 102 µm2. For L smaller than 5 µm the effective membrane area is the sum of the area of the apertures

(Fig. 1(d)).

Fig. 2. Multistep membrane fabrication procedure. (a) A PDMS thick layer is deposited on a glass cover

slip, (b) an anti-stiction agent (FOTS) is evaporated on the PDMS layer of (a), (c) a thin PDMS layer, the

final membrane, is spin coated on (b) and baked at 60 °C, (d) after oxygen plasma treatment, the PDMS

membrane is brought into conformal contact with the support of Fig. 1 and then baked in oven for 15
5
minutes, (e) the PDMS membrane (d) is transferred on the support by peeling off the substrate (glass and

thick PDMS).

2.3 SEM characterization and thickness measurement

SEM monitors the surface morphology and measures the thickness L. The images record secondary and

backscattered electrons with the In-Lens detector at an accelerating voltage of 4 kV and a working distance

of 2 mm. The parameters avoid common SEM artefacts when imaging non-conductive samples. All the

membranes studied show morphologies which are similar to those of Fig. 3, where are reported the most

significant ones for (a) L = 3 µm and (b) L = 150 nm. For L larger than § 200 nm the membranes look

smooth as the one in (a), for L smaller than § 200 nm a surface texture appears as in (b), presuming the

presence of membrane defects.

For L > 1 µm, SEM images, as that of Fig. 3(c), allow an estimation of L with an accuracy in the worst

condition better than 20%. The mean value of repeated measurements in different positions gives the

thickness estimation and the maximum error is the measurement error. For L < 1 µm the membranes are

cross-sectioned by FIB using the procedure described in [36] and the SEM images, as that of Fig. 3(d), give a

maximum error on L larger than 20%. A chemical vapor deposition (CVD) [36] coating as shown in Fig.3

(d) overcomes or reduces drawbacks of sample damages from ion beam.

Fig. 3 SEM images of membranes. (a) L = 3 µm, (b) L = 150 nm. Thickness determination of a membrane

with (c) L = 27 µm and (d) L = 500 nm. In (d) PDMS membrane is cross-sectioned by FIB after FIB-CVD

deposition of Platinum (Pt). Si3N4 support has L = 200 nm, PDMS membrane L = 500 nm.
6
2.4 Gas permeation measurements

Fig. 4 shows the final assembly to measure gas permeability. A perforated copper disk, Conflat flange

compatible, supports the PDMS membrane. A low vapor pressure epoxy (Torr Seal) ensures vacuum

sealing obtaining negligible leak losses and mechanical integrity even at pressure differences up to 1atm.

Fig. 4. Device for gas permeability measurements.

The assembly of Fig. 4 inserted in the gas permeability apparatus of reference [37], allows performing

permeability measurements. Fig. 5 sketches the apparatus. The system is similar to that assembled by

Tremblay for the same purpose, and described in ref. [38].

7
Fig. 5. Schematic representation of the experimental apparatus for gas permeation measurements. CS

capacitive sensor (Ceravac model CTR90 purchase from Leybold Vakuum,), LV leak valve, RP roughing

pump (Model XDS 10 Edwards purchase from Edwards), M membrane, QMS quadrupole mass spectrometer

(model Prisma, purchase from Pfeiffer Vacuum), IG calibrated ion gauge (Bayard-Alpert model UHV-24

purchase from Varian, Inc.), TP turbomolecular pump (model Navigator TV1001 purchase from Varian,

Inc.).

Eq. (1) gives the membrane permeability P as a function of thickness L, upstream pressure pu, downstream

pressure pd, and molar flux J through the membrane

LJ
P= (1)
( pu − p d )

The capacitive sensor (CS) on feed chamber measures the upstream pressure pu that is varied by incremental

steps up to 9 × 104 Pa. He or CO2 have purity grade N5.0. The QMS measures the downstream partial
8
pressure pd in steady-state conditions. The IG gauge calibrates [39] the QMS by means a calibrated leak. Eq.

(2) gives the molar flux J per unit effective membrane area A of the gas penetrating into the chamber trough

the membrane

pd s
J= (2)
ART

R is the gas universal constant, T is the absolute temperature and s the pumping speed of the TP pump.

Before measuring the membrane permeability, RP evacuates the feed chamber to § 1 Pa, while TP the high

vacuum chamber to § 1 × 10-7 Pa. To record the background and to obtain steady-state conditions we switch

on the QMS one hour before measuring. We repeat the measurements several times to check possible aging

effects [40] and we alternate fluxes of CO2 and He to test possible plasticization effects [41]. None of them is

observed within 10 days.

Following equation (1) we calculate the error on the permeability P by the errors on molar flux J, thickness L

and differential pressure pu – pd. Equation (2) determinates the error on J. We do measurements of pd, with

the precautions recommended in ref. [39], so that its relative error, after IG calibration of the QMS, results of

10%, the error on the membrane area A and on the TP pumping speed, s, are less than 2%, resulting in a

relative error on J of 14%. We point out that the pressure measurement performed by the IG gauge after

calibration have an accuracy much better than the 20% declared by the manufacturer. For molar flux less

than 2 × 10-11 mol s-1, the error on pd is affected by the determination of the background pressure and the

error on J results greater than 14%. The error on L, as discussed in the previous section, results for L > 1 µm

better than 20%, while below 1 µm becomes larger. Since in the present experimental conditions pu >> pd,

the error on pu – pd is determined by the error on pu which is less than 0.5%. The error on P is therefore

roughly the sum of the relative error on J and L. To taking into account possible errors arising from the lack

of reproducibility in the membrane preparation, we repeat the permeability measurements on several (up to

5) membranes of the same thickness L. The values of P reported in Figs. 6, 7, 8 are the weighted means and

the errors are the variance of the weighted means.

9
 3. Results and Discussion

3.1 Results

Fig. 6 reports the average values of P as function of L (from 1 mm to 200 nm) for CO2 and He at pu = 9 ×

104 Pa and T = 293 K.

Fig. 6. PDMS permeability P against thickness membrane L. „ He/PDMS and  CO2/PDMS. The

section 2 reports error analysis. The upstream pressure is pu = 9 × 104 Pa, the temperature T = 293 K.

Above L § 50 µm the PDMS permeability P is independent from L, in samples thinner than some tenths of

micrometers P decreases with L. The selectivity β CO2 / He ( β CO2 / He = PCO2 / PHe with PCO2 , PHe the

respective permeability) remains constant with L as in Fig. 7.

Fig. 7. Selectivity β CO2 / He as function of L. The upstream pressure is pu = 9 × 104 Pa and the temperature T

= 293 K. The section 2 reports error analysis.

10
β CO2 / He follows the reverse-selectivity behavior explained by the solution-diffusion mechanism [42]

previously observed in others dense polymers [43].

By reducing the membrane thickness below 200 nm, both permeability and selectivity change, indicating

Knudsen diffusion as flux mechanisms responsible for permeation. In this range of thickness, SEM images

reveal membrane defects (see Fig. 3(b)). The PDMS becomes porous and the solution-diffusion [42] model

generally used to explain the transport mechanism through dense membranes does not apply because the

mechanism is not the only one responsible of the permeation [35].We do not report the experimental points

since they are not meaningful for the following discussion.

3.2 Discussion

In the solution-diffusion model, based on the Fick’s and Henry’s laws, the permeability PHenry = DS, where D

is the diffusion coefficient and S the solubility and PHenry does not depend on the thickness of the membrane.

This model fits the present data only for L larger than 50 µm, where P results independent from L. In this

range, PCO2 = (1.5 ± 0.4) × 10 −12 mol m-1 s-1 Pa-1 and PHe = (2.1 ± 0.7) × 10−13 mol m-1 s-1 Pa-1 are, in fact,

within experimental errors, equal to those measured in thick PDMS membranes and reported in ref. [19, 23,

34]. For L in the range 50 µm < L < 200 nm, P shows a strong decrease with L but the selectivity does not

change significantly from its value at L = 1 mm, β CO2 / He ≈ 7 , a value very close to the one reported in ref.

[23] for thick PDMS.

Islam and Buschatz [31, 32] analyze in more details the steady state diffusion flux through a membrane,

considering non-equilibrium sorption-desorption process at the interface. The analysis concludes that for a

given system (gas and polymer), there may exist a value of L below which the surface reactions are not any

more in near equilibrium and the measured permeability becomes dependent on the membrane thickness.

Following ref. [31] one may consider the measured permeability P as an “apparent” permeability, deviating

from the product of D and S. To describe quantitatively this dependence, we compare the experimental

results with the models of ref. [32]. The authors propose two flux equations, which predict the conditions for

which the permeability would appear thickness dependent. In one model (hereafter referred as model A) they

suppose the transport coefficient ț = cost obtaining an “apparent” permeability P depending from thickness
11
P (L) as shown in equation (3) in its most general formulation. We point out in fact that the derivation of

k1 p u
model A of ref [32] made by Islam and Buschatz assumes that >> 1 and obtains for eq. (3)
J

§P·
P exp ¨ ¸ = k1 L . To give more generality to the model we calculate the flux equation without this
©a¹

approximation and eq. (3) becomes

§ §P· ·
P¨¨ exp¨ ¸ + 1¸¸ = k1L (3)
© ©a¹ ¹

κRT
Where a = and k1 is the sorption rate [32].
pu

Similarly, they propose an analogous flux equation, which assumes constant the diffusion coefficient D. This

model (referred as model B) gives the “apparent” permeability P (L) of equation (4)

§ k1 ·
¨ ¸L
P= © ¹
2
(4)
§k ·
1 + ¨ 2 ¸L
© 2D ¹

Where k1 and k2 are respectively the sorption and desorption rates as defined in ref. [32].

Fig. 8 reports the experimental data of Fig. 6 from L = 55 µm to L = 230 nm compared with model A [eq.

(3)] and model B [eq. (4)].

12
Fig. 8. Best fit of the Permeability data from L = 55 µm to L = 230 nm with model A [eq. (3)] and

model B [eq. (4)]. (a) CO2 / PDMS data compared with model A (continuous line) (b) He / PDMS data

compared with model A (continuous line) (c) CO2 / PDMS data compared with model B (dotted line) (d) He /

PDMS data compared with model B (dotted line).

2
By minimizing the χ , we obtained the best-fit parameter reported in Table 1.

Table 1. Sorption coefficient k1, transport coefficient κ, desorption coefficient k2 and diffusivity D
for He and CO2 on PDMS as determined by best-fit analysis

Parameters of Model A eq. (3) Parameters of Model B eq. (4)

k1 x 108 κ x 1013 χ2/ndf p-value k1 x 108 k2 /2D x 10-.4 χ2/ndf p-value

Gas -2 -1
(mol m s Pa ) -1
(mol2 m-4 s-1 Pa-1)
-2 -1
(mol m s Pa ) -1 -1
(m )

CO2 14±1 4±1 10.43/8 0.24 14±1 2.9±0.9 10.98/8 0.20

He 2.3±0.3 0.4±0.1 8.662/8 0.37 2.3±0.2 4±2 8.87/8 0.35

13
The χ2 analysis shows that both models have a good p-value, thus confirming the goodness of the Islam and

Buschatz’s assumption and indicating that flux models based on interface reaction could apply to describe

quantitatively the thickness-dependence of the permeability. It is difficult to distinguish which one is the

most appropriate to fit the present experiment since both fit give similar p-values. The two model seem to

give also similar values for the physical parameters. For both He and CO2 model A and model B give the

same values of k1.

The parameters obtained by model B allow to calculate PHenry and to compare it with the experimental results

k1
obtained with different experimental techniques available in literature [19, 23, 34, 38]. In fact, since S = ,
k2

results

k1 1 2 D
PHenry = D = k1 (5)
k2 2 k2

Which can be easy calculated by the best-fit parameters of Table1. PHenry is independent from the thickness

and represents the value of permeability for thick membranes, values that are available in literature.

Table 2 reports comparison among PHenry and experiments on PDMS membranes.

Table 2: Comparison of PHenry obtained by the best-fit parameters values of Model B with the
experimental measurements of permeability of thick PDMS membranes.
a
data from ref. [19], b data from ref. [21], c data from ref. [23].

Gas PHenry eq.(5) × 1012 P(1mm) this work × 1012 P from literature × 1012

(mol m-1 s-1 Pa-1 ) (mol m-1 s-1 Pa-1 ) (mol m-1 s-1 Pa-1 )

CO2 2±1 1.5±0.4 1.59 ±0.07a

1.27 ±0.02b
1.04c

He 0.3±0.1 0.21±0.07 0.13c

The agreement is very good confirming the correctness of the analysis.

14
Model B allows also to compare the results of the present paper with other experiments on polymers, like

Oxygen on wet and dry Kapton [29] and water on rubber [27]. In fact, by setting equal to 1 the dimensionless

1 2D
parameter α = [31], which serves as a measure for the deviation of the surface reaction from
L k2

equilibrium, one could determine for each gas/membrane system a specific characteristic thickness

2D P
Lc = which, when introduced in equation (4), provides a functional form for . For thicknesses
k2 PHenry

close Lc the permeability starts to differ from PHenry and becomes thickness dependent. We remind that,

according to reference [31], the surface reaction are “near equilibrium” for α << 1 , while for α ≥ 1 , P

appears dependent on L. By using Lc and PHenry of equation (5), the apparent permeability P of equation (4)

results:

L
P Lc
= (7)
PHenry 1 + L
Lc

P L
By plotting the values of the dimensionless quantity as function of one obtains the functional
PHenry Lc

form of Fig. 9, which is independent from gas and system. The values of PHenry and Lc obtained by the values

of the best- fit parameters for the present work and for the data of references [27, 29] are reported in Table 3.

Table 3: Values of PHenry and Lc obtained by the best-fit parameters values of Model B for various
gas/membrane systems.

Gas/Membrane Lc PHenry × 1012 Gas/Membrane Lc PHenry × 1012

(µm) (µm)
(mol m-1 s-1 Pa-1) (mol m-1 s-1 Pa-1)

CO2/PDMSa 30±10 2±1 O2/Kapton (dry)b,d 14 3.8 10-5

He/PDMSa 25±12 0.3±0.1 Water/Rubber c,d 78 1.0

a b c
present work Table 1, data from ref. [29], data from
O2/Kapton (wet)b,d 22 3.0 10-5 ref. [27], d from ref. [32] which contains the analysis of
the data of ref. [29, 27].

15
Figure 9 reports the behavior of the eq.7 for all the experimental data available. The parameter used in the

plot are those of Table 3.

Figure 9. Plot of P/PHenry as function of L/Lc.  He / PDMS, z CO2 / PDMS, „ O2 / Kapton (dry),

S O2 / Kapton (wet), ² Water/Rubber. The data follow closely the functional form of equation (7)

(continuous line).

We notice that the characteristic length Lc indicates the deviation of the surface reaction from equilibrium.

For L smaller than Lc the solution-diffusion mechanism is associate at non-equilibrium sorption-desorption

processes at the interface affecting the measured permeability, which becomes dependent on the membrane

thickness. For L larger than Lc these effects are less important and the permeability results independent from

L and close to the value of PHenry.

A recent paper [44] simulates the diffusive flux trough a composite membrane similar to those of the present

study. The results indicate that the effective permeation thickness Leff is bigger than the real thickness of the

membrane L. To take into account of this effect the authors proposed a corrected version of the equation first

published by Strathmann et al. [45] for calculating Leff (see equation 8 ref. [44]). The correction affects both

permeability P (equation 1) and thickness L (now Leff). In our case, as described in section 2.2, the correction

applies to the membranes having thickness lower than 5 µm. We calculate P (Leff) by using equation 8 of ref.

[44] and we fit the corrected values with models A and B. The parameters obtained are, within the

experimental errors, the same as those reported in Table 1.

16
 4. Conclusions

The paper describes a reliable technique to prepare thin PDMS membranes able to maintain 1 atm of

differential pressure without rupturing. A SEM/FIB facility allows accurate thickness measurement as well

surface characterization. A high vacuum set up equipped with quadrupole mass spectrometer measures

permeability of CO2 and He on PDMS. The investigation reveals a permeability that is thickness-dependent

for thicknesses smaller than some tens of micrometers. The models of ref. [31, 32], based on non-equilibrium

sorption-desorption process at the interface, fit quite well the experimental data. A best-fit analysis

determinates sorption coefficient k1, transport coefficient κ, desorption coefficient k2 and diffusivity D for He

and CO2 on PDMS.

2D P
The introduction of a specific characteristic thickness Lc = provides a functional form for
k2 PHenry

which describes very accurately the results of the present paper and other experiments on polymers, like

Oxygen on wet and dry Kapton [29] and water on rubber [27]. For CO2 / PDMS Lc = 30 ± 10 µm and for He /

PDMS Lc = 25 ± 12 µm.

For L < 200 nm SEM shows a porous structure of membrane, confirmed by the permeability and selectivity

data that indicate Knudsen diffusion as flux mechanisms responsible for permeation.

In the limits of validity of Islam and Buschatz’s model, it is possible to replicate the experiments of the

present paper with different types of polymers and different gas; this would enable to determine surface rate

constants not available in literature.

Acknowledgments

The authors thank very much Prof. F. Buatier de Mongeot for providing some parts of the vacuum

equipment. The work was supported by the project FIRB RBAP11BYNP_003 NEWTON of the MIUR.

17
References

[1] W.V. Chiu, I.-S. Park, K. Shqau, J.C. White, M.C. Schillo, W.S.W. Ho, P.K. Dutta, H. Verweij, Post-

syntesis defect abatment of inorganic membranes for gas separation, J. Membr. Sci. 377 (2011) 182-190.

[2] A. Brunetti, F. Scura, G. Barbieri, E. Drioli, Membrane technologies for CO2 separation, J. Membr.

Sci. 359 (2010) 115-125.

[3] W.J. Koros, R. Mahajan, Pushing the limits on possibilities for large scale gas separation:which

strategies?, J. Membr. Sci. 175 (2000) 181-196.

[4] R. W. Baker, Future Directions of Membranes Gas Separation Technologies, Ind. Eng. Chem. Res. 41

(2002) 1393-1411.

[5] R. Rego, A. Mendes, Carbon dioxide/methane gas sensor based on the permselectivity of polymeric

membranes for biogas monitoring, Sens. Actuators B 103 (2004) 2-6.

[6] S. Neethirajan, D. S. Jayas, S. Sadistap, Carbon Dioxide (CO2) Sensors for the Agri-food Industry - A

Review, Food Bioprocess Tech. 2 (2009) 115-121.

[7] L.-J. Yang, A-F. Kao, Gas Permeation in PDMS Monitored by On-site Pressure Sensors, 5th IEEE

International Conference on Nano/Micro Engineered and Molecular Systems (2010) 352-355.

[8] L.Yang, On gas permeation in PDMS, J. Micromech. Microeng. 20 (2010) 115033-115040.

[9] J.H. Ryoo, G.S. Jeong, E. Kang, S.H. Lee, Ultra-thin, hyperelastic PDMS nano membrane: fabrication

and characterization, 15th International Conference on Miniaturized Systems for Chemistryand Life Science

(2011) 686-688.

18
[10] O. C. Jeong, S. Konishi, Fabrication and drive test of pneumatic PDMS micro pump, Sens. Actuators

A 135 (2007) 849-856.

[11] M. A. Eddings, B. K. Gale, A PDMS-based gas permeation pump for on-chip fluid handling in

microfluidic devices, J. Micromech. Microeng. 16 (2006) 2396–2402.

[12] C. Hagleitner, A. Hierlemann, D. Lange, A. Kummer, N. Kerness, O. Brand, H. Baltes, Smart single-

chip gas sensor microsystem, Nature 414 (2001) 293-296.

[13] C. Hagleitner, D. Lange, A. Hierlemann, O. Brand, H. Baltes, CMOS Single-Chip Gas Detection

System Comprising Capacitive, Calorimetric and Mass-Sensitive Microsensors, IEEE Journal os Solid State

Circuits 37 (2002) 1867-1878.

[14] S. Kim, J. R. Jinschek, H. Chen, D. S. Sholl, E. Marand, Scalable Fabrication of Carbon

Nanotube/Polymer Nanocomposite Membranes for High Flux Gas Transport, Nano Letters 7 (2007) 2806-

2811.

[15] A.L. Thangawng, R.S. Ruoff, M.A. Swartz, M.R. Glucksberg, An ultra-thin PDMS membrane as a

bio/micro-nano interface: fabrication and characterization, Biomed Microdevices 9 (2007) 587-595.

[16] L. Engel, J. Shklovsky, D. Schrieber, S. Krylov, Y. Shacham-Diamand, Freestanding smooth micron-

scale polydimethylsiloxane (PDMS) membranes by thermal imprinting, J. Micromech. Microeng. 22 (2012)

045003-045011.

[17] M. Liu, J. Sun, Y. Sun, C. Bock, Q. Chen, Thickness-dependent mechanical properties of the

polydimethylsiloxane membranes, J. Micromech. Microeng. 19 (2009) 035028-035031.

[18] G. L. Jadav, V. K. Aswal, H. Bhatt, J. C. Chaudhari, P. S. Singh, Influence of film thickness on the

structure and properties of PDMS membrane, J. Membr. Sci. 415-416 (2012) 624-634.

19
[19] M. Sadrzadeh, K. Shahidi, T. Mohammadi, Synthesis and Gas Permeation Properties of a Single

Layer PDMS Membrane, J. Appl. Polym. Sci. 117 (2010) 33-48.

[20] C. K. Yeom, S. H. Lee, J. M. Lee, Study on Transport of Pure and Mixed CO2/N2 Gases through

Polymeric Membrane, J. Appl. Polym. Sci. 78 (2000) 179-189.

[21] T. C. Merkel, B. D. Freeman, R. J. Spontak, Z. He, I. Pinnau, P. Meakin, A. J. Hill, Ultrapermeable

Reverse-Selective Nanocomposite Membranes, Science 296 (2002) 519-522.

[22] H. Vinh-Thang, S. Kaliaguine, Predictive Models for Mixed-Matrix Membrane Performance: A

Review, Chem. Rev. 113 (2013) 4980-5028.

[23] K.S Houston, D.H Weinkauf, F.F Stewart, Gas transport characteristics of plasma treated

poly(dimethylsiloxane) and polyphosphazene membrane materials, J. Membr. Sci. 205 (2002) 103-112.

[24] C.-C. Hu, C.-Y. Tu, Y.-C. Wang, C.-L. Li, K.-R. Lee, J.-Y. Lai, Effects of Plasma treatment on CO2

Plasticization of Poly(methyl methacrylate) Gas-Separation Membranes, J. Appl Polym Sci 93 (2004) 395-

401.

[25] H. Lin, B. D. Freeman, Gas solubility, diffusivity, and permeability in poly(ethylene oxide), J. Membr.

Sci. 239 (2004) 105-117.

[26] B. W. Rowe, B. D. Freeman, D. R. Paul, Physical aging of ultrathin glassy polymer films tracked by

gas peremability, Polymer 50 (2009) 5565-5575.

[27] R. L. Taylor, D. B. Herbmann, A. R. Kemp, Diffusion of Water through Insulating Materials Rubber,

Synthetic Resins, and Other Organics, Ind. Eng. Chem. ACS 28 (1936) 1255-1263.

20
[28] A.M. Shishatskii, Y.P. Yampol'skii, K.-V. Peinemann, Effects of film thickness on density and gas

permeation parameters of glassy polymers, J. Membr. Sci. 112 (1996) 275-285.

[29] G. Mensitieri, M.A. Del Nobile, T. Monetta, L. Nicodemo, F. Bellucci, The effect of film thickness on

oxygen sorption and transport in dry and water-saturated Kapton® polyimide, J. Membr. Sci. 89 (1994) 131-

141.

[30] L. S. Kocherlakota, D. B. Knorr Jr., L. Foster, R. M. Overney, Enhanced gas transport properties and

molecular mobilites in nano-constrained poly[1-(trimethylsilyl)1-propyne] membranes, Polymer 53 (2012)

2394-2401.

[31] M.A. Islam, H. Buschatz, D. Paul, Non-equilibrium surface recations-a factor in determining steady

state diffusion flux, J. Membr. Sci. 204 (2002) 379-384.

[32] M.A. Islam, H. Buschatz, Assesment of thickness-dependent gas permeability of polymer membranes,

Indian J. Chem. Technol. 12 (2005) 88-92.

[33] D. Wang, S.-Y. Gou, D. Axelrod, Reaction rate enhancement by surface diffusion of adsorbates,

Biophys. Chem. 43 (1992) 117-137.

[34] T. C. Merkel, V. I. Bondar, K. Nagai, B. D. Freeman, I. Pinnau, Gas Sorption, Diffusion, and

Permeation in Poly(dimethylsiloxane), J. Polym. Sci. Part B Polym. Phys. 38 (2000) 415-434.

[35] P. Pandey, R.S. Chauhan, Membranes for gas separation, Prog. Polym. Sci. 26 (2001) 853-893.

[36] M. W. Phaneuf, FIB for material science applications – a review, in: L. A. Giannuzzi, F. A. Stevie

(Eds.), Introduction to Focused Ion Beam: Instrumentation, Theory, Techniques and Practice, Springer

Science+ Business Media Inc. USA (2005) 143-172.

21
[37] G. Firpo, L. Repetto, F. Buatier de Mongeot, U. Valbusa, Focused-ion beam fabrication of nanometer

orifices for leak detection, J. Vac. Sci. Technol. B 27 (2009) 2347-2350.

[38] P. Tremblay, M. M. Savard, J. Vermette, R. Paquin, Gas permeability, diffusivity and solubility of

nitrogen, helium, methane, carbon dioxide and formaldehyde in dense polymeric membranes using a new on-

line permeation apparatus, J. Membr. Sci. 282 (2006) 245-256.

[39] B.R.F. Kendall, Ionization gauge errors at low pressure, J. Vac. Sci. Technol. A 17 (1999) 2041-2049.

[40] K. D. Dorkenoo, P. H. Pfromm, Accelerated Physical Aging of Thin Poly [1-(trimethylsilyl)-1-

propyne] Films, Macromolecules 33 (2000) 3747-3751.

[41] N. R. Horn, D.R. Paul, Carbon dioxide plasticization and conditioning effects in thick vs. thin glassy

polymer films, Polymer 52 (2011) 1619-1627.

[42] J.G. Wijmans, R.W. Baker, The solution-diffusion model:a review, J. Membr. Sci. 107 (1995) 1-21.

[43] C. H. Laua, P. Li, F. Li,T.-S. Chung, D. R. Paul, Reverse-selective polymeric membranes for gas

separations, Prog. Polym. Sci. 38 (2013) 740-766.

[44] A. M. Peters, R.G.H. Lammertink, M. Wessling, Comparing flat and micro-patterned surfaces: Gas

permeation and tensile stress measurements, J. Membr. Sci. 320 (2008) 173-178.

[45] H. Strathmann, C.M. Bell, J. Kerres, Gas separation and perevaporation-membrane and module

development, Desalination 77 (1990) 259.

22
 Highlights

• Gas Permeability measurements of thin PDMS membranes show dependence from membrane thickness
• The data are in agreement with a model that gives the surface reactions rate constants.
• A functional form based on a characteristic thickness Lc describes present and other experiments.
• In case of ultra-thin membranes, the mechanism responsible for permeation is Knudsen diffusion

23
Graphical Abstract (for review)