Field-effect reconfigurable nanofluidic ionic diodes

Abstract

Several types of nanofluidic devices based on nanopores and nanochannels have been reported to yield ionic current rectification, with the aim to control the delivery of chemical species in integrated systems. However, the rectifying properties obtained by existing approaches cannot be altered once the devices are made. It would be desirable to have the ability to modulate the predefined properties in situ without introducing external chemical stimuli. Here we report a field-effect reconfigurable nanofluidic diode, with a single asymmetrically placed gate or dual split-gate on top of the nanochannel. The forward/reverse directions of the diode as well as the degrees of rectification can be regulated by the application of gate voltages. Compared with the stimuli-responsive tuning of the rectification properties, the electrostatic modulation offers a fully independent and digitally programmable approach for controlling the preferential conduction of ions and molecules in fluids. This device would serve as a building block for large-scale integration of reconfigurable ionic circuits.

ARTICLE

Received 19 Apr 2011 | Accepted 21 Sep 2011 | Published 18 Oct 2011 DOI: 10.1038/ncomms1514

Several types of nanouidic devices based on nanopores and nanochannels have been reported to yield ionic current rectication, with the aim to control the delivery of chemical species in integrated systems. However, the rectifying properties obtained by existing approaches cannot be altered once the devices are made. It would be desirable to have the ability to modulate the predened properties in situ without introducing external chemical stimuli. Here we report a eld-effect recongurable nanouidic diode, with a single asymmetrically placed gate or dual split-gate on top of the nanochannel. The forward/reverse directions of the diode as well as the degrees of rectication can be regulated by the application of gate voltages. Compared with the stimuli-responsive tuning of the rectication properties, the electrostatic modulation offers a fully independent and digitally programmable approach for controlling the preferential conduction of ions and molecules in uids. This device would serve as a building block for large-scale integration of recongurable ionic circuits.

Field-effect recongurable nanouidic ionic diodes

Weihua Guan1, Rong Fan2 & Mark A. Reed1,3

1 Department of Electrical Engineering, Yale University, New Haven, Connecticut 06520, USA. 2 Department of Biomedical Engineering, Yale University, New Haven, Connecticut 06520, USA. 3 Department of Applied Physics, Yale University, New Haven, Connecticut 06520, USA. Correspondence and requests for materials should be addressed to M.A.R. (email: [email protected]).

NATURE COMMUNICATIONS | 2:506 | DOI: 10.1038/ncomms1514 | www.nature.com/naturecommunications

2011 Macmillan Publishers Limited. All rights reserved.

ARTICLE

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms1514

With applications ranging from biosensing to the control of molecular transport, synthetic nanopores and nanochannels are the focus of growing scientic interest1. Analo

gous to a solid-state semiconductor diode for regulating the ow of electrons/holes to one preferential direction, nanouidic diodes are being developed to achieve the rectied ionic transport-ions favourably move in one direction and are inhibited in the opposite direction. Such rectication eect is of great importance owing to its relevance to biological ion channels2. Moreover, ionic diodes, together with ionic transistors39 represent the key building blocks for ionic circuits10,11, which would allow for regulating6, sensing12, concentrating13 and separating14 ions and molecules in electrolyte solutions, mimicking voltage-gated ion channels in a variety of biological systems.

Several nanouidic platforms based on nanopores and nano-channels were reported to produce ionic current rectication by symmetry breaking15 in geometries1619, surface charge distributions (either intrinsic material properties20,21 or chemically modied properties22,23), bath concentrations24, or a combination of them, for example, by positively and negatively patterning charged regions in conical nanopores25. Nevertheless, it has not been possible to change the predened rectifying properties obtained by these approaches once the devices are made. Although several externally tunable methods have been proposed so far, most of them aim to alter the nanochannel wall property by introducing external chemical stimuli, for example, hydronium ions (pH)2628, enzymes29 and polyvalent cations30. All these methods require changing the native environment of the solution being transported. Contrary to the chemical stimuli-responsive schemes, an electric eld normal to the nanochannel walls is able to enhance or diminish the ionic concentrations near the surface in situ39, resembling the carrier number modulation in a metal-oxide-semiconductor eld-eect transistor.

Here we report a eld-eect recongurable ionic diode by asymmetrically modulating the cation/anion ratios along the nanochannel. The eld-eect approach requires no solution replacement and

allows large-scale integration for more complex functions. A key feature of our device is that it allows the post-fabrication recongu-ration of the diode functions, such as the forward/reverse directions as well as the rectication degrees. These results may lead to the creation of recongurable ionic circuits, an ionic counterpart of the electronic eld-programmable gate array.

ResultsDevice structure. The schematic structure of the nanouidic eldeect recongurable diodes (FERD), and the experimental setup is illustrated in Figure 1. The FERD is a three-terminal device that has a similar structure to a nanouidic eld-eect transistor (FET)6,7,

yet with a critical dierence that the gate electrode of FERD is asymmetrically located near one of the microuidic reservoirs. The control devices with gate electrodes sitting symmetrically along the nanochannel were also fabricated on the same silicon wafer. We used a sacricial layer method to produce the nanochannels (20 nm in height) with a novel bond-followed-by-etch scheme, which completely avoids the nanochannel collapse problem encountered in conventional etch-and-bond schemes (Supplementary Fig. S1 and Supplementary Methods). To make a consistent notion in the following discussion, we denote the microuidic reservoir that is close to the gate electrode as Cis (C) and the other reservoir far away from the gate electrode as Trans (T).

Nanochannel characterization. We rst carried out experiments on the electrical characterization (Supplementary Fig. S2) of the nanochannels with the gate terminal oating. By measuring the nanochannel conductance as a function of KCl concentration (Cb), we observed a linear dependence of channel conductance on Cb in

the high-concentration regime and a conductance plateau in the low-concentration regime (Fig. 2). The transition concentration is around 1 mM, at which the corresponding Debye screening length (D) is around 10 nm, equal to half of the designed nanochannel height. This behaviour conrms a surface-charge governed ionic

Gate

Outlet

Inlet

Trans

Outlet

Gate

Inlet

SiO2

Nanochannels

Cis

Nanochannels

Trans

SiO2

Cis

Gate

Si

Cis

Trans

Trans in

VG

Cis in

Ag/AgCl

Trans out

VC

A

Cis out

Cu

Cis

Gate

Trans

~20 nm

Ag/AgCl

Figure 1 | Device structure and experimental setup. (a) Schematic of the nanouidic FERD. (b) Sketch of the planar layout for the assembled device. Two microuidic channels deliver the electrolyte solutions to the Cis and Trans reservoirs, formed by SiO2 trenches with supporting pillars, as shown in the magnied scanning electron microscope (SEM) image. The scale bar in the SEM image is 100 m. There are 11 parallel nanochannels connecting Cis and Trans reservoirs. The supporting pillars prevent the microuidic polydimethylsiloxane from collapsing into the reservoirs. (c) Schematic of the electrical and uidic connection congurations. The electrical contacts are made of Ag/AgCl electrodes and are integrated with the connecting tubes, serving asa low resistive loss contact. VG and VC denotes the voltage on the gate and Cis, respectively. The Trans side is referenced as ground in all measurementsin this study. The whole setup is placed in a Faraday cage to shield the electrostatic noise. (d) Photograph of a FERD device with electrical and uidic connections.

NATURE COMMUNICATIONS | 2:506 | DOI: 10.1038/ncomms1514 | www.nature.com/naturecommunications

2011 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms1514

ARTICLE

transport property31,32. By knowing the dimensions of the lithographically dened channels, the surface charge density (s) can be derived by tting the experimental data (Supplementary Methods). We tested two individual devices fabricated on the same wafer but with dierent channel width W (2 m11 and 3 m11, respectively). Both of them can be well tted by a same s ( 2 mC m 2),

suggesting the surface properties remain consistent and reliable

aer the bond-followed-by-etch process. This s value is close to those reported with the plasma-enhanced chemical vapour deposition SiO2 (~ 4.5 mC m 2)20,24.

AS the surface charge can be modied by changing the solution pH, we further tested the pH dependence of the nanochannel conductance at xed KCl concentrations (inset of Fig. 2). At high KCl concentrations (100 mM), changing the pH value has negligible impact on the channel conductance, whereas at low KCl concentrations (100 M), pH has a pronounced impact. This behaviour can be well explained by surface charge governed transport property. At high ionic concentrations, the double layer conductance (which is regulated by pH values) is overwhelmed by the bulk conductance. Therefore, the channel conductance will not change by varying pH. At low KCl concentrations, the channel conductance is dominated by the double layer and is thus pH-dependent. As [H + ] for a given electrolyte concentration decreases, the number of SiO surface charges increases33.

Field-eect recongurability. Aer verifying that the device conductance is indeed governed by the ion transport in the nanochannels (instead of leakage paths), we went on to investigate the eldeect tunability of the three-terminal FERD devices using electrical congurations shown in Figure 3a. The channel current as a function of Cis to Trans voltage (VCT) was measured under various gate voltages (VG). Figure 3b shows the representative current-voltage (IV) curves obtained with VG of dierent polarities, using a 100-M KCl solution. A clear gate-voltage controlled rectifying property is observed. At VG = 0 (middle panel, Fig. 3b), the I-V curve is symmetric with respect to the origin. At a negative gate voltage (le panel, Fig. 3b), the ionic current in the positive VCT regime is higher than that in the negative VCT regime. While at a positive gate voltage (right panel, Fig. 3b), the negative VCT conducts more ionic current than the positive VCT. Consequently, it is possible to switch the preferential direction of the ionic ow via adjusting the external gate voltage, a novel mechanism of nanouidic control that has never been shown before.

We note that an electric potential barrier will develop at the micro/nano channel interface34, which is analogous to a potential barrier at a heterojunction in semiconductor devices due to a work

107

10.0

0.6

0.5

9.5

Conductance (nS)

0.4

Conductance (nS)

108

9.0

100 mM

0.3

8.5

0.2

100 M

Conductance (S)

8.0

0.1

5

6 7 8 9

pH

109

W=33 m W=22 m

Fitting by s=2 mC m2 Fitting by s=2 mC m2

Bulk conductance

10 107 106 105 104 103 102 101 100 101

1010

11

KCl concentration (Molar)

Bulk conductance

Figure 2 | Measured conductance as a function of KCl concentrations for two devices. Both devices were fabricated using the identical process on a same wafer with gate terminal oating (Blue: W = 33 m, L = 100 m,

H = 20 nm. Red: W = 22 m, L = 100 m, H = 20 nm). Solid lines are the tting curves with s = 2 mC m 2. Dashed lines are the bulk prediction that deviates from the experimental data in the low ionic concentration region. The error bars correspond to ten measurements and are smaller than the size of the symbol used. Inset shows the pH dependence of channel conductance at two electrolyte concentrations ([KCl] = 100 M

and [KCl] = 100 mM). The pH of the electrolyte solution is adjusted by adding hydrochloric acid or potassium hydroxide into 100-M phosphate buffers. The error bars correspond to ten measurements.

200

200

VG

150

200 VG=

1.4 V

150

VG

=0 V

150

VG

=1.4 V

100

100

100

VCT

Current (pA)

Current (pA)

Current (pA)

50

50

50

0

0

0

50

VG<0

50

VG=0

50

VG>0

100

100

100

Field-effect reconfigurable diode

C

T

C

T

C

T

150

150

150

200 1.5

200

200

0.0 1.5 1.5 0.0 1.5 1.5 0.0 1.5

1.5 0.0 1.5

VCT (V) VCT (V) VCT (V)

VCT (V) VCT (V) VCT (V)

200

200

200

VG

150

VG=

1.4 V

150

VG

=0 V

150

VG

=1.4 V

150

100

100

VCT

Current (pA)

Current (pA)

Current (pA)

100

50

50

0

0

0

50

50

50

100

100

100

Control device: ionic transistor

150

150

150

200

200

200

1.5 0.0 1.5

1.5 0.0 1.5

Figure 3 | Current-voltage (I-V) curves for the FERD devices with different gate voltage (VG) polarities. (a) Testing congurations for the FERD device. (b) I-V curves for FERD devices (W = 2 m11, L = 100 m, H = 20 nm) using a 100-M KCl solution. An apparent gate voltage-controlled rectifying property is observed. The inset symbols schematically show the forward direction switched on positive and negative gate voltages. (c) Testing congurations for the control transistor device. (d) I-V curves for the control device (W = 3 m11, L = 100 m, H = 20 nm) using a 100-M KCl solution.

Gate voltage can modulate the channel conductance but no rectifying property is observed.

NATURE COMMUNICATIONS | 2:506 | DOI: 10.1038/ncomms1514 | www.nature.com/naturecommunications

2011 Macmillan Publishers Limited. All rights reserved.

ARTICLE

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms1514

function dierence35. This interfacial barrier is indeed a negligible eect in our FERD devices and is not responsible for the voltage tunable rectifying property (Supplementary Fig. S3). This conclusion is further veried by the IV characteristics of the control devices (with symmetrically placed gate electrodes), schematically shown in Figure 3c. The major structural dierence between the FERD devices and the control devices is the location of the gate terminal. No rectifying features are observed for the control device, which exhibits a symmetric I-V relationship for all VG polarities (Fig. 3d).

The control device actually shows a typical p-channel transistor behaviour, where the conductance of the nanochannel is modulated by the eld eect. In a negatively charged SiO2 nanochannel, cations are the majority carriers, which are expected to increase or decrease when subjected to negative or positive electrostatic potentials, respectively. Therefore, the channel conductance will be enhanced at a negative VG and suppressed at a positive VG. From the comparison of the FERD and the control devices, we conclude that it is indeed the asymmetric positioning of the gate electrode that introduces the rectifying behaviour, whose favourable direction is not xed but switchable via a gate potential.

Qualitative analysis. To understand the asymmetrical-gate controlled nanouidic diodes, we developed a qualitative interpretation by looking at the change of the transient dri current of both cations and anions inside the nanochannel immediately aer VCT bias is applied with dierent VG polarities (Fig. 4a). Because the ionic concentration distributions cannot change instantly, they remain the same as the steady state during this transient period. As a result, the diusion current is not considered. The solid and empty bars inside the nano-channels in Figure 4a correspond to the cation and anion concentrations, respectively. The length of the bars depicts the amount of the ionic concentration as well as the dri current (because the dri current is linearly proportional to the ionic concentrations). According to the Kirchhos current law, the sum of currents owing into any node should be equal to the sum of currents owing out of that node, we get J J J J

C C T T

VG<0 VG<0

VG>0

VCT<0

VCT <0

VCT>0

VCT>0

J KC

J CCI

J CCI J CCI

J KT J KT

J KT

J KC

J KC

Depletion

Accumulation

JTCI JTCI

JTCI JTCI

J CCI

E

E

VG >0

J KC

J KT

Accumulation

Depletion

E

E

6

4 VG/VCT= 2/2 V

10 VG/VCT= +2/2 V

VG/VCT= 2/+2 V

5

Concentration (mM)

Concentration (mM)

3

4

K+

K+

K+

2

3

Depletion

2

Accumulation

1

Cl

1

Cl

Cl

0

0

5

4

3

2

1

0

1

2

3

4

5

5

4

3

2

1

0

1

2

3

4

5

x (m)

x (m)

4

Concentration (mM)

Concentration (mM)

VG/VCT= +2/+2 V

8

3

6

K+

Cl

2

4

Depletion

2

Accumulation

1

0

4

0 5 5

4

3

2

1

0

1

2

3

4

5

5

4

3

2

1

0

1

2

3

x (m)

x (m)

Figure 4 | Qualitative and quantitative analysis of the nanouidic FERD. (a) Schematic diagram of cation and anion uxes under various conditions. The solid red and the empty blue bars represent cation and anion concentrations as well as the corresponding current ux. The length of the bars depicts their amount (not to scale). The resulting ion depletion or accumulation taking place in the centre junction of the nanochannelat steady state can be obtained by comparing the amount of ion uxes owing into and away from the dashed box. (b) Quantitatively calculated steady state proles of the ion concentrations along the channel length under different VCT and VG polarities at 1 mM KCl. The averaged K and Cl concentrations is calculated by = =

K Cl K Cl

= . Therefore, the length sum of the two bars in each side of the channel should be equal.

When the gate potential is negative (upper row in Fig. 4a), the cation concentration immediately beneath the gate electrode is enhanced by the electrostatic attraction. With a negative VCT (upper le panel, Fig. 4a), anions are driven towards the Trans side and the cations towards the Cis side, as indicated by the arrows. Under such a circumstance, the amount of the cation/anion ux owing out of the centre junction of the nanochannel is larger than that owing into it. As a result, when the system reaches the steady state again, the ions in the channel centre will be depleted, resulting in a depletion region with a lower conductance. Conversely, with a positive VCT bias (upper right panel, Fig. 4a), the ow directions of the ions are reversed. More ions ow into the centre junction of the nano-channel while fewer ions ow out. This causes an accumulation of ions in the nanochannel and hence an increase of steady state conductance. Therefore, rectication of ionic transport is established and the preferential ionic current ow direction is from the Cis side to the Trans side when VG < 0.

When a positive gate voltage is imposed (lower row in Fig. 4a), the cation concentration near the Cis side of the nanochannel is reduced by the electrostatic repulsion. Following the same analysis, it is obvious to conclude that a negative VCT conducts more ionic current than a positive VCT of the same magnitude. As a result, rectication of ionic current is also established, but the preferential ionic current direction is reversed as compared with the case of a negative gate potential.

Quantitative analysis. The quantitative analysis of the ion transport through the nanopores and nanochannels is generally calculated by concurrently solving the coupled PoissonNernstPlanck,

and NavierStokes equations3640. Since the contribution of electro-osmosis is usually very small compared with the diusive components, the NavierStokes equations can be neglected40,41. We calcu

lated the steady-state distributions of cation/anion concentrations by self-consistently solving the coupled PoissonNernstPlanck equations. To take into account the eld eect arising from the externally imposed gate potential, the electric potential inside the gate dielectric is also solved simultaneously (Supplementary Fig. S4 and Supplementary Methods). The calculated concentration proles under dierent VG and VCT polarities are shown in Figure 4b. The quantitative steady state ion prole agrees well with the qualitative transient analysis in Figure 4a. The ion enrichment and depletion eect at the micro/nano channel interface13 is also observed in the calculated results (x = 4 m). This entrance eect contributes only a small portion of the total conductance and is not responsible for the eld-eect-tunable rectifying property, as shown in Supplementary Figure S3.

2 , where H is the channel height. The surface charge density used in the calculation is s = 2 mC m 2.

y H( ) ( , ) /

/

= / d

2

c x c x y y H

y H

NATURE COMMUNICATIONS | 2:506 | DOI: 10.1038/ncomms1514 | www.nature.com/naturecommunications

2011 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms1514

ARTICLE

1.6 1 M 100 mM 10 mM 1 mM

10 M

100 M

0.90

1.0 1.2 1.2 2.1

0.30

1.4

VG>0

Cis

2.0

2.0

K

Cl

1.5

1.5

Trans

1.2

1.0

1.0

I ( V = 1 .5 V)/I ( V =1. 5 V)

Gate voltage (V)

Gate voltage (V)

1.0

0.5

0.5

0.0

0.0

0.8

VG<0

0.5

0.5

0.6

1.0

1.0

Cis

Trans

1.5

1.5

0.4

2.0

2.0

1.5

1.0

0.5

0.0

0.5

1.0

1.5

6

5 6 1 0

2

4

3

2

1 0

5

4

3

Gating voltage (V)

log (Cb) (Molar) log (Cb) (Molar)

Figure 5 | Effect of the ionic concentrations on the rectifying degree under various gate potentials. (a) Experimental results. The error bars correspond to ten measurements. The inset symbols schematically show the forward direction switched with positive and negative gate voltages. (b) Calculated ratios of cation ux (K) and ratios of anion ux (Cl) via the Cis and Trans side of the nanochannel as a function of gate potential VG and bulk concentration

Cb. The surface charge density used in the calculation is s = 2 mC m 2.

over the cation (anion) ux via the Trans side of the nanochannel as (Supplementary Methods),

r gg

K

Concentration dependence of the rectifying behaviour. Interestingly, the eld-eect recongurable rectifying behaviour shows a strong dependence on ionic concentrations. We recorded the ICT as

a function of VCT for VG ranging from negative to positive values at various KCl concentrations. The rectifying degree (R = I + V/I V, where V = 1.5 V) is plotted as a function of VG (Fig. 5a). The voltage tunability of the FERD devices (that is, the slope of the R-VG plot in Fig. 5a) becomes less prominent at both high and low KCl concentrations. With 1 M and 100 mM KCl solutions, no diode characteristics (asymmetric IV curve) were observed experimentally. The most pronounced voltage tunability happens at intermediate KCl concentrations (1 mM and 10 mM). The qualitative trend that the tunable rectifying property being most prominent at intermediate ion concentrations and weakened at both low and high ionic concentrations is also observed in the numerical calculations (Supplementary Fig. S5). The rectifying degree shows a monotonically decreasing dependence on VG when [KCl] < 100 mM. Consequently, the forward direction of the nanouidic diode can be electrically switched by the application of an external gate potential (inset symbols of Fig. 5a), consistent with the results in Figures 3 and 4.

Another intriguing feature of Figure 5a is that, at a xed gate voltage (for example, 1.5 V), the highest rectifying degree occurs at the intermediate ion concentrations rather than the lowest concentration regime where surface charge-governed ion conduction dominates. This counterintuitive phenomenon has also been observed in the systems of two terminal conical nanopore diodes18,

and nanochannel diodes2022,24,41. However, a physical picture is lacking in understanding this behaviour.

Here we develop a physical model to explain these intriguing phenomena observed in our eld-eect controlled diodes, with the aim to improve the rectication ratios. Referring to Figure 4a and considering the averaged concentrations of K and Cl ions at the two segments of the nanochannel, we denote the relation in the gated area (subscript C) and the un-gated area (subscript T) as [K]C = C[Cl]C, and [K]T = T[Cl]T, where is the cation/anion ratio. The cation/ anion ratio can be determined through the electroneutrality and the law of conservation of mass as (Supplementary Methods), g C C C b b

s s C C

(1)(1)

(2)(2)

= = ++

J J

C

T

K

K

1 1

C

T

1

1

r gg

Cl

= = ++

J

J

C

T

Cl

Cl

1

1

C

T

To observe a rectifying behaviour, K1 and Cl1 are required in the nanochannels (that is, the two solid bars, or the two empty bars in Figure 4a should be unequal in length). In fact, the further the value of K and Cl deviates from unity, the more pronounced is the rectication. Apparently, K or Cl is a function of VG, Cb and s, which is plotted in Figure 5b. At very high concentrations (Cb > 100 mM), the averaged ion concentrations in the nanochannel are dominated by the bulk rather than the surface and, hence, are barely aected by the external potential. Therefore, C1 and T1 (bulk property), which results in K = Cl = 1 and no rectifying behaviour would be observed. At very low concentrations (Cb < 100M), we obtain

K1 and Cl1. However, since K is the majority carrier that dominates the total ionic ux, the rectifying eect is less signicant. At intermediate concentrations (Cb~1 mM), where K1 and Cl1, the most profound rectifying behaviour is expected.

Discussion

The above analysis implies the strategies that can be adopted to improve the gate-tunable rectifying eect. Higher rectifying eect will happen when C1T or T1C. For an intrinsically negatively charged nanochannel SiO2 wall, it is obvious that T1. To achieve C1, the charge polarity beneath the gate electrode needs to be reversed, which requires 0oxVg/dox>2|s|. Take

s = 2 mC m 2, for example, the minimal gate voltage required for charge polarity reversal is about 9 V. This is inaccessible in our current devices owing to the gate leakage and breakdown limitations (Supplementary Fig. S6). However, if a surround-gate architecture (that is, with both the top gate and the back gate) is used, the minimal gate voltage needed to reverse the charge polarity would be reduced by half. In addition, for a better control over cation/anion ratios, a low surface charge density is desirable. This is because an inherent high surface charge density in nanochannels resembles degenerate doping in a semiconductor or high surface state density of a FET,

= + +

( ) /

2 2 2 2

4 4 , where SC = (2s0oxVg/dox)/ehNA and ST = 2s/ehNA. ox and dox are the permittivity and the thickness of dielectric SiO2, respectively. By using the total current continuity at the micro/nano channel interfaces, we can obtain the ratios of cation (anion) ux via the Cis side

2 2 2 2

4 4 and gT T T b b

s s C C

= + +

( ) /

NATURE COMMUNICATIONS | 2:506 | DOI: 10.1038/ncomms1514 | www.nature.com/naturecommunications

2011 Macmillan Publishers Limited. All rights reserved.

ARTICLE

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms1514

a b

Cis Trans

VG VG

Cis

A

Gate (VG)

Gate (VG)

VC

~8 nm

Trans

c

20 VG=1 V VG=0 V

Current (pA)

Current (pA)

201.5 0.0 1.5

20

10

0

10

201.5 0.0 1.5

20 VG=1 V

Current (pA)

201.5 0.0 1.5

VCT (V) VCT (V) VCT (V)

d

100

1 M 10 M 100 M

1 mM 10 mM 100 mM 1 M

I ( V CT=1 V )/I ( V CT=1 V )

10

1

1.0

0.01

1.0 0.8 0.6 0.4 0.2 0.0 0.2 0.4 0.6 0.8 1.0

Gate voltage VG (V)

Figure 6 | Improved eld-effect tunability over the ionic diode property with devices of a dual split-gate structure. (a) SEM image of the dual split-gate FERD devices (W = 1 m11, L = 116 m, H = 8 nm). The scale bar is 100 m. The nanochannel wall is chemically modied to be less charged and hydrophilic. The separation between the two gates is 4 m. (b) Testing congurations for the dual gate FERD device. (c) Typical ICT-VCT curves at different gate voltages VG ( 1 V, 0 V, + 1 V), for a given ionic concentration (1 mM). (d) Measured rectifying degree (referenced at VCT = 1 V) as a function of VG

for KCl concentrations ranging from 1 M to 1 M.

making the electrostatic control of the ionic concentrations in the nanochannel difficult6. By neutralizing the intrinsic surface charge through chemical modications, it would allow for more efficient gate modulation in the Cis side. A separate gate on top of the Trans segment would further permit the cation/anion modulation in the Trans side. Therefore, by placing dual gate electrodes along the channel length, one can have more exibility to tune the rectifying property via the dual gates. Moreover, reducing the nanochannel dimensions down to sub-10 nm would allow the diodes to function at a more realistic ionic concentrations (for example, physiological conditions, ~10100 mM)7. To sum up, an ideal structure for eldeect recongurable nanouidic diodes would be dual split-gates with a gate-all-around structure and a sub-10 nm nanochannels of a neutral surface.

To experimentally verify these predictions, we fabricated another wafer of devices with a dual gate structure (Fig. 6a) and an 8-nm-thick nanochannel (determined by the thickness of the sacricial Cr layer). The surface charge of the as-fabricated nanochannel is estimated as s = 1.6 mC m 2 (Supplementary Fig. S7). This native surface charge density is reduced by sequentially treating the nanochannel wall with 3-Glycidoxypropyltrimethoxysilane and ethanolamine (Supplementary Methods). The resulting nanochan-

nel surface is highly hydrophilic and the modied surface charge is estimated as s = 0.3 mC m 2 (Supplementary Fig. S7). With this surface-modied dual-gate device and an electrical setup schematically shown in Figure 6b, we recorded the ICT as a function of VCT for

VG ranging from negative to positive values at various KCl concentrations. Figure 6c shows typical ICT-VCT curves at dierent gate voltages VG, for a given ionic concentration (1 mM). The rectifying ratio (R = I + V/I V, where V = 1 V) is plotted as a function of VG for seven dierent KCl concentrations (Fig. 6d). The dual-gate device with the chemically modied surface shows a tremendously improved rectication ratio. At VG = 1 V, the rectication ratio reaches ~83 for 1 mM KCl, much higher than that of the single-gate device shown in Figure 5a (a ratio of ~1.2 under the same conditions). We note that the rectication ratio can be even higher under larger reference VCT

biases. Interestingly, the dual-gate device also exhibits a higher rectifying degree at the intermediate ion concentrations under a xed gate voltage (for example, VG = 1 V in Fig. 6d).

In conclusion, we propose and demonstrate a eld-eect recongurable nanouidic diode based on an asymmetric eld eect. This general concept could conceivably be applied to similar thin-body solid-state devices (for example, silicon-on-insulator or semiconducting nanowire). Compared with any existing nanouidic control

NATURE COMMUNICATIONS | 2:506 | DOI: 10.1038/ncomms1514 | www.nature.com/naturecommunications

2011 Macmillan Publishers Limited. All rights reserved.

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms1514

ARTICLE

systems including stimuli-responsive tuning, the electrostatic modulation platform oers the full potential for logical programming of rectied transport of ionic and molecular species. Unlike the nanouidic eld-eect transistor, where only the amount of ions/ molecules is regulated by an electrostatic potential, the FERD can be used to control both directions and magnitudes of ion/molecule transport. FERD represents a fundamentally novel system and may function as the building block to create an on-demand, recongurable, large-scale integrated nanouidic circuits for digitally programmed manipulation of biomolecules such as polynucleotides and proteins.

Methods

Device fabrication. We used a sacricial layer method to produce the nanochannels. However, there is a major dierence in our process. The conventional method etches the sacricial layer rst and then integrates the nanochannel device with a microuidic interface to form a functional device22,24,42. We nd that the nanochannels fabricated by this method are inclined to collapse because of the capillary force in the drying process and/or the pressure exerted during the bonding process. To overcome this problem, we developed a novel bond-followed-by-etch scheme. Briey, 11 thin Cr stripes are patterned on top of an insulating SiO2 layer on a 4-inch silicon wafer. Then a dielectric SiO2 layer (50-nm-thick) is subsequently deposited by plasma enhanced chemical vapour deposition, followed by a rapid thermal annealing process to improve the dielectric quality. Gate electrode is thereaer formed by a double layer li-o process. The accessing reservoirs with supporting pillars are etched simultaneously by reactive CHF3/Ar plasma at a rate of 1 nm s 1. The device is then aligned and permanently bonded with a microuidic polydimethylsiloxane stamp. The Cr etchant is pumped into the microchannel aerwards, which diusively etches the Cr sacricial layer in situ. The end-point detection for the etching process is done by looking at the colour contrast under a microscope. Deionized water is ushed through the microchannel aer the etching process is nished, and then the device is ready to use. As neither the drying process nor the external pressure is applied, the bond-followed-by-etch scheme completely avoids the nanochannel collapse problem. Moreover, the top-down process results in a uniform nanochannel dimensions along the length the channels and across the array.

Electrical characterization. The whole testing procedure is done using an automatic system. The solutions are delivered by pumps (New Era Pump Systems), controlled by a LabVIEW program (National Instruments). The current-voltage measurement is performed using HP4156B semiconductor parameter analyzer and is synchronized with the pump by the same LabVIEW program. Data postprocessing is done with MATLAB (The MathWorks) soware.

Numerical calculation. The steady-state distributions of the cation/anion concentration and the ionic current under dierent VG and VCT biases are calculated by solving the coupled two-dimensional PoissonNernstPlanck equations within the COMSOL script environment. We use three modules in the COMSOL environment: Electrostatics (AC/DC Module), NernstPlanck without electroneutrality for the calculation of K + ions (Chemical Engineering Module) and NernstPlanck without electroneutrality for the calculation of Cl ions (Chemical Engineering Module). The simulation system contains an 8-m-long, 20-nm-high nanochannel that is connected by two 11 m2 square reservoirs. The SiO2 dielectric is 50-nm-thick and 4-m-long, with surface charge density as s = 2 mC m 2 . The length of the nanochannel used for calculation is less than the real device (100 m) to reduce the simulation time.

References

1. van den Berg, A., Craighead, H. G. & Yang, P. D. From microuidic applications to nanouidic phenomena. Chem. Soc. Rev. 39, 899900 (2010).

2. Smith, P. L., Baukrowitz, T. & Yellen, G. The inward rectication mechanism of the HERG cardiac potassium channel. Nature 379, 833836 (1996).

3. Gajar, S. A. & Geis, M. W. An ionic liquid-channel eld-eect transistor. J. Electrochem. Soc. 139, 28332840 (1992).4. Fan, R., Yue, M., Karnik, R., Majumdar, A. & Yang, P. D. Polarity switching and transient responses in single nanotube nanouidic transistors. Phys. Rev. Lett. 95, 086607 (2005).

5. Karnik, R. et al. Electrostatic control of ions and molecules in nanouidic transistors. Nano Lett. 5, 943948 (2005).

6. Karnik, R., Castelino, K. & Majumdar, A. Field-eect control of protein transport in a nanouidic transistor circuit. Appl. Phys. Lett. 88, 123114 (2006).

7. Fan, R., Huh, S., Yan, R., Arnold, J. & Yang, P. D. Gated proton transport in aligned mesoporous silica lms. Nat. Mater. 7, 303307 (2008).

8. Nam, S. W., Rooks, M. J., Kim, K. B. & Rossnagel, S. M. Ionic eld eect transistors with sub-10 nm multiple nanopores. Nano Lett. 9, 20442048 (2009).

9. Joshi, P. et al. Field eect modulation of ionic conductance of cylindrical silicon-on-insulator nanopore array. J. Appl. Phys. 107, 054701, 16 (2010).

10. Maglia, G. et al. Droplet networks with incorporated protein diodes show collective properties. Nat. Nanotech. 4, 437440 (2009).

11. Ali, M., Mafe, S., Ramirez, P., Neumann, R. & Ensinger, W. Logic gates using nanouidic diodes based on conical nanopores functionalized with polyprotic acid chains. Langmuir 25, 1199311997 (2009).

12. Vlassiouk, I., Kozel, T. R. & Siwy, Z. S. Biosensing with nanouidic diodes. J. Am. Chem. Soc. 131, 82118220 (2009).

13. Kim, S. J., Song, Y. A. & Han, J. Nanouidic concentration devices for biomolecules utilizing ion concentration polarization: theory, fabrication, and applications. Chem. Soc. Rev. 39, 912922 (2010).

14. Kim, S. J., Ko, S. H., Kang, K. H. & Han, J. Direct seawater desalination by ion concentration polarization. Nat. Nanotech. 5, 297301 (2010).

15. Cheng, L. J. & Guo, L. J. Nanouidic diodes. Chem. Soc. Rev. 39, 923938 (2010).

16. Wei, C., Bard, A. J. & Feldberg, S. W. Current rectication at quartz nanopipet electrodes. Anal. Chem. 69, 46274633 (1997).

17. Siwy, Z. et al. Rectication and voltage gating of ion currents in a nanofabricated pore. Europhys. Lett. 60, 349355 (2002).

18. Cervera, J., Schiedt, B., Neumann, R., Mafe, S. & Ramirez, P. Ionic conduction, rectication, and selectivity in single conical nanopores. J. Chem. Phys. 124, 104706, 19 (2006).

19. Perry, J. M., Zhou, K. M., Harms, Z. D. & Jacobson, S. C. Ion transport in nanouidic funnels. ACS Nano 4, 38973902 (2010).

20. Cheng, L. J. & Guo, L. J. Ionic current rectication, breakdown, and switching in heterogeneous oxide nanouidic devices. ACS Nano 3, 575584 (2009).

21. Yan, R. X., Liang, W. J., Fan, R. & Yang, P. D. Nanouidic diodes based on nanotube heterojunctions. Nano Lett. 9, 38203825 (2009).

22. Karnik, R., Duan, C. H., Castelino, K., Daiguji, H. & Majumdar, A. Rectication of ionic current in a nanouidic diode. Nano Lett. 7, 547551 (2007).

23. Miedema, H. et al. A biological porin engineered into a molecular, nanouidic diode. Nano Lett. 7, 28862891 (2007).

24. Cheng, L. J. & Guo, L. J. Rectied ion transport through concentration gradient in homogeneous silica nanochannels. Nano Lett. 7, 31653171 (2007).

25. Vlassiouk, I. & Siwy, Z. S. Nanouidic diode. Nano Lett. 7, 552556 (2007).26. Alcaraz, A. et al. A pH-tunable nanouidic diode: Electrochemical rectication in a reconstituted single ion channel. J. Phys. Chem. B 110, 2120521209 (2006).

27. Yameen, B. et al. Proton-regulated rectied ionic transport through solid-state conical nanopores modied with phosphate-bearing polymer brushes. Chem. Commun. 46, 19081910 (2010).

28. Ali, M., Ramirez, P., Mafe, S., Neumann, R. & Ensinger, W. A pH-Tunable nanouidic diode with a broad range of rectifying properties. ACS Nano 3, 603608 (2009).

29. Macrae, M. X., Blake, S., Mayer, M. & Yang, J. Nanoscale ionic diodes with tunable and switchable rectifying behavior. J. Am. Chem. Soc. 132, 17661767 (2010).

30. He, Y. et al. Tuning transport properties of nanouidic devices with local charge inversion. J. Am. Chem. Soc. 131, 51945202 (2009).

31. Schoch, R. B. & Renaud, P. Ion transport through nanoslits dominated by the eective surface charge. Appl. Phys. Lett. 86, 253111 (2005).

32. Stein, D., Kruithof, M. & Dekker, C. Surface-charge-governed ion transport in nanouidic channels. Phys. Rev. Lett. 93, 035901 (2004).

33. Behrens, S. H. & Grier, D. G. The charge of glass and silica surfaces. J. Chem. Phys. 115, 67166721 (2001).

34. Schoch, R. B., Han, J. Y. & Renaud, P. Transport phenomena in nanouidics. Rev. Mod. Phys. 80, 839883 (2008).

35. Sze, S. M. Physics of Semiconductor Devices 2nd edn (John Wiley and Sons, 1981).

36. Daiguji, H., Oka, Y. & Shirono, K. Nanouidic diode and bipolar transistor. Nano Lett. 5, 22742280 (2005).

37. Daiguji, H., Yang, P. D. & Majumdar, A. Ion transport in nanouidic channels. Nano Lett. 4, 137142 (2004).

38. Constantin, D. & Siwy, Z. S. Poisson-Nernst -Planck model of ion current rectication through a nanouidic diode. Phys. Rev. E 76, 041202 (2007).39. Kalman, E. B., Vlassiouk, I. & Siwy, Z. S. Nanouidic bipolar transistors. Adv. Mater. 20, 293297 (2008).

40. Vlassiouk, I., Smirnov, S. & Siwy, Z. Ionic selectivity of single nanochannels. Nano Lett. 8, 19781985 (2008).

41. Vlassiouk, I., Smirnov, S. & Siwy, Z. Nanouidic ionic diodes. Comparison of analytical and numerical solutions. ACS Nano 2, 15891602 (2008).

42. Karnik, R., Castelino, K., Fan, R., Yang, P. & Majumdar, A. Eects of biological reactions and modications on conductance of nanouidic channels. Nano Lett. 5, 16381642 (2005).

NATURE COMMUNICATIONS | 2:506 | DOI: 10.1038/ncomms1514 | www.nature.com/naturecommunications

2011 Macmillan Publishers Limited. All rights reserved.

ARTICLE

NATURE COMMUNICATIONS | DOI: 10.1038/ncomms1514

Acknowledgements

We thank Nitin Rajan for the help of electrical characterization as well as useful discussions, and Michael Power for the help during the device fabrication process. W.G. acknowledges the support from Howard Hughes Medical Institute International Student Research Fellowship.

Author contributions

W.G. conceived the concept and designed the experiments. M.A.R. supervised the study. W.G. carried out the experiment and, together with R.F. analysed the data. W.G. and M.A.R. co-wrote the manuscript and discussed it with R.F.

Additional information

Supplementary Information accompanies this paper at http://www.nature.com/ naturecommunications

Competing nancial interests: The authors declare no competing nancial interests.

Reprints and permission information is available online at http://npg.nature.com/ reprintsandpermissions/

How to cite this article: Guan W. et al. Field-eect recongurable nanouidic ionic diodes. Nat. Commun. 2:506 doi: 10.1038/ncomms1514 (2011).

NATURE COMMUNICATIONS | 2:506 | DOI: 10.1038/ncomms1514 | www.nature.com/naturecommunications

2011 Macmillan Publishers Limited. All rights reserved.