SAMBA

Third Year Progress Report & Final Report

Delivered May 2004

2

1. Introduction The SAMBA consortium consisted of three partners:

INFM (I), including three research units at Lecce, Modena and Viterbo

LEIDEN (NL), with a subcontractor in Delft

OXFORD (UK)

The project has developed along 5 workpackages :

1) WP1: Synthesis, basic biochemical and functional characterization, structural

and spectroscopic investigation of self-assembling on metallic surfaces of

tailor-made metallo-proteins and -enzymes (WP leader LEIDEN).

2) WP2: Design, fabrication and characterization of two and three terminal hybrid

devices for transport studies of self assembled layers and single molecules

interconnecting gold nanostructures (WP leader INFM).

3) WP3: Investigation of FET and biosensor operation of the hybrid-devices (WP

leader OXFORD)

4) WP4: Theoretical modelling of the basic physical processes controlling the

biomolecular transport and devices (WP leader INFM).

5) WP5: Management and dissemination of results (WP leader INFM)

The workplan was organized according to 15 well-defined milestones, distributed

over the 5 workpackages (see the enclosed Technical Annex and Table I).

WP1

M1: (month 6) Synthesis and purification of metalloproteins -

M2: (month 12) Spectroscopic characterization of metalloproteins -

M3: (month 18) Assessment of immobilization and self-assembly of

metalloproteins on gold surfaces (e.g. ordered vs. disordered

self assembling on gold) -

WP2

M4: (month 12) Transport in molecular two-terminal devices with disordered

protein layer (gap below 100 nm) -

M5: (month 20) Transport in molecular two-terminal devices with ordered self-

assembled molecule layer (gap below 100 nm) -

3

M6: (month 24) Molecular three terminal device with ordered self assembled

molecule layer, and field effect transistor characteristics -

M7: (month 24) Sub-10 nm gate for single molecule trapping, and transport

through the single molecule -

M8: (month 30) Sub-10 nm three terminal device for single molecule field

effect transistor -

WP3

M9: (month 12) Relationship between electronic conduction and environment

in disordered layers -

M10: (month 24) Relationship between electronic conduction of the device and

the hydration of the metalloproteins -

M11: (month 36) Relationship between electronic conduction of the device and

the solvent in humid environment

WP4

M12: (month 12) single particle energy levels and wavefunctions for the ligands

around the copper center -

M13: (month 24) modelling the coupling between molecule (copper atom) and

the metallic nanogate -

M14: (month 36) modelling the FET transport and biosensor operation -

WP5

M15: (month 18) organization (or participation to the organization) of an

international workshop on bio-nanoelectronics issues in the

frame of the Phantom network –

At the end of the project all milestones have been accomplished. Concerning the

deliverables related to the various milestones, some deliverables were shifted by few

months relative to the original workplan. This was due to instrument breakdown

and/or maintenance, and to unexpected problems in the implementation of

technological processes. Problems of this nature are well known to everybody

working in the fields of future emerging technologies and nanotechnology because of

the very sophisticated and delicate instrumentation required. At any rate, all the

deliverables were released within the end of the project as reported in the following

table, at the appropriate times.

4

This report summarises the work carried out during the third and last year, and was

prepared following the list of Milestones and Deliverables promised by the different

partners at the end of the project (see Table II for a list of the deliverables expected

for month 36). In addition, we will address some open questions concerning the

completion of milestones M6 , M7, and M13 and some criticisms raised in the 2003

reviewing round (June 2003, Brussels) resulting in a number of additional

experiments not listed in the project initial description.

For details of the results concerning the activity of the first two years of the project we

refer to the last two activity and progress reports (see enclosed files)

2.General assessment of the project 2.1 Final remarks on the results achieved within the project research activity

and collaborations.

There was generally good collaboration amongst the partners, with fruitful

exchanges through project meetings and by e-mail. The collaborations resulted in a

large number of joint publications. A number of them are in press or still in preparation.

The project also offered the chance to form a number of young researchers.

The work performed during the three years of the SAMBA project has given us

the possibility to explore a new exciting field, and to develop and disseminate new

methodologies and new knowledge at the boundary between Nanotechnology and

Biology/Biochemistry. Several of the research lines will have a natural continuation,

and particularly: (i) the realisation of new-generation, single-molecule three-terminal

devices; (ii.) the computation of the electron transfer rate for the dimer proposed by

the Leiden group, combining the computed reorganization energy and transfer

integrals; (iii) the refinement of the model for the protein-layer transistor; (iv) other

specific investigations that will follow-up novel experimental outcomes related to

electron-transfer proteins; (v) ab-initio computation of electric-field-gradient

parameters that are able to shed further light onto the electronic structure of the active

site (collaboration between ONFM-Modena and a Brazilian group external to the

SAMBA partnership) Moreover, the new expertise and know-how that we have

developed has offered the chance to start new research lines and collaborations and to

propose new projects in the area of nanobiotecnology to both the Italian MIUR and

5

the EC. These will include: (i.) lab-on-chip devices for genomics and postgenomics;

(ii.) desing of novel molecular architectures through bio self-assembly;. (iii.)

molecular computation. Moreover as a consequence of the dissemination of the

results in several domestic and international conferences, schools and workshop (on

both biomolecule chemisorption onto solid substrates and electronic detection of

small bioelectronic signals) an international microelectronic Company (ST-

Microelctronics) has started an R&D project and a joint laboratory at the NNL

laboratory (Lecce) to build a new generation of LabOnChip devices.

2.2 Exchange of students/researchers and meetings in the third year

Several short visits (1-2 weeks) were exchanged among the partners. Staff

members and various students from Oxford, Lecce and Viterbo visited the Leiden

group for periods ranging from 1 week to 8 months. One of the students from INFM

finished her PhD degree in Leiden, and also spent a considerable length of time at

Oxford. Conversely, staff and students from Leiden visited Lecce, Modena and

Viterbo a number of times. One PhD student and one postdoc from Leiden have spent

a month in Oxford, and one postdoc from Oxford visited the Leiden group. Repeated

staff visits were exchanged between INFM-Modena and INFM-Lecce.

Three general meetings were organized in Brussels and Lecce, where the

partners met to coordinate the collaboration. In addition, an international workshop

was organized (see enclosed brochure for the list of speakers and titles of

presentations)

A project WEB-page was published on www.samba-project.it and regularly

updated . SAMBA gained recognition as one of the major initiatives in the

framework of the PHANTOMS network, as highlighted both on the Phantoms WWW

and on the information CD.

6

Table I

Del

iver

able

N

o.

Deliverable name

WP

No. Lea

d Pa

rtic

ipan

t

Pers

on

mon

ths

Del

iver

y D

ate

Del

iver

y T

ype

Del

iver

y

dat

e at

the

end

of th

e pr

ojec

t

D1.1 Synthesis and supply of molecules to the technology active partners 1 2 40 6 R 3

D1.2 Assessment of protein immobilization and self organization on different substrates

1 2 49 12 R 10

D2.1 Molecular two-terminal device with disordered protein layer

2 1 30 12 R 6

D2.2 Single molecule two-terminal device with transport characteristics 2 1 30 24 R 30

D2.3 Demonstration of molecular field effect transistor and measurements of the static characteristics

2 1 30 24 D 24

D2.4 Molecular field effect transistor operating with single molecule

2 1 33 30 D 36

D3.1 Testing conductivity of molecular layers under different atmospheres 3 3 27 12 D 36

D3.2 Demonstration of a biosensor based on molecular FET 3 3 30 36 R 36

D4.1 Modelling of electronic states for the groups involved in the active site and in the interaction of the protein with the electrodes

4 1 12 12 R 12

D4.2 Modelling of transport across the molecules interconnected to metallic nanogates

4 1 12 24 D 24

D4.3 Modelling FET operation with ordered self assembled molecular layers and with single molecules, in different atmospheres

4 1 12 36 D 36

D5.1 Technical report and cost statement 5 1 3 12 R 12

D5.2 Organization or contribution to the organization of a workshop on nano-biomolecular electronics, in the frame of the Phantom network

5 1 9 18 O 30

D5.3 Technical report and cost statement 5 1 3 24 R 24

D5.4 Technical report and cost statement 5 1 3 36 R 36

7

Table II

Del

iver

able

N

o.

Deliverable name WP

No.

Lea

d Pa

rtic

ipan

t

Pers

on-

mon

ths

Del

iver

y D

ate

Del

iver

y T

ype

Acc

ompl

ishe

d

D2.4 Molecular field effect transistor operating with single molecule

2 1 33 30 D OK

D3.2 Demonstration of a biosensor based on molecular FET 3 3 30 36 R ?

D4.3 Modelling FET operation with ordered self assembled molecular layers and with single molecules, in different atmospheres

4 1 12 36 D OK

D5.2 Organization or contribution to the organization of a workshop on nano-biomolecular electronics, in the frame of the Phantom network

5 1 9 18 O OK

D5.4 Technical report and cost statement 5 1 3 36 R OK

8

3.1 WP1 (Leiden)

The Leiden contribution to SAMBA concentrated on the development,

production and characterisation of protein and enzyme variants. The design of new

variants was usually based on extensive modeling. Construction at the DNA-level was

achieved by modern state-of the-art recombinant DNA techniques. Expression and

purification of the proteins/enzymes was achieved according to advanced protein

chemistry techniques. Purification protocols were developed or adjusted where

necessary. Characterisation of the proteins/enzymes was achieved by standard

biochemical techniques. Spectroscopic characterisation involved UV-VIS

spectroscopy, mass spec analysis, protein film voltammetry, cyclic voltammetry and,

when necessary, EPR and NMR spectroscopy. Activity was checked by enzymological

and kinetic techniques (stopped flow, NMR). At a later stage of the project

proteins/enzymes were also characterised for their behaviour on solid supports (gold,

mica, carbon) by means of atomic force microscopy (AFM) both under ambient

conditions and in situ.

3.1.1 DEVELOPMENT AND PRODUCTION OF VARIANTS

The group in Leiden, in collaboration with the subcontractor from Delft

University of Technology, has developed, produced and made available to the

partners within the framework of WP1, 2 and 3 the suite of electron-transfer

metalloproteins and metalloprotein enzymes listed in Table I. They were used for

immobilisation on solid surfaces, the study of their electrochemical and electronic

behaviour, and characterisation by scanning probe techniques. Both wild type proteins

and variants in their Cu, Zn, and apo forms have been expressed, purified,

characterized and shipped to INFM Viterbo, INFM Modena, INFM Lecce and

Oxford. In addition fern plastocyanin variants (w.t., F12L, G36P, F12LG36P) were

made available to the group of Prof. Takamitsu Kozhuma at Ibaraki University, Mito,

Japan, and S. antibioticus tyrosinase to the group of Prof. Shun Hiroata, Kyoto

University, to strengthen the visibility of the European research in the field, as set out

in WP5 .

A number of the ET proteins listed in Table 1 are physiological partners of the

enzymes listed in the lower half of this table. This was a consideration when choosing

the proteins for the present study, since emphasis was placed on enhancing the

chances for optimal ET rates at the electrode when using enzymes. One of the issues

9

we wanted to explore was the possibility of preparing a surface with a monolayer of

an ET protein that could act as an optimal surface for the enzyme to react with.

Examples of pairs of partners are haem containing peroxidase/cytochrome c, copper

containing nitrite reductase/peseudo-azurin, methylamine dehydrogenase/amicyanin,

and P700 plus the b6f complex/plastocyanin (Fig. 1 and 2).

To make the proteins suitable for immobilisation it was checked whether the natural S-S bridges in the protein might be used (if present) or whether cysteines should be engineered on the protein surface. To this end the Cys3-Cys26 S-S bridge in native azurin was removed by protein engineering techniques. Conversely, an endogenous S-S bridge was engineered in plastocyanin. Enzymes were made suitable for direct immobilisation onto solid substrates by engineering of surface cysteines or they were produced with the possibility in mind that they could be used as secondary partners of ET proteins already immobilised on a solid substrate (Fig.3).

Figure 1. NiR PsAzu ET couple Figure 2. cd1 nitrite reductase from Pseudomonas aeruginosa

Figure 3. Azurin Au assembly models

10

Table I : ET Metallo-proteins: wild types & variants

Azurin [Az, small cupredoxin]: wild type forms & variants * Wild type forms (Cu, Zn, apo) * Single surface cysteine: Q12C, K27C (Cu, Zn, apo), N42C, S118C * Click-on chemistry cavity mutant H117G * Removal S-S bridges: C3AC26A Plastocyanin[Pc, small cupredoxin]: wild type forms & variants

* Engineered S-S bridge: I21C, E25C (PCSS) * C-terminal extension –TCG (Thr-Cys-Gly) (PCSH) Pseudoazurin [PsAzu, small cupredoxin]: wild type forms & variants * Single & Double surface cysteines: E51C, E54C, E51CE54C Amicyanin [Ami, small cupredoxin]: wild type forms

Enzymes: wild types & variants Copper-containing nitrite reductase [NiR, multicopper oxidase]: wild type forms & variants

* Type-1 site depleted, Type-2 site depleted * Single surface cysteine: L93C, M94C * Click-on chemistry cavity H145A/G, M150G, H306A Heme/cd1 –containing nitrite reductase [cd1NiR, multi heme enzyme] :wild type Nitric oxide reductase [Nor, multi heme enzyme] :wild type Methylamine dehydrogenase [MADH ] : wild type

Others: Azurin antibodies Hotwire linker (1-(11-mercapto-undecyl)-imidazole dimer) Plasmid containing the gene of azurin

11

Finally, azurin and pseudo-azurin were engineered with more than one surface

cysteine to check if immobilisation could be improved when more than one point of

surface attachment would was available. Mutants were designed with different sets of

anchoring cysteines positioned on the side or on the south end of the proteins, thus

allowing more precise control of the orientation and flexibility of the protein.

One of the major setbacks during production of cysteine containing protein

variants is that copper can catalyse non-specific disulfide bridge formation, which

renders purification more difficult. Moreover, expressed protein, depending on the

individual mutant and the batch, may contain as much as 80% Zn in the active site

instead of the more desired Cu. Extraction of Zn ion can only be achieved by means

of unfolding, which results in large losses of protein. To increase the final yield of the

desired form of azurin, the gene of the wild type azurin and a small sub-set of the

cysteine-containing mutants were upgraded from the pUC19 vector to the pET28

vector, which usually exhibits a higher expression level. Purification protocols for

these variants have been developed taking into account the need to control the

monomeric/ dimeric ratio of the proteins. The apo- and Zn forms of the blue copper

proteins were engineered with the specific goal to investigate the effect of removing

or replacing the Cu in the active site with a redox inactive metal.

With respect to this panel of generated protein variants, those with the surface-

exposed cysteines showed a high degree of similarity with the native parent forms in

spectral/ structural behaviour as analysed with UV-vis & EPR spectroscopy and with

X-ray diffraction. In addition their redox properties in solution did not significantly

differ from their respective wild type forms.

3.1.2 CAVITY MUTANTS

Cavity mutants of azurin and nitrite reductase weredesigned to facilitate direct

connection of the active redox site to the electrode by so-called hot wires. Residues

constituting the copper sites are replaced by amino acids creating a gap in the ligand

shell of the copper. In general the cavity mutants need to be reconstituted with an

(external, click-on) ligand to obtain their native-like properties. Interestingly, the

creation of a cavity involving the axial ligand of the copper site, i.e. Met-150 in NiR

or Met-121 in Az, affects the features of the redox sites to a much lesser extent than

the ones involving the surface-exposed His-145 in NiR and His-117 in Az. However,

from the point of view of introducing the most effective (shortest) linker-ligand

12

assembly between the protein redox site and the electrode, the latter mutation, i.e., the

one involving the exposed His, is preferable.

3.1.3 CLICK-ON CHEMISTRY AND HOT WIRING

Electrode-assembly with the copper redox site of metallo-proteins and –

enzymes for amperometric sensing involves the selection and combination of linkers

and of protein variants. The main challenge of constructing such a device is

interfacing the enzyme with the electrode surface. The use of these special coatings

could provide a cushioning effect to the protein molecule. The choice of Self

Assembled Monolayers [SAM] on gold and of the appropriate hot wire to connect the

redox site of the protein with the electrode surface has become the main concern

initially. We have synthesized a number of potential linkers involving peptide or

alkyl-based spacer molecules that enable to connect the redox-enzymes to a metal

surface (see figure 5).

To explore electrode assembly involving hot wiring we choose the azurin cavity

mutant His-117Gly. We synthesized two types of linker molecules for

complementation: (1) one based on the interlinking by alkyl groups of a Cys and a His

residue [CCpep1-dimer] and (2) one with (1-(11-mercaptoundecyl)-imidazole-dimer

[MUI-d]; see Fig.6).

To complement the cavity copper site the linker had to be used in a dimerized

form; otherwise the free thiol groups of the linkers extract the metals from their sites.

The linker titration for the cavity of the His117Gly azurin variant showed only low

complementation for the Ccpep1 linker, whereas the MUI-d could indeed saturate the

copper site with increasing ligand concentrations as shown in Fig.7.

Figure 5. Protein-electrode assembly

13

Hot wiring NiR. As a second test case we chose nitrite reductase (NiR). In

copper-containing NiR (see Fig.1 and 8) a type-1 electron transfer site is present near

the surface. Deeper inside the enzyme is the type-2 catalytic copper site.

Figure 7. Cavity site [15uM of His-117 Gly Az titrated with MUI-d [0-40uM] in 20mM MES, pH6

350.0 400 450 500 550 600 650 700 750 800.00.01200.015

0.020

0.025

0.030

0.035

0.040

0.045

0.050

0.055

0.060

0.065

0.070

0.075

0.080

0.085

0.090

0.095

0.100

0.105

0.110

0.115

0.1200

NM

A

Figure 8. The partial NiR structure: type1 to type-2 (Molscript, Kraulis 1991)

HN

NH

N

NH2

O

NH

OHO

O

NH

NH

N

NH2

O

HN

OHO

O

S

S

NN

S

NN

S

Figure 6. Linker structure of Ccpep1 (left) and MUI-dimer (right).

14

The short distance (12.7 Å) between these sites results in fast electron transfer. We

created a gap in the Cu coordination shell of the type-1 site by replacing one of the

Cu-ligands (His145) by an alanine or a glycine. An external ligand (imidazole)

appears to bind to the oxidised type-1 site (Cu2+), and enables fast electron transfer. In

the M150G/A cavity mutant, the features of the redox site are much less affected in

comparison with other cavity mutants. An X-ray crystal structure of the mutant has

recently been obtained which indicates an unexpected internal reorganisation of the

side chain residues leading to a partial regain of redox activity.

However, the reduced type-1 site shows no affinity for the external ligand,

which hugely increases the type-1 sites reduction potential and prevents electron

transfer to the type-2 site. To circumvent this the external ligand was covalently

connected to the enzyme via a group that reacts with a cysteine introduced at position

93 (Fig.9a). The experimental results show covalent binding of the ligand to the

enzyme and binding to the type-1 copper site (see Fig.9b). Interestingly, at the same

time the catalytic activity of NiR is raised by a factor of 1000.

3.1.4 A NIR SENSOR

The immobilised copper containing nitrite reductase has been shown to function

as nitrite sensor. Recently, we have demonstrated that NiR can be driven to catalyse

the reverse reaction, i.e., the oxidation of nitric oxide (see Fig.10). This activity could

allow us to use the enzyme as nitric oxide sensor as well as nitrite sensor. Thus,

Wavelength 40 50 60 70 80

Absorbance

0.00

0.00

0.00

0.01

0.01

Figure 9a (Left). External ligand modeled into the type-1 site of NiR L93C/H145A. Figure 9b (Right).UV-vis spectrometry of the titration of the NiR variant L93C/H145A with increasing amounts of the linker di-thio-ethyl-imidazole reagent, which complements this cavity type 1 copper site, as can be deduced from theincrease in 600 nm absorption of the azurin copper site.

15

potentially, just one protein immobilisation strategy would allow us to build a sensing

device that can detect two different substrates, simply by changing the assay

conditions and the electronics parameter. This discovery adds a new dimension to the

impact of the SAMBA project.

3.5 AFM OF NIR

Direct immobilisation can be achieved by specific and non-specific adsorption

onto the substrate surface by, for example, formation of Au-S-Cys bond and physical

adsorption respectively. The Leiden group has successfully imaged Nir with AFM

under ambient conditions when it is non-specifically adsorbed on mica and HOPG

substrate (Fig.11). Efforts are currently under way to image Nir immobilised on gold

surface and in physiological conditions. As a next step, the effects of protein

Figure 10: Nitric oxide consumption by pseudoazurin catalyzed by NiR. The y-axis displays the readingsof the Clark electrode converted to nitric oxide concentration. The steady decrease is caused by oxygenleaking in. Pseudoazurin (20 µM) in its oxidized state was added prior to NiR (0.1 µM), and nitrite (2.5mM). The Clark electrode adapts within 5-10 s to changes in nitric oxide concentration. Panel A, pH 8.0;panel B, pH 6.5.

16

flexibility on the electrochemical properties of a cupredoxin will be studied by

conductive AFM.

3.6 OPTICAL DETECTION

Ultimately, it is hoped that we can create a device that generates signals from a single

or a very small number of molecules. We have recently demonstrated that it is

possible to detect stable fluorescent signals from a single molecule of azurin which is

labelled with the fluorescent probe cy5 using maleimide chemistry. The signal can be

switched on or off by changing the redox state of azurin. Efforts are currently

underway to investigate the possibility of immobilising the labelled azurin molecule,

and observing single switching on and off events of the fluorescent signal by

changing the redox state of azurin electrochemically.

Figure 11. Nir on template stripped gold 10 ìg/ml in 20mM MES, pH 6 TM in air

17

3.2 WP2 3.2.1 Completion of M6 : Molecular three terminal device with ordered self

assembled molecule layer, and field effect transistor characteristics (INFM-Lecce)

3.2.1a Protein devices Different molecular devices were fabricated based on: 1. commercial azurin, purchased by Sigma Aldrich, with a copper atom inside.

2. azurins supplied by the Metalloprotein & Protein Engineering Group at the Leiden

University including: i.) recombinant azurin (Az-Cu), again a copper protein but

exhibiting a higher degree of purity; ii.) a modified azurin with a Zn atom inside

(Az-Zn); iii.) apo-azurin (Az-Apo) without any metal atom, in order to assess its

role.

3. reference devices, including empty nanojunctions and devices that were silanised

or completely functionalized but with no added protein.

Protein Field-Effect Transistors based on self-assembled monolayers of copper azurin. For this study, both the source-drain separation and the oxide thickness were 100 nm

and arrow-shaped Cr/Au planar electrodes were employed. Current-voltage

experiments were carried out by using a semiconductor parameter analyzer (HP

Agilent 4155B) in the voltage range between –6 and 6 volts (drain-source voltage), at

room temperature and ambient pressure. The gate voltage (Vg) was changed in order

to investigate its influence on the current (Ids) between the source (s) and drain (d)

electrodes. Protein immobilization on Si/SiO2 substrates was achieved (as previously

described) using a two-step procedure involving (a) the self-assembly of 3-

mercaptopropyltrimethoxysilane (3-MPTS) and (b) the reaction of the free thiol

groups of 3-MPTS with the surface disulfide bridge of Az, which is broken to form a

covalent bond to the silane-functionalized Si/SiO2 substrate. As a consequence, an

oriented monolayer is formed, i.e. a monolayer in which all the proteins have the

same orientation with respect to the substrate, due to the presence of just one disulfide

bond, i.e. a unique linking site, in the protein. Since protein adsorption on surfaces

may lead to denaturation, we examined the protein shape and fold pattern after

immobilization, as described in section 3.2.1b)

18

The current-voltage characteristic (Ids-Vds) of the Pro-FET [1,2] under forward drain-

source bias (Vds) exhibits a low-current plateau at low field and then rises up to

hundreds of pA (see Fig. 1).

The 3D plot of the drain-source current as a function of the drain-source bias (Vds)

and of the gate bias (Vg) is displayed in Fig. 2. The transfer characteristic exhibits a

pronounced resonance centered at Vg=1.25V. In this region, the transconductance

changes from positive to negative values. The peak to valley ratio and the FWHM are

2 and 0.3V, respectively. This feature gradually disappears, typically after some gate

sweeps (Fig. 3). Since the fluorescence experiments in high electric field indicate that

azurin still preserves its structure, we ascribe this effect to metallic contact damage,

induced by the high electric field [3]. The ageing of nanodevices is a general issue

[see also 4, 5] that must be accurately addressed and we have carried out

0 1 2 3 4 5 60

100

200

C

urre

nt (p

A)

Voltage (V)

Cu-Az - Vg= 0.0V Cu-Az - Vg= 1.2V Zn-Az - Vg= 0.0V Zn-Az - Vg= 1.2V Apo-Az - Vg= 0.0V Apo-Az - Vg= 1.2V Silanes Silanes+Buffer

Figure 1 Current-voltage characteristics of devices made with azurin variants for two different gate voltages (0 V and 1.2 V). As a general feature, devices based on Zn-Az or Apo-Az exhibited low current (around 50 pA) with no modulation effect as a function of Vg. These results support the interpretation that conduction in the proteins and modulation in the devices is due to the reversible redox reaction Cu2+ + e- Cu1+, which reconverts continuously the Cu(II) copper oxidized state into the Cu(I) reduced state in adjacent molecules. As a reference, responses are also shown of devices that were either simply silanised or silanised and incubated in buffer without protein.

19

electromagnetic field simulations (results described later), using finite elements

methods, in order to further elucidate its mechanism.

Figure 2 Characteristic of the protein FET. 3D plot of the drain-source current as afunction of the drain-source bias (Vds) and of the gate bias (Vg) measured in the dark andat room temperature. No leakage current was observed to flow between the planarelectrodes and the back-gate (values as low as few pA and a negligible variation with Vgup to 8 V). A pronounced resonance centered at Vg=1.25 V is present (see also thetransfer characteristic at Vds = 5.5 V in the projection). In this region, thetransconductance changes from positive to negative values. The peak to valley ratio andthe FWHM are 2 and 0.3 V, respectively. This feature gradually disappears after somecycle of measurement due to the aging of the molecular layer

20

From an electronic viewpoint, the device switches from a n-MOS FET behaviour

before resonance to a p-MOS FET after resonance. This is a key result because it

would allow to exploit the advantages of a complementary logic, fabricating both p-

type and n-type devices on the same chip. For the implementation of an inverter, a

unipolar technology requires a load resistance, whereas a complementary logic -

incorporating both p-type and n-type transistor – would overcome such a limitation.

This would result in (i) a decrease of the logic gate occupation area (reduced to the

transistor size scale); (ii) a reduction in the fabrication complexity, since in integrated

circuit technology accurate resistors are harder to make than capacitors and transistors

(a) Vds= 5.5V

Cu1+ n-MOS

Cu2+

p-MOS

(b

(c

Figure 3 (a) Transfer characteristic of the protein FET. A pronounced resonance with agaussian-like shape centered at Vg=1.25 V is present. In this region, the transconductancechanges from positive to negative values. The peak to valley ratio and the FWHM are 2 and0.3 V, respectively. This feature gradually disappears after some cycles of measurementprobably due to the metallic contact damage (the red, green and blue curves were recorded insequence). Insets: Possible electronic applications of the azurin device (b) A n-MOS inverterusing a resistive load; (c) a CMOS inverting amplifier made with the protein FET. Thiswould have the advantage of consuming power only when switched.

21

and (iii) a reduction of power consumption because, as opposed to unipolar inverters

that consume power in the low state, CMOS consume power only when switching.

We ascribe these interesting results to the unique transport mechanism of our

biomolecular devices. Redox protein devices are very different from standard

inorganic semiconductors and conventional organic devices. Silicon MOSFETs and

thin-film transistor (TFTs) are based on a gate field modulating the width and the

conductance of a semiconducting channel, whereas a proposed mechanism for carbon

nanotube FETs is the Schottky-barrier dominated transport [6]. In proteins, the long-

range electron transfer (ET), which represents one of the key processes in

photosynthesis and respiration, occurs between a donor (D) and an acceptor (A) site.

Two different models for ET in proteins have been proposed [7], i.e. a superexchange

mechanism (consisting of direct quantum tunneling between the donor and acceptor)

and a sequential (incoherent) hopping between adjacent sites. The main factors

influencing the ET rate are: (1) the distance between the two redox centers (in

electron tunneling the ET rate decreases exponentially with distance, whereas in

hopping it is inversely related to the distance); (2) the nature of the micro-

environment separating the donor and the acceptor (which mediates the virtual state or

provides intermediate states, respectively), (3) the reorganization energy λ, i.e. the

energy required for all structural adjustments (in the reactants and in the surrounding

molecules) that are needed for the transfer of the electron [8] and (4) the driving force.

In particular, in the case of azurin, the essentially unchanged copper site geometry in

the Cu(II) and Cu(I) state minimizes the reorganization energy λ and favours fast

electron transfer.

The transport of electrons through systems containing redox sites (and thus in the Pro-

FET) occurs via electron hopping from one reduced (Cu(I)) molecule to an adjacent

oxidized (Cu(II)) molecule [9] (see inset of Fig. 4), which must behave as a redox

pair. For electrons to flow, therefore, both reduced and oxidized azurin must be

present and their relative proportion determines the ET rate. Let us introduce, in

analogy to solid-state physics, two functions 2Cuf + and 1Cuf + = 1 - 2Cuf + that

provide the probabilities that a copper site is in the Cu(I) and Cu(II) state, respectively

(i.e. the fraction of reduced and oxidized azurins in the layer). If Wim is the inter-

molecular transfer rate, the overall electron transfer rate WET takes the form:

22

WET(Vds,Vg) = Wim (Vds) 2Cuf + (Vg) 1Cuf + (Vg) = Wim 2Cuf + (Vg) (1- 2Cuf + (Vg))

where we have assumed that Vds and Vg influence Wim and 2Cuf + , respectively. In

other words, we propose that the inter-molecular transfer rate Wim depends on the in-

plane driving force, which is related to the bias applied between drain and source

electrodes (hopping mechanism). On the other hand, the azurin redox state is thought

to be regulated by Vg: thus, the higher Vg, the greater the fraction of oxidized azurins.

Figure 4 Three-dimensional crystal structure of the blue-copper protein azurin containing the central

Cu ion as redox site; cross-section (not to scale) and transport mechanism of the protein FET. The

site geometry of the copper site is a distorted trigonal bipyramid one. The disulfide bridge (Cys-3 –

Cys-26) opposite the copper atom, is exploited to induce chemisorptions of azurins on silane-

functionalized substrates. The field effect transistor consists of a protein monolayer connecting two

arrow-shaped Cr/Au electrodes on a SiO2 substrate. An Ag back-electrode forms an ohmic bond to

the silicon that acts as gate. As a consequence of chemisorption, the proteins sit on the surface with

the electron transfer pathway coupling the copper atom and the disulfide bridge almost

perpendicular to the substrate. In our model, transport is based on sequential electron hopping

between one reduced azurin (blue copper ion in the inset) to an adjacent oxidized one (red ion in the

inset). The gate (vertical) field influences the oxidation state of the redox site, originating the

resonance

23

This is reasonable since the protein is chemisorbed on the SiO2 surface with a known

ET route – joining the disulfide bridge to the copper site – almost perpendicular to the

surface. The probability that any two adjacent azurin molecules form a redox pair is

proportional to 2Cuf + (1- 2Cuf + ). This product is maximal when 1 – 2 2Cuf + = 0 or

2Cuf + = 0.5, i.e. when the two populations of azurin molecules are present in equal

amounts. At a particular value of Vg the fraction of oxidized molecule will equal that

of reduced molecules and therefore ET will be maximal. At higher (lower) Vg values,

the fraction of oxidized (reduced) molecules will be higher than that of reduced

(oxidized) molecules, resulting in a lower ET rate. This model explains the

phenomenon of resonance that we observe in the Pro-FET (Fig. 5) and opens the way

to a mathematical simulation of the device behaviour. Statistical considerations about

the equilibrium of the redox reaction involving the copper atom lead to a Fermi-like

functional dependence of 2Cuf + on Vg, where the midpoint potential (i.e. the voltage

at which the molecules are half oxidized and half reduced) is tuned by Vg, resulting in

a calculated probability for redox pair formation having a resonance-like shape. This

model is consistent with the interpretation of the redox peak in cyclic voltammetry

curves [10,11] and in electrochemical STM experiments [12] performed on azurins

chemisorbed on Au(111) substrates. To quantitatively simulate the device and to

identify the region involved in the ET process, Monte Carlo simulations based on this

model are currently in progress, taking into account the copper-mediated electron

transfer and the contribution from the intrinsic conductivity of the protein skeleton

(see section 3.4).

Comparison with devices based on mutant proteins The key role of the copper atom in electron transfer is further supported by a

comparison with the current-voltage curves measured in devices implemented with

two azurin variants: the first with the Cu atom replaced by a Zn atom (Zn-Az), the

second without metal atom (Apo-Az). In Fig.1, the I-V characteristics for two

different gate voltages (0 and 1.2 V, the latter corresponding to the resonance in the

transcharacteristics of the Cu-Az devices) are displayed for each protein device. The

current flowing through devices based on the two variants are significantly lower and

no modulation is observed between -6 and 6 volts. We ascribe the low conductivity of

these devices to the absence of a metal atom capable to mediate electron transfer,

24

since Apo-Az contains no metal atom, while Zn has only one oxidation state (Zn+2).

As a reference, the I-V characteristics of devices silanised or silanised and incubated

in buffer without protein are also reported in the same figure. In these cases the

currents are comparable with the noise level of empty devices.

Device aging and lifetime To complete the discussion, the issues of device ageing and lifetime have to be

addressed as a rule for molecular electronics, since the need for high throughput and

long-term stability is critical for transferring molecular electronics prototypes to

production. Device failure during operation can be accounted for by a degradation of

the molecular layer and/or a damage of nanojunctions. Even though our preliminary

data on protein fluorescence in an electric field demonstrate that the protein preserves

its structure under high fields, such fields usually lead to a damage of nanojunctions.

We observe two main phenomena concerning device damage: (1) blowing up of

nanojunctions (Figure 5a) or (2) the formation of various aggregates on the contacts

(Figure 5b). When we apply a bias, a very high electric field arises between the

electrodes (Figure 5c), because they are placed few tens of nanometers apart and there

is a drastic change in the dielectric constant at the interface between SiO2 and the

metal. This very intense electric field causes device damage.

A relationship was found between the contact separation and breakdown voltage,

(Fig. 6). The linear regression, with a slope of 1.1·105 V·cm-1, seems a valid model,

as expected for field-related phenomena, except that no intersection is predicted with

the origin. Furthermore, some devices were tested with azurin molecules in the gap,

some others without, and the presence of proteins was found to favour breakdown,

with a 50 % occurrence versus the 20 % of molecule-free devices. Finite elements

simulations were performed in order to investigate whether the breakdown could be

caused by the electric field or the current density (thermal effects can be excluded,

due to the characteristics of the phenomenon and the involved orders of magnitude).

The simulated geometry was pruned, with respect to the physical one, at a length of

few hundreds nanometers along the x-axis (joining the two tips). The silicon substrate,

underneath the SiO2 layer, was also reduced in thickness, due to the limited

penetration depth of current inside it. Dirichlet boundary conditions of fixed potential

were imposed on the outer borders of the tips, while Neumann conditions were set

anywhere else. No molecules were simulated, in this first model, in order to keep it as

general as possible; on the other hand, as mentioned, their presence is not necessary

25

for the breakdown. Contact separations were simulated, in the range 50 to 350 nm,

associating the corresponding observed breakdown voltage to each separation value,

and looking for constant characteristics in the values of local quantities, such as the

electric field and current density. The electric field magnitude is relevant outside the

metal, i.e. in the air gap; the current density, in turn, is higher inside the metal

contacts. For this reason, as the simulation output, the current density was averaged

(a) (b)

(c)

(d

(e)

Figure 5 After electrical characterization, all devices were inspected by scanning electronmicroscopy in order to assess their state. We observde two different phenomena: (a)nanojunctions blow up with a consequent insulator behaviour; (b) the formation ofaggregates between the tips, exhibiting an ohmic behaviour in the electrical characteristicsand probably consisting of a metal cluster. These phenomena are related to the high field inthe tip region when a bias is applied. Map of the electric field (c) and the current density (d)as obtained from electromagnetic field simulation using finite elements methods. Electrodesfail where the hotspots in the current density are located, near the final part of the tips. (e)To solve this problem, we are now fabricating tips with a different shape, avoidingrestrictions in the geometry.

26

over its hot-spot (the high-intensity area, containing the highest quartile as to the field

intensity distribution, inside the metal and close to the tip) as well as over the whole

tip, while the electric field was averaged over its own hot-spot, immediately outside

the metal tip. Due to the shape of the tips, that is the only place where a strong

interaction may take place between the field and the metal; consequently, there is no

point at taking the average electric field over the whole air gap.

Figs. 7 and 8 report these three quantities versus the contact separation: both the

electric field and current density, averaged over their respective hot-spots, decrease in

similar, nonlinear fashions, as the separation increases. The reason is the shape of the

field distribution: as the contacts get farther, the shape of the field and current density

near each tip tends to become more and more independent of the presence of the other

tip. The ‘knee’ between the more coupled and less coupled zone seems to be around

100 nm. Conversely, the average current density over the whole tip has a much more

constant trend; the calculated standard deviation is only 13.8%, versus more than 30%

found for the former quantities. In other words, the phenomenon seems to be

controlled by current density, and to need a minimum area of sufficiently

Figure 6 Empirical dependence between the activation voltage and contact separation, showing a linear regression

27

Figure 7 Average ‘hot-spot’ field, obtained by the simulations: for large separations, it tends toa constant value, confirming the linear relationship between voltage and separation, while withvery close tips the dependence becomes highly nonlinear

Figure 8 Average current density on the hot-spot (triangles) and over the whole tip (asterisks), as obtained by the simulations; the former behaves like the electric field, while the seconds is nearly constant

28

high density to be triggered; a small, though intensive hot-spot, is not enough.

Moreover, the ‘threshold’ values are reached just by applying the observed

breakdown voltages to the corresponding separation amplitudes. One can then extend

the breakdown law to smaller contact separations, just by finding what voltage must

be applied to any imposed gap, in order to obtain the needed average current density.

Fig. 9 shows the result of such procedure: for smaller gaps, the law becomes

definitely nonlinear, with the breakdown voltage decreasing more and more quickly

as the gap decreases: this phenomenon is due to the increasingly strong coupling

between the contacts, as already remarked, and shows that the apparently lack of

intersection with the origin is not real. On the other hand, for separations larger than

100 nm, the law tends to linearity, accounting for the behaviour observed in the

experimental data.

If breakdown has found a phenomenological description, much remains to be clarified

about its very nature. Electromigration is to be ruled out, at least in the form we know,

for the current is orders of magnitude lower than it usually is in this phenomenon. On

the other side, metal is clearly taken away from the contacts, and possible field-related

Figure 9 Activation law, as predicted from the simulations: the regime, linear for large separations, switches to nonlinear around 100 nm

29

explanations (e.g. field emission) contrast with the given description. Finally, the role

of azurin is still unclear: its capability of favouring breakdown might be attributed to

the increased conductivity of the inter-tip air channel – the same at the basis of protein

devices – but this is still to be ascertained and clarified in its dynamics. However, we

emphasize that the tip-geometry, though useful to realize very close electrodes,

introduces critical breaking points, related to the regions where the electric field is

higher (tips). An improvement can be obtained if discontinuities in the electrode

geometry are avoided to prevent field-induced damage. In order to reduce these

problems, we have started to fabricate electrodes with a trapezoidal shape (Figure 5d).

Moreover, we also fabricate more complex circuits (Figure 10) in order to study the

conductivity of several molecular devices in parallel. Since most of the designs and

prototypes for molecular electronics circuits and devices involve an interface between

a molecule and metal electrodes, the effects described must be carefully addressed

before molecular electronics can become of practical applicability.

3.2.2b Investigation of the structural stability of azurin monolayers in air

for device implementation We have shown that the azurin ET activity can be exploited for the realization of solid

state biomolecular transistors working in air and at room temperature. On the

biochemical and biophysical side, an important question is whether the protein

structure and function are conserved under non-physiological conditions. This is a key

technological point for the realization of such biodevices, but it is also an important

Figure 10 Nanojunctions for the study of the conductivity of several molecular devices in parallel.

30

issue from a fundamental viewpoint. In the Second Year Report, we provided a

complete and direct analysis of this issue, showing that the immobilization of azurin

in the solid state under non liquid conditions, by means of a specific chemisorption

process, does not result in protein denaturation. To this purpose, we carried out a

systematic investigation on the azurin monolayers, concluding that the immobilized

proteins do not undergo denaturation even after removal of the aqueous solvent. In

this Report, we show additional biophysical characterization of the protein

monolayers, demonstrating the possibility of inducing resonant tunnelling in azurin

even under ambient condition (thus demonstrating the retention of the proper ET

functionality of azurin molecules in air), and carefully testing the biomolecular films

in device-like conditions, namely (i) under high external electric fields and (ii) under

prolonged exposure to ambient conditions (ageing effects), which are crucial issues

for the reliability of protein-based devices.

ET functionality of azurin molecules under ambient conditions

The main focus of our study was to ascertain if immobilized azurin preserves its

electron transfer properties, especially in ambient conditions. STM permits to achieve

a very high resolution and to probe the electronic properties of analyte biomolecules,

providing crucial information on the integrity of the copper active site, and, hence, on

the functional state of the protein. STM experiments were carried out, both in buffer

solution and in air, on Cu azurin molecules covalently immobilized onto Au(111)

substrates via the disulphide bridge Cys3-Cys26.

Au(111) grown on muscovite mica sheets (Molecular Imaging) were flame-

annealed to obtain recrystallized terraces and immediately incubated in 70µM azurin

(50mM sodium acetate, pH 4.6) at 4°C for 1.5 h. After incubation, the samples were

gently rinsed with ultrapure water, dried with a soft jet of pure nitrogen, and

immediately imaged by STM in buffer solution or in air.

In Fig. 11, we show a high resolution three-dimensional view of azurin molecules

adsorbed onto Au (111). The proteins were clearly detectable as bright spots with a

lateral size of 4-6 nm, in good agreement with the reported crystallographic data.

In order to understand if azurin electron-transfer properties were preserved after

immobilization, we imaged the sample as a function of the bias potential applied

31

between tip and sample both in buffer solution and in air. Fig. 12 shows a sequence of

typical STM images acquired in buffer solution at different bias voltages. Importantly,

STM detection of azurin at the surface was strongly dependent on the applied bias

between the tip and the gold substrate, with protein images fading rapidly above and

below an optimal gap voltage, which enhances the tunneling through the protein

(result underlined by arrows in the figure).

FIG. 11. Three dimensional STM view showing azurin molecules adsorbed on Au (111) recorded in air. Bias voltage, -700mV, current setpoint, 1nA. Scan area, 130x130 nm2; vertical range, 0.6 nm. Scan rate 4Hz.

a b

c d

FIG. 12. STM images recorded at bias voltage of -400mV (a), -200mV (b), -20mV (c), +50mV (d)under 20mM HEPES buffer pH 4.6. Current setpoint, 1nA. Scan area, 170x170 nm2 ; verticalrange, 2nm. Scan rate 5Hz.

32

Similar results were found in ambient condition (Fig.13), although the value of the

optimal gap voltage was slightly different. This dependence of tunneling on the

applied voltage indicates the lack of gross molecular rearrangements upon

immobilization onto solid state in air. It is worth noting that this STM study

represents the first experimental demonstration of the possibility of inducing

tunneling through azurin in air.

In order to quantify how tunneling conditions are influenced by the applied bias, we

measured the apparent height of individual azurin molecules, both in buffer solution

and in air, by taking a line scan profile over groups of proteins and then averaging the

results within the same image, i.e. for each bias. The apparent height of proteins as a

function of the bias potential is plotted in Fig. 14 for both conditions. In both graphs,

a change of the measured height values of the protein is clearly visible, with values

ranging from 0 Ǻ (no protein visible, poor tunneling conditions) to 2.5 Ǻ (bright spots

corresponding to good and stable tunneling conditions across the protein). A

resonance behaviour is found, with a peak around -1 V for tunneling in air, and –0.5

V for the liquid environment. These trends can be ascribed to the occurrence of on/off

resonance conditions in the tunnelling process through the protein active site (via an

electron pathway joining the Cu atom to Cys-26, which is covalently bonded to the

Au surface via the S atom) as the tip Fermi-level position is moved on the energy

scale with respect to the molecular levels. Thus, proteins were better displayed when

the tip Fermi level was aligned with the azurin redox midpoint (defined as the

a b

c d

FIG. 13. STM images recorded at bias voltage of +100mV (a), +300mV (b), +600mV (c) and +800mV(d) in air. Current setpoint, 0.5nA. Scan area, 150x150 nm2 ; vertical range, 1nm. Scan rate 2Hz.

33

potential at which the concentration of oxidized protein equals the concentration of

reduced one). Otherwise, the proteins become hardly visible in the images, with a

progressive decrease in apparent height. The different voltages for the air and buffer

measurements are only partially due to the relative dielectric constants (εr) of the two

media (air and buffer), and can be mainly ascribed to the different tunneling current

values in the two experiments (0.5 and 1 nA, for ambient and liquid conditions,

respectively), which fix a different distance between the tip apex and the sample

3.0

2.5

2.0

1.5

1.0

0.5

0.0

Hei

ght (

Å)

-1000 -800 -600 -400 -200 0Bias (mV)

Azurin in buffer

3.0

2.5

2.0

1.5

1.0

0.5

0.0

Hei

ght (

Å)

-1500 -1000 -500 0 500Bias (mV)

Azurin in air

a

b

FIG.14. Analysis of the apparent height of individual Az molecules as a function of the bias voltage: a) in HEPES buffer; b) in air. The height was measured by taking a line scan profile over groups of proteins and then averaging the results within the same image, i.e. for each bias. The tunnelling current was set to fixed values while changing the bias.

34

surface. By increasing the current, the tip is closer to the sample and the electrostatic

field felt by the protein is larger. This experimental evidence suggests that the charge

distribution on the protein surface remains the same both in air and in a liquid

environment (likely due to the retention of the hydration shells) indicating structural

robustness of the protein in both cases. These results are in agreement with similar

experiments performed by in situ EC-STM on gold-adsorbed azurins, in which (using

a bipotentiostatic control) the substrate potential was changed with respect to a

reference electrode, while the tip-sample bias was held constant.

In conclusion, we have demonstrated for the first time the possibility of inducing a

tunneling current through azurin in air. Moreover, we have demonstrated that

chemisorption of azurin in the solid state does not alter its conformation and the redox

site structure even after the removal of the aqueous solvent, and is therefore likely to

preserve its ET function(s). Such an important result discloses very interesting

perspectives for the development of hybrid nanodevices operating in non-liquid

environments.

Azurin for biomolecular electronics: a reliability study

In this section data are presented on the resilience of the metalloprotein azurin to high

electric fields and ambient conditions, which are crucial issues for the reliability of

azurin-based devices. Concerning the effect of electric fields, two models are

discussed, which agree in indicating an unexpectedly high robustness. The predictions

were confirmed by experiments in device-like conditions: no structural modifications

occur even after a 40-minute exposure to tens of MV/m. Ageing was then investigated

experimentally, at ambient conditions and without field, over several days. Only a

small conformational rearrangement was observed over the first ten hours, followed

by an equilibrium state.

i) Conformational properties of azurin in high external electric fields As shown in this report, the metalloprotein azurin can been successfully employed in

rectifying (diode-like) and modulating (transistor-like) biomolecular devices, its key

properties being the capability of exchanging electrons through its active site, via

redox reactions. Yet, two major objections are frequently raised against electronic

application of biomolecules. First, the high electric fields applied in nanometric

devices (in the order of 107 V/m) are often indicated as potential sources of rapid

35

molecular degradation. Secondarily, the structural stability of biological

macromolecules to ambient conditions (in air) is critical, and potentially affects the

durability of devices. In this study, the reliability of azurin-based devices is

investigated in both respects.

An estimate of the inner electric fields in azurin was derived by Car-Parrinello

calculation of the potential (performed by the INFM-Modena Group). The field was

then calculated as the spatial gradient of the potential, on the same computational

grid.

Fig. 15 reports the cumulative frequency distribution of the field magnitude, over

the entire intra-molecular volume. The field magnitude is larger than 107 V/m over

nearly the whole volume, and larger than 108 V/m in about 40% of the space, with an

average magnitude |E| = 2.4 · 109 V/m and a maximum value |Emax| = 6.5 · 1010 V/m.

The three field components have similar average magnitudes, namely |Ex| = 8.21 · 109

V/m, |Ey| = 8.23 · 109 and |Ez| = 8.25 · 109. Hence the observed resilience to the usual

fields of nanometric devices is definitely plausible, since the inner forces are orders of

magnitude larger. In Fig.16, the same magnitude (|E|) is sketched along three x-, y-

FIG. 15: Cumulative frequency distribution of the electric field magnitude inside the molecule (semilogarithmic scale). The relevant range is from 106 to 1010 V/m.

|E| (V/m)1010

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8 0.9

1

10 5 10 6 107 108 109 10 11 0

Cum

ulat

ive

rel.

freq

uenc

y

36

-3 -2 -1 0 1 2 30

0.5

1

1.5

2 x 1010

x [nm]

|E| [

V/m

]

-3 -2 -1 0 1 2 30

0.5

1

1.5

2 x 1010

y [nm]

|E| [

V/m

]

-3 -2 -1 0 1 2 30

0.5

1

1.5

2 x 1010

z [nm]

|E| [

V/m

]

FIG. 16. Electric field magnitude profiles, taken along the x, y and z axes going through the protein geometric center. The field is highest in the inner part of the molecules, with peaks of the order of 1010 V/m.

37

and z-parallel axes, intersecting in the geometric center of the protein; the portion of

each axis falling inside the inner volume of azurin is between -2 and 2 nm.

The same qualitative conclusions can be reached by resorting to a much simpler

analytical model, based on the approximate global geometry of azurin and its strong

dipole (≈150 D). In particular, approximating the globular-shaped azurin to a sphere,

with uniform polarization p, a well-known result of electrostatics states that the

magnitude of the inner (uniform) electric field is:

03εpE =

where ε0 is the permittivity of vacuum. p, integrated over the volume, must yield the

known dipole moment P; hence:

30

3 4)3/4( RPE

RPp

εππ=⇒=

where R is the radius of the sphere; assuming R = 2 nm, one obtains |E| = 2.8 · 108

V/m. This value, lower than that from Car-Parrinello by one order of magnitude, is

explained by the uniform orientation of the field: in this simplified model the field

sums up concordantly throughout the volume, to give the total dipole moment. In

contrast, the true field components have variable orientation; as a proof, one can

determine their algebraic average value from the Car-Parrinello calculation: Ex = -3.3

· 106 V/m, Ey = -3.3 · 106 V/m and Ez = -2.4 · 106 V/m, which is two orders of

magnitude lower than the average magnitude calculated above. For this reason, the

analytical model is conservative; this is probably the general case, with uniform-field

approximations. The dipole moment of azurin has been determined theoretically;

however, when the geometry and dipole are experimentally known, a simple model

like the one presented here may be useful for gross evaluation of molecular resilience,

avoiding cumbersome theoretical calculations.

The theoretical predictions agree in indicating that the effect of device-level electric

fields is likely to be non-destructive. As an experimental verification, we have studied

the folding properties of azurin under an external electric field, by intrinsic

fluorescence spectroscopy. This issue was investigated by means of interdigitated

electrodes, fabricated using standard photolithographic techniques (Fig. 17). The

38

realized structure consists of 500 interdigitated lines of 1 µm width and 1 µm spacing,

resulting in an active area of 1 x 1 mm2. The electrode geometry was designed so as to

allow both the application of high electric fields to protein molecules and the

detection of a the very weak fluorescence signal over a large area. Intrinsic

fluorescence of P. aeruginosa azurin is due to a single tryptophan residue (Trp48)

[13]. Upon ultraviolet excitation, Trp48 exhibits an unusual “blue” emission (λmax

≈306-308 nm), owing to the highly hydrophobic microenvironment surrounding it

[13, 14]. In the native state, azurin, apoazurin (i.e., without Cu2+) and other metal

derivatives, such as Zn2+, exhibit identical fluorescence spectra, accounting for the

lack of any structural change besides the metal site, as also shown by crystallographic

data [14,15,16,17]. However, since the different derivatives exhibit large variations in

the fluorescence quantum yield, due to a strong quenching mechanism by the metal

ion [18,19], in this study we have used the apo-protein. As shown in Fig. 18a, the

emission spectra of apoazurin in the solid state and under high external fields exhibit

the same line-shape of the free native protein in buffer, and no emission shifts are

detectable. Since the intrinsic fluorescence on azurin is very sensitive to small

perturbations of the protein fold [20], this result indicates that the presence of such

electric fields does not affect the overall fold pattern, since the tryptophan residue

remains embedded in the same hydrophobic environment.

The retention of the native-like conformation in the protein films was also

supported by the observed independence of their emission spectra from the excitation

wavelength (not shown). This is an important point, since Az photoluminescence is

not affected by λexc if the protein is in the folded state, whereas its spectral line-shape

may be strongly influenced by the excitation wavelength if the protein undergoes

conformational transitions. Importantly, the excitation spectrum (PLE) was also

FIG. 17. Optical micrograph of the interdigited electrodes used in the fluorescence experiments.

39

unchanged upon electric field application (Fig. 18b), thus demonstrating the absence

of any relevant perturbation in the physico-chemical conditions of the chromophore

microenvironment, and confirming that such field intensities do not interfere with the

conformation of the native protein. Moreover, it is interesting to note the clear

retention of structured luminescence spectra (both PL and PLE) by solid state films,

which are peculiar features of the native Az conformation (in its native state, Trp48 is

shielded in the rigid and highly hydrophobic core of the protein). This specifically

contrasts with typical broadband, red-shifted emissions of unfolded azurins, which are

completely devoid of any spectral structure.

In the case of azurin, the investigation of the microenvironment surrounding

Trp48 is of particular relevance, since this residue is thought to play an important role

in the long-range ET processes through the molecule [13; 21]. In addition, this results

is important also because it was obtained with the apo derivative, which is

characterized by a lower structural stability with respect to the wild type copper

protein [20]. Hence, our data indicate the lack of gross molecular rearrangements even

under these extreme experimental conditions.

Although the protein inner field intensities calculated above (up to 1010 V/m) may

sound unrealistic, our values are in very close agreement with recent studies [22,23],

in which the intriguing result that the density of the first hydration shell is higher than

that of bulk water [24] was theoretically elucidated in terms of “electrostriction” of

the water molecules by the huge electric field values (~109 V/m) at the surface of the

biomolecules. In addition, molecular dynamics simulations by Xu et al. on trypsin

inhibitor [25] have demonstrated that field strengths in the range of 108 V/m do not

alter the overall structure (or temperature) of the protein (such fields are shown to be

within thermal fluctuations), and only fields higher than 109 V/m can induce

significant structural changes [25]. All these considerations agree with our

experiments in indicating that the conformational state of the protein was not

significantly perturbed with respect to the native state upon external field application.

In conclusion, the available evidence suggests that: i) proteins in the solid state are

capable of maintaining the tightly bound hydration shells and a native-like

conformation, even after solvent removal and application of high external fields; ii)

this conclusion may appear less surprising if the huge intensity of the protein inner

fields is taken into account.

40

ii) Aging effects on azurin films in air

To investigate the second reliability issue (ageing of azurin in air, at ambient

conditions), experiments were performed in a similar fashion to the previous section

except that no electric field was applied. Experimental evidence has already been

presented that immobilization of azurin in a solid state monolayer in air does not

necessarily lead to protein denaturation (see Second Year Report and [26]). Now the

ageing of azurin was monitored over periods longer than 600 hours. Fig. 19 reports

the emission maximum (λmax), full width at half maximum (FWHM) and integrated

emission, all as a function of time, from the solvent removal instant: as seen, solid

state proteins undergo some conformational rearrangement. Such rearrangement is

however of modest entity: (Fig. 19a), since there is only a slight red-shift of azurin

PL, which brings λmax to a maximum of 311÷312 nm, well beneath the typical

broadband “red” emission of denatured azurin (λmax ≈ 355 nm). Most of this

modification takes place within the first 7 hours: during this period, a small

conformational transition is observed, after which a further slight rearrangement

slowly takes place within the first 100 hours. This behaviour emerges from all three

spectral parameters, which exhibit very close agreement. Remarkably, no other

significant conformational change is observed past the first 100 hours, suggesting that

the proteins do not experience denaturation even after weeks under non-physiological

conditions. It seems that proteins at ambient conditions (room temperature, 50÷60%

relative humidity) reach an equilibrium state, in which the new conformational pattern

is unperturbed. Such stability may be attributed to the capability of proteins to retain

their hydration shells, even under these experimental conditions. It is worth noting

that upon rehydration at the end of the test proteins return to exhibit a native-like

conformation, since their PL spectrum is typical of properly folded azurin. In

conclusion azurin proves to be a much more reliable material than it might have been

argued based upon some qualitative remarks. These results should go some way

towards dispelling the scepticism surrounding the use of proteins for biomolecular

devices

41

1.0

0.8

0.6

0.4

Inte

grat

ed e

mis

sion

7006005004003002001000Time (h)

46

44

42

40

38

36

34

Spe

ctra

l wid

th (n

m)

7006005004003002001000Time (h)

312

310

308

306

λ max

(nm

)

7006005004003002001000Time (h)a

b

c

FIG. 19. Ageing of azurin (zinc derivative) at ambient conditions, over more than 600 hours: emission maximum, full width at half maximum and integrated emission, as a function of time from the solvent removal. After a weak rearrangement over the

first few-hour, azurin reaches an essentially stable configuration.

42

3.2.2 Completion of M7 : Sub-10 nm gate for single molecule trapping and transport through the single molecule (INFM-Lecce)

3.2.2a Nanojunction fabrication

The prototype structure investigated to implement protein transistors (Fig. 20)

consists of a planar metal-insulator-metal nanojunction and a silver gate electrode

deposited on the back of the p-doped Si substrate to form an ohmic bond to the silicon

that acts as back-gate in a field-effect transistor configuration.

Two Cr/Au (6 nm/35 nm) arrow-shaped metallic electrodes facing each other at the

oxide side of a Si/SiO2 substrate (drain and source electrodes) and connected by the

self-assembled protein monolayer are fabricated by a combination of

photolithography and electron beam lithography with the fine features defined in a

poly(methyl methacrylate) (PMMA) resist layer. During the first two steps of the

process, photolithography followed by lift-off is used to define the Ag 50 nm-thick

gate electrode on the back of the Si/SiO2 substrate and the contact pads (Cr/Au,

Ti/Au, Ti/Pt, thicknesses 6/60 nm) on SiO2. Subsequently, the EBL process is carried

out with a Leica LION LV1 system having a thermal field emission electron source

with a 5 nm point-size e-beam. All the fabricated nanodevices were inspected by

plane-view scanning electron microscopy (SEM) in order to establish the success of

the whole process and to estimate the separation between the nanotips. Electrical

characterization of empty nanojunctions was performed by I-V measurements using a

Figure 20 High magnification SEM image of two Cr/Au nanotips with a separation of 100 nm.

43

semiconductor parameter analyzer. Typically, about 90% of nanojunctions were of

good quality with no-leakage and open-circuit resistance higher than 1 TΩ. Using this

standard EBL process, we obtain inter-electrodes separation not smaller than 30-40

nm. The final electrode distance reflects exactly the design. Note that the progressive

reduction of the electrode width approaching the nanojunction - in a tip-like geometry

- is useful in order to reach very small inter-electrode distances, since it facilitates the

lift-off process.

If necessary, the inter-electrode separation is reduced down to few nanometers (for

single molecule trapping) using two different techniques: a brushing method [3] or by

means of an additional process consisting of Au electrodeposition [27]. The first

method consists in a partial exposure of the resist before the standard EBL process, by

brushing the PMMA layer for a precise short time, in the range of few seconds, with a

defocused electron-beam. In Figure 21a, the reduction of tip separation as a function

of the brushing time is shown. By a careful calibration of the brushing time between

10 and 60 s, the inter-electrode distance is reduced almost linearly from 50 nm down

to 10 nm; for t >60 s, the tip separation goes to zero. The inset of Figure 21a outlines

our exposure technique. Figure 21b shows a high magnification SEM image of Cr/Au

nanotips with separation of only 20 nm obtained by EBL with additional e-beam

brushing (t ≅ 40 s). Electrode distance could also be reduced by increasing the dose,

but in this case we lack a fine control of the process.

In the second approach, an additional Au electrodeposition step is performed by using

a commercial Au-cyanide electroplating set-up. High magnification (HM) SEM

images of nanojunctions fabricated by Au electrodeposition are reported in Fig. 22.

The calibration curve of the Au electrodeposition process (Fig.22a) shows the

reduction of the inter-electrode gap as a function of process duration, from an initial

value of 100 nm, to a resolvable minimum gap of only 7±2 nm (Fig.22e). The lateral

growth rate varies from 1.66 nm/s at the beginning to 2.5 nm/s when the gap is

reduced below 20 nm. Due to possible small instabilities of electrodeposition process

parameters (temperature and pH of the solution, resistance of the contacts, etc.) with

consequent changes of the Au lateral growth rate, the percentage yield for a

programmed sub-10nm gap is in the range 20-30%. To remove all PMMA residues

and contaminations between the planar electrodes, an O2 plasma treatment is carried

out at the end of the process. Both these methods to reduce the inter-electrode

44

separation have advantages and drawbacks. The brushing technique provides a higher

throughput while the Au electrodeposition allows to obtain closer electrodes, although

in this case contaminations between the planar electrodes reduce the process yield.

The calibration curve of the Au electrodeposition process (Fig.22a) shows the

reduction of the inter-electrode gap as a function of process duration, from an initial

value of 100 nm, to a resolvable minimum gap of only 7±2 nm (Fig.22e). The lateral

growth rate varies from 1.66 nm/s at the beginning to 2.5 nm/s when the gap is

reduced below 20 nm. Due to possible small instabilities of electrodeposition process

parameters (temperature and pH of the solution, resistance of the contacts, etc.) with

consequent changes of the Au lateral growth rate, the percentage yield for a

programmed sub-10nm gap is in the range 20-30%. To remove all PMMA residues

and contaminations between the planar electrodes, an O2 plasma treatment is carried

out at the end of the process. Both these methods to reduce the inter-electrode

separation have advantages and drawbacks. The brushing technique provides a higher

Figure 21 (a) Reduction of the separation between electrodes as a function of the brushing time (∆t) of the PMMA resist by a defocused e-beam. When ∆t ranges from 0 to 50 s, the tip separation is reduced almost linearly from 50 nm (the nominal one) to 10 nm; for ∆t= 60 s, the tip separation becomes zero. The insets outline our exposure technique. (i) Gaussian shape of the e-beam; (ii) The exposure overlap between two adjacent points. (iii) Our exposure strategy. First, the resist is partially exposed by means of a defocused e-beam (gray background in the inset, instead of the previous black one). Then, during the following EBL process, the Gaussian shape of the e-beam allows the resist near the edges of the tip to reach the right dose, resulting in a final electrode distance (d) inferior to the nominal exposure length (L). The evaluated deviation of d is of the order of 5 nm, due to process parameters such as the point-size of the beam, the step size of the exposure and the SEM resolution. (b) Plan-view SEM image of Cr-Au nanotips with a separation of 20 nm obtained with a brushing time of 40 s.

b) a)

45

throughput while the Au electrodeposition allows to obtain closer electrodes, although

in this case contaminations between the planar electrodes reduce the process yield.

3.2.2b Single-molecule devices

Single-molecule devices were fabricated by reducing the inter-electrode separation

down to few nanometers (for single molecule trapping) using Au electrodeposition

[28], as previously described. Azurins were directly immobilized onto gold

electrodes via the disulphide bridge Cys3-Cys26 by incubating the samples onto the

gold nanogaps for 2h. Current-voltage experiments were carried out by using a

semiconductor parameter analyzer (HP Agilent 4155B) at room temperature and

ambient pressure. The high proximity of electrodes, which results in a very high

electric field when nanotips are biased, limits to 1 volt ( negative or positive drain-

source voltage) the maximum voltage applicable to the devices without destroying the

nanojunctions. This value is in agreement with the results of our electromagnetic

field simulations, as described in Section 3.2.1a. The current flowing through such

100n

0 5 10 15 20 25 30

0

20

40

60

80

100

Sepa

ratio

n(n

m)

Process Duration (sec)

Lateral Growthrate 1.66 nm/sec

Lateral Growthrate 2. 5 nm/sec

0 5 10 15 20 25 30

0

20

40

60

80

100

Sepa

ratio

n(n

m)

Process Duration (sec)

Lateral Growthrate 1.66 nm/sec

Lateral Growthrate 2. 5 nm/sec

0 5 10 15 20 25 30

0

20

40

60

80

100

Sepa

ratio

n(n

m)

Process Duration (sec)

Lateral Growthrate 1.66 nm/sec

Lateral Growthrate 2. 5 nm/sec

(b) (a)

Figure 22 (a) Calibration curve of the Au electroplating process, showing the reduction ofseparation between electrodes as function of process duration. (b) Cross-section SEMmicrograph of electrodes with separation around 20 nm (sample tilting ≅60°). (c-e) Plan-view SEM image of Cr-Au nanotips with separation of 15 nm, 10 nm and 7 nm ,respectively.

46

devices are low, in the pA range (figure 23), and just few pA higher than those

measured in the empty devices. Thus we can not conclude if we are observing current

through a single protein or statistical noise fluctuations. It is worth nothing that since

it is difficult to trap a single molecule between electrodes separated by few

nanometers due to capillarity effects, further experiments are currently in progress to

address the issue of single molecule conduction on a statistical basis. Moreover, the

very low signal level would require a more sensitive measurement set up, with

cryogenic cooling of the sample and at least one amplification stage. A sensible

improvement could also come from a different contact and electrode geometry and/or

from the change of material used for the metallic layer.

3.2.2c Alternative routes to single protein device implementation

Since several problems arose in the protein immobilization step when the electrodes

were closer than 30 nm, we looked for alternative strategies for the implementation of

single protein devices. The major challenge in molecular electronics is to interconnect

Figure 23 Results on single-molecule devices. The observed current-voltage characteristic (Ids-Vds) of such devices exhibits low currents in the pA range. Since it is difficult – due to capillarityeffect – to trap a single molecules between two so-close electrodes, further experiments arecurrently in progress to statistically address the issue of single molecule conduction. Moreover,we are changing the electrodes geometry and metals in order to extend the available voltagerange.

-1.0 -0.5 0.0 0.5 1.0 -2 -1 0 1 2 3 4 5

Cur

rent

(pA

)

Voltage (V)

47

molecules and probe molecular conductivity in real devices working at the nanoscale

(both two- and three-terminal devices). This issue requires the fabrication of

nanometer-spaced electrodes. Since optical lithography suffers from physical

limitations (mainly related to diffraction effects), new techniques are needed for

patterning below 100 nm. But none of the proposed methods (including electron-beam

lithography) equals the advantages of photolithography for low cost and high

throughput. Moreover most of them are only appropriate for contacting a single

device.

Using the method proposed by Krahne R et al. [29], we have exploited an

AlGaAs/GaAs quantum well grown by metal organic chemical vapour deposition to

define a network of nanojunctions by means of optical lithography and wet etching.

The thickness of the quantum well and the deposited metal layer control the gap size.

In such structures, proteins can be positioned between electrodes by electrostatic

trapping or by an immobilization procedure.

The sample fabrication is illustrated in Fig.24. An AlGaAs/GaAs quantum-well

structure is grown on a GaAs substrate by MOCVD, with the Al concentration

ranging from 35% to 90%. The thickness of the two undoped AlGaAs barriers above

and below the QW are 200 and 100 nm, respectively, while the thickness of the

embedded GaAs layer is varied between 15 to 20 nm in different samples. In the first

step, 250-nm-high mesa structures are defined by optical lithography and a standard

wet-etching process using H2O/ H2O2/H2PO4 200:4:4 ml for 120 sec. In such a way,

the GaAs layer is exposed on the side of the mesa (Fig. 24a). Next, the pattern of the

electrodes is defined by optical lithography. On the exposed areas a few tens of

nanometers of the GaAs layer are removed from the side by selective wet-etching

using citric acid and H2O2 20:4 as shown in Fig. 24b. The selectivity of the etching

between GaAs and AlGaAs is nominally 100:1, hence the width of the etched layer is

the QW width. The electrodes are fabricated by thermally evaporating a 5–20 nm thin

film of Ti/Pt from a direction perpendicular to the plane of the wafer surface such that

a gap is formed exactly where the GaAs was removed (Fig. 24c). The size of the gap

is determined by the crystal structure (which can be controlled with subnanometer

precision), surface roughness of the etched AlGaAs/GaAs interface due to selective

etching (less than one nanometer for short etching times), and metal evaporation

which can be controlled on the nanometer scale. We manage to fabricate gaps that are

less than 5 nm wide and estimate the minimum gap size that can be achieved by

48

our method to be about one to two nanometers. An important advantage of this

method is that since the mesa structure and the electrode pattern are defined by optical

lithography, it allows the simultaneous fabrication of many electrodes separated by

Figure 24 (a) Sketched description of the fabrication process: a mesa structure isdefined by optical lithography and wet etching, (b) selective wet etching removessome tens of nanometers of the GaAs layer, (c) after metal evaporation twoelectrodes are separated by a gap of a few nanometers, (d) the nanocluster ispositioned by electrostatic trapping, (e) a heavily doped GaAs layer could beeventually used as a gate, and (f) the process allows for the fabrication of a networkof devices with gaps of controlled nanometer size. [Krahne R, Appl. Phys. Lett. 81,730-732, 2002]

49

nanosize gaps on the wafer surface (see illustration in Fig. 24f). The fabricated

nanojunctions can be employed to implement protein devices at low temperature

(77K), since at room temperature bulk currents through the semiconductor exceed

currents through the proteins.

To reduce leakage currents, we have developed a selective oxidation techniques

which allows to convert the AlAs very rapidly into a stable native oxide by exposing

it to water steam at elevated temperatures. The selective oxidation is carried out by

means of a flow of nitrogen, which is allowed to bubble into deionised water. The N2

flow and the temperature of the water source were optimized at values of 1.5 l/min-1

and 80°C, respectively. The oxidation was carried out at different stages of the

fabrication process: (a) after metallization, (b) just before it with the electrodes

already defined in the resist layer (in such case, the temperature was kept low, to

100°C), (c) after the selective etching and the removal of the GaAs cap layer in order

to have a direct oxidation instead of a lateral one. The temperature ranged from 100°C

to 450°C, while the oxidation time ranged from 5 to 120 min (see Figure 25). Best

results were obtained on samples having a high Al concentration (90%), with the bulk

open-circuit current at room temperature reduced from tens of µA to few pA (Fig.26).

Using these junctions, protein devices were implemented by directly immobilization

of azurin onto electrodes via the disulphide bridge Cys3-Cys26. Preliminary results

show current-voltage characteristics exhibiting peculiar peaks that are not observed in

the empty devices and could be ascribed to conduction through proteins, as reported

in Fig.27. In these devices single proteins bridge the gap between the electrodes. A

phenomenological model for the interpretation of these experimental results is under

development. Further experiments are in progress.

3.2.3 M8: (month 30) Sub-10 nm three terminal device for single molecule field effect transistor (INFM-Lecce and INFM-Modena).

3.2.3a Solid state sub-10 nm three terminal device for single molecule field effect

transistor (INFM-Lecce) . On the basis of the results achieved on two terminal single protein devices, it was not

possible to test the effect of gate modulation on the current flowing through the device

when working in the standard three terminal geometry with the back gate contact, like

the one used for protein monolayers (Section 3.2.1a)). Therefore we tried two

alternative approaches for the implementation of these sub-10 nm three-terminal

50

devices: the first one was based on the semiconductor MESA structure sketched in

Fig.24 e, where we aimed to obtain a gate contact by means of a thin doped GaAs

layer ; the second one was based on a different three electrode geometry in which

single proteins should have a higher probability to be trapped. SEM pictures of these

planar contact geometry before and after the electrodeposition process are reported in

Figure 25. Optical micrograph of two different mesa devices fabricated on an Al90Ga10As/GaAs quantum well grown by metal organic chemical vapour deposition. The selective oxidation was carried out for (a) 30 min at 100 °C and (b) 2 hours at 450 °C. The more oxidized sample exhibits a red colour.

51

Fig.28a and b, respectively.

-2 -1 0 1 210f

10p

10n

10µ

10m

11 pA

2.3

7 µA

Room Temperature - Dark

Cur

rent

(A)

Voltage (V)

Al 35 G 65As - No Oxidation Al 90 G 10As - 2h @ 450°C (G1M8) Al 90 G 10As - 1h40' @ 450°C ( γ2_(2,3)_o2 )

Figure 26 Current-voltage characteristics of differently oxidized mesa devices. Bymeans of the selective oxidation of the AlGaAs layer the bulk open-circuit currentcan be reduced from tens of µA to few pA.

0 2 4 6

200

400

600

800

1000

Cur

rent

(pA

)

Voltage (V)

, , , , Az-Cu Empty, Empty Forward, Empty Reverse

Figure 27 Preliminary results on azurin-funcionalized mesa devices. The colored curves are related to I-V measurements performed on different electrode couples on the same chip.

52

3.2.3b Single metalloprotein transistor based on a bio-inorganic electronic junction

whose transparency is tuneable by electrochemical control (INFM-Modena).

In order to implement an electrochemically addressable, nanometer-sized, bio-

inorganic junction, azurin was covalently bound to a tapered gold stylus exploiting the

Cys3Cys26 disulfide bridge. Fig. 29 summarizes the scheme of the experimental set-

up.

100nm

Figure 28 SEM picture of planar contacts geometry for the implementation of solid state Sub-10 nm three terminal device before (a) and after(b) the electrodeposition process

53

The stylus, which is insulated (see Fig. 30a) to minimize the capacitive contribution to

the measured current, is brought within tunnelling distance of a gold substrate in an

electrochemical cell filled with a saline buffer (50 mM NH4Ac, pH 4.6). The system

is equipped with a bipotentiostat that allows to vary the working electrode (stylus and

substrate) potentials while keeping their bias constant. In this configuration, the

tunnelling current can be measured versus the tip potential for different bias voltages,

by taking advantage of the direct molecular tracking, which is realized by fixing the

molecule on the tip.

As the first step, we characterized azurin immobilized on a tapered gold electrode.

Au substrate

~4 nm

d

Insulated gold tip

Reference electrode

Counter electrode

Azurin

Figure 29 Schematic representation of the electrochemically gated, singleprotein junction. Tip and substrate, which are electrically biased, are fed by abipotentiostat that controls their potential in an electrochemical cell. Azurin ischemisorbed on the insulated tip by a naturally occurring disulfide bridge(yellow); the red spot in the lower part of the globule represents the location ofthe Cu ion. The tip is brought within tunnelling distance (d) from the substrate;feedback is switched off, and tunnelling current is recorded at fixed bias whilesweeping the tip potential.

54

Fig.30b shows transmission electron microscopy (TEM) images at different

magnification, with azurin molecules chemisorbed on the apex of a gold stylus. Their

apparent shape and size is compatible with what is observed in ECSTM imaging,

(Fig.30c) where azurin appears as a bump ~4 nm in size that sustains several repeated

scans in constant current mode.

Secondly, we assessed retention of protein redox activity through cyclic voltammetry

measurements on azurin-coated, insulated tips. Fig.30d reveals the presence of azurin

oxidation and reduction waves along with a redox midpoint of +106 ± 10 mV vs

saturated calomel electrode (SCE), in agreement with reported data for both diffusion

[30] and surface-bound [31] regimes. This confirmed the presence of redox active

azurin molecules on the tip apex. Their number, estimated by the integral of the

oxidation wave assuming a single electron exchange per molecule, was ~ 106 mol.

Considering a typical solvent-exposed tip area of some tens of square microns, as

evinced from the SEM back-scattered electrons image in Fig.2a, this estimate is

consistent with about 50% coverage, as calculated for azurin on gold by reductive

desorption measurements [32]. It is worth noting that the possibility of observing

Figure 30. Tip and azurin characterization. a) Back-scattered electrons SEM image of an insulated gold tip showing the apical part not covered by the wax. b) TEM images of azurin on a tapered gold stylus, bar = 20 nm; inset bar = 50 nm. c) ECSTM image of azurin on Au(111); bar = 20 nm, It = 2 nA, Vbias = +400 mV. d) Cyclic voltammetry at the azurin-coated tip showing azurin redox activity (sweep rate 2.4 V/s).

-0.2 -0.1 0.0 0.1 0.2 0.3 0.4 -120

-80

-40

0

40

80

120

IEC

(p

Vti

vs SCE (V)

a) b)

c) d)

55

azurin redox activity on these samples depends, in our experimental conditions, on the

size of the tip area, which remains uncovered by the insulating wax . Tips showing

leakages < 1pA had an undetectable voltammetric response due to the exceedingly

small faradaic signal arising from them.

In a fashion similar to that of a conventional STM experiment, a metalloprotein-

coated tip was then brought close to a gold substrate in an electrochemical cell.

Typical current intensities not exceeding few tens of picoAmps were adopted as a

working set-point, in order to achieve large enough tunnelling gaps and thereby

preserve azurin integrity. The tunnelling current was measured as a function of the tip

potential after the system was allowed to stabilize typically for 30 min, and the

feedback was switched off. In this configuration, it is conceivable that the current

flows only through a single molecule, since there will be one molecule which will

protrude into the gap more than any other [33]. Since the tunnelling current decreases

exponentially with the electrode separation, the contribution to the overall current by

molecules positioned away from the tip apex would be negligible. This is also

plausible in electrochemical STM experiments in spite of the decreased inverse decay

of the tunnelling length (k ~ 1 Å-1) with respect to the UHV situation [34], given the

relatively large size of these proteins (3 ÷ 4 nm).

Repeated tip potential sweeps gave rise to the characteristic curves shown in figure

31a (red curves). A marked current resonance is clearly visible which is not visible in

the case of control measurements (non-electro-active Zn-azurin [35] coated tip (blue

curve), and bare tip (black curve)). The resonance is brought about thorough the

temporary population of azurin redox states in virtue of their alignment to the Fermi

level of the gold electrodes (induced by the set potential values). Zn-azurin does not

show any current resonance, consistent with the absence of redox states [35] in the

potential range of interest. This last point is of particular importance since it outlines

the role of redox states in eliciting the measured features of the tunnelling current. A

further confirmation of this phenomenon was provided by ECSTM constant current

images of a mixture of Cu and Zn azurin acquired at a substrate potential

corresponding to the maximum of the red curves in Fig.31a. Indeed, Fig.31b shows

bumps of different apparent height which are attributed to molecules bearing a

different metal ion (Cu rather than Zn) in their active site, thus contributing differently

to the molecular redox states assisting the tunnelling current [36]. It is worth noting

56

- -0.1 0.0 0.1 0.2 0.3 0.40

20

40

60

80

100

120

Itip

(p

Vtip vs SCE (V)

-0.2 0.0 0.2 0.40

4080

121620

Vti vs SCE (V)

Itip

(pa)

b)

c)

Figure 31 a) Hybrid junction characterization. Six repeated tip potential sweeps (initialconditions: Vbias = +100 mV, tunnelling current = 15 pA, tip potential = -225 mV)showing marked tunnelling current resonance compared to a bare insulated gold stylus(black curve) and to a Zn azurin-coated tip (blue curve). Inset: eight repeated sweeps(initial conditions: Vbias = +100 mV, tunnelling current = 20 pA, tip potential = -225 mV)on another sample also showing an oxidation wave (outlined by arrow) arising from thewhole set of chemisorbed molecules. b) ECSTM image of a mixture of Cu and Zn azurinon Au(111) imaged at Vs = –0.05 V, Vbias = +100 mV, It = 1 nA; image size 89 × 89 nm2.C) Two-step electron transport via the redox center. The first electron transfer (I) is from the substrate to a protein vacant redox level; while this level relaxes vibrationally andbefore it drops below the tip Fermi level, a second electron transfer event (II) takes placethat moves an electron from the molecule to the tip. Fermi levels of the substrate and thetip are separated by eVbias = +100 meV, whereas redox levels are 2λ = 400 meV apart (λvalue derived by best fitting the two-step model to the experimental data, see text).Numerical values correspond to the measurements shown in figure 31a.

57

that this image was acquired using a current set-point of 1 nA at +100 mV bias

voltage. These conditions, which are normal for imaging metalloproteins by ECSTM

[37], indicate that a large current can flow through these molecules once placed on a

metal surface in between two electrodes, a few nanometers apart. The registered

current intensities are much larger than those predictable on the bases of the electron

transfer rate estimated in solution for azurin [38, 39] and suggest that additional

pathways are elicited in the present experimental configuration for electrons to travel

from the substrate to the active site. Electrons can flow via the physical links

constituted by Cys3 and Cys26, which are involved in bonding to the Au, but can also

use different pathways arising from the close proximity of other protein residues to

the electrode surface. This last possibility may circumvent the limiting rate factors

indicated, for instance, by electron transfer measurements from the azurin disulfide to

the Cu ion [38], and may account for the large currents registered.

Fig.31c describes how electron transfer could take place in the case of Cu-azurin,

most likely in a two step process [40, 41] that involves a first transfer from the

substrate to the molecule and a further one to the tip. The densities of the oxidized and

reduced states are centred at ε0+λ and ε0-λ respectively, ε0 being the redox midpoint,

and λ the reorganization energy of azurin in the specific configuration. When the

density of the oxidized state is aligned to the Fermi level of the substrate, one electron

can tunnel from the substrate to the protein (first step); afterwards, the protein starts to

relax (dotted curve) towards its reduced configuration, and before full relaxation has

occurred, electron can tunnel from the protein to the tip (second step).

Unlike ECSTM experiments, where only differences in the apparent height of the

adsorbate in response to substrate potential variation are measured [36, 42], in the

reported configuration direct access to the current signal was achieved, enabling a

deeper insight into the mechanisms ruling the electron transfer reaction. Notably,

here the electrochemical control of the electron transfer reaction plays an analogous

role to that of the gate voltage in planar electronic devices such as single-molecule

and/or single-electron transistors [43-45]. In exceptional cases (see arrow in inset to

Fig.31a), it was possible to observe, on a single tip, both the aforementioned

resonance and the azurin oxidation wave, the latter arising from all of the

chemisorbed molecules, as in the case of cyclic voltammetry. This occurrence, which

reveals the different nature of the observed resonance with respect to the faradaic

58

contribution, can be interpreted as originating from an imperfect tip insulation that

leaves enough molecules exposed to the solution to elicit a detectable oxidation

signal.

Since the large capacitive contribution, which tends to mask the tunnelling signal,

made it impossible to evaluate I-V characteristics in aqueous media by simply

sweeping the voltage and recording the current [45], the dependence of the current on

the bias voltage across the hybrid junction was evaluated by repeating the

measurements reported in Fig.31a for several fixed bias voltages, as shown in Fig.32.

The resonance intensity and width increased and the position of the maximum shifted

towards positive tip potential values with bias voltage. This behaviour appears to be

consistent with that predicted by the Kuznetsov-Ulstrup theory [38, 46] for long-range

electron transfer in ECSTM experiments. From this set of data, the“in” and “off

resonance” tunnelling resistance could be assessed by locating the position of the

current maximum in each curve, as shown in the inset. Their values (660 ± 30 MW,

and 3.28 ± 0.25 GW, respectively) indicate an increased transparency of the

tunnelling barrier in resonant conditions, which is consistent with the fact that the

molecular redox states play a role in electron transport through azurin. This result can

Figure 32 Bias voltage dependence of the trans-characteristics. Curves(displayed with an offset of 50 pA each to avoid overlap) were obtained atdifferent bias voltage stepped by 0.05 V in the range 0.05 ÷ 0.35 V, startingfrom the lowest curve, using a sweep rate of 2.4 V/s. Inset: Itip-Vbiascharacteristics measured from the set of curves of Fig.31 at Vtip = –0.2 V(“off res”) and at each current maximum position (“in res”), showing adifferent tunnelling barrier transparency.

-0.2 -0.1 0.0 0.1 0.2 0.3 0.40.0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8

I tip (n

A)

Vti vs SCE (V)

Vbias 0.0 0.1 0.2 0.3 0.4

0.00.10.20.30.40.5

Ioff

Iin

I tip (n

A)

(V)

59

be exploited to modulate the current flow through the protein-hosting junction in a

fashion pretty much similar to that of a single particle transistor.

Furthermore, it has been discovered that the tip potential value at which the current

maximum occurs depends linearly on the applied bias. A best fit of the

aforementioned theory for two-step vibrationally coherent electron transfer in

ECSTM of redox adsorbates [40] to the observed data provided an estimate for the

numerical coefficient that determines the potential value at the redox center within the

tunnelling gap [40]. A figure of 0.40 ± 0.03 was obtained, confirming experimentally

for the first time the assumption, usually made in ECSTM data analysis [41], that the

redox couple potential lies approximately halfway through the bias voltage drop. The

model also provides an estimate for the reorganization energy λ at about 200 meV.

This figure is substantially lower than that measured for azurin in solution [47],

consistent with other ECSTM observations [42] and theoretical estimates [40].

Besides, it is worth noting that it compares well with theoretical estimates for the

reorganization energy of azurin inner shell [48]. Indeed, a reduced λ value in our

experiments may be ascribed to the particular configuration whereby two closely

spaced electrodes, sandwiching azurin, prevent most water molecules from

surrounding the protein, cancelling, thus, the solvent contribution to λ.

It is worth noting that both the present resonant-like behaviour and that previously

observed, characterized by the occurrence of a current plateau upon variation of the

substrate potential, are interpretable within the two-step electron transfer theory.

Observing either of the two regimes depends just upon the relative value of eVbias and

2λ.

The reported results demonstrate the possibility of exploiting metalloprotein redox

properties to vary the electronic transparency of a wet bio-inorganic junction made

even with a single molecule hosted in a nanometer gap. The demonstration of a

protein-based tuneable electronic gate is a critical step in the implementation of single

particle transistors made with metalloproteins and could represent a paradigm for

devising future nanobioelectronic devices.

References:

[1] G. Maruccio et al., submitted to Adv. Mat.

60

[2] We call our protein device a field effect transistor since we exploit a field effect to modulate the resistance of the protein monolayer and change the I-V characteristics of the device.

[3] J. Lee et al., Nano Letters 2003, 3, 113; C. R. Kagan et al., Nano Letters 2003, 3, 119 [4] R.M.Metzger, B.Chen, U.Hopfner, M.V.Lakshmikantham, D.Vuillaume, T.Kawai,

X.Wu, H.Tachibana, T.V.Hughes, H.Sakurai, J.W.Baldwin, C.Hosch, M.P.Cava, L.Brehmer, G.J.Ashwell, J. Am. Chem. Soc. 1997, 119, 10455

[5] S. Heinze, J. Tersoff, R. Martel, V. Derycke, J. Appenzeller, and Ph. Avouris Phys. Rev. Lett. 2002, 89, 106801

[6] O. Kuhn, V. Rupasov and S. Mukamel, J. Chem. Phys. 1996, 104, 5821 [7] IUPAC Compendium of Chemical Terminology 2nd Edition (1997) [8] D.N. Blauch and J.M. Saveant, J.Am.Chem.Soc. 1992, 114, 3323 [9] J.Park, A.N.Pasupathy,J.I.Goldsmith, C.Chang, Y.Yaish, J.R. Petta, M.Rinkoski,

J.P.Sethna, H.D.Abruna, P.L.McEuen and C.Ralph, Nature 2002, 417, 722 [10] P.Facci, D.Alliata, S. Cannistraro, Ultramicroscopy 2001, 89, 291 [11] A.Alessandrini, M.Gerunda, G.W. Canters, M.P. Verbeet, P. Facci, Chem. Phys. Lett.

2003, 376, 625 [12] P.Visconti, G.Maruccio, E.D'Amone, A.Della Torre, A.Bramanti, R.Cingolani,

R.Rinaldi, Mat. Sci. Eng. C-Bio S 23, 889-892 (2003) [13] Kroes, S.J. et al., Biophys J. 1998, 75, 2441-50. [14] Nar, H. et al., J. Mol. Biol. 1991, 221, 765-72. [15] Baker, E.N., J. Mol. Biol. 1988, 203, 1071-95. [16] Nar, H. et al., FEBS Lett. 1992, 306, 119-24. [17] Nar, H. et al., Eur. J . Biochem. 1992, 205, 1123-9. [18] Hansen, J.E. et al., Biochemistry 1990, 29, 7329-38. [19] Sweeney, J.A. et al., J. Am. Chem. Soc. 1991, 113, 7531-7. [20] Leckner, J. et al., Biochim. Biophys. Acta 1997, 1342, 19-27. [21] Larsson, S. et al., J. Phys. Chem. 1995, 99, 4860-5. [22] Merzel, F. et al., Proc. Natl. Acad. Sci. USA 2002, 99, 5378-83. [23] Danielewicz-Ferchmin, I. et al., Biophys. Chem. 2003, 106, 147-153. [24] Svergun, D.I. et al., Proc. Natl. Acad. Sci. USA 1998, 95, 2267-72. [25] Xu, D. et al., J. Phys. Chem. 1996, 100, 12108-21. [26] Pompa P.P. et al., Phys. Rev. E 69, 032901 (2004). [27] P.Visconti, G.Maruccio, E.D'Amone, A.Della Torre, A.Bramanti, R.Cingolani,

R.Rinaldi, Mat. Sci. Eng. C-Bio S 23, 889-892 (2003) [28] P.Visconti, G.Maruccio, E.D'Amone, A.Della Torre, A.Bramanti, R.Cingolani,

R.Rinaldi, Mat. Sci. Eng. C-Bio S 23, 889-892 (2003) [29] Krahne R. et al., Appl. Phys. Lett. 81, 730-732, 2002), [30] Sykes, A. G. Active-site properties of the blue copper proteins. Adv. Inorg. Chem. 36,

377-408 (1991). [31] Facci, P. Alliata, D. & Cannistraro, S. Potential-induced resonant tunneling through a

redox metalloprotein investigated by electrochemical scanning probe microscopy. Ultramicroscopy 89, 291-298 (2001).

[32] Chi, Q. et al. Molecular monolayers and interfacial electron transfer of Pseudomonas aeruginosa azurin on Au(111). J. Am. Chem. Soc. 122, 4047-4055 (2000).

[33] Klein, D. L. Roth, R. Lim, A. K. L. Alivisatos, A. P. & McEuen, P. L. A single-electron transistor made from a cadmium selenide nanocrystal. Nature, 389, 699-701 (1997).

[34] Vaught, A. Jing, T. W. & Lindsay, S. M. Non-exponential tunnelling in water near an electrode. Chem. Phys. Lett. 236, 206-310 (1995).

[35] van Amsterdam, I.M.C. et al. Effects of Dimerization on Protein Electron Transfer. Chemistry 7, 2398-2406 (2001).

[36] Alessandrini, A. Gerunda, M. Canters, G. W. Verbeet, M. Ph. & Facci, P. Electron tunnelling through azurin is mediated by the active site Cu ion. Chem. Phys. Lett. 376, 625-630 (2003).

61

[37] Friis, E. P. et al. An approach to long-range electron transfer mechanisms in metalloproteins: In situ scanning tunneling microscopy with submolecular resolution. Proc. Natl. Acad. Sci. USA 96, 1379-1384 (1999).

[38] Farver, O. & Pecht, I. Long range intramolecular electron transfer in azurins. J. Am. Chem. Soc. 114, 5764-5767 (1992).

[39] Gray, H. B. & Winkler J. R. Electron tunnelling through proteins. Q. Rev. Biophys. 36, 341-372 (2003).

[40] Friis, E. P. Kharkats, Y. I. Kuznetsov, A. M. & Ulstrup, J. In situ scanning tunnelling microscopy of a redox molecule as a vibrationally coherent electronic three-level process. J. Phys. Chem. A 102, 7851-7859 (1998).

[41] Han, W. et al. STM contrast, electron transfer chemistry, and conduction in molecules. J. Phys. Chem. B 101, 10719-10725 (1997).

[42] Tao, N. J. Probing potential-tuned resonant tunnelling through redox molecules with scanning tunnelling microscopy. Phys. Rev. Lett. 76, 4066-4069 (1996).

[43] Tans, S. J. Verschueren, A. R. M. & Dekker, C. Room-temperature transistor based on a single carbon nanotube. Nature 393, 49-52 (1998).

[44] Park, J. et al. Coulomb blockade and the Kondo effect in single-atom transistor. Nature 417, 722-725 (2002).

[45] Kubatkin, S. et al. Single-electron transistor of a single organic molecule with access to several redox states. Nature 425, 698-701 (2003).

[46] Zhang, J. et al. Electron transfer behaviour of biological macromolecules towards the single-molecule level. J. Phys.: Condens. Matter 15, S1873-S1890 (2003)

[47] Di Bilio, A.J. et al. Reorganization Energy of Blue Copper: Effects of Temperature and Driving Force on the Rates of Electron Transfer in Ruthenium- and Osmium-Modified Azurins. J. Am. Chem. Soc.; 119, 9921-9922 (1997).

[48] Ryde, U. & Olsson, M.H.M. Structure, Strain, and reorganization energy of blue copper models in the protein. Int. J. Quantum Chem. 81, 335-347 (2001).

62

3.3 WP3

The final goal of this workpackage was the implementation of metalloprotein based

biosensors. To this aim, specific deliverables (D3.1 and D3.2) and milestones (M10

and M11) were accomplished, concerning the testing of conductivity of molecular

layers under different environments, the study of the relationship between electronic

conduction and metalloprotein hydration, and the realization and characterization of a

nitrite biosensor. Different experimental methods were exploited for these studies,

ranging from protein design and synthesis (with the Leiden Group), to single molecule

characterization by scanning probe techniques, and cyclic voltammetry measurements

on the different protein layers. Spectroscopic studies and Molecular Dynamics

simulation on the metalloproteins were performed within this workpackage. Finally,

several approaches to obtain quantifiable nitrite turnover from electroanalyses of the

enzyme nitrite reductase are reported.

The experiments were performed by the groups of Oxford and INFM Viterbo, in

strong collaboration through student exchanges. The proteins were delivered by

Leiden.

3.3.1 Deliverable D3.1 (Milestones M10-M11): characterization of conductivity in

metalloproteins monolayers

Self-assembling and conductivity of Plastocyanin mutants (INFM-Viterbo)

INFM-Viterbo focused its attention on the design, spectroscopic analysis and single

molecule characterisation of copper protein plastocyanin mutants. The aim was to get

some insight into the electron transport mechanism at molecular level, which is a

fundamental task in building hydrid nanodevices/nanobiosensors, and to implement

the use of different metalloproteins (Fig.1) for interconnecting nanostructures.

The Viterbo group has closely collaborated with the Leiden group in the design and

engineering of two plastocyanin mutants able to bind selectively to a gold surface, by

exploiting either SS or SH groups. In the first mutant (PCSS), a disulfide bridge was

inserted within the protein (Fig. 1a) by replacing residues Ile-21 and Glu-25 with Cys

using structurally conservative mutagenesis. X-ray crystallography and several

spectroscopies (Resonance Raman, Electron Paramagnetic Resonance, UV-VIS)

indicated that the overall three-dimensional structure of PCSS and the redox copper

site coordination has been essentially preserved together with its electron transfer

properties [1, 5].

63

In the second mutant (PCSH) a residue tail (threonine, cysteine and glycine) was

inserted at the C-terminal end (Fig. 1b). The single surface-exposed thiol group

provided a further possibility for plastocyanin immobilisation. The spectroscopic

characterisation of this mutant confirmed that, even in this case, the copper site

coordination has been preserved [7].

The self-assembling and the conductive properties of plastocyanin immobilised on a

gold surface via either SH or SS groups were studied by using Scanning Probe

Microscopy (STM/STS, Tapping Mode AFM, Conductive AFM), cyclic voltammetry

(CV) and Molecular Dynamic Simulations (MDS).

Morphology of single plastocyanin mutants adsorbed on Au(111) (INFM-Viterbo)

The morphology of adsorbed PCSS and PCSH, in particular the height and orientation

of the macromolecules with respect to the gold substrate was examined by a

systematic analysis by TMAFM and STM [5, 7].

Fig. 2 shows representative TMAFM images recorded in buffer solution of PCSS (a)

and PCSH (b) molecules adsorbed onto Au (111). Both mutants are homogeneously

distributed over the substrate, whereas the high quality of the recorded image, even

after repetitive scans, is indicative of a stable binding to gold. Individual molecules

are well distinguishable above the substrate, and the vertical dimension of the PC

64

mutants was estimated from statistical analysis of individual cross-sections, as shown

in Fig. 2. Data taken for hundreds of molecules are represented in the histograms of

Fig. 3.

The monomodal distribution is indicative of a preferential orientation of the adsorbed

proteins on the gold substrate. The data for PCSS molecules have a mean value equal

to 2.3 nm and a standard deviation of 0.5 nm, whereas for PCSH the mean height is

1.9 nm and the standard deviation 0.6 nm. For PC mutants with the anchoring groups

to gold assumed to be ‘face down’, a vertical size

of about 2.8 nm is expected. Therefore, data for PCSS are centred at a value close to

that expected, whereas the vertical dimension of PCSH above the gold substrate

slightly differs from this value. The standard deviation, which is in both cases much

larger than the experimental error, is indicative of a significant spread in the vertical

size of the molecule. This result suggests that the disulfide bridge provides a stricter

immobilisation and a more homogeneous orientation of PCSS molecules onto gold

compared to PCSH, where the single thiol, external to the main structure, could allow

a higher flexibility of the adsorbed proteins.

0 1 2 3 40

5

10

15

(a)

h0 = 2.3 nm

σ2 = 0.3 nm

num

ber o

f mol

ecul

es (%

)

height (nm)

0 1 2 3 40

5

10

15

(b)

h0 = 1.9 nm

σ2 = 0.4 nm

height (nm)

num

ber o

f mol

ecul

es (%

)

Fig. 3: Statistical analysis of PCSS (a) and PCSH(b) height above the Au(111) substrate.

Fig. 2: TMAFM image and representativecross section profile of PCSS (a) and PCSH (b)molecules adsorbed on Au(111).

(b)

(a)

65

The adsorbed PCSS and PCSH molecules have been also imaged by STM in buffer

solution, air and under nitrogen atmosphere. No substantial differences in shape

between the two mutants could be observed in the corresponding STM images. For

both PC mutants, the STM images appear to be stable and reproducible even after

repetitive scans, thus confirming the robust binding of the protein molecules to

Au(111) substrate. Some representative STM images are shown in Fig. 4 for PCSS on

Au(111) under various experimental conditions

The lateral dimension of these single molecules well agrees with the crystallographic

values of PCSS showing a diameter of about 4.0 nm. The vertical size of PCSS and

PCSH is ranging between 0.5-0.7 nm, as measured by the tip retraction along the z

axis and shown in the cross section profile of Fig. 4. The vertical size appears to be

underestimated and significantly smaller than that expected, as generally observed for

the biomolecules imaged by STM. This recurrent characteristic of the STM images

from biological material has drawn our attention, especially regarding the possibility

that the tip may interfere with the soft biological sample during the imaging scans.

We have, therefore, estimated the tunnelling gap between the scanning tip and the

gold substrate as deeply discussed in Ref. 13. Briefly, the tip-sample separation can be

inferred from the tunnelling current and bias voltage settings, once the corresponding

tunnelling resistance has been calibrated against the gap width. In Fig. 5 the

dependence of the resistance on the tip-substrate distance is shown, in air and in

Fig. 4: STM images of PCSS molecules on Au(111) recorded in aqueous medium (a), in air (b), and in nitrogen atmosphere (c).

a) b) c)

66

water, on an Au(111) substrate by using a Pt-Ir tip. Here, the initial tip position (Z = 0

nm) refers to a tunnelling resistance of 4 x 109 Ω, while maximum tip extension

corresponds to a contact resistance of 2 x 104 Ω. From data in Fig. 5 we inferred

tunnelling distances of about 3 nm and 6 nm in water and in air, respectively, at a

working resistance of 4 x 109 Ω.

During the last year, we have also focused on the relevance of tunnelling distances

when imaging soft biological material, specifically PC mutants chemisorbed on

Au(111) [13]. A clear evidence of the possible tip-molecule interaction, once the

tunnelling distance is reduced, is shown in the sequence of STM images as recorded

in water for PCSS (Fig. 6). As inferred by the resistance-distance plot of Fig. 5, the

tunnelling gap width was initially reduced from 2.9 nm (Fig. 6 (a)) to 2.4 nm (Fig. 6

(b)). If a tip retraction of 0.5 nm is added to such distances, in both cases the tip has

enough vertical space to overcome the proteins, without interfering with their

molecular structure. In full accordance, no changes in shape or lateral dimensions

were observed for the imaged proteins. On the contrary, when the tunnelling gap was

reduced to 1.9 nm (Fig. 6 (c)) by setting the resistance to 4 x 108 Ω, the available

vertical space results to be 2.5 nm, which is below the physical height of the proteins.

As a major consequence, some PCSS molecules display a considerable enlarged

lateral dimension. For example, the molecule indicated by the white arrow increases

its lateral size from 5.7 nm (Figs. 6 (a), (b)) to 14.9 nm (Fig.6 (c)). Finally, when the

resistance is set to 2 x 107 Ω (Fig.6 (d)), which corresponds to a tip-substrate distance

of 1.4 nm, the lateral size of the indicated PCSS molecule is enlarged to a value of

-6 -5 -4 -3 -2 -1 0 10 4 10 5 10 6 10 7 10 8 10 9

10 10 R

esis

tanc

e ( Ω

) air water

Fig. 5: Tunnelling resistance measured on a flat Au(111) substrate and plotted, insemi logarithmic scale, against the vertical tip position (distance spanned by the tip).

67

16.9 nm. At the last two tunnelling resistances, the tip can either squeeze or pass

through the protein, leading to invasive measurements of the biological sample.

The strong interaction locally applied could affect either the structure or the function

of the bio-molecule, preventing any investigation of the intrinsic electron transport

properties. Such results indicate that STM is a powerful technique to image and study

single biomolecules if care is taken in setting the resistance values corresponding to

non invasive tip-substrate distances.

Functionality and redox activity of adsorbed molecules (INFM-Viterbo)

The study on the functionality of PC mutants monolayer covalently immobilised on

gold electrode was addressed by CV, and performed in collaboration with the Oxford

group [11].

(a) (b)

(c) (d)

Fig. 6: PCSS proteins adsorbed on an Au(111) substrate as imaged by constant current STM inultra pure water at decreasing tunnelling resistances: 4 x 109 Ω (a), 4 x 108 Ω (b), 4 x 107 Ω (c), 2 x 107 Ω (d).

68

In these measurements robust voltammetric responses, stable to continual scanning

(>100 cycles at 10 mVs-1), were obtained for PCSS adlayers on polycrystalline gold

as shown in Fig. 7(a). The redox midpoint potential of the PCSS adlayer was 162 ± 10

mV vs SCE, a value close to that obtained diffusively with the wild-type protein at

edge oriented pyrolytic graphite electrode. This confirms that immobilization at the

gold surface is occurring without significant perturbation of the native structure.

Typical voltammograms recorded on PCSH immobilised on gold are represented in

Fig. 8, for increasing scan rates. Independently of the sweeping rate, two well distinct

Fig. 7: Background corrected voltammogram recorded for PCSS monolayers in 100 mM potassiumphosphate pH 7.14.at 100 mV/s scan rate.

Fig. 8: Voltammograms recorded for PCSH adlayers on bare polycrystalline gold in 20 mM sodiumphosphate at pH 6.

69

peaks due to the oxidation and reduction process are evident. The high symmetry of

voltammograms and the peak separation, very close to the theoretical value, indicates

that the redox process is fully reversible and almost ideal.

The surface coverage of electroactive molecules, estimated by integrating the faradaic

response, and correcting for an AFM-determined surface roughness factor of 1.3, was

of 2-8 × 1014 molecules/cm2. This value is in good agreement with an expected

coverage for a molecule with a lateral dimension of ~3.5 nm, and indicates that

immobilisation occurs with a high degree of functional retention.

To estimate the redox midpoint and the surface coverage, the background current

generated by the gold electrode in buffer was subtracted to the cyclic voltammogram

recorded on the PCSH adlayers on gold. The redox midpoint for PCSH molecules

immobilized on bare gold was found to be +168 mVSCE ± 10 mV, a value close to that

obtained diffusively with the wild-type protein and the PCSS mutant. The integration

of the faradaic current provided an estimate of the electrode coverage value in

excellent agreement with that of a densely packed PCSH layer.

Molecular conduction of Plastocyanin mutants (INFM-Viterbo)

The redox functionality of single PC mutants, as well as the role of the redox centre in

the tunnelling mechanism, has been investigated by STM under electrochemical

control [7].

Briefly, in situ STM images for PCSS and PCSH adsorbed on Au(111) for several

substrate potentials have been obtained. The molecular features are clearly visible for

substrate potentials close to the midpoint potential, the image contrast is weaker when

the potential is far from this value and it is recovered once the initial potential is re-

established. Such findings seem to be consistent with results reported elsewhere for

AZ [3] and to support that copper site represents a preferential way for the tunnelling

process through a redox molecule, once its redox levels are properly aligned with the

substrate and tip Fermi levels. On the other hand when similar STM experiments have

been performed for the PCSH mutant no variation in the image contrast was detected.

Since the redox functionality of adsorbed mutant was demonstrated in the

corresponding CV experiments, it was hypothesized that the higher protein flexibility

resulting from immobilisation via external SH group may lead to an unfavourable

alignment of molecular redox levels with tip and substrate Fermi levels.

70

The stable binding of the PC mutants allowed studying the conductive behaviour of

individual molecules by Scanning Tunnelling Spectroscopy, which was initially

carried out in ambient condition [7, 12]. For both PCSS and PCSH, current-voltage

measurements were recorded by positioning the tip over individual molecules after

having disengaged the feedback loop.

I-V data in the ± 1 V range were compared with those obtained on Au(111) substrates

(Fig. 9). The slight asymmetry observed for gold finds an explanation in the atomic

structure of the tip apex, which has been theoretically and experimentally

demonstrated to influence the I-V relation. PCSS molecules exhibit a reproducible

and significant asymmetry when compared to gold. In contrast, only a slight

asymmetric I-V relation for PCSH is observed, being however almost

indistinguishable from that of gold within the experimental error [12].

-1,0 -0,5 0,0 0,5 1,0-1,5

-1,0

-0,5

0,0

0,5

1,0

1,5

2,0

2,5

PCSS

Tunn

elin

g cu

rrent

(nA

)

Sample Bias (V)

-1,0 -0,5 0,0 0,5 1,0

-0,5

0,0

0,5

PCSH

Tunn

ellin

g cu

rren

t (nA

)

Sample Bias (V)

Fig. 9: I-V curves recorded in ambientconditions on PCSS molecules (a) onPCSH (b) and Au(111) (c). The engagetunneling current and bias voltage are 50pA and 200 mV. The error bars representthe standard deviation of the mean whichhas been calculated at each V point.

-1,0 -0,5 0,0 0,5 1,0

-0,5

0,0

0,5

Au(111)

Tunn

ellin

g cu

rren

t (nA

)

Sample Bias (V)

71

Although both mutants seem to be electronically coupled to the electrode, the origin

of the observed differences is presently not clear.

In this connection, more recently, we have revisited the conduction properties of

PCSS and PCSH by STS and Conductive AFM (CAFM) in controlled environment

[12, 14].

STS analysis was performed under nitrogen atmosphere, to reduce the water layer at

the sample interface which may play a relevant role in tunnelling mechanism [12].

Fig. 10 shows that for both pure gold substrate and PCSH on gold, I-V curves closely

resemble those obtained under ambient conditions, whereas the asymmetry observed

for PCSS molecules has almost disappeared, with a concomitant decrease of the

recorded current.

-1,0 -0,5 0,0 0,5 1,0

-0,5

0,0

0,5PCSS

Tunn

ellin

g cu

rrent

(nA

)

Sample Bias (V)

-1,0 -0,5 0,0 0,5 1,0

-0,5

0,0

0,5

PCSHTu

nnel

ling

curre

nt (n

A)

Sample Bias (V)

Fig. 10: I-V curves recorded taken undernitrogen atmosphere on PCSS molecules(a) on PCSH (b) and Au(111) (c). Theengage tunneling current and bias voltageare 50 pA and 200 mV. The error barsrepresent the standard deviation of themean which has been calculated at each Vpoint.

-1,0 -0,5 0,0 0,5 1,0

-0,5

0,0

0,5Au(111)

Tunn

ellin

g cu

rrent

(nA

)

Sample Bias(V)

72

It can be reasonably assumed, therefore, that water molecules adsorbed at the sample-

tip interface are involved in generating the observed PCSS asymmetry. Nevertheless,

it appears intriguing that PCSH presents a symmetric I-V relation under both

experimental conditions. Indeed, it is hard to conceive that PCSH would retain a

different level of humidity with respect to the PCSS mutant.

We might perhaps invoke some differences between the orientation and interfacing of

the two mutant proteins with the electrode surface. In the case of PCSH, binding to

gold through the thiol group at the carboxy-terminal free end might lead to a less

hindered protein surface-electrode coupling than that obtained via the S-S bridge.

In order to investigate the influence of the tip-molecule gap on STS spectra, I-V

characteristics of tunnelling junctions, for different values of initial tunnelling current

were further measured [15].

I-V plots recorded at fixed location on PCSS molecules by varying the initial

tunnelling current are shown in Fig. 11. I-V response is asymmetric at all the engaging

tunnelling current values examined, even though the amount of rectification decreases

as the tunnelling current is raised. These data support a notable dependence of

conductive properties on tunnelling gap width.

We also expanded the voltage range to study the effect of the applied bias on the

electrical response. A sequence of I-V curves was recorded at the same location by

extending the bias range to ± 3 V (Fig. 12). An abrupt change was observed when

applying ± 3 V with a corresponding unusual increase in current flowing.

Nevertheless, the initial conductivity was recovered as soon as the bias range was

reduced below ± 2 V. This last finding indicates that at high voltages the bio-molecule

Fig. 11. I-V characteristicsrecorded by STS on a monolayer ofPCSS on Au(111) in ambientconditions. Each curve wasrecorded at a different initialtunnelling current.

73

was not damaged, but only subjected to transient phenomena induced by the strong

local electric field applied by the STM tip.

I-V characteristics recorded at 5 and 50 pA on PCSH molecules are shown in Fig. 13.

A highly asymmetric I-V response at low starting tunnelling current is observed, as

for the PCSS mutant. However, in contrast with what observed with PCSS, the

asymmetry completely disappears as the tunnelling gap is lowered.

In summary, the STS analysis performed on single molecule pointed out that the

protein molecule conduction properties are subtly affected by the water content and

are strongly dependent on the tunnelling gap width; with some role played by the

particular way in which the proteins assemble on the substrate.

Fig. 13: I-V characteristics as recorded by STS in ambient conditions on PCSH molecules chemisorbed on Au, starting bias 0.2V.

Fig. 12: I-V characteristics recordedby STM on PCSS molecules onAu(111) in ambient conditions, atincreasing bias voltage ranges ±1 V(green curve), ±1.5 V (purple curve), ±2 V (blue curve), ±2.5 V (red curve)and ±3 V (grey curve).

74

As concerns the tunnelling gap, the conductive properties of plastocyanin mutants

have been more controllably investigated by CAFM [14, 15]. In this case the AFM tip

is positioned in close contact with the protein molecules allowing also to probe

electronic properties of the plastocyanin mutants as a function of the applied forces.

The PCSS molecules were either immobilised on gold substrates and contacted with a

gold coated tip (Fig.14a) or alternatively adsorbed on gold coated AFM tip which was

brought in contact with a flat gold surface (Fig.14b).

I-V characteristics of the PCSS immobilised on gold substrates have been recorded in

air at room temperature. For a restricted bias region (± 0.05 V), a pure ohmic I-V

relation was found for different applied forces (Fig. 15). At similar forces, as the bias

(a) (b)

Fig. 14: Schematic representation of tip-PCSS-gold substrate junctions obtained by immobilizing PCSSwith the engineered SS group either on gold substrate (a) or on gold coated tip (b). In the first case thejunction is formed by bringing the tip into contact with the protein molecules, whereas in the second casethe PCSS modified tip is approached to a flat gold surface until a physical contact is established.

-0.04 -0.02 0.00 0.02 0.04-0.010

-0.005

0.000

0.005

0.010

Cur

rent

/nA

Bias/V

-0.3 -0.2 -0.1 0.0 0.1 0.2 0.3-0.15

-0.10

-0.05

0.00

0.05

0.10

0.15

Fig. 15: I-V characteristics over ±0.05V, measured in ambient condition, for a gold coated AFM tip in contact with PCSS molecules immobilized on Au(111) substrate at applied loads of 2nN (open circle), 4nN (dashed line), 8nN (solid line). Inset shows the sigmoidal I-V relation over ± 0.3V for 2nN(open circle), 4nN (dashed line), 8nN (solid line) .

75

range was increased up to ± 0.3 V, the I-V relation showed a sigmoidal trend (inset of

Fig. 15).

For a wider bias range ( ± 3 V), two current peaks were observed at about –1.2 V and

+1.8 V (Fig. 16). I-V spectra were quite reproducible when repeated at the same

position above the sample. The two peaks seem to be reminiscent of a resonant

electron transport, which was already observed for junction including electro-active

moieties, and related to an electron transport involving the redox centre. However, we

noticed that both peak position and current intensity were found significantly

dependent on the tip position. These changes possibly reflect local variation within the

junction, in terms of different orientation and/or number of proteins involved.

Fig. 17 shows that the conductance, as derived from the slope of the linear portion of

the I-V curves, increases immediately after the AFM probe has contacted the

-3 -2 -1 0 1 2 3

-10

-5

0

5

10

Cur

rent

/nA

Bias/V

Fig. 16: I-V curves recorded for agold coated AFM tip in contactwith PCSS molecules immobilizedon Au(111) substrate. Themeasurements are carried out atroom temperature under ambientcondition by sweeping the biasbetween ± 3V at a force load of4nN.

Fig. 17: the conductance,derived from the inverse ofthe slope for the linear biasregion (± 0.05V) of the I-Vcurve, is plotted versus timefor three applied forces (opencircle), 40nN, (filled circle)20nN, (open square) 10 nN.

0 5 10 15 20 25

0

5

10

15

20

25

R-1/M

Ω-1

Time/s

40 nN 20 nN 10 nN

76

monolayer and reaches a constant regime within 8-20 s. In addition, the stronger is the

applied force the faster is the transient response observed. This effect is attributed to

the presence of attractive capillary forces generated by a thin water layer at the

protein/tip interface, when experiments are carried out in air.

To minimise the action of capillarity forces, I-V characteristics of PCSS molecules

were repeated under nitrogen atmosphere. In these experimental conditions,

conductance was observed to remain constant as a function of time (data not shown)

and the ohmic region is now extended up to ± 0.3 V (inset Fig. 18). At higher biases

(± 1 V) I-V curves present a sigmoidal shape, which is maintained for increasing

applied forces as shown in Fig.18. I-V characteristics appear slightly asymmetric,

particularly at higher applied forces, as recently seen also for azurin (Oxford group).

This asymmetry may be due either to the redox active centre asymmetrically

positioned within the junction or also to nature of the contacts at the metal-molecule

interface. Specifically, in our measurements the protein is chemically coupled to one

electrode and physically contacted with the other one.

In Fig. 19, the resistance, as calculated from the 1/slope of the ohmic region, is plotted

as function of the applied force. The resistance estimated for PCSS proteins was

found to be 109-1010Ω at forces of 3-4 nN. Interestingly, in the corresponding plot two

distinct trends are observed. The junction resistance decreases exponentially until 8

nN, whereas for forces higher than 8 nN a more rapid exponential trend is observed.

This effect may depend on the nature of the contacts and may be further related to a

larger micro contact area achieved with increasing pressure, or to changes in the

-1.0 -0.5 0.0 0.5 1.0

-8

-4

0

4

8

3.5 nN 8.8 nN 9.7 nN 11 nN

Cur

rent

/nA

Bias/V

-0.3 -0.2 -0.1 0.0 0.1 0.2 0.3-0.2

-0.1

0.0

0.1

0.2

Fig. 18: I-V curves over ±1V, acquired under nitrogen atmosphere, for a gold coated AFM tip in contact with PCSS molecules immobilized on Au(111) substrate at increasing applied forces. Inset shows the linear I-V relation over ± 0.3V.

77

electronic properties of the deformed protein. Indeed, the discontinuity in the

dependence of resistance on applied forces is likely connected to alterations of the

three dimensional structure of the protein, which suggests the presence of a critical

pressure for PCSS conduction. Generally, the behaviour observed in Fig.19 is

reversible when the forces are reduced to the initial values (about 3 nN), thereby

indicating that the PCSS molecules are only elastically compressed, and not

irreversibly deformed.

The dependence of the current on contact force was also investigated by recording

force-distance curve and current signal at a fixed bias simultaneously. When the

measurement is carried out on bare gold, no current flows before the AFM probe has

contacted the gold; while at nominal contact (zero distance) the current jumps to

values that saturate the pre-amplifier (Fig.20a). With PCSS molecules, as soon as the

tip approaches the proteins the current increases almost linearly until a force of about

8 nN (Fig.20b). Then no further increase is recorded up to 20 nN. Above this value

the current signal quickly reaches a high value beyond the pre-amplifier limit. These

2 3 4 5 6 7 8 90.01

0.1

1

10

R /G

Ω

Applied Force/nN

Fig. 19: semilogarithmic plot ofjunction resistance vs increasingcompressional forces when the goldAFM tip is brought into contact withPCSS monolayer.

0 200 400 600 800

-10

0

10

20

30

Force

probe displacement / nm

Forc

e / n

N

-8

-4

0

4

8

Current

Cur

rent

/ nA

-200 0 200 400 600 800-10

0

10

20

30

Force Current

probe displacement / nm

Forc

e / n

N

-1.0

-0.5

0.0

0.5

1.0

Cur

rent

/ nA

Fig. 20: current/force simultaneously recorded as a conducting AFM cantilever is moved toward the gold surface (a), or toward the PCSS monolayer (b) at a bias of 1V.

78

results follow the trend observed in the preceding analysis of junction resistance

dependence on contact force, thereby confirming the presence of a critical pressure for

PCSS molecules.

The conduction characterisation was repeated for the configuration depicted in Fig. 14

(b), where the bio-molecules were adsorbed on the conductive AFM probe. All major

characteristics observed on the other experimental set-up were reproduced; this

indicating that the conduction properties of the metal-PCSS-metal junction, obtained

according to the two approaches, are very likely equivalent. Thus, it is reasonable to

assume that the distance of the copper site with respect to the gold substrate has no

effect on the conduction properties of the biomolecular junction here investigated.

The only remarkable difference observed between the two junction geometries

regards the dependence of resistance on compressional forces. In fact, the pressure

value which determines a discontinuity in the molecular resistance is spread between

a few nN and tens of nN. Occasionally, very low applied forces were sufficient to

reduce drastically the junction resistance, where an applied force of 4 nN is enough to

lower the resistance of two orders of magnitude. This variability can be due either to a

different number of molecules included in the junction or to different tip radius which

affects the microcontact area and the specific resistance.

The two well defined trends observed for PCSS resistance seem to be consistent with

an electron transport mechanism which depends on structural deformations of the

compressed protein inside the junction.

The effect of compressibility on conductance was explored on PCSH monolayer as

well. The biomolecular junction was formed by contacting PCSH monecules tethered

on Au(111) substrates with gold coated AFM probes [15].

Typical I-V characteristics as recorded at different pressures are shown in Fig. 21. All

plots appear to be highly symmetric and sigmoidal in shape. The high level of

symmetry in the current signal as function of the bias is in agreement with what

observed by STS on single molecules of PCSH under controlled atmosphere [12].

Independently of the applied pressure, I-V plots show an ohmic region within ± 0.1 V.

This is a quite restricted region as compared with that of PCSS, whose I-V linear

dependence was observed up to ± 0.3 V.

79

The resistance of the protein has been studied also as a function of the force exercised

by the conducting AFM probe. Fig. 22 shows that over a range of 15 nN, resistance of

the bio-molecular junction decreases upon increasing the force. Even for this PC

mutant, the resistance is scaling with two distinct exponential laws. As seen for PCSS,

the presence of such discontinuity in the resistance–force dependence confirms the

occurrence of an electron transport mechanism dependent on structural deformations

of the compressed proteins inside the junction [14, 15].

Optical Spectroscopy (INFM-Viterbo)

The electron transfer process depends on several parameters which sometimes are

difficult to estimate; one of which being the reorganisation energy λ of the nuclear

degree of freedom. Calculation of this energy requires a knowledge of the coupling

strength of the protein vibrations to the electron transfer event. To get some insight on

such a parameter we have previously performed femtosecond laser pump-probe

experiments on azurin [4]. More recently, we have extended such a study to copper

protein plastocyanin [10].

Due to its wide spectral content, a femtosecond pump pulse creates a population in the

excited electronic state and vibrational coherence in both the ground and excited

states. A delayed probe pulse interrogates the sample and additional information (with

respect to traditional Raman) can be obtained. The evolution of the signal is therefore

modulated by the dynamics of the vibrational states. Suitable processing of this signal

-1.0 -0.5 0.0 0.5 1.0

-2

-1

0

1

2

Cur

rent

(nA

)

Bias (V)

4 nN 6 nN 8 nN 10 nN

Fig. 21: I-V characteristics as recorded by CAFMon PCSH molecules chemisorbed on Au(111) undernitrogen atmosphere.

Fig. 22: Semi logarithmic plot of resistance vs applied forces obtained by CAFM on a PCSH monolayer under nitrogen atmosphere.

80

allows one to extract the dynamics and the vibrational feature of the ground and the

excited states of the protein active site.

The results obtained for plastocyanin indicate that the excited state, involved in the

charge transfer, decays in a non radiative fashion within about 300 fs (Fig.23).

The vibrational coherence modulating such a decay, when analysed in the frequency

domain, reveals all the frequency modes of the conventional Resonance Raman signal

(Fig.24). Additionally, some low frequency modes, of collective character and

probably of some biological relevance [4, 10], are observed in these spectra.

Moreover a 500 cm-1 mode, not present in the conventional Resonance Raman

spectrum and already observed in other copper proteins, appears. An analysis in the

time domain indicated that, very likely, it belongs to the vibrational manifold of the

excited state.

Fig. 23: Top: differential opticaltransmission of PC. The solid line isthe best fit of experimental dataobtained with a function

τ/)( teBAty −⋅+= . Inset:wavelength-resolved pump-probesignal of buffer solution containingno protein. Bottom: oscillatorycomponent (residual) obtained aftersubtraction of the exponential fitfrom experimental data of the upperplot.

0 500 1000 1500 2000-0,4

-0,2

0,0

0,2

0,4

time delay (fs)

time delay (fs)

resi

dual

(arb

.u.)

0 500 1000 1500 2000-0,2

0,0

0,2

0,4

0,6

∆T/T -400 -200 0 200 400 600 800

0,0

0,5

1,0

81

In order to confirm such findings an MDS study of the excited state of plastocyanin

was carried out [10]. According to Ref. 2, MDS of hydrated plastocyanin has been

performed by using the CHARMM package with Charmm27 as force field including

TIP3 model for water. By such approach the main Resonance Raman experimental

features around 400 cm–1 have been reproduced by calculating the Fourier transform

of the Cu-S(Cys84) distance autocorrelation functions. Similarly, the excited state of

the copper active site was modelled by increasing the equilibrium bond length in the

harmonic Cu-S(Cys84) interaction and by varying the force constant of the same

bond. According to this procedure the power spectrum obtained is shown in Fig. 25. A

0 200 400 6000.0

0.2

0.4

0.6

0.8

1.0 508

426

365

305

263

7550

22

frequency (cm-1)

inte

nsity

(arb

.u.)

Fig. 24: Fourier spectrumof the residual shown inFig. 23 bottom.

100 200 300 400 500 600 7000.0

2.0

4.0

6.0

8.0

P(f)

(arb

. u. )

frequency (cm-1)

Fig. 25: Power spectrum, as function of frequency, of the Cu-S(Cys84) bond distance fluctuation analysed for 800 ps (see Ref. 10 for details).

82

main peak at 511 cm –1 appears, thus representing a vibration mode of the excited

state of plastocyanin. The fairly good agreement between the MDS and the

experimental results gives on one hand support to the experimental findings and, on

the other, strengthens the role of MDS in modelling the electron transfer properties of

copper proteins.

Molecular Dynamics simulations (INFM-Viterbo)

In order to get additional information on the self-assembling of plastocyanin mutants

on gold substrate, we have conjugated our nanoscopic experiments with MDS studies

[8, 9, 15]. In our model, the interaction between the protein and the gold substrate

involves one or two sulfur atoms from the biomolecule, each interacting with three Au

atoms. We run a 10 ns-long MDS of plastocyanin mutants by using a Charmm force

field for the protein atoms, an TIP3P for hydration water (one layer of water molecule

as shown in Fig. 26) [8, 9].

The root mean square fluctuations (RMSF <∆r2>1/2) of both plastocyanin anchored

with a single sulfur atom (PCSS-I) (<∆r2>1/2=(0.013± 0.005)nm) and with two sulfur

atoms (PCSS-II) (<∆r2>1/2=(0.015± 0.005)nm) are significantly lower in comparison

to that of non immobilised PCSS (<∆r2>1/2=(0.03± 0.01)nm). Accordingly, the

flexibility of the macromolecule appears affected by the immobilization onto the

substrate. The observation that larger RMSF values are revealed in PCSS-II with

respect to PCSS-I suggests that the presence of two covalent bonds might yield a

larger flexibility at least in same regions of the macromolecule.

The trend with time of θ and Φ (namely the angles describing the orientation and the

precession, respectively, of the protein p axis with respect to gold surface normal),

83

calculated during the MD trajectory, are plotted, for both PCSS-I and PCSS-II, in Fig.

27. In both cases, a significant deviation of p from the normal axis is registered. This

means that both the gold-anchored systems can assume different orientations with

respect to the gold electrode. In addition, the mean value of θ is smaller for PCSS-I

than for PCSS-II (see legend of Fig. 27); this means that the protein anchored by a

single bond assumes a configuration closer to the normal. Such behaviour might be

indicative of a less extensive contact, for PCSS-I, between the protein and the gold

atoms.

Furthermore, we note that θ exhibits a larger standard deviation for PCSS-I with

respect to PCSS-II (see legend of Fig. 27); such a behaviour being in agreement with

a higher degree of freedom for PCSS-I with respect to PCSS-II. Concerning Φ, PCSS-

I and PCSS-II show different average values, reflecting a different average position of

the macromolecule with respect to the gold substrate. Interestingly, a much larger

standard deviation of Φ is observed for the PCSS-I in comparison with PCSS-II (see

legend of Fig. 27).

Such a result, together with what observed for θ, is indicative that the macromolecule,

during its dynamical evolution, can assume different arrangements with respect to the

gold surface; such a variability being more marked in the case of a single anchoring

Fig. 27: Temporal evolution, along a 10 nsMD trajectory, of θ and Φ anglesdescribing the orientation and theprecession, respectively, of the protein paxis with respect to gold surface normal,for the PCSS-I (black) and PCSS-II (red).The mean value and the standard deviationof θ are (27± 3)° and (34± 1)° for PCSS-Iand PCSS-II, respectively. The mean valueand the standard deviation of Φ are (116±10)° and (99± 3)° for PCSS-I and PCSS-II,respectively.

0 4000 8000

80 120 160 PCSS-I

PCSS-II

φ (deg

)

Simulation time

0 4000 800010 20 30 40 50

PCSS-I PCSS-II

θ

84

point. Notably, the trend of Φ for PCSS-I shows, over the fluctuations, oscillations

whose presence points out a sort of periodicity in the lateral movements of the

macromolecule anchored by a single sulfur-gold bond.

To compare MDS and AFM data, the protein anchored on gold was described by an

ellipsoid centred at the centre of protein mass and oriented according to its main

inertial axes (Fig 28). The ellipsoid and its corresponding height over the gold

substrate, extracted from the MDS trajectories, are shown in Fig. 29 by sampling

structures every 10 ps. It comes out that a single mode distribution is registered for

both PCSS-I and PCSS-II. The mean values are, in both cases, close to that expected

for the crystallographic data by assuming that protein is anchored to gold via the

disulfide bridge (h ~ 3 nm).

On the other hand, PCSS-I is characterized by a mean value slightly higher than that

of PCSS-II (see the mean values in Fig. 29). Such a finding is consistent with the

2.6 2.8 3.0 3.2 3.40.0

0.1

0.2

0.3

0.4mean=2.94 nmσ=0.02 nm

PCSS-II

Frac

tion

of m

olec

ules

Height (nm)

2.6 2.8 3.0 3.2 3.40.00

0.05

0.10

0.15

mean=3.05 nmσ=0.08 nm

PCSS-I

Fig. 29: Statistical analysis ofthe molecular height above theAu(111) substrate, as extractedfrom the MDS trajectories forPCSS-I and PCSS-II. Thevertical dimension of proteinshas been estimated from 1000molecules sampled every 10 psof the MDS trajectory.

85

previous results showing that the protein anchored by a single bond can assume an

average configuration closer to the normal to the surface.

For both systems, a spread of heights is registered (see the standard deviations in Fig.

29). Generally, this means that, during its dynamical evolution, the macromolecule,

by exploring its accessible conformations, can assume a variety of arrangements, and

then of heights, with respect to the gold substrate. On the other hand, we note that

PCSS-I is characterized by a wider distribution in comparison with PCSS-II. As a

consequence, molecules covalently bound to gold through a sulfur atom show slightly

higher flexibility than molecules anchored via two covalent bonds.

These results are in a qualitative agreement with those derived by TMAFM. However,

we note that the height distributions, as extracted by MDS, result markedly narrower

if compared to those derived by TMAFM. To explain such discrepancies, different

effects should be taken into account. First, we remark that TMAFM data refer to a

collection of molecules, likely in different starting arrangements, which, moreover,

may be anchored to gold by one or both sulfur atoms, whereas MDS considers only a

single molecule and the two different ways of anchoring separately. Additionally, it

should be taken into account that a further contribution to the structural and dynamical

heterogeneity of the adsorbed protein could arise from the complex interaction

between the protein and the gold surface. A more accurate modelling of this

interaction should be introduced in the MDS. Finally, it should be stressed that

TMAFM is sensitive to protein movements on a much longer time scale (at least

milliseconds if we refer to a tapping frequency in the range of kHz). Therefore, it can

be hypothesized that the flexibility of the molecules, on this temporal scale, above the

substrate can also contribute to the broadening of the experimental height distribution;

during the measurements the proteins assuming different arrangements. Nevertheless,

we remark that, even in the restricted temporal window as explored by MDS, our data

show that the proteins can assume a variety of orientations with respect to the gold

substrate.

Conducting Probe AFM of azurin (Oxford)

The processes associated with electron transfer (ET) through a protein matrix lie

central not only to bioenergetics reaction pathways but are also of fundamental

importance in any consideration of biomolecule based electric components.

Switchable, conformationally robust, metalloproteins of nanometer dimension

86

constitute interesting candidates in the integration of biomolecule with molecular-

scale electronic circuitry. In recent years the Oxford team been engaged in a large

program of work in which the transport properties of single molecules and molecular

associations have been analysed within a variety of proximal probe and lithographic

circuitries [11, 16-21]. This report summarises recent progress in conductive probe

AFM based analyses as funded by SAMBA. Specifically, imaging and spectroscopy

have been utilised to investigate electronic properties of a variety of mutants of the

blue copper protein azurin (K27C and zinc K27C) assembled on gold electrode

surfaces under a variety of controllable environmental conditions. Though one can

envisage the use of several methods for investigating the electrical characteristics of

protein/electrode assemblies few, if any, possess the capability of resolving the

contributions of individual molecular components and it is, indeed, these which are of

most relevance to the aim of generating reproducible devices based on single

molecules. Proximal probe analyses can be carried out with exquisite lateral

resolution, with minimal (though tightly controlled) sample preparation, and under a

variety of conditions (ambient, UHV, under fluid). Here we focus on experiments in

which a metallic AFM probe and metallic or semi-metallic substrate constitute source

and drain electrodes. Gold or platinum coated AFM tips can be brought to contact a

molecular monolayer with a controlled contact force. By applying a voltage bias

between the substrate and conducting cantilever, a current flow is generated. Since

one of the key characteristics of biomolecular assemblies is their sensitivity to surface

chemistry, mechanical perturbation and environment, an understanding of the effect

of this sensitivity on transport is required. In consideration of dielectric stability and

mechanical stability we have characterised transport as a function of mechanical

pressure and within this calculated (and simulated) perturbations of tunnel barrier,

resolved the force dependence of dielectric breakdown and calculated molecular

compressibility. An additional principal aim of this project was to investigate the role

of the redox centre in mediating current flow through a single azurin molecule.

Although interesting I-V characteristics for specific junctions have been reported,

understanding of the full spectrum of factors influencing the electrical properties of

metal-molecule-metal junction remains of great interest. Specific mechanisms are, in

many cases, largely unresolved. Examining the dependence of the junction resistance

(or conductance) on presence/absence of redox centre is one approach to examining

the mechanism of transport in potentially switchable configurations.

87

Azurin functions as an electron carrier in the respiratory chain (Pseudomonoas

aeruginosa) and has been characterised in considerable detail. By using a K27C

mutant we are able to control protein-electrode interaction/coupling (and, therefore,

electrochemical switching) through a gold-anchoring cysteine residue onto the surface

of the protein.

Azurin (Pseudomonas aeruginosa azurin wild-type and all surface cysteine mutants)

was kindly supplied by Professor Gerard Canters, Leiden Institute of Chemistry. All

measurements were made using a PicoSPM conducting AFM (Molecular Imaging,

Phoenix) with commercially available Si3N4 tips (tip radius=15±5 nm, Spring

constant=0.12 N/m) coated with Au or Pt. The tips were prepared by soaking in a ca.

5 µM azurin acetate buffer (pH=4.64) for more than 2 hours. Then the modified tip

was brought into contact with a highly oriented pyrolytic graphite (HOPG) substrate.

The schematic of experimental setup was shown in Figure 30.

The bias is applied to HOPG through the electronic circuit, so that the potential could

be expressed as the tip potential vs. substrate. During the acquisition of each current-

voltage (I-V) curve, the force was kept constant and accurately quantified. In order to

study the force dependence of the conductivity associated with the protein structure

distortion, I-V curves at different forces were collected. At every fixed force level, 20

Fig. 30. Schematic of a CP-AFM experimental configuration. The molecule of interest is effectively sandwiched between the electrical contacts formed by a metallic probe and an underlying planar electrode surface.

88

I-V curves were accumulated to give an average value of resistance. All conducting

AFM experiments were carried out under ambient conditions (20±2 °C and 40~50 %

humidity).

Low force dielectric breakdown of metalloprotein (Oxford)

At contact forces < ~ 4 nN the transport characteristics of metalloprotein junctions

differed significantly from those observed at higher (more “robust” contact) force.

Specifically, negligible current flow could be detected prior to the onset of dielectric

breakdown. Typical I-V relation profiles recorded during positive bias scans (curve i,

ii and iii) at a loaded force less than 2 nN are shown in Figure 31a.

Below the breakdown threshold, the current flow lies below our detection limits (5

pA). At breakdown, the sudden increases in transport leads to pre-amp saturation.

These I-V relations are highly reproducible but irreversible with respect to reversed

bias sweep (as shown in the curve iv). The current remains over 10 nA until zero bias,

and then quickly decreases to -10 nA, which is the lowest current measurements limit.

This behaviour is very similar to that of a direct contact between the conducting tip

and substrate. Similar behaviour, attributed to dielectric breakdown under high (>109

V/m) field has been observed with alkanethiol self-assembled monolayers under very

comparable experimental conditions [22, 23].

The scan speed dependence of the breakdown voltage has been investigated as well.

The I-V curves with different ramp speeds (dV/dt) of the applied voltage were

recorded as shown in Figs 31a. Curves i, ii and iii correspond to scan speeds of 0.5, 10

Fig. 31. I-V characters of a gold-azurin-HOPG junction at low (< 2nN) force. a) Positive bias scani) 0.5V/s; ii) 10 V/s; iii) 50V/s. b) Negative bias scan at the ramp speed of 10 V/s. All suchexperiments were carried out using a “tip modification” strategy. The tip potential vs HOPG wasramped from -1V~ 9V, then scan back to -1V in positive scan and from 1V-~-9V, then scan back to1V in a negative scan.

89

and 50 V/s, respectively. The corresponding breakdown voltage was summarised in

Table 1. Obviously, high scan speed (50 V/s) leads to remarkable positive shift of the

breakdown I-V curves. However, no obvious difference of I-V curves or breakdown

voltage was found at the scan speeds of 0.5 and 10 V/s. The breakdown occurring

within bulk materials, micrometer dielectric film for example, has been studied

intensively. It was concluded that the breakdown took place under only enough charge

accumulation at both film sides and as such occurred at higher field when charge

injection times were lower. At 10 V/s ramp speed (or lower) charge

injection/accumulation is effective enough for breakdown to be both highly

reproducible and independent of bias sweep rate. The breakdown voltage can be used

to roughly estimate the dielectric strength (the dielectric strength is the maximum

working voltage a material can withstand without breaking down. It is normally

expressed in Volts / mm of protein assuming vertical dimensions comparable to those

resolved within the protein crystal structure (~3 nm). The dielectric strength of protein

is between 1.0 ~ 1.4 GV/m, which is smaller than 2.0 GV/m for alkyl monolayers on

either metal or semiconductor, but greater than physically touched alkyl bilayers

formed between Hg drop and Ag plate (~0.5 GV/m). This value is also comparable

with that of widely used inorganic dielectric materials, e.g., SiO2 (0.8~1.3 GV/m),

indicating that, under conditions of low molecular compression (low electrode contact

force) this metalloprotein behaves as an effective molecular-scale dielectric barrier. At

increased contact force, transport characteristics are quite different.

Transport charactersitics at force between 5~70nN (Oxford)

Current-Bias (I-V) Relations

Typical I-V characteristics of a single azurin molecule confined within a CP-AFM

Table 1. Tthe ramp speed dependence on the breakdown voltage of protein under force 2nN.

90

tunnel junction at forces between 5~70 nN are show in Figure 32 (note: though the

absolute magnitude of current varies with force the symmetry and general

characteristics of the plot do not). The dependence of transport on applied electric

field can be considered within two regimes; there is a low-field region (V < 0.05

V), where current versus voltage follows a linear relationship. The electrical

resistance of the protein molecule can be thus derived from the slopes in the low field

region. At a higher electric fields, the variation of current as a function of voltage is

clearly non-linear.

The I-V characteristics resolved by conducting AFM junctions show slight asymmetry

with respect to the coordinate origin. In previous work, we have resolved greater

asymmetry in the tunnelling junctions

fabricated in an STM configuration and ascribe the difference to the more pronounced

contact asymmetry in STM junctions. In a conducting AFM junction the two contact

electrodes (tip and the underlying substrate) are associated with similar tunnelling

distances. The slight asymmetry in transport may be assignable to geometric or

chemical asymmetry within the junction (the protein is in physical contact with one

electrode and chemisorbed on the other). The I-V characteristics show weak bias scan

91

speed dependence in that no significant change is observed across a 0.1 to 40 V/s

range. Beyond this range, the charging current is considerable at speed higher than 40

V/s, and the uncertainty of the measurement makes the I-V curves unsmooth at speed

slower than 0.1 V/s. It is important to note that transport through the metalloprotein

appears, under all vertical configurations examined, to be non-resonant; that is, the

tunnelling electrons do not appear to be appreciably occupying electronic states

associated with the redox-active metallic centre. On expanding the voltage range over

which bias is swept negative differential resistance phenomena (which have been

observed by us and others with redox-active tunnelling systems) are occasionally

observed and are the subject of further investigation.

Time Dependence of transport characters under fixed force

In order to monitor the dynamic changes of protein transport under a fixed force,

resistance measurements can be made with reasonable temporal resolution (Figure

33).

The general trend is that, after an initial exponential decrease, molecular resistance

tended to become stable (typically this “equilibriation occurs over 10-12 seconds).

These temporal changes are likely to be associated with force induced compression of

the protein fold and a decrease in both tunnelling distance and tunnel barrier [21]. On

92

reaching a state of the minimum electrical resistance, transport is stable. All the data

reported herein were collected after the force has been exerted for 12 s. Simulations of

the resistance vs time data allow an exploration of the dynamics within such junctions

and the resolution of a first order rate constant of ~ 0.07 s-1.

It should be noted that, since a tip-modification strategy is used in these experiments,

data sets are less subject to the effects of lateral drift.

The force dependence of molecular transport

Though one expects the transport properties of a biomolecule to be highly sensitive to

force very little has been reported on the compressable mechanics of protein. In this

work, a dynamic evolution of protein resistance associated with the protein structure

was monitored in a wide range of force load. The extreme force that leads to complete

collapse of protein structure has been used to evaluate the mechanical strength of

protein. As shown in Figure 31, at the force less than 2 nN, almost no current flows in

the bias region between -1 and 1 V. On application of force > 5 nN reliable electric

contact and transport can be established from which molecular resistivities can be

deduced. The protein resistance (R=dV/dI V=0) corresponding to each applied force

was averaged from 20 I-V curves, and the values are given in Figure 34. It is clear that

the log(R) decreases almost linearly until 40 nN, and then, it decreases slowly. In

consideration of the factors which may contribute to the resistance decrease under

compressional force, one may, in the first instance consider changes in protein-

electrode contact area. Since current across a section is proportional to the section

area, one would expect a reciprocal decrease of resistance. Since resistance decreases

almost exponentially with time, it is likely that surface area changes are not dominant.

Molecular deformation may result in a reduced source-drain tunnelling distance, i.e.,

decreased vertical dimensions of the protein. Previous investigations have shown [20,

21] that nonresonant electron tunnelling through organic molecules exponentially

increases with decreasing tunnelling distance. When the applied force is greater than

40 nN, log(R) changes slowly, indicating the repulsion among the atoms in protein

becomes stronger. Simulations have suggested that repulsive forces within the protein

fold drive the molecule to deformation in plane. An evident of plastic protein

deformation is given in Figure 34 as force is withdrawn. The resistance trace

corresponding to increasing the force and decreasing the force cannot be overlapped.

93

Since the change of resistance is directly related to the deformation of the protein

vertical dimension, this observation is consistent with plastic molecular compression.

Similar phenomena have been observed recently in a micro slide contact measurement

[24]. When the force load exceeds ~87 nN, the current suddenly jumps from -10 to 10

nA at a bias near zero consistent with direct source-drain mechanical contact. The

force (pressure) at which this mechanical collapse occurs can be utilised in an

estimation of protein mechanical strength.

Molecular Resistance

The resistance associated with the gold-azurin-HOPG junction under 5 nN of force

(the mimimum force at which reliable transport through the molecules can be

achieved, as discussed) has been statistically analysed and is summarised in Fig. 35. A

Gaussian fit gives an average resistance of 4×1010 Ω (40 GΩ). Since this is several

orders of magnitude greater than the contact resistance associated with bringing the

tip and substrate into direct mechanical contact, we assign the resistance of the “single

protein junction” to one protein molecule. This resistance is considerably larger than

that associated with n-alkanethiol SAMs as measured by Cui et al [25]. An

assumption (backed up by the results as reported) that transport is tunnelling

dominated allows us to extrapolate a semi-logarithmic plot obtained with alkanethiols

94

to the vertical dimensions associated with a single azurin molecule. Such a plot is

shown below (Figure 36) and indicates a resistance, at 5 nN, comparable to that

associated with a C15 alkanethiol. Since the expected tunnelling distances for the

uncompressed protein lies close to double that of a C15 alkanethiol these observations

are consistent with considerable vertical compression of the molecule in the tunnel

junction.

Fig. 35. Resistance histogram (red bars) and Gaussian fit (solid blue line) of a single Cu centre azurin metalloprotein at a loading force 5nN.

Fig. 36. A semilogarithmic plot of SAM junction resistanve versus SAM thickness (number of methylenes). The measured resistance of a single azurin molecule lies close to that of a C15 chain alkyl thiol.

95

to the vertical dimensions associated with a single azurin molecule. Such a plot is

shown below (Figure 36) and indicates a resistance, at 5 nN, comparable to that

associated with a C15 alkanethiol. Since the expected tunnelling distances for the

uncompressed protein lies close to double that of a C15 alkanethiol these observations

are consistent with considerable vertical compression of the molecule in the tunnel

junction.

The resistance of octanedithiol molecule, the decanedithiol molecule and

dodecanethithiol molecule are 965 MΩ, 2.89 GΩ and 8.26 GΩ respectively [25]. The

resistance of benzenedithiol molecule and benzenedimethanethiol molecule are 1.2

MΩ, 21 MΩ respectively [26]. For a system in which transport is likely to be

dominated by non-resonant tunnelling through a saturated hydrocarbon-dominant

medium, one fully expects molecular resistance to be significantly greater than that

observed through shorter molecules or those associated with delocalised pi systems.

In an attempt to further confirm the role or otherwise of the metallic redox state in

tunnelling across a single metalloprotein molecular junction we have carried out

preliminary studies with a redox-inactive, zinc-substituted, protein. The I-V curves at

5nN loading force on the Cu centre azurin and Zn centre azurin were measured

repeatedly to determine their resistance. Although the data presented in Figure 37

indicates the absence of resonance or negative differential resistance with the zinc

Fig. 37. Typical comparative I-V curves of Cu azurin and Zn azurin at loading force of 5 nN.

96

holo protein (even across this large range of bias) a significant number of equivalent

scans with the copper protein are associated with current increase as negative bias.

Though the molecular resistances are comparable at low bias, that associated with the

copper protein drops drastically below that of the zinc protein at larger bias voltages.

The details of this phenomena are currently the subject of more detailed investigation.

Conclusions and future work

In interfacing man-made electronic components with specifically-folded

biomacromolecules, the perturbative effects of junction structure on any signal

generated should be considered. In this study we have established mechanisms by

which the transport properties of single metalloproteins can be characterised.

Specifically, we report herein on the electron transfer characteristics of the blue

copper metalloprotein, azurin, as characterized at a refined level by conducting atomic

force microscopy (CAFM). At low contact forces, the molecules behave as

nanometre-scale dielectrics prior to breakdown at fields in the region of 3~5 V. At

forces > ~ 5 nN reliable electrical contact is achievable from which a detailed study of

molecular resistance as a function of molecular perturbation is possible. Within a low

bias regime, the I-V curve is linear and the molecular resistance can be obtained by

the slope of the I-V curve. The modulation of current-voltage (I-V) behaviour with

compressional force has been examined. In the absence of assignable resonant

electron tunneling within the confined bias region, from -1 to 1 V, the I-V behaviour

has been analysed with a modified Simmons formula. In order to interpret the

variation of tunnelling barrier height and barrier length obtained by fitting with the

modified Simmons formula, an atom packing density model associated with protein

mechanical deformation was proposed and simulated by molecular dynamics. The

barrier heights determined at the minimum forces necessary for stable electrical

contact correlate reasonably well with those estimated from bulk biophysical

(electroanalytical and photochemical) experiments previously reported. At higher

forces, the tunnel barrier decreases to fall within the range observed with saturated

organic systems. The force dependence of the tunnelling distance allowed us further

analyse the stress-strain properties of protein, from which a Young’s modulus of (1.4

± 0.1) ×1010 N/m2 was derived.

Further work will be associated with refining the contributions of different

components of a protein fold to transport and the prospects associated with

97

electrostatic gating of conductance (most relevant to the concept of the “single

metalloprotein transistor”).

Deliverable D3.2: demonstration of a surface assembled enzymic nitrite sensor

(Oxford)

Typical nitrite analyses utilise spectroscopic or potentiometric ion-selective electrode

methodologies [27, 28] though these approaches are either practically slow or suffer

from the effects of interferents, particularly chloride (a problem pertinent to analyses

of waste and drinking water). The utilisation of an enzyme-based detection system has

advantages not only in terms of improved selectivity but is also of value in enhancing

our understanding of the natural electron-transfer processes associated with these

complex macromolecules. Enzymes provide substrate-specific responses to nitrite

with less likelihood of interference from other compounds. The redox activity of the

enzymes can be monitored through the current passed on to the electrode. We report

several approaches to obtain quantifiable nitrite turnover from electroanalyses of the

enzyme nitrite reductase. Analyte presence in the low µM range can be detected from

using either the inorganic compound ruthenium hexamine as an electron mediator or

the natural partner of the enzyme, pseudoazurin.

Experimental

Materials. 1-(3-dimethylaminopropyl)-3-ethyl carbodiimide hydrochloride 98+%,

hexaamineruthenium(III)chloride, 3-mercaptopropionic acid 99+%, and

mercaptosuccinic acid 97% were purchased from Aldrich. Ethanol analytical grade

was obtained from Riedel-DeHaen. Catalase E. C. 1.11.1.6 from bovine liver 30

mg/ml 30,000 units/mg prot was purchased from Sigma. All reagents were of the

highest grade available and used without further purification. Trizma Base and N-

hydroxysuccinimide 98% were obtained from Sigma, aluminium oxide 0.015 µm,

sodium nitrite, sulfuric acid 99%, di-sodium hydrogen orthophosphate 12 hydrate, and

potassium di-hydrogen orthophosphate were from BDH. Ultra pure water (18.2 M.

Elga, UK) was used to make fresh sodium nitrite solutions on a daily basis. The Lys-

Cys-Thr-Cys-Cys-Ala hexapeptide was synthesised by Dr. G. Bloomberg, (University

of Bristol, Department of Biochemistry) using a Pioneer Peptide Synthesiser (Applied

Biosystems) and subsequently purified by HPLC (using a Jupiter RPC18 column) to a

purity of or greater than 95%, and stored at -40oC under argon. Nitrite reductase from

98

Alcaligenes faecalis S-6 was provided by University of Leiden. Reductase

concentrations were determined from adsorbance at 280 nm (ε280 = 46000 M-1 cm-1).

Both proteins were stored at -80 °C in 100 mM phosphate buffer, pH 6.

Voltammetric apparatus and procedure. Electrodes were made of 3 and 4 mm

diameter gold and 9 mm diameter glassy carbon rods cast in epoxy resin as described

elsewhere [29]. Electrodes were polished with 0.015 µm alumina and sonicated in

deionised water before cycling in 1 M sulphuric acid as a systematic cleaning

procedure. A sonic bath SC-52 from Sonicor Instrument Corporation (New York) was

used in this process. Electrochemical measurements were carried out on an Autolab

PGSTAT 12 computer controlled (Dell GX110MT) potentiostat (Eco-Chemie

Utrecht, Netherlands) with a standard three-electrode configuration. The reference and

counter electrodes were a saturated calomel electrode (Radiometer Copenhagen) and

platinum gauze respectively. The contents of the working compartment and solutions

were degassed with high purity (>99.9 %) argon unless stated otherwise. All

potentials are reported with respect to the saturated calomel electrode (SCE), and all

experiments carried out at 22 °C.

Preparation of hexapeptide-modified gold electrode. The surfaces of freshly-cleaned

gold electrodes were modified by cycling +0.7 V vs SCE to –0.7 V vs SCE in the

presence of ~1 mM peptide as described previously [30]. The process of surface

modification is followed both capacitatively and through the progressive inhibition of

diffusive oxygen reduction at the electrode surface.

Electrode modification for direct electrochemistry of nitrite reductase. Freshly

polished 4 mm gold working electrodes were incubated with 3-mercaptopropionic

acid for 1 hour, and sonicated in ethanol. The modified electrodes were incubated for

30 min in a pH 8 phosphate buffer solution of 75 mM 1-(3-dimethylaminopropyl)-3-

ethyl carbodiimide hydrochloride, and 15 mM Nhydroxysuccinimide at 4 °C

according to the method described by Patel et al. [31]. The electrodes were then

rinsed with dionised water and incubated with nitrite reductase overnight.

99

Direct voltammetry of the reductase

Surface cysteine mutant assembly on bare gold.

Nitrite reductase from Alcaligenes faecalis strain S-6 L93C mutant, with its solution

exposed surface cysteine residue, was designed to self-assemble on planar gold

electrode surfaces in the formation of assembled arrays from which direct

electrochemistry would be possible (a fairly unique approach to a difficult problem).

The enzyme if found to spontaneously chemisorb on freshly cleaned gold electrodes.

Associated with this process is the appearance of a voltammetric signal which appears

to arise from the type I copper centre. This signal continues to grow in magnitude on

incubation then stabilise at a coverage, stable to rinsing in buffer, of 8.9x10-10 mol cm-2.

Two oxidation peaks are observed; the most anodic signal is observed at 0.36 V vs

SCE, consistent with the type 1 Cu system (Fig.38). A second anodic signal is observed

at –0.15 V vs SCE (not shown below). Though this is possibly associated with the

presence of peroxide, a reliable assignment is not yet possible. The reduction signals

corresponding to these anodic processes are rather broad and poorly defined. Upon

addition of excess nitrite to solution, the voltammetry associated with the type 1 centre

changes; specifically, the anodic peak diminishes in intensity whilst the cathodic trace

becomes catalytic. These observations, made only at mM concentrations of substrate,

are consistent with the assignment of the more anodic response being associated with

the type I centre.

Fig 38. Partial voltammogram of the nitrite reductase L93C mutant assembled on a gold disc electrode, before (dotted line) and after (solid line) 200 mM nitrite addition. The counter electrode is a platinumwire and the reference electrode is saturated calomel. The scan rate is 20 mV s-1.

100

Though reproducible, these responses are stable only over a period of 1-2 hours. We

conclude from this preliminary study that, though the enzyme can be engineered so as

to be conducive to both direct (bare single crystalline gold) surface immobilisation

and direct voltammetry, obtained responses are not sufficiently well defined nor

stable for robust sensor fabrication. More reliable immobilisation and voltammetry

can be obtained through the use of electrode bound self assembled monolayers.

Direct SAM adsorbed voltammetry of nitrite reductase.

Nitrite reductase covalently bound to a 3-mercaptopropionic acid SAM modified gold

undergoes reliable direct electron transfer (20 mV/s, pH 6.0). Under these specific

conditions, the midpoint potential, Em calculated as (Epred + Ep

ox)/2 is 150 mV vs SCE.

In our conditions, the peak-to-peak separation, (∆Ep = Epox - Ep

red), is 100 mV at 100

mV s-1, which is larger than expected for an adsorbed electrochemically reversible

process. However, the magnitude of the ratio of the peak currents (Ipred / Ip

ox) is close

to unity, and ∆Ep remains stable around 100 mV as the scan rate is increased over the

range 20 to 20000 mV s-1, thus allowing a good estimation of the electron transfer

rate constant we were able to plot the peak separation against scan rate and deduce the

electron transfer constant using the Butler-Volmer equation according to the method

described by Jeuken et al. [32]. The potential values are averaged from three sets of

experiments Figure 39. The calculated electron transfer constant is k0 = 9 ± 3 s-1. This

Fig. 39. Trumpet plot showing the peak separation of nitrite reductase covalently attached to 3-mercapto propionic acid modified gold against log of the scan rate, k0 is deduced from the datafitting. From the integration of the peak currents averaged from three different experiments, theprotein density at the electrode surface is 1.23x10-10 mol cm-2 this is compatible with a monolayer ofprotein molecules of 5-6 nm diameter containing 3 detectable redox sites (within the trimer) [40].

101

signal remains unchanged on the addition of nitrite to solution.

Diffusive phase pseudoazurin voltammetry

The diffusive electrochemical response of pseudoazu at a hexapeptide-modified gold

electrode is shown in Figure 40. At a protein concentration of 1 mM, no voltammetric

signal is observed in the absence of the peptide modifier. The diffusive current

response at hexapeptide-modified gold electrode shows an expected linear variance

with square root of scan rate. Faradaic peak separations were reproducibly close to 80

mV at slow scan rates (≤ 100 mV s-1), consistent with a quasi-reversible

electrochemical process. Voltammetric analyses across a pH range of 6-9 were

consistent with minimal (<20 mV per pH unit) dependence. The variance of peak

separation with scan rate can be used to calculate the heterogeneous electron transfer

rate constant according to the formulism of Nicholson and Shain [34]. Within this the

electrochemical rate constant is directly related to ∆Ep in semi reversible

electrochemical systems. The relation is given by equation (1).

0aDks

πγ

ϕα

= (1)

Fig. 40. Cyclic voltammograms obtained from a pH 6.0 phosphate buffer (100 mM) solution of 1 mMpseudoazurin (dotted) in the presence of 500 µM nitrite reductase (grey) and 5 mM nitrite (black), at a 3 mmdiameter peptide modified gold.

102

Where:

rDD0=γ (assumed to be 1) and a = nFv/RT

ϕ is a dimensionless parameter deduced from working curves. D0 is the diffusion

coefficient of the oxidized form and Dr of the reduced form and ks is the

heterogeneous electron transfer rate constant. The transfer coefficient, α, denotes the

position of the transition state associated with electron transfer with respect to the

oxidised and reduced forms, here taken to be 0.5 as the intensity of the reduction and

the oxidation currents are very similar. In all calculations, n is taken to be 1.

The values for ϕ were calculated from the experimental values for ∆Ep, using working

curves [34] (independent of salt concentration over a 20-200 mM range), ks values

were then calculated according to equation 1. By fitting experimental data to working

curves, ks values, calculated through equation 1, were ks = (3.3 ± 0.6) x 10-3 cm s-1.

This value was observed to show some dependence on the quality (determined

voltammetrically) of the peptide layer formed. At varying concentrations of protein,

no peak potential shift was observed (Astier et al. Electroanalysis in press) thus

demonstrating that no protein adsorption is taking place on the adlayer.

Pseudoazurin Interactions with nitrite reductase (“natural partner enzyme

mediation”)

A significant perturbation of the pseudoazurin diffusive voltammetric signal is

observed when nitrite reductase is added, in molar excess (0.5 mM) to the solution

(Fig. 42). Specifically, the quasi-reversible response becomes sigmoidal in appearance

– an observation fully consistent with the occurrence of an electrocatalytic reaction

with the enzyme (though the diffusive half wave potential is perturbed by interactions

with the reductase, the accompanying changes in wave shape make this difficult to

reliably quantify). In solution, both the pseudoazurin and the reductase are able to

adopt relative orientations conducive to the formation of a transient association

complex. In the absence of nitrite, and presence of low levels of dioxygen, it is likely

that these observations are indicative of peroxide generation by the reductase (the

electrocatalytic signal has a smaller peak current intensity when the cell is kept under

argon and diminishes on vigorous degassing). We see a minimum interference by

oxygen on nitrite turnover (within experimental error).

103

One can define the electron transfer steps in this electroanalytical system as follows:

At the electrode:

(i) Pseudoazurin (ox) + e- → Pseudoazurin (red)

The diffusive-phase peak current (Id in amperes) associated with step (i) is given by

the Randles-Sevcik

equation:

Id = 2.69 x 105 n3/2 A C D1/2 v1/2

For fixed values of A (here 0.07 cm2) and C (here 0.5 µM), a plot of Id vs square root

of scan rate allows a direct determination of the protein diffusion coefficient, D, as

(1.13 ± 0.2) x 10-6 cm2 s-1, a value in good agreement with previous determinations

[35].

In the solution:

(ii) Pseudoazurin (red) + NO2- →NiR Pseudoazurin (ox) + NO

Fig. 41. Schematic representation of the hexapeptide-pseudoazurin-NiR system.

104

Where “ox” and “red” denote the oxidised and reduced states respectively. The

homogeneous solution reaction is nitrite reductase (NiR) and can be deconvoluted into

two constituent steps:

Pseudoazurin (red) + NiR (ox) → Ak Pseudoazurin (ox) + NiR (red)

NiR (red) + NO2- → catk NiR (ox) + NO

From varying the experimental conditions, we isolate kA and kcat.

The enzyme kinetics were compared at different buffer concentrations. Salt

concentration doesn’t influence the direct electrochemistry of pseudoazurin (ks).

However a significant difference of kinetics was observed in the overall nitrite

turnover reaction at varying salt concentrations. When rising the buffer concentration

from 20 to 200 mM, a decrease in kcat and an increase in KM are observed.

By working under conditions of excess nitrite (15 mM), voltammetric investigations

can lead to a determination of the homogeneous electron transfer rate constant

between pseudoazurin (0.5 mM) and nitrite reductase (200 µM). By comparing the

diffusive (Id) and steady-state catalytic (Ip) pseudoazurin current responses (at the

same scan rate) an association rate constant can be determined from the following

expression [36].

Ip/Id = 0.359 kf1/2 CNiR

1/2 / (n1/2 v1/2)

Fig. 42. Cyclic voltammogram of pseudoazurin E51C mutant covalently attached to 4 mm diameter gold electrode in the absence (grey) and presence (black) of solution phase nitrite reductase (20 mV s-1).

105

where CNiR denotes the enzyme concentration and kf the diffusive (heterogeneous)

electron transfer rate

constant (all other symbols have their usual meaning). In excess of nitrite, a plot of

Ip/Id vs. 1/ν1/2 can lead to an extraction of kf (166.5 s-1) from the slope and,

subsequently, to a determination of the electron transfer rate constant between

pseudoazurin and nitrite reductase, kA (3.3 x 105 M-1 s-1 in 200 mM buffer solution

and 1.6 x 105 M-1 s-1 in 20 mM buffer solution). These values are consistent with

those reported previously from spectroscopic and voltammetric analyses [36, 37]. At

slow scan rates convective contributions to diffusive current become significant and

plots of Ip/Id vs 1/ ν1/2 become non-linear.

Under conditions of excess pseudoazurin (typically 1.25 mM to 650 µM enzyme), a

Michaelis analysis of enzyme activity, and a subsequent determination of rate of

substrate turnover, kcat, can be performed. Curve fitting the current-substrate

correlation to the Michaelis-Menten equation leads to an extraction of the following

values: kcat = 29.6 ± 2 s-1 and KM = (1.38 ± 0.2) x 103 µM in 200 mM buffer solution

and kcat = 168 ± 7 s-1 and KM = 146 ± 10 µM in 20 mM buffer solution.

Fig. 43. Michaelis-Menten representation of a substrate titration experiment; pseudoazurin 1.25 mM, nitritereductase 900 µM. The nitrite concentration was varied from 0 to 5 mM. The line shows the best fit of theMichaelis-Menten equation to the data

106

The enzyme kinetics were compared at different buffer concentrations. Salt

concentration doesn’t influence the direct electrochemistry of pseudoazurin. However

a significant difference of kinetics was observed in the overall nitrite turnover reaction

at varying salt concentrations. When rising the buffer concentration from 20 to 200

mM, a decrease in kcat and an increase in KM are observed.

On the addition of nitrite substrate to the solution, the electron transfer between the

blue copper centre of pseudoazurin and that of the enzyme (ultimately to the non-blue

copper active site) becomes catalytic. The cathodic current arising from pseudoazurin

exchange with the modified gold surface similarly becomes catalytic and linearly

proportional to the rate of substrate redox turnover (in the limit of nonsaturating

substrate levels) (Figures 44 and 45). These observations are qualitatively consistent

with those made by NMR (In the 0.3-1 mM concentration range NiR and PsAzu

associate almost 100% and the association is broken by 50 mM salt [38] .

Mediation of enzyme turnover with ruthenium hexamine

Though the mediated electrochemistry of nitrite reductase from Parococcus

denitrificans with organic mediators has been reported [39], the utilisation of

inorganic species has not been. Here we have demonstrated that ruthenium hexamine

acts as an effective electron transfer conduit with nitrite reductase, yielding

reproducible Michaelis-Menten kinetics. The kinetic plot can be fitted to the

Michaelis-Menten equation over the range 0 to 100 µM, this despite the fact that an

interaction takes place between nitrite and ruthenium hexamine at high nitrite

concentration. This interference prevents an accurate calculation of the electron

transfer rate constant between nitrite reductase and ruthenium hexamine as this

experiment requires a kinetically limiting concentration of mediator and a large excess

of nitrite. These conditions lead to large disruptions of the ruthenium hexamine

electrochemistry. As a result, only the apparent kcat and kM were calculated (Figure

43). Catalytic turnover of ruthenium hexamine by nitrite reductase in presence of

nitrite is observed on glassy carbon. The apparent kM 474 s-1 and kcat 438 s-1 values for

the ruthenium mediated system could be calculated. At nitrite concentrations above

500 µM, interference with ruthenium hexamine is observed, leading to a plateau in the

catalytic current.

107

Conclusion

Pseudoazurin surface cysteine mutants can be assembled on gold electrode surfaces in

the formation of electrochemically-addressable arrays. These arrays are responsive to

Fig. 44. The linear nitrite calibration regime (pseudoazurin mediation).

Fig 45. Calibration of nitrite by nitrite reductase 400 µM mediated by ruthenium hexamine (III)chloride 10 mM under argon. The current intensities were measured from the reduction peakintensity of ruthenium hexamine cycled from 0.1 to –0.3 V vs SCE at 20 mV s-1. The workingelectrode is a 9 mm diameter glassy carbon electrode, the counter electrode was a platinum mesh.

108

the presence of nitrite reductase and its substrate in solution but electrochemical

coupling is rather weak.

The direct electrochemistry of the enzyme nitrite reductase mutant L93C bound

directly to the gold electrode surface is reported herein. Though electrochemical

signals, responsive to substrate can be obtained, these observations are made only

with difficulty and substrate detection limits are not low. On a hexapeptide modified

gold electrode the diffusional redox signal of wild-type form of pseudoazurin can be

reliably obtained and effectively coupled to substrate turnover by the enzyme in

solution (this configuration effectively constitutes a natural partner electron transfer

relay). Under such circumstances, a calibration of nitrite in the 100 µM range can be

obtained (a level relevant to both food and water treatment industries). The electron

transfer rate constants can be calculated from the electrode to pseudoazurin,

pseudoazurin to nitrite reductase, and an apparent kM and kcat can be calculated.

The enzyme is also effectively “coupled” to an electrode through the use of an

inorganic mediator; under these conditions calibrated nitrite sensing down to 1 µM

levels is achievable. We additionally highlight the influence of the salt concentration

on the electron transfer kinetics between pseudoazurin and nitrite reductase.

References

1. M. Milani, L. Andolfi, S. Cannistraro, M. Ph. Verbeet, M. Bolognesi, Acta

Cryst. D57 (2001) 1735.

2. A.R. Bizzarri, S. Cannistraro, Chem.Phys.Lett. 349 (2001) 503.

3. P. Facci, D. Alliata, S. Cannistraro, Ultramicroscopy 89 (2001) 291.

4. T. Cimei, A.R. Bizzarri, S. Cannistraro, G. Cerullo, S. De Silvestri,

Chem.Phys.Lett. 362 (2002) 497.

5. L. Andolfi, S. Cannistraro, G.W. Canters, P. Facci, A.G. Ficca, I.M.C. Van

Amsterdam and M.Ph. Verbeet, Arch. Biochem. Biophys. 399 (2002) 81.

6. A.R. Bizzarri, S. Cannistraro, Appl. Spectrosc. 56 (2002) 1531.

7. L.Andolfi, B.Bonanni, G.W.Canters, M.Ph. Verbeet, S. Cannistraro, Surface

Science 530 (2003) 181.

8. A.R. Bizzarri, B. Bonanni, G. Costantini, S. Cannistraro, Chem.Phys.Chem. 4

(2003) 1189-1195.

109

9. A.R. Bizzarri, G. Costantini, S. Cannistraro, Biophys.Chem. 4 (2003) 1189-

1195.

10. T. Cimei, A.R. Bizzarri, G. Cerullo, S. De Silvestri, S. Cannistraro, Biophys.

Chem. 106 (2003) 221-231.

11. L. Andolfi, D. Bruce, S. Cannistraro, G.W. Canters, J.J. Davis, H.A.O. Hill, J.

Crozier, M. Ph. Verbeet, C.W. Wrathmell, Y. Astier, J. Electroanal. Chem. 565

(2004) 21-28.

12. L. Andolfi, G.W. Canters, M.Ph. Verbeet, S. Cannistraro, Biophys. Chem. 107

(2004) 107-116.

13. D. Alliata, L. Andolfi, S. Cannistraro, Ultramicroscopy (2004) submitted.

14. L. Andolfi, D.Alliata, S. Cannistraro, Chem. Phys. Chem. (2004) submitted.

15. B. Bonanni, D. Alliata, L. Andolfi, A. R. Bizzarri, S. Cannistraro, Progress in

Surface Science Research (2004) in press.

16. J.J. Davis, Phil. Trans. Royal Soc. A, 2003, 361 (1813) 2807-2825 (Invited Article).

17. J. Zhao, J.J. Davis, Nanotechnology, 2003, 14, 9, 1023-1028.

18. J.J. Davis, IEE Nanobiotechnology 2004, in press.

19. J. Zhao, J.J. Davis, Colloids and Surfaces, (2004) in press.

20. C.L. Wrathmell, J. Zhao, J. Fletcher, D. Djuricic, J.J. Davis, H.A.O. Hill, J.

Molecular Recognition, 2004, 17, 1-7.

21. Jianwei Zhao, Jason J. Davis, Mark S. P. Sansom, Andrew Hung, J. Am. Chem.

Soc. 2004, 126, 5601-5609.

22. Wold, D.J. and C.D. Frisbie, J. Am. Chem. Soc. 2001. 123 (23): 5549-5556.

23. Zhao, J.W. and K. Uosaki, Applied Physics Letters, 2003. 83 (10): 2034-2036.

24. Singh-Zocchi, M., J. Hanne, and G. Zocchi, Biophysical Journal, 2002. 83 (4):

2211-2218.

25. Cui, X.D., et al., Journal of Physical Chemistry B, 2002. 106 (34): 8609-8614.

26. Xiao, X.Y., B.Q. Xu, and N.J. Tao, Nano Letters, 2004. 4 (2): 267-271.

27. R. Perez-Olmos, Anal. Science, 1998. 14 (5): 1001-1003.

28. D.J.D. Nicholas, A. Nason, Methods in Enzymology, 1957. 3: 982.

29. L.J.C. Jeuken, J.P. McEvoy, F.A. Armstrong, J. Phys. Chem, 2002. Sect. B 106

(9): 2304-2313.

30. J. Kazlauskaite, et al., Eur. J. Biochem., 1996. 241: 552-556.

31. N. Patel, et al., Langmuir, 1997. 13 (24): 6485-6490.

32. L.J.C. Jeuken, et al., J. Am. Chem. Soc, 2002. 124 (20): 5702-5713.

110

33. M.E. Murphy, S. Turley, E.T. Adman, J. Biol. Chem, 1997. 272: 28455.

34. R. Nicholson, I. Shain, Anal. Chem., 1964. 36 (4): 706-723.

35. T. Kohzuma, et al., J. Biol. Chem.1995. 270: 25733 - 25738.

36. K. Kataoka, et al., J. Inorg. Biochem., 2000. 82: 79-84.

37. T. Kohzuma, et al., Chem. Lett., 1993. 12: 2029-2032.

38. G.W. Canters, unpublished observations.

39. B. Strehlitz, et al., Anal. Chem. 1996. 68: 807-816.

40. L.M. Murphy, et al., J. Mol. Biol. 2002. 315 (4): 859-871.

41. M. Kukimoto, Protein Eng, 1995. 8 (2): 153-158.

110

3.4 WP4

3.4.1 Completion of M13: modeling the coupling between molecule (copper atom) and the metallic nanogate.

The systems of interest for this project are very complex, and involve the

interaction among several objects with different properties, such as proteins, solid

substrates, metallic nanoelectrodes. To cope with such a complexity, we adopted in

our theoretical investigation a basic principle that can be phrased as “multi-level

modeling”.

(i) The first level of modeling operates according to a global point of view and

consists in considering the whole system as composed of simplified objects.

Each individual component is characterized by the intrinsic properties that are

important for the phenomenon under study and that can be estimated to several

degrees of accuracy. The accuracy with which such properties are known affects

the computed quantities: however, several theoretical aspects of the problem can

be unraveled independently of it, and the method to compute such properties

does not need to be specified at this level of modeling. An example of this

approach is the Marcus theory for electron transfer (ET): it gives the desired

quantity (ET rate) in terms of microscopic quantities such as the reorganization

energy and the transfer integrals, whose calculation is not directly part of theory.

(ii) The deeper second level of modeling consists in using computational

methodologies to calculate the properties required in the first-level modeling.

For example, the reorganization energies and transfer integrals that enter the

Marcus theory can be computed by quantum-mechanical methodologies.

The specific problem of the electronic transport through a single molecule connected

to two nanoelectrodes in an ElectroChemical (EC)STM setup was already treated in

the scientific literature, yielding different proposals of first-level models (with the

meaning defined above). Briefly, the advanced interpretations may be summarized in

the three following categories. (a) Two-step sequential ET [1]: in the first step the

electron passes from one electrode, say the substrate, to the oxidized protein, which

has time to fully relax; in the second step the electron moves from the protein to the

other electrode, say the tip. (b) Vibrationally coherent ET [2]: after the first ET from

an electrode to the protein, the second ET (from the protein to the other electrode) is

fast enough that the protein does not have the time to fully relax. (c) Resonant

tunneling [3]: the two ETs from one electrode to the protein and from the protein to

111

the other electrode are so fast that the protein is not even able to start the relaxation

process, and its redox level is actually never populated (superexchange-like

mechanism).

All these models can be cast in terms of two intrinsic molecular properties: the

transfer integrals (from one nanoelectrode to the molecule and then from the molecule

to the second nanoelectrode), and the reorganization energy λ . Since the proposed

first-level models seem to cover a variety of possible transfer mechanisms from very

fast (resonant tunneling) to very slow (sequential two steps ET) processes, including

intermediate cases (vibrationally coherent ET), we focused our investigation on the

calculation of the molecular quantities that enter these different theories. By

comparing the computed quantities with those obtained by fitting the various theories

to available experimental results, one can determine to which extent the different

theories are supported by simulations in various target situations.

The role of transfer integrals is very similar in all the mentioned theories: they

basically control the maximum rate of transfer but do not discriminate between viable

relaxation mechanisms at the redox center. Thus, they are not very useful in trying to

distinguish between the various theories. Furthermore, their value is extremely

sensitive to parameters such as the electrode-active site distances, whose experimental

control is rather difficult, thus hindering a reliable comparison to measured data. The

reorganization energy λ seems to be more promising to the desired purpose, because

it affects (in different ways for different theories) the shape of the current-voltage

curve and in particular the position of the current maximum.

In the second year report, we described a method for the evaluation of the outer-

sphere contribution to the reorganization energy for a protein in the ECSTM setup,

that was at that time under optimization and testing. The basic feature of the model is

to combine an ab-initio description of the protein active site with a continuum

(dielectric-like) description of the other components in the system (remainder of the

protein, solvent, STM substrate and tip). Figure 1 (reproduced from Fig. 48 of the

second-year report) shows the various features of the model. The integration of the

ab-initio treatment of the active site and of the continuum model, indicated in the right

panel of Figure 1 as a future development, was implemented and applied in the course

of the third year. The reorganization energy in the framework of such a model is

calculated (following the original idea by Marcus for ET in solution) as a difference

112

when all the degrees of freedom of the medium are in equilibrium with their

electronic density; the other is the non-equilibrium Gibbs free energy of the ET

products when the fast degrees of freedom are in equilibrium with their electronic

density and the slow ones are still equilibrated with the ET reactants. Actually, the

method that we implemented is able to compute the non-equilibrium free-energy not

only for points of the reaction coordinate corresponding to the reactants and the

products, but for a generic point between them. Thus, the adiabatic non-equilibrium

free-energy surfaces can be obtained within our method. The parameters that

introduce in our model the effects of the fast and the total (fast+slow) degrees of

freedom are the optical ( optε ) and the static ( 0ε ) dielectric constants, respectively, of

the various media in the system. For the metal components, we assumed a perfect

conductor behavior for both the static and the optical response. For the water solvent,

the dielectric constants are well-known: we used optε =1.776 and 0ε =78.39, suitable

for T=298 K. The choice of the dielectric constants for the protein is less clear: optε is

determined, as for homogeneous solvents, by the electronic polarization only and thus

we assumed 2optε = , a typical value for solvents; for 0ε , we used a non-local

dielectric constant, i.e. one that distinguishes between short range and long range

interactions through the use of a correlation length Λ [5]. The value of 0ε for short

Tip

∗ Hemisphere planar surface

∗ Perfect conductor

Solvent (light blue area)

∗ Homogeneous dielectric (εstat, εdyn)

Active site (red stick-scheleton)

∗ Set of atomic charges from ab-initio calculations

∗ Future developments: integrating ab-initio treatment and continuum

model

Protein

∗ Homogeneous dielectric (εprot)

Substrate

∗ Planar surface

∗ Perfect conductor

Pperfect conductor

Figure 1. Sketch (left) and description (right) of the framework adopted in thecomputation of the reorganization energy. The drawing is out of scale. Thedescription of the solvent follows the polarizable continuum model (PCM) [4] andthe protein medium surrounding the active site is treated on the same footing.

113

range interactions was chosen equal to optε (they are both determined by the same

electronic polarization), while for long range interactions we chose a value suitable

for a moderately polar environment, 20. The correlation length Λ was assumed equal

to 5 Å. These choices, based on physical common sense, give free-energies in the

same range of those obtained with local constant values of 3-5, that are often used for

proteins. Nevertheless, since we do not have a direct source (experimental or

theoretical) to check the chosen 0ε , before studying the reorganization energy in the

specific ECSTM setup adopted by the experimental INFM group in Modena (P. Facci

and coworkers), we compared λ computed with our model for the azurin electron

self-exchange (ESE) reaction in solution:

Az(II) + Az(I) → Az(I) + Az(II)

with the experimental estimate by Gray et al. [6], who found 0 6 0 8λ = . − . eV. Since

our model described above gives only the outer contribution outλ to λ (due to the

protein outside the active site and the solvent), to compare with experimental results

we also took into account an independent estimate of the inner term inλ . Olsson and

Ryde [7] estimated inλ by means of a QM/MM simulation of azurin, and we used

their value inλ =0.291 eV. The computed value for the total λ is 0.65 eV, in good

agreement with the experiment. Notably, this value was obtained by considering the

two azurin molecules linked by their hydrophobic patch (see Fig.2).

Figure 2. Encounter complex geometry (taken to be equal to the dimer geometry in the crystal)

used in the calculation of λ for the ESE reaction of Az. The dielectric modeling the protein is

represented in blue, the cavities that host the active sites are red.

114

We found that λ is very sensitive to the relative positions of the two proteins in

the encounter complex: in fact, when the two active sites are close enough (as in the

ESE encounter complex), the oxidized site can interact with the medium polarization

which is due to the other (reduced) site, and vice-versa. Thus, the ET reactants feel

before the ET a polarization which is already partly adapted to the ET products. In

other words, the presence of this medium polarization, that exists prior to charge

transfer, quenches the extent of the required structural reorganization, thus decreasing

the reorganization energy. The importance of the spatial proximity between active

sites to have effective protein-protein ET was often related with the need for high

electronic coupling: however, our findings show that the relative spatial arrangement

not only affects the electronic coupling (transfer integral), but is also fundamental for

diminishing the ET reorganization energy and consequently incrementing the transfer

rate.

After this preliminary calculation that allowed us to check the quality of the protein

permittivity parameters in the ESE calculations, we applied our method to the

ECSTM system. Here the ET reaction to be considered is

Az(II) + Electrode → Az(I) + Electrode

where Electrode can be the tip or the substrate. Note that, differently from the ESE,

only one Az molecule takes part in the ET reaction. Thus, one would expect a value of

λ equal to 0.65/2 eV = 0.33 eV if the variation of environmental conditions had a

small/negligible role. By considering an azurin molecule standing with the

Cys3/Cys26 disulfide bridge on the substrate (with the protein major axis

perpendicular to the substrate surface) and the tip (radius=5 nm) standing 5 Å from

the protein surface, we instead obtained a larger value of 0.43 eV. This result depends

on two opposing effects: (i) on one hand, the presence of the ECSTM apparatus

removes water from the protein neighborhood, decreasing λ ; (ii) on the other hand,

the effectiveness of the spatial proximity between the two active centers of the

encounter complex in decreasing λ , is canceled in the ECSTM setup. In the present

case, the latter effect prevails on the former.

We tried to give a lower-bound estimate to λ for Az in the ECSTM setup. To this

end, we considered the protein lying with its major axis parallel to the substrate, with

115

the tip touching the molecular surface (tip-substrate distance: 29 Å). Moreover, we

also took into account the confinement effect on the dielectric constant of water. A

recent work [8] suggests that for water confined between two layers ≈ 30 Å apart, the

static dielectric constant is around 10-20. We used the most conservative value of this

range, 10. With these assumptions, we obtained λ =0.35 eV. The INFM-Modena

experimental group performed several ECSTM experiments on Az. In one of these,

whose results were recently published [9], where the protein was adsorbed on the

substrate, they fitted their data to the resonant tunneling theory (obtaining λ =0.13

eV) and to the vibrationally coherent model ( λ =0.53 eV). Our calculated value

suggests a validation of the latter model rather than the former, which gives instead a

too small λ value. However, in a more recent experiment (documented in the present

report), where the molecule was adsorbed on the tip and different bias potentials were

applied, the fit to the vibrationally coherent model gave a smaller value of λ =0.20

eV. Although the geometry of the experimental system is somewhat different from

that assumed in our calculation (protein turned upside-down), our lowest bound

estimate (0.35 eV) should still be valid. There are different reasons that may explain

the discrepancy between the fitted value and our estimates: for example in the

experiment, the tip may actually squeeze the molecule, thus varying both its structure

(modifying the inner reorganization energy inλ ) and its dielectric properties

(important for outλ ). In addition, also the electric field in the gap between the tip and

the substrate may modify the structure of the protein and the dielectric properties of

both the protein and the solvent (electrostriction effects). Finally, we should also take

into account that the fit of the experimental data to the theory was performed by

considering only one feature of the current-voltage curve, i.e. the position of the

current maximum. More accurate fits, that will account for the overall shape of the

current-voltage curve, are in progress. Table I reports a summary of the values

computed for the reorganization energy in different simulation setups.

inλ (eV) outλ (eV) λ (eV)

ESE 0.29 0.36 0.65

ECSTM 0.15 0.29 0.43

ECSTM2 0.15 0.20 0.35

Table 1. Values of the computed reorganization energies for Az in the ESE reaction and inthe specific ECSTM geometry. “ECSTM2” takes into account the confinement effects on thewater dielectric constant and assumes that Az is forced to be lying downon the substrate bythe tip. inλ ( outλ ) is the inner (outer) contribution.

116

The proposed model can be also used to explore the dependence of λ on the detailed

geometry of the ECSTM setup. In particular, we studied the behavior of the outer

contribution to λ as a function of the vertical and horizontal tip-protein distances,

allowing for lateral tip scans. The results (for outλ only, we assume that inλ remains

the same) are reported in Fig 3.

Figure 3. Dependence of the outer contribution outλ on the tip position. The vertical distance

refers to the distance between the tip and the protein surface; the horizontal distance refers to

the projection of the tip-Cu distance (when the tip is at a lateral position with respect to the

vertical axis) on a plane parallel to the substrate.

As one can expect, the dependence of λ on the horizontal tip-protein distance is

less pronounced than the dependence on the vertical distance. In particular, the shape

of the curve for the horizontal distance shows variations of λ on the length-scale of

the tip size (tip radius: 5 nm). The variation of λ with respect to the vertical distance

is steeper, e.g. it is characterized by a smaller length scale. However, it is not dramatic

if one consider that the tip-protein distance cannot be too large (say not larger than 10

Å) in order to have a measurable current: by changing the tip-protein distance from 0

(i.e., tip-protein contact) to 10 Å the total reorganization energy changes from 0.41 eV

to 0.46, which is well within the experimental errors on this quantity.

In conclusion, we elaborated a model to compute the reorganization energy of ET

processes involving proteins, in the presence of nanoelectrodes such as those of the

ECSTM setup. This method gives results in agreement with the available

experimental data for azurin ESE. The values of the reorganization energy in the

ECSTM setup are larger than what expected simply from stoichiometric

consideration, due to physical reasons that we have discussed. By considering more

severe measurement conditions, we estimated a lowest bound of 0.35 eV for the

117

reorganization energy of azurin in the ECSTM setup. The model has been also used to

explore the dependence of λ on the detailed geometry of the system. The method can

also be extended to different experimental setups, including the transistor geometry:

preliminary simulations of a model transistor show its reliability and are promising to

foster further combined theoretical-experimental investigations. Indeed, the novel

methodological implementation and the establishment of such a combined research

activity among different groups was an important added value of this project and

represents its fall-out into the near future.

Finally, we wish to mention another relevant theoretical simulation that is currently

ongoing and that was possible thanks to the methodological achievements and

computational successes described above. For one of the dimerized structures

proposed by the Leiden chemical synthesis group [10], in which two mutated azurins

are connected to each other by means of two water molecules around the Cu centers

and through a chemical linker far from them, we are computing the reorganization

energy and the transfer integrals. The latters are obtained through an ab-initio method

recently implemented in our group [11]. The reorganization energy and the transfer

integrals will then be combined for an evaluation of the experimental observable

which is the transfer rate. The advantage of employing for this study the atomic model

proposed by Canters and coworkers rather than the encounter complex of Figure 2 is

that the presence of a chemical linker between the two molecules of the dimer allows

for kinetic experiments to measure the transfer rate: therefore, we will have a direct

comparison to consistent measured data. This is another example of activity that was

stimulated in the course of this project and was made possible by the

accomplishments of the deliverables and milestones, by both the experimental and the

theoretical groups.

3.4.2 Milestone M14 modeling the FET transport and biosensor operation.

The model presented in the previous section is a way to face the problem of

transport in a single protein transistor, as an ECSTM apparatus can be considered,

therefore addressing the final milestone and deliverable. However, we also put an

effort in modeling the working protein transistor presented in the second year report

by INFM-Lecce, based on an azurin layer. This work is an active collaboration

between INFM-Modena and INFM-Lecce.

118

To model such a complex system (a layer of proteins deposited in the empty space

between two nanoelectrodes, acting as the source and the drain, and feeling the

electric field produced by a third electrode, the gate) we adopted the same basic

principle described above, i.e. we use a first level model in which some parameters

remain undetermined and then we specify such parameters by other techniques

applied to the single components of the system. Here we first list the assumptions in

our model for the operation of the protein/based transistor, and then present the results

of an application.

Assumptions of the model.

1. The conduction between the drain and the source occurs through a sequential

hopping mechanism: from one electrode the electron is transferred to a close

oxidized protein (which thus becomes reduced). From this reduced molecule

the electron moves to a neighbor oxidized molecule and so on, until the

electron is finally transferred to the other electrode. Note that we are implicitly

assuming that the proteins have an available redox level, which is true for

copper-Az and is not true for zinc-Az or apo-Az.

2. We divide the proteins of the layer in three groups: those in contact with the

source, that can directly exchange electrons with it (group s ), those in contact

with (and exchanging electrons with) the drain (group d ) and all the

remaining proteins in the layer (group l ).

3. When no bias voltage is applied to the device, the redox energy levels of all

the proteins are aligned to a given value, called 0ε . When potential differences

such as dsV (draine-source potential difference) or gV (gate-source potential

difference) are applied to the device, the energy of the redox level of the

protein i becomes 0i ieVε ε= − , where iV is the electrostatic potential at the

position of the protein i and e is the electron absolute charge.

4. The electron-transfer rate ijk from the protein j to the protein i (assumed to

be nearest-neighbor) is given by the Marcus expression:

2( )

4i j

ijB

eV eVk exp

k Tλ

κλ

− −= −

(1)

119

where λ is the reorganization energy of the ET process, Bk is the Boltzmann

constant and T is the absolute temperature (in the numerical calculation,

T = 300 K). The maximum rate constant κ was chosen to give the same rate

constant as two azurin self-exchanging an electron in solution (max. estimate:

≈106 s-1). As for the reorganization energy λ , we calculated it by using the

model described in the previous section applied to the transistor geometry (i.e.,

two proteins adsorbed on an insulating substrate), obtaining λ = 0.63 eV (the

dielectric constants used for SiO2 were optε =2.13 and 0ε =3.9).

5. The master equation for the evolution of the probability ip that the protein i

belonging to the group l is in the reduced form (also called occupation

number) is assumed to be:

(1 ) (1 )iij j i ji i j

j i j i

dp k p p k p pdt ≠ ≠

= − − −∑ ∑ (2)

(1 )ij j ik p p− is the rate of the ET from j to i ( ijk ) weighed by the probability

that the donor protein j is reduced ( jp ) and the acceptor protein i is oxidized

(1 ip− ). Similarly, (1 )ji i jk p p− is the rate of the ET from i to j ( jik )

weighed by the probability that the donor protein i is reduced ( ip ) and the

acceptor protein j is oxidized (1 jp− )

6. We assume that the ET between the source or the drain and the proteins in

group s or d , respectively, is much faster than the ET between two proteins.

This means that the occupation number of proteins si or di in groups s or d

is always that proper for equilibrium with the electrodes, i.e. Nernst law holds:

0

0

[ ( ) ]1 [ ( ) ]s

FS i Bi

FS i B

exp eV k Tpexp eV k T

ε εε ε

− − − /=

+ − − − / (3.1)

0

0

[ ( ( ) ]1 [ ( ( )) ]d

FS ds i Bi

FS ds i B

exp e V V k Tpexp e V V k T

ε εε ε

− − + − /=

+ − − + − / (3.2)

where dsV is the drain-source potential difference and FSε is the Fermi energy

of the source (we remark that all the potentials in the device are referred to the

source, which is always grounded).

Within these assumptions, the expression for the current is given by:

120

s

s

i

i

dpI e

dt= ∑ (4)

where si runs over all the proteins in direct contact with the source (a similar

expression can be found for the drain). The sign of the current is chosen to be positive

for electrons going from the electrode to the proteins.

The set of eqs. (2) has been solved numerically, by implementing a tailored

FORTRAN90 code. In addition to the protein-protein ET rate parameters κ and λ

whose choice we discussed above, the other parameters that have to be specified are:

(a) the positions of the proteins in the device; (b) the electrostatic potential at the

position of each protein as a function of dsV and gV ; and (c) the relative redox-level

energy of the proteins 0ε with respect to FSε (i.e., the difference 0 FSε ε− that appears

in eq.(3)).

(a) As for the protein positions, we considered a square lattice with lattice vectors

parallel and perpendicular to the source-drain direction. This assumption allows us to

study easily the dependence of the current on the size of the layer. The position could

be found as well with a simulation of adsorption of the proteins on the layer from a

solution. However, this would imply to average the current results for a given device

on a large number of different layer obtained by different simulations, which are

numerically very expensive. The simple square lattice can give results appropriate for

the level of description we are aiming to with a much smaller computational demand.

(b) The distribution of the potential in the device was found from a FEM simulation,

whose details are given in the appendix A.

(c) We are thus left with just one parameter, i.e., the energy level difference 0 FSε ε− .

This value could in principle be calculated by the work functions W of the metal

composing the electrodes and the standard reduction potential 0E of azurin. Using 5.3

eV for AuW [12], 0.3 V vs NHE (Normal Hydrogen Electrode) for 0E , and 4.6 eV for

the work function of the NHE, we estimate a value of 0 0 4FSε ε− = . eV. However,

such a value should not be taken too seriously, since the system is much more

complex than the simple assumptions at the basis of these data would imply (such as

protein in bulk solution and not dried and supported on an electrode, ultra-vacuum

121

like Au surfaces). Thus, we shall use 0 FSε ε− as an adjustable parameter to be chosen

around 0 eV. This is basically the only adjustable parameter of our model.

Let us now summarize the main

results obtained from the model. We

shall compare them, whenever

possible, with the experimentally

observed behavior for the transistor

fabricated in Lecce. The data

reported here refer to a device with

the source-drain geometry reported

in Fig. 4.

• The behavior of the steady-state current I as a function of gV for dsV = 5 V is

plotted in Fig. 5 (left) for a 8-chain device (larger protein layers have larger

current magnitude, but similar behaviors). It is clear that the model is able to

reproduce the resonance-like behavior of the current, even if the shape of the

calculated curve is somewhat different from the experimental one (more

asymmetric and wider), see the right panel in Figure 5. The value of 0 FSε ε−

was adjusted to give the maximum current at the same value as the experimental

one, i.e., 1gV ≈ V. We found 0 0 45FSε ε− = . eV, which is well inside the

expected range of possible values.

• The behavior of I as a function of dsV for a fixed gV is also well reproduced by

the model: the computed curve is shown here in Fig.6 and should be compared

to Fig. 9a of the second year report. A three-dimensional plot of I as a function

of dsV and gV is reported in Fig. 7. It is evident that the position of the

maximum of ( )gI V depends on dsV . This behavior could be present also in the

experiment, but it is not easy to verify this, since in the experiment the current

for small dsV is hidden by the background, while for large dsV the device breaks

up, leaving a quite small comparable window ( 4 6dsV ≈ − V).

Figure 4. Scheme of the protein device under study.

122

• The conductivity of the device, at a given value of dsV and gV , is a function of

the size of the protein layer. In particular, it depends on the number N of

linear chains of assembled proteins connecting the source and the drain

following the relation:

( ) (1 )totI N a log bN= + (5)

which is what expected for a series of ohmic wires with increasing length with

a more or less constant resistivity. Relation (5) has been numerically verified

for N as large as 50. This means that the total conductivity does not only

Figure 5. Left: current I as a function of gV for 5dsV = V and a device with 8 chains.

0 0 45FSε ε− = . eV. Right: experimental current-voltage characteristic for the fabricated

transistor, reproduced from Figure 11 of the second-year report.

Figure 6. Current I as a function of dsV for

1gV = V and a device with 8 chains.

0 0 45FSε ε− = . eV.

Figure 7. Current I as a function of gV and dsV

for a device with 8 chains. 0 0 45FSε ε− = . eV.

123

depend on the few proteins at the center of the source-drain gap, but on a large

portion of the device.

The model also allows for the study of quantities that are difficult (or even

impossible) to probe through the experiments. For example one can study the average

population p< > (or the spatial population distribution) of the proteins in the device

as a function of gV and dsV . In Fig. 7, p< > is reported as a function of gV and dsV

for 0 0 45FSε ε− = . eV.

Fig. 7. Mean occupation number p< > as a function of gV and dsV for a device with 8

chains. 0 0 45FSε ε− = . eV

The relation between the position of the current maximum and the change of the

average protein population is evident: the maximum of the current falls within the

narrow potential range in which the proteins change from the oxidized ( 0p< >≈ ) to

the reduced ( 1p< >≈ ) form, i.e. when 0 5p< >≈ . . In other terms, the maximum of

the current corresponds to the maximum probability of having one oxidized and one

reduced molecule close to each other, as proposed in discussion of the device put

forward in the second year report.

124

In conclusion, we elaborated a model based on a sequential conduction

mechanism, which is able to reproduce the main features of the experimentally

determined conductivity and trans-conductivity. Following this model, the resonance-

like behavior of the current correspond to the maximum probability of having two

nearest-neighbor proteins in different oxidation states. So far, the model gives a

satisfactory qualitative agreement of the resonance behavior of the IV curves in the

FET, but fails in the quantitative evaluation of the current. A more refined method,

which refines the protein-lead contacts and takes into account the presence of several

layers parallel to the substrate, is under evaluation and will be tested in the future.

References.

1. A. M. Kuznetsov, P. Sommer-Larsen, and J. Ulstrup, Surf. Sci. 275, 52 (1992). 2. E. P. Friis, Y. I. Kharkats, A. M. Kutznetsov, and J. Ulstrup, J. Phys. Chem. A

102, 7851 (1998). 3. W. Schmickler, Surf. Sci. 295, 43 (1993). 4. J. Tomasi, R. Cammi, B. Mennucci, C. Cappelli, and S. Corni, Phys. Chem.

Chem. Phys. 4, 5697 (2002). 5. A.A. Kornyshev, in The Chemical Physics of Solvation; R. Dogonadze, E.

Kámál, A.A. Kornyshev, J. Ulstrup, Eds.; Elsevier,: Amsterdam, 1985, Part A, p. 85, eq. 3.13.

6. (a) A. J. Di Bilio, M. G. Hill, N. Bonander, B. G. Karlsson, R. M. Villahermosa, B. G. Malmström, J. R. Winkler, and H.B. Gray, J. Am. Chem. Soc. 119, 9921 (1997); (b) H. B. Gray, B. G. Malmström, and R. J. P. Williams, J. Biol. Inorg. Chem. 5, 551 (2000).

7. U. Ryde, and M. H. M. Olsson, Int. J. Quant. Chem. 81, 335 (2001). 8. O. Teschke, G. Ceotto, and E. F. de Souza, Phys. Chem. Chem. Phys. 3, 2761

(2001). 9. A. Alessandrini, M. Gerunda, G. W. Canters, M. Ph. Verbeet, and P. Facci,

Chem. Phys. Lett. 376, 625 (2003). 10. I.M.C. van Amsterdam, M. Ubbink, O. Einsle, A. Messerschmidt, A. Merli, D.

Cavazzini, G.L. Rossi, and G.W. Canters, Nature Struct. Bio. 9, 48 (2002). 11. (a) A. Ferretti, A. Ruini, E. Molinari, and M. J. Caldas, Phys. Rev. Lett. 90,

086401 (2003); (b) A. Ferretti, A. Ruini, G. Bussi, E. Molinari, and M. Caldas, Phys. Rev. B in press, (2004).

12. H. B. Michaelson, J. Appl. Phys. 48, 4729 (1977).