Dielectric/metal/dielectric alternative transparent electrode: observations on stability/degradation

(1)

This content has been downloaded from IOPscience. Please scroll down to see the full text.

Download details:

IP Address: 41.229.35.251

This content was downloaded on 05/09/2017 at 11:12 Please note that terms and conditions apply.

Dielectric/metal/dielectric alternative transparent electrode: observations on stability/degradation

View the table of contents for this issue, or go to the journal homepage for more 2017 J. Phys. D: Appl. Phys. 50 375502

(http://iopscience.iop.org/0022-3727/50/37/375502)

Home Search Collections Journals About Contact us My IOPscience

Dielectric/metal/dielectric alternative transparent electrode: observations on stability/degradation

L Cattin¹, El Jouad²^,⁴, N Stephant¹, G Louarn¹, M Morsli³, M Hssein¹^,⁴, Y Mouchaal²^,⁵, S Thouiri²^,⁶, M Addou⁴, A Khelil⁵ and J C Bernède²

1 Institut des Matériaux Jean Rouxel (IMN), Université de Nantes, CNRS, UMR 6502, 2 rue de la Houssinière, BP 32229, 44322 Nantes cedex 3, France

2 MOLTECH-Anjou, CNRS, UMR 6200, Université de Nantes, 2 rue de la Houssinière, BP 92208, Nantes, F-44000, France

3 Faculté des Sciences et des Techniques, 2 rue de la Houssinière, BP 32229, 44322 Nantes cedex 3, France

4 LMVR, FST, Université Abdelmalek Essaidi, de Tanger, Ancienne Route de l’Aéroport, Km 10, Ziaten, BP: 416, Morocco

5 Université d’Oran 1—Ahmed Ben Bella, LPCMME, BP 1524 ELM Naouer 31000 Oran, Algeria

6 Unité de Physique des Dispositifs à Semi-conducteurs, Université El Manar Faculté des Sciences de Tunis, Campus Universitaire 2092, Tunisia

E-mail: [email protected] Received 15 June 2017

Accepted for publication 6 July 2017 Published 24 August 2017

Abstract

The use of indium-free transparent conductive electrodes is of great interest for organic optoelectronic devices. Among the possible replacements for ITO, dielectric/metal/dielectric (D/M/D) multilayer structures have already proven to be quite efficient. One issue with organic devices is their lifetime, which depends not only on the organic molecules used but also on the electrodes. Therefore we study the variation, with elapsed time, of the electrical and optical properties of different D/M/D structures, with M = Ag or Cu/Ag. Six years after realization, it has been shown that if some structures retained an acceptable conductivity, some others became non-conductive. For a sample which remains conductive, in the case of a PET/MoO3/Ag/MoO3

multilayer structure, the sheet resistance changes from 5 Ω/sq–17 Ω/sq after six years. This evolution can be compared to that of a PET/ITO electrode that varies from 25 Ω/sq–900 Ω/sq after six years. It means that not only are the PET/MoO3/Ag/MoO3 multilayer structures more flexible than PET/ITO, but they can also be more stable. Nevertheless, if some PET/MoO3/ Ag/MoO3 multilayer structures are quite stable, some others are not. This possible degradation appears to be caused primarily by the physical agglomeration of Ag, which can result in Ag film disruption. This Ag diffusion seems to be caused by humidity-induced degradation in these Ag-based D/M/D structures. Initially, defects begin to grow at a ‘nucleus’, usually a microscopic particle (or pinhole, etc), and then they spread radially outward to form a nearly circular pattern. For a critical density of such defects, the structure becomes non-conductive.

Moreover the effect of humidity promotes Ag electrochemical reactions that produce Ag⁺ ions and enhances surface diffusivity with AgCl formation.

Keywords: transparent electrode, ageing effect, dielectric–metal–dielectric structure, metal diffusion

S Supplementary material for this article is available online (Some figures may appear in colour only in the online journal)

L Cattin et al

Printed in the UK 375502

JPAPBE

J. Phys. D: Appl. Phys.

JPD

10.1088/1361-6463/aa7dfd

Paper

37

Journal of Physics D: Applied Physics IOP

2017

1361-6463

https://doi.org/10.1088/1361-6463/aa7dfd J. Phys. D: Appl. Phys. 50 (2017) 375502 (13pp)

(3)

1. Introduction

The need for transparent conductive electrodes (TCE) is increasing rapidly due to applications in flexible optoelectronic devices. Indium tin oxide (ITO) is the most common transparent conductive material used in industrial processes.

It has already been demonstrated that large area ITO layers can be obtained by rf magnetron sputtering after annealing at temperatures above 300 °C [1]. If these processes—rf magnetron sputtering deposition and annealing at T ⩾ 250

°C—enable high performing and stable transparent conductive ITO electrodes to be obtained, they are not compatible with flexible plastic substrates such as PET with a glass transition temperature of less than 100 °C [2]. Moreover, in 1999, Minami warned of the possible shortage of indium which may occur in the future [3]. This warning has been repeated several times since (see, for instance, [4]). Even if this point of view is controversial, the scarcity of indium and the fact that it is a by-product of zinc extraction make its price fluctuate significantly [5]. Moreover, as referred to above, the intrinsic flexibility of organic semiconductors means it is envisaged that the next-generation of organic optoelectronic devices will be fabricated on flexible substrates to realize flexible and lightweight optoelectronic devices. The high temperature deposition process and the brittle nature of ITO impede its applications in flexible modules [6].

Even if some recent improvement in the deposition process of ITO enabled ITO film to grow on a plastic substrate with acceptable electrical and optical properties, the flexibility is still lacking, which means that ITO is unsuitable for use in flexible devices because it is a ceramic, and conse- quently highly brittle [7]. Actually, in order to obtain highly conductive and transparent ITO films, it is necessary to use substrate heated at temperatures above 300 °C or/and to proceed to post-annealing treatment above 250 °C. These high temperature processes not only enable conductive and transparent layers to be obtained, but also enable highly stable film performance [8–10]. For instance, in the present work, commercial ITO deposited onto glass bought six years ago still exhibits its initial performance. It is not the case with ITO deposited onto PET. Since the transition temperature of PET is Tg < 100 °C, it is not possible to work with the temperatures used for glass substrates, which makes the stability of these ITO layers deposited onto PET far lower than that of films deposited onto glass substrates. Moreover, if the usual organic optoelectronic devices described in the literature employ ITO as a bottom electrode, its application to large area devices is limited by its insufficient conductivity, mainly in the case of flexible substrates [11]. For the first time, it was proposed that a thin Ag layer be inserted between two ITO layers so that one combines the electrical conductivity of ITO and Ag with the ability of ITO to boost the transmission of Ag due to its refractive index of around two. The resistivity of such multilayer structures was lowered by at least one order of magnitude, while the optical transmittance exceeded 80% in the visible [12]. Then, because the high conductivity of these structures is due to the metal film M, it was proposed that a dielectric D be substituted for ITO, and thus D/M/D structures have been

realized. These structures are easier to make because usually the dielectric chosen is easily deposited by sublimation under vacuum and no temperature treatment is required [13]. In such structures, in order to decrease the high light reflection of the metal film, it is embedded between two dielectrics. MoO_3−x is often used as an oxide due to its capacity to perform as an anode buffer layer in organic solar cells, previous experi- ence having demonstrated that organic device performance is optim ized using such a buffer layer.

As a matter of fact, ITO is a degenerated n-type semiconductor, but the absolute value of the work function ITO is around 4.5–4.7 eV, which allows for extraction (OPVCs) (injection (OLEDs)) from (into) electron donor (hole conducting) organic semiconductors [14–17]. In order to achieve a good band match between the electrode work function (WFITO = 4.5–4.7 eV) and the highest occupied molecular orbital of the electron donor (p-type) (5–5.6 eV) and to avoid any barrier effect at the interface of the electrode/organic material, different surface treatments, such as plasma or acid treatments, enable WFITO to increase [15, 18, 19]. However, high and reproducible work functions are difficult to obtain for ITO [20]. Therefore, it is easier to obtain reproducible results through the introduction of an anode buffer layer, also called a hole transporting layer (HTL), between the ITO electrode and the organic electron donor [21, 22, 23]. Different buffer layers can be used, such as the conductive polymer PEDOT:PSS (poly(ethylene dioxythiophene doped with poly- styrene sulfonic acid)) [19], an ultra thin metal layer [24] or transition metal oxides [21]. MoO3 is most often used due to its high efficiency as a HTL. When deposited by sublimation, its band structure has been carefully studied by different research groups who have shown that sublimated MoO3 is an n-type semiconductor with large absolute values of its ioniz- ation potential, electronic affinity and work function [25–28].

More recently, it was shown that MoO3 deposited by a simple wet technique also enables an efficient HTL to be achieved [29]. Therefore, in the present work, structures using MoO3 as a dielectric have been studied thoroughly. However, we have recently shown that a metal bi-layer—Cu/Ag—enables the transmission range of the structure to be widened, while ZnS is used as the dielectric [30] with ZnS being substituted for MoO_3−x in order to prevent Cu diffusion [31]. We also present results dedicated to ZnS/Cu/Ag/ZnS structures.

In the present manuscript, we study the ageing process of the D/M/D structures. Actually, over the last few years, many publications have been dedicated to a wide variety of D/M/D structures, but the lifetime of these structures has rarely been studied. In the present work, we followed the electrical and optical properties of several types of multilayer structures— up to six years for some. These structures were characterized by x-ray diffraction (XRD), x-ray photoelectron spectroscopy (XPS), scanning electron microscopy (SEM) and energy dis- persive spectrometry (EDS). We show that the stability of the structures, MoO_3−x/Ag/MoO_3−x [32], ZnO/Cu/Ag/MoO3 [33]

and ZnS/Cu/Ag/ZnS [30], depends not only on the different layers of the structures themselves, but also on other factors, such as the cleanliness of the substrates used. If some structures became highly resistive, other structures preserved all of

(4)

their optical and electrical properties. These differing results are discussed in the light of the characterizations performed on the samples.

2. Experimental procedures

2.1. Realization and characterization of the dielectric/metal/

dielectric′ structures

The structures were deposited under vacuum onto glass or PET substrates. The deposition apparatus and the process used for D/Cu/Ag/D′ realization have already been described [30].

When D is ZnS or MoO3, it is deposited by sublimation from a molybdenum boat, while when D is ZnO, it is deposited by magnetron rf sputtering [33]. Tungsten crucibles were used for metal evaporation. The thickness and deposition rate of the different films were measured using a quartz monitor.

The surfaces of the structures were observed with a field emission scanning electron microscope (JEM JEOL 7600F).

Secondary electron images and backscattered electron images (BEI) were conducted. The composition of the films was studied by EDS with a Bruker Quantax silicon drift detector mounted to the SEM.

All XRD scans were carried out using a Siemens D5000 diffractometer using Cu Kα radiation (λKα = 0.1542 nm). A low grazing angle of x-ray incidence was set at 1°, and the detector scanned in the 2θ range from 20°–50°. A standard Bragg–Brentano θ/2θ geometry was also used for compar- ison. The center of each Bragg peak was identified using the Gaussian curve-fitting method.

The optical measurements were carried out at room temperature using a Perkin spectrophotometer. The optical transmission was measured in the 0.3 µm–1.0 µm spectral range.

The conductivity at room temperature was measured by the Van der Pauw four probes technique. The linearity of the I(V) characteristic was systematically checked on a three order of magnitude current range and possible thermoelectric effects were cancelled by inverting the polarity current. All measurements were computer controlled.

Concerning the XPS measurements, an Axis Nova ana- lytical spectrometer (Kratos) with an Al Kα line (1486.6 eV) as an excitation source was used. The core level spectra were acquired with an energy step of 0.1 eV using a constant pass energy mode of 20 eV (energy resolution of 0.48 eV). With regards to calibration, binding energy for the C 1s hydro- carbon peak was set at 284.8 eV. Then the data were analysed with CasaXPS software.

Atomic force microscopy (AFM) (JPK instruments, NanoWizard, Berlin,Germany) was used for the topography imaging and roughness measurements of the surfaces. The images were taken in air. The intermittent contact mode was used for the AFM imaging. Classical silicon cantilevers were used (NanoWorld NCHR probe). The average force constant and resonance were approximately 42 N m⁻¹ and 320 kHz, respectively. Topographic images were recorded on different areas of the surface. The SPM Image Processing software (v.2-47) from Gwyddion was used to calculate the surface roughness (analyzed area: 5 × 5 µm²).

To track the evolution with time of the electrode properties, three types of structures were studied: MoO3/Ag/MoO3, ZnO/Cu/Ag/MoO3 and ZnS/Cu/Ag/ZnS. The thickness of the different films was optimized earlier: MoO_3−x (20 nm)/

Ag (11 nm)/MoO_3−x (35 nm) [32], ZnO (20 nm)/Cu (4 nm)/

Ag (6 nm)/MoO3 (35 nm) [33] and ZnS (50 nm)/Cu (3 nm)/Ag (9 nm)/ZnS (45 nm) [30]. The different structures were characterized immediately after deposition and periodically there- after. Between each measurement, the samples were stored in ambient air in closed and transparent boxes. These boxes were stored in the experiment room. The statistics from recent years show that during the period when the structures are tracked, the relative humidity of the room was 40%–60%, while its temperature was situated between 18 °C and 30 °C. It shows that the samples were permanently submitted to a significant degree of relative humidity, but at moderate temperature.

Regarding the substrates—soda lime glass and PET—the cleaning process was as follows. After scrubbing with soap, these substrates were rinsed in running deionised water. Then the substrates were dried with an argon flow and loaded into a vacuum chamber (10⁻⁴ Pa).

3. Experimental results

3.1. Properties of structures immediately after realization The typical electrical and optical properties of the different structures immediately after realization are summarized in table 1.

As shown previously [30, 33], the insertion of a double metal film—Cu/Ag—between two dielectric films allows the transmission range to be widened, and therefore improves the average figure of merit, ΦM, of the structures [34].

Typical cross sections of the structures are shown in figure 1.

Whatever the structure, for instance, MoO3/Ag/MoO3 (figure 1(a)) or ZnS/Cu/Ag/ZnS (figure 1(b)), we can see from the

Table 1. Typical optical and electrical properties of different D/M/D structures just after realization.

Samples Sheet resistance (Ω/sq) T max (%) T averaged between

400 and 700 nm (%) Figure of merit FM

PET/MoO₃/Ag/MoO₃ [32] 13 84 74 4.21 × 10⁻³

Glass/MoO3/Ag/MoO3 [35] 5 88 77 17 × 10⁻³

PET/ZnO/Cu/Ag/MoO3 10 85 78 8 × 10⁻³

Glass/ZnO/Cu/Ag/MoO₃ [33] 7.55 83 81 16 × 10⁻³

Glass/ZnS/Cu/Ag/ZnS [30] 5 91 85^a 70 × 10⁻³

a T averaged between 400 nm and 1000 nm.

(5)

SEM images that the structures are homogeneous with a continuous metal layer inserted between the two dielectric layers. Highlighting the average Z contrast, the BEI mode chosen for these images clearly shows that the heaviest ele- ment is the brightest one.

As shown by the AFM study, the root mean square (rms) of the different structures is small; the values deduced from the AFM images are 1.05 nm, 1.35 nm, 1.56 nm for MoO₃/Ag/MoO₃, ZnO/Cu/Ag/MoO₃ and ZnS/Cu/Ag/ZnS respectively.

We have already shown that these multilayer structures, deposited onto glass or PET substrates, behave as efficient anodes in organic photovoltaic cells based on a planar heterojunction deposited under vacuum [32, 35, 36]. Since the object of the present work was to study an original electrode, we chose to use a very classical and well known planar heterojunction based on the couple CuPc/C₆₀. These products are commercial and not too expensive, allowing us to conduct repetitive tests without restriction due to the scarcity of the products. Thus, in the case of a MoO₃/Ag/MoO₃/CuI/CuPc/

C₆₀/BCP/Al solar cell, an efficiency of 1.60% was achieved with an open circuit voltage V_oc of 0.49 V, a short circuit current Jsc of 6.26 mA cm⁻² and a fill factor (FF) of 52% [9].

It must be noted that this result is very near to that obtained

with a classical ITO electrode covered by a MoO3 HTL with the same experimental conditions (η = 1.80%, Voc = 0.53 V, Jsc = 6.3 mA/cm² and FF = 54%).

The MoO3 thin films were systematically deposited by sublimation under vacuum. Their deposition rate was 0.05–0.1 nm s⁻¹. Under these conditions, our MoO3 films were amorphous with a resistivity around 10⁻⁶ Ωcm, and they are n-type. The refractive index n of our MoO3 films has been measured; it varies with the wavelength from 2.1 at 370 nm to 1.82 at 650 nm [37]. Since, as discussed in [29], the value of the work function of the MoO3

layers depends on their composition, we have checked the value of the MoO3 work function, WFMoO3, using a Kelvin probe instrument (KP Technology Model SKP5050) [38]. The value obtained, WFMoO3 = 5.3 eV, shows that the MoO3 of our multilayer structures can be used as an efficient HTL, which is confirmed by the cells’ performance presented above.

After this reminder concerning the properties of the D/M/D structures, we will focus particuarly on the evolution, over time, of their electrical and optical properties.

3.2. Evolution over time of the electrical and optical properties of the D/M/D multilayer structures

Tracking the optical and electrical properties of the structures demonstrates that the most spectacular changes concern the electrical properties. In table 2 we present the evolution over time, in years, of the sheet resistance of typical samples. It can be seen that some structures remain conductive, others lose their conductivity, while some exhibit intermediate sheet resistance values, whatever their configuration. In table 2, the values which can be measured but which are too high (⩾100 Ω/sq) are referred to as ‘intermediate sheet resistance values’ to enable us to use these structures as an electrode in the performance of the optoelectronic devices.

For comparison we have also checked the evolution with time of the properties of ITO films. In the case of glass substrate, the properties of ITO films, which have been treated at 250 °C, are remarkably stable, without any variation after six years. This is not the case for plastic substrate. In this case, the sheet resistance changes from 25 Ω/sq for new PET/ITO samples to 900 Ω/sq after six years. Therefore, while glass/ITO TCE is very stable, the stability of the PET/MoO3/Ag/MoO3

structures is better than that of PET/ITO. The object of the present work is to study the stability of D/M/D structures— structures which can replace ITO in flexible optoelectronic devices. Therefore we will essentially focus our characterization study on the six-year-old MoO3/Ag/MoO3 structures which were deposited onto PET.

The optical properties generally evolve very little. When MoO3 is the dielectric, the transmission increases by 1%– 2% during the first few months and then it stabilizes (figure 2). MoO3 thin films deposited by sublimation are oxygen deficient, inducing the presence of states in the gap. These gap states induce slight blue film coloration. The long air exposure of these films results in a progressive decrease of the density of oxygen vacancies, which explains the small transmission increase of the MoO3/Ag/MoO3 structures. Then, after a longer time period, the evolution of the

Figure 1. BEI mode SEM images of the cross section of a multilayer structure MoO₃/Ag/MoO₃ (a) and ZnS/Cu/Ag/ZnS (b).

(6)

transmission of the structures depends on their conductivity. Actually, it can be seen in figure 2 that if the shape of the transmission curve of the new and old MoO₃/Ag/MoO₃ structures are similar when the old structure stays conductive, significant differences are visible when the structure has become insulating. While the transmission of the conductive structures decreases when the wavelength increases towards the near IR range, it stays stable in the insulating sample. This decrease in the transmission in the near-IR is due to the oscillations of free electrons. On the other hand, in the case of the insulating sample, i.e. the sample without free electrons, the transmission stays high and stable even in the near-IR range (figure 2). Therefore the evolution of the electrical properties of the structures with time can be qualitatively estimated through optical measurements.

After six years of room air exposure, the slight broadening of the transmission range of the conductive MoO₃/Ag/

MoO₃ structure corresponds to a moderate increase in the sheet resistance of the structure (table 2). Similar results were obtained with the other D/M/D structures.

The tracking results of the electrical properties of the different structures are summarized in table 2. Each result given in the table corresponds to a value averaged from six samples.

We can conclude that the samples can exhibit very different behaviors.

Since some six-year-old MoO3/Ag/MoO3 structures are still conductive, we have probed them as an anode in photovoltaic solar cells. We used a cell-based planar heterojunction deposited under vacuum. The typical cell architecture includes the planar heterojunction CuPc/C60 sandwiched between two electrodes, one of them being transparent. Moreover, hole (CuI layer) and electron (BCP) transport layers were placed between the electrode and the organic layer to improve the charge collection. To realize the cells, we used the same method as described in [35], the final device being MoO3/Ag/MoO3/CuI/

CuPc/C60/BCP/Al. Unfortunately, the leakage cur rent of these devices was very important. Therefore we have added, before the deposition of CuI, a new thick oxide layer of 10 nm. The J–V characteristics then obtained in obscurity and under 100 mW cm⁻² light intensity are shown in the insert of figure 2.

The best performance is Voc = 0.35 V, Jsc = 1.37 mA/cm², FF = 35.5% and η = 0.17%. If we compare these results with the performance achieved using a MoO3/Ag/MoO3 structure just after realization, there is, of course, a significant loss of efficiency. The loss in Jsc and FF is perhaps due to the increase of sheet resistance (table 2), and the decrease of Voc can be attributed to the increase of the leakage current due to the higher surface roughness of the anode—all of which results in a relatively low cell efficiency. Nevertheless, it is positive to see that it is still possible to use a MoO3/Ag/MoO3 structure after six years of air exposure. The fact that a new oxide layer is required will be discussed after the characterization of the aged structures.

3.3. Morphological and structural study of D/M/D multilayer structures

With regards to the discrepancies visible in table 2, SEM coupled with EDS appear to be very effective tools to understand these different behaviors.

Before we proceed to a more detailed study, it must be noted that, whatever the type of structure used, it can be seen

Table 2. Evolution over time of the sheet resistance of different typical D/M/D structures.

Samples

Time in years

1.5 3.5 6

Sheet resistance (Ω/sq)

PET/MoO₃/Ag/MoO₃(C) 10 15 17

PET/MoO₃/Ag/MoO₃(I) 14 23 114^a

PET/MoO₃/Ag/MoO₃(NC) 15 150^a No conduction

Glass/ZnO/Cu/Ag/MoO₃(C) 5 12 18

Glass/ZnO/Cu/Ag/MoO3(I) 8 16 210^a

Glass/ZnO/Cu/Ag/MoO3(NC) 6 180^a No conduction

Glass/ZnS/Cu/Ag/ZnS(C) 16 44

Glass/ZnS/Cu/Ag/ZnS(I) 70 800^a

Glass/ZnS/Cu/Ag/ZnS(NC) 600^a No conduction

a Sheet resistances higher than 100 Ω cm are referred to as ‘intermediate sheet resistance values’. (C: conductive, I: Intermediate sheet resistance, NC: non-conductive)

Figure 2. Transmission spectra of MoO₃ (20 nm)/Ag (11 nm)/

MoO₃ (35 nm) structures: just after deposition , six years after deposition: conductive , no conductive . Insert: J–V characteristics of a solar cell using a six-year-old MoO₃/Ag/MoO₃ structure as an anode.

(7)

with the naked eye that some samples have become inho- mogeneous. The photo in figure 3 shows massive migration starting from the edge of the glass substrate towards the centre of the structure, which results in completely insulating and poorly transmitting samples.

As can be seen in figure 4, whatever its conductivity, each sample exhibits surface inhomogeneities. However, the size, and above all the density, of the defaults are different. In the conductive samples, the maximum diameter of the defects is of the order of 3–5 µm, while it can be more than 10 µm for the insulating samples. Moreover, it can be seen at first glance that the density of the defaults is far higher in the non- conducting samples. This is corroborated below using the open-source software ImageJ: https://imagej.nih.gov/ij/ [48].

It can be seen in the insert of figure 4(a) that defect growth often begins at something like a nucleus and spreads radially outward to form nearly circular patterns.

In order to check whether the composition of the defects is the same as that of the non-disturbed domains, we have visualized the surface of the structures in the BEI mode. In that mode, the heavier atoms appear brighter on the photographs.

As shown in figure 5, often the features in the centre of the defects appear darker than the film itself, while the disturbed zone around the central feature is brighter. This shows that the atoms constituting the central features are lighter, while those present around them are heavier than those of the rest of the film.

The SEM pictures in BEI mode of the conductive and non- conductive structures have been exploited by using ImageJ.

Such software allows us to improve the exploitation of the images themselves. The SEM images were first filtered to reduce the noise and then converted to a binary image by threshold operation to observe the area corresponding to the features (in black) and that of the substrate (in white) (figure 6). Then ImageJ can measure the coverage of the black features [24].

When the images of figure 6 are treated with ImageJ, it can be calculated that the surface coverage ratio is 12% in the non- conductive structure, while it is only 3.5% for the conducting

Figure 3. Visualisation of a depredated DMD structure.

Figure 4. SEM images of non-conductive (a), semi-conductive (b) and conductive (c) MoO₃/Ag/MoO₃ structures.

Figure 5. SEM image of the surface of a non-conductive structure (BEI mode). To aid understanding, circles have been introduced showing typical defects (continuous line) and a zone with heavy atom accumulation (dotted line).

(8)

sample, which confirms that the density of defaults is far higher in the non-conducting samples.

We can see in figure 4 that in the sample with intermediate electrical properties, the density and size of the inhomogeneities present in the films are smaller compared with the non- conductive sample, while they are higher compared to those which are conductive, which explains the different behav- iours. This first impression with the naked eye is confirmed by the treatment of the images according to the process described above using ImageJ. We calculated that the surface coverage ratio is around 6% in the intermediate structure, i.e. this value is situated between those obtained for conductive (3.5%) and non-conductive structures (12%). This shows that there is a close correlation between the conductivity of the structures and the density of the inhomogeneities they contain.

3.4. Chemical analysis of the defects embedded in the D/M/D multilayer structures

In order to check the chemical composition of the different defects embedded in the D/M/D multilayer structures, we conducted EDS analysis of the typical defects, and also large surfaces, including defects and homogeneous domains. For the first time, we studied the specific domains, central features, disordered areas, and homogeneous areas, and then we began the elemental mapping of the visualized surfaces.

Before we discuss the results, it must be noted that because the total thickness of the samples is less than 100 nm, the substrate, for instance silicon, appears systematically in the chemical analysis. Moreover, the structures themselves consist of stacked layers. This means that only qualitative results can be obtained. Nevertheless, clear tendencies can be deduced from the analysis of the specific domains. The composition varies according to the central features analyzed and will be

discussed below. Around these central defects, there is some metal accumulation by migration from the non-disturbed domains towards the disturbed ones. For instance, in the case of the ZnS/Cu/Ag/ZnS structures, the energy spectra of the different domains can be summarized as follows:

-Homogeneous domain: 1 Cu for 3 Ag, with a ratio ZnS/M (Cu + Ag) around 4, while a strong peak due to Si in the glass substrate is present.

-Disturbed area: the ratio Cu/Ag increases a little, while that of ZnS/M (Cu + Ag) decreases strongly and tends towards 1.5 and Si is less visible. This means that there is metal accumulation in these domains and that these areas are substantially thicker than the original film from which they grew.

In the centre of the disturbed area (black domains), additional light atoms are often present. These extra atoms can be different from one black domain to another, which led us to discriminate between different origins of these black defects (figure 7):

-Figure 7(a) Organic dust: increase of the relative intensity of C, O and Cl surrounded by metal accumulation

-Figure 7(b) Inorganic dust: increase of the relative intensity of Mg, Al. Here also the dust is surrounded by metal accumulation.

-Figure 7(c) No dust: break, pinhole in the top layer which induces metal accumulation.

In order to investigate more precisely the origin of these different features, we proceeded to microprobe map the visualized surfaces. The results, coupled with the BEI image, are reported in figure 8 for a ZnS/Cu/Ag/ZnS structure, and figures 9 and 10 for a MoO3/Ag/MoO3 structure.

Figures 8(b)–(d) present the map of the distribution of Ag, Cl and Cu in the film visualized in figure 8(a). It is clear that

Figure 6. SEM study of non-conductive (a) and conductive (b) MoO₃/Ag/MoO₃ structures. The image in BEI mode is treated with ImageJ (some examples of the features used to calculate the surface coverage ratio are visualized by circles).

Figure 7. SEM images of the different possible centres of the disturbed domains: organic dust (a), inorganic dust (b), no dust (c).

(9)

the bright red spots in figure 10, which are due to high Ag concentration, correspond to the bright spots of figure 8(a).

Also, figure 8(d) confirms the presence of quite high Cu concentration. These results confirm that the metals migrate from the non-perturbed domain toward the spherical defects. More surprising is the presence of Cl (figure 8(c)). Moreover it can be seen that figures 8(b) and (c) are perfectly similar, which means that Cl is probably bound to Ag. It is clearly visible in the center that there is no match between figures 8(c) and (d), which means that Cl is probably not bounded to Cu.

In order to confirm the metal migration and the presence of Cl and to identify more clearly the repartition of these different elements, we present, in figures 9 and 10, the study of a disturbed domain of a MoO₃/Ag/MoO₃ structure.

We can see in figure 10 and S1 (available online at stacks.

iop.org/JPhysD/50/375502/mmedia) that Ag and Cl correspond precisely to the bright point of the image. At the same time, the cartography of Si corresponds to the ‘nega- tive’ image. It means that the bright point of the image corresponds to the smaller contribution of Si. This confirms that metal accumulates in the spherical perturbed domains which are thicker than the structure itself.

These results are confirmed by the curves shown in figure 11. It can be seen that the concentration of Ag and Cl is far higher in the bright points, while that of Si is smaller and the concentration of Mo is stable. It means that if Ag (and Cu) diffuses, Mo (and therefore MoO3) is stable.

An examination of more than five defects in the different structures provides results similar to those presented above.

The insulating samples were characterized by XRD. When the standard Bragg–Brentano θ/2θ geometry was used, the XRD diagrams reveal some peaks (figure 12). Two of them are systematically present whatever the sample; one is situated at 32.20° and the other at 38.0°, and the second one can be attributed to Ag, while the first can be assigned to AgCl [39].

Following the JCPDS card N° 31–1238, the peak situated at 32.20° can be attributed to the plane (2 0 0) of the cubic phase of the AgCl crystal. The peak situated at 38° corresponds to the plane (1 1 1) of cubic Ag (JCPDS 65-2871).

In order to obtain more precise information, as the multilayer structures are quite thin, XRD diagrams using a low x-ray incidence grazing angle were also made (inset of figure 12). In addition to the two peaks already reported, a third peak, located at 46.2°, is present. This peak, according to the JCPDS No 31-1238 of AgCl, can be attributed to the

Figure 8. SEM image, in BEI mode, of the surface of a ZnS/Cu/Ag/ZnS structure (a) and of the corresponding Ag (b) Cl (c) and Cu (d) cartography.

Figure 9. SEM image of a perturbed zone of a MoO₃/Ag/MoO₃ structure. Insert: magnification of the studied perturbed domain.

(10)

(2 2 0) plane. Therefore the two techniques are in good agreement, the second performing better for our thin film structures.

In the disturbed regions, the XPS study confirms both the presence of AgCl and MoO3 at the surface of the film, which confirms that the dielectric is still present (figure 13).

Moreover, the binding energy of the peaks Ag 3d5/2 and Cl 2p, 368.4 eV and 198.4 eV, respectively, correspond to the values expected for AgCl [40]. The XPS quantitative analysis shows that, whatever the sample, the relative atomic concentration of Ag and Cl is such that there is an excess of Ag relative to AgCl with 30 at%–40 at% of Cl. It shows that if the formation of AgCl contributes to the degradation of the properties of the multilayers structures, it is not at the origin of Ag migration since there is an excess of Ag. It can be concluded that Ag migrates towards the inhomogeneities and then reacts with Cl.

It can be noted that the decomposition of the Mo 3d peak (figure 13(c)) shows that only one contribution is necessary to

obtain a good agreement between the experimental and theor- etical curves. The peak situated at 232.6 eV corresponds to Mo 3d5/2 and that at 235.7 eV to Mo 3d3/2 of the Mo⁶⁺ doublet.

The absence of Mo⁵⁺ confirms that, as proposed in the study of the optical properties of the non-conductive structures, with time the oxygen vacancies disappear.

Finally, we have checked the evolution of the surface of the structures by AFM. If, just after realization, the structures are smooth with an rms around 1–1.5 nm, after ageing the surface of the films becomes quite rough. For instance, when the MoO3/Ag/MoO3 structure in figure 3 is studied by AFM, the measures on the side of the sample (figure 14(a)) give an rms of 4.5 nm, while in the central disturbed domain the rms

Figure 10. BEI image of the surface of a MoO3/Ag /MoO3 structure (a) and of the corresponding Ag (b), Cl (c) and Si (d) cartography.

Figure 11. Spectra of different domains of figure 10: large area and bright spot .

Figure 12. XRD diagram of a non-conductive MoO3/Ag/MoO3

structure obtained using the standard Bragg–Brentano θ/2θ geometry. Inset: XRD diagram obtained using grazing incidence.

(11)

is 130 nm (figure 14(b)). It confirms that the disturbed areas, where the metal accumulates, are strongly perturbed.

Therefore, the metals migrate towards the perturbed domains where they react with Cl. Of course this migration of the inserted metal layer induces a progressive decrease in the conductivity of the structure. That can result in structures which are still conductive or with intermediate sheet resistance. For a critical concentration of perturbed domains, there is no more percolation of the metal films and the structure becomes non-conductive. This on–off switching effect depends on the density of the defects present in the structure.

When only 3.5% of the film corresponds to the defects, the structure retains its conductivity, but for more than 10%, the structure loses its conductivity.

4. Discussion

We have seen that defects grow at a ‘nucleus’, usually a microscopic particle (or pinhole etc), then they spread outward radially to form a nearly circular pattern. Within the affected area, the metals—Ag and also Cu—when they are present, agglom- erate to small islands and the dielectric layers are still present.

These results recall, at least partly, those obtained some years ago by Ross [41]. Following his study of Ag thin films and ZnO/Ag/ZnO structures, Ross proposed the following process for degradation:

First step: Initiation occurs almost universally at particu- late contamination or breaks in the film, such as the film edge,

after exposure to H2O vapour. Initiation primarily involves H2O contact with the Ag layer through a breach in the over- lying film. The fact that corrosion in structures starts at isolated centres suggests that breaks in the Ag film itself may promote initiation.

Nevertheless, it must be noted that, in contrast with the result obtained by Ross, the chemical reactivity of Ag (and Cu) significantly influence the degradation process since the agglomerated Ag particles have reacted, at least partially, with Cl to give AgCl. This fact will be discussed below.

Second step: Defect propagation follows initiation, and is dominated by the physical agglomeration of Ag film. The presence of water primarily controls this process, and the presence of the surrounding dielectric layers modifies it.

As a matter of fact, our study indicates a widely variable period of time between initial air exposure and the onset of defect growth (table 2). This variable incubation period may correspond to the time required for water to first penetrate into Ag through a partially sealed area in the overlaying dielectric films. That ensures that the lifetime of the structures depends on the cleanliness of the substrates and on the homogeneity of the dielectric films. For instance, as shown in figure 3, some- times the degradation process starts from the edges of the samples. That is because the film edges initiate defects due to obvious mechanical discontinuity.

This hypothesis is confirmed by Ross’ study [41]. As a matter of fact, he has observed that defects grow around the edge of the glass in ZnO/Ag/ZnO structures, while the presence of such effects is much lower in Ag monolayers, where corrosion is more uniformly distributed. This is because, in D/

Ag/D structures, Ag is isolated by the dielectric layers from the surrounding reactive media: the atmosphere and substrate.

With regards to the presence of Cl, it might be interesting to refer to the specific studies dedicated to archaeological silver objects. It is well known that old silver objects are often corroded by Cl. The corrosion of silver results from the interaction between the metal and its environment and can assume very different forms as a function of the way in which this interaction takes place—tarnishing is one of the most frequently encountered [42]. Tarnishing means that silver exposed to the atmosphere becomes covered by a dark film, which originates from an interaction with air pollutants. AgCl has been identified as a major component of tarnishing. In the case of archaeological silver objects, the origin of Cl was explained by the deposition of chloride containing airborne particles from combustion processes, the dispersion of marine salts or even from the purification treatment of water in urban areas. Similarly, in our case, Cl can result from ambient contamination, the samples being stocked in the experimental room which contains different chemical products.

In the same way, in marine silver artefacts, the main corro- sive product is silver chloride (AgCl) [43]. It must be noted that the formation of AgCl requires the absence of silver oxides.

Silver chloride, in contrast to silver oxide, does not create a protective layer, thus the metal can be completely transformed into silver chloride. The study of the electrochemical formation and reduction of a thick AgCl deposition layer on a silver substrate has shown that a dry AgCl layer has a fairly low

Figure 13. XPS spectra of a non-conductive MoO3/Ag/MoO3

structure, (a) Ag 3d, (b) Cl 2p, (c) Mo 3d and (d) O 1s.

(12)

ionic conductivity but its conductivity can be remarkably enhanced by simply wetting with water [44]. The conductivity enhancement by wetting should result from an increase in the ionic mobility (Ag⁺) due to the water molecules, which facilitates the AgCl growth. This fact corroborates the crucial importance of the presence of H₂0 in the degradation process of Ag films. It is evident that pinholes in the film covering the silver surface are a very suitable site for humidity to condense and to initiate the dissolution of the metal. Silver ions have a high ability to migrate and move quickly to the surface, where they can accumulate in the form of dendrites or filaments. The presence of Cl gases will accelerate these processes.

Following these results, we have begun a new study with samples stored in different conditions, such as in a vacuum, outdoors and in an extractor hood. We can only provide the initial results obtained after quite a short period (when compared to the six years which elapsed for the above study).

Nevertheless, significant trends can be deduced from the first few months of the study. The initial sheet resistance of the PET/MoO₃/Ag/MoO₃ structures was 13–15 Ω/sq. After two months, the sheet resistance of the structure stored outdoors was 1 × 10⁶ Ω/sq, that of the sample in the extractor hood was 51 × 10⁶ Ω/sq, while that of the structure under vacuum was stable. It means that in the presence of a high level of humidity or Cl, the electric properties of the structures rapidly degrade, which is not the case under vacuum, where these contami- nants are absent.

Finally, the necessity of the introduction of a thick oxide layer of 10 nm above the aged conductive structures before using them as an anode in OPVCs is because, even when they are conductive, the aged structures exhibit some surface defects (figure 4(b)). These defects correspond to Ag-rich regions, which induces the leakage currents encountered in the absence of this additional oxide layer.

5. Conclusion

It can be concluded that the stability of the D/M/D structures does not rival that of ITO in the case of a glass substrate.

Nevertheless, for plastic substrates, the stability of the PET/

MoO3/Ag/MoO3 structure is better than PET/ITO, which

means that D/M/D structures are better suited for flexible devices.

The disturbed region in the D/M/D structures corresponds to metal accumulation. They form around defects previously present: dusts, breaks, pinholes, etc. Their progressive growth induces the formation of metal-depleted zones, leading to the loss of continuity of the metal layers.

It is well known that Ag tends to be a highly diffusive ele- ment. Moreover, the effect of humidity promotes Ag electrochemical reactions that produce Ag⁺ ions which induce AgCl formation.

In the future, in order to suppress or at least strongly decrease the occurrence of degradation processes in D/M/D structures, some care must be taken. We have shown that defects growth usually at a ‘nucleus’, which is either organic, inorganic or some morphological default. It means that cleaning must be carried out with great care. Alkali ions on the glass surface may accelerate the corrosion of the Ag film.

In order to avoid the possibility of such a corrosion process, it would be better to wash the substrates with acid (HNO₃) which avoids the presence of Ca⁺ et Na⁺ at the surface of the substrate [45]. Moreover, it was shown that the wetta- bility of oxides by Ag increases when the dielectric band gap decreases [46]. Therefore it would be helpful to introduce a small band gap oxide between the dielectric and the Ag layer.

These structures, when used as a bottom electrode in devices, are covered by different films, necessary for their realization, which ensure they will be isolated from any contact with H₂O.

Moreover, the necessary encapsulation of the devices will effi- ciently protect the entire component from moisture contact.

Nevertheless, in order to avoid such encapsulation and obtain stable structures, i.e. avoid silver aggregation, we suggest the use of silver alloys. For instance, it has been shown that Ni can be used as a barrier diffusion between stacked metal layers [47]; therefore works dedicated to MoO₃/Ag:Ni (1–3 wt%)/

MoO₃ structures are now underway in our laboratory.

Acknowledgment

The authors acknowledge funding from the European Com- munity ERANETMED_ENERG-11-196: Project NINFFE.

Figure 14. AFM images of a non-conductive MoO₃/Ag/MoO₃ structure, (a) peripheral area of the sample shown in figure 3, (b) central zone of this sample.

(13)

References

[1] Baia L, Quintela M, Mendes L, Nunes P and Martins R 1999 Performances exhibited by large area ITO layers produced by r.f. magnetron sputtering Thin Solid Films 337 171–5

[2] Demirel B, Yara A and Elçiçek H 2011 Crystallization behavior of PET materials BAÜ Fen Bil. Enst. Dergisi Cilt 13 26–35

[3] Minami T 2008 Substitution of transparent conducting oxide thin films for indium tin oxide transparent electrode applications Thin Solid Films 516 1314–21

[4] Liu H, Avrutin V, Izyumskaya N, Özgür Ü and Morkoç H 2010 Transparent conductive oxides for electrode applications in light emitting and absorbing devices Superlattices Microstruct. 48 458–84

[5] Emmott C J M, Urbina A and Nelson J 2012 Environmental and economic assessment of ITO-free electrodes for organic solar cells Sol. Energy Mater. Sol. Cells 97 14–21

[6] Cao W, Li J, Chen H and Xue J 2014 Transparent electrodes for organic optoelectronic devices: a review J. Photon.

Energy 4 040990

[7] Park H J, Kim J, Won J H, Choi K S, Lim Y T, Shin J S and Park J-U 2016 Thin Solid Films 615 8–12

[8] Guillén C and Herrero J 2007 Structure, optical, and electrical properties of indium tin oxide thin films prepared by sputtering at room temperature and annealed in air or nitrogen J. Appl. Phys. 101 073514

[9] Bernardo G, Gonçalves G, Banquinha P, Ferreira Q, Brotas G, Pereira L, Charas A, Morgado J, Martins R and Fortunato E 2009 Polymer light-emitting diodes with amoprhous indium–zinc oxide anodes deposited at room temperature Synth. Met. 159 1112–5

[10] Granqvist C G and Hultâker A 2002 Transparent and conducting ITO films: new developments and applications Thin Solid Films 411 1–5

[11] Hssein M, Cattin L, Morsli M, Addou M and Bernède J C 2017 Importance of the electrode conductivity in organic photovoltaic solar cells J. Mater. Sci., Mater. Electron.

28 3678–84

[12] Guillén C and Herrero J 2011 TCO/metal/TCO structures for energy and flexible electronics Thin Solid Films 520 1–17 [13] Cattin L and Bernède J C 2013 Toward indium-free

optoelectronic devices: dielectric/metal/dielectric alternative transparent conductive electrode in organic photovoltaic cells Phys. Status Solidi a 210 1047–61

[14] Gassmann J, Brötz J and Klein A 2012 Sputter deposition of indium tin oxide onto zinc phthalocyanine: chemical and electronic properties of the interface studied by photoelectron spectroscopy Appl. Surf. Sci. 258 3913–9 [15] Irfan I, Graber S, So F and Gao Y 2012 Interplay of cleaning

and de-doping in oxygen plasma treated high work function indium tin oxide (ITO) Org. Electron. 13 2028–34

[16] Yamada M, Matsumura M and Maeda Y 2011 Suppression of damage to organic light-emitting layers during deposition of Al-doped ZnO thin films by radio-frequency magnetron sputtering Thin Solid Films 519 3352–7

[17]Andersson A, Johansson N, Bröms P, Yu N, Lupo D and Salaneck W R 1998 Fluorine tin oxide as alternative to indium tin oxide in polymer LEDs Adv. Mater.

10 859–63

[18] Nüesch F, Rothberg L J, Forsythe E W, Le Q T and Gao Y 1999 A photoelectron spectroscopy study on the indium tin oxide treatment by acids and bases Appl. Phys. Lett.

74 880–2

[19] Johnev B, Vogel M, Fostiropoulos K, Mertesacker B, Rusu M, Lux-Steiner M-C and Weidinger A 2005 Monolayer

passivation of the transparent electrode in organic solar cells Thin Solid Films 488 270–3

[20] Sharma A, Kippelen B, Hotchkiss P J and Marder S R 2008 Stabilization of the work function of indium tin oxide using organic surface modifiers in organic lightemitting diodes Appl. Phys. Lett. 93 163308

[21] Godoy A et al 2010 Effects of the buffer layer inserted between the transparent conductive oxide anode and the organic electron donor Sol. Energy Mater. Sol. Cells 94 648–54 [22] Yoo I, Lee M, Lee C, Kim D-W, Moon I S and Hwang D-H

2005 The effect of a buffer layer on the photovoltaic properties of solar cells with P₃OT:fullerene composites Synth. Met. 153 97–100

[23] Chan M Y, Lee C S, Lai S L, Fung M K, Wong F L, Sun H Y, Lau K M and Lee S T 2006 Efficient organic photovoltaic devices using a combination of exciton blocking layer and anodic buffer layer J. Appl. Phys. 100 094506

[24] Cattin L, Tougaard S, Stephant N, Morsli S and Bernède J C 2011 On the ultrathin gold film used as buffer layer at the transparent conductive anode/organic electron donor interface Gold Bull. 44 199–205

[25] Mayer J, Hamwi S, Kröger M, Kowalski W, Riedl T and Kahn A 2012 Transition metal oxides for organic electronics: energetic, device physics and applications Adv.

Mater. 24 5408

[26] Irfan I, Ding H, Gao Y, Small C, Kim D Y, Subbiah J and So F 2010 Energy evolution of air and oxygen exposed molybdenum trioxide films Appl. Phys. Lett. 96 243307 [27] Chiam S Y et al 2012 Investigating the stability of defects in

MoO₃ and its role in organic solar cells Sol. Energ. Mat.

Sol. Cells 99 197–203

[28] Bernède J C, Cattin L and Morsli M 2004 About MoO₃ as buffer layer in organic optoelectronic devices Technol. Lett.

1 5–17

[29] Cong S et al 2017 Modifying the valence state of molybdenum in the efficient oxide buffer layer of organic solar cells via a mild hydrogen peroxide treatment J. Mater. Chem. C 5 889–95

[30] Mouchaal Y et al 2015 Broadening of the transmission range of dielectric/metal multilayer structures by using different metals Vacuum 111 32–41

[31] Pérez Lopéz I, Cattin L, Nguyen D-T, Morsli M and Bernède J C 2012 Dielectric/metal/dielectric structures using copper as metal and MoO₃ as dielectric for use as transparent electrode Thin Solid Films 520 6419–23 [32] Abachi T et al 2013 Highly flexible, conductive and

transparent MoO₃/Ag/MoO₃ multilayer electrode for organic photovoltaic cells Thin Solid Films 545 438–44 [33] Bernède J C, Cattin L, Abachi T, Lare Y, Morsli M and

Makha M 2013 Use of Cu–Ag bi-layer films in oxide/

metal/oxide transparent electrodes to widen their spectra of transmittance Mater. Lett. 112 187–9

[34] Haacke G 1976 New figure of merit for transparent conductors J. Appl. Phys. 47 4086–9

[35] Cattin L, Lare Y, Makha M, Fleury M, Chandezon F, Abachi T, Morsli M, Napo K, Addou M and Bernède J C 2013 Effect of the Ag deposition rate on the properties of conductive transparent MoO₃/Ag/MoO₃ multilayers Sol. Energy Mater.

Sol. Cells 117 103–9

[36] Hssein M, Tuo S, Benayoun S, Cattin L, Morsli M,

Mouchaal Y, Addou M, Khelil A and Bernède J C 2017 Cu– Ag bi-layer films in dielectric/metal/dielectric transparent electrodes as ITO free electrode in organic photovoltaic devices Org. Electron. 42 173–80

[37] Nguyen D-T, Vedraine S, Cattin L, Torchio P, Morsli M, Flory F and Bernède J C 2012 Effect of the thickness of the MoO₃ layers on optical properties of MoO₃/Ag/MoO₃ multilayer structures J. Appl. Phys. 112 063505

(14)

[38] Bernède J C, Cattin L, Makha M, Jeux V, Leriche P, Roncali J, Froger V, Morsli M and Addou M 2013 MoO₃/CuI hybrid buffer layer for the optimization of organic solar cells based on a donor–acceptor triphenylamine Sol. Energy Mater. Sol.

Cells 110 107–14

[39] Zhao X, Zhang J, Wang B, Zada A and Humayun M 2015 Biochemical synthesis of Ag/AgCl nanoparticles for visible-light-driven photocatalytic removal of colored dyes Materials 8 32043–53

[40] Kaushik V K 1991 XPS core level spectra and Auger parameters for some silver compounds J. Electron Spectrosc. Relat. Phenom. 56 273

[41] Ross R C 1990 Observations on humidity-induced degradation of Ag-based low-conductivity films Sol. Energy Mater.

21 25–42

[42] Costa V 2001 The deterioration of silver alloys and some aspect of their conservation Rev. Conserv. 2 19–34

[43] Marchand G, Guilminot E, Lemoine S, Rossetti L, Vieau M and Stephant N 2014 Degradation of archaecological horn silver artefacts in burials Heritage Sci. 2 5

[44] Jin X, Lu J, Liu P and Tong H 2003 The electrochemical formation and reduction of a thick AgCl deposition layer on a silver substrate J. Electrochem. Chem. 542 85–96

[45] Park H S and Day D E 1986 Corrosion of evaporated Ag films on glass by saturated water vapor Sol. Energy Mater.

13 351–65

[46] Lazzari R and Jupille J 2001 Silver layers on oxide surfaces: morpholgy and optical properties Surf. Sci.

482–5 823–8

[47] Zhuang W D and Eagar T W 1997 Diffusional breakdown of nickel protective coatings on copper substrate in silver- copper eutectic melts Metall. Mater. Trans. A 28 969–77 [48] Open-source software ImageJ (https://imagej.nih.gov/ij/

(Accessed: 27 July 2017))