Issue
Eur. Phys. J. Appl. Phys.
Volume 90, Number 3, June 2020
Disordered Semiconductors: Physics and Applications
Article Number 30101
Number of page(s) 14
Section Semiconductors and Devices
DOI https://doi.org/10.1051/epjap/2020190368
Published online 01 July 2020

© EDP Sciences, 2020

1 Introduction

The structural flexibility and electronic structure of amorphous chalcogenides promote various photoinduced phenomena. For example, when silver/amorphous chalcogenide bi-layer films are exposed to ultraviolet or visible light, silver (Ag) dissolves and diffuses into the chalcogenide layer [1]. Such effects have been investigated for more than 50 yr [24] due in part to the recognition of possibilities for new photon-based technologies. Silver photodiffusion, which exhibits an anomalous mode of diffusion with a step-like silver concentration profile, has potential application in memory devices [5]. Deeper understanding of photokinetics will therefore lead to an improved memory devise performance. Diffusion profiles at the Ag/chalcogenide interface have mainly been investigated by Rutherford backscattering spectroscopy (RBS) [69]. Strong ion beams can also induce silver diffusion; however, in situ measurement for studying kinetics has been quite difficult. Recently, we performed neutron reflectivity measurement on Ag/amorphous Ge chalcogenide bi-layer films under light exposure. As neutron beams do not induce silver diffusion, in situ measurements of the photodiffusion process were obtained free of interference [1017]. S-rich Ge-S samples form a metastable reaction layer between the Ag and amorphous Ge-S host layers, followed by Ag diffusion from the metastable layer to the amorphous Ge-S host layer [10, 17]. Different behaviour was observed in Ge-rich and stoichiometric (GeS2) Ge-S samples. After Ag initially dissolves into the Ge-S host layer, Ag ions immediately diffuse over the entire Ge-S layer without forming a metastable reaction layer [1315]. White light from a xenon lamp was used to initiate photodiffusion of Ag in these past neutron reflectivity measurements, without investigating the effect of the source of illumination on the photodiffusion phenomena.

Two models have been proposed to explain the driving force for silver photodiffusion. The first is a band gap model related to semiconductor behaviour, and describes the process in terms of the flows of charges (the holes and the Ag ions) at the junction [1822]. According to this type of model, band-gap illumination excites electrons from the valence band to the conduction band, creating electron-hole pairs. As most chalcogenide materials are p-type semiconductors and the energy band bends upward from the chalcogenide layer to the Ag layer at the junction, the holes are assumed to flow to the Ag layer. The chalcogenide layer becomes negatively charged as the Ag layer acquires positive charge. To maintain overall charge neutrality, positively-charged Ag ions enter into the negatively charged chalcogenide layer, in response to the charge gradient, thereby producing the silver photodiffusion effect. The idea is applicable to the films with the Ag/doped reaction/un-doped chalcogenide system [1922]. Effective layer site will be clarified by the required energy for hole generation. Therefore, it is worth examining the excitation photon energy dependence of silver photodiffusion kinetics. In the second model, which is defined as reconstruction of the bond network [18,19,22], photoinduced changes in amorphous chalcogenides are attributed to changes in bonding; e.g., creation, or breaking of a bond, and bond switching [3,4]. Analogous bond network reconstruction can occur in metal/chalcogenide systems. Kludge [18] considered cases in which Ag ions became involved in the network of chalcogenide glasses, and the process was referred as an intercalation reaction. Since the band-gap light excitation is effective for most of the photo-induced changes in amorphous chalcogenide [3,4], changing the light energy will also affect the reconstruction of the bond network in silver photodiffusion system, and the excitation light energy dependence of silver photodiffusion kinetics is worth examining.

In this paper, we report on the excitation photon energy dependence of the kinetics of silver photodiffusion as measured by neutron reflectivity, using four high-power light-emitting diodes (LED) with different peak wavelengths. Based on the results, we discuss the mechanism of silver photodiffusion.

2 Experimental

2.1 Sample preparation

2.1.1 Glass synthesis

The bulk chalcogenide glasses were synthesized by standard melt quenching technique. Pure 5N elements were weighted accurately and the required amount was loaded into a fused silica ampule and then sealed under vacuum (10−4 mbar). The glass synthesis was carried out in a programmable tube furnace for 168 h (one week) at peak temperature 750 °C. The furnace was programmed to reach 750 °C within the first 24 h of synthesis. The ampules were kept at the highest temperature for 144 h. There was an essential reason for the long duration of the synthesis - good glass homogenization. Usually it is assumed that, holding the glass melt at a temperature 20–50 °C above the highest melting phase temperature for several hours would homogenize the melt. However, at equilibrium presented in the phase diagrams, glass-forming compositions are usually bordered by congruently melting crystalline phases [23], which can nucleate as melts are quenched to produce microscopic heterogeneities [24]. The continued reaction leads to these crystalline phases to dissolve and local structures characteristic of melts/glasses to emerge [24]. Avoiding the formation of microscopic heterogeneities links to slow aging of the glasses which is a warranty for the stability of their parameters, the structure/composition of the films produced from them.

2.1.2 Thin film preparation

The chalcogenide glass and silver thin films were deposited on a single crystalline silicon substrate by thermal evaporation in a Cressington 308R coating system at 10−6 mbar vacuum at evaporation rate 0.3 Å/s. The substrates were not specifically heated during the film preparation. The thickness of the deposited films was monitored during the deposition and estimated using the output from a quartz crystal microbalance. The composition of the chalcogenide thin films was studied with Energy Dispersive Spectroscopy which showed ±1.5 at.% deviation of the films’ composition compared to the composition of the source material.

2.2 Neutron reflectivity

Neutron reflectivity technique is one of the applications of optics. A neutron reflectivity profile is obtained non-destructively, and enables determination of the density, thickness, and the roughness of thin film layers [25,26]. The technique is applicable to a multi-layer system such as Ag/Ag-doped reaction layer/amorphous Ge20 S80, enabling time-dependent changes of the three layers to be simultaneously quantified in terms of density, thickness, and roughness.

Scattering length density (SLD), β, is an important physical quantity in neutron reflectivity, and is defined according to equation (1): (1)

where ρ is the mass density of the bulk, NA is the Avogadro number, ci is the concentration of each element, bi is the scattering length for each element, and Mi is the atomic molar mass for each element. The SLD of metallic silver is 3.47 × 10−6−2 and the SLD of the amorphous germanium sulphide layer in the present study was approximately 1.8 × 10−6−2. During photodiffusion, the SLD of the Ag-doped layer should increase as Ag ions migrate into the layer.

Neutron reflectivity measurements were performed on SHARAKU (BL17) at the Materials and Life Science Experimental Facility (MLF) of the Japan Proton Accelerator Research Complex (J-PARC) [27]. The proton beam power, which determines the intensity of the pulsed neutron beam generated by the spallation process [28], was 500 kW for this investigation. Neutron reflectivity (R) is calculated as II0, where I is the reflected neutron intensity and I0 is the incident neutron intensity obtained by time-of-flight from a pulsed white neutron source. The time-of-flight, tTOF, of a neutron from the neutron source to the detector is converted to neutron wavelength, λ, using the relationship: λ = htTOFmL, where h is Planck’s constant, m is the mass of a neutron, L is the length between the neutron source and the detector. The reflected angle, θr, is equal to the incident angle, θi, (θ = θi = θr) and these angles are fixed. The momentum transfer perpendicular to the layer, Q, is given by (2)

A high-power LED system (LedHUB, Omicron-Laseage Laserprodukte GmbH, Germany) was used as an excitation light source. It contains four LED modules with different peak wavelengths: 367 nm (FWHM: 9 nm)(3.38 eV), 408 nm (FWHM: 11 nm)(3.04 eV), 441 nm (FWHM: 18 nm)(2.81 eV), and 521 nm (FWHM: 44 nm)(2.38 eV). The amorphous Ge-S had composition Ge20 S80, with a band gap of 2.77 eV according to Pan et al., [29]. Hence the 367 nm and 408 nm LEDs can excite electrons in the valence band to the conduction band. The 441 nm LED has close to the required energy, whereas the 521 nm LED is lacking in excitation energy.

Figure 1 shows the experimental setup for in situ neutron reflectivity measurements under LED light illumination [12,30]. The sample on a sample stage is rotated to fix the incident angle to be θ, and the detector is rotated to fix the reflection angle to be 2θ. The LED light beam uniformly illuminates on the sample with squared shape (27 mm × 27 mm) using a lens unit composed of a rectangular prism and condensed lenses, normal to the sample as shown in the figure.

Data were acquired by an event recording system in use at the MLF. Arbitrary time-sliced TOF spectra are obtained once a complete data set is acquired. The time width of the transient data was determined on the basis of the rate of the reaction as indicated bychanges observed in the data. The neutron reflectivity profiles were fitted using the Motofit package [31].

thumbnail Fig. 1

Experimental setup for in situ neutron reflectivity measurements under LED light illumination.

3 Results and discussion

3.1 Kinetics of silver photodiffusion probed by time-resolved neutron reflectivity

3.1.1 Light excitation with the photon energy smaller than the optical gap (Eex = 2.38 eV)

Figure 2 shows the time evolution of the neutron reflectivity profile of Ag 500 Å/Ge20S80 1500 Å/Si substrate stacks during and after 2.38 eV (521 nm) illumination at intensity 35.7 mW/cm2 for 60 min. The photon energy for the light excitation is smaller than the optical gap of amorphous Ge20 S80 (2.77 eV [29]). In the lower quarter part of the figure, the neutron reflectivity profiles before the light illumination, at 50–60 min after starting the light illumination (just before stopping the light illumination) (55 ± 5 min), and at 10–20 min after stopping the light illumination (at 70–80 min after starting the light illumination) (75.1 ± 5 min) are overlaid for comparison. There is a difference in the neutron reflectivity profile before and after the light illumination, especially in the high Q region from 0.021 to 0.032 (Å−1). However, the difference is very small compared to that for the higher energy light excitation as we will see later. There is no change afterstooping the light illumination.

To find an appropriate model for the SLD depth profile, it is important to survey the changes in neutron reflectivity using model-free analytical technique. Evaluation of the critical edge, Qc, for a total reflection is one of such techniques. Below Qc, R = 1. From the theory of optics, Qc is calculatedby [25,26]: (3)

Using the relationship, Qc of silver is estimated to be 0.0132 Å−1, while Qc of amorphous Ge20 S80 is estimated to be 0.0092 Å−1. In RQ4 -Q plots, the curve follows Q4 in the total reflection region (R = 1), and drops at Qc. Figure 3 is the time evolution of RQ4 -Q plots for the neutron reflectivity profiles in Figure 2. In the figure, two critical edges are observed, P1 and P2. P2 corresponds to the critical edge of the Ag layer, which is on the top of the film. P1 is the critical edge of the reaction layer between the Ag layer and the amorphous Ge20 S80 layer, which was probably formed in the thermal evaporation process. Figure 4 shows the time variations of the positions and heights of P1 and P2. The peak positions and heights were determined by fitting assuming two Gaussian terms because two peaks are close to each other. The positions of P1 and P2 do not change through the light illumination. The height of P1 does not change, too. The height of P2 slightly decreases by the light illumination. These observationsindicate that both the Ag surface layer and the reaction layer are basically preserved through the light illumination.

Figure 5 shows the time evolution of the SLD depth profile of Ag 500 Å/Ge20S80 1500 Å/Si substrate stacks during and after 2.38 eV (521 nm) illumination at an intensity of 35.7 mW/cm2, which were obtained from the fit to the measured neutron reflectivity profiles in Figure 2. In the process of the curve fitting, we assumed that the Ag layer was basically preserved through the light illumination as observed in Figures 3 and 4. In the lower part, three SLD depth profiles, before light illumination, at 55 ± 5 min, and at 75 ± 5 min, are overlaid for comparison. By the 60 min light illumination, the Ag layer is almost maintained, and a reaction layer with step-like Ag distribution is formed. The formation of a reaction layer with step-like Ag distribution is consistent with previous RBS results [6,8]. It is noted that the buried interface underneath the Ag surface layer was found from the measurement. This is the advantage of neutron reflectivity technique. The SLD depth profiles in Figure 5 are actually obtained by giving three parameters, thickness, SLD, and roughness, for each layer. These are the parameters to fit to the neutron reflectivity profiles in Figure 2. Figure 6 shows the time variation of these fitting parameters (thickness, SLD, and roughness). The thicknesses of the Ag and amorphous Ge20 S80 host layer does not change through the light illumination, while the thickness of the reaction layer slightly increases. The SLD of the reaction layer does not change by the light illumination. The SLD of the amorphous Ge20 S80 host layer is almost constant, too. The roughness of the reaction layer decreases just after starting the light illumination and it is kept even after stopping the illumination. Overall, the Ag-doped reaction layer grows by the light illumination, but the reaction rate is very slow.

thumbnail Fig. 2

Time-evolution of the neutron reflectivity profile of Ag 500 Å/Ge20S80 1500 Å/Si substrate stacks during and after 2.38 eV (521 nm) illumination at intensity 35.7 mW/cm2.

thumbnail Fig. 3

Time evolution of RQ4-Q plots for the neutron reflectivity profiles of Ag 500 Å/Ge20S80 1500 Å/Si substrate stacks during and after 2.38 eV (521 nm) illumination at intensity 35.7 mW/cm2, which are shown in Figure 2.

thumbnail Fig. 4

Time variations of the positions and heights of P1 and P2 in Figure 3.

thumbnail Fig. 5

Time evolution of the SLD profile of Ag 500 Å/Ge20S80 1500 Å/Si substrate stacks during and after 2.38 eV (521 nm) illumination at intensity 35.7 mW/cm2, which were obtained from the fit to the measured neutron reflectivity profiles in Figure 2.

thumbnail Fig. 6

Time variation of the fitting parameters (thickness, SLD, and roughness) used for the fit to the measured neutron reflectivity profiles in Figure 2.

3.1.2 Light excitation with the photon energy comparable to the optical gap (Eex = 2.81 eV)

Figure 7 shows time evolution of the neutron reflectivity profile of Ag 500 Å/Ge20S80 1500 Å/Si substrate stacks during and after 2.81 eV (441 nm) illumination at intensity 35.7 mW/cm2 for 60 min. The photon energy for the light excitation is comparable to the optical gap of amorphous Ge20 S80 (2.77 eV). In the lower part of the figure, the neutron reflectivity profiles before the light illumination, at 50–60 min, and at 75 ± 5 min are overlaid for comparison. There is obvious change in the neutron reflectivity profile by the 60 min light illumination, and the change is preserved after stopping the light illumination.

Figure 8 shows the time evolution of RQ4 -Q for 2.81 eV illumination. Figure 9 shows the time variations of the positions and heights of P1 and P2, which are indicated in Figure 8. The height of P2obviously decreases through the light illumination. The position of P2 decreases and that of P1 increases for the first 30 min lightillumination. The height of P1 increases for the first 30 min, too. These observations suggest that the Ag layer and the reaction layer are basically preserved with some changes in the thickness or SLD, by the light illumination.

Figure 10 shows the time evolution of the SLD depth profile, of Ag 500 Å/Ge20S80 1500 Å/Si substrate stacks during and after 2.81 eV (441 nm) illumination at intensity 35.7 mW/cm2, which were obtained from the fit to the measured neutron reflectivity profiles in Figure 7. In the lower quarter part, three SLD depth profiles, before light illumination, at 55 ± 5 min, and at 75 ± 5 min, are overlaid for comparison as well. Two SLD depth profiles are plotted for the time at 75.1 ± 5 min. One assumes that the Ag layer is intact, i.e., the SLD of the Ag layer does not change (75.1 ± 5 min (an intact Ag)). The other assumes that the Ag layer is partially reacted, and the SLD of the Ag layer decreases (75.1 ± 5 min). The neutron reflectivity profiles calculated from these SLD depth profiles are shown in Figure 7 by a sky blue curve and red curve, respectively, on the measured neutron reflectivity data at 75.1 ± 5 min. As clearly seen in the figure, the latter assumption, which is a partial reaction of the Ag layer, looks better. Therefore, we assume thatthe Ag layer is partially reacted by the light illumination.

Figure 11 shows the time variation of the fitting parameters (thickness, SLD, and roughness) used for the fit to the measured neutron reflectivity profiles in Figure 7. Although the thickness of the Ag layer does not change so much, the thickness of the Ag-doped reaction layer increases and the thickness of the amorphous Ge20S80 host layer decrease for the first 30 min of light illumination. The former change is consistent with the changes of P1 in Figure 9 and indicates a growth of the Ag-doped reaction layer. The latter indicates that a part of the chalcogenide layer is affected by the Ag diffusion, forming a Ag-doped reaction layer, and that the non-affected region shrinks. The SLD of the Ag layer decreases and the roughness of the Ag layer increases through the light illumination. These changes could be related to the decrease of the P2 height in Figure 9. Since the thickness of the Ag layer does not change so much, the decrease of the SLD and the increase of the roughness would suggest either a partial loss of Ag fragments or a partial reaction in the Ag layer. Considering that the Ag-doped reaction layer grows and the Ag thickness is almost constant, the Ag ions seem to be supplied from randomly distributed Ag fragments in the Ag layer without changing its thickness. However, it is natural to assume that the Ag dissolves at the interface (boundary) and difficult to assume that Ag fragments are randomly taken from the inside of the layer. Therefore, we assume that the Ag layer is partially affected by the chalcogenide, forming dendrites patterns from some contact points at the interface to the inside of the Ag layer, for example1. The Ag layer could be thicker due to the partial reaction, and the actual loss of the Ag layer due to the Ag dissolution could be much larger than what the experimental result in Figure 11 suggests.

thumbnail Fig. 7

Time evolution of the neutron reflectivity profile of Ag 500 Å/Ge20S80 1500 Å/Si substrate stacks during and after 2.81 eV (441 nm) illumination at intensity 35.7 mW/cm2 for 60 min.

thumbnail Fig. 8

Time evolution of RQ4-Q plots for the neutron reflectivity profiles of Ag 500 Å/Ge20S80 1500 Å/Si substrate stacks during and after 2.81 eV (441 nm) illumination at intensity 35.7 mW/cm2, which are shown in Figure 7.

thumbnail Fig. 9

Time variations of the positions and heights of P1 and P2 in Figure 8.

thumbnail Fig. 10

Time evolution of the SLD depth profile of Ag 500 Å/Ge20S80 1500 Å/Si substrate stacks during and after 2.81 eV (441 nm) illumination at intensity 35.7 mW/cm2, which were obtained from the fit to the measured neutron reflectivity profiles in Figure 7.

thumbnail Fig. 11

Time variation of the fitting parameters (thickness, SLD, and roughness) used for the fit to the measured neutron reflectivity profiles in Figure 7.

3.1.3 Light excitation with the photon energy greater than the optical gap (Eex = 3.04 eV)

Figure 12 shows the time evolution of the neutron reflectivity profile of Ag 500 Å/Ge20S80 1500 Å/Si substrate stacks during and after 3.04 eV (408 nm) illumination at intensity 34.4 mW/cm2 for 60 min. The photon energy for the light excitation is greater than the optical gap of amorphous Ge20S80 (2.77 eV). In the lower part of the figure, the neutron reflectivity profiles before the light illumination, at 50–60 min, and at 75 ± 5 min are overlaid for comparison. The neutron reflectivity profile with the light illumination is quite different from that before the light illumination, and there is no change after stopping the light illumination.

Figure 13 shows the time evolution of RQ4-Q plots for 3.04 eV illumination at intensity 34.4 mW/cm2 for 60 min. Figure 14 shows the time variations of the positions and heights of P1 and P2 in Figure 13. For the light excitation, the position andheight of P2 decreases with illumination time and the peak disappears approximately at 12 min. The disappearance of P2 indicates the disappearance of the Ag layer by full Ag dissolution. The position and height of P1 increase until 12 min, but they turn todecrease after 12 min. The increase of the P1 peak position in the first stage suggests the increase of the SLD of the reaction layer due to the introduction of Ag ions, while the decrease of the P1 peak position in the next stage suggests the decrease of the SLD of the reaction layer due to the movement of Ag ions from the reaction layer to the chalcogenide host layer.

An appropriate SLD depth profile was found from a fit to the measured neutron reflectivity profile by assuming above changes. Figure 15 shows the time evolution of the SLD depth profile for 3.04 eV (408 nm) illumination at intensity 34.4 mW/cm2 for 60 min. In the lower part of the figure, the SLD depth profiles before the light illumination, at 55 ± 5 min, and 75.1 ± 5 min are overlaid for comparison. Obviously, the sample changes to one uniform reaction layer by the light illumination. In other words, silverphotodiffusion is completed by the light illumination. There are following five processes in the time evolution. (1) The Ag layer becomes thinner and thinner with light illumination time, and disappears until 5 min. (2) The Ag-rich reaction layer is formed on the film instead of the Ag layer. (3) Ag ions move from the Ag-rich reaction layer to the Ge-S host layer, resulting in a decrease of the SLD of the reaction layer and an increase of the SLD of the Ge-S host layer. (4) The reaction layer and the Ge-S host layer merge to be one uniform layer. (5) The surface of the merged reaction layer is affected, probably by oxygen in air, by extended light illumination. The processes from (1) to (4) are the same as those observed in our previous neutron reflectivity measurement of Ag/amorphous Ge20S80/Si substrate stacks under white light illumination from a Xe lamp [17].

Figure 16 shows the time variation of the fitting parameters (thickness, SLD, and roughness) used for the fit to the measured neutron reflectivity profiles in Figure 12. The thickness of the Ag layer decreases with light illumination time and disappears at 12 min light illumination. The roughness of the Ag layer increases until 8 min. There is a delay in the disappearance time from the result in Figure 15, which indicates about 5 min. This is because of alarge roughness compared to a very thin Ag layer. We assume that thin Ag layer exists until 12 min. This is consistent with the expectation of the result in Figure 14. The changes of the SLDs of the reaction and Ge-S host layers clearly show that both layers merge to one uniform reaction layer. Through the light illumination, the surface layer changes according to the photo-reaction: the Ag layer, the Ag-rich reaction layer, and the merged uniform reaction layer. Although the surface layer changes, the roughness is almost maintained. It is noted that the surface becomes bare for air when a protection Ag layer is exhausted by silver photodiffusion. In fact, a new surface layer is produced after a merged uniform reaction layer is formed. The formation of the new surface layer is attributed to photo-oxidation, in which the surface layer reacts with oxygen in air by extended light illumination. It is well-known that photo-oxidation occurs in amorphous germanium sulfide [32] and such oxidation can also occur in Ag-Ge-S glass network system. We infer that the Ge-S bond or Ge-Ag bond is broken by the light illumination, and active Ge dangling bonds in amorphous Ag-Ge-S layer capture oxygen atoms in air.

The composition of the Ag-rich reaction layer is estimated to be Ag0.66(Ge0.2S0.8)0.34, assuming that Ag ions in the initial Ag layer (380 Å) enter the amorphous Ge20S80 layer with the thickness of 300 Å (the initial thickness (1200 Å) - the thickness of the chalcogenide host layer in the second Ag diffusion process (900 Å)). On the other hand, the composition of the merged uniform Ag reaction layer is estimated to be Ag0.33(Ge0.2S0.8)0.67, assuming that Ag ions in the initial Ag layer (380 Å) enter the amorphous Ge20S80 layer with the initial thickness of 1200 Å. These estimated values are the same as those we obtained for the Ag/amorphous Ge20S80/Si substrate stacks under white light illumination from a Xe lamp [17]. From these results, we infer that Ag8 GeS6 (argyrodite) (=4Ag2S + GeS2) type structure [33] is formed in the Ag-rich reaction layer and fully Ag-doped Ag-Ge-S ternary glass with the Ag content of approximately 30% [34] is formed in the merged uniform reaction layer. Ag ions are supposed to be captured by both broken S-S and Ge-S bonds in the former one, while Ag ions are supposed to be captured by only broken Ge-S bonds in the latter one [17].

thumbnail Fig. 12

Time evolution of the neutron reflectivity profile of Ag 500 Å/Ge20S80 1500 Å/Si substrate stacks during and after 3.04 eV (408 nm) illumination at intensity 34.4 mW/cm2.

thumbnail Fig. 13

Time evolution of RQ4-Q plots for the neutron reflectivity profiles of Ag 500 Å/Ge20S80 1500 Å/Si substrate stacks during and after 3.04 eV (408 nm) illumination at intensity 34.4 mW/cm2 for 60 min, which are shown in Figure 12.

thumbnail Fig. 14

Time variations of the positions and heights of P1 and P2 in Figure 13.

thumbnail Fig. 15

Time evolution of the SLD profile of Ag 500 Å/Ge20S80 1500 Å/Si substrate stacks during and after 3.04 eV (408 nm) illumination at intensity 34.4 mW/cm2.

thumbnail Fig. 16

Time variation of the fitting parameters (thickness, SLD, and roughness) used for the fit to the measured neutron reflectivity profiles in Figure 12.

3.1.4 Light excitation with the photon energy much greater than the optical gap (Eex = 3.38 eV)

Figure 17 shows the time evolution of the neutron reflectivity profile of Ag 500 Å/Ge20S80 1500 Å/Si substrate stacks during and after 3.38 eV (367 nm) illumination at intensity 30.6 mW/cm2 for 60 min. In the lower part of the figure, three neutron reflectivity profiles before the light illumination, at 55 ± 5 min, and at 75 ± 5 min are overlaid for comparison. The neutron reflectivity profile after the 60 min light illumination is completely different from that before the light illumination. The tendency is the same as that with 3.04 eV illumination.

Figure 18 shows the time evolution of RQ4-Q plots for 3.38 eV illumination at intensity 30.6 mW/cm2 for 60 min. Both P1 and P2 shift to the lower Q-side, decreasing their height with light illumination time. Figure 19 shows the time variation of the positions and heights of P1 and P2 in Figure 18. The position and height of P2 rapidly decrease with light illumination time and the peak disappears approximately at 3 min. This indicates that the Ag layer rapidly dissolves into the chalcogenide layer, and disappears until 3 min. The position and height of P1 increase until 3 min, and then, turn to decrease until 30 min. This indicates that the Ag concentration in the reaction layer increases until 3 min, and then, it decreases until 30 min.

An appropriate SLD depth profile was found by a fit to the measured neutron reflectivity profile, assuming above changes. Figure 20 shows the time evolution of the SLD profile for 3.38 eV illumination at intensity 30.6 mW/cm2 for 60 min. In the lower part of the figure, three SLD depth profiles before the light illumination, at 55 ± 5 min, and at 75 ± 5 min. The SLD depth profiles after the light illumination are completely different from that before the light illumination, and the profiles indicate that the Ag layer fully dissolves and one uniform reaction layer is formed as observed for3.04 eV illumination. The thickness of the uniform reaction layer is much thinner than that of the initial Ag/a-Ge20S80 bi-layer. This volume contraction is supposed to occur by the involvement of Ag ions in the Ge-S glass network.

Figure 21 shows the time variation of the fitting parameters (thickness, SLD, and roughness) used for the fit to the measured neutron reflectivity profiles in Figure 17. The thickness of the Ag layer rapidly decreases with light illumination time, and disappears until 3 min. This is consistent with the expectation from the result in Figure 19. The increase of the dissolution rate, compared to the case of 3.04 illumination, is attributed to the increase of the electron transition probability, which is determined by the electronic density of states in both initial and final states. As a schematic band structure around the band gap in amorphous semiconductors clearly shows [3538], the density of states becomes larger with increasing excitation photon energy. Therefore, it is reasonable that the dissolution rate increases with higher excitation photon energy. The SLD of the reaction layer suddenly increases just after starting the light illumination, but, it turns to decrease with light illumination time. On the other hand, the SLD of the Ge-S host layer increases with light illumination time. Approximately at 30 min, both SLD become the same, and two layers merge to one uniform reaction layer. The rate of this second diffusion is also faster than that for 3.04 eV illumination. After merging two layers, a new surface layer, resulting from a reaction with oxygen in air, is formed by extended light illumination, as observed for 3.04 eV illumination.

thumbnail Fig. 17

Time evolution of the neutron reflectivity profile of Ag 500 Å/Ge20S80 1500 Å/Si substrate stacks during and after 3.38 eV (367 nm) illumination at intensity 30.6 mW/cm2 for 60 min.

thumbnail Fig. 18

Time evolution of RQ4-Q plots for 3.38eV illumination at intensity 30.6 mW/cm2 for 60 min.

thumbnail Fig. 19

Time variation of the positions and heights of P1 and P2 in Figure 18.

thumbnail Fig. 20

Time evolution of the SLD profile for 3.38 eV illumination at intensity 30.6 mW/cm2 for 60 min.

thumbnail Fig. 21

Time variation of the fitting parameters (thickness, SLD, and roughness) used for the fit to the measured neutron reflectivity profiles in Figure 17.

3.2 Mechanism of silver photodiffusion

3.2.1 Silver dissolution and diffusion

Figure 22 shows the excitation photon energy dependence of the Ag layer thickness after 60 min light illumination. The thickness abruptly drops around the optical gap of amorphous Ge20S80 (2.77 eV [29]). The decrease of the thickness of the Ag layer indicates how much silver dissolves into the chalcogenide layer. The result of Figure 22 shows that there is a threshold photon energy requirement for silver dissolution into chalcogenide, and that the threshold is the optical gap of the amorphous chalcogenide layer. This is consistent with the previous reports on the excitation photon energy dependence of silver dissolution by Kokado et al. [39] and Goldschmidt and Rudman [40] on Ag/As-S system. These results indicate that the electron excitation from the valence band to the conduction band in amorphous chalcogenides plays an important role in silver dissolution.

For Ge-S glasses, the lone-pair band forms the top of the valence band, and the anti-bonding band forms the bottom of the conduction band [4143]. Light illumination with greater photon energy than the optical gap excites electrons from the lone-pair band to the anti-bonding band. This process is common to amorphous and liquid chalcogenides, including elemental chalcogens.

According to ab initio molecular dynamics simulations by Shimojo et al. [44], the S-S bond in an isolated S8 ring is immediatelybroken by an electron excitation from HOMO to LUMO, i.e., by the excitation of lone-pair electrons to the anti-bonding band. The same S-S bond breaking occurs in liquid sulphur according to ab initio molecular dynamics simulations by Munejiri et al. [45]. We assume that S-S or Ge-S bond breaking occurs by the excitation of lone-pair electrons to the anti-bonding band in amorphous Ge20S80. Considering the energy splitting between the boding state and anti-bonding state, the required photon energy to excite to the S-S anti-bonding states is supposed to be greater than that to excite to the Ge-S anti-bonding states [43]. The broken S atoms can capture Ag ions, thereby becoming part of the Ge-S network. This process is just the same as an intercalation model of silver photodiffusion, proposed by Kluge [18]. The guests of the Ag ions are intercalated in the chalcogenide host. It is plausible to assume suchre-organisation of the network structure by the introduction of Ag ions as the first-principles molecular-dynamics simulations on Ag-containing chalcogenide glasses demonstrate the network structure composed of Ag, Ge, and S(Se) [46,47].

The result in Figures 15 and Figure 20 (or Figs. 16 and 21) suggest that silver photodiffusion takes place with the following three steps. (1) Ag ions move from the Ag layer to the Ag-doped reaction layer across the Ag layer/Ag-doped reaction layer interface, forming an Ag-rich reaction layer. (2) The Ag-rich reaction layer grows increasing its thickness with light illumination time as if the “reaction front” advances to the chalcogenide side. (3) The next Ag diffusion occurs from the Ag-rich (Ag-doped) reaction layer to the chalcogenide host layer across the Ag-doped reaction layer/chalcogenide host layer interface. Overall, there can be two potential barriers. One is lower and is at the Ag layer/Ag-doped reaction layer interface. The other is higher and is at the Ag-doped reaction layer/chalcogenide host layer interface.

These observations may be explained using the intercalation model as follows. (1) By the light illumination with the photon energy greater than the optical gap of amorphous Ge20S80, Ge-S or S-S bonds are broken in the Ag-doped reaction layer in the vicinity of the Ag/Ag-doped reaction interface. The optical gap of the Ag-doped reaction layer is smaller than that of the non-doped layer according to the optical transmission measurements by Kawaguchi et al. [34]. The optical gaps of amorphous Agx (Ge30S70)100−x are 3.40 (x = 0), 3.2 (x = 30), and 2.4 (x = 68), with higher Ge content than Ge20S80. However, the decrease of the optical gap by the Ag addition is attributed to the lowering of the anti-bonding states of weaker Ag-Ge and Ag-S bonds, and the energy levels of the Ge-S and S-S anti-bonding states are supposed to be unchanged by the Ag addition. Therefore, we assume that only the electron excitation to the Ge-S or S-S anti-bonding states, which requires greater photon energy than the optical gap of amorphous Ge20S80, is available to promote the intercalation reaction, where broken S atoms capture Ag ions. (2) By the light excitation, Ge-S or S-S bonds in “both” Ag-doped reaction and chalcogenide host layers are expected to break. However, we assume that the capturing of Ag ions takes place only at the interface where both layers contact to each other. Under the assumption, the reaction rate is higher at the Ag layer/Ag-doped reaction layer, where the concentration of Ag ions is 100%. This could be the reason why the silver diffusion across the Ag layer/Ag-doped reaction layer interface is faster than that across the Ag-doped reaction layer/chalcogenide host layer interface. In the Ag-doped reaction layer, Ag ions are supplied from the Ag layer and the Ag ions are captured by both broken Ge-S and S-S bonds until the composition reaches Ag0.66(Ge0.2S0.8)0.34. At the Ag-doped reaction layer/chalcogenide host layer interface, the reaction occurs mainly at the interface, extending its region. As a result, the “reaction front” (boundary) of the Ag-rich reaction layer advances to the chalcogenide host layer side with light illumination time. (3) By exhausting silver in the Ag layer, the content of Ag ions in the Ag-doped reaction layer saturates. Even after the saturation, the light continues to affect the Ag-doped reaction layer/chalcogenide host layer interface, leading to the bond breaking in both Ag-doped reaction layer and chalcogenide host layer. We assume here that the broken chalcogen sites can propagate, by successively giving the broken sites to neighbouring chalcogen atoms. This can be realized by bond-switching in chalcogenide glasses [3,4,48]. The shifted broken chalcogen sites will capture neighbouring Ag ions again. This can cause a simultaneous Ag diffusion over the entire chalcogenide host layer. The reaction product is stable Ag containing Ag-Ge-S ternary glass with Ag content of approximately 30%, where Ag ions are supposed to be captured by broken Ge-S bonds in the Ge-S network (Ag0.33(Ge0.2S0.8)0.67).

Kluge points out that the intercalation reaction needs a supply of electrons and ions from outside, and that they are supplied from photo-generated holes at the Ag/Ag-doped chalcogenide junction [18]. Elliott, Tanaka, and Aniya suggest that the holes generated at the Ag-doped layer/undoped chalcogenide layer interface flow to the Ag layer/the Ag-doped layer interface through the Ag-doped layer [1922]. It was found from the present study that photon energy greater than the optical gap of the chalcogenide layer was required to promote silver dissolution. Therefore, the holes are considered to be generated in the chalcogenide layer, and the hole-generation could be a driving force to produce a supply of Ag ions as Elliott, Tanaka, and Aniya suggested. However, it may be noted that it is difficult to explain two types of Ag diffusion with Ag ions localised in the Ag-doped reaction layer and diffusive Ag ions over the entire chalcogenide layer, by the hole-flow model, while our model given above can explain the behaviour.

thumbnail Fig. 22

Excitation photon energy dependence of the Ag thickness after 60 min light illumination. The initial Ag thickness before light illumination is indicated at E = 0 eV. The optical gap energy of amorphous Ge20S80 measured by Pan et al. (2.77 eV [29]) is indicated by a broken line.

3.2.2 Formation of Ag-doped reaction layer with step-like Ag distribution by the light excitation with EexEg

The Ag-doped reaction layer with step-like Ag distribution was formed by 2.38 and 2.81 eV illumination, although the Ag layer was hardly dissolved by the light excitation. The mechanism of this type of silver diffusion should be different from that of the full silver dissolution observed for 3.04 and 3.38 eV illumination.

The SLD of the Ag-doped reaction layer produced by 2.38 and 2.81 eV illumination is 2.1–2.5 × 10−6 (Å−2), whereas that produced by 3.04 and 3.38 eV illumination is 3.0–3.5 × 10−6 (Å−2). We assume that the former one is composed of Ag-Ge-S network glasses in which Ag ions are intercalated in Ge-S bonds, while the latter one is composed of Ag-Ge-S network glasses in which Ag ions are intercalated in both Ge-S and S-S bonds. This makes sense because 2.38 and 2.81 eV illumination can break only Ge-S bonds, whose energy splitting is smaller, while 3.04 and 3.38 eV illumination can break both Ge-S bond and S-S bond, whose energy splitting is greater. From the values of the SLD, the Ag-doped reaction layer produced by 2.38 and 2.81 eV illumination is supposed to be the same product as the final merged reaction layer produced by 3.04 and 3.38 eV illumination, with the Ag content of approximately 30% (Ag0.33(Ge0.2S0.8)0.67).

More details on the electron excitation may be required to understand the Ag-doping reaction. In amorphous semiconductors, the band edges tail into the band gap. The states in the band tail are localised [3538,49]. By 2.38 and 2.81 eV illumination, the electron excitation occurs between the band tail and main band, or both band tails. The density of states in the band tails are considerably small compared to those in the main bands. Thus, the transition probability for 2.38 and 2.81 eV illumination is considered to be very small. This causes a slow reaction rate, observed in Figures 6 and 11. Assuming that the reaction occurs only at the interface, where the probability of the collision between two reactants (Ag and S) is high, as well as the case for 3.04 and 3.38 illumination, the reaction rate at the Ag/Ag-doped interface would considerably be small due to the small electron transition probability in the Ag-doped reaction layer. This would be the reason why the Ag layer is hardly dissolved by the light illumination. On the other hand, the reaction rate at the Ag-doped/chalcogenide host interface could be better, because the bonds are broken at both layers; Ag-Ge and Ag-S bonds in the Ag-doped reaction layer (Ag supply) and the Ge-S bonds in the chalcogenide host layer (Ag acceptor). This would be the reason why the Ag-doped reaction layer grows with the light illumination. In addition, it may be noted that the initial/final states of the electron transition are the localised states. This means that the excited electrons or created holes are localised, and that the effect of the excitation, i.e. bond-breaking, is also localised. This can limit the reaction region near the Ag-doped reaction layer/chalcogenide host layer interface, and may help to form an Ag-doped reaction layer with step-like Ag distribution, which extends its region with light illumination time.

3.2.3 Further photo-induced structural change by extended light excitation

It is also interesting to study the effect of light illumination on Ag/amorphous chalcogenide stacks after full silver photodiffusion. In fact, we observed a macroscopic surface deformation with wavy-like patterns by extended light illumination to Ge33S67/Ag/Si substrate stacks [14]. In the present study, we observed a further structural change on the surface by extended light illumination, probably due to the photo-oxidation, which occurs by the reaction with oxygen in air after disappearance of the protecting Ag layer. We infer that Ge-S or Ag-Ge bonds in the Ag-Ge-S layer are broken by the extended light illumination. Then, broken Ge atoms would capture oxygen atoms in air, forming a new network composed of Ge, S, Ag, and O.

4 Conclusion

We investigated the photon energy dependence of silver photodiffusion into amorphous Ge20S80 using neutron reflectivity as a spectroscopic technique. It was found from the measurements that Ag photodiffusion effectively occurred by light illumination with photon energy greater than the optical gap of the chalcogenide host material. This indicates that the excitation of lone-pair electrons to anti-bonding states at the chalcogen (sulphur) sites plays a crucial role in the Ag photodiffusion. The excitation of lone-pair electrons will lead to rearrangement of the network in the glasses, including the Ag ions. Ag photodiffusion occurs in two steps. (1) the Ag layer dissolves and the Ag-rich reaction layer is formed. (2) Next Ag diffusion relocates from the Ag-rich reaction layer to the chalcogenide host layer, until the two layers merge. Such a variety of changes is realised by photo-induced flexibility in lone-pair semiconductors. We also observed the formation of an Ag-doped reaction layer with step-like Ag distribution, almost maintaining the Ag layer, by light illumination with photon energy comparable to or smaller than the optical gap of the chalcogenide host material. This is a different type of Ag photodiffusion from the full Ag photodiffusion, where the electron excitation process with the photon energy is substantially affected.

The neutron reflectivity measurements were performed on SHARAKU at J-PARC MLF under Project 2019A0196. We would like to thank S. Kasai and K. Akutsu (CROSS) for technical support on the neutron reflectivity measurement. This research was partially supported by the NASA EPSCoR grant number 80NSSC17M0029. We would like to thank Editage (www.editage.com) for English language editing.

Author contribution statement

A. A. Simon and M. Mitkova prepared the samples. Y. Sakaguchi and T. Hanashima performed neutron reflectivity measurements. Y. Sakaguchi performed the analysis of the neutron reflectivity. Manuscript was prepared by Y. Sakaguchi, and revised by all other authors, especially M. Mitkova.

References

  1. M.T. Kostyshin, E.V. Mikhailovskaya, P.F. Romanenko, Fiz. Tverdogo Tela 8, 571 (1966) [Sov. Phys. (Solid State) 8, 451 (1966)] [Google Scholar]
  2. A.V. Kolobov, S.R. Elliott, Adv. Phys 40, 625 (1991) [Google Scholar]
  3. A.V. Kolobov, Photo-induced Metastability in Amorphous Semiconductors (Wiley-VCH GmbH and Co.KGaA, Weinheim, 2003) [CrossRef] [Google Scholar]
  4. A.V. Kolobov, J. Tominaga, Chalcogenides Metastability and Phase Change Phenomena (Springer, Heidelberg, 2012) [CrossRef] [Google Scholar]
  5. M. Mitkova, M.N. Kozicki, J. Non-Cryst. Solids 299-302, 1023 (2002) [CrossRef] [Google Scholar]
  6. Y. Yamamoto, T. Itoh, Y. Hirose, H. Hirose, J. Appl. Phys. 47, 3603 (1976) [Google Scholar]
  7. J. Rennie, S.R. Elliott, C. Jeynes, Appl. Phys. Lett. 48, 1430 (1986) [Google Scholar]
  8. T. Wagner, V. Perina, M. Vlcek, M. Frumer, E. Rauhala, J. Saarilahti, P.J.S. Ewen, J. Non-Cryst. Solids 212, 157 (1997) [CrossRef] [Google Scholar]
  9. T. Wagner, G. Dale, P.J.S. Ewen, E. Owen, V. Perina, J. Appl. Phys. 87, 7758 (2000) [Google Scholar]
  10. Y. Sakaguchi, H. Asaoka, Y. Uozumi, Y. Kawakita, T. Ito, M. Kubota, D. Yamazaki, S.K.M. Ailavajhala, M.R. Latif, M. Mitkova, Can. J. Phys. 92, 654 (2014) [Google Scholar]
  11. Y. Sakaguchi, H. Asaoka, Y. Uozumi, Y. Kawakita, T. Ito, M. Kubota, D. Yamazaki, K. Soyama, M. Ailavajhala, M.R. Latif, K. Wolf, M. Mitkova, W.A. Skoda, J. Phys.: Conf. Ser. 619, 012046 (2015) [CrossRef] [Google Scholar]
  12. Y. Sakaguchi, H. Asaoka, Y. Uozumi, Y. Kawakita, T. Ito, M. Kubota, D. Yamazaki, K. Soyama, M. Ailavajhala, K. Wolf, M. Mitkova, M.W.A. Skoda, JPS Conf. Proc. 8, 031023 (2015) [Google Scholar]
  13. Y. Sakaguchi, H. Asaoka, Y. Uozumi, Y. Kawakita, T. Ito, M. Kubota, D. Yamazaki, K. Soyama, G. Sheoran, M. Mitkova, Phys. Status Solidi A 213, 1894 (2016) [CrossRef] [Google Scholar]
  14. Y. Sakaguchi, H. Asaoka, Y. Uozumi, K. Kondo, D. Yamazaki, K. Soyama, M. Ailavajhala, M. Mitkova, J. Appl. Phys. 120, 055103 (2016) [Google Scholar]
  15. Y. Sakaguchi, H. Asaoka, M. Mitkova, J. Appl. Phys. 122, 235105 (2017) [Google Scholar]
  16. Y. Sakaguchi, T. Hanashima, H. Aoki, H. Asaoka, A.-A.A. Simon, M. Mitkova, Phys. Status Solidi A 215, 1800049 (2018) [CrossRef] [Google Scholar]
  17. Y. Sakaguchi, H. Asaoka, M. Mitkova, Pure Appl. Chem. 91, 1821 (2019) [Google Scholar]
  18. G. Kluge, Phys. Status Solidi A 101, 105 (1987) [CrossRef] [Google Scholar]
  19. S.R. Elliott, J. Non-Cryst. Solids 130, 85 (1991) [CrossRef] [Google Scholar]
  20. Ke. Tanaka, J. Non-Cryst. Solids 137-138, 1021 (1991) [CrossRef] [Google Scholar]
  21. Ke. Tanaka, J. Appl. Phys. 70, 7397 (1991) [Google Scholar]
  22. M. Aniya, J. Non-Cryst. Solids 198-200, 762 (1996) [CrossRef] [Google Scholar]
  23. S. Bhosle, K. Gunasekera, P. Chen, P. Boolchand, M. Micoulaut, C. Massobrio, Solid State Commun. 151, 1851 (2011) [Google Scholar]
  24. S. Ravindren, K. Gunasekera, Z. Tucker, A. Diebold, P. Boolchand, M. Micoulaut, J. Chem. Phys. 140, 134501 (2014) [Google Scholar]
  25. J. Daillant, A. Gibaud, in X-ray and Neutron Reflectivity: Principles and Applications, Lecture Notes in Physics (Springer, Berlin, 2009), p. 770 [Google Scholar]
  26. D.S. Sivia, in Elementary Scattering Theory (Oxford University Press, Oxford, 2011), p. 93 [CrossRef] [Google Scholar]
  27. M. Takeda, D. Yamazaki, K. Soyama, R. Maruyama, H. Hayashida, H. Asaoka, T. Yamazaki, M. Kubota, K. Aizawa, M. Arai, Y. Inamura, T. Itoh, K. Kaneko, T. Nakamura, T. Nakatani, K. Oikawa, T. Ohhara, Y. Sakaguchi, K. Sakasai, T. Shinohara, J. Suzuki, K. Suzuya, I. Tamura, K. Toh, H. Yamagishi, N. Yoshida, T. Hirano, Chin. J. Phys. 50, 161 (2012) [Google Scholar]
  28. B.T.M. Wills, C.J. Carlile, in Experimental Neutron Scattering (Oxford University Press, Oxford, 2009), p. 43 [Google Scholar]
  29. R.K. Pan, H.Z. Tao, H.C. Zang, C.G. Lin, T.J. Zhang, X.J. Zhao, J. Non-Cryst. Solids 357, 2358 (2011) [Google Scholar]
  30. Y. Sakaguchi, R. Takahashi, S. Kasai, K. Ohuchi, T. Morikawa, J. Neutron Res. 21, 23 (2019) [CrossRef] [Google Scholar]
  31. A. Nelson, J. Appl. Crystallogr. 39, 273 (2006) [Google Scholar]
  32. Ke. Tanaka, Y. Kasanuki, A. Odajima, Thin Solid Films 117, 251 (1984) [Google Scholar]
  33. G. Eulenberger, Monatsh. Chem. 108, 901 (1977) [Google Scholar]
  34. T. Kawaguchi, S. Maruno, S.R. Elliott, J. Appl. Phys. 79, 9096 (1996) [Google Scholar]
  35. E.A. Davis, N.F. Mott, Philos. Mag. 22, 903 (1970) [CrossRef] [Google Scholar]
  36. J. Tauc, in Amorphous and Liquid Semiconductors (Plenum Press, London, 1974), p. 173 [Google Scholar]
  37. N.F. Mott, E.A. Davis, in Electronic Processes in Non-Crystalline Materials (Clarendon Press, Oxford, 1979), p. 458 [Google Scholar]
  38. S.R. Elliott, in Physics of Amorphous Materials, Second edition (Longman, Essex, 1990), p. 311 [Google Scholar]
  39. H. Kokado, I. Shimizu, E. Inoue, J. Non-Cryst. Solids 20, 131 (1976) [CrossRef] [Google Scholar]
  40. D. Goldschmidt, P.S. Rudman, J. Non-Cryst. Solids 22, 229 (1976) [CrossRef] [Google Scholar]
  41. J. Robertson, Adv. Phys. 32, 361 (1983) [Google Scholar]
  42. K. Hachiya, J. Non-Cryst. Solids 312–314, 566 (2002) [Google Scholar]
  43. Y. Sakaguchi, T. Hanashima, K. Ohara, Al-Amin A. Simon, M. Mitkova, Phys. Rev. Mater. 3, 035601 (2019) [Google Scholar]
  44. F. Shimojo, K. Hoshino, Y. Zempo, J. Phys.: Condens. Matter 10, L177 (1998) [CrossRef] [Google Scholar]
  45. S. Munejiri, F. Shimojo, K. Hoshino, J. Phys.: Condens. Matter 12, 7999 (2000) [CrossRef] [Google Scholar]
  46. J. Akola, B. Beuneu, R.O. Jones, P. Jóvári, I. Kaban, J. Kolář, I. Voleská, T. Wágner, J. Phys.: Condens. Matter 27, 485304 (2015) [CrossRef] [Google Scholar]
  47. D. Igram, H.E. Castillo, D.A. Drabold, J. Non-Cryst. Solids 514, 1 (2019) [CrossRef] [Google Scholar]
  48. Y. Sakaguchi, K. Tamura, Z. Phys. Chem., https://doi.org/10.1515/zpch-2020-1650 [Google Scholar]
  49. N.E. Cusack, in The Physics of Structurally Disordered Matter: An Introduction (IOP Publishing, Bristol, 1987), p. 296 [Google Scholar]

1

Elliott considered such growth of the reaction product in the chalcogenide layer as a reaction in the induction period, which is sometimes observed before the onset of fast silver photodissolution [19].

Cite this article as: Yoshifumi Sakaguchi, Takayasu Hanashima, Al-Amin Ahmed Simon, Maria Mitkova, Bond graph multiphysics modeling of encapsulating materials in power electronic modules, Eur. Phys. J. Appl. Phys. 90, 30101 (2020)

All Figures

thumbnail Fig. 1

Experimental setup for in situ neutron reflectivity measurements under LED light illumination.

In the text
thumbnail Fig. 2

Time-evolution of the neutron reflectivity profile of Ag 500 Å/Ge20S80 1500 Å/Si substrate stacks during and after 2.38 eV (521 nm) illumination at intensity 35.7 mW/cm2.

In the text
thumbnail Fig. 3

Time evolution of RQ4-Q plots for the neutron reflectivity profiles of Ag 500 Å/Ge20S80 1500 Å/Si substrate stacks during and after 2.38 eV (521 nm) illumination at intensity 35.7 mW/cm2, which are shown in Figure 2.

In the text
thumbnail Fig. 4

Time variations of the positions and heights of P1 and P2 in Figure 3.

In the text
thumbnail Fig. 5

Time evolution of the SLD profile of Ag 500 Å/Ge20S80 1500 Å/Si substrate stacks during and after 2.38 eV (521 nm) illumination at intensity 35.7 mW/cm2, which were obtained from the fit to the measured neutron reflectivity profiles in Figure 2.

In the text
thumbnail Fig. 6

Time variation of the fitting parameters (thickness, SLD, and roughness) used for the fit to the measured neutron reflectivity profiles in Figure 2.

In the text
thumbnail Fig. 7

Time evolution of the neutron reflectivity profile of Ag 500 Å/Ge20S80 1500 Å/Si substrate stacks during and after 2.81 eV (441 nm) illumination at intensity 35.7 mW/cm2 for 60 min.

In the text
thumbnail Fig. 8

Time evolution of RQ4-Q plots for the neutron reflectivity profiles of Ag 500 Å/Ge20S80 1500 Å/Si substrate stacks during and after 2.81 eV (441 nm) illumination at intensity 35.7 mW/cm2, which are shown in Figure 7.

In the text
thumbnail Fig. 9

Time variations of the positions and heights of P1 and P2 in Figure 8.

In the text
thumbnail Fig. 10

Time evolution of the SLD depth profile of Ag 500 Å/Ge20S80 1500 Å/Si substrate stacks during and after 2.81 eV (441 nm) illumination at intensity 35.7 mW/cm2, which were obtained from the fit to the measured neutron reflectivity profiles in Figure 7.

In the text
thumbnail Fig. 11

Time variation of the fitting parameters (thickness, SLD, and roughness) used for the fit to the measured neutron reflectivity profiles in Figure 7.

In the text
thumbnail Fig. 12

Time evolution of the neutron reflectivity profile of Ag 500 Å/Ge20S80 1500 Å/Si substrate stacks during and after 3.04 eV (408 nm) illumination at intensity 34.4 mW/cm2.

In the text
thumbnail Fig. 13

Time evolution of RQ4-Q plots for the neutron reflectivity profiles of Ag 500 Å/Ge20S80 1500 Å/Si substrate stacks during and after 3.04 eV (408 nm) illumination at intensity 34.4 mW/cm2 for 60 min, which are shown in Figure 12.

In the text
thumbnail Fig. 14

Time variations of the positions and heights of P1 and P2 in Figure 13.

In the text
thumbnail Fig. 15

Time evolution of the SLD profile of Ag 500 Å/Ge20S80 1500 Å/Si substrate stacks during and after 3.04 eV (408 nm) illumination at intensity 34.4 mW/cm2.

In the text
thumbnail Fig. 16

Time variation of the fitting parameters (thickness, SLD, and roughness) used for the fit to the measured neutron reflectivity profiles in Figure 12.

In the text
thumbnail Fig. 17

Time evolution of the neutron reflectivity profile of Ag 500 Å/Ge20S80 1500 Å/Si substrate stacks during and after 3.38 eV (367 nm) illumination at intensity 30.6 mW/cm2 for 60 min.

In the text
thumbnail Fig. 18

Time evolution of RQ4-Q plots for 3.38eV illumination at intensity 30.6 mW/cm2 for 60 min.

In the text
thumbnail Fig. 19

Time variation of the positions and heights of P1 and P2 in Figure 18.

In the text
thumbnail Fig. 20

Time evolution of the SLD profile for 3.38 eV illumination at intensity 30.6 mW/cm2 for 60 min.

In the text
thumbnail Fig. 21

Time variation of the fitting parameters (thickness, SLD, and roughness) used for the fit to the measured neutron reflectivity profiles in Figure 17.

In the text
thumbnail Fig. 22

Excitation photon energy dependence of the Ag thickness after 60 min light illumination. The initial Ag thickness before light illumination is indicated at E = 0 eV. The optical gap energy of amorphous Ge20S80 measured by Pan et al. (2.77 eV [29]) is indicated by a broken line.

In the text

Current usage metrics show cumulative count of Article Views (full-text article views including HTML views, PDF and ePub downloads, according to the available data) and Abstracts Views on Vision4Press platform.

Data correspond to usage on the plateform after 2015. The current usage metrics is available 48-96 hours after online publication and is updated daily on week days.

Initial download of the metrics may take a while.