Analysis of Ag Nanoparticles formed by Annealing Ag Thin Films on ITO

 

OPT407: SEM Practium

Spring 2012 Final Project

Chanse Hungerford


 

This website was created to fulfill a final project requirement for OPT407: SEM Practicum at the University of Rochester. The goal of this project was to become more familiar with the microscopy tools used for research. Use the links below to quickly navigate to each section.

 

Introduction

 

My project was a study on silver nanoparticles formed on ITO by annealing silver thin films. This technique is currently being used within my research group; the goal is to enhance thin film solar cells with metallic nanoparticles. The motivation for this project and a brief overview of how this technology works is outlined below.

Motivation

Photovoltaics have a lot of potential to generate electrical power on a very large scale. The cost of photovoltaics must be reduced by a factor of 2-5 to make them competitive with fossil-fuels. One of the major costs of conventional solar cells is for the materials. Conventional solar cells are thick in order to absorb more of the incoming light. By using thin-films the cost of solar cells could be greatly reduced. Reducing the cost of the solar cells will lead to a reduction in the cost of solar energy for consumers. While thin-film photovoltaics are cheaper than conventional cells, thin-film photovolataics still cannot compete because of their poor efficiency. [5]


Plasmonics and Solar Cells

Thin-film solar cells sacrifice efficiency because the absorbing layer thickness is much less than the penetration depth of the light [3]. A common method used to increase the efficiency of conventional solar cells is texturing. Thin-film solar cells cannot use this technique because the textures have features larger or equal to the thickness of the cell [2]. One of the proposed methods to increase light trapping for thin-film cells is through the use of plasmonics. There have been many different designs that have incorporated plasmonics in attempts to increase efficiency. Some designs use spherical nanoparticles [1], while others believe that cylindrical and hemispherical particles are better at scattering light [4]. All designs are focused on increasing the absorption and scattering of the solar spectrum. The reason why plasmonics will help is because surface plasmon excitation can increase scattering and leads to a large enhancement of the electric field. The absorption and scattering are dependent on the size of the nanoparticles. Particles that are less than 50 nm are dominated by absorption. Particles that are 100 nm scatter incoming radiation in a way that is useful for solar cells. Particles that are too large lead to multipole oscillations, which decreases scattering and is bad for solar cell applications. Certain groups have already shown that solar cells are enhanced by plasmons. For example, Derkacs has achieved an 8.3% increase in conversion efficiency by using gold nanoparticles on a thin-film amorphous silicon p-i-n solar cell [2].


This Project

One method to make metallic nanoparticle coatings is to deposit a very thin layer of the metal, then anneal the sample. This causes the film to coalesce and form semi-spherical particles. Film thickness, anneal time, anneal temperature, and even the method used to deposit the metal can affect the size, shape and distribution of the particles. For this project I will be using the current procedure being used within my research group to create nanoparticle coatings. The tools that we are using have recently changed and the effect that this change has had on the particles has not been analyzed yet.


References

[1] S. Nunomura, A. Minowa, H. Sai, and M. Kondo, Appl. Phys. Lett. 97, 063507  (2010)

[2] S. Pillai, and M.A. Green, Solar Energy Materials & Solar Cells 94 (2010) 1481-1486

[3] H. Stiebig, C. Haase, S. Jorke, P. Obermeyer, E. Moulin, and M. Schulte, Mater. Res. Soc. Symp. Proc. Vol. 1101 (2008)

[4] K. Catchpole, and A. Polman, Appl. Phys. Lett. 93, 191113 (2008)

[5] Atwater, Harry A., and Albert Polman. Plasmonics for improved photovoltaic devices. Nat Mater 9, no. 3 (March 2010): 205-213.

 

Procedure

 

Sample Preparation

There were two types of samples viewed for this project. The samples that were focused on for this project were glass slides with ITO deposited on them. Some solar cells were placed in the chamber with the glass slides and they went through the same process as the glass slides. A few of the solar cells were also imaged for comparison. These images were only qualitatively analyzed and no quantitative data was collected from the solar cell images.

Glass Slides

In order to mimic the devices that are being created by my research group, a layer of ITO was first deposited on the glass slides. This was done using an RF sputtering system at URnano, the Kurt J. Lesker PVD75. From previously calculated deposition rates the ITO was estimated to be about 40-45nm thick.

The silver (Ag) was then deposited using a very similar tool. The tool was also a PVD75 from Kurt J. Lesker, but it was configured for sputtering and E-beam evaporation. The E-beam was used to deposit Ag. Three different thicknesses were deposited: 12nm, 16nm, and 20nm. The slides were then annealed in an oven at 240C for 30 minutes.

Due to the nature of these samples, they do not require a lot of preparation to be imaged in the SEM. A small section was broken off of each slide and these sections were placed on a stub with carbon tape. The glass slide is a thick insulating layer and this may have caused the sample to charge. To avoid charging problems, a piece of copper tape was place over an edge of each sample.

TEM Grid

The glass slides and the solar cells are not easily viewable in the TEM. Luckily the process of making a sample for the TEM is relatively simple. For this project, a silicon TEM grid was placed in the E-beam during the 12nm deposition. 12nm was the only thickness chosen for the TEM because that would provide smaller particles which would be easier to view in the TEM. The TEM grid was annealed at 240C for 30 minutes with the other samples.

Solar Cells

Although the actual devices that the nanoparticles are going to be used on are not the focus of this project, the device structure is unique and interesting to see. Through the series of steps outline in Figure XXXX we end up with a 2mmx2mm free standing thin film solar cell. This is unique because most groups will only have access to one contact, the front or the back. With our structure we have access to both the front and back contacts which allows us to test a wide variety of designs. After the window is created, the rest of the steps are very similar to the glass slides. ITO is deposited on front and back of the device, then Ag is deposited on one or both of the side, and then the sample is annealed.


Imaging Techniques

SEM

The bulk of this project was done using the SEM. Both the InLens and SE2 detectors would provide good images. Standard imaging procedures were used for imaging each sample.

The FIB was also used on this project. The goal for using the FIB was to try and cut a particle to view its cross section. The size of the particles was relatively small, ~100nm, and we were unsure if they would survive the FIB process. The technique used for achieving this was to just cut through an large area and hope that a particle would be cut.

TEM

The TEM sample was prepared to try and image the oxide that may or may not be on the nanoparticles. While the particles are too thick to image through, we should be able to see the oxide at the edges if there is one present. The STEM function and EDAX were used to verify that the shell is an oxide.

 

Results

 

SEM Images

This first set of images shown in Figure 1 is a series of images at different magnifications of the three samples. Unfortunately, the 12nm and 16nm film both had problems during the E-beam deposition and the exact film thickness is not known. After imaging we can see that the two films were probably approximately the same thickness because after annealing the size and shape of the particles were very similar for the two samples. The 20nm sample is obliviously very different then the other two samples. The thicker film resulted in larger particles, which was expected.

 

12nm Starting Film

16nm Starting Film

20nm Starting Film

5kx

12nm-5kx

16nm-5kx

20nm-5kx

25kx

12nm-25kx

16nm-25kx

20nm-25kx

50kx

12nm-50kx

16nm-50kx

20nm-50kx

100kx

12nm-100kx

16nm-100kx

20nm-100kx

 

Figure 1: The three glass slide samples at various magnifications

 

There is an interesting feature that was very apparent in the 20nm images and it is those bright spots that are showing up on the particles. There are spots in the 12nm and 16nm samples, but they are more numerous on the 20nm sample.

 

ImageJ Analysis

A particle analysis was run on one of the images for each sample. The images chosen were the 25kx InLens images because these images contained a good number of particles with good contrast. This would make the analysis easier and more accurate. Table 1 shows the results of the analysis. The first three columns show the progression of the analysis, and the last column contains the resultant data.

The first column shows the original image. The scale bar was used to set the scale for the analysis. The second column is the image after making the image a binary image by using the threshold function and using the fill holes function. The third column shows the outlines of the particles that ImageJ found. For the analysis, I set the minimum area to include to 40nm^2. This is because there were a lot of single black pixels that are not actual particles and just artifacts left after the setting the threshold values. 40nm^2 is a very small value and the outlines show that the small particles were still being detected. The forth column contains a summary of some of the data collected. For this project, the diameter of the particles is of interest as well as the circularity. For diameter, the Feret diameter was calculated.

 

Original Image

Image Processing

Particle Analysis

Analysis Summary

12nm

12nm-25kx

12nm-25kx-thershold

Drawing of 12nm-25kx-thershold

 

Feret

FeretX

FeretY

Circ.

Mean

137.996

2042.905

1417.182

0.738

SD

124.08

1235.174

787.51

0.203

Min

12.057

17.052

17.052

0.116

Max

749.588

4343.872

2915.808

1

16nm

16nm-25kx

16nm-25kx-threshold

Drawing of 16nm-25kx-threshold

 

Feret

FeretX

FeretY

Circ.

Mean

127.073

2172.966

1467.83

0.752

SD

104.545

1240.53

819.823

0.2

Min

12.057

12.789

4.263

0.155

Max

577.868

4322.558

2890.231

1

20nm

20nm-25kx

20nm-25kx-threshold

Drawing of 20nm-25kx-threshold

 

Feret

FeretX

FeretY

Circ.

Mean

95.577

1993.33

1492.049

0.732

SD

116.042

1194.37

856.02

0.166

Min

12.079

12.811

4.27

0.246

Max

737.913

4334.52

2916.726

1

Table 1: ImageJ Analysis

 

Unfortunately these particles are not as spherical or as uniform as expected. Earlier experiments within the group used a thermal evaporator and deposited the silver directly onto either amorphous silicon or SiO2. We were not sure how the E-beam and the ITO would affect the particle formation and this is one of the reasons for doing this project. The 20nm film obviously has larger particles, but it is plagued with a lot of very small particles. This resulted in the 20nm sample to have the smallest mean Feret diameter, which was not expected. The 12nm and 16nm samples have a lot of non-spherical particles. They both have minimum circularities less than 0.2.

 

FIB

The FIB experiment worked better than expected. The FIB may have destroyed the small particles instead of cutting them. The sample used for FIB was a glass slide with 22nm of Ag annealed at 240C for 30 minutes. The thicker film was chosen to create larger particles, which would hopefully be better for the FIB. Fortunately, the particles survived and a few of them were cut. The method used to cut the particles was trial and error. The FIB beam was set at a high current to start; the current was then reduced to find a good setting for milling. Obtaining an image with the FIB was difficult, so instead of focusing on a particle a large area was selected for milling. Multiple attempts were made and the sample was viewed after all the milling was done. Some of the areas milled did cut through a particle. Due to the very small cross section, the images were difficult to acquire. There is no quantitative data for this test. The FIB experiment was done to determine if cutting a particle would be possible and to gain experience with the FIB. The images below show the results of the FIB experiment.

Large Hole

5pA Trap

2pA Trap

The FIB process started with a very high beam current. The high current destroyed the area that was being milled and this resulted in a large hole being dug in the sample.

The next pattern was milled using a 5pA beam current. This time the pattering is visible and the sample is intact. This pattering was milled in multiple locations to try and cut a particle.

The final test was done using a 2pA beam current. This provided the best results. One can easily see that the edges are much cleaner (sharper) then the edges with the 5pA beam. This beam current was also used to mill multiple locations.

Figure 2: FIB with various beam currents

 

The trapezoid pattern used was chosen to mitigate re-deposition problems when using the FIB. The material that is ablated away will redeposit on the sample. The trapezoid pattern provides an area for the material to redeposit and should help leave the edge of interest clean. The two images shown in Figure 3 are the two better images acquired of a cut particle.

2pA Cut Particle

5pA Cut Particle

2pA FIB

5pA FIB

Figure 3: Ag Nanoparticles cut by FIB

 

The images in Figure 3 show that the 2pA current resulted in a much cleaner cut. The cross section looks relatively clean. The 5pA current left white spots on the cut face of the particle. The bright spots are most likely an artifact created during the milling. The higher beam current may have resulted in areas of the particle annealing, making parts of the particle brighter than the rest of the cross section. A technique that would have resulted in a much cleaner cross section would have been to use the FIB to polish the area. If the FIB is set up well then we could have used a very low current to polish the milled area, which would help remove any artifact created by the high current beam.


TEM Images

The nanoparticles are formed in atmosphere and this would likely result in the particles oxidizing. This is something that has not been confirmed within my research group. The TEM images below clearly show that there is a ring around the particles. Then by using STEM with EDAX we confirmed that the ring contains oxygen, which means that the Ag nanoparticles are oxidizing during the fabrication process. Figure 4 shows a large area of the sample. Figures 5 and 6 show an individual particle.

Large Area

Single Particle

Figure 4: Large Area of TEM Grid

Figure 5: Single Ag Particle

 

Figures 5 and 6 clearly show that there is a ring around the nanoparticles. The ring is very likely a silver oxide and this was confirmed by using the EDAX. The scans are shown below.

 

STEM HAADF

Ag-L Map

Si-K Map

O-K Map

C-K Map

Large
Area

STEM HAADF Detector

Ag-L

Si-K

O-K

C-K

Single
Particle

STEM HAADF Detector

Ag-L

Si-K

O-K

C-K

Figure 6: EDAX

 

The first thing to notice is that the single particle in the STEM HAADF image is elliptical. This is an artifact that is created by the stage drifting while imaging. The Ag and Si were mapped as expected. The oxygen map displays an interesting feature that is caused by the fact that the X-ray detector is directional. The O-K map shows that there is more oxygen on the right side of the particle then the left. This is because the x-ray detector is located on the right side of the images. The single from the left side is blocked by the particle. The O-K map still shows that the oxygen concentration is larger on the particle than the surrounding substrate. This proves the hypothesis that the Ag nanoparticles are oxidizing. The C-K map can generally be ignored because it is also an artifact created while imaging. While imaging, the electron beam will cause carbon to be deposited on the sample.


Solar Cells

At the time of this project there were devices (the solar cells) that were ready for nanoparticle coatings. Glass slides with ITO are a good way to imitate what will happen with the actual device, but there can always be differences once the actual device is used. So a couple of solar cells were prepared similarly to the glass slides and imaged. This part of the project was qualitative only. Below are a few images taken of the solar cell. Images were taken with both the SE2 detector and InLens detector to show the difference between the two. Also, two of the images were colorized using some basic techniques in Photoshop.

InLens Detector

SE2 Detector

PE2-2-inlens-3mm-3kv-25kx

PE2-2-se2-3mm-3kv-25kx

PE1-5_window_50kx

PE2-2-se2-3mm-3kv-50kx

Figure 7: Solar Cell Sample Images

 

The images above show that the InLens detector was able to provide higher resolution images. For the work done in this project, the InLens detector was the preferred detector. If viewing the topography of a sample is important, then the SE2 detector may have been a better choice. The important thing to notice from these images is that the particle formation on the device appears to be different then the glass slides. The odd structures that are seen above will likely result in large losses of the electric field instead of an enhancement.

Blue Yellow Solid

Blue Red Hue

Figure 8: Two Examples of colorized images.

 

 

Conclusion

 

The new process being used does result in the creation for Ag nanoparticles. There is some control of the particles size by changing the thickness deposited by the E-beam. There due seem to be some significant issues with the new process. The 12nm and 16nm films formed mostly non-spherical particles. The 20nm film formed more spherical particles, but there were a lot of small particles that were not in the desired size range. The solar cells were plagued with many problems. The particles appeared to be mostly non-spherical and there are many patches of the very small particles. Future work on this project will require many more samples to be created. Various deposition rates for both ITO and Ag will have to be run. A step may be added to the process to pre-treat the ITO with a 240C anneal before the Ag deposition.

This project has shown that FIB on the nanoparticles does work. The samples viewed did not show anything significant. The particles were close to hemi-spherical. This may be an interesting test to try once the fabrication of nanoparticles is improved.

The one sample for the TEM provided a lot of information for this project. The particles were assumed to be oxidizing, but how badly they were oxidizing was unknown. The TEM images clearly show a ring a few nanometers thick around the particle. EDAX then showed that the particle is coated in oxygen.

Acknoledgements

Thanks to Brian McIntyre for all the help with the microscopy tools, especially for helping with using the FIB and acquiring the TEM images. Also, thank you to Josh Winans for assisting with the sample preparation.


 


 

Please enter any comments, criticisms, questions, etc. below.

Your name:

Email address: