Report of the internship done at the Laboratoire des Matériaux Avancés (LMA), Lyon, under the direction of Raffaele Flaminio

Transcription

1 Alix Cajgfinger Report of the internship done at the Laboratoire des Matériaux Avancés (LMA), Lyon, under the direction of Raffaele Flaminio Study and reduction of the dissipation in the materials used for the mirrors of the gravitational waves detector VIRGO Master 2 Sciences de la Matière 2011/2012 1

2 I would like to thank my tutor, Raffaele Flaminio, Director of the LMA, to have given me the opportunity to work on this project and see how experimental research is done inside a small lab, and who supervised my work. I would also like to thank Vincent Dolique, Massimo Granata, Nazario Morgado, Benoît Sassolas, Christophe Michel and all the members of the LMA for their help during my internship. They showed me how every experiment and program worked, and I couldn t have completed my internship without every piece of advice they gave me. 2

3 Abstract The gravitational waves detector VIRGO is an interferometer that is used to detect the gravitational waves, using mirrors as test masses. The task of the LMA is to find ways to improve the properties of these mirrors in order to improve the sensitivity of the detector. The quality of the coatings done in the lab is thus very important to the whole project. During my internship, I worked on the influence of some parameters of a coating machine on the mirror coatings that were done. The goal of this study was to determine which parameters to choose in order to decrease the thermal noise and the absorption of the coated samples. In this report, we will first have an introduction of the project and on the different elements that will help us understand the experiments done in the lab. Then, we will analyze the different setups used to do this study. In the end, we will see what results we have been able to find. Table of Contents I) Introduction to gravitational waves and to the interferometers used to detect them... 4 A) Presentation of the gravitational waves detector VIRGO... 4 B) Thermal Noise... 4 C) Mirror Conception... 5 D) Role of the index of refraction and the absorption of the coating... 6 E) Presentation of the coating machine SPECTOR... 6 F) Goal of the Internship... 7 II) Parameters of interest and methods... 9 A) Different setup parameters for the main source and the assist source... 9 B) Measurement of the quality factor of a coating... 9 C) Measurements of the refractive index and the absorption of a coating D) SIMS Analyses E) EELS and EDX Analyses III) Results A) Coating made without the help of the assist source 14 B) Coatings made with the help of the assist source 15 1) Problem of values.. 15 i) TFCalc ii) Ellipsometer ) Index of refraction and extinction coefficient ) SIMS Analyses ) Analyses done at the CLYM. 20 References

4 I) Introduction to gravitational waves and to the interferometers used to detect them General relativity is a theory of gravitation that was established by Albert Einstein in the beginning of the 20 th century. It predicted a lot of different results that turned out to be observed in the Universe and confirmed it. One of the main results that had been foreseen, however, hasn t been directly observed yet: gravitational waves. These waves are produced by the motion of massive objects of the Universe and appear as a deformation of space-time. As a result, a change in the distance between two objects could be measured. A) Presentation of the gravitational waves detector VIRGO The Virgo project is aiming at the first direct observation of gravitational waves. The detector is composed of a laser interferometer with two arms of three kilometres long each. The goal is to measure the differential arm length changes produced by gravitational wave by looking at the changes in the interference pattern. Thanks to its high sensitivity, this has great chances of success. However, improvements have to be made in order to be able to detect these waves that are very weak and hard to observe. The main limitations are the photon shot noise, the mirror thermal noise and other mechanical noises, such as those induced by the earth seismic vibration. The interferometer includes one optical cavity in each arm which are composed by two mirrors (called test masses) of about 40 kg.. These mirrors are made out of a thick cylinder of silica (the substrate) and are coated with multi-layers of two components (titania-doped tantala and silica) that are alternatively deposited on the surface of the silica substrate. These coatings are made to reach the predicted reflectivity of the mirrors, and they have to have the mechanical losses and optical absorption as low as possible. These losses are the very limitations to the sensitivity of the interferometer. B) Thermal Noise Figure 1: Sketch of the Interferometer Virgo (Ref 15) The equipartition theorem tells us that in a mechanical oscillator, the thermal vibration is given by the relationship 4

5 . However, this thermal noise is not constant with the frequency of the system, as we can see on figure 2. The biggest part of it is concentrated near the resonance frequency, which means that the higher the quality factor is, the lower the thermal noise gets far from the resonance frequency. Therefore, if we want to decrease the thermal noise that stands at our frequencies of interest, which is to say around ten to a hundred Hertz (i.e. well below the mirror resonances), it is important to improve the mechanical quality factor of the system. By doing so, one increases the amount of thermal noise at the resonance frequency, thus reducing the thermal noise elsewhere. C) Mirror conception Figure 2: Thermal noise vs Frequency The mirrors that are used in the interferometer are made out of silica substrates that are coated with thin films. To obtain a highly reflective mirror, two different materials are used: SiO 2 which has a low index of refraction, and Ti:Ta 2 O 5 which has a high index of refraction. By coating alternatively the mirrors with these two elements as shown on the figure, the result is that the reflectivity of the final sample increases. The first beam that arrives and crosses the titania doped tantala is reflected and also gets transmitted, only to be reflected again once it hits the second layer of tantala. With a few tens of /4 thick layers, the reflectivity reaches values that are high enough for the interferometer to be efficient. Figure 3: Conception of a mirror The manipulations of the substrates and mirrors have to be done in a clean room. In this area, the flux of air, coming from the ceiling and exiting from the floor only to be recycled, is finely controlled. The whole process allows people who work in this room to get clean air all the time, which reduces the risks of dust being trapped into the samples as they are treated. Then, the procedure starts with the cleaning of the substrate that is going to be coated. After that, the sample is put into a coating machine that will coat the substrate with one or many layers of one or two chosen materials by ion beam sputtering. This technique will be detailed later in the report. When the process is done, the sample is taken out of the machine and some preliminary measurements are done to get pieces of information on it. After the coating is finished and the measurements are done, the sample is put in an oven in order to anneal it. This step reduces the internal stress and strain inherent to the coating, and therefore reduces the mechanical losses of it. Then, other measurements are done to determine the properties of the final product. 5

6 D) Role of the index of refraction and the absorption of the coating The absorption of the coating has two effects on the Virgo interferometer. On the one hand, when the laser beam of the interferometer reaches the surface of the mirror, it increases slightly the temperature of the mirror where the beam spot is. Since the index of refraction depends on the temperature, an index of refraction gradient is created inside the mirror substrate, which makes more difficult the interferometer control. On the other hand, when the coating absorbs a part of the incoming light, the consequence is a loss of the total light of the system. In the end, the higher the absorption of the mirror is, the lower the power in the Fabry-Perot cavity becomes and the lower the interferometer sensitivity is. For these reasons the coating absorption must be as low as possible. E) Presentation of the coating machine SPECTOR The SPECTOR is a coating machine made out of two different ion sources, three targets of the materials we want to coat the sample with, and some masks to help controlling what is coated. The main ion source is a source of Argon. In the process, the ion beam of argon is neutralized with an electron gun and hits the target of the material we want to deposit in order to pulverize it. A flux of oxygen is also present to oxide the coating, creating the material we are looking for, Ta 2 O 5 for example. In addition to this, a source of oxygen and argon can be activated in the SPECTOR in order to help oxide the coating and to increase its density. This is the assist source. A sketch of the machine is presented bellow. Picture 1: Picture of the SPECTOR Figure 4: Sketch of the coating machine SPECTOR The method to obtain a coated sample is the following one: First of all, the substrate that is going to be coated in the machine has to have its absorption measured. This measurement, as well as the one that is done after the whole 6

7 process is finished, will tell us what the absorption of the coating is. This will give us one of the properties we need in order to determine whether one method of coating is better or not. After this measurement is done, we take the substrate and clean it. Once the substrate is clean, we put it in the SPECTOR with screws and some adequate devices, and then we have to choose what recipe to use in order to get the coating we want. This step is controlled by the software that is included in the machine. We have to select the parameters of the main source, and if needed the parameters of the assist source as well. These parameters will in the end be determinant for our study. The hope is to reach an optimal configuration, indeed. The samples are coated by ion beam sputtering. During this process, a plasma of Argon is created in the main source and an ion beam is extracted from the source with a tension that is applied to it. The first parameters we can control are then the voltage,, and the current,, of the main source. The extracted ion beam reaches the target of material and pulverizes it. The particles that are pulverized in the process get on the substrate that is rotating and form the coating. Another tension can be changed in order to stabilize the whole process, but it doesn t directly influence the coating. A neutralizer (RFN) is also there to neutralize the ion beam. When the assist source is activated, another plasma is created, made out of Argon and Oxygen. This plasma is extracted with a certain tension, -12 cm, and current, -12 cm, which can also be changed. The ion beam that comes from the assist source can help increase the density of the coating by adding energy to the particles of the material that coats the substrate. The presence of oxygen can also play a role in the oxidation process. F) Goal of the internship The goal of my internship is to study the effect of different setup parameters on the performances of the low losses coatings that are done for the interferometers. The performances of interest are the absorption, the mechanical losses and the index of refraction. The ultimate goal is then to see how to improve the samples that are made in the SPECTOR. During my internship, I had to compare different measures done with different parameters of the assist source in order to figure out which configuration is the best, and how these parameters play a role in this process. In order to do that, I was asked to try to determine how the parameters of the assist source in the SPECTOR were influencing the coated samples compared to the samples done without the help of the assist source, and how we could improve the quality of the coating materials and have a better sensitivity in the measurements done with the gravitational waves detectors. To help characterize the different samples that we obtained, a Secondary Ion Mass Spectrometry (SIMS) analysis has been done at CPE Lyon. These analyses will be detailed in the next chapter of this report. Besides that, in order to have a better understanding of the effect of the interfaces in the final product, electron energy loss spectroscopy (EELS) analyses have been done in the CLYM lab. In the material that we want to put on the substrate, a balance has to be reached between the number of interfaces which will improve the reflectivity of the final product, and the mechanical losses which are caused by the increase of material and interfaces. Of the two materials that are used for the thin films, the one that has the highest mechanical losses is the Ta 2 O 5 and the more material we put on a substrate, the more losses we get from it. The idea is thus to make better, more effective coatings in order to 7

8 reduce the mechanical losses and the number of layers needed to have the best reflectivity (thus decreasing the number of interfaces in the coating). 8

9 II) Parameters of interest and methods A) Different setup parameters for the main source and the assist source During my internship, we worked with different setup parameters to investigate the role of the assist source on the coating that is done. The following table presents the values that have been used in the SPECTOR to coat the different samples. Table 1: Summary of the setups of parameters used There are three kinds of samples that are used in the coating machine. The first type is a one inch diameter silica substrate. This substrate can have its transmission spectrum measured, as well as its absorption. The second type is a piece of silicon that can be used for other kinds of measurements, like the Rutheford Back Scattering (measurement not performed during this stage). These measurements can help us understand what the coatings are made out of. The last substrates are one third of inch diameter BK7 substrates that are used for the SIMS analyses. B) Measurement of the quality factor of a coating Coating mechanical losses are an important parameter to consider, since it is what determines the mirror thermal noise. Thermal noise comes from the internal damping of the sample, including surface and volume losses of the material. The coating contribution turns out to be the dominant source of losses. In order to determine these losses, we measure the quality factor of the samples, which gives us information about their mechanical properties. This quality factor is defined as the ratio of the energy stored in the material and the energy dissipated per cycle. 9

10 To measure this parameter, we put a welded blade made out of silica (see picture) in a vacuum chamber, held on a heavy block that prevents other noises to interfere with the measurements. This silica blade is sensitive to electric fields, so we excite it with electrodes. The point is to excite the blade at its resonance frequency and let it oscillate as the amplitude of the sample motion decreases, as we can see in the following figure. Picture 2: Welded blade Figure 5: Exponential decay of the amplitude The decrease of the amplitude with respect to time is linked to the quality factor of the sample. Indeed, the mechanical quality factor satisfies the following relationship: where is the frequency of resonance of the mode, and the time constant of the vibrating system. The equation of the envelope becomes: With the help of a program, this quality factor can then be measured and derived. Picture 3: Experimental setup to measure the quality factor In order to be able to do that, a laser beam is sent on the blade that is inside the chamber. We use it in a vacuum of at least Torr because we don t want the blade to be damped by the residual gas. The laser is reflected by the blade and reaches a position sensitive detector (PSD) that measures the amplitude of the oscillation. On the following picture, we 10

11 can see four plots. The one in the upper left corner shows the raw signal detected by the photodetector that intercepts the laser. Then, we take the fast Fourier transform of this signal and it gives us the second window on the upper right corner of the picture. This helps us determine what the resonance frequencies are, in order to excite them and get their quality factor. Once we know the value of one resonance frequency of the blade, we can excite it with the help of an amplifier. Once the excitation is turned off we can start observing the exponential decrease of the blade amplitude on the window which is under the raw signal window. A fit of this curve is done in order to obtain the graph of the quality factor as a function of time, which is shown on the bottom-right plot. Figure 6: Program to measure the quality factor At the end of the determination of the resonance frequencies and the maximum and minimum amplitudes, we can launch an automatic measurement program that will continuously measure ten runs for each mode. Then, by taking the average, we can know how to describe the behaviour of a blade. In order to get the quality factor of a coating, we must measure the one of the blade alone and the one after the coating is done, and then we can get the quality factor of the coating itself by deriving a relationship between these quality factors and the energies of the substrate and the coating. First, we have that the total energy of the sample times the loss angle of the whole sample is the sum of the energy of the substrate times its loss angle and the energy of the coating times its loss angle. (The loss angle is defined by ) However, the coating is so small compared to the substrate that we can assume that the total energy comes from the substrate, which is translated by the following equation: 11

12 Therefore, In the end, we have: With the help of this method, we can thus determine whether a change in the coating deposition parameters will improve the mechanical quality of the material or worsen it. C) Measurements of the refractive index and the absorption of a coating In order to determine the absorption of some coating, the following method is used. First, the absorption of the substrate is measured with the help of a pulsed laser and a detector. Then, the sample goes in the machine to be coated. In the end, the whole product gets measured again, and the difference between the two measures gives us the absorption of the coating. To determine the index of refraction and the thickness of a coating, we use a spectrophotometer. We have to put a coated substrate in the machine and get the spectrum of it for a range of wavelengths going from 200 to 1400 nm. What is measured by the spectrophotometer is the transmission coefficient as a function of the wavelength. Different lamps take turns one after the other in order to cover the whole range of wavelengths. The spectrum obtained with this machine can be seen in the following picture. Figure 7: Transmission spectrum of a layer of tantala Given this spectrum, we use a software designed by an engineer from LMA to get the thickness and the coefficient of the Cauchy law for the index of refraction of the coating. 12

13 In this program, the values of the maximum and minimum transmissions give an initial estimate of the index of refraction, whereas the distance between two maxima gives us the product of the thickness times the index. Based on these initial estimates a fit of the data provides the coating thickness and the index of refraction as a function of the wavelength. In our study, we are interested in the index of refraction at 1064 nm because it is the wavelength used for the laser in gravitational waves detectors such as Virgo. This means that we want to have the best index of refraction for this specific wavelength. D) SIMS Analyses These analyses help determine what the composition of the very first layer of the coating is. In addition to that, SIMS analyses are useful to detect the contaminants of the samples, since the sensitivity of the detector is very good. Therefore, it is possible with this technique to observe the contaminants coming from the deposition even if their concentration is very small. These analyses were done in a lab of CPE Lyon. The method consists in placing the sample inside a vacuum chamber at Torr and shoot an ion beam on it. The beam will ionize the molecules from the first layer and these molecules will then be detected by a mass spectrometer. The output is a spectrum graphing the intensity versus the atomic mass of the detected species. This analysis by itself cannot give a quantitative measurement of the coating composition but can be used in order to compare two samples. It tells us the level of contamination by an element, for example carbon or oxygen, which are well-known contaminants. The goal of these measurements is mainly to see if there is any kind of contamination going on when the coating is done in the machine as well as to see the differences in the composition of the samples done with different sets of parameters in the SPECTOR. For the first two measurements, which were on an uncoated substrate and a coated one, there was no contaminant except for the usual ones: carbon, oxygen and some others that we can always find in these kinds of measurements, since it is a very sensitive probe. We were able to see in these spectra the mark of tantalum and of its oxides as expected in the coating. E) EELS and EDX analyses The analyses in the CLYM lab are used to observe the pattern followed by the coated material, especially when it comes to interfaces. The CLYM is a lab that uses electron energy loss spectroscopy (EELS), energy dispersive X-ray spectroscopy (EDX) and other methods to determine the composition of a sample. The sample analyzed was a multilayer of titania-doped tantala and silica, made for the interferometer LIGO (the equivalent of Virgo in the United States). The reason we had to do these experiments was that the values of the mechanical losses that were measured for this sample were not equal to the linear combination of the values of the Ti:Ta 2 O 5 and SiO 2 ones. The first result were based on Transmission Electron Microscopy and indicated an effect of diffusion between the layers. This produces a mixed material at the layer interfaces with a rate of mechanical loss higher than expected. Also, an interesting fact which had to be verified was that in the tantala layer, the concentration of silica is not reduced to zero, but stands for ten percents of the whole atomic concentration. Besides that, it was shown that there was a dissymmetry between one side of a layer and the other, when we look at the peaks atomic density. This means that the material is 13

14 denser on the side that is close to the substrate, and becomes less dense as it goes farther away from it. In addition to these results, the lab estimated the thickness of the layers and of the whole sample to be bigger than what had been measured in our lab. In collaboration with Dr. Thierry Epicier from CLYM, we did some additional measurements in order to better understand the composition of the coatings made in the LMA. A deeper analysis of these results could tell us whether these initial results came from the machine itself and its relative precision in the measurements, or from the sample. The first step made to continue the measurements that had been done before was to get a spectrum of pure Ta 2 O 5 that would be used as a reference for the following spectra. In order to do that, we used a reference Ta 2 O 5 powder that was reduced to elements no larger than a micrometer and then placed in the microscope. In order to get characterize the sample, we need to have a spectrum of the sample, a low-loss spectrum, a white-noise spectrum and a spectrum with the dark count of the two previous spectra. The low-loss spectrum is the one we get when no electrons are deviated by the sample. It gives an idea of the behaviour of the electrons, were they alone in the microscope. The white-noise spectrum is the answer from the diodes of the detector when a constant signal lightens them. By dividing the experimental spectrum by the white-noise one, we can correct it, getting rid of the effect of the detector in the process. The second part of the analysis consisted in putting a sample made out of a multilayers coating done at the LMA in order to have a clearer idea of how it looked like. What was interesting was to see the interfaces between a layer of SiO 2 and one of Ta 2 O 5. Previous measurements and analyses had shown that there was a diffusion of the two materials across the interfaces. A mixed material appeared to stand where a clean and precise separation between the two layers was expected. However, it had to be determined whether this mixed region was as large as seen during the first measurements done using transmission electron microscopy. III) Results A) Coatings made without the help of the assist source If we want to know whether the assist source improves the properties of the coatings or not, we had to measure a sample of reference, made without the assist source being active in the chamber. By measuring its absorption, its refractive index and its mechanical losses, we can compare the values to the ones found in the other samples, made with the help of the assist source. Thus, the index of refraction used as a reference was, the extinction coefficient, which determines the absorption of the coating, was and the mechanical losses were around. In order to determine the mechanical losses of the coating independently, another experiment was done, which was to coat a silicon microcantilever with tantala and measure its thermal noise with the help of a quadrature phase interferometer for atomic force microscopy. The result that could come from the measurements would give us a precise idea of the value of the entire blade thermal noise. However, cantilevers are very difficult to manipulate and one must be careful not to break it while preparing it. Another difficulty comes from the fact that the laser used in the interferometer has a wavelength of 633 nm, 14

15 which corresponds to a region of the transmission spectrum of the coating that is low in transmission, and that the coating has a relatively low reflection at this wavelength. A first collaboration has been established between the LMA and the ENS in order to measure this thermal noise. The first results were encouraging, but at the end of my internship it was still too early to have any conclusions on the subject. B) Coatings made with the help of the assist source The first manipulation that had to be done was to do the runs that had been done with different sets of parameters of the assist source again in order to measure the absorption of the coatings and to verify the previous measurements. After that, we had to choose the best conditions and do the experiments again, but this time with blades. This allowed me to measure the Q factors that corresponded to these coatings and these parameters. The ultimate goal was to find out what influence the assist source has on the coating. The choice of runs to re do was made by looking at the absorption and at the refraction index of the coatings once the sample had been annealed. The samples which show the most interesting features were the ones to focus on. 1) Problem of values One of the problems I have been confronted to during my internship was the calculation of the index of refraction. As we have seen earlier, a software designed by a research engineer of this lab is used to compute the coating refractive index and thickness. In order to do that, we need to take the transmission spectrum of the sample as well as the one of the substrate, and an algorithm determines the Cauchy coefficient of the coating. When I arrived, we conducted new experiments which were using the same parameters as the old ones and I had to compute the refractive indices and thickness of the new samples. The goal of these new samples was to get the absorption that hadn t been measured before. By looking at the old values and by comparing them to the new ones, I noticed that there was a gap of a few percents between them. Starting from there, I took the old spectra and computed the values with the same method and the same substrate I used for the new samples, and I found out that these values were closer to the new ones than to the old ones. The issue then was to determine what influence this difference in the index of refraction has if we have to produce a mirror. i) TFCalc The software used to study this influence was TFCalc, a thin film design software that allows us to simulate mirrors. We wanted to see the effect of a change of index in different kinds of mirrors. We simulated two different mirrors, one with ten doublets, the other one with twenty (high reflective mirror), and we used two indices separated by two percent. The result in both cases was negligible. 15

16 In addition to this method, we also studied a real mirror. On this software, we can simulate a theoretical mirror with a law of index of refraction for and and compare it to the spectrum of a real one. By changing the law of index, going from the one I found to one that is multiplied by 1.01 % or 0.99 %, we can see which simulation fits the best to the experimental spectrum. The value of index that matched the best the spectrum we had for the mirror was the one found with the software and the substrate we used. On the following two images, we can see the difference between the closest law of index and one that is not good enough. Figure 8a: Fit of the transmission spectrum of a mirror (good index) Figure 8b: Fit of the transmission spectrum of a mirror (index too high) ii) Ellipsometer Another method to determine which refractive index is the one to actually consider is to use an ellipsometer. This machine uses the change in polarization of of a He-Ne laser beam upon reflection on the sample being studied. Using the change in polarization, the ellipsometer allows extracting the coating thickness and index of refraction. In order to estimate what values of indices correspond the most to the samples that had been coated, we measured their thickness and approximated index of refraction with this machine. The results, however, were higher than both the old and new values, so it was difficult to tell whether the calculations were not good or the measurement of the ellipsometer was not that accurate. 2) Index of refraction and extinction coefficient The first five runs have been done with different values of the parameters in the machine, which are mainly the bias voltage of the assist source and its current. They gave some results which were, at first sight, not decisive in the characterization of the influence of each parameter in some of the properties of the coated samples. The increase in the tension and intensity of the assist source clearly decreases the index of refraction of the coating, whereas the absorption doesn t seem to follow any specific law when it is plotted versus these parameters. 16

17 Index of refraction Index of refraction Index of refraction vs Tension of the support source 2,08 2,07 2,06 2,05 2,04 2,03 2,02 2, Vb (V) Figure 9: Index of refraction vs the tension of the Assist Source Index of refraction vs Intensity of the support source 2,08 2,07 2,06 2,05 2,04 2,03 2, Figure 10: Index of refraction vs the intensity of the Assist Source However, if we take the power of the assist source, which is its tension times its intensity, and we draw the graph of the extinction coefficient as well as the refractive index versus the power, we can see that there seems to be a general decrease in the index of refraction as well as in the extinction coefficient. One of these results is good for the final goal we want to achieve: we need an extinction coefficient that is lower than the one that we have as a reference, which is. This decrease in absorption goes along with the decrease of the refractive index. Yet, this refractive index only decreases from at most two percents, which as mentioned before doesn t influence high reflective mirrors. Ib (ma) 17

18 Index of Refraction k (10-8) Extinction Coefficient Vs Power of the Support Source Power (W) Figure 11: Coefficient of extinction vs the power of the Assist Source Index of Refraction Vs Power of the Support Source 2,06 2,05 2,04 2,03 2,02 2, Figure 12: Index of refraction vs the power of the Assist Source After looking at these results, the next step was to choose among the different sets of parameters which points we would select to do the coating of the blades and measure their mechanical losses. We choose to run the assistance source at 80 W since this was the best compromise in terms of low absorption and high index of refraction. Then the first experiment we did was to make a mirror with these conditions to check if the extinction coefficient matched the one found with the monolayer of Tantala. The goal of this experiment was also to verify the value of the index of refraction. Once the value confirmed, we put two blades in the SPECTOR and we measured their quality factors before annealing them. At this point of my internship, there were some problems with the pumps of the bench where the measurements of the quality factor were done as well as with the table anti-vibrations where the vacuum chamber containing the sample stands. The first issue was that the primary pump was not working very well, so we had to clean it and change the joints. The second one, which was related to the first one, was that the secondary pump had been affected by the malfunction of the first pump and had to 18 Power (W)

19 work harder in order to keep a good value of vacuum. When I had to measure the quality factor of the first set of blades that had been made, the secondary pump stopped working. The last problem was a leak of oil in the table, which led to a leak of compressed air and in the end we had to find a way to repair it in order to be able to do the measurements without being perturbed by the ambient vibrations. Once it had been clear that the turbomolecular pump had to be fixed, we decided to move the whole experiment to the cryogenic bench that was usually used to go down in temperature and measure the quality factors of the blades. Without using all the devices related to the cryogenic experiment, we set up the coated blades in order to be able to measure their quality factors. The first measures showed that before being annealed, the coatings had lower mechanical losses than the coatings made without the assist source. However, after we annealed the blades, the measurements showed that the mechanical losses have the same order of magnitude as the ones of the sample of reference. Every experiment that is done to measure the mechanical losses are done on two substrates in order to have a way to be sure of the results. On the following graph, we can see the average of the mechanical losses of two blades for two different experiments: one done without the assist source (in green on the graph), and one done with it (in purple on the graph). The error bar is defined by the difference between the values of the two blades done in the same conditions. With this uncertainty, it becomes unsure whether or not the assist source does decrease the losses. However, for the first mode, the assist source seems to help lower the loss angle. Figure 13: Comparison of two experiments done with or without the help of the support source 3) SIMS Analyses After all the samples had been analyzed in the SIMS machine, we have been able to compare them to one another, as well as to a Ta 2 O 5 powder that we could take as a reference. Three areas of each sample are probed during the experiment, and there are two kinds of ion detection: in the negative mode and in the positive mode. This allows us to have a general overview of what the coating we are dealing with is made out of. However, we have to keep in mind that this analysis is done at the extreme surface of the sample, which means that it is not decisive as for the inner structure of it. The goal of the study was to see if there was any difference between the coatings with respect to the power used 19

20 in the assist source. In the following graph, we can see the normalized integral of the intensity of different peaks linked to Ta 2 O 5 versus the atomic mass of these elements. The four samples are, from top to bottom: the one made with the lowest power in the assist source, the one made with the highest, the one made without the help of the assist source and the reference powder. All of them were made in the negative mode, which means that the ions detected were the negative ones. Figure 14: Normalized integral of the intensity vs the atomic mass of different elements As we can see on the graph, there is not much of a difference between the different samples and the powder when it comes to the Ta 2 O 5 peak. However, when we look at the Ta 2 O 6 one, we notice that all the coatings show a higher normalized integral of the intensity than the powder, which could mean that there is an over-oxidation of the material during the coating process. 4) Analyses done at the CLYM Finally, in order to determine the effect of the interfaces between the different layers, we had to analyze the spectra obtained in EELS and EDX. After the experiments to get the different spectra were over, we were able to average the results to have access to a reference of Ta 2 O 5 and SiO 2. The spectrum of TiO 2 was taken from a reference given in the software. However, for the first analyses, titania hadn t been used because of its low concentration in the sample. This allowed us to fit a linear combination of the two elements on a raw spectrum of the sample to see how the material changed with respect to the distance to the interface. In order to do that, we had to measure the spectra of electron energy loss at different positions, starting from the middle of the layer of titania-doped tantala. From this point, we moved the probe used in the experiment towards the silica layer with a step of one to two nanometers. On the picture, we can see the whole sample as a transmission electron spectroscopy (TEM) picture. The black layers are the doped tantala ones, whereas the white ones represent the silica. 20

21 With the help of the standards we obtained for the tantala and the silica, we were then able to determine the percentage of tantala and silica in the material that was probed with respect to the position in the material. In the following TEM picture, we can see the path followed by the probe when the acquisition was done. The step used for the scan of this line (linescan) was 2 nm. This means that every two nanometers, we acquired an electron energy loss spectrum and an EDX one, so that we were able to compare the data from both spectra. The idea of the analyses was to look at the oxygen linked to the elements and see how it behaved as we were moving the sample. By Picture 4: Interface Silica/Tantala seen in TEM comparing the different spectra, we were thus able to tell whether the material that was analyzed was made out of tantala or silica. On the following picture, we can see the superposition of the EELS spectra next to the TEM picture. Figure 15: Evolution of the spectra with respect to the position in the material noticeable bump appears. As we go down towards the interface and the tantala part of the sample, we see the evolution of these spectra: the bump coming from the oxygen of silica becomes smaller and the one coming from the oxygen linked to the tantala becomes bigger. Starting from this piece of information, we were then able to track the percentage of these elements in the material with respect to the position of the probe. In order to do that, the idea was to take two spectra of reference: one of Ta 2 O 5 which came from a powder that was analyzed for this purpose and another one of SiO 2 taken from the middle of a silica layer in the 21 The presence of the oxygen linked to the silicon can be seen on the top spectra, where a Picture 5: Mirror seen in TEM. The black layers are made out of titania-doped tantala, the white one of silica

22 sample. In a first approximation, we didn t consider the effect of the titania in the analysis. We computed a linear combination of the two elements as following, where α is the percentage of tantala in the analyzed material: By taking the difference of the values of this linear combination with the values of the raw spectrum squared, and by minimizing it with respect to the coefficient α, we fitted a curve to the raw spectrum and thus deduced the best coefficient that allowed the best fit. An example of these fits is given in the following graph. The spectrum has been acquired in the silica layer, so that we can see that the fit of the curve, which is in red on the graph, is exactly on the graph of the silica spectrum. The blue curve represents the reference of Ta 2 O 5, and the green one represents the reference of SiO 2. The black curve that is underneath the red one on this graph is the raw spectrum that has been acquired during the Figure 16: Fit of the EELS Spectrum experiment and that we want to fit. When we did the calculation to find this coefficient for every spectra of one linescan, we were then able to graph it with respect to the position of the probe in the sample, knowing that the step of the measurement was 2 nm in this case. This far, we have been able to compute only one graph that we can see bellow. Figure 17: Percentage of oxygen linked to tantala and silica in the material at the interface On this graph, we can see that the gradient of concentration seems to be about 10 nm long. However, according to Goldstein s relation (ref 16), the broadening b(t) experienced by the beam after its propagation through a thickness t is given by (with b and t in cm): 22

23 With the primary energy of the electron beam in ev, Z and A respectively the atomic number and atomic mass of the mean chemical species present in the sample, with an atomic density given by r. This relationship means that as the beam goes through the material, there is an interaction between the beam and the electrons of the material that ends up broadening the initial beam. As a result, the width of the interface we have computed earlier is bigger than the real one. In the end, the total width of the interface should have a value of about 5 nm, which contradicts the first results that had been found that said that the interface was about 35 nm wide. Further analyses have to be done in order better to understand what the interfaces look like, as well as how the diffusion between the different layers plays a role in the other measurements that are done in the LMA. To conclude, this study as given us some clues as to the influence of the assist source on coatings. Thus, we can see some differences in the properties of the samples that have been made during my internship with different setup parameters for the source. We then tried to link these differences in the index of refraction and absorption to the stoechiometry and inherent composition of the coatings, with the help of SIMS analyses and spectra analyses. However, there are still some observations to make, as well as some experiments to do in order to determine the best way to use the assist source in the Spector. 23

24 References 1) Thermal noise in mechanical experiments, P.S.Saulson, Physical review D, vol 42 number 8, ) Study of coating mechanical and optical losses in view of reducing mirror thermal noise in gravitational wave detectors, R.Flaminio et al, Class and Quantum Gravity 27, 8 (2010) 3) Internal thermal noise in the LIGO test masses: a direct approach, Y.Levin, ) Thermal noise in interferometric gravitational wave detectors due to dielectric optical coatings, A.M.Gretarsson et al, 3rd Edoardo Amaldi Conference on Gravitational Waves, AIP Conference proceedings 523 (2000) ) Effect of optical coating and surface treatments on mechanical loss in fused silica, G.M.Harry et al., Class. Quantum Grav (2002) ) Mechanical loss in Tantala/Silica dielectric mirror coating, S.D.Penn et al., ) Experimental measurements of coating mechanical loss factors, D.R.M.Crooks et al., Class. Quantum Grav. 21 (2004) S1059 S1065 8) Titania-doped Tantala/Silica coatings for gravitational-wave detection, G.M.Harry et al., Class. Quantum Grav. 24 (2007) ) Thermal noises and noise compensation in high-reflection multilayer coating, M.L.Gorodetsky, ) Mirror thermal noise in laser interferometer gravitational wave detectors operating at room and cryogenic temperature, J.Franc et al., ) How to reduce the suspension thermal noise in LIGO without improving the Q s of the pendulum and violin modes, V.B.Braginsky, Y.Levin et al., ) Thin-Film optical filters, 2 nd edition, H.A.Macleod 13) Measurements of a low-temperature mechanical dissipation peak in a single layer of Ta2O5 doped with TiO2, I.Martin et al., Class. Quantum Grav. 25 (2008) 14) Optical characterization of dielectric and semiconductor thin films by use of transmission data, Jorge I. Cisneros, Applied Optics Vol 37 No 22, ) (Schéma Virgo) 16) Introduction to Analytical Microscopy, J.L.Goldstein, Plenum Press: New York, p (1979) 24