The goal of this research is to characterize thin membranes used as transmission dynodes (tynodes) for electron multiplication. A stack of tynodes could be used in a novel ultra fast soft photon counter. A combination of a charge sensitive TimePix1 CMOS chip and a pulsed electron gun were used as part of a setup deemed Tytest. The pixel chip was calibrated using a surrogate function that is fitted to the single pixel responses of 76 different pixels. The calibration is used to perform Transmission Secondary Electron Yield (TSEY) measurements on a 20 nm thick alumina membrane. The maximum TSEY value observed is 2.76 ± 0.64 for a incident energy of 1675 eV. The maximum TSEY value seems to be at higher incident energies which cannot be reached using our setup. A unexplained spread in electrons being emitted by the tynode is observed.
The second part of this research was to create a stack of tynodes which can be used to convert a single electron into a measurable signal. The goal is to stack the tynodes using a self aligning mechanism and measure the misalignment. A ZEISS Axio-zoom V16 light microscope in combination with two µm precision manipulators placed in a clean room at Nikhef in Amsterdam is setup in this research used to stack tynodes. The aim was to have a misalignment better than 5 µm, and this setup a misalignment of 4.1±1.3 µm was measured. It’s important to note that only the misalignment between two tynodes is measured, since no more were readily available.

For this Master thesis the possibility was explored to use a monolayer of graphene as conductive layer, instead of sputtered titanium nitride (TiN), on 25 nm thick aluminium oxide and magnesium oxide membranes. These membranes will be used in a new and faster variant of a vacuum electron multiplier (e.g. the Photo Multiplier Tube (PMT)). These membranes should act as transmission dynodes, or tynodes, where electrons are multiplied while interacting when they travel through the membrane and leave the membrane on the other side instead of being multiplied by hitting the surface of a dynode, where secondary electrons are released on the same side. These membranes are constructed using insulators and can charge up, as more electrons leave the membrane than enter. To circumvent this issue a conductive layer is applied on these membranes. The advantage to use graphene over TiN as conductive layer would be that the maximum transmission electron yield is higher and occurs at a lower incoming (primary) electron energy due to the reduced total thickness of the membrane. In order to get graphene on the membranes a wet transfer method of graphene was used where a layer of Poly(methyl methacrylate) (PMMA) was applied to support and the graphene sheet and to make it buoyant.
The adhesion of graphene to alumina membranes proved to be very difficult and on many occasions the graphene was washed away during the removal of the polymer by acetone. The few remaining samples that were produced on both alumina and magnesium oxide membranes, where the conductive layer was successfully applied, had a lower transmission electron yield than the samples coated with a sputtered TiN layer. Sputtered TiN was used as conductive material in this project, before graphene was researched for this thesis. The reason for the lower yield is that the adhesion between the conductive graphene and the membranes was not good enough to conduct electrons vertically in the membranes causing them to charge up. One of the reasons for the poor adhesion is that the membranes are wrinkled and curved due to internal stresses that are caused by the production process. The graphene layer cannot follow these curves due to the way it is applied and only makes contact at the tops of the wrinkles. The primary electron energy, where the (lower) maximum yield was observed, was higher than samples with TiN, while the expectation was that graphene as the conductive layer would lower this energy, due to the reduced thickness. These observations lead to the conclusion that graphene is not a viable replacement for TiN as the conductive layer on these membranes.
The use of TiN deposited by Atomic Layer Deposition (ALD) instead of sputtering was also tested. The samples produced with this method conducted electrons well enough to prevent charging effects in the membranes. The added benefit is that the TiN can be deposited with the alumina layer in a single ALD run and that its thickness can be controlled more precisely. The positive results make ALD TiN a viable replacement for sputtered TiN, but more research should be done to find the minimum thickness of this layer where it is still conductive enough to prevent charging of the membranes.
In order to circumvent the abovementioned adhesion issues, a transfer-free method was developed to create graphene-alumina membranes, where the alumina was deposited on the graphene through ALD, instead of transferring graphene on the alumina. Since copper is used as a catalyst to grow the graphene, this put a lot of constraints on suitable wet etching techniques. This resulted in chemical baths where the temperature control was not ideal, which lead to much longer silicon etch times and loss of wafers. Near the end of the process a silicon oxide layer needed to be removed, but this step was most likely not performed correctly, leaving a thin layer of silicon oxide that blocked chemicals that should remove a deeper layer to release the membranes. Further testing of these samples fell out of the scope of this thesis and could be researched in the future. In conclusion, it is advisable to etch the silicon earlier in the process when the copper has not been deposited yet. In general the flowchart of this process needs a revision.

Laser interferometric Gravitational Wave telescopes are limited in sensitivity by seismic noise in the low frequency region (below a 10 Hz). The aim of this research is to improve performance of an advanced displacement sensor and an ultra-low noise inertial sensor, that can help measuring and correcting for seismically induced motion of the GW detetctor’s test masses. In a Rasnik, light from a back-illuminated encoded ChessField mask is projected onto a pixel image sensor, and the relative position is tracked between mask and sensor. In past experiments, Rasnik was modelled to have resolution of below ∼0.5 nm in X and Y, while RMS resolutions of ∼25 nm were measured [1], a longstanding discrepancy central to this research. It was found that, analyzing spectral information of Rasnik response data, Rasnik did in fact reach its sub-nm modelled noise floor. A resolution floor of 퐴፧ = 270 pm/√Hz was measured, shown to be limited by SNR of the pixel content and white noise aliasing. What is more, with a different setup, an optical magnification trick was shown to result in RMS resolution gain. For the forced-feedback accelerometer with interferometric readout, Heijningen [2] measured displacement sensitivities of up to 8 fm/√Hz above 30 Hz, while the instrument was modelled to be shot noise limited at 3 fm/√Hz - another discrepancy this research attempted to explain. A significant reduction of the thermal noise of the instrument’s mechanics was achieved by re-designing the feedback actuator. After that, the accelerometer’s performance was re-evaluated on a ultra-low vibration platform. A resolution of 50 fm/√Hz was measured, a factor 30 above the modelled shot noise. Sub-optimal relative intensity fluctuations in the laser, as well as laser frequency noise, were found to be responsible.