The OPA is a work in progress but has so far produced 11 dB of squeezing when corrected for dark noise as shown in Fig. 3.4. The total intra-cavity loss of0.3%was measured and attributed mainly to the nonlinear crystal. This was measured using a high finesse cavity and comparing the calculated finesse of the cavity without the crystal to the measured finesse with the crystal. Using the measured cavity parameters a normalized pump param- eter ofβn = 0.96was inferred. The theoretical curves of the squeezed and anti-squeezed
quadratures from Eq. (2.45) using these parameters are plotted along side the measured traces in Fig. 3.4. With the poor mode matching of the pump to the cavity mode the first attempt to measure the squeezing spectrum only achieved 9 dB of squeezing. A second attempt with better mode matching resulted in 11 dB. The difference between the two recorded squeezing levels is likely due to the poor locking technique introducing noise into the measurement of the squeezed quadrature from the anti-squeezed quadrature rather than loss. This result falls short of the predicted 13dB from the measured parameters. To reach 13 dB of squeezing each locking loop in the experiment will be required to be care- fully optimized to minimize noise. To measure greater than 15 dB the intra-cavity losses will need to be improved. To further characterise the OPA, the pump threshold power was measured. This was done by injecting a seed into the cavity and measuring the OPA gain while sweeping the pump power. The results of these measurements are shown in Fig. 3.5. This measurement was only made for the amplification gain. Despite best practice for experimental physics each data point was only recorded once so no meaningful error bars can be found for this data. However, two sets of measurements were made as shown in Fig. 3.5 showing the data is repeatable. From fitting the data the pump threshold power was found to be 624mW with a maximum regenerative gain of 1420. The predicted pump threshold power was calculated to be 685mW. This value was calculated with the non- linear interactionΛ = 271found using manufacturer supplied parameters and measured parameters from the SHG cavity. The discrepancy between the predicted and measured values is mostly likely due to a better alignment of the ppKTP crystal inside the OPA cav- ity and hence the actual non-linear interaction strength is higher than in the SHG system. More time will need to be invested into the method used to lock the homodyne de- tector to the squeezed and anti-squeezed quadratures. This will improve the observable squeezing. As it stands the current method requires patience for the microcontroller lock to stay locked on the correct quadrature for long enough to take a measurement. To try to reduce the fluctuations in the squeezed spectrum from the bad locking the data in Fig. 3.4 is averaged over 10 sweeps. Two other methods of locking are typically used here, chop
52 Squeezed State Generation at 1550nm 0 1 2 3 4 5 6 7 8 9 10 −10 −5 0 5 10 15 f (MHz) dB
Figure 3.4: Plot of quadrature variance of the squeezed light from the OPA. The squeezed quadrature was measured with a poorly mode matched pump (yellow) and then again with the pump optimally mode matched (purple). The anti-squeezed quadrature (green) was measured with the poor mode matching but demonstrated a normalised pump parameter ofβn = 0.96. Each trace was recorded using a spectrum analyser with a 30kHz resolution
bandwidth and 1kHz video bandwidth. Eq. (2.45) is plotted for both the squeezed (blue) and anti-squeezed (red) quadratures using the measured loss of the non-linear crystal. A maximum squeezing of 13 dB is predicted for this OPA.
locking and coherent locking. Chop locking cycles a seed on and off either using AOMs or an optical chopper. The seed is provided long enough to stabilize the cavity and homo- dyne locks. The control signals are then held when the seed is removed [47]. Coherent locking uses a frequency shifted beam locked in phase with the pump to lock the homo- dyne to a quadrature [67]. The scheme can either be built using a pair of AOMs to provide the frequency shifted beams [68] or a second auxiliary laser.
The motivation for pursuing highly squeezed and pure CV states comes from their application to advanced quantum protocols and error correcting codes. One of the pro- tocols only accessible with a low intra-cavity loss OPA is discussed in Ch. 7. Another obvious application for highly squeezed state is in error correction. For error correction to be successful it has to guarantee a low probability of failure. In quantum computing this probability relates to the reliability of the computations being performed. To phrase this another way, what would be the minimum level of squeezing required to perform reli- able quantum computing? So far the answer to this question is 20.5dB [69]. Though it is expected that other protocols and QEC will appear with a requirement for less squeezing.
Conclusion 53 0 100 200 300 400 500 600 100 101 102 103 Ppump(mW) gr
Figure 3.5: A plot of the pump power vs the regenerative gain. Two separate measure- ments were taken of the regenerative gain while sweeping the pump power (yellow and green). A curve was fitted (red, solid) to the measurements to give a pump threshold power of 624mW. The expected deamplified gain is also plotted as the red dashed line. The calculated pump threshold power was greater than the measured value (blue).
The main goal of the observation of the record breaking 15dB is to increase the sen- sitivity of the GEO 600 interferometer for gravitational wave detection [56]. A 10 dB en- hancement would require less than10%photon loss in the squeezed field. Another novel application from Ref. [56] was to use the highly squeezed states to calibrate the quan- tum efficiency of a photodiode. For their particular diode they measured an efficiency of
99.5%with an uncertainty of0.5%without the need for a calibrated light source.
High levels of squeezing can also be used to provide incrementally improved results. One project lined up will combine a squeezing gate [70, 71] with a measurement based NLA [72] to create a probabilistic squeezed gate. It is expected that with the NLA the squeezed gate will be able to achieve a fidelity of 1 for higher levels of squeezing than was achievable in the original experiment [73].