Você está na página 1de 5

Response to Referees: LK13508BR/Fong

Response to Referee 4

We are delighted that Referee 4 finds our manuscript well written and incorporates sufficient scientific advance for publication in PRB. We address all of his/her comments in the following.

Referee 4s comment (1): The cubic correction term is given as K_3 = pi^2 E/ 8 sigma L^2; is it possible that the 8 should be a 4; as easily derived form e.g. APL 94, 263104 (2009) (eq.1), which could be cited (page 4). Our response: We have double checked our calculation and we agree with Referee 4 that the nonlinear coefficient K_3 should be pi^2 E/4 sigma L^2. We have corrected it in the revised manuscript and added the suggested citation. We thank Referee 4 for pointing out the error.

Referee 4s comment (2): The phase of an oscillation as a function of driving frequency undergoes 180 degrees near resonance; the higher the quality factor, the steeper is this curve near resonance. This very well known behavior could be mentioned instead of the somewhat complicated transfer function in order to explain why high frequency oscillators display more phase noise if the resonance frequency fluctuates. (page 8) Our response: We agree with Referee 4 that the higher phase noise in higher Q oscillator can be understood as the result of steeper slope in the phase-frequency curve. We have added a sentence to describe this intuitive observation. We keep the discussion about the transfer function since it gives extra information about the oscillator dynamics in frequency domain.

Referee 4s comment (3): It is hard to determine from the figures whether the frequency fluctuates or drifts with time; a sentence clarifying this should be added. Our response: Conventionally, fluctuation can be considered as a noise oscillating around a well-defined mean value, while drift can be understood as a slow fluctuation which diverges from the mean value in long time scale. However, for any physical noise process within a finite time span, the total power must be finite and a mean value can always be redefined. Therefore, both the terms fluctuation and drift are vague descriptions of a noise process and there is no clear boundary between them. To avoid these kinds of syntactic ambiguities, a quantitative

description such as spectral power density is necessary. In our manuscript, we quantitatively show that our frequency noise spectrum has a 1/f dependence. Since it is a physical process, the total noise power cannot diverge, which means that there must be a lower cutoff frequency beyond which the power spectrum remains finite. Whether to call it a fluctuation or a drift is a matter of choice of words. Drift due to slow environmental vibrations does happen. It can be compensated in experiment by applying feedback. We deal with frequency noise in this feedback bandwidth.

Response to Referee 5

We are glad that Referee 5 finds our results interesting for those working in the field of precision micromechanical and nanomechanical systems. All of his/her comments are addressed in the following.

Referee 5s comment (1): The authors state near the top of page four that "thermomechanical noise is resolved at all temperatures concerned in this work and thus allows us to calibrate the displacement of the resonator". I interpret this statement to imply that there is no independent calibration of the resonator displacement from their measured optical response. That is, they simply assume that the resonator is in thermal equilibrium with its environment, and use the magnitude of the measured noise to infer the mechanical displacement from the measured optical response. Fine - this is done by many others in the field of cavity opto-mechanics, but doesn't this calibration method then imply that their PSD amplitude spectrum in Fig. 3a will, by definition, exactly match that of a thermal noise calculation? To use an assumption of thermal noise to calibrate displacement, and then to claim that there is good agreement between the displacement noise amplitude and thermal noise is a circular argument. Our response: We agree with Referee 5 that the calibration method we used implies, by definition, the PSD amplitude spectrum will match that from thermal noise calculation. Nevertheless, we would like to point out that throughout our manuscript, the claim of the agreement between observed amplitude noise and thermal noise is always accompanied with the claim of the disagreement between the observed phase noise and the thermal noise. The former claim is a direct consequence of the calibration procedure while the latter claim is our major result and is highly non-trivial. On the other hand, the calibration process only provides a norm to the spectrum; it does not affect the frequency dependence of the spectrum. Referee 5s comment (2): In the discussion of mechanical nonlinearity at the end of the last paragraph on p. 4, the authors describe a stress, sigma, in the beam. But in a suspended beam like this, the stress is highly anisotropic. That is sigma_xx does not equal sigma_yy does not equal sigma_zz. What value of the stress tensor are the authors using their equation for K_3? Also, Young's modulus depends on the exact type of SiN deposited. What value do the authors use? Our response: Since the beam is clamped only in the direction along the length (say, xdirection), the stress in the other two directions is relaxed and becomes negligibly small compared to sigma_xx. Therefore, the stress used in the equation for K_3 is sigma_xx. The same

equation is also derived in the newly added Ref. [21] (citation suggested by Referee 4). In literature, the Youngs modulus of stoichiometric silicon nitride is found to range from 250GPa to 300GPa. The Youngs modulus of our film is 270GPa.

Referee 5s comment (3): I do not understand why in Fig. 2 the authors use the in-phase and quadrature components of their signal instead of the more physically intuitive R and Theta, accompanied by more standard polar plots. Indeed, in Fig. 3, the authors convert to R and Theta (amplitude and phase) to compare to their noise models. Our response: The two kinds of plots that Referee 5 pointed out are, by definition, equivalent. Plots of in-phase vs quadrature components are commonly found in literature of nanomechanics (see, for example, Nat. Phys. 6, 213 (2010), PRL 67, 699 (1991)). As a matter of fact, the inphase and quadrature components are the direct measured quantities in a lock-in amplifier, which are then used to calculate amplitude and phase.

Referee 5s comment (4): In their discussion of the potential frequency noise sources to explain the discrepancy with thermal noise models, the authors do not discuss the frequency noise of their drive source. The authors should justify their belief that frequency/phase noise of their drive source plays no significant role in their measured frequency noise spectrum. Our response: Following Referee 5s suggestion, we have added a description of the phase noise of our drive source. In summary, the phase noise of our drive source (Zurich Instruments HF2LI [22]) is 6 x 10-7 rad2/Hz at f = 1Hz (measured by the company and independently confirmed by us), which is about two orders of magnitude lower than the observed phase noise in our device.

Referee 5s comment (5): In the second sentence of the last paragraph on p. 3, the authors state that the effective mass is defined as "half of the total mass". This is not the definition of the effective mass. Instead, the effective mass is defined as omega^2 = Keff/meff, where omega = 2*pi*(resonantfreq) and Keff is the effective stiffness. In clamped-clamped beams with no strain, the effective mass is close to half of the total mass (but not exactly), but the effective mass can diverge significantly from this value in the presence of strain. Our response: As is well known, there is no unique way of defining effective mass and effective stiffness. The definition of effective mass depends on the choice of the mechanical mode normalization. Details of different effective mass definitions can be found in the section 3.1 of Ref [3]. In our case, we choose the mode profile to be sin(pi z / L) and therefore the displacement of the middle of the beam (z = L/2) is normalized to 1. Under this normalization

condition, the effective mass is given by half of the total mass. To avoid confusion, we have rephrased our description to clarify our definition of effective mass in the revised manuscript.

Referee 5s comment (6): There are still instances in the manuscript of grammatical errors, such as the first sentence of the second paragraph (p. 2). Please have a native English speaker proof-read it. Our response: We thank for Referee 5 for pointing out the grammatical errors in our manuscript. We have carefully proofread it and have corrected the grammatical errors that we found.

Você também pode gostar