Signal-to-Noise Ratio improves in Critical-Point Flexure Biosensors due to Intrinsic Low Pass Filtering S1: Time Domain Simulation Framework for Force Noise The time domain response of Flexure sensor in presence of white thermo-mechanical force noise is modeled using Newton’s equation, given byπ ππ£ π0 ππΏππ΄2 + ππ£ = π(π¦0 − π¦) − + πΉπ (π‘), ππ‘ 2π¦ 2 (π1π) ππ¦ = π£. ππ‘ (π1π) Here, π is the mass of movable electrode, π£ is the velocity, π‘ is time, π = ππ0 /π is the damping coefficient. πΉπ (π‘) is the random noise force with autocorrelation function β¨πΉπ (π‘)πΉπ (π‘ ′ )β© = 2ππ΅ πππΏ(π‘ − π‘ ′ ) and one sided power spectral density ππΉ (π) = 4ππ΅ ππ. It is also important to note that, the white Gaussian noise force is πΉπ (π‘) = √2ππ΅ ππ ππ£ Using the definition ππ‘ = (1 + π£(π‘+Δπ‘)−π£(π‘) Δπ‘ and ππ¦ ππ‘ = ππ(π‘) , ππ‘ where π(π‘) is the Brownian process. π¦(π‘+Δπ‘)−π¦(π‘) , Δπ‘ Eq. S1 can be written as follows- πΔπ‘ πΔπ‘ π0 ππΏππ΄2 Δπ‘ πππ₯π‘ (π‘) ) π£(π‘ + Δπ‘) = π£(π‘) + (π¦0 − π¦(π‘)) − + √2ππ΅ ππ , 2 π π 2ππ¦ π π¦(π‘ + Δπ‘) = π¦(π‘) + π£(π‘ + Δπ‘)Δπ‘, (π2π) (π2π) where π£(π‘) and π£(π‘ + Δπ‘) are the values of velocity, at time π‘ and π‘ + Δπ‘, respectively. Similarly, π¦(π‘) and π¦(π‘ + Δπ‘) are the values of electrode positions, at time π‘ and π‘ + Δπ‘, respectively; here, Δπ‘ is the time step. Most importantly, ππΔπ‘ (π‘) is a random variable that is normal distributed with mean zero and standard deviation √Δπ‘. Knowing the initial condition i.e., π¦(0) and π£(0), Eq. S2 can be solved for any ππ΄ . At every instant π‘, a new random variable ππΔπ‘ (π‘) is generated to evaluate Eq. S2. S2: Time Domain Simulation Framework for Stiffness Noise The time domain response of Flexure sensor in presence of white stiffness noise is also modeled using Newton’s equation, given byπ ππ£ π0 ππΏππ΄2 + ππ£ = (π + Δππ (π‘))(π¦0 − π¦) − , ππ‘ 2π¦ 2 ππ¦ = π£, ππ‘ (π3π) (π3π) where Δππ (π‘) is the random noise due to stiffness fluctuations with autocorrelation β¨Δππ (π‘)Δππ (π‘′)β© = 0.5ππ (π‘ − π‘ ′ ) and one sided power spectral density ππ (π) = ππ . It is also important to note that white stiffness fluctuations are Δππ (π‘) = √0.5ππ ππ(π‘) . ππ‘ For numerical simulations, Eq. S3 can be written as follows(1 + πΔπ‘ πΔπ‘ π0 ππΏππ΄2 Δπ‘ √0.5πΔπ (π¦0 − π¦(π‘))πππ₯π‘ (π‘) ) π£(π‘ + Δπ‘) = π£(π‘) + + , (π¦0 − π¦(π‘)) − π π 2ππ¦ 2 π (π4π) π¦(π‘ + Δπ‘) = π¦(π‘) + π£(π‘ + Δπ‘)Δπ‘. (π4π) Equation S4 can be now solved for any ππ΄ to evaluate the noise response due to stiffness fluctuations. S3: Numerical Simulations Figures S1-S2 show the results of time domain stochastic simulations (see Table S1 for parameters used) for thermo-mechanical noise and stiffness noise due to temperature fluctuations, respectively. For the two cases, Eqs. S2 & S4 have been solved, respectively. We simulated the noise response at different voltages. For the specific voltage of ππ΄ = 0.9πππΌ , the results are summarized in Figs. S1a-d and Figs. S2a-d. Figures S1a-b show the fluctuations in the position of electrode on the potential energy landscape. Each symbol denotes the total energy during fluctuations. As expected, electrode does random thermal vibration around its equilibrium position, as shown in Fig. S1c. Figure S1d shows the corresponding sample average of root mean square fluctuations i.e., 2 1 π=π π Δπ¦π (π‘) = √π ∑π=1 π (π¦π (π‘) − π¦ππππ (π‘)) . π π (π‘) = π¦ππππ 1 ππ π ∑π=π π=1 π¦π (π‘). π Here, π¦π (π‘) denote the position of electrode during π π‘β simulation at time π‘ and π¦ππππ (π‘) is the corresponding mean position. ππ (1000 in this article) is the number of simulations performed to calculate the statistical average. Interestingly, Δ π¦π (π‘) starts from zero and then saturates to an equilibrium value (solid dot in Fig. S1d), which is nothing but the average noise power. Figure S1e compares the results obtained from time domain simulations (symbols) with the ones obtained from transfer function based analysis (solid line). In spite of the presence of the highly nonlinear electrostatic force, the results match because the fluctuations are small (Fig. S1d and Fig. S2d), thus justifying linearization around the equilibrium value (Eq. 2 and Eq. 3 in the main text) for transfer function based analysis. Having said that, as we go closer to the pull-in voltage, fluctuations increase considerably, eventually leading to noise initiated pull-in. Therefore, the linear transfer function based analysis is valid so long as we are below safe operating voltage. Note that, similar results and similar matching between time domain and transfer function analysis is achieved for stiffness noise as well (Fig. S2). Parameter Value π 1ππ πΏ 4ππ π» 40ππ πΈ 200πΊππ π 0.31 π 8912πΎπ/π3 π¦0 100ππ π 7.4 × 10−6 π/πΎ 1 ππ π ππ 10−3 /πΎ ππ 30ππ Table S1: Parameters used for calculation of SNR and LOD in the main text . Fig: S1: Time domain stochastic numerical simulations of thermo-mechanical noise. (a)-(b) Fluctuations of movable electrode position shown on the potential energy landscape. The region in the oval has been zoomed in Fig. S1b. Symbols denote the total energy (kinetic + potential) of the electrode. (c) Position of electrode as a function of time. π¦π denote the equilibrium position. (d) Root mean square fluctuations as a function of time. (e) Equilibrium value of root mean square fluctuations is the average noise power. Symbols denote the results from time domain numerical simulations; whereas solid line denote the calculations from linear transfer function based analysis (Eq. 5b in the main text). Fig: S2: Time domain stochastic numerical simulations of stiffness noise due to temperature fluctuations. (a)-(b) Fluctuations of movable electrode position shown on the potential energy landscape. The region in the oval has been zoomed in Fig. S2b. Dotted black curve corresponds to the maximum stiffness; whereas magenta dotted to minimum stiffness. Symbols denote the total energy (kinetic + potential) of the electrode. (c) Position of electrode as a function of time. π¦π denote the equilibrium position. (d) Root mean square fluctuations as a function of time. (e) Equilibrium value of root mean square fluctuations is the average noise power. Symbols denote the results from time domain numerical simulations; whereas solid line denote the calculations from linear transfer function based analysis (Eq. 6b in the main text). S4: Safe Operating Voltage to avoid Noise Initiated Pull-in In the main text, we argued that biasing close to pull-in improves πππ and πΏππ·. Here, we answer a very important and fundamental question regarding the stability of critical-point Flexure sensors close to pull-in point. The question is “how close to the pull-in point can one operate without making the sensor unstable?” Note that, in Figs. 3-4 in the main text, ππ΄ was swept from ππ΄ = 0 to ππ΄ = 0.995πππΌ . We should check if biasing at ππ΄ = 0.995πππΌ is feasible. To answer this question, we look at the behavior of movable electrode in response to both force and stiffness noise using time domain stochastic simulations. 1 Figure S3a shows the potential energy (π = 2 π(π¦0 − π¦)2 − π0 ππΏ 2 ππ΄ ) 2π¦ landscape of Flexure sensor at ππ΄ = 0.995πππΌ . Movable electrode is stabilized at the minimum of π. In absence of noise, movable electrode should have remained at the bottom of potential energy well in Fig. S3a (see the dotted line in Fig. S3b also). However, thermo-mechanical noise exerts random force on the electrode making it fluctuate (reason for noise characterized by Δπ¦π ) around its equilibrium position as shown in Figs. S3a-b. Symbols in Fig. S3a denote total energy (kinetic + potential) of movable electrode during fluctuations. Due to the presence of a high energy barrier Δππ = π(π¦π’ ) − π(π¦π ) ≈ 3.75 × 103 ππ΅ π, (π¦π : stable equilibrium position and π¦π’ : unstable equilibrium position) the movable electrode only fluctuates around the bottom of potential well in Fig. S3a, but cannot surmount the energy barrier to make the system unstable. On the other hand, stiffness noise due to temperature fluctuations will make the potential energy landscape fluctuate as shown in Fig. S3c. Dotted black line corresponds to the potential energy profile for maximum stiffness; whereas dotted magenta for minimum stiffness. Due to the fluctuations in the stiffness, the position of the electrode fluctuates around its equilibrium position as shown in Figs. S3c-d. Once again, the stiffness fluctuations are not strong enough to make the electrode pull-in (Figs. S3c-d). Therefore, we will classify ππ΄ = 0.995πππΌ as the safe operating voltage. (a) (b) (c) (d) Fig. S3: Results of time domain stochastic simulations of a Flexure sensor at ππ΄ = 0.995πππΌ with Δππ ≈ 3.75 × 103 ππ΅ π, due to (a)-(b) thermo-mechanical noise and (c)-(d) temperature fluctuations stiffness noise. ππ denote the potential energy at equilibrium position i.e., at the bottom of potential well. Symbols in Figs. S3a & c denote the total energy (kinetic + potential). Dotted black line in Fig. S3c correspond to maximum stiffness; whereas magenta line to minimum stiffness. Inset in Figs. S3a &c show the zoomed region around the bottom of potential well. Note that, if ππ΄ is increased further, Δππ decreases and becomes Δππ ≈ 5ππ΅ π at ππ΄ = 0.99994πππΌ as shown in Fig. S4a. In this case, the movable electrode gets the sufficient energy from the surrounding to surmount the energy barrier and gets pulled-in as shown in Figs. S4a-b. Therefore, ππ΄ = 0.99994πππΌ cannot be classified as safe operating voltage. Interestingly, pull-in at ππ΄ = 0.99994πππΌ occurs due to thermomechanical noise, and not because of stiffness noise. However, if the voltage is increased even further, pullin can occur due to stiffness noise because of temperature fluctuations as shown in Figs. S4c-d. The bottom line from this section is that as we go closer and closer to pull-in point, chances of noise initiated pull-in increases. A safe operating voltage is that does not cause noise initiated pull-in or at least not in the time duration of measurement. (a) (b) (c) (d) maximum minimum Fig. S4: Noise initiated pull-in due to (a)-(b) thermo-mechanical noise at ππ΄ = 0.99994πππΌ with Δππ ≈ 5ππ΅ π and (c)-(d) stiffness noise due to temperature fluctuations at ππ΄ = 0.999995πππΌ . π¦π corresponds to the stable equilibrium position; whereas π¦π’ unstable.