Supplementary Information
I. Trigger Circuitry
The function of trigger circuitry is to produce signals to control perturbation and trigger a
digitizer, in a manner that is phase synchronous with the deflection signal coming from the AFM
cantilever. The controller accepts as inputs the deflection signal (DEF) and a trigger request
signal (REQ). Once the trigger request is asserted, the controller delays asserting the output
signals until the deflection signal reaches a specified phase position. The controller’s phase
resolution is 8 bits, 128 values in [0°, 360°).
Trigger circuitry is composed of two separate hardware modules: the optical isolator that
receives the reference and trigger request signals from the AFM chassis, and the phase shift
controller that generates a signal (OUT) to trigger the digitizer and activate the perturbation. The
two modules are connected by a pair of optical fibers, in order to minimize feedbacks in
circuitry. The phase shift controller is implemented using a phase locked loop (PLL) as
illustrated in Figure S1.
FIG. S 1. (a) PLL circuit for phase shift controller. DEF, REQ, and OUT are the deflection signal from
cantilever, the trigger request and the signal to trigger digitizer and the perturbation, respectively. (b) The
timing diagram illustrates a case where the output is asserted 180° after the rising edge of the deflection
signal.
1
II. Data Analysis: Finding Minimum
In order to find minimum of a curve, one can use many different methods. Here, due to the
noisy nature of signal, we prefer to fit the signal with an analytical approximation of the curve
and use the fit parameters to calculate the minimum. The analytical approximation is composed
of two parts: decay and relaxation. These are defined as follows:
𝜔𝐷 (𝑡) = 𝐴𝑒
−
𝜔𝑅 (𝑡) = 𝐵 (1 −
𝑡
𝜏1
,
𝑡
−
𝑒 𝜏2 ) ,
(1)
(2)
where A and B are normalization constants; τ1 and τ2 are time constants for decay and relaxation
respectively. We fit the signal with:
𝜔(𝑡) = 𝐶𝑒
−
𝑡
𝜏1
(1 − 𝑒
The minimum of this equation is tFP and it is:
𝑡𝐹𝑃 = 𝜏2 ln (
−
𝑡
𝜏2 ) ,
where 𝑡 > 0.
𝜏1 + 𝜏2
).
𝜏2
(3)
(4)
III. Sample: Gold Electrode
In the experiments, we have used a silicon wafer sputtered with a thin gold film that is
attached to the waveform generator using conductive Silver paint (Leitsilber 200 Silver Paint).
The surface roughness of the gold electrode is measured to be 2.29 nm.
FIG. S 2. Topography image of the gold electrode surface.
2
IV. Instantaneous Frequency: Experiment and Simulation
FIG. S 3. Experimental curves for the cantilever with f0 = 272.218 kHz, k = 26.2 N/m and Q = 432. Excitation
has a rise time of 100 ns and the Vbias = 10 V at 50 nm lift height. (a) Instantaneous frequency after trigger for
excitation at different phases, (b) frequency shifts and (c) tFP’s at different phases.
FIG. S 4. Simulated curves for the cantilever with f0 = 272.218 kHz, k = 26.2 N/m and Q = 432. Excitation has
a rise time of 100 ns and the Vbias = 10 V at 50 nm lift height. (a) Instantaneous frequency after trigger for
excitation at different phases, (b) frequency shifts and (c) tFP’s at different phases.
3
V. Spectrum Analysis
We analyzed the power spectrum of different cantilevers by measuring their thermal spectra
and tuning them using their thermally measured properties. After tuning, we drove tips as we
collected data. For BlueDrive measurements, we measured properties of tips while the laser was
on, but not pulsing.
For different cantilevers, we found that their resonance frequency gets slightly smaller (~10
Hz) when we use BlueDrive. Also, BlueDrive has better signal-to-noise ratio compared to
piezoacoustic drive (bandwidth of 5 kHz around the resonant frequency).
ElectriMulti75-G (75 kHz)
ElectriTap190-G (190 kHz)
ElectriTap300-G (300 kHz)
DDESP-V2 (450 kHz)
BlueDrive
61.12 dB
64.38 dB
63.38 dB
64.20 dB
Piezoacoustic Drive
61.05 dB
62.48 dB
54.76 dB
58.81 dB
Table S 1. The signal-to-noise ratio for different cantilevers used in the experiments when they are driven by
BlueDrive and piezoacoustic drive. The calculation was done around the cantilever's resonant frequency with
5 kHz bandwidth.
FIG. S 5. Amplitude spectral density for a cantilever with f0 = 285.090 kHz, k = 25.2 N/m and Q = 451. Series
of peaks are the response of instrument’s electronics. Inset shows a zoom-in around the resonant frequency.
Note the difference between baselines.
4
VI. Frequency Dependence
FIG. S 6. Effect of different resonance frequencies on tFP measurements. (b) Zoom-in of
Fig. 10(c) showing that the 256 kHz cantilever’s standard deviations are much larger
causing minimum distinguishable rise time to be bigger than the 503 kHz tip. The scale is
shifted to make error bars visible. The cantilever parameters are f0 = 60.219 kHz, k = 1.37
N/m, Q = 150; f0 = 154.806 kHz, k = 15.8 N/m and Q = 352; f0 = 255.920 kHz, k = 17.9 N/m
and Q = 428; and f0 = 503.327 kHz, k = 72.7 N/m and Q = 499. AFM cantilevers were
excited at 180° using BlueDrive. The µtFP and σtFP values are compiled from 30 different
runs of the same experiment. Every run consists of 2000 averages of raw cantilever
deflection signal.
5
VII. Height and Voltage Dependence
FIG. S 7. Comparison of different lift heights showing experimental (a) frequency shifts and (b) tFP’s for an
AFM cantilever excited at 180°. The cantilever parameters are f0 = 255.920 kHz, k = 17.9 N/m and Q = 428.
(c) Zoom in for faster decays. The µtFP and σtFP values are compiled from 30 different runs of the same
experiment. Every run consists of 2000 averages of raw cantilever deflection signal.
FIG. S 8. Comparison of different biases showing experimental (a) frequency shifts and (b) tFP’s for an AFM
cantilever excited at 180°. The cantilever parameters are f0 = 255.920 kHz, k = 17.9 N/m and Q = 428. (c)
Zoom in for faster decays. The µtFP and σtFP values are compiled from 30 different runs of the same
experiment. Every run consists of 2000 averages of raw cantilever deflection signal.
6
FIG. S 9. (a) MDMO-PPV:PCBM image from Figure 13 (main text) with the frequency shift converted into
dFe/dz (force gradient) with units of mN/m via equation 4. (b) Instantaneous frequency vs time data for
varying light intensities with a 488 nm illumination source. Light intensities were measured with a calibrated
photodiode with an AFM cantilever in place to account for possible shadowing.
Figure S9 shows the data from the same set as used in Figure 13 in the main text. Here, the
frequency shift is shown. The frequency shift is the y-axis value at the tFP time, e.g. in Figure
S9B the frequency shift for the red curve is approximately -300 Hz. This value can be converted
to dFe/dz using equation 4 in the main text as long as the spring constant is known. Here, the
spring constant is ~17 N/m.
Figure S9B shows the intensity-dependent instantaneous frequency on the same materials. As
expected, the tFP shifts to faster times (longer 1/tFP values) as the light intensity increases. For the
data on MDMO-PPV:PCBM prepared in an identical fashion in Ref. 20 (Cox, et al., Journal of
Physical Chemistry Letters 6, 2852 (2015)), Figure S2 in that report indicates that light intensity
of only ~0.6 W/m2 could be used before the system would saturate due to using a feedback loop.
Here in Figure S9 we can easily use light intensities closer to typical solar intensities (1000
W/m2) as well as higher, an over three order of magnitude improvement (we are comparing our
use of 488 nm illumination to the 455 nm illumination in that work, where the charging rate is
similar due to the external quantum efficiency in that regime).
7