SUPPLEMENTARY INFORMATION
doi: 10.1038/nnano.2009.152
nature nanotechnology
| www.nature.com/naturenanotechnology
1
SUPPLEMENTARY
INFORMATION
:
Towards single-molecule nanomechanical
mass spectrometry
v. 9
A. K. Naik
*1
, M. S. Hanay*
1
, W. K. Hiebert*
1,2
, X. L. Feng
1
, M. L. Roukes
†1
1
Kavli Nanoscience Institute, California Inst
itute of Technology, MC 114-36, Pasadena, CA
91125 USA
2
National Institute for Nanotechnology, Natio
nal Research Council of Canada, Edmonton,
Alberta T6G 2M9 Canada
NEMS-MS System Overview
We employ proven methodologies fro
m state-of-the-art mass spect
rometry to build a novel MS
system using ultra high frequency NEMS
mass sensors. Components include a room-
temperature, atmospheric pressure electrospray i
onization (ESI) system fo
r creating protein ions
or charged nanoparticles, a two-stage different
ial vacuum system, RF hexapole ion optics to
guide the charged analytes to the detector, an
d the NEMS mass detector stage. These are
assembled to form a hybrid system comp
rising both custom built and commercial
instrumentation.
Figure S1 is a montage of images depicting our
first experimental prototype system for NEMS-
MS enabling the introduction, tr
ansport, and mass measurements
on individual proteins and
nanoparticles. Protein ions or charged nanoparticles are produced
using electrospray ionization
(ESI) and delivered to a hexapole ion guide dr
iven at radio frequencies (RF), which then
transports these species to the
NEMS mass sensor with minimal
m/z
discrimination, as desired.
The detection circuitry utilizes a bridge circuit to null the background near the NEMS
Figure
S1.
First
Generation
NEMS-MS
System.
(a) The cryostat,
its vibration isolation &
support system, and (in pit)
the
super-conducting
magnet, and its dewar. (b)
Electrospray
ioniza-tion
unit at top of system. (c)
Hexapole ion guide from
bottom. (d)(e) Sample
stage region; progres-sive
magnifications.
(f)
Hexapole ion guide; outlet
detail.
*
These authors contributed equally to this work.
©
200
9
Macmillan Publishers Limited. All rights reserved.
2
nature nanotechnology
| www.nature.com/naturenanotechnology
SUPPLEMENTARY
INFORMATION
doi: 10.1038/nnano
.2009.152
Supplementary
Information:
“Towards
single
‐
molecule
nanomechanical
mass
spectrometry”
Naik,
et
al.
resonance
1-3
and a frequency-modulated, phase-locked loop (FM-PLL) to track the NEMS
resonant frequency in real time
4
.
Biological sample preparation
Bovine Serum Albumin (BSA),
β
-Amylase and 5-nm (nominal) diameter colloidal gold
nanoparticles were obtained from Sigma-Aldric
h and used without further purification. The
solutions employed for ESI are as follows:
1.
Bovine Serum Albumin (Mol. Wt.=66kDa) : 25 μM solution in 95/5 v% H
2
O/HAc
2.
β
-Amylase (Mol. Wt.=200kDa) : 0.2 μM in pH~6.5 10mM NH
4
Ac buffer in 50/50 v%
H
2
O/MeOH
3.
Sigma-Aldrich colloidal gold nanoparticle so
lution (G1402) diluted ten-fold by 50/50 v%
H
2
O/MeOH
Electrospray Injection (ESI) an
d differential vacuum assembly
Protein ions and charged nanopart
icles are produced using a comme
rcial electrospray ionization
(ESI) system (Varian). ESI is one of two
well known “soft” ionization processes that can
reliably bring large macromolecules from
the solution phase into the vapor phase
5,6
.
These solutions were introduced using a syri
nge pump (Harvard Apparatus) and syringes
(Hamilton) to the electrospray needle (Agile
nt) by direct infusion through standard MS
components (Upchurch) to achieve typical flow rates of 4
μ
L/min. High-voltage sources (Emco)
are used to bias the ESI needle
at a constant voltage of ~2.5-3kV.
Solvated analytes delivered to
the needle are forced out in the form of charge
d droplets that repel each
other due to coulombic
forces
5,6
. The solvent within these droplets evapor
ates, reducing their volume yet preserving the
amount of charge contained. The increasingl
y unstable microdroplets eventually undergo
“coulomb fission”, fragmenting into daughter dropl
ets, and repeated cycles of this process
ultimately result in the formation of bare io
nized proteins. Although the exact mechanism of
protein ion formation from sma
ll droplets is still under debate
7
, electrospray has become a well-
established technique for produc
ing proteins in vapor form.
Our ESI delivery system is built around a commer
cial sub-assembly (Varian 1200 LC/MS) and
comprises the following components:
1) An ESI needle and gas sheath
mounted in an outer chamber at atmospheric pressure and room
temperature.
2) A 1
st
vacuum stage with a shield plate, counter
flow gas path, capillary, and pumping port.
3) A 2
nd
vacuum stage including a skimmer, collisional-cooling hexapole (“top hexapole”), and
final high vacuum orifice. This sub-assembly is embedded in a custom 2
nd
stage vacuum
chamber designed for about 10 mTorr base pressure for optimal collisional cooling
8,9
of the ions.
It is then attached to the
top plate of the high-vacuum cr
yostat chamber housing the “bottom
hexapole” (see below) and the NEMS sample stage.
All N
2
gas, vacuum port, electronic,
and fluidic lines are connected to
external instrumentation.
†
Corresponding author: roukes@caltech.edu
Version 9
-2-
Supplementary
Information:
“Towards
single
‐
molecule
nanomechanical
mass
spectrometry”
Naik,
et
al.
Ion transport
Electrospray ionization is typically performed at
atmospheric pressure and at room, or elevated,
temperature. This enables the
solvent droplets that form beyond the Taylor cone to quickly
evaporate leaving the bare protein ions they
contain. In this fi
rst-generation NEMS-MS
realization we employ physisorp
tion to capture and immobilize
the analytes on the NEMS
sensor’s surface. By contrast, this requires
that the NEMS be maintained under high-vacuum
and low temperature conditions. Our experime
ntal setup must ther
efore provide a match
between these rather different conditions; we ach
ieve this with a comb
ination of commercially-
available and custom-built cryogenic and diffe
rential-pumping components. In our first-
generation design the NEMS sample
stage is located about
2m away from the ESI source. Since
the capture area of the NEMS sensor
is small, this necessitates efficient transport of the analyte
ions created by electrospray to guide them along a
2m path to the surface of the NEMS sensor
for adsorption.
Transport of ions from atmospheric pressure to
the first differential pum
ping stage is performed
through the so called nozzle-skimmer configuration
10,11
. This produces a
highly collinear,
monochromatic beam of nitrogen molecules and
protein ions. Subsequently, the ions are
transported from this initial vacuum stage to
the NEMS sensor by a hexapole ion guide system,
having an outer radius of ~ 1.2
cm and total length of
~ 2m. The hexapol
e configuration was
chosen as a good compromise between obtaining
high transmission effi
ciency and providing
Parameter
Description
Typical Values
V
needle
Voltage on the electrospray needle
3.0kV
V
L4
Voltage on the electrostatic lens between top and bottom
hexapole
-20V
V
shield
Voltage on the shield (ESI counter electrode)
600V
V
capillary
Voltage on the capillary
200V
V
top_hex
DC offset voltage on the top hexapole
10V
V
bot_hex
DC offset voltage on the bottom hexapole
0V
f
RF
RF frequency of AC voltage applied to the hexapole
1.1MHz, 450kHz
V
RF
Amplitude of the RF voltage applied to hexapole
500V
P
int
Pressure in the intermediate collisional focusing chamber
8mTorr
T
dry
Temperature of the drying gas
180
0
C
P
dry
Pressure of the drying gas
30psi
P
neb
Pressure of the nebulising gas
15psi
R
Protein solution flow rate
4
μ
l/min
Table I.
Typical values used during ESI and ion optics
Version 9
-3-
©
200
9
Macmillan Publishers Limited. All rights reserved.
nature nanotechnology
| www.nature.com/naturenanotechnology
3
SUPPLEMENTARY
INFORMATION
doi: 10.1038/nnano
.2009.152
Supplementary
Information:
“Towards
single
‐
molecule
nanomechanical
mass
spectrometry”
Naik,
et
al.
resonance
1-3
and a frequency-modulated, phase-locked loop (FM-PLL) to track the NEMS
resonant frequency in real time
4
.
Biological sample preparation
Bovine Serum Albumin (BSA),
β
-Amylase and 5-nm (nominal) diameter colloidal gold
nanoparticles were obtained from Sigma-Aldric
h and used without further purification. The
solutions employed for ESI are as follows:
1.
Bovine Serum Albumin (Mol. Wt.=66kDa) : 25 μM solution in 95/5 v% H
2
O/HAc
2.
β
-Amylase (Mol. Wt.=200kDa) : 0.2 μM in pH~6.5 10mM NH
4
Ac buffer in 50/50 v%
H
2
O/MeOH
3.
Sigma-Aldrich colloidal gold nanoparticle so
lution (G1402) diluted ten-fold by 50/50 v%
H
2
O/MeOH
Electrospray Injection (ESI) an
d differential vacuum assembly
Protein ions and charged nanopart
icles are produced using a comme
rcial electrospray ionization
(ESI) system (Varian). ESI is one of two
well known “soft” ionization processes that can
reliably bring large macromolecules from
the solution phase into the vapor phase
5,6
.
These solutions were introduced using a syri
nge pump (Harvard Apparatus) and syringes
(Hamilton) to the electrospray needle (Agile
nt) by direct infusion through standard MS
components (Upchurch) to achieve typical flow rates of 4
μ
L/min. High-voltage sources (Emco)
are used to bias the ESI needle
at a constant voltage of ~2.5-3kV.
Solvated analytes delivered to
the needle are forced out in the form of charge
d droplets that repel each
other due to coulombic
forces
5,6
. The solvent within these droplets evapor
ates, reducing their volume yet preserving the
amount of charge contained. The increasingl
y unstable microdroplets eventually undergo
“coulomb fission”, fragmenting into daughter dropl
ets, and repeated cycles of this process
ultimately result in the formation of bare io
nized proteins. Although the exact mechanism of
protein ion formation from sma
ll droplets is still under debate
7
, electrospray has become a well-
established technique for produc
ing proteins in vapor form.
Our ESI delivery system is built around a commer
cial sub-assembly (Varian 1200 LC/MS) and
comprises the following components:
1) An ESI needle and gas sheath
mounted in an outer chamber at atmospheric pressure and room
temperature.
2) A 1
st
vacuum stage with a shield plate, counter
flow gas path, capillary, and pumping port.
3) A 2
nd
vacuum stage including a skimmer, collisional-cooling hexapole (“top hexapole”), and
final high vacuum orifice. This sub-assembly is embedded in a custom 2
nd
stage vacuum
chamber designed for about 10 mTorr base pressure for optimal collisional cooling
8,9
of the ions.
It is then attached to the
top plate of the high-vacuum cr
yostat chamber housing the “bottom
hexapole” (see below) and the NEMS sample stage.
All N
2
gas, vacuum port, electronic,
and fluidic lines are connected to
external instrumentation.
†
Corresponding author: roukes@caltech.edu
Version 9
-2-
Supplementary
Information:
“Towards
single
‐
molecule
nanomechanical
mass
spectrometry”
Naik,
et
al.
Ion transport
Electrospray ionization is typically performed at
atmospheric pressure and at room, or elevated,
temperature. This enables the
solvent droplets that form beyond the Taylor cone to quickly
evaporate leaving the bare protein ions they
contain. In this fi
rst-generation NEMS-MS
realization we employ physisorp
tion to capture and immobilize
the analytes on the NEMS
sensor’s surface. By contrast, this requires
that the NEMS be maintained under high-vacuum
and low temperature conditions. Our experime
ntal setup must ther
efore provide a match
between these rather different conditions; we ach
ieve this with a comb
ination of commercially-
available and custom-built cryogenic and diffe
rential-pumping components. In our first-
generation design the NEMS sample
stage is located about
2m away from the ESI source. Since
the capture area of the NEMS sensor
is small, this necessitates efficient transport of the analyte
ions created by electrospray to guide them along a
2m path to the surface of the NEMS sensor
for adsorption.
Transport of ions from atmospheric pressure to
the first differential pum
ping stage is performed
through the so called nozzle-skimmer configuration
10,11
. This produces a
highly collinear,
monochromatic beam of nitrogen molecules and
protein ions. Subsequently, the ions are
transported from this initial vacuum stage to
the NEMS sensor by a hexapole ion guide system,
having an outer radius of ~ 1.2
cm and total length of
~ 2m. The hexapol
e configuration was
chosen as a good compromise between obtaining
high transmission effi
ciency and providing
Parameter
Description
Typical Values
V
needle
Voltage on the electrospray needle
3.0kV
V
L4
Voltage on the electrostatic lens between top and bottom
hexapole
-20V
V
shield
Voltage on the shield (ESI counter electrode)
600V
V
capillary
Voltage on the capillary
200V
V
top_hex
DC offset voltage on the top hexapole
10V
V
bot_hex
DC offset voltage on the bottom hexapole
0V
f
RF
RF frequency of AC voltage applied to the hexapole
1.1MHz, 450kHz
V
RF
Amplitude of the RF voltage applied to hexapole
500V
P
int
Pressure in the intermediate collisional focusing chamber
8mTorr
T
dry
Temperature of the drying gas
180
0
C
P
dry
Pressure of the drying gas
30psi
P
neb
Pressure of the nebulising gas
15psi
R
Protein solution flow rate
4
μ
l/min
Table I.
Typical values used during ESI and ion optics
Version 9
-3-
©
200
9
Macmillan Publishers Limited. All rights reserved.
4
nature nanotechnology
| www.nature.com/naturenanotechnology
SUPPLEMENTARY
INFORMATION
doi: 10.1038/nnano
.2009.152
Supplementary
Information:
“Towards
single
‐
molecule
nanomechanical
mass
spectrometry”
Naik,
et
al.
minimal
m/z
selectivity, to permit simultaneous and non-
discriminative transport of a
broad range of bio-/chemical
species. The configuration
we employ actually involves
two independent hexapole ion guide stages. The top
hexapole, operating in the 10 mTorr vacuum range,
provides collisional cooling a
nd trapping of the ions and
relies on space charge for inje
ction through an orifice into
the vacuum chamber of the
main cryostat. The bottom
hexapole operating in high vac
uum plays a dual role. It
acts as an ion “pipe”, enabli
ng broadband, high-efficiency
transmission of ions over the 2m path. It is also key in
overcoming the magnetic mirror effect that would
otherwise reflect the ions back along their initial path as
they tried to enter the
high magnetic field region
12,13
. (As
described below, we have
employed well-validated
magnetomotive displacement transduction for the NEMS
in this first generation system; it provides optimal
sensitivity in a high magnetic field.) Both hexapoles are
driven by a homemade RF oscillator
14,15
that can supply
AC voltages of up to 500V
peak
. In brief, the RF power
supply is based on two vacuum tubes operating in a push-
pull oscillator configuration, driving an LC load formed by
the capacitive load of the hexapole rods and high-voltage
inductor and capacitors to t
une to the frequency of
oscillation. The significant le
ngth of the lower hexapole
sets an upper limit to the fre
quency that can effectively be
applied by the voltage source, though this upper limit did
not come into play for the large mass species probed in this
experiment. The NEMS mass
sensor is centered about
3mm below the bottom end of the hexapole guide.
NEMS Device Fabrication
The structural material for the ultrahigh frequency (UHF)
NEMS devices in this work is a 100nm thick si
ngle crystal 3-C silicon carbide (SiC) epilayer
deposited on a silicon substrat
e through molecular beam epitaxy
16
. Thin-film metal conductors
are defined by optical lithography to form wirebond
pads and a lead-frame that converges into
the central, active region of each device where the
NEMS are located. Each of these small chips
is manually diced for subsequent
individual processing.
Electron-beam lithography is used to
laterally define what ultimately become nanometer-scale NEMS features on the SiC epilayer.
Thermal evaporation is used to deposit 40nm of
Al followed by 5nm Ti on these patterns, and
standard lift-off is then employed to define th
e metallization layer. This metallization layer
connects to the larger af
orementioned lead-frame, and also serves as a mask for the subsequent
etching process. The metal-masked SiC epilayer is etched in an Ar/NF
3
plasma created by an
electron cyclotron resonance (ECR) plasma-etching system. This dry etch step removes the SiC
in unprotected regions and undercuts the silicon be
neath the masked SiC, to result in a fully
Figure S2.
(From Figure 1 of main
paper)
(a,b)
Scanning electron
micrographs showing one of the
NEMS
device
used
in
the
experiments.
(c)
Magnitude and
phase
of
the
UHF
NEMS
resonator’s response displaying a
fundamental mode resonance near
429MHz.
Version 9
-4-
Supplementary
Information:
“Towards
single
‐
molecule
nanomechanical
mass
spectrometry”
Naik,
et
al.
suspended NEMS beam
17
. The completed
devices, such as depicted in Figure S2, are
geometrically characterized by a scanning
electron microscope (SEM).
The NEMS mass sensor used in these
measurements is a 100nm thick, doubly-
clamped silicon carbide beam ~1.7
μ
m long,
~120nm wide. In addition to its function as
an etch mask, the topmost, thin-film
metallization layer subsequently enables
sensitive and well-validated magnetomotive
actuation and transduction
18
.
NEMS Mass Sensor Characterization
Data presented in this work are obtained
with devices such as described previously,
which flexurally vibrate in-plane, with a
typical fundamental mode resonance at
~450MHz and quality factor of ~2000. A typical re
sponse curve is shown in Figure S2 (panel c).
1
10
100
1000
10000
1E-7
1E-6
1E-5
L
o
n
g
te
r
m
d
r
i
f
t
1/f noise limited
Allan Deviation
Time (s)
Figure S3.
Allan deviation of a typical NEMS
resonator used in this work, as a function of
integration time. The device
is phase locked within a
control circuit to sustain its fundamental vibrational
mode at ~450 MHz.
Prior to embedding candidate devices in the UHF pha
se locked loop (PLL) circuitry each is fully
characterized electromechanicall
y, in vacuum. This involves immersion in, typically, a 7T
magnetic field and execution of the RF drive freque
ncy sweeps with a netw
ork analyzer to locate
the various NEMS electromechanical resonances. E
ach resonant mode observe
d is verified to be
electromechanical in nature by confirming its
expected dependence upon magnetic field (
)
and drive power, which is taken well
into the non-linear mechanical regime
19
.
The temperature dependence of the fundamental-mode resonance is characterized by sweeping
the sample stage temperature slowly from 25K
to 65K, while recording the NEMS frequency.
The resulting curve has an inverted-“U” shaped
curve. We find that in the temperature range
between 35K and 40K the NEMS fundamental resonance frequency is least sensitive to the
temperature changes. This reduced temperature
coefficient is an additional motivation for our
operation at 40K; in this regime we have verified
that thermal fluctuations and thermalization of
landing proteins have negligible eff
ect on the NEMS resonant frequency.
RF characterization
To determine the temporal stability of the phase
locked loop we run the system in closed loop for
extended time periods and determine the resul
ting Allan deviation. Figure S3 shows a typical
plot of Allan deviation versus measurement inte
gration time. In the fl
at region between 3-30
seconds the Allan devia
tion is limited by the
1/f
noise. For longer measurement times drift
becomes the major source of instability;
for shorter times white noise dominates.
NEMS Mass Responsivity
The dynamics of a flexural mode NEMS resonato
r can be modeled as a simple damped harmonic
oscillator
20
. Referring to the coordinate system introduced in Figure 2 in the main text and
Version 9
-5-
©
200
9
Macmillan Publishers Limited. All rights reserved.
nature nanotechnology
| www.nature.com/naturenanotechnology
5
SUPPLEMENTARY
INFORMATION
doi: 10.1038/nnano
.2009.152
Supplementary
Information:
“Towards
single
‐
molecule
nanomechanical
mass
spectrometry”
Naik,
et
al.
minimal
m/z
selectivity, to permit simultaneous and non-
discriminative transport of a
broad range of bio-/chemical
species. The configuration
we employ actually involves
two independent hexapole ion guide stages. The top
hexapole, operating in the 10 mTorr vacuum range,
provides collisional cooling a
nd trapping of the ions and
relies on space charge for inje
ction through an orifice into
the vacuum chamber of the
main cryostat. The bottom
hexapole operating in high vac
uum plays a dual role. It
acts as an ion “pipe”, enabli
ng broadband, high-efficiency
transmission of ions over the 2m path. It is also key in
overcoming the magnetic mirror effect that would
otherwise reflect the ions back along their initial path as
they tried to enter the
high magnetic field region
12,13
. (As
described below, we have
employed well-validated
magnetomotive displacement transduction for the NEMS
in this first generation system; it provides optimal
sensitivity in a high magnetic field.) Both hexapoles are
driven by a homemade RF oscillator
14,15
that can supply
AC voltages of up to 500V
peak
. In brief, the RF power
supply is based on two vacuum tubes operating in a push-
pull oscillator configuration, driving an LC load formed by
the capacitive load of the hexapole rods and high-voltage
inductor and capacitors to t
une to the frequency of
oscillation. The significant le
ngth of the lower hexapole
sets an upper limit to the fre
quency that can effectively be
applied by the voltage source, though this upper limit did
not come into play for the large mass species probed in this
experiment. The NEMS mass
sensor is centered about
3mm below the bottom end of the hexapole guide.
NEMS Device Fabrication
The structural material for the ultrahigh frequency (UHF)
NEMS devices in this work is a 100nm thick si
ngle crystal 3-C silicon carbide (SiC) epilayer
deposited on a silicon substrat
e through molecular beam epitaxy
16
. Thin-film metal conductors
are defined by optical lithography to form wirebond
pads and a lead-frame that converges into
the central, active region of each device where the
NEMS are located. Each of these small chips
is manually diced for subsequent
individual processing.
Electron-beam lithography is used to
laterally define what ultimately become nanometer-scale NEMS features on the SiC epilayer.
Thermal evaporation is used to deposit 40nm of
Al followed by 5nm Ti on these patterns, and
standard lift-off is then employed to define th
e metallization layer. This metallization layer
connects to the larger af
orementioned lead-frame, and also serves as a mask for the subsequent
etching process. The metal-masked SiC epilayer is etched in an Ar/NF
3
plasma created by an
electron cyclotron resonance (ECR) plasma-etching system. This dry etch step removes the SiC
in unprotected regions and undercuts the silicon be
neath the masked SiC, to result in a fully
Figure S2.
(From Figure 1 of main
paper)
(a,b)
Scanning electron
micrographs showing one of the
NEMS
device
used
in
the
experiments.
(c)
Magnitude and
phase
of
the
UHF
NEMS
resonator’s response displaying a
fundamental mode resonance near
429MHz.
Version 9
-4-
Supplementary
Information:
“Towards
single
‐
molecule
nanomechanical
mass
spectrometry”
Naik,
et
al.
suspended NEMS beam
17
. The completed
devices, such as depicted in Figure S2, are
geometrically characterized by a scanning
electron microscope (SEM).
The NEMS mass sensor used in these
measurements is a 100nm thick, doubly-
clamped silicon carbide beam ~1.7
μ
m long,
~120nm wide. In addition to its function as
an etch mask, the topmost, thin-film
metallization layer subsequently enables
sensitive and well-validated magnetomotive
actuation and transduction
18
.
NEMS Mass Sensor Characterization
Data presented in this work are obtained
with devices such as described previously,
which flexurally vibrate in-plane, with a
typical fundamental mode resonance at
~450MHz and quality factor of ~2000. A typical re
sponse curve is shown in Figure S2 (panel c).
1
10
100
1000
10000
1E-7
1E-6
1E-5
L
o
n
g
te
r
m
d
r
i
f
t
1/f noise limited
Allan Deviation
Time (s)
Figure S3.
Allan deviation of a typical NEMS
resonator used in this work, as a function of
integration time. The device
is phase locked within a
control circuit to sustain its fundamental vibrational
mode at ~450 MHz.
Prior to embedding candidate devices in the UHF pha
se locked loop (PLL) circuitry each is fully
characterized electromechanicall
y, in vacuum. This involves immersion in, typically, a 7T
magnetic field and execution of the RF drive freque
ncy sweeps with a netw
ork analyzer to locate
the various NEMS electromechanical resonances. E
ach resonant mode observe
d is verified to be
electromechanical in nature by confirming its
expected dependence upon magnetic field (
)
and drive power, which is taken well
into the non-linear mechanical regime
19
.
The temperature dependence of the fundamental-mode resonance is characterized by sweeping
the sample stage temperature slowly from 25K
to 65K, while recording the NEMS frequency.
The resulting curve has an inverted-“U” shaped
curve. We find that in the temperature range
between 35K and 40K the NEMS fundamental resonance frequency is least sensitive to the
temperature changes. This reduced temperature
coefficient is an additional motivation for our
operation at 40K; in this regime we have verified
that thermal fluctuations and thermalization of
landing proteins have negligible eff
ect on the NEMS resonant frequency.
RF characterization
To determine the temporal stability of the phase
locked loop we run the system in closed loop for
extended time periods and determine the resul
ting Allan deviation. Figure S3 shows a typical
plot of Allan deviation versus measurement inte
gration time. In the fl
at region between 3-30
seconds the Allan devia
tion is limited by the
1/f
noise. For longer measurement times drift
becomes the major source of instability;
for shorter times white noise dominates.
NEMS Mass Responsivity
The dynamics of a flexural mode NEMS resonato
r can be modeled as a simple damped harmonic
oscillator
20
. Referring to the coordinate system introduced in Figure 2 in the main text and
Version 9
-5-
©
200
9
Macmillan Publishers Limited. All rights reserved.
6
nature nanotechnology
| www.nature.com/naturenanotechnology
SUPPLEMENTARY
INFORMATION
doi: 10.1038/nnano
.2009.152
Supplementary Information
:
“Towards single
-molecule nanomechanical mass spectrometry”
Naik,
et al.
Version 9
-6-
Here
y
center
denotes the center
-of-beam displacement,
k
denotes
the modal stiffness coefficient, and
M
eff
denotes the modal mass of the fundamental mode,
f
nems
is the resonance frequency of the flexural mode.
At this point, i
t is illu
minating
to show how modal
mass can be calculated.
The mechanical resonator
oscillating with a certain amplitude,
y(t)
, and angular
frequency,
ω
, possesses
kinetic energy.
This kinetic
energy
can
be
calculated
by
considering
infinitesimal slices of material along the length of
the beam and summing up their infinitesimal kinetic
energies:
Here
dm
is the mass of the each infinitesimal slice
considered
,
y
is the displacement of the beam at that
point
,
is the mass density,
is the cross section of the beam and
is its length.
The
displacement at a point
x
along the beam can be expressed as:
,
where
is the mode shape.
Using this relationship:
The expression inside the large square brackets is the modal mass,
M
eff
. For point mass loading
of a doubly clamped beam the effective mass is given by
21
,
When a point mass,
, is added to the resonator, at a point
along the beam, the total kinetic
energy carried by the resonator will change, as the added mass is forced to oscillate with the
beam. The new kinetic energy is given as
:
Using
, the above equation translates into:
Thus, the resonator’s effective mass is changed
to
. This in turn changes the
resonance frequency:
0.0
0.2
0.4
0.6
0.8
1.0
0.0
0.2
0.4
0.6
0.8
1.0
Normalized frequency shift
Normalized position of the particle
along the lenght of the NEMS
Figure S4
: The position dependent
responsivity of the doubly
-clamped
NEMS resonator for the fundamental
mode. The observed frequency shift is
maximized when the particle adsorbs
at
the center of the beam. Minimal shifts
are induced for adsorption near the
clamping points.
Supplementary Information
:
“Towards single
-molecule nanomechanical mass spectrometry”
Naik,
et al.
Version 9
-7-
For
, as in our experiment, one can make a Taylor expansion of the square bracket
to a
rrive at the frequency shift caused by the added mass:
The expression in the curly bracket is the position-
dependent responsivity of NEMS:
The shape of the position dependent responsivity is shown in figure S4.
As an example, we now
calculate the responsivity for a point particle landing
at the center of the
beam.
For this case
The NEMS beam used in the experiment is 1.7 μm long, 120 nm wide and containing three
layers:
100 nm SiC (
ρ
=3.2 g/cm
3
), 40 nm Al (
ρ
= 2.7 g/cm
3
), and 5 nm Ti (
ρ
=4.5 g/cm
3
)
Thus:
M
total
=92
10
-15
and
g = 92 fg
The exp
erimentally observed mass responsivity of 12
Hz/zg
is reasonably close to this calculated
value.
Based on the experimental
ly-measured
mass responsivity,
we deduce that
mass of the beam
to be
M
total
=46fg.
Using this experimentally
-observed total mass
, and
an
upper
bound to our typical
Allan deviation,
, we
deduce the following upper bounds to our
mass
sensitivity
δ
M
and the frequency instability
for these experiments
2
:
©
200
9
Macmillan Publishers Limited. All rights reserved.
nature nanotechnology
| www.nature.com/naturenanotechnology
7
SUPPLEMENTARY
INFORMATION
doi: 10.1038/nnano
.2009.152
Supplementary Information
:
“Towards single
-molecule nanomechanical mass spectrometry”
Naik,
et al.
Version 9
-6-
Here
y
center
denotes the center
-of-beam displacement,
k
denotes
the modal stiffness coefficient, and
M
eff
denotes the modal mass of the fundamental mode,
f
nems
is the resonance frequency of the flexural mode.
At this point, i
t is illu
minating
to show how modal
mass can be calculated.
The mechanical resonator
oscillating with a certain amplitude,
y(t)
, and angular
frequency,
ω
, possesses
kinetic energy.
This kinetic
energy
can
be
calculated
by
considering
infinitesimal slices of material along the length of
the beam and summing up their infinitesimal kinetic
energies:
Here
dm
is the mass of the each infinitesimal slice
considered
,
y
is the displacement of the beam at that
point
,
is the mass density,
is the cross section of the beam and
is its length.
The
displacement at a point
x
along the beam can be expressed as:
,
where
is the mode shape.
Using this relationship:
The expression inside the large square brackets is the modal mass,
M
eff
. For point mass loading
of a doubly clamped beam the effective mass is given by
21
,
When a point mass,
, is added to the resonator, at a point
along the beam, the total kinetic
energy carried by the resonator will change, as the added mass is forced to oscillate with the
beam. The new kinetic energy is given as
:
Using
, the above equation translates into:
Thus, the resonator’s effective mass is changed
to
. This in turn changes the
resonance frequency:
0.0
0.2
0.4
0.6
0.8
1.0
0.0
0.2
0.4
0.6
0.8
1.0
Normalized frequency shift
Normalized position of the particle
along the lenght of the NEMS
Figure S4
: The position dependent
responsivity of the doubly
-clamped
NEMS resonator for the fundamental
mode. The observed frequency shift is
maximized when the particle adsorbs
at
the center of the beam. Minimal shifts
are induced for adsorption near the
clamping points.
Supplementary Information
:
“Towards single
-molecule nanomechanical mass spectrometry”
Naik,
et al.
Version 9
-7-
For
, as in our experiment, one can make a Taylor expansion of the square bracket
to a
rrive at the frequency shift caused by the added mass:
The expression in the curly bracket is the position-
dependent responsivity of NEMS:
The shape of the position dependent responsivity is shown in figure S4.
As an example, we now
calculate the responsivity for a point particle landing
at the center of the
beam.
For this case
The NEMS beam used in the experiment is 1.7 μm long, 120 nm wide and containing three
layers:
100 nm SiC (
ρ
=3.2 g/cm
3
), 40 nm Al (
ρ
= 2.7 g/cm
3
), and 5 nm Ti (
ρ
=4.5 g/cm
3
)
Thus:
M
total
=92
10
-15
and
g = 92 fg
The exp
erimentally observed mass responsivity of 12
Hz/zg
is reasonably close to this calculated
value.
Based on the experimental
ly-measured
mass responsivity,
we deduce that
mass of the beam
to be
M
total
=46fg.
Using this experimentally
-observed total mass
, and
an
upper
bound to our typical
Allan deviation,
, we
deduce the following upper bounds to our
mass
sensitivity
δ
M
and the frequency instability
for these experiments
2
:
©
200
9
Macmillan Publishers Limited. All rights reserved.
8
nature nanotechnology
| www.nature.com/naturenanotechnology
SUPPLEMENTARY
INFORMATION
doi: 10.1038/nnano
.2009.152
Supplementary
Information:
“Towards
single
‐
molecule
nanomechanical
mass
spectrometry”
Naik,
et
al.
Control runs
0
2000
4000
6000
8000
10000
12000
14000
-160
-140
-120
-100
-80
-60
-40
-20
0
Syringe pump ON
ESI needle
voltage OFF
Syringe Pump OFF
ESI voltages ON
ESI of BSA
ESI of solvent
Change in resonant frequency (kHz)
Ti me (s)
Figure S5
: The response of the NEMS frequency during the
control runs and the ESI of BSA.
To verify that the change in
resonant frequency of the
NEMS is due to mass loading
from protein molecules or
nanoparticles landing on the
NEMS,
we
perform
the
following three distinct control
runs
while
tracking
the
resonant frequency of the
NEMS:
1.
Turn off the syringe pump
delivering
the
protein
solution to the ESI needle.
Keep all ESI voltages and
ion optics control voltages
on. Proteins do not reach
the ESI needle, hence
protein ions should not be
produced (beyond rare
events arising from the
dislodging of sparse protein residue accumulated w
ithin the needle, etc. –
i.e.
from previous
NEMS-MS runs.
2.
Turn off the ESI needle voltage. All the othe
r parameters are kept the same as for ESI
injection during regula
r NEMS-MS operations. In this case although the protein solution is
ejected from the ESI needle due to the flow
pressure generated by the syringe pump and the
nebulising gas flow, the resul
ting droplets are not charged a
nd any proteins, if produced,
will not be transported to the NEMS by the ion guide.
3.
Run the ESI with the clean solvent devoid of an
alytes. Assuming that there are no leftover
proteins in the ESI system (as described in #1
above) we should see minimal change in the
resonant frequency of the NEMS.
Figure S5 shows one such time record of the res
onant frequency of the NEMS during the ESI of
Bovine Serum Albumin (66kDa) a
nd three different control runs. The resonant frequency of the
NEMS changes radically during the ESI of BSA pr
otein. The frequency change is composed of
scores of frequency jumps similar to the one
s shown in Figure 2. During the control runs,
however, the total change in res
onant frequency is noticeably sma
ller. This indicates that the
mass loading indeed arises from the protein io
ns produced during ESI. We believe the mass
loading of the NEMS during solvent-only ESI is
largely due to proteins remaining in the ESI
system from previous runs. The
resonant frequency changes during
the other two control runs are
minimal and compare well with the frequency
fluctuations observed due to background gas
molecules.
Version 9
-8-
Supplementary
Information:
“Towards
single
‐
molecule
nanomechanical
mass
spectrometry”
Naik,
et
al.
Cross-checks: ensuring th
at frequency-shift “events” indeed arise from
nanoparticle/protein adsorption
One distinguishing feature of the ESI process is
that it typically produces
species with mass-to-
charge ratios of order
m/z
~1000. This can be easily verified in our experiments. Based on the
relative cross-sectional area of
the faraday cup and the NEMS,
and using the charge-to-mass
transformation, we can deduce the
expected frequency shift due
to mass loading of the NEMS
convert from the current observed
at Faraday cup. Figure S6 shows the frequency change of the
NEMS due to mass loading and the expected
frequency change assuming several different
average
m/z
values.
Adsorption-event curve-fitting analysis
0
1000
2000
3000
4000
5000
6000
7000
-140
-120
-100
-80
-60
-40
-20
0
20
Change in resonance frequency (kHz)
Time (s)
Experimental
Estimated from current seen at
Faraday cup and m/z 600
Estimated from current seen at
Faraday cup and m/z 650
Estimated from current seen at
Faraday cup and m/z 700
Figure S6.
The frequency change observed during the experiment
(gray solid line) and the frequency change expected based on the
ionic current measured at the faraday cup. The expected curve is
calculated by assuming an average m/z of 600, 650 and 700 and
converting it into mass deposited on the NEMS.
In order to construct frequency-
shift histograms from the experimental time records of resonance
frequency while under phase-lock, one needs an objective method to identify adsorption events
and measure the magnitude of their corresponding frequency shifts,
. We have developed a
Matlab script to serve this purpose. The experime
ntal time records consist of scores of discrete
frequency jumps. We characteri
ze each of these jumps by the ma
ximum slope at the jump center
and zero slopes at the edges. Our Matlab routine
scans the experimental data for local regions of
high slope, subject to the condi
tions that our frequency-shif
t threshold is set to be
,
records the experimental data as a freque
ncy step. (As described in the main text,
is the
frequency resolution). By these
procedures the extracted frequenc
y jumps are “filtered” to reject
both impulsive noise spikes and
long term drifts. Practically,
this is achieved by putting
lower and upper limits on the
acceptable time-scale of “real”
frequency jumps. Later, each
frequency jump data in this
filtered set is fitted to the
modeled time-response of the
PLL circuitry as described in
the main text. The parameters
in this PLL model are chosen
initially from circuit analysis
and validated by separate
measurements. These circuit
parameters
are
iteratively
varied slightly to
obtain the best
fit between the experimental
characterization measurements
and calculated PLL temporal
response function (see main
text, Figure 2b). The heights of
the frequency jumps obtained
during the fitting process are
Version 9
-9-
©
200
9
Macmillan Publishers Limited. All rights reserved.
nature nanotechnology
| www.nature.com/naturenanotechnology
9
SUPPLEMENTARY
INFORMATION
doi: 10.1038/nnano
.2009.152
Supplementary
Information:
“Towards
single
‐
molecule
nanomechanical
mass
spectrometry”
Naik,
et
al.
Control runs
0
2000
4000
6000
8000
10000
12000
14000
-160
-140
-120
-100
-80
-60
-40
-20
0
Syringe pump ON
ESI needle
voltage OFF
Syringe Pump OFF
ESI voltages ON
ESI of BSA
ESI of solvent
Change in resonant frequency (kHz)
Ti me (s)
Figure S5
: The response of the NEMS frequency during the
control runs and the ESI of BSA.
To verify that the change in
resonant frequency of the
NEMS is due to mass loading
from protein molecules or
nanoparticles landing on the
NEMS,
we
perform
the
following three distinct control
runs
while
tracking
the
resonant frequency of the
NEMS:
1.
Turn off the syringe pump
delivering
the
protein
solution to the ESI needle.
Keep all ESI voltages and
ion optics control voltages
on. Proteins do not reach
the ESI needle, hence
protein ions should not be
produced (beyond rare
events arising from the
dislodging of sparse protein residue accumulated w
ithin the needle, etc. –
i.e.
from previous
NEMS-MS runs.
2.
Turn off the ESI needle voltage. All the othe
r parameters are kept the same as for ESI
injection during regula
r NEMS-MS operations. In this case although the protein solution is
ejected from the ESI needle due to the flow
pressure generated by the syringe pump and the
nebulising gas flow, the resul
ting droplets are not charged a
nd any proteins, if produced,
will not be transported to the NEMS by the ion guide.
3.
Run the ESI with the clean solvent devoid of an
alytes. Assuming that there are no leftover
proteins in the ESI system (as described in #1
above) we should see minimal change in the
resonant frequency of the NEMS.
Figure S5 shows one such time record of the res
onant frequency of the NEMS during the ESI of
Bovine Serum Albumin (66kDa) a
nd three different control runs. The resonant frequency of the
NEMS changes radically during the ESI of BSA pr
otein. The frequency change is composed of
scores of frequency jumps similar to the one
s shown in Figure 2. During the control runs,
however, the total change in res
onant frequency is noticeably sma
ller. This indicates that the
mass loading indeed arises from the protein io
ns produced during ESI. We believe the mass
loading of the NEMS during solvent-only ESI is
largely due to proteins remaining in the ESI
system from previous runs. The
resonant frequency changes during
the other two control runs are
minimal and compare well with the frequency
fluctuations observed due to background gas
molecules.
Version 9
-8-
Supplementary
Information:
“Towards
single
‐
molecule
nanomechanical
mass
spectrometry”
Naik,
et
al.
Cross-checks: ensuring th
at frequency-shift “events” indeed arise from
nanoparticle/protein adsorption
One distinguishing feature of the ESI process is
that it typically produces
species with mass-to-
charge ratios of order
m/z
~1000. This can be easily verified in our experiments. Based on the
relative cross-sectional area of
the faraday cup and the NEMS,
and using the charge-to-mass
transformation, we can deduce the
expected frequency shift due
to mass loading of the NEMS
convert from the current observed
at Faraday cup. Figure S6 shows the frequency change of the
NEMS due to mass loading and the expected
frequency change assuming several different
average
m/z
values.
Adsorption-event curve-fitting analysis
0
1000
2000
3000
4000
5000
6000
7000
-140
-120
-100
-80
-60
-40
-20
0
20
Change in resonance frequency (kHz)
Time (s)
Experimental
Estimated from current seen at
Faraday cup and m/z 600
Estimated from current seen at
Faraday cup and m/z 650
Estimated from current seen at
Faraday cup and m/z 700
Figure S6.
The frequency change observed during the experiment
(gray solid line) and the frequency change expected based on the
ionic current measured at the faraday cup. The expected curve is
calculated by assuming an average m/z of 600, 650 and 700 and
converting it into mass deposited on the NEMS.
In order to construct frequency-
shift histograms from the experimental time records of resonance
frequency while under phase-lock, one needs an objective method to identify adsorption events
and measure the magnitude of their corresponding frequency shifts,
. We have developed a
Matlab script to serve this purpose. The experime
ntal time records consist of scores of discrete
frequency jumps. We characteri
ze each of these jumps by the ma
ximum slope at the jump center
and zero slopes at the edges. Our Matlab routine
scans the experimental data for local regions of
high slope, subject to the condi
tions that our frequency-shif
t threshold is set to be
,
records the experimental data as a freque
ncy step. (As described in the main text,
is the
frequency resolution). By these
procedures the extracted frequenc
y jumps are “filtered” to reject
both impulsive noise spikes and
long term drifts. Practically,
this is achieved by putting
lower and upper limits on the
acceptable time-scale of “real”
frequency jumps. Later, each
frequency jump data in this
filtered set is fitted to the
modeled time-response of the
PLL circuitry as described in
the main text. The parameters
in this PLL model are chosen
initially from circuit analysis
and validated by separate
measurements. These circuit
parameters
are
iteratively
varied slightly to
obtain the best
fit between the experimental
characterization measurements
and calculated PLL temporal
response function (see main
text, Figure 2b). The heights of
the frequency jumps obtained
during the fitting process are
Version 9
-9-
©
200
9
Macmillan Publishers Limited. All rights reserved.