Intrinsic noise of a micro-mechanical displacement detector based on the radio-frequency single-electron transistor
Extracted source text ยท arXiv cond-mat/0109412
%&LaTeX
\documentstyle[preprint,prb,aps,epsfig]{revtex}
\textheight = 9truein
\voffset = .5 truein
\begin{document}
\draft
\title{Intrinsic noise of a micro-mechanical displacement detector
based on the radio-frequency single-electron transistor}
\author{Yong Zhang and Miles P. Blencowe}
\address{Department of Physics and Astronomy, Dartmouth College,
Hanover, New Hampshire 03755}
\date{\today}
\maketitle
\begin{abstract}
We investigate the intrinsic noise of a micro-mechanical
displacement detector based on the radio-frequency single-electron
transistor (rf-SET). Using the noise analysis of a SET by Korotkov as our
starting point, we determine the spectral density of the displacement noise
due to the tunneling current shot noise. The resulting mechanical displacement
noise decreases in inverse proportion to the increasing gate voltage.
In contrast, the displacement noise due to the fluctuating SET island charge
increases approximately linearly with increasing gate voltage. Taking into
account both of these noise sources results in an optimum gate voltage value
for the lowest displacement noise and hence best sensitivity. We show that a
displacement sensitivity of about $10^{-4}~{\rm \AA}$ and a force sensitivity
of about $10^{-16}~{\rm N}$ are
predicted for a micron-sized cantilever with a realizable resonant frequency
$100~{\rm MHz}$ and quality factor $Q\sim~10^{4}$. Such sensitivities
would allow the
detection of quantum squeezing in the mechanical motion of the micro-mechanical
cantilever and the detection of single-spin magnetic resonance in magnetic
resonance force microscopy (MRFM).
\end{abstract}
\pacs{PACS numbers: 85.35.Gv, 85.85.+j, 73.50.Td}
\section{Introduction}
Fast and ultra-sensitive displacement and force detection is of
great interest for a broad range of fundamental applications. It is an essential
requirement for the proposed detection of quantum
superposition states\cite{bose}, quantum
squeezed states\cite{blencowe}, and single-spin
magnetic resonance in
MRFM\cite{sidles} which has also been proposed for the read-out stage of a
solid-state NMR quantum computer.\cite{berman} The common
requirement of these examples is the need to perform the measurement using a
micro-to-nanoscale-mechanical detector at, or even below,
the ``standard quantum limit"
(SQL) $\Delta x_{\rm SQL}=(\hbar/2m\omega_{0})^{1/2}$,
and in the
radio-frequency range. Considering, for instance, a micron-sized cantilever
with a resonant frequency of $100~{\rm MHz}$ and quality factor $Q\sim~10^{4}$,
the detection of
quantum squeezed states requires a displacement sensitivity of about
$10^{-4}~{\rm \AA}$.\cite{blencowe} As another example,
in detecting single-spin
magnetic resonance, the interaction force between a nano-magnet with a
targeted electron spin is typically on the order of about
$10^{-16}~{\rm N} (100~{\rm aN})$.\cite{sidles}
If a micron-sized cantilever on which the nano-magnet is mounted is directly
coupled at the spin precession frequency ($100~{\rm MHz}$ or higher),
the magnetic force will cause the cantilever to oscillate at an
amplitude of about $10^{-4}~{\rm\AA}$ for a single spin.
Following the development of the
rf-SET,\cite{schoelkopf,fujisawa,aassime}
it has been proposed
to use an rf-SET-based displacement detector for measuring sub-angstrom motion of
micron-scale and smaller mechanical oscillators.\cite{blencowe2} In the rf-SET,
$1/f$ noise which limits the applications of the conventional
dc-SET is considerably improved. A shot noise limit of
$10^{-6}~{\rm\AA}/\sqrt{Hz}$ may be
achievable by using an rf-SET displacement detector,\cite{blencowe2}
provided the coupling
(gate voltage) between the SET and micro-mechanical oscillator is strong
enough.
However, the back-action of the SET onto the mechanical
oscillator must also be taken into account:\cite{schwab}
the fluctuating force acting on the oscillator
due to the fluctuating island charge produces another contribution to the
displacement noise. The total displacement noise will be the sum of
contributions from two sources - the shot noise in the tunneling current and
the force noise due to the fluctuating island charge.
In this paper, we give a calculation of the displacement noise
of the rf-SET displacement detector. By using the intrinsic noise analysis
method of Korotkov,\cite{korotkov} the spectral density of the shot
noise $S_{I}$ in the tunneling
current and the spectral density of the force noise $S_{F}$ are determined, and
the
total spectral density of the displacement noise $S_{X}$ is then deduced. We also
present two simple expressions from which good estimates of the displacement
noise due to the shot noise and the fluctuating
island charge can be obtained. Results for the ultimate intrinsic
noise limit of the detector for different parameter values are presented and
discussed. The displacement and force sensitivities are estimated for an
optimized rf-SET displacement detector and we find that they meet the
requirement for detecting quantum squeezing and single-spin magnetic resonance.
\section{Intrinsic noise of the rf-SET displacement detector}
The scheme of the rf-SET displacement detector is shown in Fig.~1.
The basic principle of the
device involves locating one of the gate
capacitor plates of the SET on the cantilever so that,
for fixed gate voltage bias, a mechanical
displacement is converted into a polarization charge fluctuation.
The
stray capacitance $C_{s}$ of the leads contacting the SET and an inductor
$L$ form a tank circuit with resonant frequency
$\omega_{T}=(LC_{s})^{-1/2}$, loaded by the SET. A
monochromatic carrier wave is sent down the cable. At
the resonant frequency the circuit impedance is small and the
reflected power provides a measure of the SET's differential
resistance $R_{d}$.\cite{korotkov2} When the gate capacitor is biased,
mechanical motion of the cantilever is converted into differential
resistance changes, hence modulating the reflected signal power.
To simplify the algebra and make the final results concise,
we shall in fact consider
a symmetric dc-SET in the calculations with junction resistances
$R_{1}= R_{2}= R_{j}$ and
junction capacitances $C_{1}=C_{2}= C_{j}$. The differences between the
rf- and dc-SET noise formulae are inessential, with appropriate
time-averages required in the former\cite{blencowe2}
resulting in only small quantitative differences in the noise values.
Referring to Fig.~1, when the drain-source voltage $V_{\rm ds}$ is small
compared to
the voltage $e/C_{\Sigma}$, where $C_{\Sigma}=2 C_{j}+C_{g}$ is the
total capacitance of the SET, and also the thermal energy $k_{B}T\ll
eV_{\rm ds}$, the peaks of the tunneling
current $I_{\rm ds}$ will be well separated in gate voltage $V_{g}$. To
derive an expression for $I_{\rm ds}$ in the tunneling region, we
consider only the two most probable island electron numbers $n$ and
$n+1$, resulting in the following approximation:\cite{blencowe2}
\begin{equation}
I_{\rm ds}=
e[b_{1}(n)-t_{1}(n)]\sigma(n)+e[b_{1}(n+1)-t_{1}(n+1)]\sigma(n+1).
\label{current1}
\end{equation}
The probabilities $\sigma(n)$ and $\sigma(n+1)$ are also given there and the
tunneling
rates $b_{i}(t_{i})$ from the bottom (top) across the $i$th junction
of the rf-SET take
the usual form [see expression (\ref{apprates}) in the appendix].
In the previous calculation of the sensitivity of the rf-SET based detector,
the shot noise formula $S_{I} = 2eI_{\rm ds}$ was
used.\cite{blencowe2,korotkov2} This expression,
which gives the maximum current noise, is only approximately valid
when $V_{\rm ds}$ is close to the tunneling threshold.
Taking as our tool the
intrinsic noise analysis of the SET by Korotkov, we obtain the following
more accurate expression for the spectral density of the current noise
$S_{I}(\omega)$ [see appendix A for an outline of the deviation]:
\begin{equation}
S_{I}(\omega)=\frac{e^{2}a(n)b(n+1)}{a(n)+b(n+1)}+\frac{e^{2}[f(n)-g(n+1)]}
{a(n)+b(n+1)}\frac{a^{2}(n) g(n+1) -b^{2}(n+1)
f(n)}{[a(n)+b(n+1)]^{2}+\omega^{2}},
\label{currentnoise}
\end{equation}
where
\begin{eqnarray}
a(m)&=&b_{1}(m)+t_{2}(m),\cr
b(m)&=&b_{2}(m)+t_{1}(m),\cr
f(m)&=&b_{1}(m)-t_{2}(m),\cr
g(m)&=&b_{2}(m)-t_{1}(m).
\label{def1}
\end{eqnarray}
The spectral density of the displacement noise due to the shot noise in
the tunneling current is:
\begin{equation}
S_{X}^{I}(\omega)=S_{I}(\omega)/(dI_{\rm ds}/dx)^{2}.
\label{SXI}
\end{equation}
The dependence on the displacement $x$ enters through the gate
capacitance which is approximately $C_{g}\approx C_{g}^{0}(1-x/d)$
for $|x|\ll d$ where $d$ is the cantilever electrode-island electrode
gap.
The fluctuation of the island charge will produce a
fluctuating force on
the cantilever, therefore leading to another contribution to the
displacement noise. The ultimate force noise is only determined by
the stochastic character of the tunneling process. By using the
same method of intrinsic noise analysis,\cite{korotkov}
the following expressions are
obtained for the spectral density of the force noise $S_{F}(\omega)$
[see appendix B for an outline of the deviation]:
\begin{eqnarray}
S_{F}(\omega)=&&\frac{4A^{2}e^{4}a(n)b(n+1)}{a(n)+b(n+1)}\cr
&&\times \frac{(2C_{j}/C_{g})^{2} (2n+1+2\Delta n)^{2}
+(4C_{j}/C_{g})(2n+1)(2n+1+2\Delta
n)+(2n+1)^{2}}{[a(n)+b(n+1)]^{2}+\omega^{2}},
\label{SF}
\end{eqnarray}
where \begin{equation}
A=C_{g}(2C_{j}-C_{g})/2dC_{\Sigma}^{3}
\end{equation}
and
\begin{equation}
\Delta n=\frac{C_{g}V_{g}}{e}-\frac{C_{g}V_{\rm ds}}{2e}-n-\frac{1}{2}.
\end{equation}
We model the cantilever as a simple harmonic oscillator, so that the
spectral density of the displacement noise due to the fluctuating
force is:
\begin{equation}
S_{X}^{F}(\omega)=\frac{S_{F}(\omega)/m_{\rm
eff}^{2}}{(\omega^{2}-\omega_{0}^{2})^{2}+
\omega^{2}\omega_{0}^{2}/Q^{2}},
\label{SXF}
\end{equation}
where $m_{\rm eff}$ is the motional mass of the cantilever,
$\omega_{0}$ is its resonant frequency, and $Q$ is
the quality factor. For a cantilever with geometry $l\times w\times
t$ (length$\times$width$\times$thickness) and made from material of
density $\rho$ and Young's modulus $E$, the resonant frequency of the
lowest-order flexural mode is\cite{sidles}
\begin{equation}
\omega_{0}=3.516 \frac{t}{l^{2}} \sqrt{\frac{E}{12\rho}}.
\label{frequency}
\end{equation}
The motional mass of the cantilever is given by $m_{\rm eff}=\rho
lwt/4$ and the effective spring constant $k_{\rm eff}$ at the
cantilever's tip is $k_{\rm eff}=m_{\rm eff}\omega_{0}^{2}$.
The total displacement noise will be given as the sum of the above
two contributions (\ref{SXI}) and (\ref{SXF}):
\begin{equation}
S_{X}(\omega)=S_{X}^{I}(\omega)+S_{X}^{F}(\omega).
\label{SXtot}
\end{equation}
In addition, the following expressions are given here only for simple
estimates of the displacement noise at the resonant frequency of the
cantilever [see appendix C for an outline of their derivation]:
\begin{equation}
S_{X}^{I}=K\frac{2R_{j}}{e V_{\rm ds}}\left[\frac{2.5
k_{B}TdC_{\Sigma}}{C_{g}(V_{g}-V_{\rm ds}/2)}\right]^{2},
\label{approxSXI}
\end{equation}
\begin{equation}
S_{X}^{F}=\frac{Q^2 e^{3}R_{j}}{k_{\rm eff}^{2}V_{\rm
ds}}\left[\frac{C_{g}(2C_{j}-C_{g})(V_{g}-V_{\rm ds}/2)}
{d C_{\Sigma}^{2}}\right]^{2},
\label{approxSXF}
\end{equation}
where $K\approx 40$ is a constant.
\section{Results and Discussion}
In this paper, we consider crystalline Si cantilevers with mass
density $\rho= 2.33\times 10^{3}~{\rm kg/m}^{3}$ and Young's modulus
$E_{100} = 1.33\times 10^{11}~{\rm N/m}^{2}$.
To make the picture clear, we compare the displacement noise for
different cantilevers. Table I lists the resonant frequencies $f_{0}$ and
masses $m$ (the effective mass of a cantilever is $m_{\rm eff}= m/4$) of
cantilevers
with the same width and thickness but different lengths. The resonant
frequencies $f_{0}= \omega_{0}/2\pi$ are obtained from expression
(\ref{frequency}). We
mainly focus our attention on the detector using the cantilever
with geometry $1.25\times 0.6\times 0.15$ in $\mu{\rm m}$, quality factor
$Q=10^{4}$, resonant
frequency $f_{0}= 117~{\rm MHz}$, and effective spring constant
$k_{\rm eff} = 3.6\times 10^{9}~{\rm aN/\AA}$. The
standard quantum limit of the fundamental flexural mode of this cantilever is
$\Delta x_{\rm SQL}= 3.3\times 10^{-4}~{\rm \AA}$, so that the minimum detectable
force at the resonant frequency is $\Delta F= k_{\rm eff}\Delta x_{\rm
SQL}/Q= 120~{\rm aN}$.
Fig.~2 shows the dependence of the displacement noise $\sqrt{S_{X}^{I}}$ on
the gate voltage $V_{g}$ at the resonant
frequency $f_{0}= 117~{\rm MHz}$ of cantilever . The result
assumes a symmetric SET with junction capacitance $C_{j}= 100~{\rm aF}$,
junction resistance $R_{j}= 50~{\rm k}\Omega$, gate capacitance
$C_{g}= 50~{\rm aF}$, gate
capacitor gap $d= 0.1~\mu{\rm m}$, source-drain voltage
$V_{\rm ds}= 0.38~{\rm mV} (0.6e/C_{\Sigma})$, and
temperature $T= 74~{\rm mK}(0.01e^{2}/C_{\Sigma})$, which are all
routinely achievable.
The solid line is obtained by using expressions
(\ref{current1}), (\ref{currentnoise}), and (\ref{SXI}) and the dashed
line using (\ref{approxSXI}). The minimum displacement noise occurs near the
edges of the tunneling current $I_{\rm ds}$, and decreases approximately in
inverse proportion to increasing gate voltage. The displacement noise diverges
in the center of tunneling current peaks and in the Coulomb blockade region where the
tunneling current is suppressed. The result here is similar to Fig.~2
in Ref.\ \onlinecite{blencowe2}.
The dependence of the displacement noise $\sqrt{S_{X}^{F}}$ on the gate voltage
$V_{g}$ at the resonant frequency $f_{0}= 117~{\rm MHz}$ is illustrated in
Fig.~3. The result
assumes the same parameters as above. The solid line is obtained by
using expressions (\ref{SF}) and (\ref{SXF}), and the dashed line using
(\ref{approxSXF}). The maximum displacement noise increases linearly with
the increasing gate voltage.
Fig.~4 shows the total displacement noise
$\sqrt{S_{X}}$ as a function of gate voltage.
The tunneling current is shown as the dashed lines.
Considering the opposite dependences of $S_{X}^{I}$ and $S_{X}^{F}$ on
gate voltage, we will expect an optimum gate voltage value for the
lowest displacement noise and best sensitivity.
But there are two problems involved in the determination.
The first problem is that near the edges the tunneling current $I_{\rm
ds}$ may be too small to be measured accurately.
The second is that because $S_{X}^{I}$
and $S_{X}^{F}$ change in different ways with increasing gate
voltage $V_{g}$, it is rather involved finding the most optimum value of
$V_{g}$ at which the total
noise is minimum; the positions of the minima in the displacement noise
doublets will change relatively to the tunneling current peaks with
increasing gate voltage. An alternative, simpler strategy is to investigate the
displacement noise at the gate voltage for which the tunneling current is
fixed and measurable, say between $30\%$ and $60\%$
of its maximum value $I_{\rm ds, max}$.
Fig.~5 shows the total displacement noise $\sqrt{S_{X}}$ for our rf-SET
displacement detector using different cantilever parameter values
listed in Table I. The above method
is utilized with the displacement noise determined where the tunneling
current is half its maximum value. For the detector using a cantilever
with resonant frequency $f_{0}= 117~{\rm MHz}$, the minimum displacement noise
is $4.3\times 10^{-6}~{\rm \AA}/\sqrt{Hz}$, corresponding to an absolute
displacement sensitivity of $5\times 10^{-4}~{\rm\AA}$ and force sensitivity
$180~{\rm aN}$. Note that the minimum displacement
noise increases and shifts to lower gate voltage for increasing
cantilever length. As is evident from (\ref{approxSXF}), this is a direct
consequence of the effective spring constant dependence:
obviously a stiffer cantilever will be displaced less by the same applied force.
Fig.~6 shows the detector displacement noise for different
junction capacitance values. Both $S_{X}^{I}$ and $S_{X}^{F}$
decrease with decreasing $C_{j}$.
Because small displacements are
detected by monitoring changes in tunneling current, taking a smaller
junction capacitance serves another advantage
for the experiment: with $C_{\Sigma}V_{\rm ds}/e$
kept fixed, a smaller junction capacitance corresponds to a larger
source-drain voltage, so that the maximum tunneling current $I_{\rm ds, max}$ increases
according to expression (\ref{currentmax}).
Fig.~7
shows the displacement noise when the tunneling current is biased at
various values between
$30{\%}$ and $60{\%}$ of $I_{\rm ds, max}$.
We see that performing detection at a smaller percentage of the maximum
tunneling current becomes another effective way to further decrease the
displacement noise. The lowest displacement noise we
obtain is $\sqrt{S_{X}} = 3.2\times 10^{-6}~{\rm\AA}/\sqrt{{\rm Hz}}$,
assuming $C_{j}= 100~{\rm aF}$, $C_{g}= 50~{\rm aF}$, and $I_{\rm ds}=
0.57~{\rm nA}$, biased at $30{\%}$ of its maximum value.
The optimum gate voltage is $1.7~{\rm V}$ in
this case.
In conclusion, approaching the SQL will become possible for a carefully
engineered rf-SET displacement detector. A sensitivity at, or even
beyond the
SQL may be achievable by using the same parameters as in the above
analysis but with
the smaller junction capacitance $C_{j} = 75~{\rm aF}$.
Such a sensitivity would enable us to observe quantum squeezed states in
micro-mechanical systems. The proposed
displacement detector also meets the crucial requirement for advances in MRFM.
The radio-frequency micro-mechanical cantilever used in our model allows
direct coupling at the spin precession frequency, therefore allowing fast
read-out, while the predicted force sensitivity would allow the measurement of
single-spin resonance.
\acknowledgements
The authors thank Keith Schwab and Rob Schoelkopf for pointing out the
need to include the backaction on the cantilever due to the fluctuating island
charge and Martin Wybourne for helpful
discussions. This work was supported in part by the National Security
Agency (NSA) and Advanced Research and Development Activity (ARDA) under
Army Research Office (ARO) contract number DAAG190110696, and by an award from
Research Corporation.
\appendix
\section{Tunneling current shot noise}
The tunneling current shot noise is given by expression (30) in Ref.\
\onlinecite{korotkov}. In our notation, it is:
\begin{eqnarray}
S_{I}(\omega)=&&2U+e^{2}
\sum_{m,m'}\left[t_{2}(m')-b_{2}(m')+t_{1}(m')-b_{1}(m')\right]
B_{m',m}\cr
&&\times\left\{\left[t_{2}(m-1)-b_{1}(m-1)\right]\sigma(m-1)+
\left[t_{1}(m+1)-b_{2}(m+1)\right]\sigma(m+1)\right\},
\label{appSI}
\end{eqnarray}
where
\begin{equation}
2U=(e^{2}/2)\sum_{m}\sigma(m)\left[t_{2}(m)+b_{2}(m)+t_{1}(m)+b_{1}(m)\right]
\label{U}
\end{equation}
and $B_{m',m}$ is the real part of the inverse of the tridiagonal
matrix $(i\omega I -\Gamma)$, where $I$ is the unit matrix and
\begin{equation}
\Gamma_{nm}=\delta_{n-1,m}a(m)+\delta_{n+1,m}b(m)-\delta_{n,m}c(m),
\label{Gamma}
\end{equation}
with
\begin{eqnarray}
a(m)&=&b_{1}(m)+t_{2}(m),\cr
b(m)&=&t_{1}(m)+b_{2}(m),\cr
c(m)&=&b_{1}(m)+t_{2}(m)+t_{1}(m)+b_{2}(m).
\label{abc}
\end{eqnarray}
The tunneling rates are
\begin{eqnarray}
b_{i} (n)&=&\left[\Delta E_{i}^{-}(n)/e^{2} R_{j}\right]/
\left[1-e^{-\Delta E_{i}^{-}(n)/k_{B}T}\right],\cr
t_{i}(n)&=&\left[\Delta E_{i}^{+}(n)/e^{2}R_{j}\right]/\left[1-e^{-\Delta
E_{i}^{+}(n)/k_{B}T}\right],
\label{apprates}
\end{eqnarray}
where
\begin{eqnarray}
\Delta E_{1}^{\pm}(n)&=&\left[-{e}/{2}\ \pm\left(en-C_{j}
V_{\rm ds} -C_{g} V_{g}\right)\right]{e}/{C_{\Sigma}},\cr
\Delta E_{2}^{\pm}(n)&=&\left[-{e}/{2}\ \mp\left(en+(C_{j}+C_{g})
V_{\rm ds} -C_{g} V_{g}\right)\right]{e}/
{C_{\Sigma}}.
\label{energies}
\end{eqnarray}
A convenient form for the tunneling rates is
\begin{eqnarray}
b_{1}(m)&=&\frac{1}{R_{j}C_{\Sigma}}\frac{n-m+\Delta
n+\tilde{V}_{\rm ds}}{1-\exp\left[-(n-m+\Delta n+\tilde{V}_{\rm
ds})/\tilde{T}\right]},\cr
b_{2}(m)&=&\frac{1}{R_{j}C_{\Sigma}}\frac{m-n-1-\Delta
n+\tilde{V}_{\rm ds}}{1-\exp\left[-(m-n-1-\Delta n+\tilde{V}_{\rm
ds})/\tilde{T}\right]},\cr
t_{1}(m)&=&\frac{1}{R_{j}C_{\Sigma}}\frac{m-n-1-\Delta
n-\tilde{V}_{\rm ds}}{1-\exp\left[-(m-n-1-\Delta n-\tilde{V}_{\rm
ds})/\tilde{T}\right]},\cr
t_{2}(m)&=&\frac{1}{R_{j}C_{\Sigma}}\frac{n-m+\Delta
n-\tilde{V}_{\rm ds}}{1-\exp\left[-(n-m+\Delta n-\tilde{V}_{\rm
ds})/\tilde{T}\right]},
\label{appconvrates}
\end{eqnarray}
where
\begin{eqnarray}
\Delta n&=&\frac{C_{g}V_{g}}{e} -\frac{C_{g}V_{\rm ds}}{2e} -n -
\frac{1}{2},\cr
\tilde{V}_{\rm ds}&=&\frac{C_{\Sigma}V_{\rm ds}}{2e},\cr
\tilde{T}&=&\frac{C_{\Sigma}k_{B}T}{e^{2}},
\label{dimdef}
\end{eqnarray}
and $n$, $n+1$ are the most probable island electron numbers,
determined by the condition
\[
n<\frac{C_{g}V_{g}}{e}-\frac{C_{g}V_{\rm ds}}{2e}<n+1,
\]
which is equivalent to $-0.5<\Delta n<0.5$.
Under the condition $k_{B}T\ll eV_{\rm ds}$ (equivalently
$\tilde{T}\ll\tilde{V}_{\rm ds}$), we have that $t_{1}(n)$,
$b_{2}(n)$, $b_{1}(n+1)$, $t_{2}(n+1)\approx 0$ for small drain-source
voltage, and furthermore $b(m)\approx 0$ for all $m\leq n$ and
$a(m)\approx 0$ for all $m\geq n+1$. Because the probabilities
$\sigma(m)\approx 0$ for all $m\neq n$,~$n+1$, only the terms
$m=n$ and $m=n+1$ need be considered in the summation over $m$
in eqs.~(\ref{appSI}) and (\ref{U}). From the form of the
matrix (\ref{Gamma}) and the above approximate conditions on the probabilities
and tunneling rates, we then also have that only $m'=n$ and $m'=n+1$
need be considered in the sum over $m'$. The matrix elements
$B_{n,n}$, $B_{n,n+1}$, $B_{n+1,n}$, and $B_{n+1,n+1}$ contain only
the tunneling rates $a(n)$, $b(n)$, $c(n)$, and $c(n+1)$ and it is
valid to just consider $2\times 2$ matrices for $\Gamma$ and $B$:
\begin{equation}
B={\rm Re}\left(\begin{array}{cc}
i\omega-c(n) & -b(n)\\
-a(n) & i\omega -c(n+1)
\end{array}
\right)^{-1}.
\label{matrix}
\end{equation}
The probabilities $\sigma(n)$ and $\sigma(n+1)$ are given in Ref.\
\onlinecite{blencowe2} as
\begin{eqnarray}
\sigma(n)&=&b(n+1)/[a(n)+b(n+1)],\cr
\sigma(n+1)&=&a(n)/[a(n)+b(n+1)].
\label{rhon}
\end{eqnarray}
Substituting (\ref{matrix}) and (\ref{rhon}) into (\ref{appSI}) and
(\ref{U}) and neglecting all small terms, we get
\begin{equation}
S_{I}(\omega)=\frac{e^{2}a(n)b(n+1)}{a(n)+b(n+1)}+\frac{e^{2}[f(n)-g(n+1)]}
{a(n)+b(n+1)}\frac{a^{2}(n) g(n+1) -b^{2}(n+1)
f(n)}{[a(n)+b(n+1)]^{2}+\omega^{2}}.
\label{appcurrentnoise}
\end{equation}
\section{Force noise due to the fluctuating island charge}
The electrostatic interaction between the fixed island
and cantilever electrodes of the gate capacitor is
\begin{equation}
F=A[C_{j}(V_{\rm ds}-2V_{g})-ne+Q_{p}]^{2},
\label{appforce}
\end{equation}
where $Q_{p}$ is the background charge and
\begin{equation}
A=C_{g}(2 C_{j}-C_{g})/2dC_{\Sigma}^{3}.
\label{A}
\end{equation}
Define
\begin{equation}
\tilde{Q}=C_{j}(V_{\rm ds}-2V_{g})=-(2n+1+2\Delta n)e C_{j}/C_{g},
\label{Q}
\end{equation}
where the second equality follows from (\ref{dimdef}). Neglecting
$Q_{p}$ in (\ref{appforce}), the force noise is given as
\begin{equation}
S_{F}(\omega)=A^{2}\left\{4\tilde{Q}^{2}e^{2}S_{nn}(\omega)+e^{4}
S_{n^{2}n^{2}}(\omega)-2\tilde{Q}e^{3}\left[S_{n^{2}n}(\omega)+
S_{nn^{2}}(\omega)\right]\right\}.
\label{appSF}
\end{equation}
Similarly to expression (27) in Ref.\ \onlinecite{korotkov}, we have
\begin{equation}
S_{n^{\alpha}n^{\beta}}(\omega)=4\sum_{m,m'}
m'^{\alpha}B_{m',m}m^{\beta}\sigma(m).
\label{correlations}
\end{equation}
Using expressions (\ref{matrix}), (\ref{rhon}), and
(\ref{correlations}) to obtain $S_{nn}(\omega)$,
$S_{n^{2}n^{2}}(\omega)$, and $\left[S_{n^{2}n}(\omega)+
S_{nn^{2}}(\omega)\right]$, the final result for the force noise is
\begin{eqnarray}
S_{F}(\omega)=&&\frac{4A^{2}e^{4}a(n)b(n+1)}{a(n)+b(n+1)}\cr
&&\times \frac{(2C_{j}/C_{g})^{2} (2n+1+2\Delta n)^{2}
+(4C_{j}/C_{g})(2n+1)(2n+1+2\Delta
n)+(2n+1)^{2}}{[a(n)+b(n+1)]^{2}+\omega^{2}}.
\label{appSF2}
\end{eqnarray}
Note that both $S_{I}(\omega)$ and $S_{F}(\omega)$ have a Lorenzian
spectrum.
\section{Simple estimate for the displacement noise}
A simple, approximate formula for the tunneling current
is\cite{beenakker}
\begin{equation}
I_{\rm ds} =I_{\rm ds, max}/\cosh^{2}\left[e C_{g}(V_{g,{\rm
res}}-V_{g})/2.5k_{B}TC_{\Sigma}\right],
\label{appcurrent}
\end{equation}
where $V_{g,{\rm res}}$ is the value of the gate voltage at which the
tunneling current is maximal:
\begin{equation}
V_{g,{\rm res}}=\frac{V_{\rm ds}}{2}+(n+\frac{1}{2})\frac{e}{C_{g}}.
\label{vgres}
\end{equation}
Subsituting (\ref{appcurrent}) and the shot noise limit formula
$S_{I}=2eI_{\rm ds}$ into (\ref{SXI}), we obtain
\begin{equation}
S_{X}^{I}=\frac{1}{2eI_{\rm ds, max}}\left[\frac{2.5 k_{B}T d
C_{\Sigma}}{C_{g}(V_{g}-V_{\rm ds}/2)}\right]^{2}\frac{\cosh^{4}
\left[e C_{g}(V_{g,{\rm
res}}-V_{g})/2.5k_{B}TC_{\Sigma}\right]}{
\sinh^{2}\left[e C_{g}(V_{g,{\rm
res}}-V_{g})/2.5k_{B}TC_{\Sigma}\right]}.
\label{appSXI}
\end{equation}
Using expression (\ref{current1}) for the tunneling current with
$\sigma(n)=\sigma(n+1)=0.5$ and
expressions (\ref{appconvrates}) for the tunneling rates with $\Delta
n=0$ gives for
the maximum tunneling current:
\begin{equation}
I_{\rm ds, max}=V_{\rm ds}/4R_{j}.
\label{currentmax}
\end{equation}
To make a simpler estimate of the displacement noise, we might
replace the dimensionless, parameter-independent function
$\cosh^{4}[x]/\sinh^{2}[x]$ by a constant $K$, say its minimum value which is 4. However, this
function poorly approximates the correct doublet shape and it is
found numerically that the $K\approx 40$ provides a good fit.
Thus we obtain the final result for the displacement noise due
to the shot noise in the tunneling current:
\begin{equation}
S_{X}^{I}=K\frac{2R_{j}}{e V_{\rm ds}}\left[\frac{2.5
k_{B}TdC_{\Sigma}}{C_{g}(V_{g}-V_{\rm ds}/2)}\right]^{2}.
\label{appapproxSXI}
\end{equation}
A simple approximation to the maximum force noise is obtained by
setting $\Delta n=0$ in (\ref{appSF2}) and neglecting small terms:
\begin{equation}
S_{F}(0)=(A^{2}e^{5}R_{j}/V_{\rm
ds})\left[C_{\Sigma}(2n+1)/C_{g}\right]^{2}.
\label{approxsf1}
\end{equation}
Substituting in the definition (\ref{A}) for $A$ and using
$2n+1=C_{g}(2V_{g}-V_{\rm ds})/e$ at maximum tunneling current, we
obtain:
\begin{equation}
S_{X}^{F}=\frac{Q^2 e^{3}R_{j}}{k_{\rm eff}^{2}V_{\rm
ds}}\left[\frac{C_{g}(2C_{j}-C_{g})(V_{g}-V_{\rm ds}/2}
{d C_{\Sigma}^{2}}\right]^{2},
\label{appapproxSXF}
\end{equation}
at the resonant frequency $\omega_{0}$ of the cantilever, which is
assumed to be much smaller than the tunneling rates in magnitude.
\begin{references}
\bibitem{bose}S. Bose, K. Jacobs, and P. L. Knight, Phys. Rev. A {\bf
59}, 3204 (1999).
\bibitem{blencowe}M. P. Blencowe and M. N. Wybourne, Physica B {\bf
280}, 555 (2000); XXXIV Rencontres de Moriond: Quantum Physics at
Mesoscopic Scale, edited by D. C. Glattli, M. Sanquer, and J.
Tr\^{a}n Thanh V\^{a}n, (EDP Sciences, France, 2000), pp. 43-7.
\bibitem{sidles} J. A. Sidles, J. L. Garbini, K. J. Bruland, D. Rugar, O.
Z\"{u}ger, S. Hoen, and C. S. Yannoni, Rev. Mod. Phys. {\bf 67}, 249 (1995).
\bibitem{berman}G. P. Berman, G. D. Doolen, P. C. Hammel, and V. L.
Tsifrinovich, Phys. Rev. B {\bf 61}, 14694 (2000).
\bibitem{schoelkopf} R. J. Schoelkopf, P. Wahlgren, A. A. Kozhevnikov, P.
Delsing, and D. E. Prober, Science {\bf 280}, 1238 (1998).
\bibitem{fujisawa} T. Fujisawa and Y. Hirayama,
Appl. Phys. Lett. {\bf 77}, 543 (2000).
\bibitem{aassime}A. Aassime, G. Johansson, G. Wendin, R. J.
Schoelkopf, and P. Delsing, Phys. Rev. Lett. {\bf 86}, 3376 (2001).
\bibitem{blencowe2}M. P. Blencowe and M. N. Wybourne, Appl. Phys.
Lett. {\bf 77}, 3845 (2000).
\bibitem{schwab}K. C. Schwab (unpublished).
\bibitem{korotkov}A. N. Korotkov, Phys. Rev. B {\bf 49}, 10381 (1994).
\bibitem{korotkov2} A. N. Korotkov and M. A. Paalanen, Appl. Phys. Lett.
{\bf 74}, 4052 (1999).
\bibitem{beenakker}C. W. J. Beenakker, Phys. Rev. B {\bf 44}, 1646 (1991).
\end{references}
\begin{table}
\caption{The resonant frequency $f_{0}$ and mass $m$ of crystalline
Si cantilevers with different lengths ($l$) and fixed width
($w=0.6~\mu{\rm m}$) and thickness ($t=0.15~\mu{\rm m}$).}
\label{Table 1}
\begin{tabular}{ccccc}
$l$ in $\mu{\rm m}$ & $1.25$ & $1.5$ & $2$ & $4$\\
$f_{0}$ in MHz & $117.2$ & $81.36$ & $45.77$ & $11.44$\\
$m$ in pg & $0.262$ & $0.315$ & $0.419$ & $0.839$\\
\end{tabular}
\end{table}
\begin{figure}
%\mbox{\epsfig{file=Fig1.eps, width=6in}}
\vskip 1in
\caption{Scheme of the rf-SET displacement detector.}
\label{Fig.1}
\end{figure}
\begin{figure}
%\mbox{\epsfig{file=Fig2.eps, width=6in}}
\vskip 1in
\caption{Displacement noise due to the tunneling
current shot noise as a function of gate voltage (upper, doublet curves).
The estimate of the lower limit on the displacement noise vs. gate
voltage is also shown (lower, dashed line). The noise analysis
assumes a
symmetric SET at $T=74~{\rm mK} (0.01e^{2}/C_{\Sigma})$ with
junction capacitance $C_{j}= 100~{\rm aF}$,
junction resistance $R_{j}= 50~{\rm k}\Omega$, and static gate
capacitance $C_{g}= 50~{\rm aF}$
(the gate capacitor gap is $d=0.1~\mu{\rm m}$). The drain-source voltage is
$V_{\rm ds}=0.38~{\rm mV} (0.6e/C_{\Sigma})$. The displacement noise is
determined at 117~MHz-the resonant frequency of the cantilever.}
\label{Fig.2}
\end{figure}
\begin{figure}
%\mbox{\epsfig{file=Fig3.eps, width=6in}}
\vskip 1in
\caption{Displacement noise due to the force noise from the
fluctuating island charge as a function of gate voltage (lower curves).
The estimate of the upper limit on the displacement noise is also shown
(upper, dashed line). The noise analysis assumes the same parameters as
in Fig. 2. }
\label{Fig.3}
\end{figure}
\begin{figure}
%\mbox{\epsfig{file=Fig4.eps, width=6in}}
\vskip 1in
\caption{The total displacement noise as a function of gate voltage
(upper, doublet curves). For illustrative purposes, the sample voltage range
is chosen such that the force and current displacement noise contributions
are comparable. The tunneling current is given by the lower,
dashed line.}
\label{Fig.4}
\end{figure}
\begin{figure}
%\mbox{\epsfig{file=Fig5.eps, width=6in}}
\vskip 1in
\caption{The total displacement noise as a function of gate voltage
for the different cantilever lengths listed in Table I. The
same parameter values as in Fig. 2 are assumed. The displacement noise
at the resonant frequency of the cantilever is determined at the gate
voltage for which the tunneling current is fixed at half its maximum
value.}
\label{Fig.5}
\end{figure}
\begin{figure}
%\mbox{\epsfig{file=Fig6.eps, width=6in}}
\vskip 1in
\caption{Total displacement noise for different junction
capacitances. Except for the junction capacitances, the same parameter
values as in Fig. 2 are assumed and the displacement noise is
determined at the tunneling current half-maximum.}
\label{Fig.6}
\end{figure}
\begin{figure}
%\mbox{\epsfig{file=Fig7.eps, width=6in}}
\vskip 1in
\caption{The displacement noise for different percentages of the
maximum tunneling current. The same parameters as in Fig.~2 are used.}
\label{Fig.7}
\end{figure}
\vfill
\eject
\mbox{\epsfig{file=Fig1.eps, width=6in}}
\mbox{\epsfig{file=Fig2.eps, width=6in}}
\mbox{\epsfig{file=Fig3.eps, width=6in}}
\mbox{\epsfig{file=Fig4.eps, width=6in}}
\mbox{\epsfig{file=Fig5.eps, width=6in}}
\mbox{\epsfig{file=Fig6.eps, width=6in}}
\mbox{\epsfig{file=Fig7.eps, width=6in}}
\end{document}