The millipede, a very dense, highly parallel scanning-probe data-storage system

Abstract

Ultrahigh storage densities of up to 1 Tbit/in². can be achieved by local-probe techniques to write, read back, and erase data in very thin polymer films. The thermo-mechanical scanning-probe-based data-storage concept called Millipede combines ultrahigh density, terabit capacity, small form factor, and high data rate. After illustrating the principles of operation of the Millipede, we introduce a channel model for the analysis of the read back process, and compare analytical results with experimental data.

Full text

Page 1

Reprint









The Millipede, a Very Dense, Highly Parallel Scanning-
Probe Data-Storage System

G. Cherubini, T. Antonakopoulos, P. Bachtold, G. K. Binnig, M. Despont,
U. Drechsler, A. Dholakia, U. Durig, E. Eleftheriou, B. Gotsmann, W.
Haberle, M. A. Lantz, T. Loeliger, H. Pozidis, H. E. Rothuizen, R. Stutz,
and P. Vettiger









The 28th European Solid-State Circuits Conference –
ESSCIRC2002


24 - 26 SEPTEMBER 2002, FLORENCE, ITALY


Copyright Notice: This material is presented to ensure timely dissemination of scholarly and technical
work. Copyright and all rights therein are retained by authors or by other copyright holders. All persons
copying this information are expected to adhere to the terms and constraints invoked by each author's
copyright. In most cases, these works may not be reposted or mass reproduced without the explicit
permission of the copyright holder.

Page 2

The Millipede, a Very Dense, Highly Parallel Scanning-Probe
Data-Storage System
G. Cherubini, T. Antonakopoulos, P. Bächtold, G. K. Binnig, M. Despont, U. Drechsler,
A. Dholakia, U. Dürig, E. Eleftheriou, B. Gotsmann, W. Häberle, M. A. Lantz,
T. Loeliger, H. Pozidis, H. E. Rothuizen, R. Stutz, and P. Vettiger
IBM Research
Zurich Research Laboratory
8803 Rüschlikon, Switzerland
cbi@zurich.ibm.com
Abstract
Ultrahigh storage densities of up to 1 Tbit/in.2 can be
achieved by local-probe techniques to write, read back,
and erase data in very thin polymer films. The thermo-
mechanical scanning-probe-based data-storage concept
called Millipede combines ultrahigh density, terabit
capacity, small form factor, and high data rate. After
illustrating the principles of operation of the Millipede,
we introduce a channel model for the analysis of the
read back process, and compare analytical results with
experimental data.
1. Introduction
Techniques that use nanometer-sharp tips for imaging
and investigating the structure of materials down to the
atomic scale, such as the atomic force microscope
(AFM) and the scanning tunneling microscope (STM)
[1–3], are suitable for the development of ultrahigh-
density storage devices [4–9]. As the simple tip is a very
reliable tool for the ultimate local confinement of
interaction, tip-based storage technologies appear as
natural candidates for extending the physical limits that
are being approached by conventional magnetic storage.
Areal densities achievable by today’s magnetic record-
ing technologies are limited to about 100 to 150
Gbit/in.2 by the well-known superparamagnetic limit.
Several proposals have been formulated to overcome
this limit, for example the adoption of patterned
magnetic media, for which, however, the biggest
challenge remains the patterning of the magnetic disk in
a cost-effective manner. On the other hand, data rates
well above 800 Mbit/s are achieved by magnetic record-
ing, whereas the mechanical resonant frequencies of the
AFM cantilevers limit the data rates of a single canti-
lever to a few Mbit/s for AFM data storage. The feed-
back speed and low tunneling currents limit STM-based
storage approaches to even lower data rates. The
solution for substantially increasing the data rates
achieved by tip-based storage devices is to employ
arrays of cantilevers operating in parallel, with each
cantilever performing read/write/erase operations over
an individual storage field [7–9].
In this paper, we consider the “Millipede” concept, as
described in [7–9], for the realization of highly parallel
scanning-probe data storage, characterized by areal den-
sities up to 0.5 to 1 Tbit/in.2, far beyond the expected
superparamagnetic limit, and parallel operation of very
large two-dimensional (32×32) AFM cantilever arrays
with integrated tips and write/read functionality. After
illustrating the principle of the Millipede, we introduce
an equivalent model for the characterization of the read-
back signal from a thermomechanical sensor, and
compare analytical results with experimental data.
2. Principles of operation of the Millipede
The Millipede device shown in Fig. 1 is a highly
parallel scanning-probe data-storage system, where in-
formation is stored as sequences of “indentations” and
“no indentations” written on nanometer-thick polymer
films using an array of AFM cantilevers. Each cantilever
performs write/read operations over an individual
storage field with size on the order of 100×100 !m2 [7–
9]. Thermomechanical writing is achieved by applying a
local force by the cantilever/tip to the polymer layer, and
simultaneously softening the polymer layer by local
heating. Initially, the heat transfer from the tip to the
polymer through the small contact area is very poor, but
2D cantilever array chip
Polymer storage medium
on xyz scanner
x
z2
z3
z1
y
Multiplex-driver
Figure 1. Illustration of the Millipede concept. From [7].

Page 3

improves as the contact area increases. This means that
the tip must be heated to a relatively high temperature of
about 400ºC to initiate the softening. Once softening has
initiated, the tip is pressed into the polymer, and hence
increases the bit size. Figure 2a shows data bits of 40 nm
in diameter that have been written using a 1-!m thick,
70-!m long, two-legged Si cantilever, whose legs are
made highly conducting by high-ion implantation,
whereas the heater region remains low-doped. Figure 2b
shows that 40-nm bits can be written very close to each
other without merging, implying a potential storage
areal density on the order of 400 Gbit/in.2. For example,
a (32×32) cantilever array with 1024 storage fields, each
having an area of 100×100 !m2, has a capacity of about
6 Gbit on a chip area on the order of 3×3 mm2. More
recently single-cantilever areal densities up to 1 Tbit/in.2
have been demonstrated, although currently at a some-
what degraded write/read quality, as illustrated in Fig.
2c.
To read the written information, the heater cantilever
is given the additional function of a thermal readback
sensor by exploiting its temperature-dependent resis-
tance. In general, the resistance increases nonlinearly
with heating power/temperature from room temperature
to a peak value of 500–700ºC. The peak temperature is
determined by the doping concentration of the heater
platform, which ranges from 1×1017 to 2×1018 cm–3.
Above the peak temperature, the resistance drops as the
number of intrinsic carriers increases because of thermal
excitation. For sensing, the resistor is operated at about
350ºC, a temperature that is not high enough to soften
the polymer as in the case of writing. The principle of
thermal sensing is based on the fact that the thermal
conductance between heater platform and storage sub-
strate changes according to the distance between them.
The medium between the heater platform and the
storage substrate, in our case air, transports heat from
the cantilever to the sample. When the distance between
cantilever and sample is reduced as the tip moves into a
bit indentation, the heat transport through the air
becomes more efficient. As a result, the evolution of the
heater temperature in response to a pulse applied to the
cantilever is different and, in particular, the maximum
value achieved by the temperature is smaller than in the
case in which no bit indentation is present. As the value
of the variable resistance depends on the temperature of
the cantilever, the maximum value achieved by the
resistance will be smaller as the cantilever moves over
an indentation. Therefore, during the read process, the
cantilever resistance reaches different values whether it
moves over an “indentation” (bit “1”) or a “no indenta-
tion” (bit “0”). The thermomechanical cantilever sensor,
which transforms temperature into an electrical signal
that carries information, is electrically equivalent, to a
first degree of approximation, to a variable resistance. A
detection circuit must therefore sense a voltage that
depends on the value of the cantilever resistance to make
a decision whether a “1” or a “0” is being written. The
relative variation of thermal resistance is on the order of
10–5/nm. Hence a written bit “1” typically produces a
relative change of the cantilever thermal resistance
ΘΘ
∆
RR /
of about 10–4 to 5 × 10–4. Note that the relative
change of the cantilever electrical resistance is on the
same order of magnitude. As a consequence, one of the
most critical issues in detecting the presence or absence
of an “indentation” is the high resolution required to
extract the signal that contains the information about the
bit being “1” or “0”. The signal carrying the information
can be viewed as a small signal superimposed to a very
large offset signal.
Write/read operations depend on a mechanical
parallel x/y scanning of either the entire cantilever array
chip or the storage medium. The tip-medium contact is
maintained and controlled globally, i.e., not on an indi-
vidual cantilever basis, by using a feedback control for
the entire chip, which greatly simplifies the system.
Early results demonstrating the concept of the entire
chip approach/leveling [10] indicate that overall chip
tip-apex height control to within 500 nm is feasible. The
stringent requirement for tip-apex uniformity over the
entire chip is determined by the uniform force required
to reduce tip and medium wear due to large force
variations resulting from large tip-height nonuni-
formities [11]. As the Millipede tracks the entire array
without individual lateral cantilever positioning, thermal
expansion of the array chip has to be small or well
controlled. For a 3×3 mm2 silicon array area and 10-nm
tip-position accuracy, the chip temperature has to be
controlled to about 1ºC. This is ensured by four
temperature sensors in the corner of the array and heater
elements on each side of the array. Thermal-expansion
considerations are a strong argument for a two-dimen-
sional array arrangement instead of one-dimensional,
which would make a chip 32 times longer for a (32×32)
array of cantilevers.
120 nm
200 nm
(c)
0.5
1 Tb/in2
N
2
IBITT
1
/
(a)
(b)
Figure 2. Series of 40-nm data bits formed in a uniform
array with (a) 120-nm and (b) variable pitch, resulting in
bit areal densities of up to 400 Gb/in.2. (c) Ultra-high-
density bit writing with areal densities approaching 1
Tb/in.2. From [9].

Page 4

Parallel operations of large two-dimensional arrays is
achieved by a row/column time-multiplexed addressing
scheme similar to that implemented in DRAMs. In the
case of Millipede, the multiplexing scheme is used to
address the array column by column with full parallel
write/read operation within one column [9]. In particu-
lar, readback-signal samples are obtained by applying a
read pulse to the cantilevers in a column of the array,
low-pass filtering the cantilever response signals, and
finally sampling the filter output signals. This process is
repeated sequentially until all columns of the array are
addressed, and then restarted from the first column. The
time between two pulses corresponds to the time it takes
for a cantilever to move from one bit position to the
next. For a (32×32) cantilever array, a distance between
the centers of two consecutive indentations (pitch) of 40
nm, a constant velocity of the x/y scanner of 250 !m/s,
and pulses of duration of 5 !s that are periodically
applied to each cantilever at a rate of 6.25 kHz, a data
rate of 6.4 Mbit/s is achieved using the multiplexing
scheme described above. A full parallel operation would
increase the data rate to 204.8 Mbit/s.
3. The read-channel model
In this section, we consider the readback channel for
a single cantilever, scanning a storage field where bits
are written as indentations or no indentations, which will
also be referred to as “logical marks”, in the storage
medium. The model of the read channel, which we
consider for the analysis of the detection system, is
illustrated in Fig. 3. As discussed in Section 2, a canti-
lever is modeled as a variable resistance that depends on
the temperature at the cantilever tip.
To evaluate the evolution of the temperature of a
heated cantilever during the read process, we resort to a
simple RC-equivalent thermal circuit, as illustrated in
Fig. 4, where
)1 (
x
R
η+
and
resistance and capacitance, respectively. The parameter
ΘΘ
∆=
RRxx
/
η
indicates the relative variation of thermal
resistance resulting from the small change in the air gap
width between the cantilever and the storage medium, as
compared to the case of a flat surface without
indentations. Subscript x indicates the x-distance in the
direction of scanning from the initial point. Therefore,
the parameter
x
η will assume the largest absolute value
when the tip of the cantilever is located at the center of
an indentation. The heating power that is dissipated in
the cantilever heater region is expressed as
ΘΘ
C denote the thermal

,
)),((
)(
)),(,(
2
xtR
tV
e
xttP
C
Θ
e
=Θ
(1)
where
Θ
the temperature-dependent cantilever resistance.
As the heat-transfer process depends on the value of
the thermal resistance and on the read-pulse waveform,
),( xt
Θ
depends on time t and distance x. However, as
the time it takes for the cantilever to move from the
center of a logical mark to the next is much larger than
the duration of a read pulse, we assume that
does not vary significantly as a function of x during the
period a read pulse is applied, and that it decays to the
ambient temperature
0
Θ before the next pulse is
applied. Therefore the evolution of the cantilever
temperature in response to a pulse applied at time
vxt/
0
, at a certain distance x0 from the initial point
of scanning and for a certain constant velocity v of the
scanner, obeys a differential equation that is expressed
as
1
) , (
0
R
x
+
η
)(tVC
is the cantilever temperature, and
is the voltage across the cantilever,
),( xt)),((xtReΘ
is
),( xt
Θ
0=

.
)),((
)
x
(
t
1
)),((
)1 (
0
2
00
0
R
tV
Θ
C
xt
C
xt
e
C
=
Θ−Θ+Θ′
Θ
ΘΘ
(2)
With reference to the block diagram of the read
channel illustrated in Fig. 3, the source generates the
read pulse
)()(tVtV
CP
=
that is applied to the cantilever
variable resistance. Furthermore, the active low-pass RC
detector filter, where Rlpf and Clpf denote the resistance
and capacitance of the low-pass filter, respectively, is
realized using an ideal operational amplifier that exhibits
infinite input impedance, zero output impedance, and
infinite frequency-independent gain. The readback
signal
),,(tVo
which is obtained at the low-pass filter
output in response to
)),/ )rect(()(
0
τ
ttAtVP
−=
where
0 x
the applied voltage




≤≤
=




otherwise0
0 if1
rect
τ
τ
t
t
,(3)
and A denotes the pulse amplitude, obeys the differential
equation

V
.)(
)),((
),(
1
C
),(
0
lpf
t
0
lpflpf
0






Θ
+−=
′
tV
xR
R
xtV
R
xt
P
e
oo
(4)
Clock rate
1/T
Reference
cantilever
V (t)
p
Rlpf
Clpf
Sensor
cantilever
Threshold
decision
V (t)
p
Figure 3. Block diagram of the detection circuit.
"
R (1+#x)
C"
Heating
power
Figure 4. RC-equivalent thermal model of the heat
transfer process.

Page 5

As the voltage at the output of the low-pass filter
depends on the variable resistance value
the readback signal is determined by solving jointly the
differential equations (2) and (4), with initial conditions
),(
Θ=Θ
xt
and
,(
0
xtVo
comparison between experimental and synthetic read-
back signals is given in Figs. 5 and 6 for a time constant
of the low-pass filter
lpf
τ
=
the duration of the applied rectangular pulse.
Assuming that ideal control of the scanner is
performed, such that the time of application of a read
pulse corresponds either to the cantilever being located
at the center of an indentation for detecting a bit “1”, or
away from an indentation for detecting a bit “0”, two
possible responses are obtained at the output of the low-
pass filter as solutions of (2) and (4), which we denote
by
) 1|, (
0
0
xo
xtV
and
o
V
ly. By sampling the readback signal at the
instant
,
0
τ+= tts
simple threshold detection may in
principle be applied to detect a written bit, where the
value of the threshold is given by
)),((
0 xtReΘ
,
000
. 0)
0
=
As an example, a
s, 18. 1
µ
and two values of
=
α
) 0|, (t
0
0
=
x
x
α
, respective-

V] ) 0|,() 1 |,(½[
00
00 Th
=+==
xsoxso
xtVxtV
αα
.
(5)
As mentioned in Section 2, one of the most critical
issues in detecting the presence or absence of an inden-
tation is the high resolution required to extract the small
signal
) 1|,(
0
0
xso
xtV
contains the information about the bit being “1” or “0”,
superimposed to the offset signal
) 0|,(
0
0
=−=
xso
xtV
αα
which
).0|,(
0
0
=
xo
xtV
α
As
illustrated in Fig. 3, a solution to this problem consists in
subtracting from the readback signal a reference signal
that is obtained by applying at time
)(tVP
to a cantilever scanning a storage field where no
0tt =
the read pulse
indentation is written. The readback signal is thus given
by
~
00
−=
oo
VxtVxtV

).0|,(),(),(
0
0ref,
=
xo
xt
α
(6)
In this case, the threshold is set at
~
Th
=
so
tVV

] ) 0|,( ) 1|,( ½[
00
00
=−=
xsox
xtVx
αα
. (7)
The implementation of the detection scheme analyzed
here is presented in [12].
We consider now read pulses that are periodically
applied at the instants
tn=
sufficiently low, i.e.
, τ>>
T
temperature and the output voltage achieve the above
initial conditions at the instants tn. We assume that ideal
timing recovery is performed, i.e. the period T is equal
to the time it takes for the cantilever to move from the
center of a logical mark to the next, and that at the
instants tn the cantilever is located at the center of logi-
cal marks. Setting
0x
, the readback signal samples
that are obtained in response to N pulses applied to the
cantilever for detecting a sequence of N binary symbols
are expressed as
, nT
where the rate 1/T is
such that the cantilever
0=

, 1, 0,),,(
~
V)(
,,
1
0
,
−=+==∑
n
~
o
V
−
=
NiiT txtts
isnis
N
ois
K
τ
(8)
where
), ( t
n
x
is given by (6) for pulses applied at time
xn=
from the initial point of
scanning. Note that the functions
=
n xno
xtV
in (6) are given by the solution of
the differential equations (2) and (4) for
=
x
η
respectively.
The readback signal (6) at the output of the low-pass
filter is observed in the presence of additive noise.
Therefore, the readback signal for detection of the i-th
binary symbol is given by

()()(
,,isis
twtstr
+=
tn and at distance
nTv
), (
o
V
n
xt
and
) 0 |, (
ref,
α
ΘΘ
n x
∆=
RR
x
/
η
and
, 0
),
,is
(8)
where w(t) denotes the noise signal. The components of
the noise signal that must be taken into account are
thermal noise (Johnson’s noise) from the sensor and the
reference cantilever resistances, which during the read
process achieve a temperature of about Θ1 = 350ºC, and
from the low-pass filter resistance, as well as noise from
equivalent noise sources in the operational amplifier.
To determine system performance, we evaluate the
signal-to-noise ratio at the detection point, expressed as

SNR,
~
V
σ
log10
2
w
2
Th
10






=
(10)
where the variance of the noise is approximated as

σ
,
21
2))(/)](()(4 [
lpf
lpf0
2
1lpf
f
τπ
OA11
2
w ∫
df
j
RkRRf
+
WRk
ee
+∞
∞−
Θ+Θ+ΘΘ
=
(11)
(a)(b)
Figure 5. (a) Experimental and (b) synthetic readback
signals for
τ =
s.25.10
µ
(a)(b)
Figure 6. (a) Experimental and (b) synthetic readback
signals for
τ =
s.25.15
µ

Page 6

where
and
power spectral density of the operational amplifier. For
typical values of the system parameters, an SNR in the
range 14 to 20 dB is obtained.
Assuming that the indentations have a regular shape,
which may be derived by the visco-elastic model of bit
writing described in [9], or approximated by simple
functions of the raised-cosine type, a synthetic model for
the simulation of the readback signal to evaluate
detection and servo/timing recovery algorithms is
obtained by applying oversampling pulses to the
cantilever at the rate K/T, provided the condition
τ>>
KT/
is satisfied. A comparison between the read-
back signal obtained along a data track and that obtained
by the synthetic model is shown in Fig. 7.
J/K1038
denotes the equivalent input voltage noise
. 1
)
23
−
×=
(
k
is the Boltzmann constant
OAfW
4. Conclusion
The Millipede has the potential to achieve ultra high
storage areal densities, on the order of 1 Tbit/in.2. The
high areal storage density, small form factor, and low
power consumption make Millipede very attractive as a
candidate future storage technology for mobile applica-
tions, offering several gigabytes capacity at data rates of
several megabytes per second. The read-channel model
introduced in this paper provides the methodology for
analyzing the system performance and assessing various
aspects of the detection and servo/timing algorithms that
are key to achieving the system reliability required by
the applications envisaged.
References
[1] G. Binnig, H. Rohrer, Ch. Gerber and E. Weibel,
“7×7 Reconstruction on Si(111) Resolved in Real
Space,” Phys. Rev. Lett., 50 (1983) pp. 120-123.
[2] G. Binnig, C. F. Quate, and C. Gerber, “Atomic
Force Microscope,” Phys. Rev. Lett., 56 (1986) pp.
930-933.
[3] H. J. Mamin, L. S. Fan, S. Hoen, and D. Rugar,
“Tip-based Data Storage Using Micromechanical
Canti–levers,” Sensors and Actuators A, 48 (1995),
pp. 215-219.
[4] C. F. Quate, “Method and Means of Data Storage
Using Tunnel Current Data Readout,” US Patent
No. 4 575 822 (1986).
[5] H. J. Mamin, B. D. Terris, L. S. Fan, S. Hoen, R. C.
Barrett, and D. Rugar, “High-Density Data Storage
Using Proximal Probe Techniques,” IBM J. Res.
Develop., 39 (1995), pp. 681-700.
[6] H. J. Mamin, R. P. Ried, B. D. Terris, and D. Rugar,
“High-Density Data Storage Based on the Atomic
Force Microscope,” Proc. IEEE, 87 (1999), pp.
1014-1027.
[7] M. Despont, J. Brugger, U. Drechsler, U. Dürig, W.
Häberle, M. Lutwyche, H. Rothuizen, R. Stutz, R.
Widmer, G. Binnig, H. Rohrer, and P. Vettiger,
“VLSI-NEMS Chip for AFM Data Storage,” in
Technical Digest 12th IEEE Int’l Micro Electro
Mechanical Systems Conf. “MEMS ‘99” (IEEE,
1999), pp. 564-569.
[8] P. Vettiger, M. Despont, U. Drechsler, U. Dürig, W.
Häberle, M. I. Lutwyche, H. E. Rothuizen, R. Stutz,
R. Widmer, and G. K. Binnig, “The ‘Millipede’ -
More than One Thousand Tips for Future AFM
Data Storage,” IBM J. Res. Develop., 44 (2000), pp.
323-340.
[9] P. Vettiger, G. Cross, M. Despont, U. Drechsler, U.
Dürig, B. Gotsmann, W. Häberle, M. A. Lantz, H.
E. Rothuizen, R. Stutz, and G. K. Binnig, “The
‘Millipede’—Nanotechnology
Storage,” IEEE Trans. Nanotechnol., vol. 1, no. 1
(2002, in press).
[10] M. Lutwyche, C. Andreoli, G. Binnig, J. Brugger,
U. Drechsler, W. Häberle, H. Rohrer, H. Rothuizen,
P. Vettiger, G. Yaralioglu, and C. Quate, “5×5 2D
AFM Cantilever Arrays: A First Step Towards a
Terabit Storage Device,” Sensors and Actuators A
73 (1999), pp. 89-94.
[11] B. D. Terris, S. A. Rishton, H. J. Mamin, R. P.
Ried, and D. Rugar, “Atomic Force Microscope-
Based Data Storage: Track Servo and Wear Study,”
Appl. Phys. A, 66 (1998), pp. S809-S813.
[12] T. Loeliger, P. Bächtold, G. K. Binnig, G.
Cherubini, U. Dürig, E. Eleftheriou, P. Vettiger, M.
Uster, and H. Jäckel, “CMOS Sensor Array with
Cell-Level Analog-to-Digital Conversion for Local
Probe Data Storage,” these proceedings.
Entering Data
0 0 0 1 1 1 1 1 0 1 0 0 0 1 0 1 0 1 0 0 1 0 1 1 1 0 0 0 1 0 0
(a) Experimental data
0 1 1 1 1 0 1 1 1 0 1 1 1 1 1 0 1 1 1 0 0 0 1 0 0 0 0 0 0 1 1
(b) Synthetic model
Figure 7. Comparison between (a) the readback signal
obtained experimentally along a data track and (b) the
readback signal obtained by the synthetic model.