Polyethylene Nanofibers with Very High Thermal Conductivities · the nanofiber and QC the heat...
Transcript of Polyethylene Nanofibers with Very High Thermal Conductivities · the nanofiber and QC the heat...
SUPPLEMENTARY INFORMATIONdoi: 10.1038/nnano.2010.27
nature nanotechnology | www.nature.com/naturenanotechnology 1
Supplemental Information
Polyethylene Nanofibers with Very High Thermal Conductivities
Sheng Shen1, Asegun Henry1, Jonathan Tong1, Ruiting Zheng1,2 and Gang Chen1*
1Department of Mechanical Engineering, Massachusetts Institute of Technology, Cambridge, MA 02139, USA
2Key Laboratory of Radiation Beam Technology and Materials Modification of Ministry of Education, College of Nuclear Science and Technology, Beijing Normal University; Beijing, 100875, China
* Email: [email protected]
1. Fabrication of ultradrawn polyethylene nanofibers
A decalin solution, containing 0.8 wt. % ultrahigh molecular weight polyethylene,
is prepared by heating the polymer-solvent mixture at 145 oC. To prevent the degradation
of polyethylene, the dissolution was carried out under nitrogen. The solution is then
quenched in water to form a gel. The nanofiber pulling system is illustrated in Fig. S1.
The fabrication of the ultradrawn polyethylene nanofibers in the current work includes
two steps. First, a small sample of wet gel is heated by heater 1. After reaching 120 oC,
the heater is turned off to reduce the evaporation of the solvent from the gel and a 100-
200 μm long suspended fiber is rapidly drawn using a sharp tungsten tip or an AFM
cantilever, which is fixed on a motorized stage. Second, heater 2, located underneath the
fiber and heater 1, is used to heat the fiber and surrounding air to ~ 90 oC. After several
seconds, when the two ends of the fiber dry out and solidify, further drawing is conducted
by moving the tungsten tip or the AFM cantilever at a speed of ~ 1 μm/s to achieve the
higher draw ratios.
© 2010 Macmillan Publishers Limited. All rights reserved.
2 nature nanotechnology | www.nature.com/naturenanotechnology
SUPPLEMENTARY INFORMATION doi: 10.1038/nnano.2010.27
2. Stretching ratios of the ultradrawn nanofibers
To estimate the draw ratio in Step 1, we approximate the polyethylene gel, before
deformation, as a thick cylinder with diameter D1 and length L1 (Fig. S2), and assume
that the deformation caused by drawing in Step 1 only occurs in the thick cylinder. In
reality, the two ends are conically shaped, but for approximation purposes, we neglect the
change in volume within the conical ends. The end of the AFM cantilever used in the
drawing process is ~ 4 μm wide. We assume the diameter of the initial cylinder is D1 = ~
6 μm, accounting for the heated polymer gel flowing over the cantilever end due to the
capillary force when the AFM cantilever dips into it. After the drawing in Step 1, the
length of the polyethylene fiber was measured to be L2 = 100-200 μm and the diameter,
D2=1.5-2.5 μm (Fig. S2). Based on volume conservation within the cylindrical section,
we can calculate L1 by 44
22
212
1 LDLD ππ= . Hence, the draw (stretching) ratio in Step 1
can be estimated to be 2
2 12
1 2
6 ~ 16L DL D
= ≈ . In Step 2, the final length of the nanofiber L3
has been successfully drawn in the range of 1,000-10,000 μm, thus giving a draw ratio of
3
2
10 ~ 50LL
≈ . The overall draw ratio in the two steps is thus 3 3 2
1 2 1
60 ~ 800L L LL L L
= ⋅ ≈ for
ultradrawn nanofibers. Such a large range in draw ratio is indicative of our ability to
control the geometrical parameters of our nanofiber samples; though certainly, higher
draw ratios are more difficult to achieve.
The overall draw ratios for three samples presented in the paper are estimated to
be: ~ 410 for sample 1, ~ 270 for sample 2 and ~ 160 for sample 3. Based on the
© 2010 Macmillan Publishers Limited. All rights reserved.
nature nanotechnology | www.nature.com/naturenanotechnology 3
SUPPLEMENTARY INFORMATIONdoi: 10.1038/nnano.2010.27
accuracies of length and diameter measurements of polymer fibers during the drawing
process, we estimated that the uncertainty in the draw ratios is around 10 %.
3. Thermal conductivity measurement system
The thermal conductivity measurement system can be represented by a three-
junction thermal circuit with two thermal resistances corresponding to the polyethylene
nanofiber and the AFM bi-material cantilever, as shown in Fig. S3. The two quantities
that are varied during the measurement are the laser power absorbed by the cantilever QB
and the temperature of the thermocouple T
B
A. The specific thermal resistance of the
nanofiber is ~10 K/W, while typical thermal contact resistance between two solids is
usually 10 -10 K/W range. In addition, we have also used silver epoxy to join the
nanofiber and the thermocouple, thus increasing the contact area. Hence, the thermal
contact resistance at the two ends of the nanofiber is neglected in the following analysis.
Note that if there is thermal contact resistance, the fiber thermal conductivity will be
higher than reported values. The nanofiber is assumed to be uniformly cylindrical along
its length. Although irregularities at the ends are possible, they are neglected in this
analysis. Our SEM images of the nanofiber justify this assumption. All the measurements
are done under high vacuum (< 50 μTorr) and therefore heat convection from air is
negligible. Based on the low emissivity (~ 0.1) of the microfiber , we estimate that the
emissivity of the nanofiber is smaller than 0.1 according to Rayleigh scattering theory
where emissivity is proportional to volume. Due to the very low emissivity (< 0.1) of the
nanofiber and the small temperature difference between the nanofiber and the
surrounding, the radiation loss from the fiber is estimated to be < 1 nW, which is
9
7 8 1
2,3
© 2010 Macmillan Publishers Limited. All rights reserved.
4 nature nanotechnology | www.nature.com/naturenanotechnology
SUPPLEMENTARY INFORMATION doi: 10.1038/nnano.2010.27
negligibly small compared to the heat conduction (~ 100 nW) through the nanofiber. As
an approximation, we also assumed that the heat input from the laser can be modeled as
an input at point B. But, in reality, the distance between the cantilever tip and the laser
spot on the cantilever was estimated to be ~ 20 μm. Since the cantilever is very thermally
conductive, we neglect the thermal resistance between the cantilever tip and the laser spot
in the heat transfer model. This assumption will be justified in subsequent sections.
The two-step process developed in our paper can be understood as follows. In the
first step, we calibrate the bending of the cantilever by varying the laser power. In the
second step, we measure the heat transfer via the nanofiber by varying the thermocouple
temperature. These two steps are further explained below.
Step 1: Calibration via Varying of Laser Power
At steady state, the heat balance of the three junctions at point B in Fig. S3 is,
, (1) CAB QQQ +=
where QB is the laser energy absorbed by the cantilever. QB
C
)
A is heat conduction through
the nanofiber and QC the heat conduction through the cantilever. Since the thermal
conductance of the AFM cantilever GC (~ 10 μW/K) is around three orders of magnitude
larger than the thermal conductance of the nanofiber GF (~10 nW/K), the heat conduction
through the nanofiber (~ 100 nW) can be ignored compared to the change of the
absorbed power (~ 10 μW in Fig. 3 (c)) at point B. Equation (1) can then be written
in differential form as follows,
AQΔ
BQΔ
, (2) BQ QΔ = Δ
In Fig. 3 (a), we can obtain the bending1 2(P PB −Δ when the absorbed power on the
cantilever tip changes from P1 to P2. Since the bending signal measured by the
© 2010 Macmillan Publishers Limited. All rights reserved.
nature nanotechnology | www.nature.com/naturenanotechnology 5
SUPPLEMENTARY INFORMATIONdoi: 10.1038/nnano.2010.27
photodiode is effectively a representation of the cantilever temperature, which is
proportional to the heat transfer through the cantilever, , it can also be shown that, CQ
1 2 1 2( ) 1 (C P P P PQ )Bα−Δ = ⋅Δ − , (3)
where α1 is a proportionality constant, determined by the properties and dimensions of
the cantilever.4 From Eqs. (2) and (3), we have,
1 2
1 2
(1
( )
B P P
P P
QB
α −
−
Δ=
Δ) , (4)
Step 2: Variation of Thermocouple Temperature
The second step measures the heat transfer changes through the nanofiber when
the temperature of the thermocouple is changed. The laser input to the cantilever is held
constant. In Fig. 3 b, when the temperature of the thermocouple TA changes from TA1 to
TA2, the heat fluxes through the nanofiber can be expressed as,
) , (5) ( 11 ABFA TTGQ −=
) , (6) ( 22 ABFA TTGQ −=
where GF is the thermal conductance of the nanofiber. Since GC is much larger than GF,
the temperature of cantilever tip TB is assumed as a constant during the temperature
change of the thermocouple. The heat conduction through the nanofiber in the experiment
is estimated to be ~100 nW which corresponds to a ~ 10 K temperature change on T
B
2
-2B.
This justifies our assumption. Thus, subtracting Eq. (6) from Eq. (5), we have,
1 2( ) 1A AA T T F A AQ G T−Δ = − Δ −
)
, (7)
where can be obtained by the thermocouple measurement. In Fig. 3b, we can also
obtain the bending of the cantilever caused by the temperature change from T
1 2A AT −Δ
1 2( A AT TB −Δ A1
© 2010 Macmillan Publishers Limited. All rights reserved.
6 nature nanotechnology | www.nature.com/naturenanotechnology
SUPPLEMENTARY INFORMATION doi: 10.1038/nnano.2010.27
to TA2. Similarly, we can relate the heat flux through the cantilever and the bending of the
cantilever by,
1 2 1 2( ) 2 (A A A AC T T T TQ B )α−Δ = ⋅Δ −
)
, (8)
where is the change of the heat flux through the cantilever due to the
temperature change from T
1 2( A AC T TQ −Δ
A1 to TA2, and α2 is a proportionality constant when changing
the temperature of the thermocouple.
In Step 2, the absorbed laser power does not change (BQ 0=Δ BQ ). This leads to
the following,
, (9) CA QQ Δ−=Δ
Thus, based on Eqs. (8) and (9), we have,
1 2 1 2 1 2( ) ( ) 2 (A A A A A AA T T C T T T TQ Q B )α− −Δ = −Δ = − ⋅Δ − , (10)
From Eqs. (7) and (10), the proportionality constant α2 can be expressed as,
1 2
1 22
( )A A
A AF
T T
TGB
α −
−
Δ= ⋅
Δ, (11)
In these two cases, the proportionality constants α1 and α2 are equal, which will be
proved in the next sections. In terms of Eqs. (4) and (11), the conductance of the
nanofiber can be expressed by,
1 2 1 2
1 2
( ) (
1 2 ( )
//
A A
B P P P PF
A A T T
Q BG
T B−
− −
Δ Δ=Δ Δ
)− , (12)
where )()( 2121PPPPB BQ −−
ΔΔ and )()21( 21 AA TTA BT −− ΔΔ can be obtained from the slopes of
Figs.3 c and d, respectively. The thermal conductivity of the fiber can then be found by
assuming an one-dimensional heat transfer along the fiber,
© 2010 Macmillan Publishers Limited. All rights reserved.
nature nanotechnology | www.nature.com/naturenanotechnology 7
SUPPLEMENTARY INFORMATIONdoi: 10.1038/nnano.2010.27
2
4d
LGK FF π= , (13)
where L and d are the length and the diameter of the nanofiber, respectively. It should be
emphasized that though it may appear that the fiber conductance is independent of the
cantilever conductance in Eq. (12), the effect of the cantilever conductance is actually
embedded within the measured bending signals.
From the above discussion, we see that the differential heat transfer through the
nanofiber is directly measured from the change of the bending in the cantilever, since the
initial bending of the cantilever is well recorded by the photodetector [Figs.3 a and b in
the manuscript]. Although the nanofiber thermal conductance is much smaller than that
of the cantilever, our technique allows direct measurement of very small heat transfer
through the nanofiber.
4. Temperature distribution along the cantilever due to laser heating
During the experiment, the area of the laser spot (~ 20 μm × 20 μm) on the
cantilever is observed, while our model assumes heating occurs at one point. In this
section, we will show that for the experimental conditions, the finite laser spot size has a
negligible effect. Using general beam theory, the deflection of a bi-material strip with
different thermal expansion coefficients can be solved from the following differential
equation ,
( )021
21122
2
)()(6 TxTKtttt
dxZd
−+
−= γγ, (14)
where Z(x) is the vertical deflection at a location x, γ is the thermal expansion coefficient,
K is a constant defined by the thickness ratio and the Young’s modulus of the layers, t is
© 2010 Macmillan Publishers Limited. All rights reserved.
8 nature nanotechnology | www.nature.com/naturenanotechnology
SUPPLEMENTARY INFORMATION doi: 10.1038/nnano.2010.27
the thickness of the layers (t1= 650 nm, t2= 70 nm), T(x) is the temperature distribution
along the cantilever, T0 is the reference temperature at zero deflection. The boundary
condition for the differential equation is Z(l) = 0, and dZ(l)/dx = 0 since the base of the
cantilever is fixed. Subscripts 1 and 2 refer to the two layers of the strip: “1” for Si3N4
and “2” for Au. Thus, the temperature distribution of (T(x)-T0) must be determined to
solve the deflection Z(x).
Based on equation (14), it can be observed that variations in heating input will be
reflected by the temperature distribution across the cantilever. To show the validity of our
assumption that all the incident heat by the laser is absorbed at a point, we will consider
two cases: (1) the assumed case where all the heat is absorbed at the end and (2) a finite
distribution of heat across some length, l1, from the tip. These cases are displayed in Figs.
S4 (a) and (b). Note that this model is only one-dimensional. Because the relative area
that the incident heat impinges upon the cantilever is a smaller proportion with respect to
the surface area of one side of the cantilever, the current model can be assumed to
provide a conservative estimate.
For case 1, the temperature distribution follows the previously published
distribution,
0 11 1 2 2
( ( ) ) ( )( )
PT x T l xw k t k t
− = −+
, (15)
where P is the total heat incident on the cantilever, l is the length of the cantilever, w is
the effective width of the cantilever, and ki is the thermal conductivity for layer i.
For case 2, the temperature distribution can be described by a piece wise
distribution where (T(l)-T0) = 0 and continuity is assumed at l1,
© 2010 Macmillan Publishers Limited. All rights reserved.
nature nanotechnology | www.nature.com/naturenanotechnology 9
SUPPLEMENTARY INFORMATIONdoi: 10.1038/nnano.2010.27
[ ]
21
11 1 2 2 1
0 2
11 1 2 2
- - ;0( ) 2 2
( ( ) - )- ;
( )
lP xl x l w k t k t l
T x TP l x l x l
w k t k t
⎧ ⎡ ⎤≤ ≤⎪ ⎢ ⎥+⎪ ⎣ ⎦= ⎨
⎪ ≤ ≤⎪ +⎩
, (16)
The temperature distribution for this case is shown in Fig. S4 (c), where it is assumed l =
200 µm and l1 = 20 µm. It can be observed that a distinct parabolic region exists near the
tip of the cantilever, as expected. But overall, it can be qualitatively observed that this
distribution is approximately linear.
Given these temperature distributions, integration of Equation (14) will yield a tip
displacement Z(0), which can be compared between both cases to test how close the real
case can be approximated by the assumed case. For case 1, integration using (15) will
yield,
31(0)Z Cl= , (17)
Similarly, for case 2, integration using Eq. (16) yields,
3 31
2 (0) 624 6l lZ C
⎛ ⎞= − +⎜
⎝ ⎠⎟ , (18)
where,
1 22 1 2
2 1 1 2
( )( )
t t PCt K w k t k t
γ γ += −
+ 2
Upon substitution of known values, the tip displacements become Z1(0) = 8E6·C
and Z2(0) = 7.998E6·C. As observed, the assumed case where all incident heat is incident
at the tip only differs by the more realistic case by 0.002E6 m3 or 0.025 % uncertainty.
Hence, based on this analysis, the assumption that all the heat is incident at a point is
validated for approximation.
© 2010 Macmillan Publishers Limited. All rights reserved.
10 nature nanotechnology | www.nature.com/naturenanotechnology
SUPPLEMENTARY INFORMATION doi: 10.1038/nnano.2010.27
5. Proportionality constants α1 and α2
In calibrating the cantilever bending, the laser spot inputs heat at a point to the
cantilever at some distance x1 (~20 µm) from the tip. The calibrated bending constant is
α1. In the second step of the experiment, heat is transferred at the tip-nanofiber junction
(~x=0) and the corresponding bending constant is α2. In both cases, the bending of the
cantilever is measured at x = x1, where the laser spot is located. We will show that α1 is
equal to α2.
For case 1, when the laser power is delivered to the cantilever, as shown in Fig.
S5 (a), the temperature profile at steady state is,
GP
lxTxT 1
0 1)( ⎟⎠⎞
⎜⎝⎛ −=− , for lxx ≤≤1 , (19)
where l is the effective length of the cantilever (l = 200 μm), G is the effective thermal
conductance of the cantilever and P1 is the absorbed power. The temperature is constant
for because of the adiabatic boundary condition at10 xx ≤≤ 0=x . By solving Eq.(14)
using Eq. (19), we have,
( ) 13
21
2112 )()( Pxl
KGttttxZ −
+−−= γγ , for lxx ≤≤1
Thus, the deflection measured at x = x1 (the location of the laser spot) is,
( ) 1113
121
21121 )()( PPxl
KGttttxZ αγγ =−
+−−= , (20)
For case 2, when the heat is conducted through the nanofiber, as shown in Fig. S5
(b), the temperature profile at steady state is,
© 2010 Macmillan Publishers Limited. All rights reserved.
nature nanotechnology | www.nature.com/naturenanotechnology 11
SUPPLEMENTARY INFORMATIONdoi: 10.1038/nnano.2010.27
GP
lxTxT 2
0 1)( ⎟⎠⎞
⎜⎝⎛ −=− , for lx ≤≤0 , (21)
By solving Eq. (14) using Eq.(21), we have,
( ) 23
21
2112 )()( Pxl
KGttttxZ −
+−−= γγ , for lx ≤≤0
Thus, the deflection measured at x = x1 is,
( ) 2223
121
21121 )()( PPxl
KGttttxZ αγγ =−
+−−= , (22)
As can be observed, the two proportionality factors, α1 and α2, are equal from Eqs.
(20) and (22),
( )31 21 2 2 1 1
1 2
( ) t t l xt t KG
α α γ γ += = − − −
6. Radiation heat transfer between the cantilever and the heated needle
Upon completion of the measurements, both varying the absorbed power of the
cantilever and the temperature of the thermocouple, the motorized control stage is used
(100 nm step resolution) to move the needle with the thermocouple backwards until
breaking the nanofiber. Then, the needle is moved forward to the original location where
both measurements were previously conducted and the bending of the cantilever is
measured again by varying the temperature of the thermocouple. The observed bending
signal without the nanofiber is caused only by the radiation transfer between the
cantilever and the heated needle (Fig. S6). The maximum influence of thermal radiation
is ~ 25 %. The deflection signals in Fig. S6 and Fig. 3 b are not in the same range due to
the thermal drift of the cantilever. In Fig. 3 d, we have normalized the deflection signals
© 2010 Macmillan Publishers Limited. All rights reserved.
12 nature nanotechnology | www.nature.com/naturenanotechnology
SUPPLEMENTARY INFORMATION doi: 10.1038/nnano.2010.27
in Fig. S6 to ambient temperature (T1) and subtracted them from the deflection signals in
Fig. 3 b. When varying the absorbed power by the cantilever in Fig. 3 a, no correction of
radiation is needed because the conductance of the cantilever is much larger.
7. Laser power absorbed by the cantilever
The output of the photodiode is converted into an X or Y signal corresponding to
the position of the laser spot and a sum signal proportional to the incident laser power. A
power meter (Newport, Model: 818-UV) is used to measure the radiant power in the
incident beam (1.809 mW, 650 nm wavelength), the reflected beam (1.222 mW) and the
strayed beam (0.427 mW). The strayed beam is defined as the beam passing through the
unblocked area of the cantilever and is measured behind the cantilever perpendicular to
the laser incidence direction. We also put the power meter very close to the cantilever at
different locations around the cantilever to measure the scattered light from the cantilever.
The scattered light intensity was measured to be negligibly small because the cantilever is
very thin and flat. Thus, the absorbed power by the cantilever is calculated to be 0.160
mW by (1.809 mW-1.222 mW-0.427 mW). Due to the strayed light (0.427 mW), the
light power incident on the cantilever is 1.382 mW by (1.809 mW-0.427 mW). Thus, the
absorptivity of the cantilever (gold film) is calculated to be 0.116 by the ratio of 0.160
mW and 1.382 mW, which is consistent with the literature value. The accuracy of each
power measurement is ~ 0.002 mW which is obtained by multiple measurements. From
uncertainty propagation, that would yield an overall error on the absorbed power of
~0.004 mW. The uncertainty of the absorbed power by the cantilever is thus ~ 2.5 %.
During the experiments, it is not the incident light that is measured but the reflected light.
© 2010 Macmillan Publishers Limited. All rights reserved.
nature nanotechnology | www.nature.com/naturenanotechnology 13
SUPPLEMENTARY INFORMATIONdoi: 10.1038/nnano.2010.27
The ratio of absorbed to reflected light is 0.131. These numbers can vary depending on
the shape of the cantilever and optical arrangement. In Fig. S7 (a), the photodiode sum
signal of the reflected light is plotted as a function of the reflected laser power. The linear
relationship between them corresponds to a slope of 0.6434 mW/V. The absorption of the
cantilever can be calculated from the photodiode sum signal as 0.131 × 0.6434 ×
(photodiode sum signal). In Fig. S7 (b), we also plot the deflection signal as a function of
sum signal.
8. Uncertainty analysis
The total uncertainty incurred during measurement is a composition of multiple
uncertainties propagating throughout the experiment. The general formulation for KF is,
⎟⎟⎠
⎞⎜⎜⎝
⎛ΔΔ
⎟⎟⎠
⎞⎜⎜⎝
⎛ΔΔ
⋅==
2
122
44
BTBQ
dL
dLG
KA
B
FF ππ
From uncertainty propagation rules, the total uncertainty in KF is,
( ) ( ) ( )
2
2
2
2
1
1222
2
⎟⎟⎟⎟⎟
⎠
⎞
⎜⎜⎜⎜⎜
⎝
⎛
⎟⎟⎠
⎞⎜⎜⎝
⎛ΔΔ
⎟⎟⎠
⎞⎜⎜⎝
⎛ΔΔ
+
⎟⎟⎟⎟⎟
⎠
⎞
⎜⎜⎜⎜⎜
⎝
⎛
⎟⎟⎠
⎞⎜⎜⎝
⎛ΔΔ
⎟⎟⎠
⎞⎜⎜⎝
⎛ΔΔ
+⎟⎠⎞
⎜⎝⎛+⎟
⎠⎞
⎜⎝⎛=⎟⎟
⎠
⎞⎜⎜⎝
⎛
BTBT
e
BQBQ
e
dde
LLe
KKe
A
A
B
B
F
F , (23)
where the uncertainties of the nanofiber length and diameter are ~ 3% and ~ 8 %,
respectively. Therefore, to completely determine the total uncertainty, the uncertainty of
each bending measurement must be derived.
In the first measurement where the incident laser power is varied, only the
deflection signal and sum signal of the photodector were directly measured. Hence, a
© 2010 Macmillan Publishers Limited. All rights reserved.
14 nature nanotechnology | www.nature.com/naturenanotechnology
SUPPLEMENTARY INFORMATION doi: 10.1038/nnano.2010.27
conversion is needed to change the sum signal term into absorbed power. This can be
accomplished as follows,
(reflectedabsorbedB
reflected
PPQ SumP Sum
⎛ ⎞ ⎛ ⎞Δ = ⋅ ⋅ Δ⎜ ⎟ ⎜ ⎟⎜ ⎟ Δ⎝ ⎠⎝ ⎠
) , (24)
where this expression is intentionally written as the product of three terms to utilize
previous measurements. The ratio of absorbed to reflected power can be calculated by the
power meter data, as previously discussed, and the ratio of the reflected power to the sum
signal can be obtained by previous data shown in Fig. S7 (a). Thus, the ratio of the
absorbed power to the deflection signal is,
2 2
reflectedabsorbedB
reflected
PPQ SumB P Sum B
⎛ ⎞ ⎛ ⎞⎛ ⎞Δ Δ= ⋅ ⋅⎜ ⎟ ⎜⎜ ⎟⎜ ⎟Δ Δ⎝ ⎠ ⎝ ⎠⎝ ⎠
⎟Δ, (25)
where the ratio of the sum signal to the bending signal reflects what was actually
measured. The estimated error for the absorbed power and the reflected power is thus
0.004 mW and 0.002 mW with averages of 0.160 mW and 1.222 mW, respectively,
( ) ( )0025.0
222.1002.02
160.0004.02 2222
=⎟⎠⎞
⎜⎝⎛ ⋅
+⎟⎠⎞
⎜⎝⎛ ⋅
=⎟⎟⎠
⎞⎜⎜⎝
⎛+⎟⎟
⎠
⎞⎜⎜⎝
⎛=
⎟⎟⎟⎟⎟⎟
⎠
⎞
⎜⎜⎜⎜⎜⎜
⎝
⎛⎟⎟⎠
⎞⎜⎜⎝
⎛
reflected
reflected
absorbed
absorbed
reflected
absorbed
reflected
absorbed
PPe
PPe
PPPP
e
In the measurement relating the reflected power to the sum signal (Fig. S7 (a)),
the standard deviation and range of the reflected power are 0.004 mW and 0.18 mW.
Likewise, for the sum signal, they are 0.4 mV and 0.28 V. Note that the reflected power
in this instance is different than what was previously used since this quantity is
contingent on the sum signal.
© 2010 Macmillan Publishers Limited. All rights reserved.
nature nanotechnology | www.nature.com/naturenanotechnology 15
SUPPLEMENTARY INFORMATIONdoi: 10.1038/nnano.2010.27
( ) ( )
2
2 2 2 22 0.004 2 0.0004 0.00190.18 0.28
reflected
reflected
reflected reflected
Pe e P e SumSum
P P SumSum
⎛ ⎞⎛ ⎞⎜ ⎟⎜ ⎟ ⎛ ⎞ Δ⎛ ⎞Δ ⋅ ⋅⎛ ⎞ ⎛ ⎞⎝ ⎠⎜ ⎟ ⎜ ⎟= + = + =⎜ ⎟ ⎜ ⎟ ⎜ ⎟⎜ ⎟ ⎜ ⎟ Δ ⎝ ⎠ ⎝ ⎠⎝ ⎠⎝ ⎠⎜ ⎟Δ⎝ ⎠
Similarly, in the measurement relating the deflection signal of the cantilever to the
sum signal (Fig. S7 (b)), the standard deviation and range of the deflection signal is 2.21
mV and 823.3 mV. For the sum signal, they are 0.4 mV and 0.125 V. Again, the sum
signal considered in this case is different from what was previously used.
( ) ( )
2
2 2 2 222
2
2
2 0.0004 2 2.21 6.98 50.125 823.3
Sumee Sum e BB
ESum Sum BB
⎛ ⎞⎛ ⎞Δ⎜ ⎟⎜ ⎟ Δ Δ⎛ ⎞ ⎛ ⎞Δ ⋅ ⋅⎛ ⎞ ⎛ ⎞⎝ ⎠⎜ ⎟ = + = + =⎜ ⎟ ⎜ ⎟ ⎜ ⎟ ⎜ ⎟⎜ ⎟Δ Δ Δ ⎝ ⎠ ⎝ ⎠⎝ ⎠ ⎝ ⎠⎜ ⎟⎜ ⎟Δ⎝ ⎠
−
The total uncertainty of equation (24) is thus,
22 22
2 2
22
absorbed reflectedB
reflected
B absorbed reflected
reflected
P PQ Sumee eePB BSumQ P P Sum
BB P Sum
⎛ ⎞⎛ ⎞⎛ ⎞ ⎛⎛ ⎞⎛ ⎞ ⎛ ⎞⎛ ⎞Δ Δ⎜ ⎟⎜ ⎟⎜ ⎟ ⎜⎜ ⎟⎜ ⎟ ⎜ ⎟⎜ ⎟⎜ ⎟Δ ΔΔ⎜ ⎟⎝ ⎠ ⎝ ⎠ ⎝ ⎠ ⎝ ⎠⎜ ⎟ ⎜⎜ ⎟= + +⎜ ⎟⎜ ⎟ ⎜⎜ ⎟Δ Δ⎜ ⎟⎜ ⎟ ⎜⎜ ⎟⎜ ⎟ ⎜ ΔΔ Δ⎜ ⎟ ⎝ ⎠⎝ ⎠ ⎝⎝ ⎠
⎞⎟⎟⎟⎟⎟⎠
2
2
2
0.0025 0.0019 6.98 5 0.00447
B
B
QeB
EQB
⎛ ⎞⎛ ⎞Δ⎜ ⎟⎜ ⎟Δ⎝ ⎠⎜ ⎟→ = + + − =⎜ ⎟Δ⎜ ⎟⎜ ⎟Δ⎝ ⎠
, (26)
In the second measurement where the temperature of the heated needle is varied,
the temperature is directly measured by an attached thermocouple with a resolution of 0.1
K, as shown in Fig. 3d in the manuscript. The standard deviation of the deflection signal
is ~ 0.6 mV. Because the deflection signal in Fig. 3b is found by subtracting the radiation
signal, caused by the heated needle, from the total signal, an additional error is incurred.
© 2010 Macmillan Publishers Limited. All rights reserved.
16 nature nanotechnology | www.nature.com/naturenanotechnology
SUPPLEMENTARY INFORMATION doi: 10.1038/nnano.2010.27
Thus the final standard deviation of the deflection signal in Fig. 3(b) is ~ 2 0.6⋅ mV. The
range of the temperature and deflection considered, is 40 K and 7 mV, respectively, from
Fig. 3d. Note that it was assumed the error was normally distributed, hence two times of
standard deviations were taken for a 95% confidence interval. This will be assumed for
all subsequent parameters. This results in the following uncertainty,
( ) ( )
2
22 2 222
2
2
2 0.1 2 2 0.6 0.05940 7
A
A
A A
Tee T e BB
T T BB
⎛ ⎞⎛ ⎞Δ⎜ ⎟⎜ ⎟ ⎛ ⎞Δ Δ⎛ ⎞ ⎛ ⎞Δ ⋅ ⋅ ⋅⎛ ⎞⎝ ⎠⎜ ⎟ = + = + =⎜ ⎟⎜ ⎟ ⎜ ⎟ ⎜ ⎟ ⎜ ⎟⎜ ⎟Δ Δ Δ ⎝ ⎠⎝ ⎠ ⎝ ⎠ ⎝ ⎠⎜ ⎟⎜ ⎟Δ⎝ ⎠
,(27)
Upon substitution of the uncertainties in (26) and (27) into (23),
( ) ( ) ( ) %8.27278.000447.0059.008.0203.0 22 ==++⋅+=⎟⎟⎠
⎞⎜⎜⎝
⎛
F
F
KKe
The values of thermal conductivities of three samples in the main text were averaged
based on two or three repeated measurements on one individual sample, and the largest
error among the repeated measurements on each sample was chosen as the error of the
sample.
9. Scanning electron microscope (SEM) images of nanofiber samples
Figures S8 (a), (b) & (c) show the measured diameters of samples 1, 2 and 3 by
SEM, respectively. The uniformity of the samples is evaluated by measuring the diameter
at ~ 10 different locations (with varying intervals 2 - 20 μm) along one sample. The
standard deviation of the diameter measurements is ~ 12 nm except at the end which is
connected with the cantilever.
© 2010 Macmillan Publishers Limited. All rights reserved.
nature nanotechnology | www.nature.com/naturenanotechnology 17
SUPPLEMENTARY INFORMATIONdoi: 10.1038/nnano.2010.27
10. Molecular Dynamics Simulations
Equilibrium molecular dynamics simulations were conducted, using the adaptive
intermolecular reactive empirical bond order (AIREBO) potential, which was
implemented in LAMMPS – a parallel molecular dynamics software package developed
at Sandia national labs. A time step of 0.25 fs was used and the simulations were run for
10 ns. Periodic boundary conditions were applied in all directions and the minimum
energy unit cell dimensions were 7.1 Å × 5.05 Å × 2.58 Å, which are within 4% of the
values obtained experimentally5. All simulations were initialized with each atom at its
equilibrium (minimum energy) lattice site and random velocities corresponding to room
temperature. In contrast to our previous work6, we have used the more widely employed
temperature definition, which is based on equipartition of the kinetic and potential
energy7. The size of the systems ranged from 6 × 5 × 40 unit cells (7200 atoms) to 8 × 6
× 40 unit cells (11,520 atoms); however, no size dependence was observed in the thermal
conductivity results. The simulations were run in parallel on processor grids ranging from
30-48 processors, using Sandia computing facilities. Each simulation required
approximately 20 days of computational time to complete. Other details associated with
computing the heat flux can be found in our previous works8, which are based on the
operator originally derived by Hardy9.
Figure 4a shows the full normalized heat flux autocorrelation function for a
simulation of a single polyethylene chain, which was calculated up to t = 5 ns. The inset
shows the first 35 ps in greater detail and provides a better view of the strong symmetric
oscillatory contributions close to ~ 50 THz. These oscillations correspond to
contributions from the C-C optical phonon bending motions ~ 50 THz, which do not
© 2010 Macmillan Publishers Limited. All rights reserved.
18 nature nanotechnology | www.nature.com/naturenanotechnology
SUPPLEMENTARY INFORMATION doi: 10.1038/nnano.2010.27
contribute significantly to the thermal conductivity, because of their symmetry about zero.
Figure 4b, however, shows the normalized heat flux autocorrelation function, when these
oscillations are suppressed by averaging over 0.025 ps intervals. This is done to highlight
the asymmetric features of the function which decay over a longer time scale and assist in
delineating the difference between the single chain and bulk crystalline behaviors.
References
1. Shi, L. & Majumdar, A. J. Heat Transfer 124, 329 (2002). 2. Marla, V. T., Shambaugh, R. L. & Papavassiliou, D. V. Ind. Eng. Chem. Res. 46, 336 (2007).3. Fujikura, Y., Suzuki, T. & Matsumoto, M. J. Appl. Polym. Sci. 27, 1293 (1982). 4. Shen, S., Narayanaswamy, A., Goh, S. & Chen, G. Appl. Phys. Lett. 92, 63509 (2008). 5. Sperling, L. H., Introduction to Physical Polymer Science (Wiley-Interscience, New Jersey, 2006)6. Henry, A. & Chen, G. Phys. Rev. Lett. 101, 235502 (2008).7. Turney, J. E. McGaughey, A. J. H. & Amon, C. H. Phys. Rev. B, 79, 224305 (2009).8. Henry, A. & Chen, G. J. Theoretical and Computational Nanosci. 5, 2, 141 (2008). 9. Hardy, R. J. Phys. Rev. 132, 168 (1963).
Figure Captions:
Figure S1 Schematic diagram of experimental setup to fabricate ultradrawn
nanofibers.
Figure S2 Two-step drawing process for fabricating the nanofibers.
© 2010 Macmillan Publishers Limited. All rights reserved.
nature nanotechnology | www.nature.com/naturenanotechnology 19
SUPPLEMENTARY INFORMATIONdoi: 10.1038/nnano.2010.27
Figure S3 Three-junction thermal circuit model for analyzing heat transfer in the
experiment.
Figure S4 (a) Case 1: Total heat at tip, (b) Case 2: Heat distribution at tip, (c)
temperature distribution in Case 2.
Figure S5 (a) Case 1: configuration and temperature distribution, (b) Case 2:
configuration and temperature distribution.
Figure S6 Experimental data of the radiation heat transfer between the heated
needle and the AFM cantilever.
Figure S7 (a) Measured reflected power versus the sum signal of photodiode, (b)
Deflection signal versus the sum signal of photodiode.
Figure S8 SEM images of measured nanofiber samples.
© 2010 Macmillan Publishers Limited. All rights reserved.
20 nature nanotechnology | www.nature.com/naturenanotechnology
supplementary information doi: 10.1038/nnano.2010.27
Heater 2
Heater 1
Gel Nanofiber
Tungsten tip
Microscope
Motor
FIG1
© 2010 Macmillan Publishers Limited. All rights reserved.
nature nanotechnology | www.nature.com/naturenanotechnology 21
supplementary informationdoi: 10.1038/nnano.2010.27
FIG2
© 2010 Macmillan Publishers Limited. All rights reserved.
22 nature nanotechnology | www.nature.com/naturenanotechnology
supplementary information doi: 10.1038/nnano.2010.27
TB
A CB
TA TC
Nanofiber AFMCantilever
QC
QB
QA
FIG3
© 2010 Macmillan Publishers Limited. All rights reserved.
nature nanotechnology | www.nature.com/naturenanotechnology 23
supplementary informationdoi: 10.1038/nnano.2010.27
FIG4
© 2010 Macmillan Publishers Limited. All rights reserved.
24 nature nanotechnology | www.nature.com/naturenanotechnology
supplementary information doi: 10.1038/nnano.2010.27
x=l
T -T0
x=l
T -T0
x =x1
T0
Vacuum
Power, P2
x =l x =0
Laser beam
T0
Vacuum
Power, P1
x =l x =0
x =x1
FIG5
© 2010 Macmillan Publishers Limited. All rights reserved.
nature nanotechnology | www.nature.com/naturenanotechnology 25
supplementary informationdoi: 10.1038/nnano.2010.27
0 500 1000 1500−4874
−4872
−4870
−4868
−4866
−4864
−4862
−4860
Time (arb. units)
Def
lect
ion
Sign
al (m
V)T1
T3T4
T1T2
FIG6
© 2010 Macmillan Publishers Limited. All rights reserved.
26 nature nanotechnology | www.nature.com/naturenanotechnology
supplementary information doi: 10.1038/nnano.2010.27
1.3 1.35 1.4−4600
−4500
−4400
−4300
−4200
−4100
−4000
−3900
−3800
−3700
Sum Signal (V)
Def
lect
ion
Sign
al (m
V)
Experimental dataLinear fitting
1.5 1.6 1.7 1.8
1
1.05
1.1
1.15
Sum Signal (V)
Pow
er (m
W)
Experimental datay=0.6434x+0.001
FIG7
© 2010 Macmillan Publishers Limited. All rights reserved.
nature nanotechnology | www.nature.com/naturenanotechnology 27
supplementary informationdoi: 10.1038/nnano.2010.27
c
a b
Sample 1 Sample 2
Sample 3
FIG8
© 2010 Macmillan Publishers Limited. All rights reserved.