Yao N. Focused Ion Beam Systems: Basics and Applications
Подождите немного. Документ загружается.
work without it causing damage to the sample surface. When the nozzle
location is too close to the beam location the incident ion beam can be
perturbed by the nozzle due to extra field effects near the nozzle. Typically,
most focused ion beam units use a mechanical system to insert and retract the
gas nozzle. The location of the inserted nozzle must be consistent so that the
gas flux does not change over time. The ideal nozzle height is generally a few
hundred micrometers above the sample surface, although the optimum
location varies with different systems.
The temperature of the gas crucible also affects the gas flux in the nozzle. A
higher temperature causes a higher gas flux. However, if the gas crucible
temperature is too high the gas chemistry will be changed.
3.3.2 Definition of beam overlap
Now we need to understand how the focused ion beam scans the sample
surface. The operator sets the ion beam current acceleration voltage, focus,
and stigmatism to define the beam diameter (D). The ion beam spot with
diameter D scans within a pattern that is drawn by the operator. The pattern
contains the digital pixels that indicate the location of the beam spot. At each
pixel the ion beam stays for a certain period of time, called the dwell time, and
decomposes the gas precursor on the surface to produce a deposition. Then the
ion beam is blanked and moves with step size (S) to the next pixel. Depending
on the step size of the beam we can also define the beam overlap (OL). The
overlap of the scanning beam can be expressed as the following equation:
OL ¼
D S
D
· ð3:8Þ
The beam overlap is one of the critical parameters in order to achieve
successful deposition in FIB. If the step size is half the beam diameter, the
first and second beam spots overlap by 50%. A negative overlap happens
when the raster beam moves from one location to another with a larger step
h
z
Figure 3.5 Schematic diagram of a gas needle.
Focused ion beam systems78
size than its spot size. Figure 3.6 shows an example of beam overlap sche-
matically. When we set up the beam overlap in the system we need to take
into account the Gaussian beam profile. Even though the beam is controlled
digitally, the Gaussian beam profile of the beam spot contributes an extra ion
beam current in the space between beam spots. In general, positive overlap is
used for milling, zero overlap is used for etching, and negative overlap is used
for deposition.
3.3.3 Beam control for deposition pattern
The beam scans over a patterned area through all individual pixels in one
loop time. Once the ion beam has reached the last pixel of the pattern it
moves back to the first pixel of the pattern. This ion beam scanning cycle
repeats a number of times until it reaches the input setup time or achieves the
correct deposition thickness. Figure 3.7 shows a digital pattern of an ion
beam scan with an area of L · W.
50% overlap
0% overlap
–50% overlap
Figure 3.6 Examples of beam overlap.
SDOL
W
L
Figure 3.7 Schematic diagram of a digital pattern showing the step size (S),
overlap (OL) and diameter (D ) of the beam spot.
Gas assisted ion beam etching and deposition 79
The deposition rate in focused ion beam systems also depends on the
amount of ion beam current as well as the scan area and loop time. The
relationship between loop time, overlap, and dwell time can be understood
from Figure 3.7 . The total number of steps (N
s
) in the area L · W can be
expressed as follows:
N
s
¼
LW
S
2
· ð 3:9 Þ
In practical situations the beam scans fast enough over the pattern so that the
beam blanking time between pixels can be ignored. The loop time (t
l
)is
directly related to the pixel dwell time (t
d
) and the total number of step in the
pattern. The loop time can be shown as
t
l
¼ t
d
N
p
¼
t
d
LW
ð Dð 1 OLÞÞ
2
· ð3 :10Þ
Equation ( 3.10) shows that the loop time is determined by two operational
parameters, the beam dwell time and the overlap, once the area of pattern is
defined.
3.3.4 The effects of dwell time and overlap
Now we will take account of the effects of beam dwell time and overlap on
deposition rate and quality. Figure 3.8 shows a series of SEM images of
tungsten deposition on Si substrate with various overlaps and dwell times.
The size of the drawn pattern was 6 m m · 1.5 m m and the target thickness was
1 mm. For all depositions, a beam current of 1000 pA was used. Figures 3.8(a)
to (d) show tungsten deposition for a beam overlap of 50%, 0%, 150%,
and 500%, respectively. For the depositions with positive and zero overlap,
the thickness of the deposition appears uneven and does not reach to the
target thickness. For positive beam overlap the high beam current density
between beam spots can provide a strong sputtering yield during the
deposition process. Figure 3.8 (c) shows a good deposition with 150%
overlap with target and even thickness. Figure 3.8 (d) shows that the target
deposition thickness can obtained with a deposition of 500% overlap.
However, the deposition surface is not smooth. The space between the ion
beam spot diameters is too large so that neither gas decomposition nor extra
sputtering occurs in the space. Figures 3.8 (e) and (f) show the deposition with
an overlap of 150% and dwell times of 2 ms and 20 ms, respectively. All gas
precursors are consumed in the first few tenths of a microsecond of exposure
to the beam. In addition, extra long dwell time provides a sputtering effect on
Focused ion beam systems80
the deposited surface. If the dwell time is too long the deposited materials are
sputtered away leaving a milling hole on the surface of the sample, as shown
in Figure 3.8(f). The examples show that the proper setting of dwell time and
overlap are essential to achieve a successful deposition in FIB systems. The
optimum overlap and dwell time for good depositions can be achieved by
series of experiments. A sophisticated system can store the optimum overlaps
and dwell times for different deposition conditions so that the operator can
easily load the optimum parameters when required.
(a)
(c)
(d)
(b)
(e)
(f)
Figure 3.8 Images of a series of different overlaps and dwell times, with a
viewing angle of 52
. (a) 0.2 ms dwell time, 50% OL, (b) 0.2 ms dwell time,
0% OL, (c) 0.2 ms dwell time, 150% OL, (d) 0.2 ms dwell time, 500% OL,
(e) 2 ms dwell time, 150% OL, (f) 20 ms dwell time, 150% OL. (Courtesy
of FEI Com pany.)
Gas assisted ion beam etching and deposition 81
3.3.5 Ion beam current density on deposition
The ion beam current density must be taken into account for the success of
deposition of material in the focused ion beam system. Figure 3.9 shows three
distinct ranges in the graph of the deposition rate (mm/min) and the beam
current density (A/mm
2
). At the low beam current density both the decom-
position rate and sputtering rate are low, and result in a low deposition rate.
At intermediate beam current density values, both the gas decomposition and
sputtering rates increase. However, the decomposition rate reaches its max-
imum and is dominant so that the maximum deposition rate occurs. With a
high ion beam current density the decomposition rate does not increase but
the sputtering rate keeps increasing.
The range of intermediate beam current density is most efficient for the
deposition rate, with minimum sputtering. As an example, it has been
accepted that the most efficient deposition rates are within the current density
range of 2–6 pA/mm
2
for C
9
H
16
Pt (Pt deposition) and 100–150 pA/mm
2
for
W(CO)
6
(W deposition) [12].
In practical situations it can be easier to use the deposition time rather than
the beam current density. Once we know the time of maximum beam current
density we can avoid unwanted sputtering as well as obtain the maximum
deposition rate during deposition process. In this calculation the time of the
intermediate current density range (2–6 pA/mm
2
) for Pt deposition is intro-
duced. This calculation can be based on a Pt volume deposition yield of
0.5 mm
3
/nC at 50% overlap and 0.2 ms dwell time. For a simple calculation
CBA
Lower beam current;
larger patterns
2-6 pA/m
2
medium patterns
Higher beam current;
smaller patterns
Current Densit
y
(
p
A/
m
2
)
Rate
m/min
(thickness/time)
High efficiency
deposition per ion.
Ion dose on each
pass does not
decompose all gas.
Slow layer growth
rate & long
deposition time.
Most efficient
growth rate.
Each beam
scan uses up
nearly all gas.
All gas is used up
by only part of ion
dose at each point.
Remaining ions
then sputter
surface.
Deposition stops
and milling begins
Figure 3.9 Deposition efficiency of Pt [12].
Focused ion beam systems82
let’s assume that we would like to deposit 100 mm
3
( ¼10 mm · 10 mm and
1 mm thick) of platinum. In this case, we need 0.6 nA ( ¼6 pA/ mm
2
·
100 mm
2
) beam current to have the beam current density of 6 pA/ mm
2
over 100 mm
2
. Based on the definition of deposition time the calculation
will be:
Deposition time ¼
volume ðmm
3
Þ
beam current ðnAÞ · volume deposition yield ðmm
3
=nCÞ
;
ð3:11Þ
and the calculated deposition time for 6 pA/mm
2
is
mill time ¼
100 cubicmicrometers
0:6 nanocoulombs per second · 0:5
¼ 333 s ¼ 5 min 30 s:
Similarly, the deposition time for the boundary of 2 pA/mm
2
is about
16 min 37 s. So, an approximate corresponding time for a current density
between 2 pA/mm
2
and 6 pA/mm
2
will be between 16 min 37 s and 5 min
30 s, respectively. So to minimize the unwanted sputtering during Pt
deposition the deposition time must be set longer than 5 min 30 s for a
thickness of 1 mm.
3.4 Filling metal in high aspect ratio vias
As the semiconductor industry moves toward smaller dimensions and greater
numbers of charge carriers, the need for high aspect ratio vias increases.
Aspect ratio can be defined as the ratio of the depth of a via to the width of
the via at some given point. In today’s integrated circuits, it may be necessary
to make contact to a metal line that is more than 10 mm deep, while at the
same time avoiding making electrical contact to other metal traces that are in
close proximity. This could mean that one would have to create a platinum
filled via that is 10 mm deep and only 0.50 mm wide (an aspect ratio of 20:1).
Certain challenges exist when attempting to make high aspect ratio vias:
reaching the required depth;
keeping critical dimensions small to avoid shortin g with other traces;
getting uniform, solid, metal deposition in the via;
achieving low resistivity in FIB deposited metal within a high aspect ratio via;
knowing when the desired track is contacted (end-point detection).
As you attempt to mill a deep via, it becomes difficult to reach the desired
depth. It can be thought of as digging a trench while throwing dirt over your
shoulder. As you get deeper and deeper, it gets more difficult to continue
Gas assisted ion beam etching and deposition 83
making progress. Another major concern, of course, is avoiding hitting other
metal traces. This becomes challenging as the number of metallization layers
increases and device planarization increases. As more time is spent milling
a via, more edge rounding and widening of the via will occur. It can be
challenging to avoid contact with other metal lines. Milling the via at low
beam current decreases the effects of edge rounding and increases the aspect
ratio of the via. Low beam current milling, however, decreases the quality of
the end-point signal.
If the via cannot be filled solidly with metal, then it will have such a high
resistivity that it will be of no use. To increase the chances of a solid platinum
fill, the deposition is slowed by lowering the stage relative to the needle.
Another option would be to decrease the gas flow from the injector. An
organic carrier typically transports the metal that is deposited. When
depositing platinum on the surface of a sample, there is a much greater solid
angle for escape of the organic carrier than within a high aspect ratio via. As
more of the organic carrier gets trapped in the platinum, the resistivity
increases (due to the hydrocarbon-based carrier compound). By slowing
down the deposition rate, we also increase the amount of time for the carbon-
based material to escape from the hole, thus decreasing the resistivity.
Selecting the area for the via to be milled can be done by using a CAD
layout, optical images, or by other means of referencing a feature visible in
the FIB. Once the area is selected, the procedure outlined below can be used
as a guideline. It is often difficult to obtain a sharp and definitive end-point
trace when milling a high aspect ratio via. End-point detection done by
measuring the stage current, which is essentially a measure of the amount of
secondary electrons that are coming off the sample. When milling a very
narrow and deep via, it is often difficult to get many secondary electrons out
of this hole, thus making end-point detection more challenging. However, the
end-point may not be as pronounced as an end-point for a via of lower aspect
ratio. If possible, it is often helpful to characterize the amount of time it takes
to reach a given metal line, and use this time as a gauge in conjunction with
the end-point detection trace. Gas chemistry is essential for the via milling
process. Through chemical reaction of the gas with the milled material, the
material is removed and not redeposited. If gas is not used then only rela-
tively low aspect ratio holes are possible due to the sputtered material not
being able the escape the milled hole once the depth is approximately four
times the width of the hole. Once this condition is reached further milling
does not increase the depth and just results in sputtered material being
pushed around inside the milling area by the ion beam. It has been found
that, on most samples, the best results are obtained using insulator enhanced
Focused ion beam systems84
etch gas such as XeF
2
. It has also been found that the use of a stovepipe
needle is a necessity when milling a high aspect ratio via. The beam must be
paused periodically on the deposition process. During the beam pausing time
the byproduct gases in the via come out and fresh deposition gas precursors
go in. Without beam pausing the deposition of the large void will be left
inside the void. The desired beam on/off time is about 10/90. Figure 3.10
shows the results of milling a high aspect ratio via in an integrated circuit and
filling platinum in the via.
References
[1] H. F. Winters and J. W. Coburn. Surface Sci. Rep., 14 (1992), 161–270.
[2] S. Joyce, J. G. Langan and J. I. Steinfeld. Surf. Sci., 195 (1988), 270–82.
[3] M. A. Loudiana and J. T. Dickinson. J. Vac. Sci. Technol. B, 3 (1985), 1393–6.
[4] Y. Y. Tu, T. J. Chuang an d H. F. Winters. Phys. Rev. B, 23 (1981), 823–35.
[5] H. F. Winters. J. Vac. Sci. Technol. B, 1 (1983), 927–31.
[6] J. J. Hren. Introduction to Analytical Electron Microscopy, ed. J. J. Hren,
J. I. Goldstein and D. C. Joy (New York: Plenum, 1979), p. 481.
[7] T. J. Stark, G. M. Shedd, J. Vitarelli, D. P. Griffis and P. E. Russell. J. Vac. Sci.
Technol. B, 13 (1995), 2565–9.
[8] U. Gerlach-Meyer, J. W. Coburn and E. Kay. Surf. Sci., 103 (1981), 177–88.
[9] J. Orloff, M. Utlaut and L. Swanson. High Resolution Focused Ion Beams
(New York: Plenum, 2003).
Figure 3.10 High aspect ratio hole milled in SiO
2
to an aluminum contact.
Image has been stretched to reflect edge-on viewing. (Courtesy of FEI
Company.)
Gas assisted ion beam etching and deposition 85
[10] A. D. Dubner an d A. Wagner. J. Appl. Phys., 66 (1989), 870–4.
[11] K. Gamo, N. Takakura, N. Samoto, R. Shimizu and S. Namba. Jpn. J. Appl.
Phys., 23 (1984), L293–5.
[12] Platinum Deposition Technical Note (Hillsboro, OR: FEI Company, 2003), PN
4035 272 21851-B.
[13] R. J. Young, J. R. A. Cleaver and H. Ahmed. J. Vac. Sci. Technol. B, 11 (1993),
234–41.
[14] J. D. Casey Jr, A. F. Doyle, R. G. Lee and D. K. Stewart. Microelectron. Eng.,
24 (1994), 43–50.
[15] M. L. Thay er. Tut orial. IEEE Int. Reliability Phys. Symp. (1996).
[16] A. Yamaguchi and T. Nishikawa. J. Vac. Sci. Technol. B, 13 (1995), 962–6.
[17] R. J. Young. Vacuum, 44 (1993), 353–6.
[18] M. E. Gross, L. R. Harriott and R. L. Opila, Jr. J. Appl. Phys., 68 (1990),
4820–4.
[19] R. L. Kubena, F. P. Stratton and T. M. Mayer. J. Vac. Sci. Technol. B, 6 (1988),
1865–8.
[20] K. Gamo, N. Takehara, Y. Hamaura, M. Tomita and S. Namba. Microelectron.
Eng., 5 (1986), 163–70.
[21] D. K. Stewart, L. A. Stern and J. C. Morgan. Electron Beam X-ray and Ion Beam
Technologies: Submicrometer Lithographies VIII, Proc. SPIE, 1089 (1989), 18–25.
[22] Z. Xu, T. Kosugi, K. Gamo and S. Namba. J. Vac. Sci. Technol. B, 7 (1989),
1959–62.
[23] G. M. Shedd, A. D. Dubner, C. V. Thom pson and J. Melngailis. J., Appl. Phys.
Lett., 49 (1989), 1584–6.
[24] P. G. Blauner, J. S. Ro, Y. Butt and J. Melngailis. J. Vac. Sci. Technol. B, 7
(1989), 609–17.
[25] T. Tao, J. S. Ro, J. Melngailis, Z. Xue and H. Kaesz. J. Vac. Sci. Technol. B, 8
(1990), 1826–9.
[26] J. Puretz and L. W. Swanson. J. Va c. Sci. Technol. B, 10 (1992), 2695–8.
[27] A. D. Della Ratta, J. Melngailis and C. V. Thompson. J. Vac. Sci. Technol. B, 11
(1993), 2195–9.
[28] L. R. Harriott, K. D. Cummings, M. E. Gross and W. L. Brown. Appl. Phys.
Lett., 49 (1986), 1661–2.
[29] R. J. Young and J. Puretz. J. Vac. Sci. Technol. B, 13 (1995), 2576–9.
[30] H. Komano, Y. Ogawa and T. Takigawa. Jpn. J. App. Phys., 28 (1989), 2372–5.
[31] D. K. Stewart, A. F. Doyle and J. D. Casey, Jr. Electron-Beam, X-Ray, EUV, and
Ion-Beam Submicrometer Lithographies for Manufacturing V, Proc. SPIE, 2437
(1995), 276–307.
[32] P. J. Heard and P. D. Prewett. Microelectron. Eng., 11 (1990), 421–5.
[33] Carbon Deposition Technical Note (Hillsboro, OR: FEI Company, 2003), PN
4035 272 27241-A.
[34] T. Shiokawa, Y. Aoyagi, P. H. Kim, K. Toyoda and S. Namba. Jpn. J. Appl.
Phys., 23 (1984), L232–3.
Focused ion beam systems86
4
Imaging using electrons and ion beams
kaoru ohya
The University of Tokushima
tohru ishitani
Hitachi High-technologies Corporation
4.1 Overview of imaging using FIB
Sc an ni ng ele ctr on micr os cop es (S EM) , tr an smiss ion e lec tro n m ic ro sco pe s
(TEM) , and scanning TEM (STEM) have been used for s truct ural
observation of microdevice s, adva nce d mate ria ls, and bi ologi cal spec i-
mens. In r ece nt y ea rs, a g all i um (Ga) f ocuse d ion be am (FI B) h as bee n
us ed f or p repar ing the i r c ros s-s ect ional s ampl es [1 , 2 , 3 ,andChapter9 in
this book]. Here, F IB works both as a mi lling beam and as a pr obe f or a
scanning ion microscope (SIM). SIM images are u sed d ur ing t he whole
milli ng processes, t hat is , dr awing th e m il ling ar ea , m illi ng mo ni tor i ng ,
confirmat ion o f the fina l mi lling, and obse rva tion of the FI B mille d sec -
tio ns of i n ter es t. As SIM i ma ge re so lut i on ha s be en i mp ro ved an d ab ou t
5 nm is a chi evabl e at present, S IM obse rv ation h as bee n incr eas ing ly use d
in pl ace of SEM obse rvatio n when there is no es pecia l need for hi gh image
resolution.
However, we have found that the properties of SIM images are somewhat
different to SEM images. For example, SIM and SEM images of identical
FIB milled cross sections are shown in Figures 4.1(a)–(d): (a) and (c) show
a solder (Pb-Sn) on copper (Cu) and (b) and (d) show a Si device (static
random access memory) [4]. Black-and-white contrast among materials is
opposite between the SIM and SEM images. In addition, grain contrast (or
channeling contrast) can be observed more clearly in the SIM image and its
contrast is sometimes stronger than the material contrast. Figure 4.2 shows
Focused Ion Beam Systems: Basics and Applications, ed. N. Yao.
Published by Cambridge University Press. ª Cambridge University Press 2007.
87