Yao N. Focused Ion Beam Systems: Basics and Applications

Подождите немного. Документ загружается.

working distance with an original screen magniﬁcation of 215 000 · (67.14k ·

in 4 · 5 format) and a ﬁeld width of 1.7 m m. Using dual channel mode both

images were acquired simultaneously during a single scan where the image in

Figure 6.4(a) was captured on the Everhart–Thornley SE detector and

the image in Figure 6.4 (b) was recorded using the in-lens SE detector. As

both images were acquired simultaneously under identical primary beam

(a)

(b)

Figure 6.4 Semiconductor cross section image acquired at 890 eV at an

original screen magniﬁcation of 215 000 · and a horizontal ﬁeld width of

1.74 mm. Both images were acquired simultaneously under identical beam

conditions. Panel (a) was recorded with the Everhart–Thornley detector and

panel (b) was recorded with the in-lens SE detector. The improved

resolution in the lower in-lens image is due to emphasis on SE

detection

and the reduced contri bution of SE

and SE

image signals.

Focused ion beam systems158

conditions the images compare the character of the different signals captured

by the two detectors. The signal-to-noise ratio is somewhat lower in the ET

image relative to the in-lens image due to the short working distance. It is

clear that the in-lens detector reveals greater detail and improved resolution

with respect to the ET detector. This example visually illustrates the effect of

electron–solid interactions discussed above and the difference between a high

spatial resolution SE

in-lens detection signal and an ET detector convolved

with a contribution of SE

and SE

signals in the image. In the 3D recon-

struction application we will primarily employ the in-lens detectors to take

advantage of the superior spatial resolution.

Low voltage on-axis in-lens energy selective backscatter (EsB) detection

The principle of on-axis in-lens electron projection detection is extended to

partition and extract low energy BSE

from SEs. The following discussion is

with reference to Figure 6.5. It is possible to separate the BSE

signal from

SEs and BSE

electrons because the BSE

are distinct both in terms of energy

distribution and trajectory distribution. Secondary electrons pass through an

electrostatic focus and diverge en route to the on-axis in-lens SE detector,

as shown on the left-hand side of Figure 6.5. A small number of secondary

electrons pass through the annular opening in the SE detector and are

repelled by an adjustable ﬁeld generated by an electrostatic grid. Low loss

BSE

electrons which follow a nearly axial trajectory pass through the annular

EsB detector

BSE

In-lens SE

detector

Filtering

grid

SE detection

BSE detection

Figure 6.5 Schematic representation of the electron projection paths deﬁning

the in-lens SE detection and in-lens EsB detection. Secondary electro ns are

projected onto the in-lens SE detector while the BSE

follow an axial

projection onto the upper in-lens energy selective backscatter detector.

High-density FIB-SEM 3D nanotomography 159

portion of the in-lens detector, shown on the right-hand side of Figure 6.5.

Those BSEs with energy higher than the grid voltage navigate across the ﬁeld

and are directly detected by the EsB detector. The grid voltage is not only

used to reject secondary electrons but it is also an energy ﬁltering element

that can be used to create an energy window between the primary and grid

voltages. Using this type of BSE

detection system it is possible to collect a

low voltage high spatial resolution z-contrast image at short working distance

with a 25 eV energy image window. Isolating the BSE

signal at low voltage

coupled with energy ﬁltering to concentrate on the elastic peak is a beneﬁt in

3D reconstruction.

High scattering angle BSE

images will also display minimum topography in

addition to strong material contrast, as illustrated through the image pair in

Figure 6.6. The sample is an uncoated geological material. The images were

acquired at 750 V with a horizontal width of 7.7 m m. Both images in Figure 6.6

were acquired simultaneously during a single scan and captured using two

detectors. The side chamber mounted SE

detector was used to collect the

image shown in (a) and the energy selective backscatter (EsB) detector was

used to collect the image in (b). The SE

detector image shows strong topo-

graphic contrast while the EsB image displays minimum topography and

strong z-contrast, in a dramatic example of the different information content

extracted by the two detectors. Other contrast mechanisms such as Schottky

barrier contrast, voltage contrast and charging contrast are also suppressed

from an EsB image. Enhanced z-contrast with minimum topography is a

beneﬁt in FIB-SEM 3D reconstruction to separate different materials and

phases throughout the 3D volume during data reduction. Similarly, the elim-

ination of edge contrast, charging contrast, and other contrast mechanisms

deleterious to FIB-SEM nanotomography is also a signiﬁcant beneﬁt. The

ability to extract this information at low voltages of 1 kV and less means that

depth resolution is also optimized. Moreover, this type of in-lens BSE detector

requires no insertion, alignment, or adjustment. Thus, in many respects the in-

lens EsB is an ideal detector for the purposes of FIB-SEM 3D nanotomo-

graphy. It is expected that future work will continue to highlight the beneﬁts of

FIB-EsB nanotomography in both materials science and emerging biological

applications.

Electrostatic–electromagnetic objective lens

The Gemini column derives it name from the ‘‘twin’’ element conﬁguration of

the objective lens. With attention to Figure 6.3(b), the upper portion of the

objective lens is an electromagnetic element while the lower portion of the

objective lens is an electrostatic element. Collectively, these elements combine

Focused ion beam systems160

to produce the electron optical analogy of an achromatic lens triplet,

represented schematically in the lower right portion of Figure 6.3 (b). The ﬁrst

lens of the triplet is an electromagnet element and the ﬁnal two lenses of the

triplet are electrostatic. Traditional electron optics systems consisting only of

focusing elements lead to a lens system with positive aberration coefﬁcients.

(a)

(b)

Figure 6.6 Comparison of images acquired simultaneously on an EsB

CrossBeam highlighting the difference in information content from an SE

detector and an in-lens BSE detector. The images were acquired at 750 V

and collected simultaneously using two different detectors. The image

depicted in (a) was formed by the side chamber mounted SE

detector and

shows strong topographic co ntrast. The lower image (b) produced with the

in-lens EsB detector displays minimum topographic contrast and strong

z-contrast.

High-density FIB-SEM 3D nanotomography 161

However, the action of the decelerating electrostatic ﬁeld functions as a

negative aberration lens, represented by the central element of the triplet. The

ﬁnal electrostatic lens element is the pole piece focusing element. This unique

objective lens design leads to excellent low energy performance because the

chromatic and spherical aberration coefﬁcients actually decrease with pri-

mary electron voltage.

Real-time SEM imaging during FIB milling

The electrostatic component of the Gemini objective lens results in the

absence of any appreciable magnetic ﬁeld at the sample surface and this leads

directly to a unique ability to image in real-time at ultra-high-resolution

during the FIB milling process. Blurring and beam distortion result during

any attempt to simultaneously operate a gallium FIB column and an SEM

column in the presence of a magnetic ﬁeld. This interference is due to

interaction between the magnetic ﬁeld with the isotopes and multiply charged

ions within the gallium ion beam. The electrostatic elements on the lower

portion of the Gemini objective lens contain the vast majority of the elec-

tromagnetic ﬁeld generated by the upper portion of the objective lens,

creating an opportunity for operation of simultaneous focused ion beam

milling and scanning electron beam imaging.

Another notable consequence derived from the ability to simultaneously

image with the SEM during the FIB milling process is the formation of a

hybrid contrast mechanism that combines the SE image contrast produced by

the primary electron beam and the channeling contrast produced by the ion

beam on crystallographic samples. This operation mode is often desirable

due to strongly enhanced orientation contrast that complements the electron

beam induced SE contrast in the collected image. This hybrid contrast is only

present while both beams are operating simultaneously.

It should be also noted that the real-time imaging during the ion milling

process will not inﬂuence the BSE image contrast generated by the primary

electron beam because ion bombardment does not produce backscatter

electrons. Therefore collecting the in-lens EsB signal can also be advanta-

geous when it is desired to suppress orientation contrast. Dual channel

operation allows the operator access to display both these signals simulta-

neously on two separate computer screens during the milling operation.

6.4 Data acquisition parameters for FIB-SEM 3D reconstruction

The preceding established a rationale for using low voltage operation in

order to optimize depth resolution and a strategy to extract SE

and BSE

Focused ion beam systems162

information content to achieve high-resolution FIB-SEM nanotomography.

Electron detection principles that allow ultra-high-resolution imaging at low

voltage and simultaneous collection of both in-lens SE and in-lens BSE

signals have been described. Finally, high density data collection of these

signals is facilitated by the ability to image in real-time during the FIB

milling process by taking advantage of the electrostatic objective lens

design.

The following is a general guideline for data speciﬁc acquisition parameters

described for various types of FIB-SEM 3D reconstruction including

FIB-SE, FIB-BSE, and FIB-EBSD. Obviously there are nuances to the

techniques and methods described and the analyst is encouraged to adapt the

methods described here according to their own circumstances. In practice,

once the operation parameters are considered acquisition is executed with

relative ease by an operator oriented with both SEM and FIB instrumenta-

tion. The setup process can also be assisted through a computer-aided expert

system to facilitate the automated procedure.

6.4.1 Electron beam imaging current

An important parameter under the operator’s control to affect the signal-to-

noise ratio in the recorded image data is the electron probe current. The

Gemini column does not suffer the same tradeoff between probe current and

probe size as encountered with a traditional electron column. The details are

related to the lack of crossover in the beam path and the fact that ‘‘spot size’’

is not adjusted by changing the condenser lens setting as on a system with

crossover electron optics. So it is possible to maintain high resolution and

high current while enhancing depth of focus. The high current/high depth of

ﬁeld mode on the Gemini column approximately doubles the current while

also increasing the depth of ﬁeld due to a commensurate reduction in the

convergence angle. Therefore, the high current mode is generally a good

choice for a given aperture as an increased depth of ﬁeld is a beneﬁt that is in

addition to improved signal intensity. Probe currents can be in the range of

less than 10 pA to greater than 20 nA, as desired. Software also allows an

option to automatically track working distance during the milling process to

maintain focus over large cut depths.

The real-time SEM imaging conditions during ion milling used to produce

the 3D reconstruction data in most of this work yielded an incoming count of

approximately 20 000 electrons averaged per pixel per frame and represents a

moderate current. The calculations are derived from the dwell time per pixel,

the probe current, and the frame averaging parameters.

High-density FIB-SEM 3D nanotomography 163

6.4.2 FIB milling current

The FIB milling current will dictate the quality of the cut face and the

amount of time required to mill the selected volume. The current and dwell

time of the ion beam should be designed to ensure that each milling element

cuts completely through all materials in the cross section to the desired depth.

A ‘‘mill to depth’’ option allows a cut depth to be numerically input. The

milling calculation is based upon a materials ﬁle that calibrates milling rates

derived from measured or theoretical values input by the user.

The amorphous damage depth due to the milling process is usually not an

issue for SEM imaging. However, be aware of the amorphous damage layer

associated with the FIB milling process and ion beam sensitivity of the

material. A lower FIB voltage will reduce the amorphous depth while

a reduced milling current may reduce the damage rate, particularly on

organic or biological specimens. Amorphous damage is a potential issue

when collecting FIB-EBSD tomographic data. The analytical volume of

EBSD is estimated to be in the range of 30 nm and good EBSD have been

generated from 30 kV FIB using a 2 nA beam. However, the analytical

volume of the EBSD data and the tolerable range of amorphous damage will

be highly material dependent and even grain orientation dependent. In the

case of FIB-EBSD it is most practical to run tests to conﬁrm good quality

patterns for a given FIB milling current.

An acceptable current range for FIB-SEM data collection can be anywhere

from 10 pA–20 nA. The speciﬁc ion beam current will depend upon the

milling properties of the material, the physical volume of the data cube and

the time allotted for the acquisition. Assuming a nominal data volume

consisting of 10 mm in width (x-axis), a z-depth of 5 mm and a y-depth of less

than 10 mm, the acquisition can be completed in approximately one hour or

less on a silicon based material with a current of 20–50 pA (recall with

reference to Figure 6.1, that the z-axis represents the axis along which the

cutting plane (x–y plane) will advance).

6.4.3 Acquisition time

One hour is quite reasonable for the typical sample volumes of approximately

500 mm

, but there is no serious compromise associated with longer acqui-

sition time. Some FIB-EBSD experiments have operated continuously for up

to four days. Sample drift also is not an issue since the data set requires image

alignment even with zero image drift, as described in the data reduction

section. In sequential acquisitions, as in FIB-EBSD and FIB-XEDS, automatic

Focused ion beam systems164

image registration is employed between data acquisition cycles for both FIB

and SEM alignments.

6.4.4 Image collection frequency

The depth resolution along the cutting plane has been estimated as 5–20 nm

while the data acquisition time has been deﬁned as 3600 seconds. Together

with knowledge of the linear dimension along the cutting plane (x–y plane

and z-axis, with reference to Figure 6.1) it is possible to determine the

appropriate image acquisition rate. It is a good practice to acquire the data at

a slice thickness that corresponds to half the estimated resolution limit.

Assuming a 20 nm depth resolution, an image will be collected every 10 nm.

Assuming this 10 nm value and a linear z-axis dimension of 1.5 mm, the

number of required frames is 150. The total number of required frames to

achieve a 10 nm resolution in this example corresponds to an image recorded

every 24 seconds. Using the frame averaging noise reduction mode the system

can automatically average approximately 200 frames every 20 seconds. It is a

sound practice to collect an image at twice the frame averaging cycle rate.

Continuing with this example, collecting an image every 10 seconds yields

a 4 nm slice thickness and a total of approximately 350 frames. In this

instance the data acquisition parameters are such that we are over sampling

by nearly a factor of three to ﬁve relative to the estimated resolution limit. In

the data reduction process one may choose to work with all the data frames

or cull the data using any index parameter, i.e., every other frame. Thus, in

general, an image recorded every 10–20 seconds will produce a high quality

FIB-SEM nanotomographic reconstruction at a 10–20 nm resolution over

volumes of 250–500 mm

If very high milling rates are desired for milling larger volumes or smaller

volumes in a shorter time period then faster AVI capture rates may be used.

The system software on the CrossBeam has a built-in AVI capture capability

that allows the images to be recorded as frequently as every 100 ms.

6.4.5 Pixel size

The image pixel size is a function of both the SEM magniﬁcation and the

image store resolution. At a screen magniﬁcation of 10 000 · it is reasonable

to employ a 1 K · 1 K store resolution but it is possible to save the images at

a3K· 2 K store resolution. It is interesting to note that the maximum

achievable resolution can be recorded without any a priori knowledge of the

structures contained in the data volume. Following data acquisition it is of

High-density FIB-SEM 3D nanotomography 165

course possible to post-process any sub-volume of interest and digitally zoom

in on any feature in the volume, but that operation won’t ultimately increase

the information content contained in the data.

6.4.6 Storing images

The storage volume required will depend upon the number of AVI movies

recorded over time, the store resolution, and the length of those movies.

A typical movie will require 200–300 Mb of disk storage and a comparable

amount disk space for post-data reduction ﬁles. As a practical matter it is a

good idea to have a separate storage volume to accommodate the movie data

if one intends to complete a large number of reconstructions, process the

data, and create inspiring animations. Fortunately, data storage is cheap.

6.5 Quantiﬁcation of FIB based 3D tomographic data

The continuous nature of the milling process, in combination with a method

devised to determine the slice thickness between image frames, facilitates

direct quantiﬁcation [15]. The continuous nature of the process also means

that the data slices are more uniform in nature and less subject to thickness

variation, charge ﬂuctuation, and re-alignment errors associated with stop-

ping and restarting the tomographic slice process.

The procedure involves patterning one or more features into or on the

surface of the region of interest that can also be viewed in cross section during

the FIB tomographic process. This feature may be etched into the surface by

either the FIB itself or by another means. The feature could also be deposited

onto the surface through the FIB or other method. In other words, the feature

(s) can be either recessed (i.e., etched into) or raised (i.e., deposited) with

respect to the sample plane. If deposited, the pattern can consist of any suitable

material such as an oxide, metal, organic, inorganic, or any combination

thereof. The feature must simply have a known geometric relationship between

the top view and the cross section. The most direct example is a pattern

consisting of at least two straight lines forming a known angle etched into the

surface perpendicular to the cross section, as depicted in Figure 6.7.

Since the angle deﬁned by the lines is known, a mathematical relationship

can be established that is generally described by:

¼ L

 A

nþ1

;

 A

nþ1

tan 

;

ð6:2Þ

Ontheright-handsideofFigure6.7 is an image showing actual ﬁducial lines

etched into a platinum protective layer on top of a multilayer semiconductor

Focused ion beam systems166

device. As the cut face progresses the angled line(s) move toward the center

line. If the angle is 45



then the change in distance between the center line and

the angle line is equal to the slice thickness. This ﬁducial system allows

quantiﬁcation through the mathematical relation above based upon either

the average image slice thickness over the entire acquisition or the thickness

of individual image slices. The ability to image the quantitative mark on the

cross section in a single view improves speed and accuracy of the data.

The data can be processed after the image data acquisition and do not require

any prior knowledge of sample sputter rates (or other properties of the

samples) nor a calibration or measurement during the data acquisition pro-

cess. The ﬁducial marks are not seen in the processed ﬁnal data set as they are

usually placed to one side of the image just outside the region.

6.6 Data reduction

The typical steps involved in data reduction are summarized as follows:

(i) image alignment; (ii) selection of the sub-volume representing the region

of interest from the original data volume; (iii) image processing (optional);

(iv) reconstruction of the volume; (v) any additional volumetric analyses

(i.e., porosity, phase density, etc.); (vi) volume visualization, which could

include application of color and transparency mapping; and (vii) animation

of the results as desired.

As mentioned above, image alignment is required even if there is no sample

drift because the image plane is translating in the þy-direction in the SEM

view as the cutting plane progresses. This geometric effect occurs because the

Figure 6.7 The geometrical relationship above is applied to quantify the

average slice thickness in each frame of the FIB-SEM reconstruction

volume. On the left is a schematic related to Eqn. ( 6.2). On the right is an

example of the ﬁducial lines etched using the FIB into a platinum layer

created using the deposition capabilities of the FIB. The dashed box around

the central ﬁducial line is used to perform automated image stack alignment.

As the cut face progresses the angled lines move toward the central line

allowing quantiﬁcation of the slice thickness based upon the known

geometrical relationship.

High-density FIB-SEM 3D nanotomography 167