Yao N. Focused Ion Beam Systems: Basics and Applications

Подождите немного. Документ загружается.

lamellae to be prepared along its length relatively quickly (Figure 8.2(b)). The

most widely used method to prepare a slice is hand polishing and as this can

be readily combined with the other methods and only requires polishing

pastes or sheets, it is outlined in detail below.

The ﬁrst step in preparing a slice by hand polishing consists of identifying the

ROI under an optical microscope. If it is too small to be observed, FIB milling

Ion

beam

Tilting sample

+tilt

–tilt

Line

milling

(e)

Tapered

sample

(b)

TEM lamellae

Dice

(a)

(c)

(d)

TEM grid

Figure 8.2 (a) T-shaped slice prepared by dicing four cuts, (b) an SE image of

a tripod polished slice into which numerous TEM lamellae have been milled,

ﬁnished H-bar sample ready for TEM, and (e) schematic of the tilting of a

sample into the ion beam to produce a lamella with parallel sidewalls.

Focused ion beam systems218

or Pt deposition can be used to render it visible. An area of approximately

2.5 mm

surrounding the ROI is then cleaved, cut, or polished. This is then

mounted, using wax, onto a microscope slide such that it protrudes over the

slice’s edge and is polished using diamond paste or Al

sheets with a series of

polishing grades (30 m mdownto0.5m m). A good support for the Al

sheets

is a wet glass pane as this keeps them ﬂat and prevents them slipping. During

polishing the sample is repeatedly examined under an optical microscope and

the polishing is stopped when the ROI is to within 30–50 mm of the edge of the

slide. The slide is then heated on a hot plate to melt the wax and the sample is

removed and remounted with the other side protruding over the edge of the

glass slide. It is then polished again until the slice is 50–100 m m wide. Typically,

this hand polishing takes about two to three hours but the exact time depends

on the hardness of the material and the initial dimensions of the square. The

slice is then removed from the slide (again by melting the wax), washed in

acetone to remove any residual wax and ﬁxed using conductive paint to a

TEM slot grid which has been cut in half (Figure 8. 2(c)). The sample is then

mounted in a suitable clamp and placed into the FIB or FIB/SEM system

chamber with the top of the slice perpendicular to the FIB.

In the FIB or FIB/SEM system, the ROI is identiﬁed and a 1.5 mm wide

and 1–3 m m thick Pt strap is deposited over it using ion beam assisted metal

deposition [ 36,37]. This planarizes the surface and prevents rounding and

milling of the top of the sample during FIB milling. If the sample is not

planarized, height steps can lead to striations in the face of a cross section

(the ‘‘curtain effect’’) due to changes in the sputter yield [38,39]. (The

required thickness of the Pt deposited is dependent on the ion beam proﬁle

and the topography of the surface.) If the deposited Pt strap obscures the

ROI, lines for positioning the TEM lamella during the FIB milling can be

milled outside the area where the Pt will be deposited.

After depositing the Pt strap, two trenches are milled at a distance of

1.5 mm from each side of the Pt out to the edges of the slice, using a large

beam current (2 nA to 20 nA) and by scanning the ion beam in a serpentine

pattern. The exact dimensions of the trenches (and the Pt strap) are dictated

by time constraints, the size of the feature or ROI, and the angles of tilt or

rotation required in the TEM (see Section 8.2.4).

Having milled the trenches the remaining material between them is then

thinned to create an electron transparent lamella. For this step the magnitude

of the ion beam current is reduced to 1000 pA or 500 pA (to decrease the spot

size to give better control of the subsequent thinning/milling steps). In

addition, the milling regime is changed. The ion beam is scanned in lines and

is moved inwards in steps of half a beam diameter; a process frequently

Preparation for physico-chemical analysis 219

referred to as FIB polishing. This decreases the amount of material

redeposited onto the sides of the lamella and reduces the probability of

accidentally milling the lamella due to ion beam or stage drifts. As the sputter

rate is higher when milling at the edge of a step than when milling into a ﬂat

surface the ‘‘depth’’ to which the lines are milled should be at least 30% less

than the ‘‘depth’’ that the trenches were milled, if the same milling para-

meters are used (i.e., beam overlap, sputter yields, and dwell times). Dis-

cussions on the effect of different milling regimes on the sputter rates can be

found in the papers by Ghandhi and Orloff [40] and Santamore et al.[41].

Furthermore, as a result of milling by the tails of the ion beam distribution

and the increase in the sputter yields as the angle of incidence of the ion beam

to a surface increases, the sidewalls of the remaining material are tapered

(Figure 8.2(e)). (For a mathematical description of the evolution of this taper

see Boxleitner [42].) If a lamella with parallel sidewalls is required, for

example for quantitative chemical analysis (and throughout the rest of the

chapter it is assumed that this is always the case), this taper can be removed

by tilting the sample into the ion beam while it is being thinned using the line

milling procedure (Figure 8.2(e)). The tilt angles used have ranged from 1



, depending on the material, the magnitude of the beam current, and the

depth of the lamella being milled.

Once the remaining material is approximately 500 nm the magnitude of the

ion beam is further reduced to 300 or 100 pA (again to produce a smaller spot

size) and it is thinned, using the line milling and tilting procedure, until it is

approximately 100 nm thick. Some materials are under intrinsic stress and as

such may buckle during the ﬁnal thinning to electron transparency. Walker

[43] showed that a cut made at the side edge of such lamella can relieve the

stress, resulting in the straightening of the lamella and enabling it to be

further thinned to electron transparency.

During the ﬁnal thinning, if a single ion beam system is being used the

sample is repeatedly imaged and the real-time SE signal is monitored for any

sudden increases in intensity to ensure that neither beam nor stage drifts

cause the lamella to be milled. If an FIB/SEM system is being used the

electron beam can be used to monitor and end the thinning once the thick-

ness of the remaining Pt has been reduced to a few hundred nanometers.

Techniques to thin the lamellae below 100 nm are discussed in Section 8.6.3.

A ﬁnished sample, through some CMOS transistor gates, ready for TEM, is

shown in Figure 8.2(d).

As FIB milling is relatively slow to remove large volumes of material it is

not practical to use it to prepare lamellae of features that are hundreds of

micrometers beneath the surface of a sample. To mill away a trench that is

Focused ion beam systems220

100 mm deep, 50 m m wide, and 60 m m long would take over 8 hours when

using a beam current of 20 nA. To overcome this limitation to enable

lamellae to be prepared from crack tips located hundreds of micrometers

beneath the surface of the sample both Deshais et al. [44] and Huang et al.

[45] combined the H-bar technique with BIB milling. They both used

conventional dicing and polishing to prepare the slices for the H-bar

technique, but then used BIB milling to sputter the slice to decrease the

distance of the crack tips from the surface (which also reduced the width of

the slice) before the sample was placed in the FIB system. Mechanical

polishing to the crack tips was not used as there were concerns that this

might affect the integrity of the cracks.

8.2.2 Ex-situ lift-out

The ex-situ lift-out technique was ﬁrst described by Overwijk et al. [46] and

involves cutting a lamella free from the sample, lifting it out, and placing it

onto a TEM support grid using a needle, micromanipulator, and optical

microscope. From the mid 1990s until recently (circa 2005) this has been the

main technique used.

In this technique, the whole of the sample (if it is of suitable dimensions,

otherwise it has to be diced or polished) is placed in the FIB chamber. As

with the H-bar technique, the ROI is identiﬁed and coated with a Pt strap.

Staircase-shaped cuts, positioned 1.5 m m from either side of the Pt strap, are

then milled using a large beam current (2 to 20 nA) (Figure 8.3(a)). Staircase-

shaped cuts rather than rectangular ones are milled to reduce the milling time

as in the subsequent steps the slice/lamella is only ever milled/viewed at 45



The staircase cuts are typically milled 15–20 mm wide and 8–12 m m deep. If

the dimensions are much smaller than this, it becomes difﬁcult to lift out the

lamella with the needle and micromanipulator.

The material between the staircase cuts is then thinned to approximately

300 nm by using the line milling and tilting procedure, outlined for the H-bar

technique. The sample is next tilted to 45



and the base and side edges (one

completely and the other only 80% along its length) of the slice/lamella are

milled through (Figure 8.3(b)). The changes in intensity of the SE signal

during the milling can be used to determine when the lamella has been cut

through. It is very important that the base and sidewalls are completely cut

through otherwise the lamella can not be lifted out from the FIB cuts. To

ensure this the box used to cut through the base should be positioned at least

1 mm above the base of the slice and the depth to which it is milled should be

Preparation for physico-chemical analysis 221

at least twice the thickness of the slice. (As the box is milled at 45



redepo-

sition onto its sidewalls will occur which reduces the sputter rate.) In a single

beam system if there is any uncertainty as to whether the slice/lamella has

been cut through the sample should be rotated by 180



and the opposite side

of the slice/lamella inspected for the cuts. If using an FIB/SEM system it is

not necessary to rotate the sample. Once the FIB has cut through the slice/

lamella it will mill a U-shaped cut into the opposite staircase which can be

seen by electron beam imaging as this is perpendicular to the surface of the

sample (see Figure 8.5(c)).

(c)

(a)

Staircase cuts

(b)

Cuts to the side

wall and base

(d)

TEM lamella

(f)

TEM lamella

(e)

TEM lamella

Figure 8.3 SE images of (a) staircase cuts milled on either side of the Pt,

(b) sample tilted to 45



whose sidewal ls and base have been cut, (c) an

ex-situ micromanipulator and optical microscope, (d, e) a needle being

pushed against a lamella, and (f) a lamella being placed onto a carbon

support membrane.

Focused ion beam systems222

The sample is then tilted back to zero degrees, and again the tilting and line

milling procedure and a beam current of 300 pA or 100 pA is used to further

thin the lamella to 100 nm. The remaining material at the holding sidewall is

then cut to free the lamella from the sample. Again, if a single ion beam

system is being used the intensity of the SE signal can be used to determine

when the lamella has been cut free from the sample. Usually, the lamella falls

into the path of the ion beam resulting in a sudden increase in the intensity of

the SE signal. Once cut free, in a single beam system care must be taken not

to image the sample at too high a magniﬁcation (>5000 times) in order to

prevent further ion implantation and damage to the lamella. If an FIB/SEM

system is being used electron beam imaging can be used to check that the

lamella has been cut free from the sample. This advantage and other

advantages of using an FIB/SEM system for the preparation of TEM lamellae

are discussed in detail in Section 8.5.

The next step involves lifting the ‘‘free’’ lamella away from the FIB cuts and

placing it onto a carbon coated TEM grid. As this can be as time consuming as

preparing the lamella and more difﬁcult to do it is described in detail.

The sample is removed from the FIB chamber and placed under an optical

microscope equipped with long working distance lenses (typically 13 mm or

18 mm), to enable a micromanipulator to be used beneath them (Figure 8.3(c)).

(An invaluable aid when teaching the lift-out step is to have a TV camera

attached to the microscope.)

Although both glass and metal needles have been used for lifting out the

lamellae the majority of researchers use glass needles as it is easier to get the

lamellae to attach to these. Ebel et al.[47] investigated using glass and metal

needles for lifting out GaAs lamellae. They reported the highest success rate

when using ﬂoating glass needles and for the metal needles it was necessary to

bias them to get the lamellae to adhere.

Glass needles can be made by melting the central part of 1 mm diameter

glass rods and pulling the two ends at a controlled rate. Metal needles can be

prepared by a modiﬁcation of the electro-polishing methods used for making

scanning tunneling microscopy tips [48]. A recipe that produces suitable

tungsten needles consists of dipping 50 mm diameter wire in 2 M KOH for

15 minutes while applying 20 volts and then for 6 minutes while applying

6 volts. Dipping [49] the needles into the electrolyte produces needles with

long shanks which enables their tips to reach to the base of the FIB cuts.

Using a low magniﬁcation (50·), the FIB cuts are placed at the center of

the optical axis and the needle is positioned about 5 mm above them. The

optical magniﬁcation is then increased to 500 times and the needle tip brought

into focus. To prevent crashing the needle into the sample as it is lowered, the

Preparation for physico-chemical analysis 223

microscope is over-focused and the needle lowered and brought into focus.

This process is repeated until the needle’s tip is positioned above the FIB

cuts. The sample is then rotated to align the lamella parallel to the needle’s

shaft. To illustrate the lift-out steps SE images, from when the technique has

been performed within an FIB/SEM system, are used (this alternative

approach is discussed at the end of this section) as these are more informative

than the equivalent optical images. The needle is further lowered and

repeatedly swept through the FIB cuts and pushed against the lamella until

the latter sticks to it through electrostatic forces (Figures 8.3 (d) and (e)).

Once the lamella is attached to the needle, the needle is raised away from the

sample and the sample is replaced with a carbon coated TEM grid. The

needle and lamella are then lowered (again by focusing on the needle tip,

over-focusing, and lowering the needle into focus) and then repeatedly swept

across the surface of the carbon membrane until the lamella sticks to the

carbon. Generally, the adhesion between the lamella and carbon membrane

is greater than that between the needle and lamella. If necessary, the needle is

then used to push the lamella ﬂat onto the carbon (Figure 8.3 (f)). The

adhesion between the lamella and carbon is sufﬁciently strong that only

rarely does a TEM lamella fall off during transportation to or mounting in

the TEM. As well as continuous carbon membranes, holey and lacy carbon

ﬁlms can be used [50, 51]. However, adhesion of the lamella to these supports

is not as good (due to the smaller contact area) and typically 1 in 30 samples

may be lost while transferring the grids to the TEM.

The success rate for the lift-out step ranges from 60% to 90%. Failure can

occur when the lamella is being lifted out from the FIB cuts or as it is being

placed onto the supporting membrane. During the lifting out step, the

lamella may ‘‘jump’’ up the shaft of the needle, away from the tip, such that it

cannot be brushed against the carbon membrane. Several approaches have

been used to reposition the lamella at the needle’s tip. These include (i) using

a second micromanipulator, (ii) brushing the needle against the rim of the

TEM grid, and (iii) tapping the micromanipulator so that the lamella jumps

off the needle and onto the support membrane. Of these three methods,

tapping the micromanipulator is probably the quickest and easiest one to do.

The lift-out can also fail if the carbon membrane is accidentally punctured

and ripped by the tip of the needle as the lamella is being placed onto it

(Figure 8.4 (a)). This can result in (i) the carbon support membrane rolling

around and encasing the lamella, (ii) the lamella on the ripped carbon

membrane being at a large angle to the electron beam in the TEM such that it

cannot be tilted onto a crystal axis (Figure 8.4(b)), and (iii) the lamella falling

through the resulting hole and being lost. To prevent these problems TEM

Focused ion beam systems224

grids that have 10 m m sized holes and no support membranes can be used

[52]. Also, as the needle’s tip can pass through the grid holes, this makes it

easier to get the lamella off the needle. The absence of the carbon support

membrane is also advantageous for high-resolution electron microscopy

(HREM). Once the lamella is on this type of grid (Figure 8.4 (c)) it is possible

to manoeuvre it over a grid hole (Figure 8.4(d)) by pushing it with the tip of

the needle. As with the carbon coated TEM grids adhesion of the lamellae to

these grids is very good. Typically, only about 1 in 40 lamella is lost during

transport to and mounting in the TEM.

(d)

(a)

(b)

TEM lamella

(c)

TEM lamella

(e)

TEM lamella

(f)

Pt strap

Figure 8.4 SE images of (a) a carbon suppo rt membr ane punctured by the

tip of the needle, (b) a lamella sat on a torn membr ane, (c) a Cu grid without

a support membrane onto which a TEM lamella has been placed and (d) the

same lamella after being pushed over a hole, (e) a lamella that has moved

against the sidewall of an FIB cut, and (f) a slice with a Pt strap that has

been used to ‘‘weld’’ the different layers.

Preparation for physico-chemical analysis 225

The lift-out step may also fail if the lamella jumps (due to electrostatics)

away from, or moves against, the sidewalls of the FIB cuts (Figure 8.4 (e)).

As the lamella is symmetrical in shape it can readily be distinguished from

any dirt or debris and thus, if it does jump away, it is possible, with patience,

to ﬁnd it and to lift it up. If the lamella moves against the side of the cuts it

becomes difﬁcult to see it with the optical microscope and to get it to attach

to the needle. Repeatedly swinging the needle back and forth against the side

of the FIB cut may knock and move the lamella such that it can then be lifted

out. Unfortunately, when doing this the tip of the glass needle may shatter

showering the FIB cuts and lamella in glass fragments.

Samples containing cracks or numerous layers that are poorly adhered can

also be difﬁcult to lift out as these may fracture or the layers delaminate as

the needle tip is being pushed against them. Huang et al.[45] prevented the

fracturing of lamellae containing cracks by leaving a supporting frame

around the lamellae. For samples with poor adhesion between the layers, a Pt

strap may be deposited onto the face of the lamella (by tilting the sample to



) to ‘‘weld’’ the different layers together (Figure 8.4 (f)) before it is thinned

to electron transparency.

The lift-out step has also been performed in the chamber of FIB/SEM

systems using SE imaging [ 53]. This enables higher magniﬁcations to be used

than is possible optically giving more control over positioning the needle

relative to the lamella and making it easier for new users to learn. In this

approach it was found that the highest yield occurred if a tungsten needle

with a radius of curvature of about 1 m m and an accelerating potential of

5 kV was used. Unfortunately, even though higher magniﬁcations are pos-

sible the success yield of this approach (70%) is comparable to that of the

ex-situ lift-out method. This, when coupled with the fact that the FIB system

is being used for this step, means there is limited value in using this approach

relative to the ex-situ method or the in-situ slice method which is discussed in

the next section.

8.2.3 The in-situ slice technique

This technique can be viewed as a hybrid of the ex-situ lift-out and H-bar

techniques. It was ﬁrst reported by Ohnishi et al.[29] but for many reasons

(such as patent issues and availability of suitable in-situ micromanipulators)

it has taken many years to become accepted within the FIB community and it

is only recently (circa 2005) that it has become the main FIB based technique

being used. The chief reason why this method has superseded the ex-situ lift-

out technique is because its success yield can be close to 100%. This high

Focused ion beam systems226

success yield is because Pt ‘‘welding’’ is used to ﬁx a slice of the sample onto

the needle’s tip as against relying on electrostatic forces.

The in-situ micromanipulator is an integral part of this technique. These

have been mounted in the gas injector ports, on the stage, or on the sidewalls

of the chamber (Figure 8.5(a)). An advantage of mounting the micro-

manipulator independently of the stage is that the sample can be moved to

different locations while the tip of the needle remains ﬁxed on the electro-optic

axis. In addition to manipulating TEM lamellae in-situ micromanipulators

(a)

(e) (f)

(b)

Base and side cuts

(c)

Needle Pt welded to a

slice

(d)

Slot grid

Figure 8.5 (a) Image of in-s itu micromanipulator mounted on the stage,

(b) SE image of a 500 nm thick slice that has had its base and sidewalls cut

through, (c) SE image of a needle ‘‘welded’’ to a slice, (d) SE image of a slice

being attached to a TEM slot grid, (e) SE image of a TEM grid with both

tabs and V-slots, and (f) SE images of thinned lamellae mounted on a

V-shape groove and a tab (insert) on a TEM grid.

Preparation for physico-chemical analysis 227