Geckeler K.E., Nishide H. (Eds.) Advanced Nanomaterials

Подождите немного. Документ загружается.

21.2 Near-Field Raman Imaging Techniques 679

Raman spectra. Although extensive efforts have been devoted to fabricate ﬁ ber tips

with a higher throughput, the throughput is still very low, especially for tips with

an aperture < 200 nm. In addition, the Raman signal is intrinsically weak (typically

less than 1 in 10

photons). A typical Raman spectrum takes several minutes to

record when using the conventional near - ﬁ eld Raman method, which makes it

prohibitive to construct a Raman image in this way. For example, when Webster

et al . [7] used the aperture technique to study the stress distribution of a scratch

on a Si crystal wafer, a period of about 9 hours was required to obtain a Raman

image of 26 × 21 points.

An alternative approach to near - ﬁ eld Raman microscopy is to use an aperture-

less conﬁ guration (which is also known as tip - enhanced Raman scattering ; TERS )

[8 – 10] , employing a metal or metal - coated tip (see Figure 21.2 ). In this conﬁ gura-

tion, the laser impinges on the apex of the metal tip, inducing a strong local electric

ﬁ eld enhancement in the vicinity of the metal tip apex. In this case, the Raman

Figure 21.2 (a) Schematic diagram of

tip - enhanced Raman scattering in reﬂ ection

geometry. The metal tip is brought to the

sample surface by tuning fork or cantilever.

The Raman signals can also be collected by

the focus lens instead of the collection lens;

(b) Magniﬁ ed view of the tip region. The

far - ﬁ eld Raman signal comes from the whole

laser - illuminated area (a few μ m) and is

featureless, while the near - ﬁ eld Raman signal

comes from only the tip region ( < 100 nm) and

carries information related to the device

features.

680 21 Near-Field Raman Imaging of Nanostructures and Devices

enhancement is a result of local ﬁ eld enhancement by the tip, or increased polariz-

ability of the sample due to an interaction between the tip and sample, which is

similar to surface - enhanced Raman scattering. As the laser is delivered to the

sample surface by an objective lens rather than by a metal - coated ﬁ ber tip, the low

optical throughput of conventional ﬁ ber tips can be overcome. With a suitable

metal tip (usually Ag or Au), the Raman signal from the sample close to the tip

can be enhanced by several orders of magnitude.

Besides the higher overall Raman throughput, TERS has several other merits

over the apertureless conﬁ guration:

1 . A higher spatial resolution : At least in principle, it is much easier to fabricate

a sharp tip than a small aperture. A spatial resolution < 10 nm has been achieved

using TERS.

2 . Fewer topographic artifacts : In aperture conﬁ guration the tip is large, although

its aperture may be small; therefore, the tip cannot follow the sample surface

exactly. This is normally not a problem, and is less severe in TERS.

3 . A variety of potential tip materials : In addition to noble metals, silicon and

its oxides, nitrides, nanowires and nanotubes may represent good candidates

for tips.

4 . A wider spectral response range : Unlike the aperture technique, which usually

functions in the visible region due to a need to use optical ﬁ bers to deliver the

laser, TERS can extend its applications to a much wider range, for example

from ultraviolet to IR.

Unfortunately, TERS also has some major disadvantages, the main challenge

being to obtain reproducible near - ﬁ eld images. It is very difﬁ cult to achieve a

consistent near - ﬁ eld enhancement in TERS, due to difﬁ culties in controlling the

geometry of the metal tip. The technique also encounters problems of wear and

tear, oxidation, and contamination of the tip apex (e.g., by carbon) that may have

severe adverse effects on the near - ﬁ eld enhancement. Another problem is that the

laser spot, when focused on the tip apex, causes an intense background (far - ﬁ eld

signal) that should be eliminated in order to achieve a better SNR [11, 12] . As a

result, TERS is rarely used for routine characterization.

A third approach, developed by the present authors ’ group, uses a dielectric

microsphere to focus the laser to a spot size that is smaller than the diffraction

limit. Besides being used as the excitation source for Raman spectroscopy, the

incident laser beam (linearly polarized Gaussian TEM00 mode) is also used to trap

the microsphere just above the sample surface, using the well - known “ optical

tweezers ” mechanism. In these experiments, the sample and polystyrene micro-

spheres (3 μ m diameter) were placed in a sample cell ﬁ lled with deionized water.

One microsphere was trapped at the center of the laser beam and was in contact

with the surface of the sample during scanning. The experimental set - up is shown

schematically in Figure 21.3 a and b.

21.2 Near-Field Raman Imaging Techniques 681

Figure 21.3 (a) Schematic diagram of the near - ﬁ eld Raman

microscope with microsphere; (b) Detailed description of the

sample cell: 1 = water immersion lens (60 × NA = 1.2 water

immersion), 2 = water, 3 = focused laser, 4 = cover glass,

5 = polystyrene microsphere (3 μ m), 6 = sample, 7 = sample

cell ’ (c) Typical Si

–

Si Raman intensity versus position with the

ﬁ tted data to determine the laser spot size.

682 21 Near-Field Raman Imaging of Nanostructures and Devices

21.3

Visualization of Si

–

C Covalent Bonding of Single Carbon Nanotubes

Grown on Silicon Substrate

In this section, it is shown that Raman microscopy can be used to characterize

nanoscale samples, using Si

–

C bonding between the carbon nanotube s ( CNT s)

and the Si substrate as an example. Recently, CNTs have attracted a great deal of

interest on the basis of their extraordinary electrical, chemical, thermal, and

mechanical properties [13 – 15] , all of which make them promising materials for

applications in nanometer - scale electronics and devices. For example, a unique

property of CNTs is that they may be either metallic or semiconducting, depending

on their chirality and diameters [16] . The tuning of electrical properties of single

CNTs by inducing covalent bonding between the CNTs and the Si substrate has

been predicted [17 – 22] . Such direct integration of CNTs into well - developed silicon

technologies offers a new strategy of building nanoscale, or even single - molecule,

electronic devices by selectively forming covalent bonding between the CNTs and

the Si substrate.

Several groups have achieved Si – CNT bonding in ceramics with good mechani-

cal properties by annealing Si nanoparticles together with CNTs, which served as

a mechanical reinforcement [23] . The direct observation of Si

–

C covalent bonding

on single CNTs on Si substrate samples has not been reported, due partly to a lack

of suitable characterization techniques. However, it has been predicted that the

interaction between Si and the CNT results in a rich variety of changes in

the electronic structures of the CNT [19] . For example, when Miwa et al . studied

the single - walled CNT adsorbed onto partially and fully hydrogenated Si (001)

surfaces [19, 20] , it could be shown that the removal of a few H - atoms along the

adsorption sites would enhance the metallic character of the CNT. However, the

removal of all H - atoms would transform the CNT to the semiconducting state. By

using ﬁ rst - principle calculations, Peng et al . were able to determine the binding

energies of CNTs adsorbed onto different sites on Si (001) surfaces [22] . In fact,

their results showed that the absorption sites at the surface trench for CNTs paral-

lel to the Si dimmer rows, and between Si dimers for CNTs perpendicular to the

Si dimer rows, were stable.

Such stability was also shown to depend heavily on the size (diameter) of the

CNTs. When preparing these Si – CNT devices the ﬁ rst step is to fabricate Si – CNT

covalent bonding on a large scale. Irrespective of which technique is used to char-

acterize the sample, it must be capable of identifying those CNTs with Si

–

C bonds.

In these studies, a micro - Raman imaging technique [24] was used that was capable

of probing down to hundreds of nanometers, to investigate isolated single CNTs

and their interaction with the Si substrate. The CNTs were grown on a silicon

substrate at high temperature to achieve Si – CNT bonding, with Raman imaging

showing that some of the single CNTs were attached covalently to the silicon wafer

surface.

The CNTs were grown on the Si substrate using a thermal chemical vapor depo-

sition ( CVD ) technique, the substrate having been cleaned using trichloroethylene,

21.3 Visualization of Si

–

C Covalent Bonding of Single Carbon Nanotubes Grown 683

acetone, and ethanol in turn, by ultrasonic agitation. An ultrathin Fe ﬁ lm to be

used as the catalyst was deposited in vacuo (10

− 5

Torr) by ion beam sputtering. The

substrates were pre - annealed at 850 ° C in a hydrogen - ﬁ lled quartz tube furnace for

90 min. The CNTs were then grown at 1000 ° C for 20 min using methane, hydro-

gen, and argon with ﬂ ow rates of 75, 20, and 100 sccm (standard cm

min

respectively. The sample was cooled below 300 ° C before being exposed to ambient

air, in order to avoid any damage caused by rapid cooling.

A JEOL JSM - 6700F scanning electron microscope was used to study the surface

morphology of samples and locations of interest for further Raman studies. The

Raman spectroscopy and imaging were carried out using a WITec CRM200 confo-

cal Raman system ﬁ tted with an Olympus microscope objective lens (100 ×

NA = 9.5). A double - frequency Nd:YAG laser (532 nm, 100 mW laser; CNI Laser)

was used as the excitation source. The Raman scattered light was dispersed using

a grating (1800 grooves per millimeter), and detected using a thermoelectrically

cooled charge - coupled - device ( CCD ) cooled to − 64 ° C. A piezo - stage was used to

scan the sample for the Raman imaging. The spatial resolution of the Raman

imaging was approximately 250 nm.

The SEM image of the sample is shown in Figure 21.4 a. It can be seen that all

the CNTs were grown horizontally on the Si substrate, with random orientations.

The various diameters of the CNTs may be due to the different sizes of the Fe

catalyst. With assistance from markings on the Si substrate, Raman imaging was

performed in the same region shown in the SEM image. A careful analysis of the

Raman spectra in the Raman images revealed that the Raman spectra of the CNTs

shown in Figure 21.4 a could be indexed as two groups: (i) those consisting of only

normal CNT peaks; and (ii) those with additional peaks, which can be assigned to

–

C bonds (this point is discussed later in the chapter). For simplicity, these two

types of spectra are referred to as CNT spectra and Si

–

C spectra, respectively.

Figure 21.4 b shows the typical Raman spectra of the two groups. The CNT spec-

trum shows only Raman bands of normal CNTs: the D band (1344 cm

− 1

), G band,

and 2D band (2681 cm

− 1

) [25] . The G band can be further identiﬁ ed as the G – band

at 1570

− 1

and the G+ band at 1590 cm

− 1

. For certain CNTs (such as the CNT high-

lighted by the broken line in Figure 21.4 a), the Raman spectrum shows extra bands

belonging to Si

–

C bonds at approximately 1510 cm

− 1

(band A), 1692 cm

− 1

(band

B), 1921 cm

− 1

(band C), 1981 cm

− 1

(band D), 2392 cm

− 1

(band E), and 2620 cm

− 1

(band F). While several regions were tested in this study, it should be noted that

the extra bands (A – F) were observed only for certain CNTs, which suggested that

the Si

–

C bonds formed only for certain CNTs.

Raman imaging was performed in the sample region shown in Figure 21.4 a

with a scanning step size of 50 nm, and with the incident laser polarized in the

plane of the substrate. Raman images constructed using different peak intensities

are shown in Figure 21.5 a – d. Figure 21.5 a was generated by the G band (ca.

1580 cm

− 1

) intensity of the CNTs. Brighter regions represent CNTs, which show

an excellent correlation with those obtained using SEM. Images were also con-

structed using the intensity of the extra band A; the resultant image is shown in

Figure 21.5 b. The shape and position of the bright region matched perfectly the

684 21 Near-Field Raman Imaging of Nanostructures and Devices

CNT, emphasized by the dashed curve in the SEM image (Figure 21.4 a). Images

generated using bands B and F (not shown) exhibited near - identical patterns to

that of band A, clearly indicating that the three bands were all correlated to the

same CNT marked in the SEM image.

In order to conﬁ rm that peaks A – F in Figure 21.4 b were indeed related to Si

–

covalent bonds, the Raman spectra were compared with those of bulk SiC, which

have been well studied [26, 27] . Bulk SiC has several polytypes, which can give rise

to the following ﬁ rst - order Raman peaks: TA mode between 200 – 300 cm

− 1

, LA

mode at around 610 cm

− 1

, TO mode at around 790 cm

− 1

, and LO mode at near

970 cm

− 1

[28] . The TO mode at 790 cm

− 1

is by far the strongest Raman peak in all

polytypes, which is absent in the present sample (the reasons are given later). The

peak positions and relative intensities of the SiC second - order Raman bands varied

for different polytypes, with generally two characteristic bands: a broad band

Figure 21.4 (a) Scanning electron microscope image showing

the areas of interest. Raman bands belonging to Si

–

C bonds

can be detected on a carbon nanotube (CNT), as indicated by

the broken red line; (b) Comparison of Raman spectra

obtained on normal CNTs (black) and the special CNT (red).

The dashed lines represent the peak positions of the CNTs ’ D

band, G band, and 2D band.

between 1510 cm

− 1

and 1540 cm

− 1

, and a narrower band around 1710 cm

− 1

. Both

bands corresponded to multibands, resulting from the overlapping of different

overtone phonons of the SiC vibration [27, 29, 30] . As shown in Figure 21.4 b, both

band A (1510 cm

− 1

) and band B (1692 cm

− 1

) observed in these Raman spectra

showed a good agreement in shape and intensity with those two SiC characteristic

bands. The slight shift to lower wavenumbers ( Δ ω ∼ 10 cm

− 1

) may be due to the

small - size effect [31] . In addition, the weaker bands E and F also showed an agree-

ment with the broad band of SiC at 1930 cm

− 1

. As there have been no reports on

the Raman band of SiC above 2000 cm

− 1

, it was not possible to compare band F

with that of bulk SiC. However, the excellent agreement between the Raman image

using band F and those of band A and B showed that band F also belonged to the

–

C bonds in this sample.

The ﬁ rst - order Raman peak at 790 cm

− 1

is the strongest among all the Raman

peaks in all polytypes of SiC, but was absent from these experiments. This can be

explained by the substantial structural difference between the bulk SiC and

the present CNT/Si system, where all the Si

–

C covalent bonds were formed at

the Si – CNT interface. Such an unique an arrrangement deviates considerably

from that in the bulk SiC, where the Si

–

C covalent bonds are arranged in three

dimensions with a long - range order; in contrast, the bonds in the CNT/Si system

showed a long - range order in only one dimension – that is, along the long axis of

the CNTs.

Figure 21.5 Raman images constructed using peak intensity

of: (a) G band of CNTs and (b) Si

–

C band A at 1510 cm

− 1

The Raman images generated using bands at 1692 cm

− 1

and

2620 cm

− 1

are identical to those in panel (b). The green arrow

indicates the polarization of the incident laser. The blue

dotted lines serve as a guide to the eye, and show the

position and shape of the special CNT.

21.3 Visualization of Si–C Covalent Bonding of Single Carbon Nanotubes Grown 685

686 21 Near-Field Raman Imaging of Nanostructures and Devices

21.4

Near - Field Scanning R aman Microscopy Using TERS

The NSRM system used in these experiments consisted of a standard Nanonics

NSOM system and a Renishaw micro - Raman spectrometer. The Nanonics NSOM

system uses a bent optic ﬁ ber tip (cantilevered tip), which makes the scanning

head very compact. The compact design allows the easy integration of the

spectrometer and NSOM by placing the scanning head under the microscope

objective of the Renishaw system [32] . An argon ion laser (488 nm line) is focused

onto the metal tip, which is made from tungsten wire through electrochemical

etching [similar to the method used in scanning tunneling microscopy ( STM ) tip

preparation], and then coated with silver through radiofrequency ( RF ) - sputtering

[33] .

The interaction between the tip and the laser enhances the local electrical ﬁ eld

near the tip, which in turn enhances the Raman signal nearby. The same objective

lens collects the Raman signal in the backscattering geometry, which is then

directed into the Renishaw spectrometer by the coupling optics. In the conﬁ gura-

tion described above, both the near - ﬁ eld and far - ﬁ eld Raman signals are collected.

In order to reduce the far - ﬁ eld component, the laser beam must be focused tightly

to form a spot on the sample which is as small as possible. The collection efﬁ ciency

of the optical system should also be high in order to obtain a good SNR, since the

Raman signal is intrinsically weak. These two aspects require a lens or objective

with a large numerical aperture ( NA ). In addition, the polarization direction and

incident angle have strong effects on near - ﬁ eld enhancement, and these should

be tuned carefully. The polarization direction is tuned by a λ /2 plate, which is

placed just before the laser and adjusted to be perpendicular to the beam. As

shown by computer simulations, the near - ﬁ eld enhancement is not maximum

when the laser beam is parallel to the tip axis. To maximize the near - ﬁ eld enhance-

ment, two approaches can be used: the ﬁ rst approach is to set the tip oblique to

the sample and use backscattering geometry; the second is to use a perpendicular

tip and an oblique incident laser.

To observe the near - ﬁ eld enhancement, an experiment on single crystal

silicon was performed. For this, the single crystal silicon was placed on the

piezo scanning stage of the NSOM system, and the silicon Raman peak at

520 cm

− 1

recorded for two tip positions: (i) with the tip “ touching ” the silicon

surface to record the Raman spectrum, which includes both the near - ﬁ led

and far - ﬁ eld components; and (ii) with the tip lifted from the surface so as to

record the far - ﬁ eld Raman spectrum. These are shown as spectra in Figure 21.6 a

and b, respectively. The near - ﬁ eld component is approximately 50% of the far - ﬁ eld

signal.

It should be noted that the near - ﬁ eld Raman signal comes from a small

region on the sample, while the far - ﬁ eld signal comes from a much larger volume

(about 2.5 μ m in diameter for our instrumental setup and the laser penetration

of Si is about 0.5 μ m). Assuming that the diameter of the metal tip is several

tens nanometers, say 70 nm, and the near - ﬁ eld enhancement occurs at a depth

21.4 Near-Field Scanning Raman Microscopy Using TERS 687

Figure 21.6 The silicon Raman spectra with the tip “ touching ”

the sample surface (spectrum A) and the tip withdrawn

(spectrum B).

of 20 nm of the Si sample. The far - ﬁ eld signal derives from a volume of

far

()

×π

2500

500

nm , while the near - ﬁ eld signal is derived from a much

smaller volume

near

()

×π

. The enhancement factor can be

estimated as

near

far

near

× , where I

near

= 0.35 I and I

far

= (1 − 0.35) I are the measured

near - and far - ﬁ eld Raman intensities, respectively. Such a simple estimation shows

that the near - ﬁ eld enhancement factor is more than 10

. In another experiment,

the improvement in spatial resolution was demonstrated by using a featured

sample: a silicon device, which consists of SiO

lines of 380 nm width, with 300 nm

separations. The SiO

lines were 30 nm higher than the silicon substrate ( Figure

21.7 a). By comparing the width of the SiO

strips in the AFM proﬁ le and the real

width (380 nm), the topographic spatial resolution of the set - up was derived as

34 nm (Figure 21.7 b). Raman mapping (128 × 20 points) was performed on a strip,

and the Raman image constructed using the Raman peak intensity, as shown in

Figure 21.7 c. The Raman spectra were collected with a 1 s integration time. In this

experiment, the Raman spectrum contained both far - ﬁ eld and near - ﬁ eld compo-

nents when the tip was on a Si - single crystal, but had only the far - ﬁ eld component

when the tip was on SiO

, where the SiO

prevented the tip from coming close

enough to the Si to excite its near - ﬁ eld Raman signal. Hence, the intensity varia-

tion of the Si Raman peak represented the near - ﬁ eld contribution, while the far -

ﬁ eld contributed a constant background. The Raman intensity image shown in

Figure 21.7 c has a clear correlation with the Si device, as might be expected. This

was the ﬁ rst two - dimensional ( 2 - D ) near - ﬁ eld Raman image of a real nanometer -

sized device. The 1 s integration time required to record a good quality Raman

spectrum is short enough to make NSRM a useful technique for imaging

purposes. It should be noted that the SiO

lines in the NSRM image are

not straight, due to hysteresis and creep of the piezoelectric scanner, which is

688 21 Near-Field Raman Imaging of Nanostructures and Devices

calibrated and linearized in normal AFM scanning. In Raman mapping, however,

the scanning rate is much lower than that in normal AFM scanning, as the

scanner must remain at each point for several seconds while the spectrometer

records the Raman spectrum. Thus, the creep will become more obvious during

the mapping process.

21.5

Near - Field R aman Imaging Using Optically Trapped

Dielectric Microsphere

Recently, a new approach to near - ﬁ eld imaging has been developed, whereby the

laser is focused to a spot size which is smaller than diffraction limit, by using a

dielectric microsphere. In addition to being used as the excitation source for Raman

spectroscopy, the incident laser beam (linearly polarized Gaussian TEM00 mode)

is also used to trap the microsphere just above the sample surface, through the

well - known “ optical tweezers ” mechanism [34, 35] . Simulation studies have shown

that subdiffraction - limited focusing can be achieved when the diameter of the

dielectric microsphere is comparable to the wavelength of laser [36, 37] . In the

experiments, the sample and polystyrene microspheres (3 μ m diameter) in solution

were placed in a sample cell ﬁ lled with deionized water. One microsphere was

trapped at the center of the laser beam, and was in contact with the surface of the

sample during scanning. The schematic diagram of the experimental set - up is

shown in Figure 21.3 a. The near - ﬁ eld Raman microscopy set - up with polystyrene

microsphere is based on the WITec CRM200 confocal Raman microscopy system

Figure 21.7 (a) Schematic cross - sectional diagram of the Si

device sample; (b) Atomic force microscopy image of the Si

device sample; (c) Raman intensity image of the Si device

sample.