Binh1996 PDF

ADVANCES IN IMAGING AND ELECTRON PHYSICS. VOL.
95
Electron Field Emission from Atom-Sources: Fabrication,

Properties, and Applications of Nanotips
VU THIEN BINH
Laboratoire d'Emission Electronique, DPM-URA CNRS, Universite Claude Bernard
Lyon I , 69622 ViNeurbanne. France
N . GARCIA
Fisica de Sistemas Pequeiios, CSIC, Universidad Autonoma de Madrid, CIII, 28049
Madrid, Spain
AND
S . T. PURCELL
Laboratoire d'Emission Electronique, DPM-URA CNRS, Universite Claude Bernard
Lyon I , 69622 Villeurbanne, France
1. Introduction . . . . . . . . . . . . . . . . . . . .
. 63
11. Electron Emission from a Metal Surface: Summary of the Basic Results. . 64
A. Metalivacuum Barrier . . . , . . . . . . . . . . . 64
B. Emission Currents . . . . . . . . . . . . . . . . . . 66
C. Energy Distribution of Emitted Electrons . . . . . . . . . 72
D. Current Density Distribution , . . . . . . . . . . . . . . 74
E. Current Stability . . . . . . . . . . . . . . . . . . 78
111. Electron Emission from Nanotips . . . . . . . . . . . . . . 81
A. Experimental Setup and Procedures. . . . . . . . . . . . . 82
B. Confinement of the Field Emitting Area . . . . , . . . . . . 84
C. Field Emission Characteristics from Nanotips: Experiment . . . . . 97
D. Field Emission Characteristics from Nanotips: Discussion . . . . . . 104
IV. Applications . . . . . . . . . . . . . . . . . . . . . 112
A. Atomic Resolution under FEM . . . . . . . . . . . . . . 112
B. Monochromatic Electron Beam . . . . . . . . . . . . . . 115
C. Local Heating and Cooling by Nottingham Effect . . . . . . . . 118
D. Fresnel Projection Microscopy , . . . . . . . . . . . 124
E. Ferromagnetic Nanotips: Atomic Beam Splitter . . . . . . . . . 145
V. Conclusions . . . . . . . . . . . . . . . . . . . . . 149
References . , , . . . . . . . . . . . . . . . . . . 150
1. INTRODUCTION
A major instrumental development in the history of electron optics and

electron microscopy occurred in the early 1960s when the field emission
Copyright 1996 by Academic Press. Inc.
63 All rights of reproduction in any Form reserved.
64 VU THIEN BINH ET AL.
gun (FEG) replaced the thermionic emitter as the electron source in the
scanning electron microscope. In 1965, Crewe [l] made the first experi-
mental demonstration that a dramatic improvement in resolution could
be made by the use of the FEG, because this permitted the electron beam
to be focused into a probe area of only a few angstroms in diameter.
Further developments can be expected if new improvements of the field
emission (FE) tip can be realized, in particular by decreasing (1) the size
of the emitting area, (2) the angular dispersion of the emitted flux, (3) the
extraction voltage, and (4)the width of the energy distribution of the
emitted electrons; and by increasing the stability of the emission. This
chapter summarizes the improvements in all of these aspects that can be
realized by the use of a nanotip as a FE source. These nanotips are
single-atom sharpness nanoprotrusions, 2 to 5 nm in height, on top of
hemispherical base tips for which the whole FE current is emitted from
the topmost apex atom [2, 31.
The object of this chapter is to present the consolidated results obtained
with controlled field emission from nanotips and in particular to discuss
the specific effects related to the fact that the source is atomic size. This
chapter does not present an exhaustive review of field emission. A good
number of review papers and books exist in the literature about field
emission and its applications, even at its very beginning [4-81. However,
in Section I1 we summarize the basic results of field emission theory,
including discussions of thermionic emission, in order to place the field
emission from nanotips in the global context of electron sources. Charac-
teristics of field emission from nanotips are then presented and discussed
in Section 111.
The usefulness of the very specific properties of nanotips is convincing
only if they permit new advances to be developed. The demonstration
of their utility will be explored in Section IV. Among other examples,
nanometric-resolved images of synthetic polymers and RNA-based biolog-
ical molecules will be presented and discussed. They were obtained with
the Fresnel projection microscope (FPM), using the nanotip as an atom-
size electron source.
11. ELECTRON
EMISSION FROM A METALSURFACE:
SUMMARY OF THE BASICRESULTS
A . MetallVacuum Barrier
Within a metal, an electron current density of roughlyj, = enou, impinges

on the inner surface, where no is the electron density, e is the electron
ELECTRON FIELD EMISSION FROM ATOM SOURCES 65
Position (A>
FIGURE1 . Potential energy for an electron in the vicinity of a metal surface with and
without applied fields. The decrease in the effective barrier due to the Schottky effect is
3.8 F”z.
charge, and urnis the electron velocity. For no between 10” and loz3cm-3
and u, = lo8 cm * s-‘ near the Fermi energy E F ,j, is -10I2 A . cm-2.
Only a small fraction of this current escapes from the metal due to the
surface tunneling barrier which is presented schematically in Fig. 1. This
barrier is higher than EF by the value of the work function #, about 2 to
5 eV at zero applied field, and it is modified by the application of an
electric field F, which decreases the potential energy by an amount of
-eF,x outside the metal, x being the distance from the tip surface. Near
the surface the emitted electron experiences an image force, which comes
from the attraction of the induced positive charge in the metal. The stan-
dard form for potential energy of an electron V ( x ) at a distance x from
the cathode with the zero of energy at the bottom of the conduction band is
V ( x )= EF + # -eF,x - - - E F
e2
+ # -eF,x 3.6
-- forx >x,
4x X (1)
V ( x )= 0 for x < x,
with the energies in eV, the fields in VIA, the distance x in A,and x, =
3.6/(EF + #) = 0.3 A is chosen such that V(x,) = 0.
The potential energy given by Eq. (1) shows a maximum at
giving an effective work function in the presence of the applied field of

+en = + - e3l2F:l2= 6 - 3.8FA1*. (3)
This reduction of the height of the barrier by 3 . 8 F f 2 is called the Schottky
effect [9].The barrier width, Ax, for electrons at EF can be obtained from
Eq. (1) to be
Extracting electrons from metal surfaces can be done by two main

processes, which are represented schematically in Fig. 2:
1. The emission of thermally excited electrons having energy greater
than the barrier height. This process includes thermionic emission
for zero applied field [Fig. 2.I(a)], and also the Schottky emission,
which is thermionic emission in the presence of an applied field
[Fig. 2.I(b)].
2. The tunneling emission of electrons through the barrier with energy
lower than the barrier height (Fig. 2.11). This process includes cold
field emission for E < EF at T = 0 K and also thermal field emission
for thermally excited electrons having EF < E < +eff at T > 0 K.
There is an overlap between these two regimes where the field and temper-
ature are such that electrons are emitted both over and through the barrier.
As described in the following sections, the main features of the tempera-
ture and field dependence of the emission currents and the energy distribu-
tions of emitted electrons can be understood by considering the electrons
inside the metal to be a free electron gas.
B . Emission Currents
1. Thermionic Emission
In zero applied field, the thermionic current J, is given by considering the
flux from all the electrons that have energy normal to the surface greater
than the barrier height 4:
wheref(p) = 1/{1 + exp[(E - E F ) / k B T ]is} the Fermi-Dirac distribution

function for electrons with momentum p and Eminis the minimum normal
energy to pass the barrier and kB is the Boltzmann constant. Considering
the fact that
I. Thermionic Emission
Vacuum
11. Field Emission (FE)
FIGURE2. Schematic diagrams for electron emission from a metal, with the respective
energy distributions of the emitted electrons. I. Thermionic emission: (a) without applied
field; (b) in the presence of applied field (Schottky ernmission). 11. Field emission (FE).
where m is the mass of the electrons, then
J =?!I
h3
+X
-=
--m
The thermionic emission condition for electrons of the metal is

2
Px EF+ 4.
-2 (7)
2m
Thus the minimum energy of an electron to be emitted is
Emin= EF + 4 + P: + PZ
-. 2m
Integration gives
Jx = - h3
2ekBTexp
(-")kBT I+Xexp (-&)
--m2 m k ~ TdpyI:mexp (-L)
2mkBT dpz, 00
J41r mek2
x2 = d
h3
T exp (-&)
= AoT2exp (-A),
kB T
(9)
Equation (9) is the Richardson-Dushman equation. Some parameters for

several standard emitters are given in Table I.
Though Eq. (9) is for a step barrier, it has the same form for a field-
+
modified barrier. For this case, has to be replaced by 4eff given by Eq.
(3). The applied field lowers the height of the barrier by A+ = 3.8FA'2 and
then the thermionic emission current will be enhanced by exp(A+/kBT).
The lowering becomes noticeable for Fo> lo5 V/cm, as can be seen from
the values in Table 11.
This effect is used in the so-called Schottky emission cathodes, which
are basically ZrO-covered W(100)tips with radii -1 pm. The presence
of ZrO also lowers the intrinsic work function from 4.5 eV to 2.8 eV,
TABLE I
PARAMETERS OF SOME THERMIONIC
EMISSION
CATHODES AT
ZEROAPPLIEDFIELD
W LaB6 ZrOIW
Work function 4 (eV) 4.5 2.4-2.7 2.8

Richardson constant A. (A cm-* K-2) 75-120 30 100
Emission current density j , (A cnC2) 1-3 25 500
Working temperature T (K) 2600-2900 1400-2000 1800
TABLE I1
VARIATION
OF THE WORKFUNCTION AND CORRESPONDINGEMISSION
CURRENT
DENSITIESWITH APPLIED
FIELDI N THE SCHOTTKY
REGIME
Fo (V/cm) 105 1 06 107 108

(VIA) 0.001 0.01 0.1 1 .o
A 4 (eV) 0.12 0.38 1.2 3 .O
j(Fo)/j(Fo= 0) (4 = 4.5 eV, T = 2000 K) 2 8 800 1.5 x 109
which permits an emission current density of more than two orders of

magnitude larger (-500 A/cm2) at a temperature of 1800 K, in contrast
to clean W thermionic cathodes at -2600 K (see Table I). Although the
potential barrier is lowered by the Schottky effect, the electrons still have
to overcome the barrier by their thermal energy, so it is a little confusing
to call this cathode a field emission Schottky gun.
2. Field Emission
Application of an external electric field lowers the potential barrier and
also modifies the position of the maximum x, and the barrier width as
shown in Table 111. For Fo < lo7 V/cm, the barrier width is large and it
is practically opaque to the electrons. Appreciable tunneling occurs for
F, > 2-3 x LO’. The tunneling current was named field emission because
there is no need to heat the cathode in order to deliver the electron current.
The Fowler-Nordheim (F-N) model describing the electron emission
from metals by application of a high electric field was developed by assum-
ing that the temperature of the metal is 0 K, the free-electron approxima-
tion applies inside the metal, the surface is smooth and planar, and the
potential barrier closing the surface in the vacuum region consists of an
image force potential and a potential due to the applied electric field F,.
The emitted flux is found by considering the product of the supply function
for the flux of electrons impinging on the barrier and the transmission
TABLE I11
MAXIMUM,
xo, AND WIDTH,Ax (ATEF) OF THE TUNNELING
(4 = 4.5 eV)
BARRIER
~
Fo (V/cm) 1 os 106 10’ 3 x 10’ 108

xo (A) 60 20 7.0 3.5 2.0
Ax (A) 4500 450 45.0 15.0 4.5
probability, D(E,). D(EJ depends on the height and width of the barrier,
and the subscript x specifies that the transmission probability depends
only on the component of energy normal to the surface. D(E,) for a free
electron gas in the WKB approximation is given by [lo]
Considering the fact that the electrons tunnel in the narrow range of energy
near E F , Eq. (10) results in
with
and
e 3 / 2 I/?.
~ F
w = o = for F, in V /A and 4J in eV.
4J 3-8T
Do corresponds to the transmission probability at the Fermi level, and
t ( w ) and u ( w ) are nondimensional, slowly varying functions derived from
elliptic integrals [l 11 to take into account the image forces during the
tunneling process. For FE, t(w) = 1 and u(w) ranges from 0.4 to 0.8.
The tunneling current for a given function D(E,) is
Substituting D(E,) from Eqs. (1 1) into the integral gives
J , = 1.55 x 10”- Fi
4t2(W)
with J , in A/cm2, 4J in eV, and F, in VIA. In Table 1V we give some
numerical estimates of the applied fields necessary to achieve current
densities of lo6 and lo7 A/cm2.
As the current density impinging on the metal surface from the inside
of the metal is about 10” A/cm2, applied fields for density currents of
TABLE IV
CURRENT
DENSITIES
AND CORRESPONDING
FIELDSFOR VARIOUS
WORKFUNCTIONS
j , = lo6 A/cm2 j , = lo7 Alcm2
+=2eV Fo = 1.5 x lo7 V/cm F, = 1.8 x lo7 Vlcm

u(w) = 0.3964 u(w) = 0.2933
f ( w ) = 1.0751 t(w) = 1.0849
+=3eV Fo = 2.9 x lo7 V/cm Fo = 3.5 x lo7 V/cm
u(w) = 0.4791 u ( w ) = 0.3890
r(w) = 10668 t ( w ) = 1.0758
4=5eV Fo = 6.3 x lo7 V/cm F, = 7.6 x lo7 Vlcm
u ( w ) = 0.5749 u(w) = 0.5013
t ( w ) = 1.0568 t ( w ) = 1.0646
106-107 A/cm2 correspond to a transmission probability of the deformed

potential barrier of the order of 10-6-10-5, respectively.
3 . I-V Characteristics: The Fowler-Nordheirn Equation

In experimental field emission, the current I is measured as a function of
the potential difference V between the tip and the screen. These quantities
are related toj, and Fo as
V
I = J,A and F o = ~ = p V
where A is the emitting surface area, K and p are geometric factors

determined by the local geometry of the electron emitter, and r is the tip
radius. Equation (13) becomes
4 2v2A
t (w) [
I = 1.55 x 1 0 - 6 q e x p -0.685 -
43'2 v(w)
pv
]. (14b)
with p in AT', A in A2, and I in A.

The curve obtained by plotting ln(IlV2)versus 1 / V is called the Fowler-
Nordheim plot. It is practically a straight line whose slope is a function
of 4 and p. This behavior is observed experimentally for hemispherical
and buildup tips (see Section 111). F-N plots are used in such cases to
determine experimentally the tip parameters (4, p, and A) or to follow
in-situ tip sharpening due to a variation of 0.
At temperatures above 0 K, the electron emission from the thermal tail
near EF cannot be ignored. Within the low-temperature approximation,
(i.e., T I 1700 K for W),
which implies a negligible thermal tail in the electron distribution at V =

EF + 4, the current is [12]
I - I,
-z (1.28 X 10s)t2(w)+
I0
where I, and I are, respectively, the currents at temperatures of 0 (K)

and T (K). The variation affects the preexponential term of the F-N equa-
tion and shows that the increase in the current is proportional to T 2 . It is
of the order of 5% when the temperature increases from liquid nitrogen
temperature (-78 K) to room temperature (RT = 300 K). However, the
temperature dependence does not alter the linear variation of the F-N plot.
C . Energy Distribufion of Emitted Electrons
Let us now discuss the energy distribution of the emitted currents for
both thermionic and field emission. The width of the distribution is one
of the main parameters of importance in the use of the electron beam
in microscopy.
1. Thermionic Emission
At high temperature the thernal tail of the Fermi-Dirac distribution,
f ( E ) , becomes
As the electrons are emitted in all directions within the half-space, the
normalized total energy distribution, F,(E), is given by
F,(E) dE = -exp
(kB n2 (-")kB T dE.
The half-width (FHWM) of this distribution is BE = 2.45kBT with a

mean energy (E) = 2kBT and the maximum in the distribution occurs at
Em,, = kBT with respect to the vacuum level. For a cathode temperature
of -2800 K, this gives an energy spread of around 0.6 eV. The widths
of the experimentally measured distributions are generally much higher
values, of about 2 eV. The difference between these two values is due to
additional mechanisms and the experimental setups used. They are mainly:
The roughness of the cathode emitting area

The voltage drop across the emitting area of the cathode when it is
heated by electrical resistivity
The Boersch or space charge effect [13], which is the result of Coulomb
interactions among the electrons inside the emitted e beam
The stability of the high-voltage power supply, in particular when
very high voltages are used
Thermionic emission is then characterized by a wide energy distribution.
This is a severe handicap for some applications, and this is one of the
reasons that field emission sources, which have narrower energy distribu-
tions. have come into use.
2. Field Emission
The energy dependence of the electron density emitted in the field emission
process, J ( E ) , is described by the total energy distribution (TED), origi-
nally derived by Young for a free electron gas [14]. It turns out to depend
simply on the product of a transmission probability factor and the Fermi-
Dirac distribution function:
with
[
and
B = 1.58 x 10" exp
-6.85 X
1
:7~(~)43n
1 t(w)4'/2
-z= 1.025-.
d Fo
The maximum in the energy distribution relative to EF occurs for
and the half-width at T = 0 K is given by

AE(0) = d In 2
For T # 0 K, the expression for AE becomes too complex to be useful.
Representative values of the current densities and the peak positions and
TABLE V
CURRENT
DENSITIES
(INA/cm2) AND TED PEAKPOSITIONS
AND FWHMs (INeV) FOR
VARIOUSFIELDSAND TEMPERATURES (a = 4.5 eV)
Fo (V/cm) 107 3 x 107 s x 107 8 x lo7 I 0'
77 K
J 1.65 X lo-'' 4.65 X 10' 5.27 X lo5 1.26 X 10' 8.44 X lo8
Emx 0.0114 -0.0 195 -0.0230 -0.0262 -0.0280
AE 0.0612 0. I38 0.212 0.323 0.396
300 K
J 2.56 X lo-" 4.86 X 10' 5.35 X lo5 1.27 X 10' 8.47 X 10'
Emm 0.00942 -0.0364 -0.0520 -0.0654 -0.0716
AE 0.136 0.203 0.281 0.396 0.472
1000 K
J - 8.43 x 10' 6.41 x 10' 1.36 x 10' 8.85 x lo8
Emax - 0.0555 -0.0373 -0.0966 -0.121
AE - 0.458 0.487 0.593 0.669
1500 K
J - - 8.43 x lo5 1.50 X 10' 9.41 x lo8
Emax - - 0.0471 -0.0694 -0.113
AE - - 0.680 0.746 0.816
widths are tabulated in Table V. Note that the predicted FWHMs are
-0.3 eV, which is generally in agreement with experiment.
A graphical representation of Eq. (18) is given in Fig. 3. Several charac-
teristics of the TED to note are as follows.
I . The high-energy slope (a) is mostly temperature-dependent.
2. The low-energy slope (b) is mostly field-dependent.
3. At a temperature T* = d/2kB,the average number of FE electrons
under E F is equal to those coming from over the Fermi level and
Em, = EF. The temperature T* is called the inversion temperature.
For T < T * , most of the field emitted electrons are under EF and
Em,,< EF. Conversely, for T > T * ,there are more electrons emitted
with energy higher than E F , and the maximum in the energy distribu-
tion is over the Fermi level.
4. For useful current densities (>lo5 A/cm-*), the width of the energy
distribution has a lower limit of -0.2 eV at 77 K and -0.3 eV at 300 K.
D . Current Density Distribution
An important parameter for the use of emitted electron beams is the

current density in the beam. This has two aspects: the current density at
Energy Relative to EF (eV)

FIGURE3. Plot of the theoretical TED from Eq. (18) for 4 = 4.5 eV, Fo = 0.5 V/A,
and T = 300 K.
the emitting surface and the current density in the beam at some distance
from the emitter, which has been influenced by the local field of the
whole tip.
1 . Thermionic Emission
As most metals melt before they reach a sufficiently high temperature to
obtain thermionic emission, the most widely used thermionic cathode
consists of a W wire, 100-200 pm in diameter, bent like a hairpin. Only
the bent tip of the filament contributes to the emission. The emission area
is in the range of 10-'-10-* mm2. In order to reduce the emitting area,
sharpened filaments are used either by direct electropolishing of the hairpin
wire or by soldering to a heating wire a small electropolished tip with
small radius of curvature at the apex, but the size of the emission source
remains much larger and the current density much lower than FE sources.
2. Field Emission
The density distribution inside the field emitted beam is determined by
the field distribution over the emitting area, which means the tip apex.
To obtain a high electric field F, at the emitter apex, we use the property
that F, near a charged conductor is inversely proportional to the radius
of curvature r of its surface [15].
In practice, p has to be calculated by taking in account the exact geometry

of the blunt tip after each thermal treatment [ 161. However, an estimation
of the electric field at the apex of the tip can be made, within an accuracy
of a factor of 2, by using either the hyperboloidal approximation [ 171
2v
F -
- r ln(4Dlr) '
or the paraboloidal approximation [ 181

2v
F, =
r ln(2Dlr) *
These equations are valid for r -e D,where r is the tip radius and D is
the cathode-anode spacing. It is then easy to estimate that to have field
emission (0.3 < F, < 1 V/&, a voltage of few lo3 V is enough if the tip
radius is of the order of a few tenths of a pm, for cathode-anode distances
in the order of cm.
To estimate the variation of the current density over the tip, it is neces-
sary to determine how p vanes as a function of angle from the tip apex.
As the tip is usually needle-shaped, it can be usually modeled by a cone
with a hemispherical tip end of radius r, as can be seen by the image of
the simulated tip in Fig. 4 [16]. P(d)/P,, where Po is the apex value, for a
similar tip has been given in ref. 15 and is reproduced in Fig. 5 . The field
variation over the emitting area at the tip end induces a varying current
density distribution, J ( d ) / J ( O ) ,controlled by Eqs. (13)-( 14), and is also
piotted in Fig. 5 for a constant work function and a field of 0.5 V/A
at the apex. It shows that the FE e-beam density distribution is roughly
a Gaussian shape with a total opening angle OC of -200" for F, = 0.5
VIA, which increases to -240" when the field increases to 0.7 V/A. The
figure demonstrates that the e-beam source size is controlled principally
by the tip geometry. Superimposed on this Gaussian distribution, the
current density variation is also affected by the modification of the work
function over the emitting area. For simplicity, however, this variation,
which is dependent on each specific crystallographic and adsorption state
of the tip end, is not considered here.
The second effect of the emitter shank on the density distribution is
the compression of the lines of force toward the tip axis, which means
+- -2000 A 4
FIGURE4. 3D geometry of a FE tip. The shape is the result of a numerical simulation
of the morphological changes by surface diffusion for a tip with a cone angle of 14" [16].
I I I
- 0.8
a 0.4 -
0.2 -
0 45 90 135 180
Angle from apex (deg)
FIGURE5. Variation of p(O)lp(O)(from ref. [IS]) and of the current density, J(O)/J(O),
given by Eqs. (13)-(14) away from the apex of a hemispherical FE tip.
71
FIGURE6. Schematic representation for the virtual radial projection point source V tor
a microtip relative to the surface apex and its geometric center C .
that the electron trajectories are not radial. Calculations of the electron
trajectories [19] have shown that the full beam opening, BC, decreases to
0" with a ratio of
01,
0.5 I- < 0.7.
0,
Actually, this means that the FE e beam is radially emitted from a virtual
radial projection point source, V, which is the intersection point of all the
asymptotes of the electron trajectories far away from the tip (Fig. 6), with
virtual source size roughly of
A = 2.rrr2(1- cos 0,-).
The virtual point V is situated on the tip axis and is shifted behind the
hemispherical apex geometric enter C , by a distance of at least the value
of the radius of the tip. This compression phenomenon induced by the
tip shank can also be treated as the expression of a refractive index, in
electron optics terms, with the tip playing the role of an intrinsic electron
lens [20]. The effective beam opening angle from a hemispherical tip is
then in the range of 45 to 80".
E . Current Stability
For both thermionic and field emission sources, the reproducibility and
stability of the emission current are determined primarily by the reproduc-
ibility and stability of the cathode work function. This can be seen from
Eqs. (9) and (13), which show that the thermionic and FE current densities
are exponentially dependent on the work function.
1 . Thermionic Emission
For thermionic emitters, various low-work-function surface treatments
are often employed because of the greatly increased currents-for ex-
ample, by depositing on the W surface (+ = 4.5 eV) either a layer of
LaB, ( n = 4 , 6 , 9 ) to lower the work function to 2.52 to 3.35 eV depending
on the boron concentration [21], or a layer of ZrO to reach a value of
+ = 2.8 eV [22]. The decreases in the work function arise from the presence
of adsorbed surface double layers. In these cases a dipole moment pin,,
can be associated with each adsorbate atom. The corresponding change
of the work function due to the adsorbed layer is given, in the first approxi-
mation, by
A$ = 2TpindNaea (25)
where N , is the maximum number of adsorption sites per unit area, 8, is
the fraction of occupied sites, and pin,,is the adsorbed atom moment. This
equation implies a linear relation between A 4 and the value Ba, and pin‘,,
i.e., changes in the chemical composition of the first monolayer of the
emitter surface. The stability of the emitted current is therefore controlled
by the stability of the adsorbed layer. This is a very demanding prerequisite
for hot cathodes working inside an electron gun environment and espe-
cially when the largest possible current density is drawn. Holding the
cathode work function to a constant value is thus a very complex techno-
logical problem which, added to the difficulty of fabricating homogeneous
emitters, has impeded the extensive use of such techniques, as the use
of single crystal LaB6 cathodes, for example [23].
2. Field Emission
FE currents depend exponentially on 43‘2 [Eq. (14)], and thus the reproduc-
ibility and stability of the FE current are strongly influenced by adsorption
during operation of the emitter. This is the main cause for the regular
“regeneration” of the tips in actual FE guns.
Figures 7 and 8 show, as illustrative examples, the variations for hemi-
spherical and buildup tips due to the adsorption of gases on the surface
from the UHV environment. There is a rapid regular decrease in the FE
currents until a few percent of the initial values during the first 10 to 20
min as a consequence of the formation of adsorbed layers. The actual
duration is a function of the surrounding working pressure and is traced
by a smooth continuous variation of the FEM pattern. The following
I I
W<111>Hemispherical Microtip
-
40 -~-~ -
3
-
20
Time (min)
FIGURE7. Total FE currents in UHV for a hemispherical microtip at fixed applied
voltage measured as a function of time from a flash cleaning.
increase of the current is due to the formation of multiple localized emitting

areas coming probably from the field-induced formation of small protru-
sions and local changes in the work function of the adsorbed layer. The
origin can be either the surface diffusion of the adsorbed atoms under
field gradient and/or ion bombardment. The long-term behavior is then
unpredictable. It leads to the appearance of erratic local high emitting
zones with subsequent destruction of the tip. In order to avoid the adsorp-
L ' ' I
0.3
* 0.2
8
U
9
0.1
0.0
0
1 2 4 6 8 10 12 14 16
Time (rnin)
FIGURE 8. Total FE currents in UHV for a buildup microtip at fixed applied voltage
measured as a function of time from a flash cleaning.

FIGURE9. Measured TEDs from a W microtip at 300 K (FWHM = 0.25 eV) and at
1400 K (FWHM’ = 0.58 eV), showing the large increase in energy spread with temperature.
tion process, some FEGs have a working temperature in the range of

1000 K. The tip is quite insensitive to contamination when operating
at high temperatures. Continuous FE of several hours is then possible.
However, the results of this thermal treatment are an increase in the
energy dispersion as shown in Fig. 9 and a geometric instability due to
surface diffusion.
EMISSION
111. ELECTRON FROM NANOTIPS
Passing from thermionic to field emission cathodes principally allows re-

duction of the emitting area and the energy dispersion of the e beam. This
section summarizes the further appreciable improvements over normal
microscopic field emission cathodes that can be made by using nanotips as
field emission sources. These nanotips consist of 2- to 5-nm-high pyramidal
nanoprotrusions of single-atom sharpness on top of hemispherical base
tips [2, 31. Due to the atomic size of the emitting area, the field emission
characteristics present very specific measured properties compared to the
conventional field emission behavior presented above, some of which
have not yet been explained. Table VI summarizes the main differences
between field emission characteristics from microscopic tips and nanotips.
This comparative table shows clearly that nanotips have most of the
qualities that can lead to a serious improvement in the FEG. These experi-
mental characteristics as well as the physics of the field emission from
82 V U THIEN BINH ET AL.
TABLE VI
COMPARISON OF ELECTRON
SOURCE PROPERTIES OF MICROSCOPIC
TIPS A N D NANOTlPS
Microscopic tips Nanotips
Emitting area, A 2nrz ( 1 - cos 0,) = 7rr? Apex atom

with r 2 2.5 nm
Beam opening, 0" 45-80" 4-6"
Stability Minutes/regular decrease Hourddiscrete jumps
I-V characteristics F-N straight line Current saturation
TED From conduction band + From localized band(s) -+
a peak at EF localized peak(s) and shift
Energy dispersion, AE AE 2 0.3 eV increasing AE 2 0.06 eV peak shifting
with T and Fo with F,,
one atom will be discussed in order to explain these specific properties.

The discussion is organized into four main parts:
II1.A. A description of the experimental system used for these studies
which includes field electron microscopies (FEM), field electron emis-
sion spectroscopy (FEES), and field ion microscopy (FIM) in the
same chamber.
1II.B. A comparison of nanotips with other tips that also exhibit a
confinement of the field emission area.
1II.C. An overview of the experimental characteristics of the field emit-
ted beams from nanotips, pointing out the specific properties that are
attached to the atomic size of the emitting area.
I11.D. Discussions about the physics of the observed emission properties
taking into account the atomic nature of a nanotip.
A . ExperitnPntal Setup and Procedures
Most of the experimental results that are presented in this chapter were
obtained with the experimental installation whose schematic diagram is
shown in Fig. 10. This installation includes, in the same ultrahigh-vacuum
chamber, the possibilities of in siru tip treatments, field electron emission
microscopy (FEM), field electron emission spectroscopy (FEES), and
field ion microscopy (FIM). The tip mounting includes both a mechanical
movement and electrostatic and magnetic deflection systems, which allow
transfer among FEM, FEES, and FIM measurements at will and within
the same environment. The whole device is inside a chamber which has
FIGURE 15. 3D calculated field distribution with atomic resolution over a n equipotential m-face for a nanotip (p,) = 4 n m and base diameter =
4 nm). ( a ) A complete view of the nanoprotrusion o n top of the 50-nm-radius base tip. (b) A close view of the nanoprotrusion apex to show the local
field enhancement over the topmost atom. The color x a l e represents the I-ange of variation of the fi f x t o r . which is between lo4 and 10' cm-'.
(From ref. [34].)
F I G U R E 21. A M l E spots observed o n the screen for W a n d Au nanotips. These metallic ion beams come
from the ionization of the fast diffusing atoms toward the apex of the protrusions. After quenching. E'EM
and FIM show the superposition of the three emitting spots due to gas imaging ions ( F I M ) . electrons ( F E M ) .
and metallic ions (AMIE). This is shown o n the right-side image obtained with a W nanotip.
Fluorescent scr
Electrostatic le
FIGURE10. Experimental setup used for studying the fabrication and emission proper-
ties of nanotips. The UHV system contains FEM, FEES, and FIM facilities in the same
chamber.
a base vacuum of -5 X lo-'' torr and has a controlled gas introduction

system.
The tips used in these studies were electrochemically etched [24] from
Pt and Au polycrystalline wires, Fe single-crystal whiskers, and W( 111)
single-crystal wire. The etched tips were spot-welded onto W loops to
allow the control of the tip temperature by joule heating and cooling with
liquid nitrogen. The temperatures were determined by a combination of
optical micropyrometer measurements on the conical tip shank, the heat-
ing loop resistivity values, and by fitting the experimental TED spectra.
The controlled temperature range available was 80-3500 K.
For FEES, a fluorescent screen was placed at 2.5 cm from the tip, with
a I-mm-diameter probe hole in its center. Any region of the FEM pattern
could be studied by the electron energy analyzer by using the tip displace-
ment movement and visual control of the pattern on the screen. The TEDs
were measured with a commercial 135" hemispherical energy analyzer

with nominal resolution of 10 meV, positioned behind the probe hole, in
which the entrance lenses had been adapted for the FEES measurements.
The tip mounting and deflection systems allowed the choice of the local
zone of the apex region of the tip to be analyzed concomitant to the
alignment of the e beam to the analyzer axis.
FEM and FIM observations were done with the standard technique,
i.e., a microchannel plate (MCP) in front of a fluorescent screen located
5 cm away from the tip. FEM and FIM patterns were followed by a video
camera connected to a tap recorder and a numerical image treatment
system.
B . Confinement of the Field Emitting Area
The first problem to be confronted in improving tip performance is the

reduction of the field emission area. As the FE area is governed principally
by the tip geometry and in particular by the apex structure and composi-
tion, three directions can be foreseen for narrowing the FE area at the
apex as depicted in Fig. 11:
1. To decrease the whole tip radius, i.e., to produce ultrasharp tips
2. To confine the emission over a small area by modifying the atomic
structure and/or the work function
3. To confine the field over a small protruding zone, i.e., to fabricate
buildup tips and nanotips
FIGURE1 1 . Schematic diagram showing the three possibilities for FE tips to exhibit
confinement of the emission area to nanometer dimensions.
a b
FIGURE12. FIM images of an electrochemically etched ultrasharp tip with an estimated
radius of about 2 nm. (a) The best image voltage (BIV) is on the apex three-atom facet ( I 11).
(b) The BIV is on the zones underneath, with consequently a loss in the resolution of the
ending trimer, in order to show the structure underneath the hemispherical tip end.
The results of the studies of these mechanisms presented in the following

section clearly show the advantages in the use of the nanotips for confining
the whole emission area over the last apex atom of the nanometric pro-
trusion.
1 . Ultrasharp Tips
Ultrasharp tips are tips with ending radii of about a few nanometers. The
field emission confinement is then simply a result of the reduction of the
high field apex area. The fabrication techniques to decrease the tip radius
to a few nm can be either an ex situ electrochemical etching technique or
an in situ mechanism using ion bombardment.
The electrochemical tip etching technique [7, 81 can be controlled to
produce very sharp tips [24]. In Fig. 12 we show the FIM image of a
W(111) tip with a radius of about 20-25 A.
It was obtained after an
electrochemical etching in NaOH (2N) with a controlled pulsed AC cur-
rent, followed by a very gentle field evaporation of the first adsorbed layer
after the introduction into the vacuum. One can notice, from this example,
that the (1 11) plane at the tip apex for this tip is a three-atom plane.
Another possibility is the in situ etching of initial tips having radius of
-100 nm by ion bombardment during FE under a pressure of to
ton of Ar or Ne [25]. This technique can be pushed toward the obtention
of ultrasharp tips with a radius in the range of 10 nm (Fig. 13).
80 I I I I
70 -
Sputter Voltage 500 V
-
W
m Sputter Current lOpA
2 60
3G
' 50 -
g 40 -
3rd 30 - -
3w" 20 - 0
-
10 1 I I I I I
0 10 20 30 40 50
Sputter Time (min)
FIGURE13. Evolution of the tip radius with argon sputtering time. The inset is the
FEM image at the end of the sputtering cycle. The radii were estimated from the voltages
needed to have a fixed FE current of 1 X lo-'* A.
The use of these ultrasharp tips for FE applications gives rise to the
following comments :
1. The production of such a tip, either by electrochemical etching or
in situ sputtering, needs a priori a tip radius control by FIM, which
is not very convenient for FE gun settings.
2. With the electrochemical fabrication, the tip must be etched as
shortly as possible before its introduction into vacuum due to possible
tip evolution and blunting by corrosion outside the vacuum chamber,
especially if one wants to keep the value of the radius in the range
of several nanometers.
3. The source size for electron emission is roughly determined by the
radius of the tip, which is still at best -20 X 20 atoms.
4. The beam opening angles are those of a hemispherical cap ending
tip that is in the range of 45 to 80" as discussed in Section II.D.2.
Therefore, a diaphragm is needed to collimate the beam, which re-
sults in a relatively low current density available for use in the fi-
nal beam.
5 . A very unstable emission if the ultrasharp electrochemical etched
tip is used. To keep this nanometric radius, the initial tip cannot
be thermally cleaned, so the adsorbed layer on the shank diffuses

instantaneously to the apex when the field is cut and is the cause of
the nonreproducibility and large fluctuations during field emission.
2. Local Work Function Decrease

The second technique is based on a significant lowering of the local work
function by an external deposition of appropriate foreign atoms andlor
reorganization of the surface atomic structure [4, 26-29]. Products such
as copper phtalocyanine, Ba, Cs, or Zr compounds, for example, are used
in order to enhance the FE over single adsorbed molecules or atoms. This
technique has been discussed in detail by these authors and presents the
following characteristics:
1. The stability of the adsorbate, especially under field emission (see
below), due to the energy transfer between the emitted electron and
the substrate (Nottingham effect). This unstability becomes critical
in particular for individual atom or molecule deposition at a selected
site for a controlled local enhancement [28, 301.
2. The local field enhancement may be created at one atom or molecule
[4,29], but it does not cancel the field emission from the surrounding
regions. This means that the emitted electron beam contains simulta-
neously the background FEM pattern superimposed on the field
emission spots of the adsorbed particles. This behavior is clearly
visible from the FEM patterns [41.
For better control of the local modification at the apex, the combined
action of work function decrease with tip geometry gives a more localized
FE area. This is the technique utilized to obtain ZrOlW tips [22,31], used
in Schottky emission guns.
3 . In Situ Field Sharpening

A procedure for fabricating in situ tips that have localized FE over atomic-
size areas is the thermal field shaping method, i.e., using the diffusion at
high temperature of surface atoms in the presence of large electric field
gradients. Two cases have to be considered:
1. When the applied electric field F is in the range of few 0.1 V/& then
buildup tips are obtained [32].
2. For F > 1 V/& a field surface melting is produced and leads to the
fashioning of nanometric protrusions on the top of the base tips [2,
31. Due to their specific protruding geometry, these cathodes were
named teton tips or nanotips.
b
FIGURE 14. Equipotential lines for a hyperbolic tip (radius = 50 wand applied voltage =
175 V) and a point charge at 2 A from the apex (from ref. [33b]). (b) Equipotential lines
corresponding the superposition of the potentials given in (a). The dashed region corresponds
to the tunneling region. (c) The same as in (b) including the image force correction. In this
case, the tunnel bamer is lower and the equipotentials near the protrusion are almost flat.
The basic mechanism for the protrusion formation will be considered

in the following paragraphs. The first step is to analyze the enhancement
of the local field with the protrusion geometry.
a . Local Field Enhancement. The protrusion technique is based on
the property of local field enhancement over nanoprotrusions [ 191 leading
to the confinement of FE to their apexes. The presence of the nanoprotru-
sions distorts and compresses the equipotentials in their vicinity. To esti-
mate the field enhancement, let us first consider an analytical approach
that uses the superposition properties [19, 33a] of a point charge or a
dipole on top of a microscopic tip whose potential distribution is described
by the hyperboloid screen geometry. The resulting equipotential line for
V = V,, will then define the whole tip, including the protrusion at the
apex (Fig. 14). The field distribution along the tip axis in the presence of
such a protrusion was shown to be
where po is the height of the protrusion, z is the distance from the micro-
scopic base-tip apex, and F, is the field at the apex without the protrusion.
The field at the top of the protrusion (z = p,,) is -3 times that of the
substrate in its vicinity; this value does not depend on the protrusion
height in this model.
Numerical calculations based on the superposition principle give a more
precise 3D potential distribution at the atomic scale of the whole tip,
which consists of a base tip with a nanoprotrusion at the apex [34]. The
base tip is also described by a hyperboloid function, but the protrusion
is now modeled on the atomic scale by a cluster of electrostatic charges
placed at the center of the spheres at the atomic sites that shape the
protrusion. The value of each of the charges is given by minimizing the
electrostatic energy of the complex capacitor consisting on one hand of
the hyberboloid and the cluster of spheres, all of them at the same potential
Vtip,and on the other hand of a plane orthogonal to the tip axis, describing
the screen, located a few centimeters away at another fixed potential. The
3D potential distribution is then calculated for a given voltage between
this tip and the screen by summation of the different potentials created
by the hyberboloid plus all the charges. The field is derived afterward from
the calcualted potential distribution. This method, assuming no symmetry
except for the base tip, allows calculations of the potential and field distri-
butions in 3D with atomic resoluton for any protrusion shape.
Figure 15 (see color plates following page 82) is an example of such a
calculation of the 3D field distribution over a conical protrusion with
( I 1 I) axis, base radius 2 nm and height 4 nm, placed on top of a 50-nm-
radius base tip. The variation of the parameter p of Eq. (21), which is
equivalent to the field distribution, is plotted in Fig. 16 for a cross section
through protrusions with different heights and a fixed cone angle of 53"
(Fig. 16a), and for different heights and cone angles (Fig. 16b). For small
protrusions a value of 3.2 for the enhancement factor is found forp, = 1 nm.
The numerical results confirm the estimated value of the field enhancement
given by Eq. (26), (see Fig. 16a). This /3 value is practically constant for
po < 1 nm, which emphasizes the role of the electrostatic screening from
the base tip.
A second result must be noted. For a given geometry of the protrusion
end (i.e., a constant cone angle of -53O) and for p, 2 2 nm, the field at
the apex is -9.5 times that of the substrate in its vicinity. This enhance-
ment factor is much larger than the value of 3 estimated by the analytical
approach [Eq. (26)]. As the FE is an exponential function of PV, protru-
sions with po 2 2 nm result in a confinement of the emission area exclu-
sively over the top of the protrusions. This is the essential reason for the
choice of nanotips as advanced FE sources.
6 . Tip Sharpening in the Presence of Applied Field. In this approach,

the role of the applied field is to induce a gradient in order to define a
01
-4 -3 -2 -1 0 1 2 3 4
Distance from Apex (nm)
a
I I I I I I I
-4 -3 -2 -1 0 1 2 3 4
b
FIGURE16. (a) Field distribution over the apex of nanotips with conical protrusions of
different heights and a fixed cone angle of 53". (b) Field distributionover the apex of nanotips
with conical protrusions of different heights and a fixed base radius of 2 nm. (From ref. [34].)
direction of surface diffusion. As this driving force is effective only if it

is applied to diffusing atoms, the protrusion formation has to be performed
at temperatures high enough to create mobile atoms on the metal surface.
Under the conditions of elevated temperature and field, the surface atoms
will migrate from low-electric-field regions toward higher-field regions.
The final geometry of the tip apex is then governed by the equilibrium
between two opposing diffusion processes, the first driven by the gradient
of the electric field, and the second driven by the capillary forces (gradient
of the surface chemical potential) [35].
Depending on the value of the applied electric field, two different tip
end geometries can be obtained: buildup tips for F around 0.5 V/A and
nanotips for F larger than 1 VIA. the specific properties attached to each
of these two profiles are described next.
Buildup Tips. Consider the case when the applied field is around 0.5
V/& this procedure is termed the buildup technique [321. The applied field
induces a gradient across some low-index facets with, as consequence, the
enlargement of these facets. It can be performed either with positive or
C d
FIGURE17. FEM of a buildup sequence. (a)-(d) Evolution of the FEM pattern from a
hemispherical W(11 I ) tip (a) to a buildup tip (d), due to the enlargement of the three facets
{ 112) under temperature and field. All the patterns are at the same scale in order to show
the confinement to one spot of the emission area during the buildup.
negative polarity. With negative polarity one can follow in time the varia-
tion of the apex geometry by FEM because the field value is in the range
of FE. For W(111) tips, the facets which are enlarged are the three (112)
planes around the tip axis. Taking the facet enlargement to its limit ends
in the intersection of the three facets with a comer at the (1 11) apex (Fig.
17). This corner can end in one or three atoms and becomes a small
triangular facet after a controlled field evaporation.
This local region of high curvature creates a predictable local field
enhancement [19]. The calculation of the field distribution at atomic scale
gives an enhancement factor in the range of 1.4 compared to the surround-
ing field, as shown in Fig. 18 [34]. This enhancement factor is small because
the angle between (112) and (111) is only -20". It is enough to allow
preferential FE over the protruding apex, but without being exclusive as
indicated by FEM patterns and local current measurements over the tip
end cap. In Fig. 19 is shown the same FEM pattern but with increasing
FE voltages and MCP gain, which clearly illustrates the apparent con-
finement is partially an artefact due to signal detection sensitivity.
Thus, for buildup tips the FE current is not confined to the apex atoms.
The buildup only enhances preferentially the FE over the intersection
corners of some facets by a ratio of ZlZ, = 15, without being exclusive.
Derived methods to increase the local angular beam confinement by
using concomitantly the buildup and the selective work function reduction
(with oxygen processing or ZrO coating, for example) [36] are now cur-
rently used in commercial FE guns.
I I I I I I I
-4 -3 -2 -1 0 1 2 3 4
FIGURE18. Field distribution over the apex of a ( 1 1 1 ) buildup tip (dashed line) and a
nanotip (height po = 2 nm and cap diameter 4 nm). (From ref. [34].)
a b C
FIGURE19. (a)-(c) Comparison of the FEM pattern of a buildup tip (a) with the corre-
sponding FIM pattern (b). It shows the one-atom boundary between the {112} facets and
the three-atom comer forming the apex of the tip. An increase of the FE voltage and the
MCP gain shows that the emission area is not confined to the topmost atoms, as revealed
by the FEM pattern (c) of the same buildup tip shown in (a).
Teton Tips or Nanotips [2,31. To obtain exclusive FE from the protru-

sion apex, the calculations indicated that a minimum protrusion height of
about 2 nm is necessary. A very high mobility of the atoms is needed to
obtain such a protrusion height, as for example in the Taylor cone forma-
tion with liquid layers [37]. However, in this latter case the apex of the
protrusion is in the micrometer range. To obtain a very sharp apex, the
protrusion formation by surface melting mechanism has been introduced
[2, 31. Under these conditions, the surface atoms are very mobile but the
underneath protrusion is still solid, and this is the main difference from
the classic Taylor cone formation. It is the very high mobility of the
surface atoms driven by the field gradient over the solid substrate that
leads to the formation of nanoprotrusions ending with atomic sharpness.
This process is detailed in the following paragraphs.
Field surface melting mechanism. In order to increase the mobility of
only the surface atoms, they must be under an action which lowers their
activation energy but which does not affect their underneath neighbor
atoms. This is what happens when a large electric field is applied to a
metal surface. For a flat surface, the effect of the field on the reduction
of the activation barrier for surface diffusion is negligible even at very
large applied field. The reason is that the dipole induced by the applied
field is small. However, if the surface is rough-with adatoms, vacancies,
kinks, steps, etc., as due to thermal treatment-the values of the perma-
nent dipole moments are different at each point. This difference is in-
creased by the spreading out of the surface charge [38]. The action of the
field is then enhanced on the protruding parts of the surface.
The estimation of surface diffusion in the presence of an applied field
can be made by considering the activation energy for surface diffusion in

the presence of a field, Q ( F ) [81:
Q ( F ) = Q, - f a F 2- p F (27)
where Q, is the activation diffusion barrier at zero field [39], and a and
pare the atomic polarizability and the permanent dipole moments, respect-
ably. The surface diffusion coefficient in the presence of F is given by
where a is the jumping distance of diffusing atoms taken to be the unit

cell -3 A and v, is the attempt frequency (10’2-1013s-I).
For a field value of 2.55 V/A, which is approximately the value used
for the fashioning of W nanotips, D , = 3 x cm2/s at 1200-1500 K.
One can also estimate D , from the atom flux supply needed to obtain the
-
experimental atomic metallic ion emission (AMIE) beam of lo6 ions/s
(see below). The value obtained is also in the range cm2/s. As the
criterion for surface melting is a diffusion coefficient larger than 2 X 1O-j
cm2/s [40], the surface in the presence of very high field is then melted
locally at about one-third of the bulk melting temperature.
Growth and formation of nanoprotrusions. The high diffusivity facili-
tates an increase of the height of some existing protrusions due to the
field gradient driving force over the thermally induced corrugations leading
to the formation of nanoprotrusions. The geometry of the formed nanopro-
trusions is determined by the equilibrium between the pulling-up by the
electric field gradient force and the blunting due to the capillary force. A
schematic drawing of this mechanism is given in Fig. 20. When the field
enhancement over the apex of these protrusions is high enough, i.e., for
a certain height, the last atom is ionized. This gives rise to a metallic ion
beam which is regulated by the supply of diffusing surface atoms to the
apex under the field gradient. The appearance of such atomic metallic ion
emission (AMIE) is detected by the presence of a spot on the screen
placed in front of the tip (Fig. 21, see color plates following page 82).
By adjusting the two parameters, F and T , the high protrusion formed
during AMIE could end in one atom. Note that the AMIE mechanism
has been experimentally observed for W, Pt, Au, and Fe and therefore it
can be used for all metallic emitters.
Of crucial importance is that the high protrusion geometry remains
intact upon quenching. After cooling, the resulting protrusions are gener-
ally ( I 11) pyramids of 2 to 3 nm dimensions (for the base and the height)
ending in one atom. The FIM analysis of such a nanoprotrusion is pre-
sented in Fig. 22(I). The sequence shows the FIM of the apex atom and
the structures underneath obtained by progressive field evaporation.
1 Atomic Metallic Ion Emission

and Nanotip Formation
FIGURE20. Schematic of the field- and temperature-driven formation of nanometric

protrusions on a metal surface. Atomic metallic ion emission (AMIE) from the protrusion
apex occurs under a positive field > 1 V/A and T = one-third the melting point. A rapid
quenching preserves the nanoprotrusion with a one-atom apex.
As mentioned above, 3D calculations of the field distribution for teton

tips showed a field enhancement factor of 7 to 10 over the apex atom
compared to the substrate tip [34], which means that all the FE current
comes exclusively from the single apex atom. FEM observations of the
protrusion tip showed only F E from the protrusion zone, which means a
FE spot of <2 mm on the screen at -5 cm. No other pattern out of this
spot is observed over the whole range of applied voltages for observing
the nanotips [Fig. 22(II)1.
From the above comparisons, the nanotip or teton tip geometry is best
11. (a) 11. (b) 11. (c)

FIGURE22. 1. FIM analysis of a W nanotip on top of a ( 1 11) base tip. Images (a)-(f)
are FIM patterns [Ne is imaging gas except for (e), which is with He] during a progressive
field evaporation. They show the topmost atom (a), then a trimmer (b), a small facet (c),
the crystalline structure of the nanopyramid (d), and the triangular base of the nanoprotrusion
(e). The FIM pattern (f) is obtained with BIV on the base tip and shows the location of the
nanoprotrusion (the blurred central zone) within the ending-cap base tip. 11. FE spot from
the W nanotip at different voltages (a) 525 V, 1.5 x A, (b) 620 V, 2 X A, (c) 700
V, 1.3 X lo-* A. The spot geometry is the same for the three FEM patterns, and this is an
indication of an exclusive FE from the apex of the nanotip.
suited as an electron source in view of the size of the emitting area and
the confinement of the electron beam. The confinement of the FE area
over the apex atom implies specific FE characteristics, which will be
described below.
C . Field Emission Characteristics from Nanotips: Experiment
In this section the measured field emission properties of the nanotips are
presented. The measurements include: (1) the beam opening angle, (2)
the current stability in time, (3) the I-V characteristics, and (4) the total
energy distribution. To clarify the presentation, the experimental measure-
ments presented in this section are separated from their interpretations,
which will be presented in Section 1II.D.
1. Beam Opening Angle

Typical images of the FEM patterns of a nanotip are shown in Fig. 22(II).
The emitted electron beam was self-collimated to opening angles of 4-6"
[2, 31. From a practical point of view, this beam gives rise to a single,
nearly round spot with a diameter in the range of 0.5 cm on a fluorescent
screen situated 5 cm away. The electron current distribution across the
beam was characterized either by a direct measurement with a probe hole
or by an image processing of the intensity of the spot obtained on the
fluorescent screen. In both cases the experimental measurements have to
be deconvoluted by the measurement function (probe hole diameter or
MCP gain and fluorescent screen response) in order to have the actual
distribution. For different extraction voltages it is generally a Gaussian
distribution, as shown by the example in Fig. 23, and the size of the
emission spot does not change dramatically for a large range of FE volt-
5 m
4 m
u 2oooo
1OOOO
0
-10 -5 0 5 10
Beam Angle (deg)
FIGURE23. Measurement of the angular distribution of the FE current of a nanotip by
the probe hole technique.
.o
11.2 ~
7
c
0.8 -
W
= 0.25 %
g 0.6 -
Y
5
U 0.4 -
0.2 - 6.0 65 70 -
0.0
0
' I
2
I I I
4
I
6
I I I
8
Time (hour)
FIGURE24. Total current emitted by a W nanotip as a function of time for current of
about I nA.
ages. This indicates that the total FE'current comes exclusively from the
nanoprotrusion apex even at higher voltages.
2 . Field Emission Stability

A characteristic example of the nanotip FE current stability behavior at
fixed FE voltage (VApp),starting with a clean tip, is shown in the Fig. 24.
From multiple experimental results, three properties can be highlighted:
1. The high stability in time of the current over periods of hours for
currents I1 nA. In the example of Fig. 24, the variation is less than
1% for -10 h of continuous emission.
2. For higher currents, up to 0.1 pA (Fig. 25), we observed reversible
and irreversible jumps of the current between different levels, fol-
lowed by periods of relative stability. These changes have been
shown to be due to current induced local heating [see Section IV.C].
3. For FE current higher than 0.1 PA, the probability to destroy or to
irreversibly change the protrusion is very high, and this constitutes
the third characteristic of FE from nanotips: The useful current
coming from the apex of the nanoprotrusion has an upper limit of
around 0.1 pA.
However, even if the stable working current is -1 nA, the brightness
of the beam is still exceptional if one considers that the whole current is
coming from atom-size source with a 4-6" beam opening.
' I
I I
I1
t
2o
'O
01
1 I I I I I I I
0 10 20 30 40 50 60 70 80
Time (min)
FIGURE25. Total current emitted by a Pt nanotip as a function of time for current in
the 50-nA range.
3 . I-V Characteristics
The variation of the field emission current from a nanotip versus the
applied voltage is plotted as In(ZlV2)versus 1/V. An example of such I-V
characteristics is shown in Fig. 26. The main difference from the Fowler-
Nordheim relation of Eq. (14) is the current saturation at high voltages
[2]. This means that, relative to field emission from a metal surface, there
is a progressively weaker emission current on increasing to higher applied
I I I I I I
- 10-'0 -
z
%
&
-
4
> 10-12 -
-
I I I I
10-l~ ~ I I
6 7 8 9 10'~
1/v (volt-')
FIGURE
26. FE I-V characteristics from a nanotip.
10-14 . I I I
W< 11 I> Build-up Tip

-
n
1 0 - l ~-
*0s
2
W
10-16 - -
J 10-17 -
-
I I I
lo-'*
fields, This current saturation behavior is not observed for buildup tips
(Fig. 27), and is therefore a signature of an exclusive FE from the apex
of the nanotip when localized FE is observed.
4. Total Energy Distribution [41]

The experimental procedure for measurements of the TEDs of nanotips
was as follows; (1) in situ fabrication of a clean microscopic tip of less
than 100-nm radius; (2) FEES measurements of this tip centred on the
(1 11) region; (3) fabrication of a single-atom protrusion on top of this tip;
(4) FEES study from this protrusion; and ( 5 ) destruction of this protrusion
by a controlled heating of the tip and FEES measurements of the resulting
microscopic tip at the (111) region.
The TED spectra recorded for microscopic tips during steps (2) and (5)
had the same shape, were located at the Fermi level (EF),and showed
the well-known behaviors for clean W( 11 1) tips [42]. Two example spectra,
measured after the destruction of the protrusion, for different values of
the applied voltage are presented in Fig. 28. Basically, one strong peak
was observed with a sharp edge at EF for any applied voltage, VApp.An
increase of VApp did not change the position of the Fermi edge, only
causing a broadening of the peak on the low-energy side. To fit the spectra
we have used the classic Eq. (18) for the tunneling current from a free-
electron metal which was developed by Young [ 141. Excellent agreement
between the experimental data and theory was found, as is shown in Fig.
I I I 1 I I I
1 .o
B
-
.-NLd
P) 0.8
0.6
0
v
C
y 0.4
C
s
0
0 0.2
0.0
-1.2 -1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2
Energy Relative To EF (eV)
FIGURE28. TEDs of the W(111)microtip before formation of the nanotip. The fits by
Eq. (18) are represented by the solid curve over the experimental points.
28. The value of EF for microscopic tips was constant and was then taken
as the reference level for all the spectra. The measured FWHMs for the
spectra in Fig. 28 were 0.23 eV and 0.30 eV for 1650 V and 2250 V,
respectively, a variation which is a consequence of the increase of the
slope of the tunnel barrier.
The TED spectra were recorded for different VAppafter formation of
protrusions on top of the macroscopic tip. Four salient features, which are
not present for microscopic tips, can be discerned in the experimental
observations.
1 . The spectra are composed solely of well-separated peaks. To show

clearly the relation between the peaks and the protrusion height, the
evolution of the TEDs versus height protrusion on the same base
tip is plotted in Fig. 29. The TED spectra were recorded for different
stages during the formation of a protrusion, that is, at increasing
height, on top of the microscopic tip. As the height of the protrusion
increased, the contribution from the localized band became predomi-
nant (curve 2+ 3), then exclusive (curve 4). For BU or small-radius
tips, standard TEDs were measured (curve 1). In general, the number
of peaks and their relative intensities depends on the protrusion
geometry and on V,,,. In Figs 30a and 30b, examples of TED spectra
observed from different nanotips are shown with one and two
peaks, respectively.
I I I I I I 1
1.0 -
- ~~
w (111)
0.8 - (1) : Build-up tip
-
E2 0.6 -
(2) - (4) : Nanotips
with increasing
-
v
m
Y
protrusion height
C
0.4 -
U
-
0.2 -
-
-2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5
FIGURE 29. Evolution of the TEDs versus height of nanoprotrusion, showing the ap-
pearance of localized peaks not at EF.
2. The peaks are not pinned at the Fermi level, and a linear shift of the
whole spectrum is observed as a function of VApp.This is illustrated
in Fig. 31, where the dependence of the TEDs with a two-peak
spectrum are given. The positions of the peak maxima as a function
of VAppare plotted in Fig. 32. All the data fall on parallel lines,
*
with slopes of 1.65 0.02 meV per applied volt, showing that the
separation between the peaks remains constant. The total shift of
the peaks for the range of V,,, in this experiment was -0.7 eV. Note
that no shift was detectable for microscopic tips and for similar
changes of VAYp.The shifts and the intensity variations of the peaks
during their shlft were reversible; they could be varied reproducibly
by changing VApp.
3. None of these peaks could be fitted satisfactorily by Eq. (18). For
the same FWHM, the spectra from the microscopic tip have wider
tails on the low-energy sides and sharper maxima than the peaks
from the protrusion.
4. The FWHMs of the peaks vary little with V,,,. For example, for
the TEDs of Fig. 31, the FWHMs remain -0.24 eV for all values
Of vApp*
A direct relationship exists between FEM patterns and FEES spectra.

The presence of adsorption from the background gases on top of the
1.0
F
3 0.8
3
88 0.6
c
v
2m 0.4
1
0
u 0.2
0.0
-2.0 -1.5 -1.0 -0.50 0.0 0.50
b
FIGURE30. Examples of TEDs from high nanotips with (a) one and (b) two peaks not
pinned at EF.
protrusion was easily characterized by large instabilities of the emission

current and strong modifications of the TEDs. The adsorption could be
removed by application of field and temperature, after which the emission
properties were again stable. The FEES spectra could be the same as
before the cleaning or could show changes in the positions, number,
and relative intensities of peaks for the same VApp,probably reflecting a
different structure in the geometry of the protrusion and the nature of the
atoms forming the apex.
1200 I
looo 1 I I
Single Atom Protrusion

I I I
53q+(a!
800
WJ
600
3
400
200
--2.5
n
-2.0 -1.5 -1.0 -0.5 0.0 0.5
Energy Relative to EF(eV)
FIGURE31. Evolution of a two-peak TED from a nanotip as a function of applied
voltage, showing displacement with applied field. (From ref. [41].)
D . Field Emission Characteristics from Nanotips: Discussion
In this section we wish to connect the measured specific FE properties

of the nanotips to the atomic size of the protrusion apex, indicating by
the presented arguments that the nanotips are atom-sources for field emis-
sion of electrons.
We
O
CI
-0.4
u
>
Y
.r(
cd
2 -0.8
d
E
0
?
P
' Z -1.2
a"
%u
L -1.6
900 lo00 1100 1200 1300 1400 1500
v.4pp
FIGURE32. Position of the peaks of Fig. 31. The shift is 1.65 meV per applied volt and
is reversible. (From ref. [411.)
1 . Self-collimation of the e beam to 4-6"

The resulting beam opening that determines the spot size measured at the
projection screen is controlled by two mechanisms: (1) the intrinsic angular
spread 8, of the emission current just after the tunneling barrier at the
apex, and (2) the compression of the electron trajectories due to the
influence of the emitter shank on the potential distribution. The second
effect reduces the initial angular spread by at least a factor of about 2
[Eq. (24)]. This means that the measured angular opening 0, of the e beam
of 4-6" at the screen corresponds to an angular spread 8, at the emitting
apex atom of the order of 8-12".
Two factors are of importance in the determination of the value of 8,:
1 . The geometric effect, i.e., the radius of curvature of the protruding
emitting area
2. The diffraction effect, i.e., the size of the tunneling region restricted
to one atom versus the wavelength of the incoming electrons inside
the tip A, (-4A for EF = 8 eV).
a. Geometric Effect. Consider first the field emission from a smooth
hyperboloidal surface having a small radius of curvature which simulates
the apex of the nanotips. Semiclassical calculations neglecting the electron
Ja.
diffraction at the tunneling opening [43] give a full angular spread 8, of
8, = 4
This relation is obtained by using the WKB approximation for the tunnel-
ing probability T ( E ) ,given by
evaluated at E = EF and the field distribution F(8) over the apex, given by
F(8)=zFo(I -:).
For Fo = 0.5 V/A and 6 = 4.5 eV, Eq. (29) gives an angular spread
8, = 52". To verify the validity of this approach, the problem was solved
using the time-dependent Schrodinger equation (TDSE) of a Gaussian
wave packet moving toward a constriction which represents the tunnel
apex atom. The TDSE was solved numerically by means of an algorithm
based on a fourth-order Trotter formula [44]. It gives a value for 8, = 50".
In conclusion, the geometric factor attached to the nanoprotrusion ge-
ometry gives a reduction of the value' of OC of about a factor of 2. The

resulting beam opening 8" is then of the order of 25" in the case of a
nanoprotrusion, instead of 45-80' for microscopic tips. However, this
does not explain the values of 4-6' for the 8" measured from the nanotips.
b. Diffraction through a Tunnel Barrier. When the tunneling emission
comes from a region which has dimensions of the same order as the
wavelength of the incoming electrons, the diffraction through the tunnel
barrier must be considered. This is the situation of field emission from
the last atom of the nanotips. In the diffraction problem without the
presence of a tunneling barrier, it is possible to estimate the diffraction
by using the Heisenberg uncertainty principle. In the presence of a tunnel
barrier, the diffraction process requires the solution of the TDSE [45]of
the transmission function T,(E). The angular spread BC is then defined as
the angle for which the tunnel intensity J ( 8 ) is lie of the axial intensity
J ( 0 ) . J ( 8 ) is given by
The summation over i runs over all quantized levels.

As a first approximation, each scattered plane wave is considered to
be filtered incoherently by the tunneling barrier. Under such assumptions,
electrons having a total energy of E will have a transmission function Ti
given by
Ti(E)= I Fn(ksin 8)I2T(Ecos' 8) (32)
where k = and Fn(k sin 8) is the Fourier transform of the slit
function which is the diffraction function of the constriction. The calcula-
tions give, for a constriction opening between 8 and 20 A, EF = 8 eV,
n = 1, and a field value of F = 0.5 V/A, an angular spread of 8, 5 20".
This result expresses two properties which are intrinsic to the tunnel-
ing process.
1. Only the electrons within a small energy range AE will contribute to
the current, due to the filtering effect of the triangular tunnel barrier.
2. The transmission probability decays exponentially with the angle.
The angular spread BC can then be estimated from
(h'k';;' 8,) 1
z-
(33)
e
which gives, for small values of angular spread,
ec=2
:/-. (34)
This relation gives, for EF = 8 eV and AE in the range 0.1-0.3 eV,

values of 8, between 15 and 20", which are in agreement with the exact
TDSE calculations.
The values of Bc in the range 15-20" are the consequence of the diffrac-
tion process though an atom-size slit with a triangular tunneling barrier.
Taking into account the geometric effect and the convergent lens effect
of the tip shank, one can expect these values to be divided by a factor of
2 or more. The resulting values for 8, are then in the same range as those
measured experimentally from the nanotips, which is from 4 to 6".
2 . Stability
The long-term stability behavior of FE from nanotips can be assessed by
comparison with hemispherical and buildup tips. For these latter cases
the current stability is explained by considering the adsorption of the
residual gas over the emitting area. As noted in Section II.E, hemispherical
and buildup tips have a limited stability due to the formaton of an adsorbed
layer which varies the work function (see Figs. 7 and 8). The FE stability
of the nanotips presented in Fig. 24 is simply explained if one considers
the very small probability of having an adsorbed atom on the apex atom
coming from the surrounding gas phase (-5. lo-" to lo-'' torr). An estima-
tion of the impinging frequency v can be calculated, just by considering
the gas kinetics equations, which gives v = s-l for a surrounding
pressure of torr, i.e., a time interval of -3 h. This value is in the
range of the experimental measurements.
When the FE current is increased, the temperature at the protrusion
increases due to the Nottingham effect within the localized band structure.
Note that this effect has been used to measure the local energy exchange
during the FE process from a nanotip, which will be presented in the
following section 1V.C. The probability of a rearrangement of the nanotip
apex on an atomic scale increases at higher temperature, or equivalently
higher FE current, leading to abrupt changes in the total current. This is
the cause of the observed reversible and/or irreversible discrete jumps in
the current in Fig. 25. The upper limit for the FE current is also explained
by the very high increase of the temperature during FE, leading to the
destruction of the protrusion by surface diffusion or by local melting.
3. Localized Band Structure

The existence of well-separated peaks in the single-atom TEDs shows
that the electrons do not tunnel directly from the bulk Fermi level to the
vacuum. The peaks and their shifting suggest the presence of a localized
band structure at the tip apex. Furthermore, this idea is supported by the
constant spectra widths and energy gaps observed experimentally. The
. . .
FtGURE 33. Simplified model of FE from single-atom nanotips for two values of VApp.
The lightly and darkly shaded bands signify the position of the band for two different values
of V A p p .
peaks in the TED spectra then occur because the emitted electrons reso-
nantly tunnel to the vacuum only through these bands. This situation is
depicted schematically in Fig. 33 for the case of a one-band TED.
Resonant tunneling through atomic energy levels of adsorbed atoms,
which have been broadened due to interaction with the surface under-
neath, was first introduced by Duke and Alferieff [46] and later developed
more fully by Gadzuk [47, 481. This was used to explain the small bumps
added to the energy distributions of the clean microscopic tips observed
in FEES experiments [27] with chemisorbed atoms on metallic surfaces.
It must be emphasized that the presence of chemisorbed atoms in these
experiments only slightly modified the standard peak of a clean micro-
scopic tip, in contrast to the spectra from the protrusions which consist
solely of well-defined peaks. This latter behavior could have its origin in
the atomic size and shape of single-atom protrusion tips, and in particular
on the reduced coordination number of the atom that constitutes the apex
compared to a single atom on a surface.
The shifts of the peaks run counter to a metallic behavior of the topmost
atom. The linearity of the shifts versus V,, shown in Fig. 32 means linear
shifts versus applied field F at the cathode surface because F = PV,,,.
This shift and its linearity versus the applied voltage are explained by a
charge confinement in the region of the topmost atom, which implies a

field penetration into the tip.
The charge confinement and the penetration of the field can be estimated
by the Thomas-Fermi model of screening [49]. To estimate the field pene-
tration x, for the protrusion, the expression for the potential of the electric
field penetration into a flat subsurface region is used as a first approxi-
mation:
V, = x,Fexp (-:) (35)
where x is the distance from the surface to a position within the cathode.
Thus the energy of the emitted electrons (x = 0) varies linearly with
vApp as
AE = exOpVApp (36)
by taking into account the relation between F and VApp.
Applying Eq. (36) to our experimental results of AEIVA,, = 1.65
meV/V and taking p as 5 to 10 X lo6 m-l for a protrusion of 2-3 nm height
[34], gives x, of 2-3 A for the single-atom protrusion. This value should
be compared to the screening length of a metal surface, which is less than
0.5 [49], and also with the estimation of field shift with single adsorbed
Ba, which is 1.3-1.7 A [27]. It is also roughly the dimension of an atom
and this strongly supports the idea that the observed peaks in the TED
spectra are related to localized levels at the topmost atom.
Calculations of the electronic structure over metal protrusions for differ-
ent materials, structures, and geometries (height) [50] have been recently
developed using the tight-bonding formalism. The main advantage of this
semiempirical method is that complex objects containing nonequivalent
atoms can be calculated. The calculations were done for single-atom end-
ing pyramidal protrusions for different metals (W, Fe, and Cr). Different
heights and crystallographic orientations were considered. Figure 34 is
an example of the results, showing the local denisty of states of the topmost
atom. The main points of interest related to this work are as follows:
1. The local density of states over the apex atom of the pyramidal
protrusion evolves toward a peak structure when the height, p,, of
the protrusion is increased layer by layer, starting with one adsorbed
atom on a surface. This electronic structure comes to a steady state
for po 2 4 layers.
2. The final steady-state electronic structure is characterized by a pre-
-
dominant peak localised 1 eV over the Fermi level. However, the
2.5
2.0
9
9 1.5
*
P)
0
*
vl v3
Y
1.0
3
E: 1.0
0.5
0.0
-5 -3 -1 1 3 5 -5 -3 -1
E[eVI WeVI
a b
FIGURE34. (a) Local density of states of a W surface atom of a semiinfinite W (001)
crystal. (b) Local density of states of a W atom at the apex of a (001) pyramid of height 2
4 atomic planes above the (001) surface from ref. [50b]. The vertical line corresponds to
EF.(From ref. [50bl.)
calculations have been done for zero applied field and do not take
into account a possible field shifting. The experimental measurements
of many nanotip TEDs show that the extrapolated zero-field position
of the bands are found quite generally to lie in the 1 to 2 eV range
above E F , in agreement with the calculations.
3. This local density of states distribution is specific to a single-atom
ending protrusion. The conventional surface LDOS structure is re-
covered when the last atom is stripped off.
It is premature to compare too strictly the calculations with the experi-
mental results and to expect a fit between the two sets of values. However,
the theoretical results confirm most of the previously specific experimental
characteristics of the TED spectra from the single-atom nanotips.
All these results, experimental and theoretical, assert that the free-
electron behavior for field emission is not valid for the atomic-scale emit-
ting source of a nanotip. Consequently, the local density of states specific
to each nanoprotrusion must be considered in interpreting the experiments
with an atomic-scale probe as in scanning tunneling microscopy and scan-
ning tunneling spectroscopy experiments with atomic resolution, instead
of the commonly used free-electron model [51, 521.
4. Current Saturation in the I-V Characteristics

The current saturation related to the presence of a protrusion was mea-
sured at the very beginning of the study on nanotips [2], and it is the
signature of the presence of a high nanoprotrusion ending in one atom.

Different interpretations [2, 33a, 531 were proposed to account for the
observed discrepancies between I-V characteristics for nanotips and the
conventional Fowler-Nordheim analysis.
Since in the conventional Fowler-Nordheim analysis the model of the
tip is a planar surface, the field everywhere outside the tip is a constant.
For protrusion tips with atomic sharpness, the field in the region around
the emitter and away from the apex varies on the scale of the protrusion.
10-l~ I I I
1 0 - l ~- -
z
h
*w
-
cw10-17 -
J
10-18 -
10-l~ I I 1
1.6 l o 3 1.8 2.0 2.2 1 0 ‘ ~ 2.4
1/v (Volt-‘)
a
I I I I
-
-
-
-

b
FLGURE 35. Nanotip FE I-V characteristics (a) and accompanying TEDs (b) for three
different extraction voltages measured concomitantly: (1) 420 V, (2) 475 V, and (3) 525 V.
The TED for 420 V is above E F . They show the direct relationship between the current
saturation and the peak structure.
It is found, for example, that the field away from the apex decreases rapidly
[34], whereas the field is constant for a planar geometry. Consequently, the
analysis of the I-V characteristics for nanotips have to take into account
the protruding geometry in the calculation of the tunneling barrier. The
results [33a, 531, plotted as In(J/F2)versus 1/F, show a saturation of the
current but only in the very high-field region.
A second parameter is the presence of localized peaks for FE from
nanotips. The relation between the presence of the localised peak(s) in
the TED and the current saturation for the I-V characteristics is shown
in Fig. 35. They show clearly that the presence of the peaks and their
shifting (Fig. 35b) are concomitant with the current saturation (Fig. 35a).
The presence of localized bands that shift with the applied voltage dramati-
cally changes the supply function, which could be the predominant factor
for the current saturation.
Conventional Fowler-Nordheim analysis is then not valid, because con-
sidering the tip as planar and the electrons as a free-electron gas are not
valid assumptions for nanotips.
IV. APPLICATIONS
As was pointed out in the introduction, the level of interest in nanotips

depends essentially on the new possibilities that they can open due to the
specific field emission properties attached to the atomic size of their emit-
ting area. Five actual subjects will be presented, which have been devel-
oped with the nanotips: FEM with atomic resolution; monochromatic
electron sources; energy exchange and its analysis at the atomic scale;
low-energy ,high-resolution microscopy with the Fresnel Projection micro-
scope; and finally, several new phenomena with ferromagnetic nanotips.
The results, which are presented below, were obtained in the last few
years and are examples of the breakthroughs made possible by the use
of nanotips.
A . Atomic Resolution under FEM
Atomic resolution in FEM has been a subject of interest in the FE commu-

nity since its inception [6]. The conventional approach to estimating the
resolution in FEM [6,7] considers the distribution of momenta transverse
to the normal emission direction and the wave nature of the electron.
Taking these into account gives an effective resolution of the order of
2 nm, well above the size of one atom.
Consider the resolution problem within a three-atom apex of a nanotip.

Because of the field enhancement in the vicinity ofthe three atoms forming
the apex of the protrusion, the tunneling barrier will present three minima,
located on top of each of these atoms. However, as the distances between
these minima are of the order of the wavelength of the tunneling electron,
the three emitting beams cannot be considered independent. Electron
diffraction and interference between these beams are important, and thus
a full quantum mechanics approach is necessary. This has been performed
with a 2D model which mimics as close as possible the three-atom apex
of the nanotip and is solved by exact numerical integration of the corre-
sponding TDSE (Fig. 36) [33]. The initial electron wave packet is a Gauss-
ian packet moving in the direction of the nanotip axis and the field-emitted
e beam is described by the probability distribution of the transmitted wave
packet. The calculations show that if the field emission comes from all
the atoms of the protrusion apex, rigorous conditions on the protrusion
geometry are necessary in order to obtain atomic resolution in the FEM
patterns. The first of these conditions is that the radius of curvature of
the equipotential around the apex should be of the order of the interatomic
Ilb
FIGURE36. (continued) 11. Intensity of the wave packet reflected and transmitted by
the tunneling model in I for a tilt angle of 30". (a) In the presence of the tunnel barrier, the
two coherent waves emitted by the two atoms do not merge and the atomic resolution is
observed. (b) In the absence of the tunnel barrier, the two transmitted waves interfere and
no atomic resolution can be observed. (From ref. [33b].)
distance. This requires a protrusion ending in a trimer with a height po

much larger than 1 nm. The second condition is that the tilting angle O1
with the presence of the tunnel barrier must be in the range 20-30".
Although the first condition is fulfilled by nanotips, the second condition
is attained only for three-atom ending protrusion whose height is in the
range of 4 to 6 nm [34]. This is illustrated in Fig. 37.
Experimental realisation of atomic resolution in FEM for a three-atom
ending protrusion is shown in Fig. 38. The three emitting spots of Fig.
38a are related directly to the three ending atoms. The stability of the
current over a 5-h duration demonstrated that the spots were coming from
the atomic position, because larger emitting areas would have certainly
shown current decreases due to adsorption (cf. discussion about stability).
Moreover, we show in Fig. 38b the FEM pattern of the same nanotip with
an adsorbed atom between two of the initial three atoms, which is indicated
by the arrow in the figure. The presence of this additional spot was accom-
panied by a discrete jump in the FE current, which is another experimental
proof of atomic resolution.
B . Monochromatic Electron Beam
The existence of localized bands in the TEDs from nanotips is in contrast

to FE from microscopic metal emitters, where the electrons come from
the wide conduction band and have a distribution fixed at the Fermi level
(EF). In this case, the distribution width is defined essentially by the
tunneling barrier, which fixes the lower limit of about 0.3 eV (cf. Section
II.C.2). Conversely, for nanotips the energy distribution of the FE elec-
trons is now governed not only by the tunneling barrier but also by the
localized band structure. Modifying this last parameter allows us to narrow
the FE energy spread well under the 0.3-eV limit [54].
Examples of TEDs measured for W and Pt nanotips at 293 and 80 K
are shown in Fig. 39. The nanotip TEDs are characteristically narrower.
For the W nanotip the measured FWHM was 120 meV at room temperature
and 110 meV at 80 K (Fig. 39a). For the case of the Pt nanotip, the
measured FWHM was 100 meV at 293 K, and this decreases to 64 meV
at 80 K (Fig. 39b). This has to be compared with the energy dispersion
from microtips plotted in the same figures, which was about 0.3 eV and
decreases only modestly with temperature. In Fig. 40 the FWHM of the
Pt nanotip TEDs is plotted as a function of temperature. It shows a linear
decrease with temperature with a zero temperature value of 51 meV.
The above values are the experimentally measured values without any
correction for the instrumental broadening. Using the conventional decon-
l5 r
- 0
-2 -1 0 1 2
b
FIGURE37. Variation of the tilt angle between the tip axis and the direction of the
maximum field over each of the atoms forming the trimer apex of a protrusion as a function
of the protrusion height. (a) 3D calculated field distribution with atomic resolution over an
equipotential surface for a trimer nanotip (po = 4 nm and base diameter = 4 nm). It shows
a local enhancement of the field just over each of the atoms forming the trimer apex. (b)
Plot of the /3 factor showing the local field enhancement over each of the three atoms of
the trimer apex for three different heights of the nanoprotrusion. The indicated angles are
the tilt angle over the apex atoms. Tilt angles > 20" are obtained for protrusions with a
height >4 nm. (From ref. [34].)
a b
FIGURE38. (a) FEM patterns showing three emitting spots from a nanoprotrusion. (b)
FEM pattern of the same nanotip showing the presence of an absorbed atom indicated by
the arrow. (From ref. [33al.)
-1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2

a
1 I I I I I I
-
-
1.0
-
--
9v)0 . 8
.-
- Bulk tip (293 K)
FwHM=270 meV
Nanotip (293 K)
-
d
t
d
FWHM=lOOmeV
-
0.6 Nanotip (80 K)
0
G
W
-
2
fz 0.4 -
-
"
I
0
0.2 -
-
0.aI
.1.0 -0.8 -0.6 -0.4 -0.2 0.0 0.2
b
FIGURE39. TEDs of microtips and nanotips for different temperatures: (a) W; (b) Pt.
110 I I I I I
-
-
-
-
I I I I I
40
0 50 100 150 200 250 300
Temperature (K)
FIGURE40. Variation of FWHM of the TEDs from the Pt nanotip with temperature.
The experimental data falls on a straight line.
volution technique [55], the FWHMs for 293 K, 80 K, and extrapolated

to zero temperature are 90 meV, 43 meV, and 20 meV, respectively. An
energy dispersion in the range of 20 meV should then be achievable by
cooling the tip to liquid helium temperatures, a factor at least 10 times
narrower than for the standard FE microtips. Thus the use of nanotips
allows us to break through the limiting value of about 0.3 eV for cold
field emission.
C . Local Heating and Cooling by Nottingham Effect
During the field electron emission process, energy exchanges take place
between the emitted electrons and the cathode surface. These exchange
processes, or so-called Nottingham effect [56], can cause a heating or
cooling at the emitter surface when the average energy of the replacement
electrons, which is near EF,is different from that of the emitted electrons.
The Nottingham effect is negligible for macroscopic tips with apex radii
greater than a few tens of nanometers and FE current < 1 pA [57]. This
is partly because both the replacement electrons and the emitted electrons
come from energy levels close to EF and thus the energy exchange per
emitted electron is limited. The localized peak distribution in the TED
and its shifting lead to questions about the local heating or cooling at the
nanotips. Experimental measurements show that there is considerable
heating induced by the FE process in the case of nanotips and, further-
more, they show that local temperatures of areas with atomic scale can
be measured [%I.
1. Localized Peaks under EF :Heating Effect

The energy exchange in a nanotip during FE is depicted in Fig. 41 for the
case of two localized bands. The replacement electrons coming from
near EF must fill the levels in each localized band emptied by emission
(processes 1 and 2 of Fig. 41) and also from the upper localized band level
to the lower (process 3 of Fig. 41). In both processes, an amount of energy
is lost by the electrons which depends strongly on the number and position
of the bands with respect to EF. This implies a dependence on the protru-
sion geometry and the V,,, as well. Since the energy exchange per emitted
electron could have values of the order of electronvolts, this will lead to
much larger increases in the temperature at the single-atom apex of a
nanotip compared to the conventional Nottingham effect.
The experimental problem in the studies of such temperature increases
is how to measure the local temperature at the apex of a nanotip during
FE. This temperature may be very different from the temperature of the
whole tip because of the very small emitting area. It is necessary to
have a local probe of the temperature giving atomic-scale resolution. The
determination of the local temperatures is based on two effects. The first
FIGURE41. Simplified potential diagram for a nanotip emission which depicts the addi-
tional energy-exchange paths during field emission in the presence of localized bands at the
apex of the nanotip.
effect is the possibility of having a repetitive, back-and-forth motion of a

single atom between neighboring atomic sites at the nanotip apex, termed
“flip-flop” [59], whose frequency is dependent on the local temperature.
The second effect is that the shape of the TEDs depends on temperature.
Using procedures based on these effects, the local temperatures at the
apex of a nanoprotrusion for electron emissions were determined in the
range of to A [%I.
In the first procedure the current fluctuations due to the flip-flop of one
adsorbed atom between two neighboring sites at a nanotip apex were
measured versus the FE currents. For single-atom protrusion tips, the
total FE current switches between two fixed discrete values which depend
on the atomic configuration of the protrusion. Each current level is associ-
ated with a particular TED. An example of two TEDs measured during
the two states of a flip-flop is shown in Fig. 42. The number of peaks and
their relative positions are preserved during the flip-flop, but the TEDs
shift as a whole and the relative peak intensities change. The switching
between the two spectra is repetitive as long as the flip-flop continues.
This phenomenon allows very easy detection of the flip-flop even for total
FE currents from the single-atom tips in the range of A. The variation
of the number of counts at a fixed energy during a flip-flop process is
FIGURE 42. Effect of a flip-flop process on the TED from a nanotip presenting two
bands. The inset shows change in the total number of counts at one particular energy during
a flip-flop. (From ref. 1.581.)
I I I I
"
950 975 1000 1025 1050 1075
FIGURE43. Frequency of a flip-flop at a nanotip apex versus V,,,. (From ref [58].)
shown in the inset of Fig. 42. The effect of the emission current on the
flip-flop frequency is shown in Fig. 43. For an increase of the FE voltage
from 950 V to 1070 V, the flickering frequency increases from -0.1 Hz
to -11 Hz. This corresponds to an increase in the temperature in the
range of 30 K for a FE current increase from -3 x A to -9 x lo-'* A.
The second method is based on the following experimental observation:
The shape of the peaks in the TED from single-atom protrusion tips is
temperature-dependent . Figure 44a shows a broadening of the high-energy
edge of the TED of a nanotip for controlled increasing temperature by
using the heating loop at fixed FE voltage and current. As shown in Fig.
44b, for increasing applied voltage and F E current there is a broadening
of the high-energy side due to the emission-induced temperature increase,
in addition to the shift of the spectra characteristic of the nanotips. The
local temperature increases at the apex of the protrusion tip found by
fitting the spectra for different FE currents are shown in Fig. 45. The
temperature increase can reach a value of -210 K for -1 X A. For
higher FE currents the temperature increase is even larger. The protrusion
becomes unstable and it can be destroyed by a local melting for I >
-10-7 A.
2. Localized Peaks over EF: Cooling Effect

The position of the localized bands of nanotips relative to EF can be
controlled by the applied voltage because of the field shifting. In particular,
1.o
G?
0
2 0.8
8 0.6
v
c
2 0.4
v,
1
0
u 0.2
0.0
-1.0 -0.8 -0.5 -0.3 0.0 0.3 0.5
a
b
FIGURE 44. (a) TEDs from a protrusion tip with one localized band: room temperature
and 590 K. The higher temperature is created by loop heating current. (b) Spectra from the
same nanotip for different applied voltages and different emission currents. The spectrum
at higher voltage shifts to lower energy [4]. It has been numerically shifted by A E to the
position of the lower voltage peak (small dots) to show the broadening of the high-energy
side of the spectrum that is related to the temperature increase. (From ref. [ 5 8 ] . )
this allows emission from localized levels well above EF of the support
tip. This phenomenon is explained by the partial filling of the bands by
the tail of the Fermi sea, which acts as a supply function [60]. In this
case, energy conservation in the FE process is obeyed.
Experimentally, the linear shifting is typically -0.5 eV for the range of
250 I I 1 1 I
-
-
0.0 0.2 0.4 0.6 0.8 1.0 1.2

1 (nA)
. 45. Variation in temperature versus FE current as determined by fitting the
FIGURE
TEDs from nanotips.
possible applied voltages, which permits us to scan the bands completely

through EF in a controlled way. As an example, Fig. 46 shows a narrow
band shifting linearly with applied field through EF. It does not change in
form or width, but its intensity drops rapidly as it crosses E F .
The experimental results show that emission from a peak above EF is
possible if the peak or related band is sharper than the Fermi edge itself.
I I I I I

FIGURE 46. FEES spectra as a function of applied field from a W nanotip. For the
lower voltages the peak shifts to above E F . (From ref. 1601.)
Under such conditions, the peaks above EFcan provide significant cooling
of the tip due to Nottingham energy exchange effects, because the emission
comes exclusively from electrons above the Fermi level.
A side consequence of a band being positioned on the high-energy tail
that has been noticed is that the FE is strongly dependent on the tempera-
ture, in complete contrast to normal FE, which has a very weak tempera-
ture dependence [see Eq.(15)]. This is simply because the supply function
increases as the temperature is raised and the total current is then propor-
tional to exp(-AElkT) for AE S- kT.
D. Fresnel Projection Microscopy

In electron microscopy, efforts to increase resolution have focused mainly
on reductions of the spherical aberration, the wavelength, and the energy
spread of the electron beam. These approaches give excellent results for
specimens which are not sensitive to radiation damage caused by the
interactions with a high-energy e beam. However, there still remains much
to be done in obtaining high-resolution images of organic specimens.
Using the nanotip as an atom-source of electrons in a projection micro-
scope is another approach to achieving observations of carbon and organic
nanofibers such as synthetic polymers and RNA [61,62,75]. This combina-
tion takes advantage of the simplicity and low working voltages (50-300
V) of the projection microscope, and the unique properties of the field-
emission electron beam from the nanotips that are related to the atomic
size of the sources. The experimental images of nanometric fibers were
interpreted as Fresnel diffraction patterns from opaque objects, even for
-
fibers whose diameters were down to 1.4 nm.
1 . The Fresnei Projection Microscope
Projection microscopy was proposed in 1939 by Morton and Ramberg [63]
with their point projector electron microscope. In 1968, E. W. Muller
introduced the field ion shadow projection microscope [8, 641 based on
the same principle, which is the following. The greatly magnified shadow
of an object (magnification factor -lo6) can be obtained by making use
of the quasi-radial propagation of field emitted electrons or ions coming
from a tip when the object is inside the beam path. The projection or
shadow microscope is then essentially a lensless microscope based on the
radial propagation of an e beam from a point source (Fig. 47). The image
has a magnification factor M given by
i
M=-zDld (37)
0
ELECTRON FIELD EMISSION FROM ATOM SOURCES I25
Projection
microscope
4-- -4- -
MCP
magnification = i/o = D/d

FIGURE47. Schematic description of the Fresnel projection microscope. The projection
coherent source is a field-emission W nanotip emitting in the range of 200 to 300 V. The
image magnification is given in the first approximation by the ratio Dld, where d and D are
the distances between the virtual projection point to the object and to the screen, respectively.
where i and o are the image and object dimensions and D and d are the
distances of the projection point to the screen and to the object, respec-
tively. Equation (37) shows that the magnification increases by approach-
ing the object to the projection point and could reach values in the range
107-106for projection point-object distances between 10 nm and 100 nm,
with the screen located 10 cm away. With recent technological develop-
ments due to scanning tunneling microscopy (STM) [65], tip-sample dis-
tances of less than 1 nm can now be routinely handled by using piezodrives
for controlled nanometric displacements. This has given rise to renewed
interest in this projection microscope [61, 66, 671.
Among the nanotip characteristics, two are of particular interest for
the projection microscope: the atomic size of the emitting area and the
protrusion geometry of the nanotips. We show hereafter that both play a

role in the image formation and, therefore, in the analysis of the interaction
between the coherent nanosource and the nano-objects.
a . The Virtual Projection Point. The distribution of the electric field
in the apex region of a nanotip induces trajectory distortions of the emitted
electrons, and thus the center of the real source at the apex does not
correspond to the projection point or virtual point source [20].The virtual
projection point is defined as the intersection of the asymptotes of the
trajectories from the distortion-free zone far away from the tip. This is
drawn schematically in Fig. 48 for a nanotip and has to be compared with
the virtual source of a conventional tip in Fig. 6. It is assumed as a first
approximation, even if the distortions depend on the exact geometry of
the tip end, that the tip behaves like a lens with a value of the ratio BJB,
around 0.5. This means that the minimum distance dminfrom the virtual
source to the apex is greater than 2r. From the schematic drawings in
Fig. 48, it can be seen that nanotips give smaller dmin and therefore higher
possible magnifications compared to hemispherical microtips (Fig. 7), due
to their protruding geometry. This allows us to work experimentally under
Fresnel conditions, as we discuss below.
Moreover, as the distance of the tip to the object is in the range of 100
nm and less, and because in the projection microscope configuration the
object also plays the role of an extracting anode, the FE voltages needed
to have field emission current are in the range of 50 V to 300 V [68]. Thus,
FIGURE48. Schematic representation of the virtual radial projection point source Vfor
a nanotip relative to the surface apex and its geometric center C. Both C and V are much
closer to the apex than for the hemispherical tip (Fig. 6). (From ref. [62].)
for high-magnification working distances, the projection microscope is

intrinsically a low-energy electron microscope.
6 . Fresnel versus Fraunhofer Diffractions. The above approach of
defining the magnification considered the projection microscope only
within the “geometric” point of view. However, as the FE beam from
nanotips is corning from the last single atom, the interaction between a
coherent beam with an object must also be considered [69], in other words,
the diffraction of the beam by the object and the interference. Electron
interference and holography are intimately related. However, the essence
of holography [70] is a two-stage process: first the formation of an interfer-
ence pattern by adding an intense reference beam with a beam modulated
by the specimen, and second, the extraction of information about the
object from this interference pattern. The exact mechanism for the diffrac-
togram formation in a projection microscope has to be settled first, and
this requires diffraction and interference theory. This step is necessary
before information from the object itself can be extracted with confidence
from the diffractograms. In the interpretation of the images presented
here, we take into consideration only the diffraction mechanisms between
a source and an object with the different related parameters (source size,
object size, source-object distance, wavelength, etc.) defining the incident
wavefront geometry which are basic to the understanding of the inter-
ference images. Within the projection microscope configuration, the
object-screen distance is typically about 10 cm. It is therefore the distance
between the tip and the object and the sizes of the source and the object
which will determine the nature of the resulting diffraction [62].
Because the object dimensions (>1 nm) are much larger than the wave-
length of the electrons X (-0.1 nm), let us consider the classical electron
optics wave theory, which provides a precise formalism for describing
the scattering. For illustration, imagine that we have an object 0 having
a transmission function T ( y , , z,), illuminated by a point source V which
gives a beam illumination B(y,, z,) at the object. Under these conditions,
the wavefunction 9 ( P S )at each point P,of an image of the object projected
onto the screen is given by the Fresnel-Kirchhoff formula with the Helm-
holtz-Sommerfeld boundary condition [7 I] :
where Tois the amplitude of the incidence wave, yo and z, are the trans-
verse coordinates in the object plane, k is the wave vector, rloand r,, are
tip-object and object-screen distances, respectively, n is the unit vector

perpendicular to the object plane ( y , z ) , and the term [cos(n, ros) +
cos(n, rt,)]/2 = K(n, ros, rro)is the obliquity factor. The exponential factor
describes the spherical waves impinging and scattering from the object
with their respective director cosines. B ( y , , z), is the beam shape known
from the experimental measurements (Fig. 23) to be Gaussian-like:
where w = zt, sin(a) is the illumination of the beam at the object, and a
is the half-beam opening.
T(y,, z,) describes the transmission function of the mask object. For
example, for an opaque object, T(y,, z,) is 0 inside the object and 1 outside
the object. If the object is three-dimensional, the x component is averaged
because it is the direction of propagation.
The intensity at each point on the screen is then
W,)= lWP,>l2 (40)
Within the experimental situation,
x:, (y2 + z’,). (41)
This implies that
which is precisely the condition for the Fresnel approximation.

Introducing Eq. (42) in Eq. (38) gives
x exp ( - i k ( y o y s + zozs)) exp (ik (Y2 + 2 3 ) dy, dz,.

XOS 240
There are two limiting cases for the diffraction. First, under experimen-
tal situations where the electron source is small compared to the object
and for small source-object distances, the small angle approximation or
Fresnel conditions can be applied and the result of Eqs. (43) and (40) is
a projection image on the screen which is clearly recognizable despite
fringes around its periphery. This is known as Fresnel or near-field diffrac-
tion when the wavefront can be considered as spherical within the ob-
ject dimension.
An increase in the source-object distance results in a continuous change

in the fringes. For large source-object distances, the projected pattern
will shrink considerably and the fringes will bear little or no resemblance to
the actual object. Thereafter, moving the source-object distance changes
mostly the size of the diffraction pattern and not its shape. This is Fraun-
hofer or far-field diffraction. The incoming wave is then nearly planar
over the extent of the diffracting object. As a practical rule of thumb,
Fraunhofer diffraction will prevail over Fresnel when
Y2 Z2
x,*% 2 and X,**? (44)
h A
where yo and z, are the object dimensions. Similarly, if the size of the
source is of the order of the object dimension, then the incoming wave
is almost a plane compared to the object dimensions, which means Fraun-
hofer diffraction.
In experimental situations where the electron source is atom-size as
with the nanotip, the small-angle approximation and Fresnel conditions
can be applied for small tip-sample distances, i.e., high magnification, so
the resulting projection diffractograms bear the contour of the mask object.
This is a big advantage for the experimental observations.
2. Experimental Procedures
In the Fresnel projection microscope [61, 621, the electron point source
is a W( 111) single-atom nanotip spot-welded to a joule heating loop and
in contact with a liquid nitrogen reservoir. The sample and the object
holder are attached to a nanodisplacement system which is composed
of a commercial piezomotor for the x direction and home-made inertial
movement driven by a piezotube for the y - z displacements. The overall
displacements are in the range of centimeters in the x , y , z directions.
The resolution in the displacements is given by the minimum bending and
elongation of the piezotube, which are in the range of 0.1 nm. The projec-
tion image is formed -10 cm away from the tip on a multiple-channel
plate coupled to a fluorescent screen. These images are visualized and
analyzed through a numerical image acquisition system.
The entire microscope system is vibration-isolated with a simple pneu-
matic system, without any internal antivibration system as is normally
employed in STM microscopy. Shielding of the stray magnetic field is not
needed to obtain nanometric resolution [61, 621.
The absolute dimensions of the samples and the scales given in the
figures are measured directly by following the displacement of the projec-
tion image on the screen versus the motion of the object due to the
FIGURE49. Fresnel diffraction patterns (VFE= 300 V) by nanometric carbon holes and
fibers. The illuminated area corresponds to the e beam coming from a W nanotip. (From
ref. [61].)
deflections of the sample-holder piezotube with applied voltages. The

dimensions of the object are then determined directly for any nanotip-
object distances with an accuracy given only by the calibrations of the
piezodrives whose behaviors are now very well known [65]. This proce-
dure removed the uncertainty in the determination of the object dimen-
sions, because the position of the virtual projection point source V is not
known with accuracy due to the deformation of the electric field lines
near the tip apex.
3 . Experimental Results
a . Nanometric Carbon Fibers. An example of the effects of the na-

notip geometry on image formation is illustrated in Fig. 49, which shows
the diffraction patterns obtained with carbon nanofibers. A comparison
with calculated Fresnel diffractograms is given in Fig. 50. The calculated
Fresnel diffraction pattern is obtained with the following parameters: A =
0.7 A,diameter of the wire = 14 A and point source-object distance =
280 A. This last value corresponds to the distance of the virtual source
(see Fig. 48) to the object and not to the actual distance from the nanotip
FIGURE50. Fresnel diffraction patterns (VFE= 300 V) by a nanometric carbon fiber.

A nanometric structural defect is indicated by the arrow. (1) Image with the Fresnel projection
microscope. (2) Calculated Fresnel fringes from a 1.4-nm-diameter wire illuminated by a
beam of A = 0.7 A coming from a point source at 28 nm from the wire. (3) Diameter of the
wire. (From ref. [62].)
apex to the object. The observations of nanometric details present along

the fibers and the similarity between experimental and calculated Fresnel
diffraction patterns indicate that the nanotips used were nearly ideal coher-
ent point projectors. This is consistent with the simple approach given in
the upper paragraph. Comparison between these results and those of refs.
[66, 671 gives rise to the following considerations:
In the case of carbon fibers, for example, a direct comparison with
diffraction under the Fresnel conditions, as shown in Fig. 50, already
gives good agreement. Interpretations in terms of holography [72,73],
although fancy, could be misleading about the nature of the diffractive
object. This was recently amply proved when the former experimental
observations [66], which had been interpreted by holographic theories
as images of the atomic lattice of the substrate [72, 731, were shown
to be only Fraunhofer diffractograms of multiple -20 nm carbon
holes 1741.
Among the diffraction patterns of fibers presented by different authors
[66], some of them cannot be interpreted as Fresnel diffraction. In
these images the underlying FEM patterns of the tips used to obtain
these fringes were composed of multiple spots over the whole screen,
which means an actual source that was not limited to one atom. The
diffractograms presented in ref. [66] must be interpreted as Fraunhofer
diffraction. The same conclusion was reached when some diffraction
patterns presented in refs. [66,72] were reinterpreted in ref. [74]. This
confirms that conventional FE tips with extended electron source area
have to be considered as plane wave sources.
b . Organic Molecules. The FPM is then a low-voltage, high-resolution

microscope giving nanometric resolution in the hundred-volt energy range.
It is a perfect tool for observations of organic materials such as synthetic
and biomacromolecules. This prediction is assessed by the observations
with nanometric resolution of synthetic polymers (PS-PVP) [62] and bio-
logical molecules of RNA [75] with the FPM.
Sample Preparation Procedure. The object preparation in FPM differs
from conventional electron microscopies (transmission or scanning) be-
cause of the low-energy observation beam of less than 300 V. Within this
range of energy, the samples are opaque objects when their thickness are
greater than 1 nm. This means that the samples have to be prepared as
standalone fibers across holes.
The following two-step procedure has proved to be valuable for organic
polymer fibers, synthetic 1621 or biological [75].
1. Dissolution of the macromolecules in a solution at a concentration
around a few mg/ 1 . The solvent has to be specific to each sample.
For example, chloroform is used for the PS + PVP polymers [62]
or NaCl solution in the case of RNA [75].
2. Deposition of a drop of 2 p1 of this solution on a holey-carbon grid.
After evaporation of the solvent, the probability of having polymers
stretching across a hole is rather large, allowing observation by FPM
as shown in Fig. 51. The polarization of the holey-carbon grid is
used during the deposition of RNA to assist the anchoring of the
molecules on the substrate due to the negative polarization of the
phosphate groups.
Note that no other specimen preparation such as staining or metal
coating, for example, is done.
Polymers [621. The polymers were a mix of polysulfone of bis-phenol-
A (PS) (95%) and polyvinylpyrrolidone (PVP) (5%). They are the constit-
uents for the fabrication of the hollow fibers used in commercial fibers
for human dialysis. Figure 52 is an overview image of the main characteris-
FIGURE51. Low-magnification FPM image of an RNA network stretching across a

micrometer carbon hole. The black hole in the middle of the image is the blind direction of
the channel plate. Notice the similar diameter of most of the fibers. (From ref. [75].)
FIGURE52. Overview of a supramolecular network of the polymers (PS-PVP) with the

indications of some specific structures: 1, polyhedric shape of the network; 2, periodic
structure along the fibers; 3, clew. The imaging voltage is 280 V. (From ref. [62].)
tics of the polymer network. Other examples of each of the designated

specific characteristics are shown with higher magnification in Fig. 53.
Some conclusions on the polymer behavior can be highlighted.
1. Observations of polymers with details less than a nanometer can be
achieved with an e-beam energy in the range of 200-300 V without
any observable degradation of the sample under the beam even after
1-hr-duration observation.
2. The polymer chains are self-organized into polyhedral superstruc-
tures with fibers of different lengths and different diameters, with
special mention of the presence of the nanofibers sitting across the
polymer holes (Figs. 52 and 53a).
3. When the polymers are not stretched over two anchoring points,
they form a clew (Fig. 53c). For the polymer this feature should be
its minimum energy conformation and is observed only for polymers
as opposed to carbon fibers.
4. The Fresnel diffraction patterns show a periodic variation along the
length of the structure (Fig. 53b). This periodic variation also has
an echo in the surrounding fringes. The comparison between Figs.
49 and 50 and Fig. 53b shows clearly the differences between the
experimental diffraction patterns of a carbon fiber and a polymer
fiber. The periodic structure for the polymer fibers, which induces
modulated diffraction fringes, is also present in the diffractograms
of the network. This then raises the question of the formation of
periodic supramolecular structures from the initial polymer solution.
Figure 53b, for example, suggests strongly the presence of a twist
shape for the supramolecular fiber structure.
RNA [751. The capability of FPM for high-resolution analysis of soft
materials is also largely confirmed by the observations of the as-deposited
A-RNA molecules. For some images comparisons are made with simulated
Fresnel patterns of the masks sketched and shown as insets in the figures.
The objective of these simulations is not to find out the exact real experi-
mental geometry of the objects, but only to show what kind of mask
geometry can give the observed diffractograrns. From the experimentally
observed diffractograms, which are presented in Figs. 54 to 60, the follow-
ing points can be highlighted.
The periodic structure along the .fibers. The fibers, whose diameter is
around 2 nm, present a periodic variation of the fringe intensity along the
longitudinal direction. This periodic variation of the fringe intensities could
be observed along the whole fiber length (Fig. 54a). In this figure, the
length of the fiber is about 30 nm and the periodicity is -30 nm. Compari-
sons between the experimental images (Fig. 54b) and calculated Fresnel
b C
F ~ G U R53.
E Detailed observations of some characteristic features of the polymers. Im-
aging voltages are between 260 and 275 V. (a) Self-organization into polyhedral supramolecu-
lar structures. The polyhedric shape of the holes reflects the presence of nanometric struc-
tures of the polymer fibers constituting the network. These structures are also echoed
in the rich diffractograms inside the polymer holes, as shown by a comparison with the
diffractograms of carbon network in Fig. 49. (b) Periodic supramolecular structure of a
polymer fiber. The diffractogram suggests the presence of a twist shape. (c) A polymer clew
with its surrounding diffraction pattern. (From ref. [62].)
b
FIGURE54. (a). FPM images of an RNA free-standing fiber at two different magnifica-
tions, showing the presence of a periodic variation of the fringe pattern (the high-magnification
part is framed inside the low-magnification image). (b). Comparisons with the Fresnel diffrac-
togram: The upper image is the FPM image; the middle diffractogram are simulated diffraction
from a 2D mask presented in the lower part, which mimics the shadow of the A-RNA
conformation (diameter of 2.3 nm and period of 3 nm). The numerical simulations use a
wavelength value of 0.7 8, and a projection virtual source-sample distance of 500 A. (From
ref. [75].)
diffraction from a 2-nm-diameter fiber that mimics the periodic variation

of the helix pitch in the A-RNA structure [76] show that the periodic
variation of the fringes can only be interpreted by the presence of a
periodic structure along the fibers which is very near to that of RNA.
Secondary structures. Besides the above periodic variation along the
fibers whose diameters were mostly around 2 nm, different other confor-
mations were also noticed.
1. The formation ofnetworks. The linear periodic fiber shown in Fig.
54 is observed when it is stretching across a small carbon hole. When
the dimensions of the carbon holes are larger, the characteristic
features observed were not single fibers but networks.
The network units are of polyhedric shape with the constituent
fibers showing a periodic structure and having a diameter mostly
in the range of 2 nm. The image in Fig. 51 and the high-magnification
images of the networks presented in Fig. 55 are illustrations of this
configuration. A cumulative histogram of the measured angles at
the crossing points of these networks reveals a peak around 120".
This value is also confirmed by the very high percentage of three-
b
FIGURE 55. High-magnifications FPM images given the characteristic details of RNA
networks. They are indicated by the numbered arrows. (1) 2-nm fiber network and connec-
tions without extra material at the crossing point. Figure 56 compares such connections to
a simulated Fresnel diffractogram. (2) High-density connection zone. (3) Fibers less than
2 nm in diameter with a connection without extra material. (4) A loop inside a network.
This has to be compared with a loop along a fiber shown in Fig. 60. (From ref. [75].)
branch crossing points. This indicates that the network observed

is not due to the superposition of individual fibers during the flatten-
ing of a 3D distribution. In this latter case, the distribution of the
crossing-point angles must be random and the number of four-
branch links must be predominant. This polyhedric-shape net struc-
FIGURE56. Fork separation structure showing a connection without extra material at

the crossing point. (a) Calculated Fresnel diffractogram (A = 0.7 A, point source-object
distance = 500 A); the mask of the object is shown as an insert (not at the same scale). (b)
FPM image o f a network connection without extra material shown in Fig. 55. (From ref. [75].)
ture, concomitant with the presence of a periodic structure along

the constituent fibers, suggests more an intrinsic molecular struc-
ture, a notion which is asserted by the structure of the crossing
points that will be discussed just below.
At the crossing points between the fibers, their size remains con-
stant (Fig. 56). That means there is no supplementary matter pres-
ent inside this crossing zone. In order to verify this assertion, we
have made Fresnel diffraction simulations for two different cases,
with and without supplementary matter at the connection point.
The results, presented in Figs. 56 and 57, show very specific fringe
patterns for each of the two cases. This proposed secondary struc-
ture is very similar to the one indicated by Noller et al. [77] for
RNA, in which the connections between the different fibers are
done without matter surplus but through a splitting and a prolonga-
tion of the two strands.
Connections between the fibers could be very dense; in other words
in some places the length between two connections could be very
small, as indicated by arrows in Fig. 55.
2 . Supercoiled structures
The diameter of some fibers could be greater than 2 nm. Moreover,
they can present a very complicated profile as illustrated by Fig.
ELECTRON FIELD EMISSION FROM ATOM SOURCES I39
a b
FIGURE57. Four-branch connection point with extra material showing a diffraction
image very different from Fig. 56. (a) Calculated Fresnel diffractogram (A = 0.7 A, point
source-object distance = 500 A); the mask of the object, not at the same scale, is shown
as an insert. (b) FPM image of a connection with extra material showing fringes inside the
connection zone. (From ref. [75].)
58. This conformation suggests a supercoiled structure. This is

strongly comforted by the presence of a smaller-diameter fiber
extending out from these features.
When a fiber is cut (Fig. 59), it develops a crooked end at the free
extremity. This free-end equilibrium conformation is very different
from the clew structure developed by synthetic organic polymers
as shown in Fig. 53C.
Loop structures can also be observed. This conformation can be
inside a network (Fig. 55) or as a free-standing loop along a fiber
(Fig. 60). To assess the hypothesis of a nanoloop, a Fresnel simula-
tion has been performed. Results show good agreement between
the simulated diffractogram and the observed image. Such loop
structures were also proposed to depict the super twisted aspect
of RNA analyzed by electron microscopy [78].
6. Discussion. From the point of view of electron microscopy, the
following points should be stressed.
Coherence. The whole beam area is covered with sharp diffraction
patterns. This is an experimental indication of the high coherence of the
field emitted beam from nanotips at 200-300 V . Moreover, these patterns
A B
FIGURE58. (A) Low-magnification FPM image of an RNA network; the arrow indicates
the presence of a supercoiled fiber. (B) High-magnification FPM images of the supercoiled
fiber of the framed zone. (a) The initial structure; the arrow indicates a small-diameter fiber
extending from the supercoiled conformation. (b) The same fiber modified after e-beam
irradiation, showing a decrease in its diameter. (From ref. [751.)
are indications that nanometer fibers have to be considered as opaque

objects under the e beams in this energy range.
High Resolution. The nanometric resolution of the images obtained
with the FPM is within the theoretical limit for the visual detectability of
small objects in a statistically noisy image. Using the Rose equation [79],
which for the purpose of electron microscopy is do 2 51C(fn)’’2,where
do is the characteristic object size, C is the contrast factor relative to the
immediate surroundings (in our case l), f is the efficiency of “electron
utilization” (assumed to be l), and N is the number of incident electrons
per unit area. The images were taken with an exposure time of i s, so do
is in the range of 2-3 A.
FIGURE 59. (A) Low-magnification FPM image showing a cut RNA fiber with a crooked
free end. (B) The FPM image of the free-end frame zone (a), and comparison with the
calculated Fresnel diffractogram of a crooked-end fiber shown in the insert. (C) From (a)
to (c), FPM image evolution of the free-end conformation under increasing e-beam irradiation
doses. (From ref. [75].)
C
B
FIGURE60. Loop structure along an RNA fiber. (A) Low-magnification FPM image.
(B) Detail of the loop which is framed (a), and comparison with the Fresnel diffractogram
(b) (A = 0.7 A, point source-object distance = 500 A); the insert represents the loop mask,
not at the same scale, that was used for the simulation. (From ref. [75].)
Magnetic Stray Field. The sharp diffraction figures obtained are exper-
imental proofs that the projection microscope using a nanotip as coherent
nanosource does not need magnetic protection in order to perform Fresnel
diffraction. This is confirmed by the following estimation of the image
blurring due to the stray magnetic field. The measured permanent magnetic
field is about 0.5 gauss (-0.5 X tesla) with AC stray field B(o)in
the range of 1 to 5 milligauss -5 x tesla) near the microscope
chamber. Under these experimental conditions, simple calculations [80]
of the deviations of the image at the screen by the Lorentz force effects
and/or the change of the phase due to the vector potential due to the stray
fields gives
A(i) = 2 x lo2 x B (45)
with A(i) in meters and B in tesla. For the measured range of the stray
field B ( o ) , the deviations are from 20 to 100 pm. They are substantially
smaller than the fringe widths at the screen, which were in the millimeter
range. Thus the blurring will not prevent the observation of the interfer-
ence fringes, as is fully supported by the experimental results.
Irradiation Effects. Irradiation effects are consequences of collisions
between the incoming electrons and the atoms of the specimen. Different
main processes can be envisaged:
Elastic scattering, the atom remains in its ground state, and the elec-
tron conserves its energy but changes direction
Inelastic scattering that excites the atom
Inelastic scattering that ionizes the atom
Capture of the incident electron by the atom, followed by a multielec-
tron excitation as in the Auger process, for example
The primary damage process is inelastic scattering, which causes either
molecular excitation or ionization. The energy dissipated is either con-
verted to molecular vibrations with temperature increases or causes bond
scissions as the loss and diffusion of hydrogen and the production of
radicals. The damage depends on the energy dissipated in the specimen
per unit volume (J ~ m - or~ electron
) dose ( q = f7 = e n , in C cm-2), which
is proportional to the number of incident electrons per unit area. However,
the knowledge of the individual damage processes is very poor because
the range of primary and secondary processes is very broad and complex.
For practical electron microscopy, damage processes and in particular
the loss of mass can be observed by following the evolution of the images
or, in our case, the diffraction patterns under the irradiation.
From the values of the currents and the dimensions of the objects,
during the observations of the polymers (PS-PVP) the electron exposures
were in the range of 1016-10*7electrons cm-2 s-’. For these polymers, no
damage or charging effects were noticed during 15 min- to 1-h-duration
observations, i.e., for an electron dose in the range of lot9to 3.6 X 10”
electrons cm-*.
Observations of the RNA show also that low-magnification imaging

could generally be carried out for extended periods of hours without
apparent evolution of the fiber structures. The transformations under
irradiation happen only when the object is close enough to the
nanotip, in other words, at high magnification, which induces a greater
irradiation flux. In practical observations, this means that the modifica-
tions are observed when the irradiated sample area covered by the beam
opening is inside a circle of diameter of less than 50 nm. When these
transformations are observed on the screen, the field emission from the
nanotips becomes very unstable, thus indicating the presence of adsorbed
species on the nanotip. This last phenomenon is an indication of the
presence of an evaporation of materials from the fiber and during its
morphological changes. In other words, the transformations observed are
accompanied by a matter loss from the fibers. The images in Figs. 58-59
are examples of the observed modifications of RNA general structure
under irradiation. Figure 58 shows a supercoiled structure becoming thin-
ner under the irradiation. Modifications can be observed more clearly
when the free extremity of a fiber is exposed to irradiation, as shown in
Fig. 59.
Studies by electron diffraction and by electron loss spectroscopy 181,
821 indicate, for complete destruction of the different bases of the nucleic
acid (adenine, cystosine, guanine, and thymine), electron doses in the
range of 5 X to 5 X lo-’ C cm-* are needed and with an incident
energy of the order of 20 keV at 300 K. This means an irradiation dose
of 3 X lOI7 to 3 x 10l8 electrons cmP2.Very few studies exist for energy
in the range of 50 to 300 V, but let us use these values and some assump-
tions to interpret the results. In the experimental procedure, the total
current of the incident beam is of the order of A. Let us assume
now that the destruction doses are deposited within 1 s. This means that
with the total current of lo-’’ A, the circular area irradiated by the e beam
must be smaller than a circle of -150 nm (for a 3 x 10’’ dose) to 50 nm
in diameter (for a 3 x 10l8 dose). These values are in the range of our
experimental observations.
A strict comparison between the given values and the experimental
observations is not realistic because the differences, first in energy of the
incident beam and second in the energy transfer process to the sample
itself during the irradiation, which depends strongly on the specimen
supporting device. For example, in FPM the RNA fibers are standalone
nano-objects, while in the other experiments the fibers are deposited or
embedded on solid substrate. This comparison tells us that the organic
fibers during FPM observation undergo irradiation damages, due to the
incident electron beam, only from a certain threshold flux.
E. Ferromagnetic Nanotips: Atomic Beam Splitter
As the last application of nanotips, we present now the field electron

emission and atomic metallic ion emission (AMIE) studies of Fe nanotips.
This was performed to investigate the possible magnetic interactions and
beam properties from ferromagnetic nanotips [83]. The Fe tips are obtained
from (1 11) or (1 10) Fe whiskers by an electrolytic sharpening. The experi-
mental results show unique emission properties specific to the magnetic
atomic-scale protrusions. In particular, in the case of FE, the electron
beam from an Fe nanotip splits when the temperature is lowered from
above T, (1042 K) to below T,, and this process is reversible upon reheating
to above T,. In addition, the AMIE patterns coming from single protrusions
consisted of sharp, multiple spots. Both of these effects were not found
with nonmagnetic tips. These results show that magnetic nanotips can be
used as an atomic-scale beam splitter for electrons and metallic ions when
operated for temperatures under T,. Possible explanations of these phe-
nomena are discussed below.
The observed FEM patterns from Fe nanotips are presented in Fig. 61
in two consecutive sequences: cooling (a to e), then heating (f to j). The
FE voltage in this example is -2870 V. The experimental procedure
was as follows. A stable single spot pattern coming from a nanoscale
protrusion was obtained at -1100 K; the heating current was then
cut, leading to the cooling of the tip while the FEM patterns were
recorded on the video. The sequence, Fig. 61a to 61e, are examples
illustrating the variation of the FE pattern during the cooling of the nanotip
from -1100 K (Fig. 61a) to liquid nitrogen temperature (Fig. 61e). This
sequence shows a progressive splitting of the initial singlet spot into a
stable doublet during the temperature decrease. The tip was then heated
back up to -1100 K. Figures 61f to 61j show a progressive merging of
the doublet into one singlet spot during the heating sequence. The splitting/
merging from a singlet toward a doublet spot are reversible processes.
In detail:
1. The splitting/merging of the electron beam(s) occurs when the na-

notip temperature crosses a critical temperature which is near the
Curie temperature. This temperature is well under the crystallo-
graphic phase-transition temperature Ta-y, which was checked by
FEM observations of an abrupt change in the FE patterns with
standard Fe microtips of approximately 100-nm radius.
2. The splitting/merging rates depend on the coolinglheating speed.
3. During the splitting/merging process, the two beams are not of equiv-
alent intensity, although at the final temperature the intensities of
e f
FEM observations of Fe nanotip
process
FIGURE 61. Evolution of the field emission pattern from an Fe nanotip as a function
of temperature. (a)-(e) Cooling sequence from (a) - 1100 K to (e) liquid nitrogen temperature.
(f)-(g) Consecutive heating sequence from (f) liquid nitrogen to (j) 1100 K. The splitting
and merging are reversible and are illustrated schematically on the right-hand side. (From
ref. [83].)
the two beams could be very similar (i.e., Fig. 61e). The same asym-
metry in the intensity occurs during the splitting and the merging.
4. As long as the protrusion has a certain height and the structure
underneath is not destroyed, the process is reversible with tempera-
ture without measurable change in the localization of the spot.
Fe - AMIE, sextuplet spot

FIGURE62. Sextuplet AMIE spot from an Fe nanotip for two different extracting volt-
ages with the schematic illustration of the six bright spots which constitute the pattern
(AMIE voltages were, respectively, -9.5 kV and -10.5 kV at -800 K).(From ref. [83].)
5 . The splitting angle of the doublet depends on the temperature of the

nanotip, with a maximum in the range of 4-6".
It is important to note that none of these observations took place for
W and Au nanotips, nor for standard magnetic Fe microtips. In this last
case, this may be because the splitting of the beam occurs only for the
atomic-scale geometry of the magnetic protrusion, or its observation is
masked by the spatial resolution limit of the magnetic microtips, which
is 2 nm [6].
There are also effects on the AMIE beams that are unique to the mag-
netic nanotips, which could have the same physical cause as for the
electron beams. The big difference with W or Au are the particular patterns
of the Fe AMIE spots (Fig. 62). Figure 62 shows specific Fe AMIE pat-
terns. A characteristic one is a triangular sextuplet spot shape with sharp
edges. The opening angle between two extremities of the triangle is -6",
and between each spot it is -2". The intensity of the triplet inside is higher
than for the triplet at the outskirt. This difference in intensity between
the internal and the external triplets can be easily distinguished when the
AMIE voltage is lowered. This is illustrated in Fig. 62. For W and Au,
only single spots with opening angles of 2-4" were observed for the whole
AMIE temperature range (see Fig. 21).
The multiplet-spot AMIE can move due to a gradual displacement of
the protrusions over the surface of the base tip. During their displacements
their whole initial patterns are conserved. The progressive movement of
two AMIE multiplet-spots toward each other can lead to a partial overlap-
ping of their patterns. The whole initial patterns for both the AMIE spots
were conserved even during their partial overlapping. This characteristic
b
FIGURE63. Sequence showing the progressive moving toward each other of two sextu-
plet Fe AMIE spots (-10.5 kV, -800 K). The sextuplet pattern of each spot is conserved
during the displacement and the partial overlapping. The duration of this sequence is a few
minutes. (From ref. [83].) (a) Fe-AMIE: experimental observations of the overlapping of
2 sextuplet spots. (b) Schematic representation of the above overlapping process.
is illustrated in Fig. 63, in which two sextuple-spot AMIE patterns move

toward each other until their partial overlapping.
To explain these data, two general mechanisms, both based on the
magnetic phase transition, have been proposed. Interpretations based on
crystallographic phase transition has not been taken into account, not
only because 1100 K is well under To-,,, but also that the splitting/merging
are progressive processes during the temperature variation, which is in
contradiction with crystallographic phase transition, which has to be
abrupt. These two proposed mechanisms are as follows.
1. Geometric interpretation. The splitting of the e beam is due to the
reversible formation of two protrusions during the magnetic phase
transition, and AMIE multiplet spots are emitted from an aggregate
of protrusions. The aggregation or splitting of the protrusions are due
to a rearrangement of the structure caused by magnetic interaction
for temperatures under T,. FIM analysis did not give unequivocal
answers because of the difficulties of getting stable and controlled
progressive field evaporation images. However, we find this hypothe-
sis difficult to handle in view of the classical weak magnetic energy
(order of meV) versus the structure modification energy (order of
eV). Furthermore, the conservation of the AMIE sextuplet patterns

during their movement and also during their partial overlapping (Fig.
63) would not occur in the case of aggregates of protrusions.
2. Magnetic interpretation. First of all, calculations show that classical
explanations taking into account the deviation of the emitted beams
by the magnetic field of the bulk tip can be ruled out. The beam
deviations in that case are orders of magnitude too small and also
cannot produce the observed patterns. This leads to the conclusion
that the effect must be due to a very strong magnetic interaction at
the atomic scale. This could be either a scattering process, a magnetic
diffraction (Aharonov-Bohm like) [84], Stern-Gerlach-like spin selec-
tion [ 8 5 ] , or Lorentz force beam deviations under very large and
localized magnetic gradients existing at the apex of the ferromagnetic
nanotips. The different patterns observed then reflect the magnetic
state of the different particles (ions or electrons), the 3D field distribu-
tion at the atomic scale apex region, and the nature of the interaction.
At present we cannot assess which of these mechanisms, alone or
in concomitance, are the proper ones for explaining the presented
phenomena.
V. CONCLUSIONS
The size reduction to one atom of the field emission area, which is the
main characteristic of the nanotips, is obtained by taking advantage of
the protrusion effect to enhance locally the field over the topmost atom
of the nanoprotrusion. The field emission beam, which comes exclusively
from this atom, manifests specific properties that are attached to the
atomic size of these nanosources. Applications of the intrinsic physical
properties of the nanotip lead to the possibilities of having atomic resolu-
tion under FEM, or to measure the energy exchange down to atom size
area. The nanotip is a coherent, monochromatic e-beam source. The use
of the nanotip as a point source in a projection microscope transforms
it into a versatile, low-energy, high-resolution electron microscope: the
Fresnel projection microscope.
Most of the main physical properties of the nanotips are explained by
taking into consideration the physical mechanism of electron tunneling
through an atom with a field-deformed triangular barrier. However, some
properties observed experimentally are still under consideration such as,
for example, the splitting of the e beams and the AMIE with ferromag-
netic nanotips.
In this chapter the nanotips are presented as e-beam nanosources. They

are also AMIE sources, which means metallic ion sources, with all the
specific properties attached to their atomic dimensions (Fig. 20). Further-
more, among the applications which take advantage of these properties
and not quoted in this chapter are those related to the use of microguns.
Microguns are constituted of the integration of the nanotips, used as
atom sources of electrons or ions, inside microlens systems [86], with the
advantages of a drastic reduction of the size and the aberrations. These
microguns can be standalone field emission gun systems or inside an
array used, for example, as tools for parallel nanowriting or metallic
nanodeposition.
ACKNOWLEDGMENTS
It is a pleasure to thank V. Semet for his important participation in this

work as well as L. Bitar for his contributions and fruitful discussions.
The contributions of Dr. R. Semet, Dr. Pham Quang Tho, and Prof. E.
Taillandier for the choice of the samples (polymer and RNA) and discus-
sions about the Fresnel projection microscope images are highly appreci-
ated. We acknowledge the technical assistance from the Service Central
d’Analyse du CNRS-DCpartement Instrumentation. This work has been
supported by European Union Contracts (SCIENCE, HCM, and BRITE),
by French and Spanish government agencies.
REFERENCES
1. A. V. Crewe, Conf. on Non-Conventional Electron Microscopy, Cambridge, England

(1965); A. V. Crewe, J. Walls, and L. M. Welter, J. Appl. Phys. 39, 5861 (1968).
2. Vu Thien Binh, J. Microsc. 151,355 (1988); Vu Thien Binh and J. Marien, Surface Sci.
202, L539 (1988).
3. Vu Thien Binh and N. Garcia, J. Physique I1,605 (1991);Vu Thien Binh and N. Garcia,
Ultramicroscopy 42-44, 80 (1992).
4. E. W. Muller, Ergeb. Exackt. Naturwiss. 27, 290 (1953).
5 . (a) L. W. Swanson and A. E. Bell, in “Advances in Electronics and Electron Physics,”
XXIII, L. Marton (Ed.), p. 193, Academic Press, New York (1973); (b) A. Modinos,
“Field, Thermionic and Secondary Electron Emission Spectroscopy,” Plenum Press,
New York (1984).
6. R. H. Good and E. W. Muller, in “Handbuch der Physik,” XXI, p. 176, Springer
Verlag, Berlin (1956).
7. R. Gomer, “Field Emission and Field Ionisation,” Harvard Univ. Press, Cambridge,
Mass. (1961).
8. E. W. Muller and T. T. Tsong, “Field Ion Microscopy, Principles and Applications,”

Elsevier, Amsterdam (1969); E. W. Muller and T. T. Tsong, Prog. Surface. Sci. 1,
l(1974).
9. W. Schottky, Z. Physik 14, 63 (1923).
10. R. H. Fowler and L. Nordheim, Proc. Roy. SOC.Lond. A 119, 173 (1928).
1 I. L. Nordheim, Proc. Roy. Sac. Lond. A l21, 626 (1928); H. C. Miller, J . Franklin Inst.
282, 382 (1966).
12. E. L. Murphy and R. H. Good, Phys. Rev. 102, 1464 (1956); S. G. Christov, Phys.
Status Solidi 17, 1 I (1966).
13. H. Boersch, Z. Phys. l39, 115 (1954).
14. R. D. Young, Phys. Reu. 113, 110 (1959).
IS. W. P. Dyke and W. W. Dolan, in ”Adv. in Electronics and Electron Phys.,” VIII, L.
Marton (Ed.), p. 89, Academic Press, New York (1956).
16. Vu Thien Binh, A. Piquet, H. Roux, R. Uzan, and M. Drechsler, Surface Sci. 25, 348
(1971); Vu Thien Binh and R. Uzan, Surface Sci. 179, 540 (1987).
17. C. F. Eyring, S. Mackeown, and R. A. Millikan, Phys. Rev. 31, 900 (1928).
18. J. A. Becker, Bell System Tech. J . 30, 907 (1951).
19. D. J. Rose, J. Appl. Phys. 27, 215 (1956).
20. A review is given in P. W. Hawkes and E. Kasper, “Principles of Electron Optics,”
Vol. 2, Applied Geometrical Optics, Academic Press, London (1989).
21. D. M. Goebel, Y. Hirooka, and G. A. Campbell, Reu. Sci. Instrum. 56, 1888 (1985).
22. D. W. Tuggle and L. W. Swanson, J . Vac. Sci. Techn. B3,220 (1985).
23. A. N. Broers, J. Appl. Phys. 38, 1991 (1967).
24. Vu Thien Binh, A. Piquet, R. Uzan, and M. Drechsler, Rev. Phys. Appl. 5,645 (1970).
25. E. W. Muller, Z. Phys. 106, 132 (1937); A. P. Janssen and J. P. Jones, J. Phys. D: Appl.
Phys. 4, 118 (1971); H. W. Fink, IBM J . Res. Deveiop. 30, 460 (1986).
26. L. W. Swanson and L. C. Crouser, in G. A. Somorjai (Ed.), “The Structure and
Chemistry of Solid Surfaces, p. 60-1, Inorganic Materials Research Division Series,
John Wiley, New York (1969).
27. E. W. Plummer and R. D. Young, Phys. Rev. B 1, 2088 (1970).
28. L. W. Swanson and L. C. Crouser, Surf. Sci. 23, 1 (1970).
29. H. W. Fink, Phys. Scr. 38, 260 (1988).
30. Ch. Kleint and K. Mockel, Surface Sci. 40, 343 (1973).
31. D. W. Tuggle, J. 2. Li, and L. W. Swanson, J. Microsc. 140, 293 (1985).
32. I. L.Sokolovskaia,J. Tech.Phys. (URSS)26,1177(1956);P. BettlerandC. Charbonnier,
Phys. Rev. 119,85 (1960). These authors were the first to apply an electric field to have
buildup tips.
33. (a) J. J. Saenz, N. Garcia, V u Thien Binh, and H. De Raedt, “Scanning Tunneling
Microscopy and Related Methods, NATO-AS1 Series E: Appl. Sci., Vol. 184, 409,
R. J. Behm, N. Garcia, and H. Rohrer (Eds.), Kluwer, Dordrecht (1990);(b) H. DeRaedt
and K. Michielsen, in “Nanosources and Manipulations of Atoms Under High Fields
and Temperatures: Applications,” NATO-AS1 Series E: Applied Sciences 235, 45. Vu
Thien Binh, N. Garcia, and K. Dransfeld (Eds.), Kluwer, Dordrecht (1993).
34. D. Atlan, G. Gardet, Vu Thien Binh, N. Garcia, and J . J. Saenz, Ultramicroscopy
42-44, 154 (1992).
35. Review papers on surface diffusion under different driving forces can be found in Vu
Thien Binh (Ed.), “Surface Mobilities on Solid Materials, Fundamental Concepts and
Applications,” NATO-AS1 Series B: Physics, Vol. 86, Plenum Press, New York (1983).
36. L. W. Swanson and L. C. Crouser, J. Appl. Phys. 40,4741 (1969); L. H. Veneklaasen
and B. M. Siegel, J . Appl. Phys. 43, 1600 (1972).
37. G. I. Taylor, Proc. Roy. SOC. Lond. A 280, 383 (1964).

38. R. Smouluchowski, Phys. Rev. 60,661 (1941).
39. P. Bettler and C. Charbonnier, Phys. Rev. 119, 85 (1960).
40. G. Neumann and G. M. Neumann, in “Surface Self Diffusion of Metals,” Diffusion
Monograph Series, No. 1, F. H. Wohlbier (Ed.) (1972) USA; J. G. Dash, Contemp.
Phys. 30, 89 (1989).
41. VuThien Binh, S. T. Purcell, N. Garcia, and J. Doglioni, Phys. Reu. Lett. 69,2527(1992).
42. C. E. Kuyatt and E. W. Plummer, Reu. Sci. Instrum. 43, 108 (1972).
43. P. Serena, L. Escapa, J. J. Saenz, N. Garcia, and H. Rohrer,J. Microscop. 152,43 (1988).
44. H. De Raedt, Comp. Phys. Rep. 7, 1 (1987).
45. N. Garcia, J. J. Saenz, and H. De Raedt, J. Phys.: Condens. Mutter 1, 9931 (1989).
46. C. B. Duke and M. E. Alferieff, J . Chern. Phys. 46, 923 (1967).
47, J. W. Gadzuk, Phys. Reu. B 1,2110 (1970).
48. J. W. Gadzuk and E. W. Plummer, Rev. Mod. Phys. 45, 487 (1973).
49. See, for example, C. Kittel, “Introduction to Solid State Physics,” Wiley, New York
(1968).
50. (a) F. Gautier, H. Ness, and D. Stoeffler, Ultramicroscopy, 42-44, 91 (1992); (b) H.
Ness, Thesis, UniversitC Louis Pasteur, Strasbourg, France (1995); (c) H. Ness and F.
Gautier, J . Phys. Condens. Mutter 7, 6625 (1995).
51. C. J. Chen, Phys. Reu.Lett. 65, 448 (1990); 69, 1656 (1992).
52. N. Garcia, Vu Thien Binh, and S. T. Purcell, Surface Sci. Lett. 293, L884 (1993).
53. P. H. Cutler, J. He, N. M. Miskovsky, T. E. Sullivan, and B. Weiss, J . Vuc. Sci.
Technol. B 11(2), 387 (1992).
54. (a) Vu Thien Binh, N. Garcia, S. T. Purcell, and V. Semet, in “Nanosources and
Manipulations of Atoms Under High Fields and Temperatures: Applications,” NATO-
AS1 Series E: Applied Sciences, Vol. 235, p. 59, Vu Thien Binh, N. Garcia, and K.
Dransfeld (Eds.), Kluwer, Dordrecht (1993); (b) S. T. Purcell, Vu Thien Binh, and N.
Garcia, Appl. Phys. Lett. 67, 436 (1995).
55. R. D. Young and C. E. Kuyatt, Rev. Sci. Instrum. 39, 1477 (1968).
56. W. B. Nottingham, Phys. Rev. 59, 908 (1941).
57. L. W. Swanson, L. C. Crouser, and F. M. Charbonnier, Phys. Reu. 151, 327 (1966);
for a review on energy exchange during FE, see ref. pa].
58. Vu Thien Binh, S. T. Purcell, G. Gardet, and N. Garcia, Surface Sci. 279, L197 (1992).
59. J. R. Chen and R. Gomer, Surfuce Sci. 79,413 (1979).
60. S. T. Purcell, Vu Thien Binh, N. Garcia, M. E. Lin, R. P. Andres, and R. Reifenberger,
Phys. Rev. B 15, 17259 (1994).
61. Vu Thien Binh, V. Semet, and N. Garcia, Appl. Phys. Lett. 65, 2493 (1994).
62. Vu Thien Binh, V. Semet, and N. Garcia, Ultramicroscopy, (1995) in press; Vu Thien
Binh, N. Garcia, and V. Semet, Phil. Trans. R . SOC.Lond. A 350 (1995) in press.
63. G. A. Morton and E. G. Ramberg, Phys. Rev. 56,705 (1939).
64. E. W. Muller, 15th Field Emission Symposium, Bonn (1968).
65. For a review of STM techniques, see C. Julian Chen, “Introduction to Scanning Tunnel-
ing Microscopy,” Oxford Series in Optical and Imaging Sciences, Oxford Univ. Press,
New York (1993).
66. H. W. Fink, W. Stocker, and H. Schmid, Phys. Rev. Lett. 65, 1204 (1990);J . Vuc. Sci.
Technol. BS, 1323 (1990).
67. J. C. H. Spence, W. Qian, and A. J. Melmed, Ultrumicroscopy 52,473 (1993).
68. R. D. Young, Reu. Sci. Instrum. 37, 275 (1966).
69. N. Garcia and H. Rohrer, J . Phys. Condensed Mutter 1, 3737 (1989).
70. D. Gabor, Nature 161,777 (1948); P. W. Hawkes and E. Kasper, “Principles of Electron
Optics, Vol. 3, Wave Optics,” Academic Press, London (1994).
71. Joseph W. Goodman, in “Introduction to Fourier Optics,” McGraw-Hill Physical and

Quantum Electronics Series, McGraw-Hill, New York (1968).
72. H. W. Fink, H. Schmid, H. J. Kreuzer, and A.Wierzbicki, Phys. Rev. Leu. 67, 1543
(1991); H. J. Kreuzer, K. Nakamura, A. Wierzbicki, H. W. Fink, and H. Schmid,
Ultramicroscopy 45, 381 (1992).
73. J. C. H. Spence and W. Qian, Phys. Reu. B 45, 10271 (1993).
74. G. M. Shedd, J . Vac. Sci. Technol. A 12, 2595 (1994).
75. Vu Thien Binh, L. Bitar, V. Semet, N. Garcia, and E. Taillandier, submitted.
76. See, for example, D. Voet and J. G. Voet, “Biochemistry,” Wiley, New York (1990).
77. H. F. Noller, Annu. Reu. Biochem. 53, 134 (1984); R. R. Gutell, B. Weiser, C. R.
Woese, and H. F. Noller, Prog. Nucleic Acid Res. Mol. Biol. 32, 183 (1985).
78. W. R. Bauer, F. H. C. Crick, and J. H. White, Sci. Am. 243, 118 (1980).
79. A. Rose, Ado. Electronics 1, 131 (1948); “Vision: Human and Electronics,” Plenum
Press, New York (1973); R. M. Glaser, “Introduction to Analytical Electron Micros-
copy,” p. 423, Plenum Press, New York, 1979.
80. R. P. Feynrnan, R. B. Leighton, and M. Sands, “The Feynman Lectures in Physics,”
Vol. I1 Addison Wesley, London (1964).
81. L. Reimer and J. Spruth, J. Microsc. Spectr. Electron 3, 579 (1978).
82. A. V. Crewe, M. Isaacson, and D. Johnson, in “Proc. 28th Annual Meeting of EMSA,
Baton Rouge, La. (1970), p. 264, Claitor’s Publ. Div.
83. Vu Thien Binh and N. Garcia, Surface Sci. 320, L69 (1994).
84. S. Olaru and I. Iovitzu Popescu, Reu. Mod. Phys. 57, 339 (1985).
85. See N . F. Mott and H. S. W. Massey, “Theory of Atomic Collisions,” p. 210, Clarendon
Press, Oxford (1965); J. Kessler, “Polarised Electrons,” Vol. 1 , Springer Verlag, Ber-
lin (1976).
86. D. Pribat, Vu Thien Binh, and P. Legagneux, Electrode de focalisation inttgrte pour
r6seaux de microcathodes a effet de champ et proctd6 de fabrication, Patent 90-
14287 (1990).

Binh1996 PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Binh1996 PDF

Uploaded by

Copyright:

Available Formats

ADVANCES IN IMAGING AND ELECTRON PHYSICS. VOL.

Electron Field Emission from Atom-Sources: Fabrication,

A major instrumental development in the history of electron optics and

Within a metal, an electron current density of roughlyj, = enou, impinges

giving an effective work function in the presence of the applied field of

Extracting electrons from metal surfaces can be done by two main

wheref(p) = 1/{1 + exp[(E - E F ) / k B T ]is} the Fermi-Dirac distribution

11. Field Emission (FE)

where m is the mass of the electrons, then

The thermionic emission condition for electrons of the metal is

Equation (9) is the Richardson-Dushman equation. Some parameters for

Work function 4 (eV) 4.5 2.4-2.7 2.8

Fo (V/cm) 105 1 06 107 108

which permits an emission current density of more than two orders of

Fo (V/cm) 1 os 106 10’ 3 x 10’ 108

Substituting D(E,) from Eqs. (1 1) into the integral gives

+=2eV Fo = 1.5 x lo7 V/cm F, = 1.8 x lo7 Vlcm

106-107 A/cm2 correspond to a transmission probability of the deformed

3 . I-V Characteristics: The Fowler-Nordheirn Equation

where A is the emitting surface area, K and p are geometric factors

with p in AT', A in A2, and I in A.

(i.e., T I 1700 K for W),

which implies a negligible thermal tail in the electron distribution at V =

where I, and I are, respectively, the currents at temperatures of 0 (K)

C . Energy Distribufion of Emitted Electrons

The half-width (FHWM) of this distribution is BE = 2.45kBT with a

The roughness of the cathode emitting area

and the half-width at T = 0 K is given by

D . Current Density Distribution

An important parameter for the use of emitted electron beams is the

Energy Relative to EF (eV)

In practice, p has to be calculated by taking in account the exact geometry

or the paraboloidal approximation [ 181

increase of the current is due to the formation of multiple localized emitting

Energy Relative to EF (eV)

tion process, some FEGs have a working temperature in the range of

Passing from thermionic to field emission cathodes principally allows re-

Microscopic tips Nanotips

Emitting area, A 2nrz ( 1 - cos 0,) = 7rr? Apex atom

one atom will be discussed in order to explain these specific properties.

A . ExperitnPntal Setup and Procedures

a base vacuum of -5 X lo-'' torr and has a controlled gas introduction

were measured with a commercial 135" hemispherical energy analyzer

B . Confinement of the Field Emitting Area

The first problem to be confronted in improving tip performance is the

The results of the studies of these mechanisms presented in the following

be thermally cleaned, so the adsorbed layer on the shank diffuses

2. Local Work Function Decrease

3 . In Situ Field Sharpening

The basic mechanism for the protrusion formation will be considered

6 . Tip Sharpening in the Presence of Applied Field. In this approach,

direction of surface diffusion. As this driving force is effective only if it

Teton Tips or Nanotips [2,31. To obtain exclusive FE from the protru-

can be made by considering the activation energy for surface diffusion in

where a is the jumping distance of diffusing atoms taken to be the unit

1 Atomic Metallic Ion Emission

FIGURE20. Schematic of the field- and temperature-driven formation of nanometric

As mentioned above, 3D calculations of the field distribution for teton

11. (a) 11. (b) 11. (c)

C . Field Emission Characteristics from Nanotips: Experiment

1. Beam Opening Angle

2 . Field Emission Stability

W< 11 I> Build-up Tip