REPORT DOCUMENTATION PAGE OMB No. · To measure the tribological properties of surfaces under well defined conditions, a new ... Friction Between a Diamond Tip and Polycrystalline

UN CLASSIFIEDS:CURITY CLASSIFICATION OF THIS PAGE

ForM ApproveddREPORT DOCUMENTATION PAGE OMB No. 0704-0188

la. REPORT SECURITY CLASSIFICATION lb. RESTRICTIVE MARKINGS

UNCLASSIFIED2- cr ... 3. DISTRIBUTION /AVAILABILITY OF REPORT

Approved for public release; 4,SAD- A264 647 distribution is unlimited.

4 5. MONITORING ORGANIZATION REPORT NO(SN4.FS

_____ ____ ____ ____MCI_ MESR 9 3 0 f6a. NAME OF PERFORMING ORGANIZATION 16b. OFFICE SYMBOL &a. NAME OF MONITORING ORGANIZATIO

International Business (If applicable) AFOSR/NCMachines Corp I

6c. ADDRESS (City, State, and ZIP Code) 7b. ADDRESS(City, State, and ZIP Code)

Building 410, Bolling AFB DCSan Jose, CA 95120 20332-6448

8a. NAME OF FUNDING/SPONSORING 8b. OFFICE SYMBOL 9. PROCUREMENT INSTRUMENT IDENTIFICATION NUMBER

ORGANIZATION (if applicable)

AFOSR INC F49620-89-C-006887c. ADDRESS (City, State, and ZIP Code) - 10. SOURCE OF FUNDING NUMBERS

Building 410, Bolling AFB DC PROGRAM PROJECT TASK WORK UNIT20332-6448 ELEMENT NO. NO. NO ACCESSION NO.

61102F 2303 A2

I (.JTTLE (include SecurityClassification) Fundamental Studies of Friction, Lubrication, and Wear byAtomic Force Microscopy

12. PERSONAL AUTHOR(S) I I f

13a. TYPE OF REPORT 13b. TIME COVERED 114. DATE OF REPORT (Year, Month, Day) t S. PAGE COUNT

/Finall FROM 7/1/89 TO 9/30192 1993, April, 14 516. SUPPLEM ENTARY NOTATION 9 3 1 1 9

17. COSATI CODES i8. SL i3e1t2 5 b k numb1er)FIELD GROUP $SU-GROUP

19. ABSTRAC To probe friction, lubrication, and wear on an atomic scale, a capacitance-based ultrahigh vacuum

force microscope was developed. It measures both parallel and perpendicular forces between a tip and a surface.Friction between a diamond tip and chemical-vapor-deposited (CVD) diamond films at loads below onemicronewton showed a sublinear dependence of friction on load, with the effective friction coefficient varyingbetween 0.2 and 0.8 depending on load and location. Stick-slip behavior resulting from both surfaceheterogeneity and static/dynamic friction was observed. A CVD process was developed for growing single crystaldiamond tips with radii as small as 30 nm. The normal force between these tips and diamond (100) and (Il 1)surfaces agrees with calculated dispersion forces. The frictional force variation on the (100) surface are tentativelyassociated with a reconstructed geometry convoluted over an asymmetric tip shape, while the (11) surface ex-hibits features which cannot be simply associated with the surface structure. Friction is approximately 3nanonewtons independent of loads up to 100 nanonewtons. A field emission technique was developed for con-tinuously observing the motion of individual adsorbed atoms and molecules. The hopping of individual Cs atomsbetween sites on a tungsten tip was observed with 2 ps and 5 A resolution.

20. DISTRIBUTION / AVAILABILITY OF ABSTRACT 21. ABSTRACT SECURITY CLASSIFICATION

OIUNCLASSIFIED/UNLIMITED 0 SAME AS RPT. 0 DTIC USERS UNCLASSIFIED

22a. NAME OF RESPONSIBLE INDIVIDUAL 22b. TELEPHONE (Include Area Code) 22c. OFFICE SYMBOL

Captain Thomas E. Erstfeld (79O 797-49j5 1d/lOD Form 1473, JUN 86 Previous edirions are obsolete. SECURITY CLASSIFICATION OF THIS PAGE

9 3 5~1 0 4. ( UNCLASSIFIED

or

Final Report: Fundamental Studies of Friction, Lubrication, and ad

Wear by Atomic Force Microscopy ..........By ..............

Gary M. McClelland Distribution/IBM Research Division Availability CodesAlmaden Research Center I avip and/orSan Jose, CA 95120 Dist Special

Objectives ist ial

The objective of this work is to obtain an atomic-level understanding of t e nrct -tween well-characterized surfaces by using atomic force microscopy (AFM), scanningtunneling microscopy, and other surface analysis techniques.

Status of the Research Effort

(A comprehensive description of this work is given in the attached reprints and pre-prints.)

An Ultrahigh Vacuum Bidirectio;aal Capacitance AFMTo measure the tribological properties of surfaces under well defined conditions, a newtype of atomic force microscope was designed and constructed. This was the first vac-uum force microscope, the first microscope to use capacitance detection, and the firstmicroscope to measur-e forces on the tip both parallel and perpendicular to the samplesurface. A comprehensive mathematical analysis of the performace of the capacitancedetection system shows that the sensitivity is ultimately limited by the need to avoidsnapping of the lever into the sensor electrode by the attractive electrostatic force. Avery stable and sensitive capacitance-sensing electronic circuit was designed. Usingcurved wire cantilevers and flat plate capacitance sensors, a noise level from 0.01 to 1000I lz of 0.03 A RMS is achievable with the capacitance circuit, but the best performanceobserved was a factor of 3 higher, due to roughness of the cantilever.

Friction Between a Diamond Tip and Polycrystalline Diamond FilmsThe capacitance AFM was used to study the friction of a diamond tip on a polycrstallinediamond film at loads below 10.6 N. The tip was formed by fracturing a diamond in air.The film was grown by hot filament vapor deposition from a mixture of I % methanein hydrogen at a pressure of 6 kPa, resulting in crystallites 2- unm across. Experimentswere performed without cleaning the sample and tip at a pressure below 10-9 Torr.

Friction was measured on a a single crystallite where topographic measurements showeda roughness of several A over regions of several 100 A. The dependence of friction onload was sub-linear, as expected for elastic single-asperity contact. Stick-slip behaviorwas observed, which in some cases could be correlated with features on the surface, butin other cases appeared to arise from a difference between static and dynamic friction.The local static and dynamic friction coefficients could be deduced from the shape of thestick-slip behavior. Depending on the point of contact, static friction coefficients couldvary from 0,2 to 0.8, perhaps due to local crystal orientation.

Approvea ror' ptt4 , ruieulotdJ"tr"but"o• un±lmited.

Fabricating Diamond Force Microscope Tips by Chemical Vapor Deposition

Previous force microscope studies have used tips of uncertain geometry and surfacechemistry. We have developed a chemical vapor deposition method to manufacturesingle crystal diamond tips of known orientation and surface chemistry. The diamondsare grown onto the end of etch-sharpened tungsten wires on which a seed crystal is de-posited onto an etched-sharpened tungsten wire by carefully controlled contact with adiamond abrasive block. Diamond crystals t- Ipm in diameter are grown in a I 12/Cl 14mixture near a hot Ta wire. Of ten tips grown simultneously in a single holder, the bestis chosen using scanning electron, microscopy. Extremely low electron dosage is used toavoid damage to the hydrogen-terminated surface. This surface is very inert, and canbe cleaned of physisorbed contaminants by flash desorption. After the friction exper-iment, the tip is examined by high magnification SEM to determine its orientation andapex radius (typically 30 nm).

Atomic Scale Friction of a Diamond Tip on Diamond (100) and (111)SurfacesThe vacuum AFM was modified to incorporate a dual fiber force sensor using a noveltilted conliguration to avoid interference betwcen the fiber and sample. Friction andnormal forces were measured between single crystal diamond AFM tips and single crys-tal diamonds samples, both cleaned in UIIV. This is the first AIM study involving awell-characterized tip of known orientation. The attractive normal force between the tipand surface agreed well with calculated dispersion interactions. The frictional force ex-hibits periodic features, which on the (100) surface are tentatively associated with a2x I reconstructed surface convoluted over an asymmetric tip shape. The (I i) surfaceshows features which cannot be simply related to the surface structure. As the tip isscanned back and forth along a line, the same features are observed in each direction,but offset, suggesting the presence of a conservative force independent of the directionof motion as well as a non-conservative force. The friction is approximately C-3 x 10-9N, independent of loads up to I x 10-7 N. The differential friction coefficient for thissystem is less than 0.01 at the largest loads, and similarly small values have been meas-ured for oxidized tungsten tips sliding on diamond surfaces. These values areunprecedentedly small for unlubricated sliding in vacuum, suggesting that we are closeto attaining the zero friction limit predicted for incommensurate rigid sliding surfaces.

Observing the Motion of a Single Adsorbed Aton with Picosecond andSubnanometer Resolution

To characterize metal tips for AFM experiments, we became interested in applying fieldion and field emission microscopy. We soon realized that field emission provides aunique opportunity to continuously observe single adsorbed atoms and molecules on thespatial and temporal scales of molecular vibration. We developed a new instrument, thefemtosecond field emission camera, to record the time dependence of field emission froma sharp tip. In this instrument, the field-emitted electrons are tightly focussed into abeam, which is electrostatically swept across a detection screen as in a streak camera.

In our first experiments, hopping of single Cs atoms between sites on a W< I I>tip was observed. Atomic resolution field ion microscopy was used to form sharp tipsending in a plane of three atoms. Individual Cs atoms are deposited onto a 80 K tip,and a field emission image is recorded. The tip is pulsed negatively for 2 ns while the

2

focussed emission beam is swept across the screen. The increased field emission excitesthe Cs, which hops to a nearby site. This event is recorded with 2 ps resolution by thechanging field emission intensity. After the sweep, the new position of the Cs atom isreflected in the field emission image. By defocussing the e beam perpendicular to thesweep direction, simultaneous temporal and one dimensional spatial resolution of thedynamics is achievable. Because the accelerating field at the tip is so high, the time-of-flight spread of the field-emitted electrons, which determines our instrument's time re-solution, can be shorter than 100 fs. Work underway is aimed at observing thevibrational motion of individual adsorbed molecules.

Publications

"Tribology at the Atomic Scale," G. M. McClelland and S. R. Cohen, in Chemistry andPhysics of Solid Surfaces VIII, eds. R. Vanselow and R. Rowe (Springer, Berlin, 1990),p. 419.

"Force Microscopy with a Bidirectional Capacitance Sensor," G. Neubauer, S. R.Cohen, G. M. McClelland, D. ltorme, and C. M. Mate, Rev. Sci. Insirum. 61 2296(1990).

"Nanotribology of Diamond Films Studied by Atomic Force Microscopy," G. Neubauer,S. R. Cohen, G. M. McClelland, and H4. Seki, in Thin Films: Stresses and MechanicalProperties II eds. M. F. Doerner, W. C. Oliver, G. Pharr and F. R. Brotzen. MRSSymposium Proceedings 188, 219-224 (1990).

"Friction at the Atomic Scale," Gi. M. McClelland and J. N. Glosli, in Fundamentals ofFriction: Macroscopic and Microscopic Processes , eds. I. L. Singer and I-. M. Pollock,(Kluwer, Dordrecht, 1992), p. 405.

"Diamond Force Microscope Tips Fabricated by Chemical Vapor Deposition," G. J.Germann, G. M. McClelland, Y. Mitsuda, M. Buck, and It. Seki, Rev. Sci. Instrum. 63,4053 (1992).

"Atomic Scale Friction of a Diamond Tip on Diamond (100) and (11) Surfaces," G. J.Germann, S. R. Cohen, G. Neubauer, G. M. McClelland, II. Seki, and 1). Coulman, J.Appl. Phys. 73, 163 (1993).

"Observing the Motion of a Single Surface Atom with Picosecond and Sub-NanometerResolution," Ii. Heinzelmann, F. Watanabe, and Cy. M. McClelland, Phys. Rev. Lett.,in press.

Technical PersonnelPrincipal investigator: Gary M. McClelland.

Technician: Bruce A. I loenig.

Visiting Scientists (postdocs): Gabi Neubauer, Yoshitaka Mitsuda, I larry 1leinzelmann,Geoffrey Germann.

3

Papers Presented"Tribology at the Atomic Scale," Eighth International Summer Institute on SurfaceScience, Milwaukee, August 22, 1989.

"Nanotribology Using a 2D AFM with Capacitative Detection,' AVS Meeting, Boston,October 26, 1989.

"Tribology at the Atomic Scale," AFOSR Chemical Dynamics Contractors Meeting,Captiva Island, Florida, November 1, 1989.

"Friction at the Atomic Scale," U. of Akron Physics Seminar, November 29, 1989.

"Atomic Scale Tribology of Graphite and Diamond," AFOSR Tribology ContractorsMeeting, Dayton, Ohio, December 1, 1989.

"Tribology at the Atomic Scale," Sandia National Labs, February 9, 1990.

"Nanotribology of Diamond Films Studied by Atomic Force Microscopy," MRS Meet-ing, San Francisco, April, 1990.

"The Atomic Dynamics of Friction," Gordon Conference on Energetic Materials, June1990.

"The Atomic Dynamics of Friction," International Symposium on Industrial Tribology,Chicago, August, 1990.

"Friction at the Atomic Scale," West Coast Spectroscopy Conference, Asilomar, Febru-

ary, 1991.

"Tribology at the Atomic Scale," March APS meeting, Cincinnati, 1991.

"Friction at the Atomic Scale," Mclivain Lecturc at University of Wisconsin ChemistryDepartment, April 24, 1991.

"Friction at the Atomic Scale," NATO ASI on Fundamentals of Friction, Braunlage,Germany, July, 1991.

"Bidirectional AFM Study of Diamond-Diamond Friction in UlIV," International STMConference, Interlaken, August, 1991.

"Bidirectional AFM Study of Diamond-Diamond Friction in UI IV," AFOSR contrac-tors meeting, Irvine, October 22, 1991.

"Friction and Picosecond Dynamics at the Atomic Scale," 12th International Vac.Cong./8th Intl. Conf. on Solid Surfaces, The Ilague, Netherlands, October 12-16, 1992.

"Bidirectional Force Microscope Study of Diamond-Diamond Friction in UIIV," ACSmeeting, San Franciso, April 5-9, 1992.

4

"Bidirectional Force Microscope Study of 1)iamond-Diamond Friction in UIIV,' Inter-national Conference on Metalurgical Coatings and Thin Films, April 5-9, 1992.

"Direct Observation of Picosecond Dynamics of Individual Surface Atoms,' Interna-tional Field Emission Symposium, Halifax, August 11-15, 1992.

Patent"Subpicosecond Atomic and Molecular Motion Detection and Signal Transmission byField Emission,' G. M. McClelland, U. S. patent 5151594, issued Sept. 29, 1992.

5

m > u .0

~ 0 E tj C 0

be C CL u .~Csg ~ ~ I 1-1- 1

.1~C~ -S=0%;0ta 9 .. m41 30 0

S c ba 0

ba.,~ U l

:5 E

-0 0 ca .20.E0 :2 . - 0

0 M.-

MU

Sq

00 b oIS , :,. -

0) !2 0 L)Ec0 . u.Z>.g

u0 0 7 0~ 'c:ýd~

- 0-E1!IC

t-i ýd0 0. 95 3. 0.--

m

00 g

S- - 't~o .

=c 0 ' IQb u 0 -. v cc E - r-C C

-,

04 8 . r- 0E

6 ) 0 ) 6 ~C 6 . £ . C ' 0 0

0 Z- E0

0 0 Z4 Z; 8 6

Fa 64 .,- Z 0 64 =0 45. ~ . c9 r - ' 2 = W~ i5 6 ) S =Z 0

r> 0 4>-u 0786 o u u24 v6)

0 ) >c 9 6 ).' ... 1 1 c j

.E .~ I> 8'oI - 1

.3 c: 0 S. 4> > _u C6

be o.g §-215 ýo4 .Ji 300 0 . L E 0J - . .z 6.M

- - 0C -0 E

p~ - . -

5'oK E,

.t

0~H -s E a~

od - 0)qa tj 75U,1

C3. '1 i 1

-- o C

- 0.I -

,-o~ Uc :i --

00,S

0 A OU

fa t3 be EC60

00 10

du 0.- c~E 72..

Z3 -. .0 '4 .,v, u 71 2

.c a 52 - 1: g2 a! E ;

CQ z;o- 0. ~ 0* 4u

0 o E =-

wu Do

:3tl ct) C~ -

2 2

~ Cu

Am 0 co 0 m-E1 a:" p= E o wo13E m0j >-

Ic 301 Ct 0. >, 4

0 ~ 2* E 8

/J >

/j .0 2

0 0 E0 rdi(00 o E ~

00__ _ _ _ _ _ --o 0 > V ý X.

0~c 0 0 0 0 0 0 e 2~~~)~ 41-- ~ . O(wu)~~~ oi(u ~~~

X-r IO v -

4ý C U U>o u~ 0

00

0 E~

9 E

-co 0 -

u 000 r

. S 2 5 4 0O

u 0 0 20o.- *

E E t ý 32-w 4, ,--It

w ODw .- 0j 0 -0 - 0 I.)

fI 0-s 's1 - 2 -c =~ -S

E ~ 2 0 0

0E0-M. 0 -e

00 0 0

ba >

0 - -S! 2 m. y 5~ j :acZ

2 r- 00q0

S . ~ ~~O

Cd~ >

c g~ > <5 -~

E r. 4i 8-S :m~

O> 8 E~ -Rt Or-!

40. 0 - =2 u

'8 tO0

45 0 - E-~ gO

tj t= C :

12 E -5c ~at C -= U a '' ,-

-2 .0- =2 ',

-5 tj 00

o Z c U()u5u.

C t- ao

.5 -

u u ~ LM t.

-r* boR:s 0 > - E

C3 M 0

c 20 o0 -or 2. u 00 0 0A

C13 0~, C6 0

- = -a-. 9 ou -

mO r -S -20 .

-> c S4 g ,EF

= _.:2 C t4 ;a u j I.

Z g S E 10 = - 0E ~ o =~ g

0 _0

A

.0 0 E 00 0 go~

.a 2. , 6-.g.~0 0~ 5~-UuU- ~ 0~~

0 es-

-o E- 8 o u -~: *- t- e 'a Z

0 9% ce 00 M

Cý F 2 1 1> ,.0w1 wý*

04 .- 00 t, 2 .

o E E bo 'A ;.7 0

0 . 0 ~5 -v0,

I., 0 c O 9 ;V -. ,

Cd~ 06 0 0

- s- 5 0 00 0 >,2 U 2 ~ .

t!P uu s. .5g w> 9ý -5 '4> *i

*Z 9'~ E g 0 o.9

-9 0 E 7.2 9c .

~~~~~; ~ U r 0 7S80 es < 04jg

m -2 E

E >~

or -j

u_

00 U .-

Al 2A~

--ub

45 Tg- 8>1

0 tc a -

0 s~

ac-. JrO .5 > " 0

0~

-- u raj ita1 I .

1

4>z ca I ; Yg- 8 "cV9~~~ r.o 81

2 0 CL

- ~ ~~ P b

U Iy~ J*691

#I-j u4t v rHo os u -5

CkC

AA 0. co 0

be &E " 5

a- .a

0.

EP >

ad 0 § a' -S

0 > o 0U~

B 0 r -S 9 -

z' =u*~ 'o~ u

z~ 0, L' a.IS*~~5~ U >

0 ........ .............. o

1 R4:7 L L

*0 0 ;0C. .5

> E u

0 00 5 C 20

C

uu E SO

0 0 E ca

0 0 0 b5 "0a Z 2 o

0 0 0400~00 0 q0 r

00 000 c 0

%0 0 0 A AA A . ~--0000 1]o-s*000 oI E )0%0 0 08

000 0 - Lz00 0 00 c - E :K'.2

0 0o 3I)rC0 to -:- C. C

0 _C 0.000b

0 C 5 :3 5. E0 E t~

C 0

-~ r- 1

ýt C

2,~~~ ~ ~ ~ 51. r s4

to ý' .4 , P, E .

0 00~

-0~~ L>O Uz -1xL:0-0.. 0

72 0

04 U 0 ~

U, Cc 79

0 - . 0 -u a:r

n E t4;7 c E E i3 E

.2 06 it ~C: 0 0 ~

-0~

E2 e -a lu

E4 -4J~ a- Is ,- iic N , 0-S 8 ~0u 0 3c

0o 'o Eed 4..

0~~~~- E >lo'bAi.

to gj 0 0- L 2.5 -r. g

0 '.0 &'

8.5 -9 .2 .

"0' ý4 -CE 43,

04S

4))V C t

al-C-

o Ci *0

'o. -a* o - ) 5 2

-0 '0 OF 1 0 0 H

'0~ . U-

0 o

ocE *0a

u o-

> O"c 0-§s~ 00 -_ - T.:

00

0 . g .

SS

90 06 bga~ 0.S 4> -0 4' 4

Do N~ .. j 1> O'OCa .. U 0 '

_i 0

0''

E S L 8-9 ftI "It

-S 41

0)~-m oc...-0 e, -

* 009

000k *0-'- cog

~ - 0

-~ ~ Ch

0-~ ~-3 O .NO ~'e' rz :I*

eV, IM 2 r.4 Ss .

--f A'

0 RA t ~

us 0 ý '- - 'w:& Ot ad.U -. aa a d

- - - - -- - LOL OL

.- ~§ E 0. 0 - ý !

E E = K. I '

Sm 22 --P

~ ~ ~ _ ~S;~~fA~ ~ -Ur

-0 1o go 'Q -ttcv EA0 0 S1 '5'~

It 2

to 2 0. w g .'a0N;9

Alb S 11 C4

;A .9 co -.

8ýR~&S 46.0.S

ILS

I A-i F - a.jk1

tip~

CK t:

in at li

0 q~I ~Dfill"]

a#.:Ls1 '1k I

11 - 80 s 0

A~ ell .g~

s fAll. 0.09 -

If " I! i ,1 ýI 0I 10j

Ij ~ ~ ) Ale to

ilj1JfjIJJifal 4!ZZ' m.du~w~ush~~Z.0 0 D"o0 90 l& ~ dzz

;! :R 4; A

c~~~a~~~m~~~s~~--- -----e e ma mas m . m m c m a

Reprinted from

REVIEW OFSCIENTIFICINSTRUMENTSa publication of the American Institute of Physics Vol. 61, No. 9, September 1990

Force microscopy with a bidirectional capacitance sensorGabi Neubauer, Sidney R. Cohen, Gary M. McClelland, Don Home,and C. Mathew MateIBM Research Division. Almaden Research Center, 650 Harry Rd., San Jose, California 95120-6099

pp. 2296-2308

Force microscopy with a bidirectional capacitance sensorGabi Neubauer,a) Sidney R. Cohenb) Gary M. McClelland,c) Don Home,and C. Mathew Mated)IBM Research Division, Almaden Research Center, 650 Harry Rd., San Jose, California 95120-6099

(Received 5 March 1990; accepted for publication 10 April 1990)

A new method for sensing cantilever deflection in the atomic force microscope (AFM), basedon capacitance measuremrrent, is described. Parameters governing the design of such aninstrument are considered in detail. Two different geometries are compared, wire on plate andan integrated flat plate sensor. The electronic circuitry, providing 6X 10 - '" F noise in a 0.01-1000 Hz bandwidth, is also described. Implementation of the design ideas into a working AFMin ultrahigh vacuum is demonstrated. This AFM allows simultaneous measurement ofcantilever deflection in two orthogonal directions, necessary for our nanotribology studies. Thetheoretical sensitivity of 5 X 10- ' F/m is not achieved due to roughness. The bidirectionalsensing and imaging capabilities are demonstrated for an Ir tip on cleaved graphite, and adiamond tip on diamond films. The capacitance detection technique is compared andcontrasted with other AFM sensors.

1. INTRODUCTION it should be operable at frequencies near zero (i.e., dc

Since its introduction in 1986,' the atomic force micro- mode), while recording the instantaneous cantilever deflec-

scope (AFM) has proven to be an invaluable tool in obtain- tions. Many AFMs surmount problems of low-frequencying atomic scale, and near-atomic scale topographical im- noise by rapidly oscillating the sample or cantilever (i.e.,ages of nonconducting surfaces. In addition, it has been used operating the AFM in the ac mode) and using phase-sensi-to measure forces as weak as the van der Waals attraction tive detection. For force measurements that are made duringbetween a sharp tip and a nearby surface. In the AFM, forces hard contact between tip and surface, such oscillations are

between a tip and a surface are detected by measuring the not allowed as they may perturb the mechanics of the system

deflection of a flexible cantilever upon which the tip is under study.

mounted. In the first AFM, the cantilever deflection was Keeping these requirements in mind, we will, in Sec. II,monitored by tunneling from a second tip; subsequent de- examine the applicability of different existing AFM cantile-signs have used optical detection. In this paper, we report the ver sensor techniques, and summarize the role of capaci-implementation of a new scheme: using capacitance to detect tance in scanning microscopies. Section III presents in detailcantilever deflection in an AFM. Our new AFM, incorpor- our new detection method and considers various possibleating this sensor, was designed for ultrahigh vacuum detector designs. The electronic circuitry will then be intro-(UHV) experiments in tribology (the study of friction, lu- duced, paying particular attention to the low frequencybrication, and wear), at the subnanometer level. In such noise behavior. Section IV briefly describes the incorpora-studies, the imaging capabilities of the AFM are combined tion of the instrument into a UHV chamber with other sur-with force measurements in the nanonewton range, perpen- face probes. In Sec. V th. instrumental performance will bedicular and lateral to the surface. This is the first AFM capa- evaluated using data obtained for iridium tips and diamondble of sensing the two directions of force simultaneously. tips sliding on highly oriented pyrolytic graphite and dia-First results obtained with this instrument have been pub- mond films, respectively. Section VI proposes refinements

lished elsewhere.2'- Gdddenhenrich et al. have also recently and extensions to the capacitance technique.described a capacitance-based AFM.4

Whereas our AFM was designed specifically for atomic II, AFM CANTILEVER SENSOR TECHNIQUESscale tribology, it has promise for broad applications as aprobe for interfacial forces. To be useful for this range of The first AFM employed tunneling from an additionalapplications, three requirements are of particular impor- tip to the back side of the lever for detecting the cantilever

applcatons thee equremnts re f prtiula imor- deflection.' The high sensitivity resulting from the exponen-tance. First, the AFM should be operated inside an ultrahigh delcin'Tehgsnitvyrsuigfomhexpe-tancuum Fiste, the r wshouldbel-pharactezd isuriaes can b tial distance dependence of the tunneling current led to avacuum system, where well-characterized surfaces can be wide application of this technique, mostly in imaging stud-prepared and maintained. Next, it must be capable of mea- ies. 1.5,6 However, because of its low dynamic range, the tun-suring frictional force and load simultaneously. And finally, neling technique requires the use of feedback circuits to fol-

low large cantilever deflections and thermal drifts, therebylimiting the available bandwidth. As the major disadvan-

"Current address: Lawrence Berkeley Laboratory, Center for Advanced tage, the tunneling detection method suffers from contamin-Materials, Mailstop 66/200. 1 Cyclotron Rd., Berkeley, CA 94720. ation-induced instabilities of the tunneling junction. whichChaim Weizmann Postdoctoral Fellow."Author to whom correspondence should be addressed, can make its performance very unreliable, even under vacu-Also affiliated with thc IBM Magnetic Recording Institute. um conditions. Finally, the tunneling tip exerts extraneous

2296 Rev. Sai. Instrum. 61 (9), September 1990 0034-6748/901092296-13S02.00 c 1990 American Institute of Physics 2296

forces on the cantilever, both van der Waals and electrostatic total RMS noise was measured to be 0.01 A in a bandwidthin nature, which can significantly distort the actual force of 1 Hz to 20 kHz.measurements. A bidirectional sensor based on the tunnel- In a recently reported GaAIAs diode laser feedback sys-ing technique has been operated in our laboratory7 by moni- tem, 8 the amplitude of cantilever vibrations is detected bytoring the tunneling current from the cutting edges of two an integrated photodiode, which measures intracavity inten-orthogonal razor blades to the back side of a lever.8 How- sity changes due to interaction of the reflected part of theever, the low reliability of the tunneling technique, now com- laser light with the gain medium. At the lever resonancepounded by the simultaneous operation of two independent frequency, a sensitivity of 0.03 X/`H z was reported. 1 "tunneling junctiolus, was unsatisfactory. Finally, an optical method using deflection rather than

The need for a more reliable detection method led to the interferometry is being used in several laboratories. 2̀ '2 1introduction of a variety of optical techniques, which have a Here, the deflection of a laser beam off a small mirrormuch greater dynamic range and exert no extraneous force mounted on the cantilever is detected with a dual-photo-on the cantilever. In homodyne optical interferometry, ini- diode position-sensitive detector. This approach was alsotially demonstrated in our laboratory, a HeNe laser beam is demonstrated in vacuum,2 " with the HeNe laser and thereflected off a cantilever and interferes with a part of the detector located outside the UHV chamber, and a smallestbeam reflected from a reference flat. 9" Despite its simpli- detectable cantilever displacement of 4 X 10- 4 k/j`Hz duecity and higher reliability as compared to the tunneling to shot noise was calculated by the authors. Deflections ofmethod, this homodyne interferometer suffers from its high the laser beam due to air path fluctuations were clearly re-noise level at frequencies below 100 Hz. For example, in a 0.5 cognized as the major source of noise in the setup in air.2Hz to 3 kHz bandwidth a cantilever displacement noise level This problem can very likely be avoided by placing diode andof 0.2-A RMS was found. detector in vacuum, or by running fibers to these elements,

By introducing optical fibers into the homodyne setup, placed outside the vacuum.the air path length of the interferometer was lowered to that Although the use of rppacitance variation to measureof the fiber-level spacing, _- 10 Mim, thus reducing the low lever deflection in a force microscope is new, the idea offrequency noise."2-14 A major improvement of the homo- using capacitance to measure displacement is well known,22

dyne techniques was recently achieved through an all-fiber and general purpose sensors are commercially available. Mi-configuration containing a multimode GaAIAs diode laser croscopes which directly sense the capacitance variation be-with a direct single-mode fiber output and a single-mode tween a sharp tip and a surface have been described, 23 -2 anddirectional coupler.13 In addition to the advantages of very this principal is the basis of a video disk player.2"compact, UHV adaptable instrumentation and reduced heat Scanning microscopy has also been demonstrated wheredissipation, optical path length fluctuations due to air turbu- the electrostatic force that exists between the two electrodeslences are almost completely avoided, the only remaining of a capacitor, here a sharp tip and a conducting surface, isspacing being the one between fiber and cantilever. Because the output of the feedback loop. This technique is indepen-of the short coherence length of the diode laser, the low- dent of the ineans of detecting lever deflection. This princi-frequency stability is further improved, since interference pal has been used to profile conducting surfaces with andeffects due to stray reflections are eliminated. Thus, in a dc to without dielectric overlayers, 11.28 and to detect minute static1-kHz bandwidth, the peak-to-peak noise is less than 0.1 A. charges.2

1

Above I kHz a noise spectral density of 5.5X 10- A/ýHz The new cantilever-sensing technique, which will be de-is reported, close to the shot noise limit. A fiber design using scribed in the next section, uses a compact, vacuum-compa-a HeNe laser and a single mode directional coupler has also tible sensor, which measures forces in both directionsbeen reported." through changes in capacitance and, in addition, enables

In the heterodyne technique"' a frequency-shifted part changing of cantilevers in situ. This is the first AFM capableof a HeNe laser, formed by a radio frequency (rf) Bragg-cell of measuring two directions of force simultaneously. Bidir-modulator, interferes with the phase-shifted part reflected ectional optical sensors are possible, but have not yet beenoff the cantilever. This technique requires extensive rf elec- demonstrated. This method provides high sensitivity and atronics as compared to the homodyne setup, but it is inher- high dynamic range while maintaining low noise levels in theently insensitive to laser intensity fluctuations. This method dc mode; e.g. 0.03-4 RMS in a 0.1 Hz-to- I kHz bandwidth.has only been implemented in the ac mode, where the inter- Recent publications by ourselves 2̀ and another group"ferometer noise at the lever resonant frequency was much have demonstrated the use of a capacitance AFM but haveless than the thermal noise of the lever, which amounted to not elaborated on the many important factors involved in the0.01 A•./,z. construction and operation of such an instrument, which we

Another interferometric scheme, very well suited for present below.both ac and dc measurements, was reported recently.' 7

From a HeNe laser two orthogonally polarized light Ill. THE NEW DETECTION METHODbeams-a reference and a sensing beam-are formed, whichare reflected off the base and end of the cantilever, respec- A. Experimental design considerationstively. The relative phase of the two beams is a measure of the The specific geometry chosen for the capacitance sensorcantilever deflection. This method greatly reduces sensitiv- proves to be an issue of some complexity. The obvious con-ity to optical path length or laser intensity fluctuations. The cern of high sensitivity is opposed by the more subtle subject

2297 Rev. Scl. Instrum., Vol. 61, No. 9, September 1990 Force microscopy 2297

of snapping. The rate of change in capacitance with distance dC A (2)(sensitivity) increases as the distance between two bodies dd d 2

decreases. However, the attractive electrostatic force be- The spring constant k and resonance frequencyfR of a rec-tween the bodies also becomes large at small distances. The tangular bar fixed at one end (i.e., the top plate) are"geometry must be chosen to achieve maximum sensitivitywithout causing snapping together, which occurs when the k = (Ebh 3/4l 3), (3a)gradient of the attractive force exceeds the spring constant. R ( 1 .02/1r) (k lblhp)", (3b)This will be discussed below. During typical operation, tips where E is the Young's modulus of the material, b, h, and Iare changed on a frequent basis. The design must be able to are the width, thickness, and length of the bar, respectively,accommtcdate different sizes and shapes of tips and cantile- andp is the material density. The electrostatic force attract-vers, anO provide for ease in exchanging them. ing the plates is the partial derivative with respect to d of the

Two design geometries were evaluated, and are depicted electrostatic energy W,.1 (Q 2/2C) in a capacitor holdingin Fig. 1. The integrated sensor [Fig. I (a) I detects capaci- charge Q = VC:tance between two flat plates. The bottom plate consists oftwo electrodes. For example, we tried preliminary studies F =- -W = _ _ (1/C)where the top plate consisted of a doped silicon strip that had ad 2 ddbeen micromachined3 ° to a thickness of several microns, and =C2 W A A (4)attached to the glass plate using field-assisted bonding.' 2 c 2 dV (4)The area of overlap of the two plates and the distance be- 2 e 2 detween them determines the capacitance. Tlý ý sensor is cou- From (4) we obtain the force derivative:pled to the force-sensing cantilever. The bidirectional detec- F,' = AeV 2 /d 3. (5)tion capability of this design would operate as follows: an Here A = bl /3 because the sensor is only free to bend on oneincrease in load brings the upper plate closer to both of the end, and hence only part of the area contributes to changes inbottom electrodes and thus increases the capacitance. Fric- capacitance. In order to have similar sensitivities in the twotion results in a twisting of the upper plate, which would dimensions, the width and length of the upper plate must bebring it closer to one of the bottom electrodes and further approximately equal so that A = 12/3 = b 2/3. The exact di-from the other. An increase would be seen in the difference mensions of the device are dependent on the desired charac-signal from the two bottom electrodes. The parallelism and teristics. Spring constant and resonance frequency mustflatness of the two plates is crucial to this technique. This is have values that satisfy the instrumental operating condi-to prevent shorting out for the small separations used. tions. From Eqs. 3(a) and 3(b) it is derived that

The capacitance between the plates is ( E ":S • k 1/"

C = eA/d. (1) 3/= 4l.17r3/4P3/i,/f,/, (6a)Here A is the plate area, d the distance between plates, arid f k 1 /2

the permittivity in vacuum, 8.85 X 10- 2 F/m. Ultimately, h= (ik) (6b)the sensitivity is governed by the capacitance derivative 0.701Tr'2E 1/

The maximum allowed voltage is then set by the snappingcriterion, which requires k tobe greater than the force deriv-ative. In practice, this means that F, is some fraction a of"the cantilever force constant. From (5) and (3a),

h h 3d I 12c (7)

V 2 A 2 E

"Substituting from (6b) gives

S b4 V2A 2 = .31/4 (8)

"Since the detector noise is generally independent of the rfvoltage applied, but the detectable current through the ca-pacitor increases linearly with V, a corollary to (2) is thatthe signal-to-noise ratio varies as:

M d dC VŽ ± _. -- S/N ddd d2 (2a)

Sample S/N can be improved by increasing V, increasing A. or de-K .•..2 creasing d. The limitations on this "tuning" process are de-fined by the equations above. A lower limit for d will most

FIG. 1. Schematic views of the two sensor configurations: (a) integrated likely be defined by the precision of the micromachining pro-sensor with tip; (b) wire on plate with sharpened end pointing toward the cess. This means that the only way to improve S/N is tosample. Parameters indicated are those referred to in the text. increase Vor A. Equation (7) shows that this can only be

2298 Rev. Scl. Instrum., Vol. 61, No. 9, September 1990 Force microscopy 229"

done in conjunction with increasing h, which will adversely dC 1.05x. 10__(r,r.)'14(ktu)'affect k andfR. These tradeoffs can be incorporated into the dd Vsensitivity equation by substituting VA from Eq. (8) into where ka ='. Remembering from above that the S/v ik(2a): proportional to V( dC/dd), we see that the S/N is indepen-

(5) (Ec"allE I's ý ' 9 dent of varying V and d. as long as these parameters areJ•f ood.78p " J fX4d- ̀ varied proportionately according to Eq. ( 13) to maintain a

This is the S/N for displacement measurement. For most particular F'.force: Specifying k andfR, as was done for the first configura-

purposes a more important figure is the S/N for fotion, fixes two parameters, in this case the wire radius and

(S) (SIN), (e-'a11 2E"19 I length:N,. k 4.78p"'! k "4f;1 4

( 10 ) J '416

The significance of these two equations deserves empha-

sis. For a given material and length-to-width ratio, there are r I A• - 17)four adjustable parameters: 1, h, d, and V. By requiring a 2.98p'"E ') fk'certain cantilever force constant and resonant frequency and The snapping condition givesusing a voltage limited by snapping, three of these variableshave been eliminated. For example, choosing a desired k of d_ (3.71x10 121. r41)50 N/m and fR of 5 kHz, we find that ! = b = 2.3 mm, and V \ irEa I k.(11h = 19 um, using E = 1.7 X 10 Pa and p = 2300 kg/m 3 , For the Ir wire we commonly use, E = 5.2 X 10'' Pa andvalues for Si. By setting the plate separation d to a realistic p 2.25 X 104 kg/m'. We find for a desired k of 50 N/m andminimum of 2/.tm, we obtain a capacitance derivative (sensi- fJ of 5 kHz that r, = 33 ym and I = 3.0 mm. Since the cur-tivity) of =-3.9 X10-6 F/m. From Eq. (8), the working vature r, is necessary only for alignment of the wire with thevoltage is then 6.4 V RMS. plates, it must be of the order of the plate dimension. We

The second configuration [Fig. l(b)], which we have choose2 mm, which for V= 3.5 V RMSanda = 0.5 givesaimplemented, is that of a curved wire near two orthogonal working distance of 590 A', corresponding to a sensitivity ofplates. The capacitance is then measured between the wire 2.4X 10 ' F/m. Clearly, there are a number of consider-and each of the two plates. This device is somewhat simpler ations that will determine the final choice of parameters. Forthan the first as the curved wire is itself the cantilever, so no instance, there is a minimum wire diameter that can be easilymechanical connection need exist between the cantilever/tip handled. A final consideration not discussed above is theand any part of the sensor providing greater ease in inter- dynamic range: in order to relieve the constraints of oper-changing tips. In addition, no microfabrication is required. ation with feedback control on the sensor plates, d must beThe disadvantages are that the cantilever must be conduct- large enough to prevent snapping of the wire into the sensoring and have the geometry of a curved wire. The wire is plate as the cantilever is pushed back. If d is made too large,curved in order to ease its alignment with respect to the however, the sensitivity is unacceptably low.plates. Due to handling constraints in the current mode of op-

Analogous to the flat plate geometry, a series of equa- eration, our typical wire is 125 jm in diameter and 6 mmtions describes the choice of parameters. First we obtain the long, giving k = 90 N/m and fR = 2.5 kHz. With r, = 2capacitance derivative by approximating the wire-plate ge- mm, F' = 50 N/m, and V = 3.5 V RMS, the operating dis-ometry by that of an ellipsoid separated from the plate by a tance is 490A I Eq. (14) 1, giving a sensitivity of 4.0 X 10distance much smaller than either of its principle radii: F/m.

dC.dCC= 5.56X10 1t'ri: l)dd d B. Electronic circuitry

Here, r, is the wire radius and r, is the radius of curvature. The above calculations for the model sensor geometriesThe spring -onstant and resonance frequency of the cantile- indicate that capacitance noise levelson the order of 10 " Fver are given by"2 are required to achieve deflection sensitivities of O01 A.

k = 31rIE1/41', (12a) These levels are small not only on an absolute scale, but alsocompared to the total input capacitance of several pF for

fR = 0.284(r, E/pl) ", (12b) typical circuits. The basis for the capacitance measurement,

and the force derivative is applying an oscillating voltage across the capacitor and mea-suring the reactive current, is well known. Very precise mea-

F' = 2.78x 10 '( V/d 2 )•r~rl. (13) surements are possible in principle, since a capacitor has no

As for the parallel plate geometry, the sensitivity increases Johnson noise (associated with resistors) or shot noise (as-with decreasing d. but in practice the operating distance d,,r sociated with semiconductor junctions). The sensitivity ofshould he set so that F' is a fraction a of k, giving the capacitance circuit is optimized by measuring the capaci-

S..2........ ..... /.. tance at a high frequency to maximize the sensor current andd.,•,P =2,78x 10 V2 rbryka). (14) h using a balanced configuration to null out the effect of

Combining Eqs. ( I I ) and ( 14), we find that at this distance amplitude variations in the alternating voltage.

2299 Rev. Scl. Instrum., Vol. 61, No. 9, September 1990 Force mict oscopy 2299

As illustrated in Fig. 2. the instrument uses a grounded into account). When larger series capacitors were usd tocantilever, and the circuit measures at 6.4 MHz the capaci- provide a large balancing range, the varactors apparentlilance C, of the sensor plate with respect to ground. The made somecontribution to the I/J'noise described b.-low A,,transformer input corliguration (using a Mini Circuits T!-I presently configured, when the varactor bias is varied he-transformer) has three advantages. It allows one terminal of tween 8 and 27 V. the varactor capacitance changes from ti tothe sensor capacitance to be at ground, keeps the sensitivity 2.5 pF, corresponding to a circuit offset range of 1.75 pFof the circuit independent of the total input capacitance, and Figure 3 presents a schematic of the rfsource used. Thisaffords a simple means for balancing the circuit through the circuit generates a sine wave of very well-regulated frequen-trimmer capacitor or the varactor. Negati-,e feedback keeps cy and amplitude, and we hae no evidence that it contribut-the base of the MPSH07 near rf ground, so that the sensor ed to the noise of our instrument.current flows through the feedback capacitor C,. giving an The two principal sources of white noise in this circuitrf signal at the collector of (C, - C,) Ve./C,. where Vf is are the voltage and current noise of the MPSHO7. With adcthe rf supply voltage and C, is the balancing capacitance. base current i, = of 2.6/iA, the shot noise current is \ 2qi,After the emitter follower 2N3904, the signal is demodulat- = 0.9 pA/! Hz, where q is the electronic charge. (In thised at approximately unity gain by a mixer (Mini-Circuits paper all spectral densities are root mean square signal perTFM-3). It is then filtered before being amplified by 100, unit bandwidth.) This noise current pa',ses through C,', andgiving an overall circui: gain of dV,,,,/dC, = IOOV 1,/C,. is demodulated around the rf frequency giving a noise Nol-With a different output amplifier and filtering, this circuit tage of 770 nV/- Hz at the circuit output. The M! SH07could be operated at a bandwidth up to 1 MHz. Using an rf tage of 77 nV eciHzeat te cru output. th a lsevoltage V, of 3.5 V RMS, and a C.. of about 2 pF, we mea- voltage noise is specified to be 1.7 nVio Hz. With a total basesure a gain of 1.2x 10"4 V/F, in good agreement with this capacitance C, (including the sensor and stray capacitanceequation. of the lead in wires) of 5 pF. the resulting noise at the collec-

The + 10-V output swinggivesarangeof0.16pF. Pre- tor is 1.7 Y (C, + C, )/C, - 6 nV/,. Hz. giving 600 nVi

cise balancing is therefore essential just to bring the output ;Hz at the circuit output.on scale, and is also important to minimize the effect of V1, The net noise is thus calculated to be "0.67 t- 0.777 = Iamplitude noise on the output. The balancing is provided by /iV/• Az. We measure 1.7 uiV/I Hz, at frequencies ,5()0the varactor. Because the varactor performance conforms Hz, in reasonable agreement. As expected the measuredclosely to published specifications, it also provides a simple white noise spectral density is independent of V,.means to calibrate and check the performance of the circuit, Since the mixer demodulates in only a narrow band-since the circuit responds in the same way to variations ofthe width around 6.4 MHz, we would expect any current orsensor capacitance as it does to variations of the varactor voltage noise originating before the mixer to appear at thecapacitance (after the two 20-pF series capacitors are taken circuit output with a frequency-independent noise spectrum.

+15

~0.1

Sensor Plate Cr S25Kand Cantilever 2N3904 5O

MPSH07 2K 0n2 0.033 0.15 20K

0.1.08

- 20 200120p

20• pF 20 pr F

I IO 110< M B156,4 MHz Balance -15

3.5 V 8-27 V

3 Rei 2 Sc l i .imtc dIigrain outicm. t c. tit I uNo. i, detect eapaci 9 an9 e c Forgec Iet em t he meisor os ale a tid cally t 2I0 cc I file's iid cated i1lc-i isst-d t) ill" j reT0',t Mince intt

2300 Rev. Sci. Instrum., Vol. 61. No. 9, September 1990 Force microscopy 2300

+15

+4 +5 4 0.10. 1 -6.8 -L0 2C -

I . Ti- i

6.4. MHz TiT. i04t 2KCrystal t10 ý4(frrite

r -- 'LHOO02C.-S TI-1 I t

-200 3000.

W IN41 4

1K, 2,2 nF1K 1K

-154

FIG. 3. Schematic diagram of the radio ý-equency generator used for capacitance measuremen t. Unless indicated ot herwise, ca pacitance ks expressed in 11,F.resistance i, fl.

From Fig. 4, we see that this is the case above 500 Hz, but last component is very dominant at low frequencies it onlythat below this frequency the noise rises steadily. This addi- contributes 2OuV of noise per decade of frequency. Integrat-tional noise component is unchanged by offietting the circuit ing the above equation,with the varactor, but it is proportional to V,,. For this rea-son we believe that the noise arises from fluctuations of the N, = I.Nd f, (20)reactance of the portion of the circuit before the MPSHO7. -

The noise could be altered substantially by soldering and wefnthtoanisNvraIto10Hzbdithounsoldering trimmer capacitors, but never in a reproducible we find, I the total oiseover aere to 100 Hrebandwidthito

wa.The observed noise spectrum can be fit by the equation figure would drop only slightly, to 53,uV. Over a range of____0.0 1-l1000Hz, the total noise is 68 INor 5.7 x. 10 "' F. Ata

Nf =~ -v.70' + ( 1347y ) tV/<H_1z, (19) sensitivity of 2 X~ 10 'F/rn, this is equivalent to a dispiace-

indicating th *the noise consists of a constant component ment noise of 0.028-A RMS.plus acomponent with a l/fpower spectrum. Although this The actual traces of the output voltage show~n in Fig. 5

are consistent with these noise figures. The INy.pical drift rateis less than 0.5,uV/s. corresponding to less than 0.01 A/min.Since this is better than the positioning drift rate due to ther-

15 mal variations and piezo creep achieved in most AFMs, the0.02 * ~ actual drift in the sensor displacement measurement is not

10T limited by the capacitance measurement.> Although the capacitance circuit is mounted in a tube

0.01 several centimeters from the AFM inside vacuum, the 0i.45 . MHz rf source is located outside the vacuum; its signal is

z L I brought into the vacuum and passes down to the capacitanceC I 0 00dctector circuit through a coaxial connector. Probably be-1 10 100 1000 (0

Freqency[Hzlcause the circuit works near nulling, the capacitance signal isFrequecy HzIfairly insensitive to betiding or vibration of the rfcable. The

FI~ 4.Meaurd rot nea suar nose~eoratd~nitVvsfreueny fr 15-V supply is al Iso locat Ied outside of vacuum.

the capacitance circuit. The left vertical axis represents c'ircnit output vol. h w aaiac inl oehrwt eeec

tagc. while the righi axis represents the. corresponding lever uhisplaclmieni. ground were brought to two Tektronix AM 5021 differential.r~ssiirni.g a scnsiiivrty of 2., 10t' F/rn. amplifiers operated at a gain of 1(W. Because the circuit out -

2301 Rev. SOi. tostrum., Vol. 61, No. 9, September 1990 Force microscopy 2301

looo (a) When we first implemented our capacitance detection800. 0.2 reset" scheme, an anomalous signal was noticed with conducting

samples. When the tip came into hard contact with the sam-"600. ple, both the load anm friction signals showed an abrupt I-

S1.15 X 10 ", F increase in capacitance. This effect is due to,,400-the stray capacitance of the sample to the sensor plate, which

S200. 0.1 A acts in series with the admittance (impedance of the_0_ sample to ground to generate a stray sensor-ground capaci-

0 ii t`0 1'5 -20 tance. Although the derivative of this capacitance with re-TIME (n s~e) spect to the sample position is much sialler than the deriva-

10()0. tive of the cantilever-sensor capacitance with respect to the

"800- M (0) cantilever position, the iotal sample-sensor capacitance is oft7 = 10 msec the same order as the total cantilever-sensor capacitance.

600- The anomalo is signal arises because touching the grounded

Co 400- 0.1 A cantilever to the sanipie increases the capacitance ofthe sen-sor plate to ground.

200. Obviously this problem could be eliminated by perma-nently grounding the sample, so that tip conact would have

0 0.5 1.0 1.5 2.0 2.5 no effect, 1,ut this solution would not allow tunneling mi-TIME (spc) n fet u hsslto ol o lo unln icroscopy. which requires that the sample be electrically bi-ased with respect to the tip. The solution is to use a resonant

FIG ..Noise traces recorded from the capacitance circuit, with two differ- The at the 6.4-

ent time constants. The 0.1-4 calibration bar assumes a sensitivity of circuit to permanently ground the sample only at the 6.4-2 x 10 ' F/m: (aI Using a 0.2-ms time constant low-pass filter, corre- MHz operating frequency of the capacitance circuit. A 200-sponding to a bandwidth of I kHz: (b) 10-ms filter, bandwidth 20 Hz. pF capacitor in series with a custom-wound 3-pH inductor

were housed in 18-mm-diam stainless steel can sealed by awelded two-conductor feedthrough. Before sealing, the in-

put is very sensitive to the line frequency induced and other ductor winding was adjusted to tune the circuit resonance to

variations of the varactor voltages, these voltages were pro- the 6.4 MHz, so that a low 241 impedance was achieved at

vided by three 9-V batteries in series divided by a potentiom - this frequency. After the filter was mounted on the AFM

eter. base plate, and attached between the sample holder and the

Aside from the fundamental limitations of the circuit AFM body (ground), the anomalous signal was eliminated.

discussed above in some detail, a chief source of noise is thevariation of stray capacitance of the circuit or sensor plates. IV. EXPERIMENTBecause of the extreme sensitivity of the circuit, changes of A primary design goal was good vibration isolation.the capacitance to ground of even the output or supply wires which was attained by mounting the AFM on a suspensionof the circuit due to swinging ofthe instrument in its suspen- of very low resonance frequency and by raising the lowestsion can generate noise. To circumvent this problem, all resonance frequency of the AFM through the use of small,wires running through vacuum to the instrument run stiff, lightweight parts. Further matters of concern were re-through shielded cables (shield grounded) over most of duction of stray capacitance and accessibility to all of thetheir length. The vibration problem is most severe for wires devices mounted on the UHV chamber (airlock. LEED,going directly to the sensor. For this reason, these wires are Auger, heater, evaporation source).kept as short and as rigid as possible. Figures 6 and 7 diagram the AFM and the UHV

A particularly persistent interfering noise was at 120 Hz chamber, respectively. The AFM sits on a base plate. whichan] synchronized with the power line. This signal was lar- is suspended by two stages of spring support (450 mm totalgest )rnly when the piezo supplies were turned on, but be- length when fully extended) to the top flange of a 150-mam-cause it was proportional to the rfvoltage, it seemed to repre- diam tube extending upwards from the chamber center. Thesent a true capacitance signal on the order of 10 ' Fp-p. It spring system, with an intermediate mass of about I kg be-was concluded that the signal resulted largely from the tween the first and second stages, has a lowest vertical reso-

0. -pF stray capacitance ofthe sensor plates to the piezoe- nance frequency of 1-.2 Hz. The sample and sensor arelectric tubes. The 120-1tz signal could result if the imped- mounted opposite one another on piezoelectric tripods. Theantce at 6.4 MHz of the piezo driving circuits to ground was tripod consists of three piezoelectric tubes (PZT5A. 25 mmmodulated at 120 Hz. This could result front the switching hog, 3 mm diameter) each having a scanning range of -- 4on and off of the rc:tifier diodes in the piezo power supply. A tim. For coarse xyz translation, each tripod unit is adjustedsimilar effect scented to arise from the ±- 15-V supply of the by three mechaiical screws that are accessed by a rotatablecapacitance circuit itself. In agreement with this theory, we wobblcstick. A translation of I /imn results from a I' Screwfound that the 120-11z interference disappeared when 1.2- rotatIion in the x and "' directions, arnd spur gears provide amtl inductors were inserted (outside of vacuum) in nearly further factor of 7 reduction in the z directions. In normalall wires going to the AFM, including the piezo and capaci- operation, the cantilever remains fixed and the sensor platestance circuit wires, but excluding the rf supply. and sample are moved toward it first with coarse mechani-

2302 Rev. Sci. Instrum., Vol. 61, No. 9, September 1990 Force microscopy 2302

_J The bottom of the mount is a post that slides into a socketA" -• and is locked into place by a set screw.1 ~ t ->Most of the experiments were done using iridium as the

cantilever/tip material because of its hardness and inertness--- 3 with respect to adsorption of residual gases. The end of the ir

wire is bent by 900 towards the sample, and the tip is fabri-cated by electrochemically etching this end in molten NaCi.The NaCI is contained in a graphite crucible, which is held ata potential of - 2.5 V with regard to the specimen for c-- 2

8,-- s. " Tip radii of < 1000 A are obtained, as estimated fromSEM pictures. A typical example is shown in Fig. 8. Beforebeing transferred to the AFM, the tip is heated to 1300 K by

_ an electron gun.0.• The sensor plates are spring mounted on a Macor block.

0 _.which is screwed into the vertex of the piezo tripod. Theplates are cut using a diamond pen from doped Si (I ll(Pensilco). They are coated with an evaporated gold filmover a chromium underlayer in the region of the spring

mounting to eliminate problems of contact resistance.

FIG. 6- Schematic of AFM. showing (1) spring suspension, (43 mounting The two plates are chosen close enough to be coupled by

ring, (2) eddy current damping, (3) mechanical adjust screws for sample a capacitance C, of 0. 1 pF, which mixes the signals for theand sensor, (5) piezotripods for sample and (8) sensor, (6) sample mount, cantilever deflections in the y and z directions. For example(7) tip mount on piezo tube scanner, (9) macor mounting block for sensor the effective capacitance C,. of the y sensor plate includes theplates, (10) leads to capacitance circuit, and (It) scaled tube containingcapacitance circuitry. For clarity, sample and sensor are shown mechani- capacitance C, of the gap between they sensor and the canti-cally retracted from the wire. lever and the capacitance of C,. acting in series with the ca-

pacitance C, of the gap between z sensor and the cantilever:C =C,+, + +--- (21)

cal, and then fine piezoelectric adjustments. The cantilever,

however, is mounted on a piezoelectric single tube scan- Since CY and C. are about 1 pF, dC./dC. ý0.01, and thener,33 which may be used to oscillate it in any direction. Both interaction between y and z is only about 1%, much smaller

cantilever and sample are attached to small mounts with than interactions due to inexact relative positioning of thewire hooks so that a wobblestick can carry them to all de- sample, wire and sensor plates. In any case, the interactionvices-airlock with linear translation stage, LEED, Auger, can be measured by moving the sensor with the piezos in

heater, evaporation source-without breaking the vacuum, well-defined directions, and later this effect can be removed

Rear-viewLEED

TurbomolecularPump z - Translatable

e-Gun Heater - /'-- Ml, M2 X-)r-ZS a-p"e/ T. .. . T r a n s l a t a b l e

Sample/Tip - 0 MountStora e MI

Ro .~~ ughing Carousel"'I -t" Valve \

Magnetically-coupled - I_"LinearTransporier . A ..

-]' , -, 1M3 " AFM ,..c " ..

WobblestickGate M2 #2"Valve

OMS Wobblestick X_n 1. , LeakValve

Ml- z- Translatable [orMount Auger Gun

[:[(i 7 S'chcCma1c top 1,tW (If the tJ IV Chamlner, sio%. rip I ihe -irf:icet ilý}id lntritrurneillltt ic, ile di t .fffrcnt ".,ainplte niouututt lhit .illhV. their r .c ,o,hIh.l-slit'k Ilo I I,, f•ed to mcne cirl ple\/tlp.- hetweeri Ole carouiwl., I E)D mn tllut. Auger/ion pul1 floutnl. llt li- A NM lItc h iiieat tri el-parlei ,tei hetivcithe :ir h ,. k. heater. 95M1 nnmu iitit. ;cc lict1 ciroun , l W ohblc ,ticlk tioc 2 i,,i ed t ofo nlcc llah icacl I gri ,,,, I.ct , tm c t '. ti r hc A IMl a:mclplc •|m d • ncIic c u i' tcc1cll .N15( 1 /,' Ioi pumip pro, ide_, purm ping for the i•amiIi t al:itcher Bloth OW uIi Ic k cr lCh k r.gicII ,lold 11 cI11L elhia h r oi•, roiighed hI. a t I' " ttlhollolcc 1lu1r lrtntci'

2303 Rev. SO. Instrum., Vol. 61, No. 9, September 1990 Force microscopy 2303

by applying a linear transformation to they and z data. insensitivity of the capacitance signal to sensor position or byThe electronics circuitry is mounted in a 38-mm-diam a sudden off-scale signal. The plates are then taken back to a

tube, which is sealed with an atmosphere of argon to aid leak position as close as possible to the wire without achievingdetection and clamped directly into the base plate of the snapping of the wire into the sensor plates or touching of anAFM. The wire (teflon-clad copper wire with 0.15 mm outer asperity on wire or plates. To calibrate the relation of detec-diameter from the California Fine Wire Co.) leading from tor output to cantilever displacement, the sensor plates artthe sensor plates to the circuit is about 80 mm long. All oscillated by the piezoelectric tubes over a known distance,electrical connections to the AFM are made through feed- and the amplitude of the signal modulation is recorded. Eventhroughs in the top flange. To reduce stray capacitance, all before snapping occurs, the attractive force of the sensorsignal lines are shielded from each other and especially from plates on the cantilever reduces the effective cantilever forcethe wires carrying the piezo voltages. To avoid vibrational constant by adding a downward curving potential to the up-coupling of the AFM to the outside through these relatively ward curving parabolic potential of the cantilever. The ex-stiff wires, a connector board on top of the AFM connects tent to which the force constant is modified can be deter-them to the weaker wires described above, which finally lead mined by vibrating the cantilever and measuring theto the AFM. Other wires, such as those inside the tube be- resonant frequency shift. Two nearly equal, low resonancetween the circuit board and UHV feedthroughs, are stiffer. It frequencies are found, because the doubly degenerate reso-was found that vibration of circuit parts and connecting nance frequency of a straight wire is split when it is bent.wires could contribute to the noise. For conducting samples, the coarse approach of the

All the piezo scan voltages, capacitance sensitivities, sample toward the tip can be monitored electrostatically bytunneling current, and y and z cantilever motions are record- applying a 30-V p-p triangle wave to the sample. Then, whileed on video tape using an A. F. Vetter Co model 4000 16 bringing the sample toward the tip using alternately me-channel analog-digital processor and Panasonic model chanical and piezoelectric motion, the normal electrostaticAG1230 VCR. For analysis, this data is transferred to an force'"6 is monitored through phase-sensitive detection ofIBM PC-AT. the cantilever deflection at twice the frequency of the trian-

gle wave. By comparison of the cantilever deflection withA. Experimental procedure that calculated from the tip radius, the sample is brought to

To achieve optimum sensitivity, the two sensor plates within 500-2000 A of the tip. For a friction measurement

are first brought into contact with the cantilever (for unidi- (with both plates close to the wire), the sample is then ad-

rectional force detection only one plate is used). Depending vanced slowly (several A/s) up to and beyond the point

on whether the cantilever-sensor contact is insulating (due where hard contact occurs, and then backed out to separa-

to contaminations) or conducting, contact is detected by an tion. Simultaneously, the sample is oscillated in the friction-sensitivey direction at 1-10 Hz. Because the change in z isslight for each y oscillation, each such loop gives an averagefrictional force for a particular average load.

V. RESULTS

Our AFM has been used to study several different sys-tems: Iridium tips on highly oriented pyrolytic graphite(HOPG) and Au( lll) films, tungsten on SiO,, and dia-mond tips on diamond films. Detailed analysis of these re-suits is published elsewhere. 2 ' Here, the bidirectionalityof the instrument in the dc mode will be shown with theHOPG data. and the imaging capabilities demonstratedthrough pictures of diamond films.

A. Friction of an iridium tip on highly oriented pyrolyticgraphite

The graphite sample used in our experiment was cleavedin air by peeling a thin layer of graphite from the sample with

adhesive tape immediately before loading it into vacuum.Figures 9 and 10 represent a friction experiment performedas described above. The z advance is 8 A/s and they oscilla-tion is 20-A p-p at 2 Hz. A typical loop obtained during one

such y oscillation of the sample is shown in Fig. 9, displaying-> - 75() A the deflection ofthe cantilever in the lateral (a) and normal

direction (b) versus the sample y displacement. In (a)

IIi x I ltcrr 'm r ,ý gr ah •i ir Iti, tl,.hcd AfIcirovhcmiuu i. n wi o ,,,c numbers indicate different regions of the loop: Just after theNa('I w•uh tip rmuihu. u,f 7M1 A sample change.s direction the loop shows a sticking region


7 I iI 74 4 (a)•,32 :3::6 6 F'

00

0 -2

-j 2 t22oLo-- 3- approach--..retraction..-4 0 4 1

Sl l 0S25 -() 2.5 :'.(b) ,S20 2 20 2.

* 0 'a30 0>

215 2.5 (b)- 4

10 2.0 T w 20 2.

SN I=

e0

0.5 Vi0 /.10 1. _j 20 " p =1 2

a 10a ~approach- A-retrct'la~ný--0 5 10 15 20

y Sample Displacement [A) 0 20 40 60 88 60 4( 200

z Sample Oisplacemert [A]

FIG. 9. (a) Lateral and (b) normal deflection of the cantilever (k = 100N/m), as thesample undergoesone complete oscillation in the lateral direc- FIG. 10. (a) Average lateral and (b) normal deflectionsof the cantilever vstion. Numbers in (a) indicate different regions of the friction loop: 1-2, the sample z displacement. Arrows indicate approach and retraction.sticking region; 2-3, sliding region.

(1-2), with a slope (i.e., cantilevery deflection per sample y the noise in this experiment is too high to clearly distinguish

displacement) of one. When the restoring force of the bend- a periodicity.

ing wire overcomes the static friction (point 2), the tip slides The most striking feature in Fig. 10 is the slow onset of

across the surface (2-3) until the turnaround at point 3. the normal cantilever deflection as the sample is advanced

Figure 10 shows a friction experiment in which a freshly towards the tip. Only at higher loads is the cantilever even-

cleaved HOPG surface is contacted with a new Ir tip with a tually bent backwards at the same rate as the sample. result-

radius of 1500 A, and spring constant of 100 N/m. Here, the ing in a slope of I in curve b at higher sample z displace-

utility of bidirectional sensing is demonstrated. The load and ments. The reversibility of this behavior, as indicated by the

friction values for this figure have been measured simulta- subsequent retraction, indications that the deformation is

neously, obviating the need to make assumptions about sam- elastic. As discussed elsewhere,2 the elastic response to a

pIe compliance or elasticity. Average lateral (a) and normal given load is much greater than predicted from a model

(b) deflections of the cantilever, deduced from each loop, based on the bulk elastic constants of graphite. an effect that

are plotted versus the z displacement of the sample, as it is may be due to delamination and bending of a graphite flake

advanced towards the tip and then retracted again at the normal to the surface. From Fig. 10 we derive a friction

same rate. The average frictional force is obtained from half coefficient of 0.25, much higher than the values previously

of the average loop height multiplied by the wire spring con- reported for metal tips on graphite. Since this value was ob-

stant. The average load is computed from the average nor- tained on the first approach of the sample to the tip. it is

mal cantilever deflection which, in these experiments (Fig. likely that much lower friction coefficients for single tips on

9), does not change by more than 4 A. during oney oscilla- graphite are attained only after a flake has broken off the

tion. In Figs. 9 and 10, the dc deflection of the cantilever in surface and attached to the tip.

the two orthogonal directions is computed directly from themeasured capacitance change, without the employment of B, Diamond tip/diamond film interactionsfeedback electronics to adjust the sensor position. As can be The imaging capabilities of the AFM in the ac modeseen from the loop in Fig. 9(a), the cantilever deflection were tested in both the contacting (repulsive) and noncon-noise is approximately 0.3-A p-p. The effective bandwidth, tacting (attractive) profiling mode. In the attractive modedetermined by the sampling rate, is dc to 200 Hz. The struc- the cantilever was vibrated at I Hz above its resonance fre-ture in the sliding part of the loop in Fig. 9(a) could be an quency at an amplitude of 20-A p-p. Since the compliance ofindication for a modulation of the frictional force with the the cantilever is changed due to the attractive force deriva-underlying sample lattice, as it was found for graphite in air." tive between tip and sample, the vibration amplitude of the

2305 Rev. Scl, Instrum., Vol. 61, No. 9, September 1990 Force microscopy 2305

cantilever is reduced as a consequence of a lowering of theresonance frequency.'"' Profiling of the surface was ac-T ------complished by a feedback circuit that kept the tip-sample 35A -

spacing constant by maintaining the vibration amplitude of 1. .. - .the cantilever at 50% of its resonant value. In the repulsivemode the tip was vibrated with an amplitude of 5-20 Ap-p ata frequency of several hundred Hz, depending on cantileverspring constant and resonance frequency. When the sampleis brought to within the range of the tip vibration, the ampli-tude is reduced due to the hard contact of tip and sample. In 400 _ __ . - -this application profiling was accomplished by a feedback _____ __ _-__

circuit that kept the tip-sample spacing constant by preserv-----____ing a reduction of :- 20% in the tip modulation amplitude.-.-.------- -

The diamond tip was obtained by gluing a small dia- _-----mond fragment, 200/pm across, to the end of a cantilevermade from tungsten wire. A spring constant of 20 N/m wascalculated from the geometry of the wire and confirmed Soo Afrom the resonance frequency. The diamond films used inour experiments were prepared in two different ways: Film 1 FIG. II. Attractive mode line-scan image ofdiamond film I. Scanned areawas deposited in another vacuum chamber onto a silicon was 500 A x 400•A (xy). The offset between lines is 5 A•in y. The bar showssubstrate, at 1100-1300 K by hot filament chemical vapor the vertical (z) dimension corresponding to 35 A. The lateral resolution.subtrteatdemonstrated by the narrowest stepwidth seen, is 30 .A.

deposition. The film was characterized by Raman spectros-

copy and SEM photographs, which showed diamond crys-tals typically =2 ,um across. 36 Film II was prepared on asilicon substrate, in a microwave plasma deposition attached to wire or plates. Particulates in the size range 0.1-chamber."7 To learn about the film topography at the film- 10 pm represent a significant fraction of laboratory parti-/substrate interface, the silicon substrate was etched away, cles. " While the polished silicon wafers are smooth to betterand the resultant coherent film glued upside down onto a than 100 X, the wire is manufactured by an extrusion pro-second silicon substrate. Both films were kept in laboratory cess, and SEM pictures of uncleaned wires show bothair for a couple of weeks before being loaded into the AFM grooves along the length of the wire, and particles < 10umwithout further treatment. seemingly stuck to the surface. To alleviate the smoothness

Figure !1 shows an example of an attractive mode image problem to some extent, a rather high voltage of 3.5-V RMS

of film I over an area of 500 A X 400 A. The image reveals asmall scale roughness not visible in the SEM pictures, asfeatures of only a couple of A high can clearly be resolved.We assume that over a such small scanning area the tip is . , , .

actually profiling a single-crystal plane of a diamond crystal-lite. A detailed account of our investigation of frictionalforces of this film will be deferred to a later publication.. 5

Figure 12 shows an example for a repulsive mode image offilm If over an area of - 5000 A >x 7000 A. Relatively smallfeatures are again resolved, with heights ranging from 10 to --- -

100 A. -

VI. DISCUSSION 7000, :' - L -------- --The theoretical performance of this capacitance AFM is

comparable to, or better than, that of other AFMs in currentoperation, although due to roughness our observed noise lev-el is nearly one order ofmagnitude worse than best reported---results using optical fibers. From the measured signal,1X10 ' V/,. taking the calibration factor of the circuit of __.___ 8O-1.2/ 10" V/F into account, we find an actual sensitivity 8--"..."0-during the experiment of only 1 I 10 ' F/m, severaltimes lower than the expected sensitivity (see Sec. III). Aclose examination of wire and capacitor plates showed that. 0 A -'in practice, our minimum working distance and consequent-ly our maximum sensitivity is not determined by the theo- IG, 12 Rcpukke modeline-wan inageofdianiond film H Scaoncd arca

retical operating distance d,,, ((i - 0.5), but rather is limit- ý-:.• 5(X)A . 7(X A ()t-A hbei een lives is I(M1 A the bar shliw the

ed by the roughness of the wire and/or dust particles eCrlical dinlCrsi.•MI, coIC,,pOrnding to 80 Aý 3he lateral resolution is 60t A

2306 Rev. Sci. Instrum., Vol. 61. No. 9, September 1990 Force microscopy 2306

was applied to the sensor plates which sets the snapping dis- and one for repulsive forces. Alternatively, a single sensortance d,, (a= 0.5) to 490 A and the sensitivity to plate could be used by applying a nonzero dc bias at zero tip4.0X 10-7 F/m (Sec. III). Still, we usually have to work at force, so that the lever is pulled by the electrostatic forcean actual distance of up to 5 times bigger than d,,p, so that against its own mechanical restoring force.theoretically a sensitivity of 1 I X 10 ' F/m is to be expect- In conclusion, we have presented a new detection meth-ed. This is found experimentally. Polishing methods for ob- od, based on capacitance measurement, forsensing the canti-taining smoother wires to work closer to the snapping dis- lever deflection in an atomic force microscope, and em-tance are currently under study in our laboratory. Ideally, ployed it for the investigation of tribological properties (i.e.,dust-free conditions should also be maintained both while friction, lubrication, and wear) of various samples. The newpreparing cantilevers and sensor for operation, and within AFM offers a low noise level at frequencies near dc, is UHVthe UHV chamber. compatible, and bidirectional: Frictional force and load are

Whereas optical techniques for sensing AFM cantilever determined simultaneously by measuring the capacitancedeflection are fundamentally constrained by photon shot change between the cantilever and two orthogonal plates.noise, no such limit applies to the capacitance technique. Theoretically, the instrument should achieve 0.03-A RMSThe essential limitations of our detection scheme arise from noise in the 0.01 to 1000 Hz bandwidth. In actuality, wethe properties of the input transistor. One way to overcome observed 0.1- A RMS. It is hoped that by reducing the sensorthese limitations is to use a resonant input stage, so that the roughness we will be able to approach the theoretical sensi-sensor current changes more rapidly than linearly with re- tivity. The bidirectional operation of our instrument in thespect to the capacitance. Such a circuit would best operate dc mode has been demonstrated by the frictional data ob-with a FET input which, due to its high impedance, would tained using an iridium tip on HOPG. Imaging capabilitiesdamp the resonant circuit minimally. The higher voltage of our instrument have been shown in both the contactingnoise of the FET input would be outweighed by the resonant and noncontacting profiling mode through images obtainedsignal enhancement. Such a resonant circuit would probably with a diamond tip on diamond films.decrease the noise of our circuit by an order of magnitude, sothat even with the crude wire plate geometry, the noise levelwould match that of the best optical techniques. Using amicrofabricated flat geometry, much better sensitivity could ACKNOWLEDGMENTSbe attained.As dtissed. aWe are pleased to acknowledge S. Chiang, R. J. Wilson.u sdiscussed frc acit e sensingdetail, increasee te d etase- and C. T. Rettner for helpful discussions, J. E. Schlaegel andused for capacitance sensing increases the displacem ent sen- B .H e i o e h i a si t n e n .B d i n a d Tsitivity of the sensor, but the voltage must be limited to avoid BA. Hon for techni aianc and A. adian an -letting the electrostatic force snap the lever into the sensor aia forp nd fil ResaTh wor wsp-plate. In a microfabricated cantilever-sensor system, this tially supported by the Office of Naval Research under Con-problem can be avoided by balancing out the electrostatic tract No. N00014-88-C-0419, and by the AFOSR underforce with two sensor plates, one on the sample side of the Contract No. F49620-89-C-0068. S. R. C. thanks the Myron

A. Bantrell trust for support through a Chaim Weizmanncantilever, and one on the opposite side. Then the sensor-cantilever distance and voltage can be adjusted to give a postdoctoral fellowship.much better S/N than would be possible for one plate alone.

One disadvantage of all AFMs used to date is that todetect the force on the tip, the force must be allowed to dis-place the tip. The relative position of the tip with respect tobeteenthemeaure leer . Binnig, C. F. Quate, and Ch. Gerber. Phys. Rev. Len. 56,.930 (1986).the sample is the difference between the measured lever posi- -G. Neubauer, S. R. Cohen, and G. M. McClelland, Mater. Res. Soc.tion and the controlled sample position. This complication Symp. Proc. 153, 307 (1989).can be avoided by using a zero displacement force sensor. In 'S. R. Cohen, G. Neubauer, and G. M. McClelland, J. Vac. Sci. Technol. Athis scheme, the force on the tip would be balanced by an (to be published)."4T. Gdddenhenrich. H, Lemke, U. Hartmann. and C. Heiden, J. Vac. Sci.opposing force, which would be adjusted by a feedback loop Technol. A 8, 383 t 1990).to keep the lever displacement zero. This method could in 'Y. Kuk and J. Silverman, Rev. Sci. Instrum. 60, 165 (1989).principle be applied with any method of deflection detection, "P. K. Hansma, V. B. Elings, 0. Marti. and C. E. Bracker. Science 242. 209but it is particularly well suited to the capacitance technique, (1988).

"G. Neubauer and G. M. McClelland (unpublished work).in which the sensor electrode is already available for apply- 'S Gould, 0. Marti, B. Drake. L. Hellemans. C. E. Bracker. P. K. Hansma.

ing the opposing force. The opposing force could be applied N. L. Keder. M. M. Eddy, and G. D. Stucky, Nature 332, 332 (1988).to the sensor plate by a variable dc bias superimposed on the "C. M. Mate. G. M. McClelland, R. Erlandsson. and S. Chiang. Phy% Re%

rfsine wave. The effective force constant of the lever could be Lettý 59, 1942 (1987).'R. Erlandsson, G. Hadziioannou. C. M. Mate, G. M. McClelland. and S.

varied by adjusting the gain of the feedback loop controlling Chiang, J. Chemr. Ihys. 89. 5 190) (1988).

the opposing force. Using this method, the dynamic range of R. Erland'son. G. M McClelland. C M. Mate. and S Chiang. J. Vac Sci

measurable forces is not limited by the range of measurable Techniul. A 6. 266 (1988.

lever displacements. [I. J. Mamin. D. Rugar. J 1: ,, Stern. B 1) rcrris. and S. [ lambrl, ApplI'hyv Lett. 53. 1563 (11188).

This zero displacement technique could he implemen- D. Rugar, ft. J Marnin. R. Erp d' 1:n..! Stern, and 1B 1) Tcrrs,. Re%ted by using two opposing sensor plates, one for attractive SOi lnttrum. 59,2337 (9SS)


"D. Rugar. H. J. Mamin, and P. Gunther, AppI. Phys. Lett. 55, 2588 27R. C. Palmer, E. J. Denlinger, and H_ Kawamoto. RCA Rev. 43. 194(1989). (1982).

"~S. Breen, B. E. Paton, B. L. Blackford. and M. H. Jericho (to be pub- 'hY. Martin. D. W. Abraham, and H. K. Wickramasinghe. Appl. Phys.lished)- Lett. 52. 1103 (1988).

`~Y. Martin, C. C. Williams, and H. K. Wickramasinghe, 3. Aj.pl. Phys. 61, "'J. E. Stem, B. D. Terris. H. 3. Mamin, and D. Rugar. AppI. Phys. 1-eU 53,4723 ( 1987). 2717 (1988).

"C. Schbrnenberger and S. F. Alvarado, Rev. Sci. Instrum. 60. 3131 "K. E. Peterson, Proc. IEEE 70. 420 (1982).(1989). ~'W. Y. Lee, R. Sequeda, J. Salem. and D. Chapman. Appl. Phys. Lett. 50,

~D. Sanid, D. Jams, and V. Weissenberger. Opt. Lett- 13, 1057 ( 1988). 522 (1987).~D. Sarid, D. A. [ams, J. T. Ingle, V. Weissenberger, andJ1. Ploetz, J. Vac. '"J. P. Den Hartog, Mfechanical Vibrations (McGraw-Hill, New York,Sci. Technol. A 8, 378 (1990). 1956). pp. 148 ff,

"G. Meyer and N. M. Amer, Appl. Phys. Lett. 53, 1045 (1988). HG. Binnig and D. P. E. Smith. Rev. Sci. Instrum. 57, 1688 (1986).2S. Alexander. L. Hellemans, 0. Marti, 3. Schneir, V. Elings. P. K. 1

4 C. J. Smithells, Metais Reference Book-, 3rd ed. (Butterworths, Washing-Hansma, M. Longmire. and!J. Gurley, J. AppI. Phys. 65, 164 (1989). ton, DC. 1962). Vol. 1. p. 256.

:R. V. Jones and 3. C. S. Richards. J, Phys. E 6, 589 (1973). '5G. Neubauer, S. R. Cohen, G. M. McClelland, and H. Seki, MRS Symp."ZJ. L. Garbini L, J. Albrecht, J. E. Jorgensen, and G. F. Mauer. J. Dynam- Proc. (in press).

ic Systems, Measurement, and Control, Trans. ASME Ser, G 107, 192 ~M. Buck, T. 3. Chuang, J. H. Kaufman, and H. Seki, MRS Fall Meeting(1985). Symp. Proc. (to be published).

2'J. R. Matey and J. Blanc, J. Appi. Phys. 57. 1437 (1985). "'Sample was prepared at Penn State and provided by A. Badzian and T.2, C. D. Bugg and P. J. King,.J. Phys. E 21, 147 (1988). Badzian.7' C. C. Williams. W. P. Hough, and S. A. Rishton, Appi. Phys. Lett. 55,203 ~eP. R Austin. Design and Operation of Clean Rooms (Business News.

(1989). Tray. MI, 1970).


.4.

4- 0

in ILL

a~ z< 111

C' -

ONo

>..

0U24

U -

e -

0C *aj

4.1

. & , - i 4 + +,'+ ,+-3'a3 .. . 2'.

, +.- o -. N.. A. t ' a.

I\-. ~~C ---•++" .-++,+-SI--

o . ,=,m. +-,.,,'=a ...=: +. o+S++•- [-={ + I• - a l]I+ +.'' +,

405' A T.jO g'8

do A id •- e0' • ,e z, 1 J

"oo o= o ), "; , o e-=2, i+.,=31

"a!o 8' " " -, =0 W

C' mttj+ -+.+

S.=,.,=, .., m+ • ..4-6;+." -.] 1 "+" a• -.. ~ =.=-..

-'• 1. I .I 0... .. , ..

a~ A, s a~w 4 E *A

IW"O OS.5-- -3

0 1:. Ago l.. aa L-

o"32 S•t v- 345 + al " r .3 r - "

~-U8

IL A I g

-. 2a - - ... o

~8 MA 8If ! ;Z rm+ = ,- + = .6[;-:• . +. "P -+: + .= ". .. ,

S -.3 -, 0 "" . , , , ,

93 '.--- z " I 0 0++ • .+'"+ = = +• = •.+• ,= _• . . = + • - . . -,I

4,", ..,,,* 0 v' o.- - I--0

0 F0

& -a- - I t

v. _J#3 .; 1. 1-0 a I lb

a.9 1 16.0-

j 0 'b -a Is j.2 9@AA~~ A1 --

sf Oil V~~ad~ i v. .Aa~~az

- -~ -~ 23 OWS ~ ~ 3.1. ~!1 ~ ~ s ~ 33 ~a'U ~A

0- 0-0

u*A j~ -. 4jo

Aa .a . .

d~ 2

:V\3 :0 L

~~1~ pj~W:Ia 3 00 S-a,3Ol~ 02I~W~u~t~WiQJ M 4~U0.2

Z.a 01 1~ 50 0 0*C *a n0 0- - C a

a -a.- -a - b .- '5 0 '-- a l- C 5

* ~ ~ ~ ~ 0 aa.'a -s n u a - C 0. cc *

40 w a .C C a 0- z - C ~ a.-

to -. 0 10 CC sZ- 0 5. . . U O

a.~~~~0 .0 0C 0C 5 - -- s aa

o4 = - n. . 0 O a-ac a 0 a- .. t. -.

-~~~ ea a 0a a- a s 0S.- ;Q I -0.0. .0 0 -a w 0C b C . - 5 9

3.. .5i C C C *OC C*k.C a .C-

-~~~ a- < a- a. -. 0 .k n -C - . t

C6 Z E! 01 te ai -I ta s c C . O

O ~ ~ ~ M 0r - 'wCm ~ O 5 - - - 5 0 5

-3 - 4 S. A S -i ; I~ -a- - -S; -

'31 .0 .- A.C . .4v .. W a C C C C a.a 3

A4 - ) -C aao 0 5 n cs - o 50* e - - . S. '

- .,W*W ~C - .C~a~~a as .* o c

Diamond force microscope tips fabricated by chemical vapor depositionG. J. Germann, G. M. McClelland, Y. Mitsuda,a) M. Buck,b) and H. SekiIBM Research Division, Almaden Research Center, 650 Harry Road, San Jose, California 95120

(Received 6 April 1992; accepted for publication 7 June 1992)

A chemical vapor deposition method is described for fabricating force microscope cantileverswith single-crystal diamond tips. The = l-Am-diamn diamond tips have corner radii of 30 nm,and have been used to study diamond-diamond friction on well-characterized surfaces in UHV.The tip size and orientation can be determined by electron microscopy without altering thesurface atomic structure.

I. INTRODUCTION To begin forming the cantilevers, the end of a 1.0-cm

Atomic force microscopy (AFM) requires the use of length of 0.005-cm-diam tungsten wire is crimped into a

sharp tips for imaging forces on the atomic scale.' How- 0.4-cm length of 0.040-cm o.d. stainless-steel tubing, which

ever, sharp tips are often fragile, especially at higher loads, will become the base post of the cantilever. The wire is bent

such as occur in indentation experiments. 2 Several groups to an angle of 9g0 at a distance of 0.2 cm from the base and

have used tips made of diamond, the hardest substance electrochemically etched with 8-V ac in a 10% aqueous

known. Diamond tips have been formed by fracture3 or by KOH solution, until a t-200-/pm length remains after the

polishing, 4 and ion-impregnated conducting diamond tips bend. The 0.2-cm length gives a force constant of =-30

have been developed for scanning tunneling microscopy. 5 N/m and a res. nant frequency of =5600 Hz. The etchedAlthough a fractured diamond tip can probably be quite end radii vary from _0.1 to 2 pm, with a cone anglesharp, its geometry and surface chemistry are not well de- between = 10' and 25'.fined. In this paper we present a method for preparing A diamond nucleation site is introduced at the tip ofdiamond AFM tips by chemical vapor deposition (CVD). 6 the W wire to facilitate diamond growth. While viewingThese single-crystal tips have a well-defined crystal struc- under an optical microscope, a diamond abrasive blockture and can have quite sharp corners. Furthermore, their (General Electric) mounted on a micropositioning stage isdiamond surfaces, which are grown in a hydrogen-rich at- approached to the W tip. The block is pressed to the tip formosphere, are hydrogen terminated. They therefore have a 1 s at 22X 10-4N, a force monitored by the deflection oflow surface energy, and should become contaminated less a weak wire spring on which the cantilever holder isreadily than other tip materials such as Si, SiO 2, or mounted. Significantly less force than this does not intro-Si3N4. duce a nucleation site on the tip (no diamond growth),

Apart from its utility in AFM applications, CVD dia- while much greater forces cause multiple nucleation sites,mond film shows great promise for the construction of resulting in multiple diamond cryrtals and poorer tips. Ev-improved electronic devices. 7 Diamond can be made to be idently the nuclei deposited from the block are no biggersemiconducting and, in addition to its hardness, has very than = 10 nm, because they could not be observed by scan-high thermal conductivity and a low coefficient of fric- ning electron microscopy (SEM).tion. " As manufacturing techniques become more ad- Diamonds are grown at these nucleation sites in thevanced, its role in production of devices will become more CVD apparatus diagrammed in Fig. 1. The wire cantileverprevalent. Because of this, an investigation of diamond on tips are mounted f 3 mm from a Ta filament consisting ofthe nanometer scale is warranted. We report elsewhere on 0.25-mm Ta wire wrapped in a 4-mm-diam coil = 13-mmthe use of diamond-tipped cantilevers to investigate long. The apparatus is evacuated and CH 4 and H2 arediamond-diamond interactions on the atomic scale.10 To flowed over the filament toward the cantilevers at rates ofaddress diamond's important tribological applications, 0.3 and 60 sccm, respectively. The flows are regulated bythese experiments probe the forces both perpendicular mass flow controllers, and a vacuum pump keeps the pres-(load and topography) and parallel (friction) to the sur- sure constant at 40 Torr. These conditions provide flowface. stability and concentration uniformity during CVD.

Slower flow rates reduce the probability of forming singleII. FABRICATION crystals on the tip and increase the probability of forming

Although we have chosen tungsten for the cantilever less desirable multiple crystals.

material for its stiffness and ease of etching, other cantilev- Monitoring the temperature by an optical pyrometerer materials are suitable for diamond tip deposition, includ- (Wahl DHS-52), the Ta filament is heated to =2100 'C tong materialswhich is commonly used for AFM cantilevers, effect dissociation of the CH 4 and H2 and to heat the can-ing silicon, is com mon d for tilever tips to a temperature appropriate for diamond dep-and as a test substrate for diamond CVD.

osition, = 800 'C. This temperature was not measured di-rectly, but under these conditions the temperature of the

"Present address: Institute of Industrial Science, The University of assembly which holds the cantilevers, as measured by aTokyo, 7.22-1 Roppongi, Minato-ku, Tokyo 106, Japan.

"'Present address: U. of Heidelberg, Institute of Applied Physical thermocouple, is 500 *C. After CVD has continued for I h,Chemistry, D-69fM0 Heidelberg, Germany. the filament and the CH 4 flow are stopped and the canti-

4053 Rev. Sol. Instrum. 63 (9), September 1992 0034-6748/92/094053-03$02.00 Cc) 1992 American Institute of Physics 4053

VacuumPump*

Thermocouple

W-WireCantilevers

-- Ta(~) Filament

CH4 =E- H2

WIFC MFC

FIG. I. Schematic of chemical vapor deposition (CVD) apparatus used

to deposit diamond on AFM cantilevers. The H, and CH 4 flow rates are

regulated by mass flow controllers (MFC)

levers are allowed to cool in flowing H 2. This process al-most always produces a single =lj-m-diam diamond crys-

tal or aggregate of several such crystals at the very tip ofthe cantilever. Generally, fewer than 10 diamonds are seenon other regions of the cantilever which have not beenintentionally nucleated. Since the diamond surfaces aregrown from a gas rich in H atoms, their surfaces are be-lieved to be terminated with hydrogen.'t

Il1. CHARACTERIZATION AND USE

While using a SEM to characterize the diamond tip, it

is important to avoid damage by the 10-keV electron beam.We minimize damage by using a frame grabber to capturea low magnification, 6000 X SEM image of the AFM tip in5 s with a beam current of 0.02 nA. To estimate the beamdamage, we note that for polyethylene, a 10-keV electronwill inflict on average one damage site every 300 A alongits path.12 If the probability of damage to a diamond sur-face, with its single C-C and C-H bonds, is similar to (b)

that for polyethylene, the instance of damage to our dia- FIG. 2. Scanning electron micrographs of diamond AFM tips a tow.

mond tip in the SEM would be 0.3% per (3.5 A)3 unit cell. resolution image after experiments involving 3 • 10 ' N loads Ih) High-We conclude that the SEM imaging does not significantly resolution micrograph of another tip. The outward pointing corncr, " hich

alter the surface bonding of the CVD deposited tip. has an estimated radius of - 300 A, is used for imaging.

The orientation of the crystal facets of the diamond tipon the cantilever cannot be controlled. The "ideal" AFM 0.25 cm. We found that only those 10 cantilevers (the twotip crystal possesses a sharp corner pointing along the W outermost rows of 5) positioned as shown in Fig. I grewtip axis. We have found that a suitable tip, one with a sharp acceptable diamonds. This was probably because the centercorner oriented outward from the AFM cantilever, can be cantilever deposition positions, being closest to the Ta fil-obtained on W once in about 20 attempts. To facilitate the ament, were too hot to support deposition.production process, we deposited diamond tips onto 10 Two sample cantilevers are shown in Fig. 2. High mag-nucleated tungsten wire tips simultaneously. Originally we nification, > 6000X, photos are taken after rather thanattempted to grow diamonds on 25 tips simultaneously, by before the AFM experiments, in order to avoid damagemounting the cantilevers on a square 5 X 5 array spaced by which may result from the high electron intensity inpolved.

4054 Rev. Scl. Instrum., Vol. 63, No. 9, September 1992 Force microscope tips 4054

The gray scale in the. 3 represents not a topographic image"but rather the force parallel to the surface, reflecting thefriction and dynamics along the raster scan direction

In summary, a straightforward method of fabricatingdiamond-tipped AFM cantilevers has been developed. ThisCVD method produces sharp, hard, and inert tips, which

FIG. 3. AFM friction image (57.9x 12.5 A) of a hydrogen-terminated are found to be mechanically robust and wear-resistantsingle-crystal (1O0) diamond surface taken using a diamond-tipped can-r

tilevei. The gray scale range. from 0 (black) to 1.5x 10 ý N (white). Itrepresents the deflection of the cantilever parallel to the surface as the tip the hydrogen-terminated (100) surface of a single-crystalis scanned from left to right and is a measure of the forces along the scan diamond, obtaining friction images with sub-nm resolu-direction. tion.

Note added in proof Visser et aL. have also recently

From these SEM images we find tip radii of = 300 A. The described a CVD technique for fabricating semiconducting

diamond tips form in a variety of shapes and sizes ranging diamond AFM and STM tips. [E. P. Visser, J. W. Gerrit-

from 0.1 to 2 plm, and exhibit almost exclusively ( 100} and son, W. J. P. van Enckevort, and H. van Kemp, Appl.

({11 I} planes. A typical corner used for AFM imaging is a Phys. Lett. (in press)].

four-sided pyramid, with two opposite {1 001 planes at anincluded angle of 90, and two { Il l } planes at an angle of ACKNOWLEDGMENTS109.5'. The sharpest corners are cube corners formed by{ 100} plane junctions. We are grateful to V. S. Smentkowski and J. T. Yates,

The diamond-tipped cantilever is introduced into the Jr. for help with the early experiments, G J. VKubiak andvacuum AFM chamber through a sample transfer/airlock J. Pacansky for helpful discrssions, and J. Vazquez forsystem, and the cantilever is cleaned by heating in a Ta obtaining the electron micrographs. This work was par-oven to 500 °C, under vacuum, for 30 min. This heating tially supported by Office of Naval Research Contractprocess is sufficient to clean the hydrogen terminated dia- N00014-88-C-0419 and Air Force Office of Scientific Re-mond surface of physisorbed contaminants.i'114 After search Contract F49620-89-C-0068.

cleaning, the cantilever is mounted in the AFM and itsmotion is detected with a bidirectional fiber optic interfer- 'For a review, see D. Rugar and P. Hansma. Phys. Today 43. 23 (1990)

ometer. Details of the basic experimental apparatus,'5 and 'P. J. Blau and B. R Lawn, eds.. Mficroindentation Techniques in Ma-

the dual fiber detection scheme16 are described elsewhere. terials Science and Engineering (ASTM, Philadelphia. 1985)'O. Marti. B. Drake, and P. K. Hansma, Appl. Phys. Lett. 51. 484

The normal force between the diamond tips and (~ I 1) (1987); T. R. Albrecht and C. F Quate, 1. Appl. Phys. 62. 2599

or (100) diamond surfaces exhibits van der Waal's attrac- (1987); E. Meyer, H. Heinzelmann, P Grutter. Th. Jung, tlh,

tive forces that are appropriate for the estimated tip radii,'t Weisskopf, H.-R. Hidber, R. Lapka. H. Rudin, and H.-J. Giinth,.odt.

Measuring the maximum attractive force from the curve of 1. Microscopy 152, 269 (1988): G. Neubauer. S R. Cohen, G. M.McClelland. and H. Seki. in Thin Films. Stresses and Mechanical Prop-

normal force versus separation, we find 7.5 x 10 " N. The erues if. edited by M. F. Doerner. W. C. Oliver, G Pharr, and F Rattractive dispersion force"' between a sphere of radius R a Brotzen (Materials Research Society. Pittsburgh, 1990). MRS Symp.distance D from a flat is F .. 4..R/6D2 , where A is the Proc. 188.Hamaker constant, estimated from the diamond optical 'T. Miyamoto. R. Kaneko. and S. Miyake. 1. Vac. Sci. Technol B 9.

1336 (19Q0).constants to be 3.5 10 . s......in estimated tip radius R. Kaneko and S. Oguchi, Jpn. I. Appl. Phys. 29. 1854 (1990).

of 30 nm, and the observecd .+, 'ctive force, this 'R. C. Devries, Ann. Rev. Mater. Sci. 17. 161 (1987): B. V. Spitxyn. L

gives 0.47 nm for the effe ý tip from the L. Bouilov. and B. V. Derjaguin. Prog. Crystal Growth Charact. 17. 79

sample at its potential mi ,tle value, con- (1988).A. T Collins, Se•mcond. Sci. Technol. 4. 605 (1989); K. Shenai an( B,

sidering the effect of surfaL .- J. Baliga, Proc. Inter. Symp. on Diamond and Diamond-Like Films

The diamond - tungsten .... a , risingly strong, (Electrochemical Society, NJ, 1989).and we have never dislodged the tip from its nucleation site 'J. E. Field, ed,, The Properties of Diamond (Academic, London, 1974)

1 N. By SEM we 'B. Samuels and 1. Wilks. J. Mater. Sci. 23. 2846 (1988), and reterenceson the tip, even at loads up to 3 0 yStherein.observe no evidence of wear after weeks of AFM experi- m"G. J. Germann, G. Neubauer, S. R. Cohen. G. M. MeClelland, and H

ments and no progressive change in the attractive approach Seki (to be published).curve for our diamond tip-diamond surface experiments M. Frenklach and H Wang, Phys Rev. B 43. 1520 (1991); P I)eak, I

Giber, and H. Oechsner. Surf, Sci. 250, 287 (1991),during this period.. "T. Okada and L. Mandelkern. J. Polym. Sci A-2 5. 239 (1967); L

To illustrate the use of the diamond tip, we show an Mandelkern. in The Radiation Chemistrv of Macromolecules, edited by

image in Fig. 3 of the friction force between a diamond tip M. Dole (Academic, New York, 1972). Vol. I. p. 287.and a single crystal diamond (100) surface.') The image "A V. litamza. G. D. Kubiak, and R. If Stulen, Surf. S.1 237. 35

( 1990).clearly shows structure in the form of rows along a (210] "If. B. Pate, Surf. Set 165. 83 1198h1direction of the bulk diamond lattice, spaced by 2.3 k. "G. Neubauer, S. R. Cohen. G. M McClelland, 1) Home. and C M

These are consistent with a 2X I monohydrogen termi- Mate. Rev Sci Ins•rum. 61, 2296 (tOQO).

nated reconstructed surface.13 We believe that averaging "'D. Rugar. It J. Mamin. and P. iiinthet, Appl. Phys Ltett 52. 2588(1989)

over an asymmetric AFM tip prevents the observation of '1,. N. Israclachvmli. lntermolecular and Surjface Forres (Academrc,structure perpendicular to the rows shown in t!ie figure. Iondon. 1985)

4055 Rev. Sci. Instrum., Vol. 63, No. 9, September 1992 Force microscope tips 4055

Atomic scale friction of a diamond tip on diamond (100) and (111) surfacesGeoffrey J. Germann, Sidney R. Cohen,a) Gabi Neubauerb) Gary M. McClelland,c)and Hajime SekiIBM Research Division, Almaden Research Center. 650 Harry Road, San Jose. California 95120

D. CoulmanE. I. Dupont, Experimental Station, P. 0. Box 80326, Wilmington, Delaware 19880-0356

(Received 1 June 1992; accepted for publication 22 September 1992)

The friction of a clean diamond tip on diamond (I ll ) and (100) surfaces is studied using anultrahigh vacuum force microscope that simultaneously measures forces parallel andperpendicular to the surface. The 30 nm radius diamond tip is fabricated by chemical vapordeposition. The attractive normal force curve between the tip and surface agrees well withcalculated dispersion interactions. The frictional force exhibits periodic features, which on the(100) surface are tentatively associ'ated with a 2 X I reconstructed surface convoluted over anasymmetric tip shape. The (I I ) surface shows features that cannot be simply related to thesurface structure. As the tip is scanned back and forth along a line, the same features areobserved in each direction, but offset, suggesting the presence of a conservative forceindependent of the direction of motion as well as a nonconservative force. The friction isapproximately =3x 10-9 N independent of loads up to I X 10-7 N.

INTRODUCTION to measure the lateral force directly,"' the fiber must be

Interfacial solid-solid sliding friction is customarily as- positioned very close and parallel to the surface. Our de-

sociated with wear, but atomic force microscope (AFM)I sign (Fig. 1) avoids this problem by symmetrically posi-and surface force apparatus 2 experiments show that sliding tioning two fibers 90* apart in a plane perpendicular to the

without wear is feasible, and AFMs have directly measured sample and the cantilever. From the normalized sum and

atomic scale frictional force variations.3- 7 With some ex- difference of these two signals, the deflection normal

ceptions, 8 little attention has been given to the tip morphol- (load) and parallel (friction) to the surface is determined.ogy or composition, and, as discussed by Cohe9 many Forming these linear combinations has no effect on the

den, noise of the deflection measurement. Because variations inprevious measurements of tip-surface forces have probably laser diod- nower are subtracted out,15 the 0.15 A (rms)been affected by interfacial contamination. Here we present noise measured over a scan line was caused mainly bya study of atomic level friction between (100) and ( 111) novemeng e overia s l a e ydiamond surfaces and a diamond tip in ultrahigh vacuum Avlentvriation.Thisis he irs obervtionof he tomc sale All experiments used the diamond tip displayed in Fig.(UH'v). This is the first observation of the atomic scale 2(a). As described elsewhere,16 the tip was grown onto thestructure of friction in UHV, although several vacuum end of the tungsten cantilever (force constant, 30 N/i) bystudies, including our work on films formed by chemical CVD. Before the AFM measurements, it was examined atvapor deposition (CVD),10 have probed a somewhat larger low maanigelectro microcopscale.7 °0 11 Because of its unique hardness, diamond is an low magnification by scanning electron microscopy

ideal material on which to observe wearless friction. We (SEM), using low electron exposure to insure negligible

observe variations in the lz1eral force associated with the alteration of the surface bonding.' 6 By examining high-resolution SEMs obtained from several viewing angles after

diamond crystal lattice, and find that the frictional force the AF M meas ne tip was found togbe for

varies little with load. The lateral tip-surface force shows a two A lM a nd two tip pas found t(be and

strong conservative component, which can be as large or oreteo that the sam place norma was are

larger than the nonconservative frictional component that te (1 tino th e tip.aTe fl sige rystal sm

dissipatesthe (110) direction of the tip. The flat single-crystal sam-pies were polished with diamond grit in olive oil, rinsed inseveral solvents, and blown dry. After being placed in the

EXPERIMENT ultrahigh vacuum (UHV) chamber, both surfaces and tip

To probe frictional forces, several bidirectional AFMs were raised to 500 *C for 30 min and returned to 25 *C, a

have been designed that measure both lateral (frictional) process known to leave cledn ordered hydrogen-terminatedand normal (loading) forces. These designs have used ca- surfaces.171 8 The vacuum was limited by outgassing of the

pacitance measurement,12 laser beam deflection,"'7 and op- fiber cladding to 5 I0 Torn

tical fiber interferometry. '3 " If fiber interferometer is usedRESULTS

"*Present address: Department of Chemical Physics, Weirmann Institute The t-p-surface force normal to the surface shows aof Science, Rehovot 76200, Israel.

h'Present address: Intel Corp. P. 0. Box 58119, Santa Clara, CA 95052. broader and deeper attractive well upon retraction than

"•'To whom correspondence should be addressed. during approach (Fig. 3). As demonstrated by molecular

163 J. Appl. Phys. 73 (1), 1 January 1993 0021-8979/93/010163-05$06.00 c 1993 American Institute of Physics 163

In te r - ( a ) 1 P - - -

Normal

force

tLateralforce

.Sample feo't"

FIG. 1. Dual fiber interferometer for measuring lever deflection normal(load) and perpendicular (friction) to the surface. The two interferom-eters are optically independent.

dynamics (MD) calculations, 19 no plastic deformation orchemical processes are expected at these loads forhydrogen-terminated diamond surfaces. The hysteresisthus indicates the presence of physisorbed molecules orflexible groups chemically bound to the sample or tip. Fora sphere of radius R a distance D from a flat, the attractivedispersion force is approximately F= -AR/6D , where Ais the Hamaker constant, estimated from the diamond op- (b)tical constants to be 3.5 X 10-' J. This function is plotted (100)in Fig. 3(a) for the 30 nm tip radius estimated by SEM.The observed maximum attractive force of 7.5x 10 - Ngives 0.48 nm for the effective distance of closest approach. (111)a reasonable value considering probable roughness of thetip. Comparing Figs. 3(a) and 3(b), the approach curvevaried little over the weeks required to do the experiment,suggesting little tip wear.

Hertz showed that the radius a of the elastic contact ofa spherical tip of radius R with a flat is [3PR(I-v2)/

2E]1" 3 , where P is the load and E and v are the Young'smodulus and Poisson's ratio.20 Using R = 30 nm, the con-tact radius a is 1.6 nm at the largest 10 7 N loads used,giving an elastic deformation of each of the two solids ofa 2/2R =0.042 nm, and MD simulations indicate that the FIG. 2. (a) Scanning electron nucrograph of the diamond tip) grown on

additional deformation at the interface is well below 0.1 the end of an etched tungsten tip. The corner pointing ouiward along the

nm.1 Surface roughness probably leads to a smaller con- wire axis was used for imaging. (h) iometri: model ofthe tip

tact area and larger deformation.An image [Fig. 4(b)] of the frictional forces on the temperature range extrapolated, it is possible that recon-

(100) surface was obtained at 1.5 X 10 - N load and a scan struction could occur under our heating conditions. Fric-rate of 2.3 nm/s while oscillating the sample 0. 15 nm peak- tion on the (Ill) surface at 1.5 ,, 10 N load shows lessto-peak (p-p) at I kHz along the scan direction to increase well-defined rows [Fig. 4(c)]. bearing no clear relation tocontrast. The image shows rows along the (021) direction the diamond ( I l l) lattice. The approximate parallelism ofspaced by a single lattice constant along the (001) direc- the predominant features on the (100) and ( 111 ) surfacestion, a pattern consistent with a 2x I reconstruction. The suggests that the tip geometry may play a role in empha-lack of structure along (021) may be du,: to averaging over sizing this direction. Simultaneous measurement of thethe asymmetrical contact area of the tip, whose shape and normal deflection of the lever shows no atomic level struc-orientation is shown in Fig. 4(a). Although the 2 X I re- ture within the 0.07 nm p-p noise, but on the (100) surfaceconstruction was previously observed only at much higher we observed 0.3--1.5 tnm high striations approximately 5temperatures than we used, the measured Arrhenius pa- tim apart. That no topographic features narrower than 2rameters for the associated hydrogen desorption' 8 predict nm were observed may reflect the size of the diamond tip.that the reconstruction would go 5% toward completion Recently, atomic-level structure of CVD diamond filmsduring our 30 min heating. Considering possible errors in has been observed by STM21 and AFM..2

measuring the temperature of the diamond surface, and The "'friction loops"' of Fig 5 (without small ampli-deviation from the strict Arrhenius relation over the large tude oscillation) on the (1(XI) surface are obtained by

164 J. Appl. Phys., Vol. 73, No 1, 1 January 1993 Gprmann el al 164

8 - . .- . scanning the sample alternately back and forth along thez (a) same path in opposite directions. The sample position isS *, subtracted from the measured tip position to generate the

-" 0 _ abscissae of Fig. 5. Measuring the force parallel to the scano ,direction [Fig. 5(a)]. the loop shows very similar features

a . - Retract along opposite scanning directions, but the force offset be-E°.- . Approach tween the two directions reveals a net dissipation of me--8 * Calculated chanical energy. The diagonal straight lines in the loop

-12 4 6 8 .. 1 14.. 6 indicate "slipping" of the lever, which occurs when the0 2 4 6 8 10 12 14 16derivative of the tip-surface force changes from being

Z (b) greater than to being less than the lever force constant.,4- The shape of the loop, which changes little during consec-

- utive scans, suggests that the lateral force is the sum of a0 4 - - position-dependent conservative part, independent of slid-

S-4 ., -, ing direction, and a position-independent dissipative force,a ' *" °" whose sign reverses with sliding direction.

Ez -8 When the surface is scanned at an angle 45' with re-

-12 . .. ... .. . . . . . . .. spect to the layer axis [Fig. 5(b)], approxim ately parallel0 2 4 6 8 10 12 14 16 to the features of Fig. 4(a), very little hystersis is observed.

Tip - Sample Separation (A) In fact, during a scan in a particular direction, the lever is

deflected laterally to both positive and negative sides of itsequilibrium position. The decrease in hysteresis can be par-

FIG. 3. Normal force between the tip and the surface as the sample is first eq attriu teapproached and then withdrawn from the tip. Positive forces are repul- tially attributed to the fact that the sliding direction is 45'sive. Curves (a) and (b) were recorded 40 days apart. with respect to the force sensing direction. Also, for sliding

[101]..L/(a) (10]) 1

(b)

(110-0• • (01

[110] (0111

(c)

FIG. 4. (a) Schematic of the tip, viewed as if looking through the tip along the sorface normal, with the proper experimental orientation relative to thegrey scale images. (b) 5.79X 1.25 nm image of the frictional force between the diamond tip and a (I10()) surface. The gray w-ale, ranging from (1 to1.1 / 10 ' N. represents the deflection of the cantilever parallel to the scan direction as the tip is scwanned front left to right. (c) Friction on the ( I I)surface, area 5.8 X 1.25 rim. grey scale range 0-11 x 10 ý N. The sample orientation is determined by x-ray diffraction

165 J. Appl. Phys., Vol. 73, No. 1, 1 January 1993 Germann et al 165

S- -,* Fritio

°" 4 ,(•AomcFatuej

3. -.. ' .-

5.

o 1.. 0"- 0]

31

0 5 10 0 15 0 25 30 0 25 a5 N 5u10be2

Z 50Co

lip~~~ ~ PosrincntSipeon,

4I.6 ocsmaue oer frAton tomic Fst ea( t 1 urfaeisfis

0 460

E 04-40

It20

z 10

-20 - --- ~--- -r - i.-. 0 25 50 75 100 1?

0 10 20 30 40 50 Scan Nmelip Position on Sample (4)

FIG. 6. Forces measured over friction loops as the (Ill)I surface is firstadvanced toward, then withdrawn from the tip while being scanned back

FIG. 5. Lateral tip-surface force on the (100) surface as the sample is and forth. The horizontal a'.i: is the sequential scan number. (a) Averagemoved back and forth a:.-; a single line. Several sequential loops are friction force (note inverted vertical scale) and size of force- features. (b)shown, offset %ertically. The solid trace represents the right-to-left scan, Average force normal to the surface (load)the data points, left-to-right. Positive (up) forces oppose right-to-left mo-tion. (a) (100) surface with zero net load, scanning in the same directionas in Fig. 4(b), along the force-sensing direction. (b) ( 100) surface at1.5 x 10-' N loa(', scanning in the direction rotated r-/4 counterclockwise at a particular place on the surface. The average frictionalfrom the scan direction of Fig. 4(b). The force component is measured hysteresis was the same within 10%.along the scan direction of Fig. 4(b).

DISCUSSION

in this direction, the lever cannot become "stuck" on the The minimal load dependence of the frictional forcesurface because there is a component of the sample motion observed here is in sharp contrast to macroscopic systems,along the axis of the cantilever, a direction in which it is in which friction increases linearly as the load increases,not compliant. These results emphasize the existence of a causing more asperities to contact. Our experiments corre-conservative force parallel to the surface, dependent on spond more to the contact of a single asperity. For a singleposition but unrelated to the sliding process itself. "Hertzian" elastic contact of a sphere on a plane, the con-

The average friction, defined as half the difference be- tact area increases as P213, a stronger variation than wetween the average lateral force along the scanning direction observe. If the lateral force were due to the slope of thein the two directions of the friction loop, is plotted versus local topography, the conservative lateral force would beload in Fig. 6. As the surface is brought into contact with proportional to load by a proportionality constant equal tothe tip, friction increases suddenly at the attractive force the local slope of the surface. Even for atomically flat sur-maximum, but it changes little in the repulsive regime, as faces, the shear strength of the interface sometimes de-the load varies between 0 and I X 10-7 N. The magnitude pends linearly on load.23 Since none of these models areof the rms lateral force deviation from its average along a consistent with our results, it may be that the load is car-single scan direction (labelled atomic features in Fig. 6), ried by a different region of the tip than that which deter-which we associate with the conservative potential, also mines the lateral tbrces. At the highest loads, the ratio ofdepend little on load. frictional force to load is 0.05, and the friction is essentially

Although the qualitative frictional properties described independent of load, so that the "differential friction coef-were observed repeatedly, variations of up to a factor of 2 ficient" is essentially zero. It has been predicted that fric-in frictional forces were observed from day to day on dif- tion between strong smooth solids interacting by nonchem-ferent areas of the surface. Within this range, the average ically bound forces should disappear unless the solids arefrictional forces observed on the (100) and (111) surfaces commensurate, 24 even at zero temperature. Recent MDwere the same. The dependence of lateral force on sliding calculations of diamond friction show zero friction along adirection was probed by comparing the sliding in the two particular sliding direction. 25 Thermally activated pro-directions 45' with respect to the lever axis [as in Fig. 5(b)] cesses should act to reduce friction.

166 J. Appl. Phys., Vol. 73, No. 1, 1 January 1993 Germann et a[ 166

The largest loads we used correspond to a contact 'For a review. see D. Rugar and P Hansma. Phy% Today 43. 23 (1990)

loading stress of 20 GPa. a value also attained in experi- See, for example, M L, Gee. P. M. McGuiggan. J. N. sraelachs il, and

ments using a much larger 60 jum radius tip.26 At this A. M. Hornola, J. Chem. Phys. 93. 1895 (1Q9W)) S. Granick, Science283. 1374 (V991),

stress, the ratio of friction to load is similar to the value we 'C. M. Mate, G. M, McClelland. R. Erlands.so,. and S Chiang. Phy,

measure, but in the larger scale experiments, the friction Rev. Lett. 59, 1942 (1987).

increases strongly with load, in contrast to our observa- 'R. Erlandsson, G. Hadziioannou, C. M, Mate. G M. McClelland. andS_ Chiang, I. Chem. Phys, 89, 5190 (1988); E. Meyer, R. Oerney. L

tions. Howald. D, Brodbck, R. Lurhi, and H.-), Guntherod., in Fundamen-Concerning the friction of diamond, there is an exten- ials of Friction. edited by 1. Singer and 11 Pollock (Kluwer. Dordrecht)

sive literature, which has been summarized by Samuels and (in press).

Wilks. 2 6 They speculate that energy dissipation during fric- 'O. Marti, 3. Colchero. and .. Mlynek. Nanotechnol. I. 141 (1990)'S R. Cohen, G Neubauer, and G. M. McClelland, J. Vac. Sc. Technol.

tion occurs by two mechanisms. In the "ratchet mecha- A 8, 3449 (1990).nism," energy is released by the sudden transfer of the 7G. Meyer and N. M. Amer. Appl. Phys. Lett 57. 2089 (1990).

loading force from one asperity to another. In the elastic 'Q. Guo, J. D. J. Ross, and H. M. Pollock. MRS Symp Proc. 140. 51

mechanism, the released energy comes from elastic strain (1989): M. D. Pashley, J. B. Pethica. and D. Tabor, Wear 100. 7(1984); D. Maugis and H. M. Pollock, Acta Metallurg. 32. 1323

in an asperity. The sudden slips we observe could be con- (1984).

sidered a type of ratchet mechanism, in which the lever 'S. R. Cohen, Ultramicroscopy 42. 66 (1992).itself supplies a lateral loading. Yet the lateral force shows "'G. Neubauer. S. R. Cohen. G. M. McClelland, and H Seki, in Thinhysteresis even where the ip is not sticking. In the elastic Films: Stresses and Mechanical Properties 11. edited by M. F. Doerner.

W. C. Oliver, G. Pharr. and F. R. Brotzen (Materials Research Society.

asperity mechanism, for a sudden slip to occur between Pittsburgh. 1990). pp, 219-224.neighboring minima of the interfacial potential, the total 15G. Neubauer. S. R. Cohen. and G. M. McClelland, in Interfaces Be.

lateral distortion of the asperities would have to be at least tween Metals, Polymers. and Ceramics. edited by B. M. DeKoen. A- .1

0.1 nm, the minimum (atomic) scale over which the inter- Gellman. and R. Rosenberg (Materials Research Society. Pittsburgh.1989). p. 307.

facial potential varies. Using the fact that the lateral stiff- "G. Neubauer, S. R. Cohen, G. M. McClelland, D. E. Horn, C. NI. Mate,

ness20 of a contact of radius a is to a good approximation and C. M. Mate, Rev. Sci. Instrum. 61. 2296 (1990).

aE, for the typical observed lateral force of 10- N, the "G_. M. McClelland and J. N. Glosli. in Fundamentals ofFriction, editedby H. M. Pollock and 1. L. Singer (Kluwer, Dordrecht) (in press)

contact would then need to be less than 0.1 nm in radius, 14C. M. Mate, Phys. Rev. Lett. 68, 3323 (1992).

too small to be considered an asperity. Thus, the deflection "5D. Rugar, H. J. Mamin, and P. Gunther. Appl. Ph•s. Lett. 52. 2588

of individual bonds is involved in the friction. For example (1989).

the bending potential of a C--C-C bond is weak enough ."G J. Germann, G. M. McClelland, Y. Mitsuda. M. Buck. and H Seki.Rev. Sci. Instrum. 63, 4053 (1992).that a 0.1 nm deflection can be attained with a force of '7B. B. Pate, Surf. Sci. 165, 83 (1986).

4X 10-9 N. From these considerations, pendant groups On '"A, V. Hamza, G. D. Kubiak. and R. H. Stulen, Surf Sci 237. 35

the surface, which act to decrease the effective force con- (1990).stant for surface atoms, may enhance friction.24 "'J. A. Harrison, D. W. Brenner, C. T. White. and R. 1 Colton. Thin

Solid Films 206, 213 (1991); J. A. Marrison. C. T. White. R J. Colton.

and D. W. Brenner, Surf Sci. 271,57 (1992).2"K. L. Johnson, Contact Mechanics (Cambridge University Press. Cam-

ACKNOWLEDGMENTS bridge, 1985).2'T. Tsuno, T. Imai, Y. Nishibayashi, K. Hamada, and N. Fujimori, Jpn.

1. Appl. Phys. 30. 1063 (1991).We thank V. Smentkowski and J. T. Yates, Jr. for 22L. F. Sutcu, M. S. Thompson, C. J. Chu. R. ff. Hauge, J. L. Margrave,

encouraging our interest in diamond friction and for help and M. P. D'Evelyn, Appl. Phys. Lett. 60, 1685 (1992).with our early experiments. B. A. Hoenig assisted with 2

-B. J. Briscoe and D. C. B. Evans, Proc. R. Soc. London A 380. 389

computer programming. We enjoyed helpful discussions (1982).'J. B. Sokoloff, Phys. Rev. B 42, 760 (199); G. M. McClelland, inwith G. D. Kubiak and J. A. Harrison. S. R. C. thanks the Adhesion and Friction, edited by M. Grunze and 14 3. Kreuzer,

Myron A. Bantrell Trust for a Chaim Weizmann Postdoc- Springer Series in Surface Science 17 (Springer, Berlin, 1989), p. 13 M.

toral Fellowship. This work was partially supported by the Hirano, K. Shinjo. R. Kaneko, and Y Murata, Phys. Rev. Lett. 67.Office of Naval Research contract N00014-88-C-0419 and 2642 (1991).

'J. A. Harrison, C. T. White. R. J. Colton. and D. W. Brenner. Phys.the Air Force Office of Scientific Research contract Rev. B 46 (1992) (in press).F49620-89-C-0068. 26

B. Samuels and J. Wilks, J. Mater. Sci. 23. 2846 (1988).

167 J, Appl. Phys., Vol, 73, No. 1. 1 January 1993 Germann et aL 167

RJ 9215 (81523) February 9, 1993Surface Science

Research Report

OBSERVING THE MOTION OF A SINGLE ADSORBED ATOM WITHPICOSFCOND ,ND SUB-NANOMETER RESOLUTION

II. Ifein7eImannF. Watanabc

(. McClclland

IBM Research DivisionAlmaden Research ('Center650 flarry RoadSan Jose. California 95120-6099

LIMITED DISTRIBUTION NOTICE

This report has bepn submitfed for publication outside of IRM and will probably t- copyrighted if aicvlped for publication It has been issued asa Reso., -h Report for ealay dissemination of its contents In view of tihe transfer of copy•rin1t to tlia outside publisher its dislfibution otsidle ofIBM prior to publication should be limited to tpeer communications and $pecific rali$lý Aill., ot utidn iptublicatin roquosts should be filled

only by rpprintS nr lpoally obtained ropiae of the articlet, to . payment of rnyilt,'sl

Research Division

7. = Yorktown Heights. New York 0 San Jose, California * Zurich, Switzerland

RJ 9215 (81523) Februry 9, 1993Surface Science

Observing the Motion of a Single Adsorbed Atom withPicosecond and Sub-nanometer Resolution

Harry IIcinzelnmantFumiya WatanabeGary M. Mzf'!llandt

IBM Research DivisionAlmaden Research Center650 Harry RoadSan Jose, California 95120-6099

ABSTRACT

A new method is described for recording the motion of a single adsorbcd atom or moleculein real time, with an ultimate achievable resolution of lO1l sq. he t1me dependence ofelectron field emission through an atom adsorbed on a mctal tip is measured by !ocuecingthe electrons into a beam, which is swept across a detector screen. Simultaneoiis temporaland one-dimensional spatial resolution of the dynamics is possible. lhe motion of zingleCs atoms between sites on atomically characteri7ed W( lll) tips is observed at 2 ps resol-ution.

"iCurrcnt addrcss;Institute of PhysicsUniversity of Bascl4056 BasclSwitzerland

STo whom corresponldence shohld be addressed

The physical and chemical properties of matter are determined by atomic motion, which

occurs on the length scale of Angstroms and on the 10" 4- _1012 s time scale of vibration.

Until recently, temporal information on this scale has been deduced only indirectly from

scattering experiments or from spectral data in the frequency domain. Pump-probe ia.er

techniques 12 have now been applied to large ensembles of molecules to achieve

temporal resolution good enough to resolve molecular vibrations in stable and reacting

molecules,I and time-resolved electron diffraction techniques have achieved somewhat

poorer resolution.3 Optical techniques are generally incapable of continuously observing

the motion of single atoms and molecules on picosecond and shorter time scales, because

they are insufficiently sensitive, and because the optical probes transitions drastically

perturb the motion by changing the electronic state. On a much slower microsecond

time scale, scanning tunneling microscopy4 (STM) and field emission microscopy,5 .6 .7

which can also be applied to thermal ensembles,8 have been used. The jumping of single9

atoms can be traced by field ion microscopy (FIM), but with resolution of only - I s.

STM has been combined with pump-probe laser techniques to monitor electronic relax-

ation with sub-nanometer resolution, and an STM switch based on the motion of a

single atom has been demonstrated.11

We have developed a method for continuously observing the motion of single

adsorbed atoms or molecules by recording the variation of field emission intensity from

a metal tip. The jumping of a Cs atom between adsorption sites on a tungsten tip has

been recorded with 2 ps and sub-nanometer resolution. In our femtosecond field emis-

sion camera (FFEC),12 electrons field-emitted from a tip and selected by a lens aperture

are focussed into a spot, which is swept electrostatically across a detector screen (Fig.

I a). The spatial variation of the detected intensity along the sweep records the temporal

variation of the field emission intensity (1 mm = 10 - 100 ps). Field emission (FE) is

a tunneling process, which occurs when the width of the potential barrier at the surface

of a metal tip is reduced by applying a negative potential (Fig. 2). Electrons originate

from the metal interior, but the tunneling rate depends exponentially on the local barrier

2

height, which is determined by the identity and location of surface atoms. FE thus

provides a method for monitoring the position of surface atoms without strongly per-

turbing them by electronic transitions.

In our instrument, which has a base pressure of 2 x 10"11 Torr, a lens optimized

for field emission14 is positioned 7 mm from the tip. The beam electrons are multiplied

by a dual microchannel plate before sti•riing a phosphor screen, which is imaged by a

vidicon and frame grabber. A krytron pulse generator15 supplies the - 2 kV/ns sweep

voltage. To imagc the tip by field ion or field emission microscopy,13 the lens voltage

is turned off or reduced, so that the angular distribution of emitted electrons or ions is

projected onto the detector (Fig. lb).

Fink16.17 showed that the triatomic end of a tungsten tip can continuously emit

currents as large as 10-5 A, or 6 x 1013 e/s, which gives many electrons during the

10"13 to 1012 s vibrational period of adsorbed heavy atoms. To begin our experiments,

we sputter-sharpen a chemically etched <111> W tip to a - 300 A radius by field

emitting in 5 x 10- Torr Ne.16 Annealing briefly at 1100K forms a corner of three

{ 2111 planes.16 Using field ion microscopy13 (FIM) to image the tip atoms, we field

evaporate to form an apex plane composed of a triangle of three W atoms16 (Fig. 3).

Individual Cs atoms can be observed in FE, because the partially ionized adatom

greatly lowers the local work function18 (Fig. 2). A Cs atomic beam is directed at very

low flux onto a 90 K tip, so that the average time between the arrival of atoms detected

in the viewed area (= 20 A diameter) is - 60 s. To ensure that no more than one Cs

atom resides near the apex, frequent positive voltage pulses evaporate Cs atoms from

the tip, while leaving the W atoms unmoved. Deposition is ended when a bright spot

(Fig. I b) appears near the image of the apex, increasing the FE by a factor of 5 to -

10' 'A, and signaling the arrival of a single Cs within I nm of the apex. At 90 K, the

atom remains at this site.

3

A series of alternating sweeping and imaging experiments are now performed, so

that the position of the Cs atom is determined after each sweep. To sweep, the tip is

pulsed with a negative voltage for several ns, generating a 5 pA e- beam, which is fo-

cussed and swept across the screen to record the emission intensity displayed in Figs.

3 and 4. The high current heats the adsorbed Cs to - 800 K, presumably by inelastic

electron tunneling6,9 or the Nottingham effect, so that during the sweep, it may jump

to another site.2 1 Following each sweep, FE images the Cs after it has cooled (Fig. lb).

During the brief tip pulse the strongly bound W atoms do not jump.

Because the tip pulse is rounded, if no Cs motion occurs, the FE will be bell-

shaped, as in sweep 6 of Fig. 3. Judging from the FE images (Fig. 3) at 90 K, the Cs

occupies one of two sites in these sweeps. (Some other sites were observed in other

sweeps.) In sweeps 1, 7, and 8, the Cs jumps between sites. During sweep 1, an abrupt

increase in FE is observed, with a rise time within 20 ps. The atom has moved from

the position giving lower field FE, occupied in sweep 6 and designated "L", to a position

designated "H," resulting in higher FE. The images indicate that Cs has moved from

the upper left hand portion of the image to the center, increasing in brightness because

the field is stronger at the tip apex. Between sweeps I and 6, the atom has moved back

to position L. In sweep 7 an L-+H transition is observed at about the samL !ime as in

sweep 1. Although the jumping of the atom is a statistical process, in both sweeps, the

jump occurs at the time when the FE is sufficiently strong to heat the Cs vibration high

enough to activate the atom. During sweep 8, Cs moves H--L. The FE changes back

to the curve of sweep 6 characteristic of the L position, and the FE image correspond-

ingly returns to that after sweep 6. After the experiments, FIM confirmed that the atomic

geometry of the W tip was unchanged by either the negative or positive pulses.

When the sweeping e- beam is not perfectly focussed, the location at which an

electron strikes the detector screen will depend on the position of its emission from the

tip. This spatial information is lost along the sweeping direction, but the .ime depend-

ence of the spatial distribution of the emitted electrons perpendicular to the sweep di-

4

rection can be recorded. Fig. 4 shows such a measurement, at increased sweep speed.

Four traces from adjacent detector pixels are presented, each displaying a distinct time

dependence from a portion of a 20 A diameter region of the tip. The two upper traces

from the upper portion of the tip show a sudden increase in the FE at about 300 ps, while

the two lower traces, representing the bottom, show no increase. The sudden increase,

with an apparent instrumental rise time of 2 ps, is consistent with a Cs atom moving

suddenly from the edge to the upper portion of the viewed region of the tip.

Compared to photoabsorption, which switches atomic motion to a new potential

surface, perturbation of the atomic trajectories by a single field-emitted electron is weak,

because the electron mass is so small compared to the Cs mass. Yet the net result of

many such interactions over many vibrational periods is substantial heating. The field

emission of single electron could strongly perturb atomic motion if it left a localized

empty state which is not quickly refilled. The heating of ultrasharp field emission tips

by high DC currents has been observed recently.6

22.

The FFEC is essentially an electronic streak camera in which the photoemitting

electron source has been replaced by a field emitter, which is itself the sample of interest.

The time resolution achievable by our instrument is in principle limited by four factors.

The most obvious is the ratio a/v involving the velocity v and the size a of the focus spot

at the detector. In practice this effect is not limiting, because the tip is a nearly perfect

point source for sharp focussing, and high sweep rates of 10 kV/ns are readily achieva-

ble. Inhomogeneities in the deflection field are not a problem, because we distinguish

only relatively few (256) time points at small deflections.

A fundamental resolution limit is the distribution of the time of flight (TOF) be-

tween the tip and deflecting field caused by the = 0.5 eV23 kinetic energy spread. of

the emitted electrons. Two electrons emitted simultaneously but arriving at the de-

flecting field at different times will be deflected to different positions on the detector.

In a spatially uniform accelerating field, most of the TOF spread occurs just after the

electrons are emitted, when the electrons are moving slowly and the transit time is

5

strongly affected by the energy. But in field emission from a point, most of the potential

drop between tip and lens occurs within 0.5 mm of the tip, and the accelerating field at

the tip ,s so strong (= 1 V/A) that the TOF spread contribution due to initial acceleration

is only I x 10"14 s. In our instrument, most of the effect of the energy spread on the TOF

occurs after the region of strong initial acceleration, but before the lens, where the

electron kinetic energy is - 1000 eV. This region is traversed in 370 ps, so a - 0.5 eV

energy spread causes a TOF spread of ('A)x370x.5/1000 = 0.090 ps. Over the much

longer path between the lens and the deflecting field, the TOF spreads only 0.050 ps, a

smaller value because the electron energy is a much higher 8 kV. A TOF spread esti-

mated at 0.3 ps arises from the dependence of the flight time of electrons of a given

energy on the off-axis angle with which they leave the tip.

The net results of these effects is that, at the highest sweep rates, the time resolution

of our instrument in 0.4 ps, a number which can be reduced by a factor of 10 by scaling

down the size of conventionally manufactured electron optics. Microfabricated optics24

could enable achievement of the 10"14 s resolution limit imposed by the initial acceler-

ation.

Our measurements exceed by a factor of 106 the speed of the fastest previous

continuous measurements of single atom motion.4'7 In addition, one-dimensional spatial

resolution has been demonstrated, and with more complex electron optics, full temporal

and two-dimensional spatial resolution can be achieved. In future experiments, our in-

strument will be used to time resolve the trajectories, including the vibrational motion,

of individual atoms and molecules undergoing diffusion, adsorption, desorption, and

reaction.

This work was partially supported by the Office of Naval Research and the Air

Force Office of Scientific Research. H. H thanks the Swiss National Science Foundation

and the Treubelfonds, Basel, Switzerland, for fellowships.

6

References

1. M. Dantus, H. H. M. Janssen, and A. H. Zewail, Chem. Phys. Lett. 181, 281 (1991);

L. Khundkar and A. H. Zewail, Ann. Rev. Phys. Chem. 41, 15 (1990).

2. F. Budde, T. F. Heinz, M. M. T. Loy, J. A. Misewich, F. de Rougemont, and H.

Zacharias, Phys. Rev. Lett. 66, 3024 (1991); J. D. Beckerle, R. R. Cavanagh, M.

P. Casassa, and E. J. Heilweil, J. Chem. Phys. 95, 5403 (1991).

3. H. E. Elsayed-Ali and G. A. Mourou, Appl. Phys. Lett. 52, 103 (1988); J. W.

Herman and H. E. Elsayed-Ali, Phys. Rev. Lett. 68, 2952 (1992); J. C. Williamson,

M. Dantus, S. B. Kim, and A. H. Zewail, Chem. Phys. Lett.

196, 529 (1992).

4. G. Binnig, H. Fuchs, and E. Stoll, Surface Sci., 169, L295 (1986).

5. Ch. Kleint, and K. Mtckel, Surface Sci. 40, 343 (1973).

6. V. T. Binh, S. T. Purcell, G. Gardet, and N. Garcfa, Surface Sci. 279, L197 (1992).

7. I. Brodie, Surface Sci. 70, 186 (1978).

8. Ch. Kleint, Surface Sci. 25, 394 (1971); J.-R. Chen and R. Gomer, Surface Sci.

79, 413 (1979).

9. G. Ehrlich and F. G. Hudda, J. Chem. Phys. 44, 1039 (1966); T. T. Tsong, Rep.

Prog. Phys. 51, 759 (1988).

10. R. J. Hamers and D. G. Cahill, J. Vac. Sci. Technol. B 9, 514 (1991).

11. D. M. Eigler, C. P. Lutz, and W. E. Rudge, Nature 352, 600 (1991).

12. G. M. McClelland, USA patent 5151594; H. Heinzelmann, G.M. McClelland, and

F. Watanabe, proceedings of the NATO ARW on Manipulation of Atoms under High

Fields and Temperatures, NATO ASI Proceedings Series E ed. V. T. Binh, N.

Garcia and K. Dransfeld. (Kluwer, Dordrecht, 1993), to be published.

13. D. P. Woodruff, and T. A. Delchar, Modern Techniques in Surface Science

(Cambridge U. Press, Cambridge, 1986).

14. M. M. El Gomati, M. Prutton, and R. Browning, J. Phys. E 18, 32 (1985).

7

15. Lasermetrics Co., Model 8612.

16. H.-W. Fink, IBM J. Res. Develop. 30, 460 (1986); H.-W. Fink, Physica Scripta

38, 260 (1988).

17. see also V. T. Binh, S. T. Purcell, N. Garcia, and J. Doglioni, Phys. Rev. Lett. 69,

2527 (1992), and references therein.

18. C. J. Todd and T. N. Rhodin, Surface Sci. 42, 109 (1974).

19. E. L. Wolf, Principles of Electron Tunneling Spectroscopy (Oxford, New York,

1985).

20. H. Bergeret, A. Septier, and M. Drechsler, Phys. Rev. B 31, 149 (1985).

21. R. Morin, Surface Sci. 162, 109 (1985).

22. for a collection of recent articles, see Proc. SPIE - Int. Soc. Opt. Eng. 1358 (1991).

23. E. W. Plummer and R. D. Young, Phys. Rev. B 1, 2088 (1970).

24. T. H. P. Chang, J. Vac. Sci. Technol. B 6, 1698 (1990); L. S. Hordon and R. F.

W. Pease, ibid., p. 1686.

8

Figures

Fig. 1. Femtosecond field emission camera. Not to scale; total length, 60 cm. a)

Time resolving mode, in which the electrons are focussed and swept across

the detector screen. b) Imaging mode, in which the emitted electrons are

projected onto the detector to spatially resolve their emission point on the

tip (field emission microscopy). For FIM imaging, the tip voltage is made

positive.

Fig. 2. Potential energy diagram for electron field emission from a sharp tip,

showing the effect of electric field and the adsorption of a Cs atom.

Fig. 3 Series of sweeping and imaging experiments documenting the jumping of a

single Cs atom between two adsorption sites on a tungsten tip. The num-

bered plots present the time-resolved field emission intensity for four of

eight consecutive sweeps. The time axis is referenced to the tip voltage

pulse, and a smoothed sweep 6 is superimposed on all plots. The field

emission images to the right are recorded after each sweep. The lower right

diagram presents the FIM-determined position of the three W atommr, spaced

by 4.5 A, of the apex layer of the tip.

Fig. 4 Simultaneous temporal and one dimensional spatial resolution of field

emission, showing the jumping of a single Cs atom. The four plots, recorded

simultaneously, present field emission from four adjacent regions of the tip.

Consecutive data points, separated by 2 ps, are connected by a straight line.

The schematic indicates a possible motion of the adsorbed Cs atom consist-

ent with the field emission.

9

(a) Time Resolve

DeflectorLens .- Sweep Detector

Pulse

-1 kV -7--7 kV

(b) Image H *e

-500 V

Fig. I. Femtosecond field emission camera Not to scale; total length, 60 cm. a) Time

resolving mode, in which the electrons are focussed and swept across the detector

screen. b) Imaging mode, in which the emitted electrons are projected onto the

detector to spatially resolve their emission point on the tip (field emission

microscopy). For FIM imaging, the tip voltage is made positive. 10

al

LE Field Of f

4.5eV! \' Field On

Fermi

[]= Field nCC

Adsorbed

0 Cs 10ADistance from Surface

Ftg. 2. Potential energy diagram for field emission from a sharp tip. showing the effect of

electric field and the adsorption of a Cs atomn.

01

SL. m•m m

II tI

Cn

"4--

0)

Cn"A 7

0 123 45Time (ns) FIM

Fig. 3 Series of sweeping avid imaging experiments documenting the lumping of a

single Cs atom between two adsorption sites on a tungsten tip. The num-bered plots present the time-resolved field emission intensity for four of eightconsecutive swe ,s The time axis is referenced to the tip voltage pulse, anda smoothed sep6is superimposed on all plots. The field emission imagesto the right are recorded after each sweep. The lower fight diagram presentsthe FIM-determined position of the three W atoms, spaced by 4.5 A, of theapex layer of the tip. 12

CS

-J-

~0

EI U

0 100 200 300 400 500

Time (ps)

Fig. 4 Simultaneous temporal and one dimensional spatial resolution of field

emission, showing the jumping of a single Cs atom. The four plots, recorded

simultaneously, present field emission from four adjacent regions of the tip.

Consecutive data points, separated by 2 ps, are connected by a straight line.

The schematic indicates a possible motion of the adsorbed Cs atom consist-

ent with the field emission. 13

Documents

REPORT DOCUMENTATION PAGE OMB No. · To measure the tribological properties of surfaces under well defined conditions, a new ... Friction Between a Diamond Tip and Polycrystalline