Encyclopedia of Smart Materials (Vols 1 and 2) - M. Schwartz (2002) WW Part 14 docx

It can thus be seen that when the structure is large and/or the frequency of interest high or wave-length relative to the structure short, many smart skin elements are required, implying

800Frequency (Hz)

PlatePassive controlActive/Passive control

4005

101520253035404550

600 Figure 4 Radiated power for a broadband1I1O case using multiple smart foam modules

operating in phase.

reference signal for the LMS control approach was taken

from the internal signal generator used to drive the

distur-bance (termed internal reference)

Figure 4 presents the radiated power with and without

control, when all the smart skin cells are wired together

in phase as a single channel of control The error signal is

provided by a single microphone located close to the smart

foam surface and at the plate-foam center Also shown is

the passive effect of the smart skin when it is located on

the plate but not activated It is apparent that the passive

effect of the skin is good at high frequencies above 1000 Hz

but is limited to resonant frequencies of the base plate

be-low this value Turning on the active control provides

rea-sonable attenuation at low frequencies, though there are

some frequency ranges where the control is negligible, for

instance near 900Hz In this case the smart skin transfer

function is reduced in level We now extend the controller

so that the six smart skin modules can be controlled

in-dependently with a six by six LMS control arrangement

Figure 5 presents these new results for the low frequency

range For the results of Figure 5 three different reference

signals control configuration are also studied: one using

300

−20

−10

010203040

Figure 5 Attenuated SPL for broadband 6I6O

case using multiple-independent smart foam modules.

an internal reference, one using an external referencesignal taken from an accelerometer located on the plate(representing a more realistic arrangement), and an ex-ternal reference signal with feedback (FB) from theactive component of the smart skin removed (1) It is appar-ent that much improved performance is achieved over theSISO case of Fig 4, particularly for the internal referencecase, due to the multicell active skin being able to matchthe complex radiation impedance load near 900 Hz, for ex-ample (3) In this case the smart skin transfer function isalso modified in a distributed manner

Using an external reference signal also provides able attenuation; however, it is reduced from the internalcase implying that the system is acausal (1) Some of thelost performance is recovered when feedback removal isemployed, indicating that the smart foam vibration hassome input to the plate system

reason-Recently the smart foam skin has been used to strate control of interior noise in aircraft (4) Figure 6shows a smart foam skin covering four panels in thecrown section of the fuselage of a Cessna business jet Theapplication is focused toward reducing cockpit noise due

demon-Error microphones

M3

C3Top euselage ribs Microphone traverse

C1

M4

Figure 6 Cessna crown panel control arrangement.

to exterior flow separation over the crown of the aircraft

Error microphones were located as shown at the ear

lo-cations of the crew and a microphone traverse was used

to measure the sound pressure levels in a plane at the

crew head height The flow noise disturbance was

simu-lated by an exterior speaker located just over the crown

of the aircraft and driven by band-limited white noise A

four by four feedforward LMS control approach was

im-plemented using a realistic reference signal taken from an

interior mounted accelerometer located on the fuselage at

the aircraft crown (i.e., just under the excitation location)

Figure 7 presents tabular results of the attenuation

achieved at the error sensors (near the crews’ ears) with

an excitation band of 500 to 900 Hz The reference speaker

refers to the use of a reference signal from the signal

driv-ing the disturbance The reference accelerometer refers to

the use of a fuselage-mounted accelerometer as a

refer-ence sensor and in this case attenuation of the order of 2

to 4 dB are achieved The global attenuation measured

us-ing the microphone traverse was found to be 2.5 dB with

the active skin turned on However, the active skin also

provides a passive attenuation of 4 dB when it is installed

over the bare fuselage panels and not turned on Thus the

total global attenuation of the smart foam active skin is

around 6.5 dB, a significant difference It also apparent

from Fig 7 that one of the main limitations on achievable

attenuation is the causality of the controller when using an

accelerometer as a reference signal When the reference

signal is taken from the speaker drive signal the control

path delay is less, and the performance increases markedly

The results do, however, demonstrate the potential of the

smart foam skin in reducing structurally radiated sound

in a realistic application

PIEZOELECTRIC DOUBLE AMPLIFIER SMART SKIN

Piezoelectric transducers tend to be high-force,

low-displacement devices (1) In contrast, active noise control

applications in air require high-displacement actuators,

particularly at very low frequencies Thus much of the

work in developing piezoelectric based actuators for

ac-tive noise control applications has been in designing

de-vices that amplify their displacement This amplification

Error mic 1

0510152025

Error mic 2Reference speakerReference accelerometer with feedback removalReference accelerometer without feedback removal

Error mic 3 Error mic 4

Figure 7 Averaged attenuation microphones for band-limited

500 to 900 Hz excitation.

is usually based on a geometric lever-type principle, andthus results in lower output force More explicitly, the ac-tuators are designed to have the correct source impedancerelative to their load In our application, the load is air with

a relatively low impedance, thus the device needs to have

a low source impedance for maximum power output.Figure 8 shows a schematic diagram of a piezoelectricdouble amplifier actuator, which is the basis of the secondactive skin concept (5) The legs of the element consist

of piezoelectric bimorphs or unimorphs In this case, thepiezoelectric transducers are manufactured from the ce-ramic material PZT (1) These devices are amplifiers inthat due to their asymmetry, small in-plane motions areamplified to larger transverse tip motions at the top of thelegs The tops of the legs are connected to a triangular orcurved stiff, lightweight diaphragm as shown Thus as thelegs move in, the diaphragm is squeezed upward Sincethe diapraghm axis is transverse to the tip motion, verysmall tip motions cause very large diapraghm motions (i.e.,amplify it) in a vertical direction Thus the complete struc-ture comprises a double amplifier actuator and gives am-plification ratios of diaphragm to piezoelectric elementin-plane deflection of the order of 20 : 1 The wholeconfiguration can be built in heights typically ranging from

1.3

234

PZT-Brass-PZTBimorph leg

Speaker paperdiaphragm

0.56

Figure 8 The active skin element (domains in mm)

Vibrating plate surface

PZT Bimorphs

Active-Skindiaphragm

Figure 9 Smart skin constructed from piezoelectric

double-amplifier elements.

3 to 6 cm, leading to a fairly compact device In

construct-ing an active skin of such devices, a number of them are

positioned to completely cover the surface of a structure as

shown in Fig 9 The devices can be either located directly

on the structure as shown or positioned just above it with a

small air gap In addition, the devices can be wired together

as one channel of control or independently controlled,

de-pending on the complexity of the base structural response

Figure 10 shows an actual device designed and

constructed by the Materials Research Laboratory at

Pennsylvania State University The device is 50× 60 mms,

34 mms high, and was found to have a maximum cover

displacement of 300µm at 100 Hz Figure 11 shows six

of the devices arranged to completely cover the surface of

a 170× 150 mm aluminum plate of 1.5 mm thickness In

this test arrangement, the active skin cells are located on a

perforated aluminum sheet which is located 5 mm from the

surface of the radiating plate Thus the active skin has a

small air gap between its bottom surface and the radiating

surface of the structure (5) Small accelerometers located

on each active cell diaphragmn are also apparent in Fig 11

These accelerometers are used to provide time domain

estimates of the radiated pressure in the far-field from the

Figure 10 A single active-skin cell.

Figure 11 The active-skin in a top-mounted SAS configuration.

measured surface vibration data, termed structural tic sensing (SAS) and described in (6) Such approachesallow integration of the sensors into the smart skin itself.The test plate and the active cells were mounted in arigid baffle located in the anechoic chamber at VAL Anoise disturbance to the plate was provided by a smallshaker attached to the back of the plate and drivenwith band-limited random noise The radiated sound fromthe plate-skin structure was measured using an array of

acous-16 microphones located on a hemispherical tube structure

as described above and a microphone traverse that couldmeasure the sound directivity in the horizontal midplane

of the plate The total radiated power from the plate could

be calculated from the 16 pressure levels measured by themicrophone hemispherical array (5)

Figure 12 depicts a schematic of the experimental rigand the control arrangement The control approach usedwas the Filtered -x LMS algorithm (1) implemented on

a TMSC40 DSP The shaker was driven with band ited noise of 175 to 600 Hz The Filtered -x algorithm was

lim-Amplifier

Shaker

Aluminumpanel

Baffle

Microphones

AccelerometersActive-skin

C40 DSPFiltered-x LMScontroller

C30 DSP SASFilter controller

Figure 12 The active-skin experimental setup.

−90°

0°θ

90°

Sound pressure level (dB)

Before controlAfter control

Figure 13 Total in-plane acoustic directivity (SPL), top-mounted

accelerometer configuration with microphone error sensing.

executed with a 2000 Hz sample rate, and 175 and 96

tap FIR filters were used for the control and system

iden-tification paths, respectively Since six independent cells

were located on the structure to comprise the active skin,

a six by six controller was implemented (5) Two tests were

performed using different error sensors In the first test, six

microphones evenly distributed over the microphone array

were used as conventional pressure error sensors located

in the radiated far field In the second test, the diaphragm

accelerometer signals were used in the structural acoustic

sensing approach, described in (6), to estimate the

pres-sures at the same locations as the previous error

micro-phones These estimates were then used as error signals

for the LMS algorithm

Figure 13 presents experimental results of the

directiv-ity of the total radiated sound power measured using the

far-field microphone traverse before and after the control

using the active skin elements It is apparent that the

active skin provides global sound pressure level

atten-uation of the order of 10 dB, which is impressive since

the excitation band encompasses multiple modes of

vi-bration of the radiating plate (5) Figure 14 shows the

corresponding radiated power versus frequency Good

con-trol is seen over the complete bandwidth of 170 to 600 Hz

except around 350 and 530 Hz, where anti-resonances

occur in the plate-active skin system The overall sound

power reduction for the results of Fig 14 is 10.9 dB

Fur-ther experiments were conducted using the accelerometers

in the SAS approach, and the results are presented in

Fig 15 Good attenuation is evident across the frequency

band, except near the system anti-resonance The

over-all reduction is now 9.5 dB, which is still impressive

Thus the results demonstrate that it is possible to utilize

an active skin that can provide significant attenuation of

sound radiated from a structure vibrating in complex

re-sponse shapes The successful use of the accelerometers

is significant in that it shows that an active skin with

completely integrated actuators and sensors can be

con-structed to provide very significant broadband

attenua-tion of sound radiated from structures under broadband

excitation (5)

200303540455055606570

250 300

Before controlAfter control

350 400Frequency (Hz)

450 500 550 600

Figure 14 Radiated sound power spectra, top-mounted

accelero-meter configuration with microphone error sensing.

SMART SKINS FOR SOUND REFELECTION CONTROL

It should be noted that the above mentioned smart skinapproaches could also be used to absorb sound imping-ing on structures by coating the structure with the smartskin However, in this application, a modified sensing ap-proach is needed in which the reflected or scattered wavecomponents are independently (than the total pressurefield) sensed and minimized by the controller Fuller et al.(7) discuss such approaches using the smart foam notedabove, and a combination of two microphones located nearthe smart foam surface are used to separate out the re-flected wave information from the total pressure field (7).Figure 16 shows a schematic of the experimental testing

in a plane wave acoustic standing wave tube The noise isgenerated by a speaker at the right end of the tube and im-pinges on the smart foam The two microphones are used

to separate out the reflected and incident wave responsesfrom the total pressure field The reflected wave signal isused as error information to the LMS controller The con-troller thus provides a control signal to the smart foam tominimize the reflected signal

Figure 17 presents the measured intensity of the dent and reflected wave intensities versus frequency withthe control off and on With the control off, the incident andreflected intensities are almost equal at low frequencies

inci-20030354045505560657075

Figure 15 Radiated sound power spectra, top-mounted

ac-celerometer configuration with SAS error sensing.

Reflectedwave

Disturbanceinput signal

ComputerwithLMScontroller

Control signal

Signal generator signal

error signal

AcousticsourceLP

filter

Figure 16 Smart skin reflection

con-trol experimental arrangement.

below 300 Hz, implying that the smart foam is acting like

a rigid surface with very little sound absorption Above

300Hz, as is expected, the foam provides increasing

pas-sive sound absorption, and the reflected intensity is less

than the incident When the active control is turned on,

the incident intensity remains the same, but the smart skin

leads to a significant reduction in reflected sound energy

below 300 Hz This reduction in reflected sound due to the

smart skin is apparent over the complete frequency range

of Fig 17 The two microphones can also be used to measure

the acoustic impedance of the smart foam When the

con-trol is turned on at low frequencies, the normal acoustic

impedance of the foam falls from very large values to be

almost identical to the characteristic impedance of the air

Thus the active element in conjunction with the controller

150 200 300 400 500

Frequency (Hz)

Incidentwave

Reflectedwave

Reflected wave undercontrol

600 700 800 900 1000

Figure 17 Reflection control using a smart

skin.

of the smart foam have modified the smart foam dynamics

so that it looks like a perfectly sound absorbing surface

ADVANCED CONTROL APPROACHES FOR SMART SKINS

The conventional control approaches used with a smartskin can be divided into two types; multi-channel feedfor-ward, which is generally used when access to a coherentreference signal is available, and multiple input-multipleoutput state space feedback methods, which are often usedwhen such a convenient reference signal is not available.These approaches are summarized in (1) As discussedabove, the smart skin approach relies on covering a ma-jor part of the structure with independently controllable

Centralizedprocessor

Local control rules

Multiple independent control signals

Figure 18 Biological control approach.

elements It can thus be seen that when the structure

is large and/or the frequency of interest high (or

wave-length relative to the structure short), many smart skin

elements are required, implying a control approach with

a very high number of control channels In this case, the

conventional approaches are likely to be unsuitable due

mainly to computational limits on the control processor

and stability/performance aspects There are two different

approaches suitable for high sensor/actuator count

sys-tems (8, 9) Both approaches are hierarchical and are

in-spired by biological systems of muscle control They are

thus termed BIO controllers

In the first approach, the smart skin elements are

ar-ranged into groups of “slave” actuators under the

con-trol of a “master” actuator A schematic of the concon-troller

is shown in Fig 18 A top-level centralized controller is

used to send signals to the master actuators Simple local

control laws are used to modify and apply the same

sig-nal to nearby slave actuators For example a very simple

local law discussed in (8) would be take the same control

signal, apply it to an in-phase, out-of-phase, or off-phase

H2

G

H1FIR

filterReference

Figure 19 BIO controller with phase local control law.

Radiated sound

Structure

ActuatorSensor

Controllaw

Localcontroller

Local controllercommandsignal

Top levelcontrollerAveraged

performancemetric

Figure 20 Schematic of a BIO controller arrangement.

slave actuator via simple analog switches and keep thesetting that gives the lowest cost function value Figure 19shows a block diagram realization of such a control systemfor a feedforward approach The process then continues

to the next slave actuator, and so on, in a predeterminedpattern For the system of Fig 19 the top-level controllercould be digital, while the local control changes occur viasimple analog-switching circuits The approach in effecttakes many independent actuators and connects them to-gether via the local controller to create a suboptimal dis-tributed actuator driven by one (or few) channel of controlfrom the top-level centralized controller The net result ofsuch approaches is a large reduction in control channels

to the top-level digital controller, and thus the tional overhead requirements are vastly reduced The BIOapproach in effect takes advantage of some limited knowl-edge of the dynamics of the distributed system to be con-trolled in order to reduce the extensive number crunchingrequired in fully coupled optimal approaches

computa-In the second approach, local analog feedback loops areclosed around individual smart skin elements and associ-ated sensors as shown in Fig 20 The analog local feedbackloops have programmable feedback gains that are adapted

by a higher-level digital controller in order to minimize aglobal cost function (obtained from an array of sensors)such as radiated sound power from the structure covered

by the smart skin (9) Such approaches have been used

to control sound radiation from very large structures Aswith all feedback approaches, stability is an important is-sue Thus work has also been performed to increase thestability margins via using directional feedback sensors topartially de-couple each local feedback loops In addition,specialized distributed actuators are used that rolloff inlevel in the higher-frequency regions where the local openloop transfer function becomes non-minimum phase

CONCLUSION

The results presented have demonstrated the high tial for the implementations of a smart skin approach forreducing sound radiated from vibrating structures whenthe radiating structure is massive, stiff (i.e., low mobility),

poten-or the source vibration pattern is complex The smart skin

has also demonstrated the possibility of combining active

and passive control approaches in order to increase the

control bandwidth and the efficiency of the active portion

A configuration has been demonstrated that further shows

that the error sensors can be integrated directly into the

skin and still result in a far-field sound reduction

BIBLIOGRAPHY

1 C.R Fuller, S.J Elliott, and P.A Nelson, Active Control of

Vibration Academic Press, San Diego, CA, 1996.

2 C.A Gentry, C Guigou, and C.R Fuller JASA 101(4): 1771–

1778 (1997).

3 C.A Gentry, C Guigou, and C.R Fuller Submitted to JASA,

1999.

4 C Guigou and C.R Fuller Proc SPIE Smart Structures and

Materials Conf., San Diego, CA, SPIE Vol 3044, pp 68–78,

1997.

5 B.D Johnson, M.S Thesis VPI& SU Blacksburg, VA, 1997.

6 J.P Maillard and C.R Fuller JASA, 98(5): 2613–2621

(1995).

7 C.R Fuller, M.J Bronzel, C.A Gentry, and D.E Whittington

Proc NOISE-CON 94, pp 429–436, 1994.

8 C.R Fuller and J.P Carneal JASA, 93(6): 3511–3513 (1993).

9 M Kidner and C.R Fuller, Proc 8th Conf on Nonlinear

Vibra-tions, Stability and Dynamics of Structures Blacksburg, VA,

July 2000.

SPIN-CROSSOVER MATERIALS

University of Utah, Chemistry

Salt Lake City, UT

Smart materials respond to their environment as

illus-trated by photochromic eyeglasses, that darken upon

ex-posure to ultraviolet light to attenuate additional

ultra-violet light Hence, materials that have fast reversible

responses to environmental stimuli are sought as

compo-nents of smart systems Similar to photochromic

materi-als, thermochromic materials reversibly respond to heat

and exhibit substantial color changes upon small changes

in temperature Spin-crossover materials (1) are a class

of thermochromic materials that possess fast, reversible

color changes amenable to display and memory devices (2)

These color changes can also be induced by light

(photo-chromic) or pressure (piezo(photo-chromic) as well as heat Due

to the nature of the mechanism of their thermo-, photo-, or

piezochromic responses (i.e., redistribution of the electron

density at a metal ion site within the molecule), they are

extremely fast and reversible As a consequence of the (1)

fast color change, (2) strong contrast between colors, and (3)

the intermolecular interactions within the solid, the

differ-ing colors can be maintained for a long period of time, and

(4) due to the lack of moving parts (i.e., no bond breaking

or forming), these materials are completely recyclable and

amenable to fast, low power-consuming, high-data-density

display (2,3) and storage devices and “smart” materials and

systems of the future

Low spin

1A1g

Spectrochemical seriesincreasing ligand field, ∆

be induced by light or pressure.

Thermochromism results from transition-metal plexes, such as Fe(II), which can be thermally stimulated

com-to change from a colored low-spin electronic state com-to a quently colorless high-spin state (1a) (Fig 1) The high-spin 5T2g ground state for Fe(II) has a t2g–egsplitting,

fre-of <11,000 cm−1, and the low-spin1A1g excited state forFe(II) has a of >21,000 cm−1. ∼16,000 cm−1for Fe(II)surrounded with six unsaturated nitrogen-bound ligands,FeN6, can be induced to switch between the high-spinand low-spin states Upon switching between the high-and low-spin states on cooling, FeN6has a significant de-crease in Fe–N distances by 19± 5 pm, and an increase

in the magnetic susceptibility χ due to a change of four

in the number of unpaired electrons From a namic perspective, the enthalpy H is 10 ± 6 kJ/mol,

thermody-and the entropyS is 52 ± 13 J/Krmol; hence, the tion is entropy-driven (1a) Additionally and importantly,the color changes from deeply colored red/purple to color-less upon switching to the high-spin state (see later) Con-comitantly, the unit cell typically changes significantly.The color change of the spin state switch is similar tothat of liquid crystal displays (LCD) prevalent in digitalwatches; however, as a consequence of the mechanism, thethermochromic metal complexes change colors much fasterwithout degradation upon cycling with respect to LCDs(3) Due to the change from low to high spin, this class ofmaterials is referred to as spin-crossover materials Inaddition to the technologically important color changes,spin-crossover materials also exhibit a small, but mea-surable change in magnetic susceptibility This sharptransition in the change in the magnetic properties

transi-is illustrated by the temperature dependence of the

magnetic susceptibility–temperature product for

Fe(o-phenanthroline)2(NCS)2, which undergoes a first-orderphase transition from a low- to a high-spin state at−97◦C(4), (Fig 2)

Materials that can be easily and reversibly stimulated

to change colors for an innumerable number of cycles havebeen exploited for display devices Liquid crystal displays(LCD) found in digital watches, are a common example(3) Materials that have greater switching speeds, sharpercontrast, and enhanced stability enabling more dutycycles may lead to improved display and memory devices

in the future Spin-crossover materials can exhibit sharpcolor changes from small changes in temperature (i.e., theyare dramatically thermochromic) As a consequence of thethermochromic mechanism (redistribution of the electrondensity within the molecule without either bond breaking

2

1

150 200 250Temperature, T, K

Figure 2 Temperature dependence of the magnetic

suscepti-bility–temperature product for Fe(o-phenanthroline)2 (NCS) 2 ,

which undergoes a first-order phase transition from a low- to a

high-spin state at 176 K ( −97 ◦C) (4a).

or forming), they are extremely fast and recyclable and

hence are candidates for high-data-density display and

storage devices of the future

For display/memory devices, it is necessary that the

transition temperature (T) is near room temperature,

∼22◦C This is, however, insufficient because the ambient

temperature fluctuates and hence the transition needs to

be effected over a broad temperature range, 17± 27◦C To

achieve this, the system must exhibit history-dependent

behavior (hysteresis) such that the transition temperature

for color change upon increasing temperature (T↑)

ex-ceeds the transition temperature for color change upon

de-creasing temperature (T↓) ideally by at least 50◦C, that is,

T ↑ − T↓ > 50◦C Molecules cannot exhibit hysteretic

ef-fects, but in a solid or film, interactions between molecules

can lead to hysteretic effects Hysteresis has been reported

for FeL2(NCS)2(L= (N2(CH)2N–)2], where T↑ = −128.7◦C

and T↓ = −149.5◦C (Fig 3) (5) Thus, although the

transi-tion and T↑ − T↓ temperatures are too low to be practical,

the necessary phenomena have been demonstrated, and

new systems that exhibit higher temperatures are needed

Using a mixture of triazole, HN(CH)2N2 (trz), and

aminotriazole, H2NN(CH)2N2 (H2Ntrz) ligands

coordi-nated with Fe(II), a polymer of [Fe(trz)3 −3x(H2Ntrz)3x]

23

Figure 3 Temperature dependence of the magnetic behavior

of FeL2(NCS)2 showing the low-moment (purple) behavior

be-low−149.5◦C (T↑) and high-moment (colorless) behavior above

Figure 4 Temperature dependence of the magnetic behavior

of Fe(trz) 2.85(H2 Ntrz) 0.15](ClO4 ) 2rnH2O showing the low-moment(purple) (Figs 5 and 6) behavior below 39 ◦C (T↑) and high-moment (colorless) behavior above 13 ◦C (T↓) (6).

(ClO4)2 r nH2O composition has been isolated, which for

x= 0.05 exhibits T↑ = 39◦C and T↓ = 13◦C (2,6,7 ) (Fig 4).These values bracket room temperature and demonstratethe feasibility of room temperature applications In ad-dition to the change in magnetic behavior, the color con-comitantly as with hysteresis occurs (Fig 5), from pur-ple to colorless at 21◦C (Fig 6) Solid solutions of triazoleand aminotriazole can be blended to lead to a systematic

change in the transition temperatures: T↑ = 296 − 160x

and T↓ = 313 − 180x in units of Kelvin.

Smart materials for future applications need to respond

to environmental stimuli, and spin-crossover materials (1)are a moderately large class of materials that respond toheat, light, and/or pressure This summary focuses on theuse of heat to change the electronic structure of a material,which in turn leads to substantial and reversible color,magnetic, and structural changes Most of the materialsdiscussed in this context are inorganic coordination com-plexes demonstrating that (1) reversible first-order transi-tions occur, (2) such materials exhibit the technologicallyimportant property of hysteresis, and (3) both the transi-tions and hysteresis can occur at room temperature

320 340 360 380 400 420Temperature, T, K

Figure 5 Temperature dependence of the optical density of

Fe(trz) 2.85(H 2 Ntrz) 0.15](ClO 4 ) 2rnH2O at 520 nm showing sis (purple) (1a).

hystere-Figure 6 Dramatic color change for the dark purple low-spin

state of [Fe(trz) 2.85(H 2 Ntrz) 0.15](ClO 4 ) 2rnH2O below 21◦C to the

colorless high-spin state at 21 ◦C (6).

ACKNOWLEDGMENTS

The author acknowledges continued partial support

by the Department of Energy Division of Materials

Science (Grant Nos FG02-86ER45271.A000,

DE-FG03-93ER45504, and DEFG0296ER12198) and helpful

discussions with Prof O Kahn

BIBLIOGRAPHY

1 (a) P G ¨utlich, A Hauser, and H Spierling, Angew Chem 33:

2024 (1994) (b) E Konig, G Ritter, and S.K Kulshreshtha,

(1981).

2 (a) O Kahn, E Codjovi, Y Garia, P.J van Koningsbruggen,

R Lapouyade, and L Sommier, ACS Symp Ser. 644: 298

(1996) (b) O Kahn and C.J Martinez, Science 279: 44 (1998)

O Kahn, J Kr¨ober, and C Jay, Adv Mater 4: 718 (1992).

3 C Esher and R Wingen, Adv Mater 4: 189 (1992) R Bissell,

N Boden, Chem Brit 31: 38 (1995).

4 (a) B Gallois, J-A Real, C Hauw, and J Zarembowitch, Inorg.

Chem 29: 1152 (1990) (b) M Sorai and S Seki, J Phys Chem Sol 35: 555 (1974).

5 W Vreugdenhil, J.H van Dieman, R.A.G de Graaff, J.G Haasnoot, J Reedijk, A.M van der Kraan, O Kahn, and

THERMORESPONSIVE INORGANIC MATERIALS

JOHNS.O EVANS

University of Durham

Durham, UK

INTRODUCTION

The encyclopedia of Chemical Technology (1) defines a

smart material as one that “responds to its environment

in a timely manner” and “receives, transmits, or processes

a stimulus and responds by producing a useful effect. .”

One of the most common everyday stimuli is, of course,

tem-perature The effect of temperature on the vast majority of

materials is well known—as materials are heated, they

ini-tially expand in volume before eventually melting,

sublim-ing, or decomposing Thermal expansion is often viewed as

a deleterious property Engineers have to build expansion

gaps into structures such as railway lines and bridges; the

design of components in a car engine must be somewhat

compromised so that they function both during start-up on

a cold morning and when the engine is hot; historically,

clock pendulums had to be carefully engineered to prevent

slow running on warm days Thermal expansion can,

how-ever, also be put to good use It has long been known that

the walls of bowed buildings can be pulled back into shape

by a cooling iron bar; steel tires can be shrink-fitted onto

the wheels of railway carriages Although these examples

represent a technologically useful response to the stimulus

of increased (or decreased) temperature, they do not quite

fall into the category of “smart.” However, the very simple

concept of coupling together two materials, one that has

a large and one a smaller coefficient of thermal expansion

to produce a bimetallic strip can certainly be considered to

create a smart composite body Here, the strains induced

by the higher expansion of one material cause the strip

to bend as temperature is increased, leading to a simple

temperature-sensing/responsive device

The vast majority of materials known and used in

tech-nological applications have a positive coefficient of thermal

expansion; the reasons for this behavior are discussed

later Certain materials, however, display the opposite

property and contract in volume when heated These

mate-rials thus have a negative coefficient of thermal expansion,

description as “negative thermal expansion” (NTE)

mate-rials The properties of these materials and the origin of

these effects form the main topic of this article

Materials that have this unusual thermoresponse have

a number of important technological applications

Applica-tions related to specific materials are discussed later; more

general comments are appropriate here The first major

area of application is in producing composite bodies By

mixing normal materials with a negative thermal

expan-sion phase, one can achieve a composite that has a precisely

controlled positive, negative, or even zero coefficient of

expansion Second, certain of the materials discussedlater can be chemically doped to control their expansionproperties Thus, one can envisage a single material thatcould be adjusted to have zero overall expansion Suchmaterials are of use where repeated thermal shock mightlead to mechanical failure (an everyday example is oven-to-table cookware) or in optical devices such as high precisionmirrors where any thermal expansion might distort opti-cal properties Materials that have strong intrinsic thermalcontraction are most likely to be used as compensators forthe positive expansion of other phases For example, there

is obvious interest in the electronics industry in producingcircuit boards and heat sinks whose expansion propertiesmatch those of silicon; the dental industry would like ce-ramic fillings whose properties match those of teeth; the ex-pansion properties of materials used in optical componentssuch as refractive index gratings and optic fibers must beprecisely controlled

The observation that normal materials expand whenheated can be explained at the most simple level by refer-ring to Fig 1 which shows a potential energy well for a typ-ical diatomic molecule As molecules are warmed, one ex-cites higher and higher energy vibrational levels Becauseinteratomic potentials are typically asymmetrical (bondsare more extensible than compressible), this leads to anincrease in bond length with temperature A more formalexplanation for the thermal expansion of solids is given

by the Gruneisen relationship that relates the thermal pansion of a material [α ν = (1/ V) ( ∂V / dT)P to its volume (V), specific heat at constant volume (CV) and isothermal com-

anharmonicity of a typical potential in a solid

The frequencies of most modes in a normal crystaldecrease as the volume increases, and the Gruneisenparameter of normal materials is typically positive in therange of 1 to 3 and only weakly temperature-dependent.Thus, the thermal expansion of normal materials may beexpected to have a temperature dependence similar to theirspecific heat capacity Therefore, one would expect thermalexpansion to be zero at absolute zero, to increase rapidly astemperature is increased, and to approach a constant value

at temperatures above the Debye temperature of the rial This simple behavior is at least approximated in manymaterials In “negative thermal expansion” materials con-sidered in the remainder of this article, there will always be

mate-an underlying expmate-ansive component caused by vibrationalmodes that tend to increase bond distances In certaincircumstances, however, these modes may be dominated

by other more exotic effects It is well known that strongchemical bonds (for example, those between highly chargedelements and oxygen: Si–O, W–O, etc.) expand significantlyless than weaker bonds (e.g., Na–O, K–O, etc.) (2) Thus

1040

Interatomic distance

Figure 1 A typical interatomic potential well.

materials in which the unusual effects dominate normal

expansion will often contain strongly bonded atoms

There are two simple expressions commonly found in

the literature to describe the thermal expansion of

mate-rials The linear coefficient of thermal expansion may be

defined by the differential form α l = 1/l(dl/dT) or by

the average coefficient of thermal expansionα l = (l − l0)/

l0(T − T0) Because the expansion coefficient generally

changes as a function of temperature, it is important to

quote the temperature range for any value ofα l For

aniso-tropic materials,α lhas been defined as 1/3α V Values ofα l

for typical materials are quoted in Table 1 and are plotted in

terms of % length extension for selected examples in Fig 2

ORIGINS OF NEGATIVE THERMAL EXPANSION

Phonons

The previous section described how thermal expansion can

be related to the specific heat capacity of a material and, in

turn to the population of vibrational modes In general, this

Table 1 Thermal Expansion Coefficients

0.501.001.502.00

500 600Temperature (K)

700 800 900 1000

ZrW2O8

Al2O3SiMgOQuartz

Figure 2 Thermal expansion curves for selected materials.

leads to positive thermal expansion However, it was firstpointed out in the 1950s that certain vibrational modes canlead to the opposite effect and negative thermal expansion(14–16) The simplest example of such an effect can bereadily understood by referring to Fig 3 If one considers

an metal–oxygen–metal linkage in a structure, a dinal vibration (e.g., along an M–O–M bond) will lead to

longitu-an extension of the M–M distlongitu-ance During a trlongitu-ansversevibration, however, if the M–O distance remains essen-tially unchanged, the M–M distance decreases This can

be understood with respect to the Gruneisen constant forsuch a mode A simple analogy for a transverse mode in

a M–O–M linkage is a guitar string If one plucks a note(excites a transverse vibrational mode) on a guitar string,then stretches the string using the tuning forks, the notemoves to a higher frequency Therefore, the Gruneisen pa-rameter,γ = −d(nν)/d(nV) for such a mode is negative.

The contribution to thermal expansion for such a mode willalso be negative

Equation (1) is, of course, a simplification of the erties of a real material In a real structure, differentvibrational modes have different energies and are pop-ulated at different temperatures In reality, one mustconsider the actual population of modes and use theaverage Gruneisen parameter at a given temperaturewhere γ a ν =c i γ i /c i, and ci weights the contribution

prop-of a mode to the overall specific heat CV In general, verse modes have lower energy than longitudinal modesand are preferentially populated at low temperature Thus,

trans-Temperature Temperature

Figure 3 Schematic representation of the effect of different types

of vibrational modes on thermal expansion Shaded circles sent metal atoms and the open circle an oxygen atom Longitudi- nal vibrations tend to expand metal–metal distances; transverse vibrations can lead to contraction.

250 300 350 400

Figure 4 Thermal expansion curves of CuFeS2 , using data of

Knight (17) The solid line represents a fit to a simple two Debye

type model At low temperatures, the population of modes that

have negative Gruneisen parameters leads to negative thermal

expansion At higher temperatures, modes that have a positive

Gruneisen parameter dominate.

they may dominate the overall Gruneisen parameter at low

temperatures and lead to negative thermal expansion The

population of low-energy transverse modes is the origin of

negative thermal expansion at very low temperatures in

a number of materials Examples of this phenomenon are

shown in Figs 4 and 5 Figure 4 shows the cell volume of

CuFeS2derived from powder neutron diffraction data as a

function of temperature (17) Here, the thermal expansion

is essentially zero at low temperatures [in accordance with

expectations of Eq (1)], is negative at temperatures below

100 K, where modes that have negative Gruneisen

param-eters presumably dominate, & approaches a constant

posi-tive value at high temperatures Similar phenomena occur

in many materials, including compounds that have rock

salt, diamond, and zinc blende structures Barron and

co-workers have provided extensive data in this field (4)

In-teresting correlations can be made between the magnitude

of the negative thermal expansion and factors such as the

openness of the lattice and the rigidity (covalency) of the

chemical bonds

In the context of phonon-induced negative thermal

ex-pansion, the properties of ice are worth mentioning The

128.0128.5129.0129.5130.0130.5

Figure 5 Thermal expansion of ice using data of Rottger (5).

increase in volume of water when freezing is one of thebest known examples of an unusual thermoresponsiveeffect Both the normal hexagonal and cubic forms of icehave a density (0.92 g/cm3) that is significantly lower thanthat of water This leads to the costly winter phenomenon

of frozen pipes bursting then leaking during a thaw, and italso means that ice floats on water For this reason, pondsand lakes always freeze from the top down, giving rise to

a protective layer of ice on top of the water This process isessential for the survival of aquatic life This expansion ofwater can also be incorporated into simple, cheap devices

to monitor the temperature history of frozen food and othercommodities Water can be sealed in a thin glass vial, andsurrounded by paper impregnated by a water-soluble dye.When the water freezes, it expands and breaks the glassvial Subsequent warming of the vial causes the water tomelt, flow into the dye, and color the surrounding area.Thus, one can monitor whether the temperature of a com-modity has ever exceeded 0◦C during its storage/shipment.Less well known is the fact that ice shows negative ther-mal expansion in the solid state at temperatures belowabout 60 K Figure 5 shows the temperature dependence

of the cell volume of ice as determined by synchrotron ray diffraction (5) This cell volume contraction is againcaused by the population of low-energy (<50 cm−1) trans-verse vibrational modes at low temperatures (5,18)

X-Rigid Unit Modes

The previous section described how vibrational modes maycause negative thermal expansion at low temperatures incertain materials The population/importance of individualmodes are intimately related to a material’s structure Vi-brational modes can, however, give rise to NTE across farmore extensive temperature regions in certain categories

of materials

Figure 6 shows the structure of quartz (7) Its mal expansion properties are included in Fig 2 At low

ther-Figure 6 Structure of quartz viewed down the c axis showing

corner-sharing SiO 4 tetrahedra The ab plane of the unit cell is shown in bold lines.

temperatures, quartz has a positive coefficient of thermal

expansion; at high temperatures (>573◦C), the thermal

ex-pansion becomes low or even negative The ideas that led

to an explanation of this phenomenon can be traced to the

structural description of phase transitions in quartz given

by Megaw (19) The structure of the high temperature (β)

form of quartz can be described as a network of

corner-sharing SiO4 tetrahedra that form paired helical chains

parallel to the crystallographic c axis (a 64or 62screw axis)

The intertwined chains give rise to the hexagonal channels

(important in the structure ofβ-eucrypite described later)

through the structure that are visible in Fig 6 The

struc-ture of the low temperastruc-ture (α) form of quartz can be

de-rived from that ofβ quartz by a coupled rotation of these

tetrahedra through an angleθ about the <100> axes In

the more symmetricalβ-quartz, θ is 0◦; inα-quartz, θ is

16.3◦ at room temperature (20) This coupled static

rota-tion of essentially rigid polyhedral groups leads to a rapid

reduction in volume on cooling and a large positive

expan-sion coefficient on warming Similar dynamic rotations of

such groups at high temperature can explain the negative

thermal expansion ofβ-quartz.

Dove and co-workers built on the concepts of rigid

tetrahedral building blocks to produce models for

nega-tive thermal expansion in terms of such distortions (21)

Figure 7 shows a schematic representation of a simple

two-dimensional structure that can be considered a

logi-cal extension to the 1-D schematic of Fig 3 If this figure

represents a hypothetical 2-D structure containing metal

atoms strongly bound to a square coordination

environ-ment of oxygen atoms, it can be readily appreciated that

the structure has inherent flexibility Certain distortions

of the structure are possible in which (strong) M–O

dis-tances and (rigid) O–M–O bond angles remain unchanged

Such distortions are likely to be energetically facile and

thus are good candidates as low energy vibrational modes

of the material Dove and co-workers termed such modes

“rigid unit modes” (RUMs) and developed methodologies

for identifying them in real 3-D structures

Temperature

Temperature

Figure 7 Schematic representation of rigid unit mode vibrations

in a framework oxide structure as polyhedral and atomic views.

Metal atons are shaded: oxygen atoms are open circles.

It can also be seen that this type of mode is directlyanalogous to the negative Gruneisen parameter transversevibration of Fig 3 In fact, the area of the unit cell of Fig 7

is proportional to the tilt angleθ The thermal average of

this angle <θ> T increases with temperature so that the

overall cell area A(T) is given by

A(T) = Aocos2θ ≈ (1− < θ2> T) (2)

If one assumes simple harmonic motion and the principle

of equipartition of energy, A(T) can be related to the perature T, the moment of inertia I of the rigid body, and

tem-the vibrational frequencyν by

Concepts of RUMs can also be used to understanddisplacive phase transitions of the type displayed byquartz and many other framework materials It is in-teresting to note, and it is a general feature of manyframework materials that display negative thermal ex-pansion, that displacive phase transitions and NTE be-haviour are intimately linked Many framework materialspossess a large positive coefficient of thermal expan-sion as they are warmed This may be viewed as evi-dence that the rigid units are “unwinding” in the struc-ture Once materials reach their high symmetry (maxi-mum volume) structures, they display negative thermalexpansion

Phase Transitions

As suggested by Fig 1, chemical bonds usually expand

as temperature increases Consequently, the majority ofmaterials have positive expansion coefficients There are,however, unusual circumstances, often in the region ofstructural phase transitions, in which at least averagebond distances can actually decrease as temperature in-creases The ideas of bond valence which originate in thework of Pauling and were developed extensively by Brownand O’Keeffe (23,24) showed that the contribution of agiven bond to the valence sum of any atom depends ap-

proximately exponentially on the bond length v = exp(r0−

r/0.37), where r0is a constant for a given E-X combination

of elements and v is the valence contribution due to bond length r This observation can be thought of at the sim-

plest level as short bonds being proportionately strongerthan long bonds From the form of this relationship, onecan readily appreciate that if a metal coordination environ-ment changes from being distorted to being more regular

at a phase transition, the average bond length decreases

Figure 8 Cell parameters of PbTiO3 as it approaches its

tetragonal–cubic phase transition.

For example, the bond length r in an undistorted MO6

octahedron is always shorter than the average bond

dis-tance in a distorted octahedron that has, for example, three

short and three long bonds

One example of this effect will be discussed here

Figure 8 shows the temperature dependence of the cell

parameters of PbTiO3 as it approaches its

ferroelectric-paraelectric phase transition at 490◦C (25) PbTiO3

con-tains highly distorted Ti octahedra at room temperature,

Ti–O bond lengths are 1.766, 4× 1.979, and 2.390 ˚A (26).

When warmed, these octahedra become less distorted, and

the average Ti–O bond length decreases from 2.012 to 1.983

˚

A (this represents a bond length expansion coefficient of

α l= −3 × 10−5K−1) The overall coefficient of thermal

ex-pansion of PbTiO3from 30–400◦C isα l = −3.3 × 10−6K−1

(27) Presumably, the contraction of the Ti–O bond lengths

in this material dominates the normal expansion of other

bonds

Negative volume changes at phase transitions have also

been described in materials such as Co2P2O7 As described

later, such phases have been included as fillers in glass

bonding frits to help match thermal expansion coefficients

of materials to be joined Control of displacive phase

transi-tions is also important in the thermal expansion properties

of ZrV2 −xPxO7and A2(MO4)3families and will be discussed

later

Magnetic Transitions

Magnetostrictive phenomena in the region of magnetic

phase changes can also give rise to materials of low

mal expansion The “normal” phonon-driven positive

ther-mal expansion of materials that have a significant

mag-netoelastic coupling can be compensated for by a large

contraction driven by changes in the magnetic structure

This is the case for alloys such as “Invar”, Fe0.65Ni0.35(see

later), and transition metals such as Cr andα-Mn

Cor-rectly processed, Invar has been quoted as having a

ther-mal expansion coefficient of 0.02 × 10−6K−1, though across

a restricted temperature range (28,29) Materials such as

Lu2Fe17 and Y2Fe17 also exhibit negative thermal

expan-sion below approximately 400 K (30)

Linking Themes

Perhaps the best framework in which all of these ena may be described is to consider the change in the in-ternal pressure of a solid caused by heating This internalpressure, will in turn, gives rise to a volume change Thevolume coefficient of thermal expansion can be related to

phenom-the isophenom-thermal compressibility K and phenom-the internal pressure

Thus the sign of the coefficient of thermal expansion can

be related to the sign of

∂ S∂VT The entropy of normalmaterials (or the amount of structural disorder) decreases

as pressure increases and volume decreases; materialsthat have NTE, however, increase in disorder as volume isreduced

A further link between many of the materials of thenext section is their proximity to a structural phasetransition Many of the Invar phases described in the nextsection are close to a structural phase transition to a body-centered structure Cubic zirconium tungstate discussedlater is thermodynamically unstable at temperatures be-low 1000◦C with respect to more condensed binary phasesand undergoes a number of phase transitions as a function

of applied pressure and temperature These instabilitiesare relevant to its unusual properties Interestingly, metal-lic plutonium, which has negative thermal expansion in its

δ and η phases has no less than six allotropes (α, β, γ , δ,

η, ε) at atmospheric pressure; the NTE δ phase has an

atomic volume 26% larger than theα.

Thus, it is possible to identify several criteria that might

be expected to give rise to unusual thermoresponsive fects in solids:

ef-1 proximity of the material to a structural phasetransition

2 proximity of the material to a magnetic or electronictransition

3 the presence of a framework structure containingstrong metal anion bonds and two coordinate anions

4 the absence of interstitial cations in a frameworkstructure

5 framework structures in which volume-reducing placive phase transitions are frustrated

dis-Finally, it is interesting to note that it is possible tomake composites that exhibit negative thermal expansionwithout relying on any of these mechanisms Sigmund andTorquato showed that it is theoretically possible to pro-duce three-phase composites that exhibit isotropic nega-tive thermal expansion by engineering a topologically spe-cific combination of two materials whose thermal expan-sion is positive (one high, one low) and empty space (31)

In theory, such composites can be designed so that whenheated, the bimaterial interfaces of the composite bend intothe void space and the overall composite contracts—thematerial literally folds in on itself

MATERIALS THAT DISPLAY NEGATIVE

THERMAL EXPANSION

Invar

Probably the first technologically exploited example of an

“unusual” thermoresponsive effect occurred in 1897 when

Swiss physicist Charles Edwarde Guillaume working at

the International Bureau of Weights, Sevres, discovered

that an Fe/Ni alloy of approximate composition Fe0.65Ni0.35

showed almost zero thermal expansion across a broad

tem-perature range This material was christened “Invar,” and

because of its discovery, Guillaume was awarded the Nobel

prize for Physics in 1920 Invar is used in manufacturing

many high precision devices Perhaps the most widespread

current application of Invar is in television and computer

screens where it is used as a mask to prevent the electron

beam from unintentionally hitting phosphor spots of the

wrong color that would lead to blurred images The low

ex-pansion of Invar is essential in this application due to the

possibility of heating by the electron beam Historically,

Invar was used in clock pendulums where its low

expan-sion prevented losses in accuracy as temperature changes

Other areas of application include waveguide tubes,

sur-veying tapes, molds for shaping composites in aircraft

prod-ucts, and thermostat controls Invar is also used in both

tankers and tubing for liquid natural gas facilities In

typ-ical applications, Invar has an expansion less than

one-tenth that of steel, yet retains strength and is typically

cheaper than a material such as Ti Guillaume is also

cred-ited with the invention of “Elinvar” which has extremely

small changes in elasticity and hence, mechanical strength

as a function of temperature Thus, this material was used

in producing watch springs

It was realized very quickly that the unusual expansion

properties of Invar were intimately connected to its

mag-netic properties Invar itself is a disordered face-centered

cubic alloy that has a magnetic phase transition at around

500 K Invar has an extremely low coefficient of thermal

expansion below this temperature and a more normal one

above it

The connection between magnetism and thermal

ex-pansion led Weiss to propose the “2γ state” model to

ex-plain the Invar effect (32) There are two different magnetic

states in this model, separated by a low energy barrier: a

ferromagnetic ground state in which magnetic spins are

parallel and the volume is larger, and an antiferromagnetic

(antiparallel spins) state at a slightly higher energy, but

with a smaller volume As temperature is increased, the

higher energy state becomes increasingly populated, and

its smaller volume compensates for the usual expansion

effects caused by thermal vibrations Very recently, van

Schilfgaarde et al described ab initio calculations of the

volume (and hence temperature) dependence of the

mag-netic and thermodynamic properties of Invar (33) Their

work describes how noncollinear spin arrangements are

crucially important in the Invar system The magnetic

structure undergoes a continuous transition from a high

volume ferromagnetic state to a disordered noncollinear

spin arrangement at lower volumes This noncollinearity

leads to an unusual dependence of the binding energy as a

function of volume and hence, to unusual thermal sion properties

expan-Many other materials (both crystalline and disordered)exhibit the Invar effect, including Fe3Pt, Cr,α-Mn, Lu2Fe17,and Y2Fe17 There is also a related phenomenon known asthe anti-Invar effect in which a magnetic phase transitionleads to an anomalously high thermal expansion Such aproperty could be used as an activator in a micromachine

K−1( 10) No other material displays such large isotropiccontraction across such a wide temperature range Eventhough this material has been known since the late 1950s,only in the last few years have its structure and unusualproperties been fully understood

The structure ofα-ZrW2O8, determined by powder tron diffraction, is shown in Figure 9 It is a cubic material

neu-(space group P213) and can be described as a network

of corner-sharing ZrO6 octahedra and WO4 tetrahedra.Each octahedron shares all six corners with a WO4tetra-hedron; each tetrahedron shares only three of its fourcorners with an octahedron leaving one oxygen strictly

(a)

(b)

Figure 9 The structure of cubic ZrW2O8 (a) A polyhedral sentation of corner-sharing ZrO6octahedra and WO4tetrahedra (b) A ball-and-stick view of a section of the framework structure Zr small dark circles, W small light circles, oxygen large grey circles The one-coordinate oxygen is shaded a lighter gray.

9.09

3009.10

9.119.129.139.14

Angstroms 9.159.169.179.189.19

Temperature/K

DilatometerData

Neutron DataNIST

1200 1500

Neutron DataHRPD

Figure 10 Thermal expansion behavior of ZrW2 O 8 The plot

shows the cubic unit cell parameter of ZrW 2 O 8 from 2 to 1443

K Data are derived from powder neutron diffraction and

dilato-metric studies (see text for details) The dilatodilato-metric data, where

the physical length of a ceramic bar is measured, have been scaled

to the cell parameter derived from neutron diffraction data at 298

K Regions of overlap have been omitted for clarity ZrW 2 O 8 is

ki-netically stable up to 1050 K, unstable in the dotted region of the

curve, and thermodynamically stable at 1443 K.

one-coordinate The temperature dependence of the cubic

unit cell of ZrW2O8is shown in Fig 10 Data points from 2

to 520 K were determined by powder neutron diffraction,

where the unit cell parameter is measured directly Data

from 520 to 960 K were determined by dilatometric

mea-surements on a ceramic block of the material, whereas the

single data point at 1443 K was determined by neutron

diffraction The only region of this graph where ZrW2O8

is thermodynamically stable is the single data point at

1443 K Below approximately 1050 K, however, ZrW2O8

is metastable for extended periods of time The origin of

this negative thermal expansion can be directly traced

to the structure of the ZrW2O8 framework Dove and

co-workers showed that the topology of the structure is such

that ZrW2O8 has intrinsic flexibility and can support a

large number of rigid unit modes (35) Thus, ZrW2O8is a

three-dimensional analog of the simple 2-D model of Fig 7

Their calculations predict that ZrW2O8will support a large

number of low-energy vibrations that tend to contract the

lattice

Experimental evidence for this model has been obtained

by a number of techniques David et al showed how

infor-mation about the Gruneisen parameter weighted

vibra-tional density of states can be extracted directly from the

cell parameter information of Fig 10 by using maximum

entropy techniques (Fig 11) (36) This work shows that the

negative thermal expansion can be attributed to a family

of modes whose energy ranges from 3 to 8 meV Ernst et al

used inelastic neutron scattering to probe the phonon

den-sity of states directly (Fig 12) (37) These measurements

show that there are a significant number of low-energy

phonons in ZrW2O8, and again led the authors to conclude

that modes in the energy range of 1.5–8.5 meV are most

im-portant for NTE Specific heat capacity measurements by

Ramirez and Kowach again highlighted the importance of

low-energy vibrations in ZrWO (38) Recent calculations

Figure 11 A maximum entropy reconstruction of the Gruneisen

parameter weighted phonon density of states extracted from the cell parameter data of Fig 10 The negative peak around 55 K (4.7 meV) corresponds to a family of low-energy modes that tend

to contract the lattice; those centered at 850 K (73 meV) tend to expand the lattice.

provide further support for the importance of these tions (39)

vibra-It has been shown that the structure of ZrW2O8givesthis material a number of other highly unusual proper-ties as a function of both applied temperature and pres-sure As the material is warmed to 450 K, it undergoes

a phase transition toβ-ZrW2O8in which the lighter shadedoxygen atoms of Fig 9 become dynamically disordered(34,40) Oxygen migration at such low temperatures ishighly unusual Under an applied pressure of around0.3 GPa, ZrW2O8undergoes a further phase transition to

γ -ZrW2O8, which also involves the migration of oxygenatoms to new sites in the structure (41–43) When heated

to 100◦C, theγ form reverts to the low pressure α form.

This series of phase transitions it has been suggested, is

a potential method of energy storage At higher pressures,

TOFFA

60 80 100 1200

0.00.51.01.52.0

1 per unit call)

2.53.0

140

hω– (meV)

Figure 12 Generalised phonon density of states g(w) of ZrW2O8

at 300 K Data recorded by inelastic neutron scattering using of-flight (<40 meV) and filter-analyzer spectroscopy (>40 meV)

time-[reproduced from article by Ernst et al., with permission(37)].

the material becomes amorphous (44) Recently Wilkinson

et al successfully prepared a new room temperature

polymorph—trigonal ZrW2O8 Taken together, these

ob-servations suggest that the unusual thermoresponses of

ZrW2O8 are related to the metastability of the material,

its unusual topology, and its proximity to a number of

dif-ferent phase transitions

Other materials such as Hf W2O8 and ZrW2 −xMoxO8

have been prepared recently and characterized (45–48)

AM 2 O 7 and Related Phases

It has been known for a number of years that members

of the cubic AM2O7family (e.g., ZrV2O7 and ZrP2O7)

ex-hibit unusual thermal expansion properties Their

struc-ture (Fig 13) is closely related to that of ZrW2O8and again

contains corner-sharing ZrO6 octahedra and AO4

tetra-hedra In contrast to ZrW2O8, however, the tetrahedra

form A2O7pyrophosphate/vanadate units and thus share

all four of their corners; therefore, all oxygens are

two-coordinate This basic structural type is known for a range

of A metals, including A= Si, Ge, Ti, Re, Mo, Nb, Sn, Zr,

Mo, W, Pb, Ce, U, and Th, and for M= P, V The structural

properties of many of these materials as a function of

tem-perature are complex At high temtem-peratures, they have the

simple cubic structure of Fig 13 When cooled, however,

many members of this family undergo displacive phase

transitions In ZrP2O7, the unit cell parameter as a

func-tion of temperature shows a single phase transifunc-tion to

what is apparently a 3× 3 × 3 cubic supercell at room

tem-perature (49) ZrV2O7, on the other hand, undergoes two

(a)

(b)

Figure 13 Structure of ZrV2O7 (a) ZrO6octahedra share corners

with VO4 tetrahedra (b) VO4tetrahedra share corners to form

V O pyrovanadate groups.

clear phase transitions, initially to an incommensuratelymodulated ∼3 × 3 × 3 superstructure, then, finally to anordered 3× 3 × 3 superstructure (50) Both SiP2O7 andTiP2O7show similar behavior (51,52)

Thus, these structures show the classic behavior offramework materials, as described previously: a strongpositive coefficient of thermal expansion as they approach

a displacive phase transition, followed by low or tive thermal expansion thereafter High temperature co-efficients of thermal expansion areα l = −7.1 × 10−6(400–

nega-500 K) and+5.4 × 10−6 K−1 (600–700 K) for ZrV2O7 andZrP2O7respectively

Interestingly, Sleight and co-workers showed how theformation of solid solutions of the type ZrV2 −xPxO7can beused to suppress these phase transitions and extend therange of negative thermal expansion to include room tem-perature (53) Figure 14 contains dilatometric data for se-lected members of the family Materials such as ZrVPO7have low or negative thermal expansion at room tempera-ture and below This represents an important way in whichthermal expansion properties may be effectively controlledand material properties fine-tuned

NZP Phases

The NZP or Nasicon family of materials is one of the mostwidely investigated for controllable low or negative ther-mal expansion properties This family derives its name

from the parent composition Na Zr2(P O4)3and has generalformula M1 M2 A2(BO4)3, where M1, M2, A, and B can be

a variety of metal cations The name Nasicon is usuallyreserved for the Na1 +xZr2(SiO4)x(PO4)3 −x family that hasbeen widely investigated as a solid-state electrolyte due toits high ionic conductivity [see, for example, Breval et al.for a review (54)] Nasicon derives its name from Na supe-rionic conductor

The structure of NZP is shown in Fig 15 It can be scribed as a framework of fully corner-sharing ZrO6 oc-tahedra and PO4 tetrahedra Two octahedra are linked

de-by three PO4tetrahedra to form Zr2(PO4)3 − units Thesegroups align in chains that run parallel to the c axis andare cross-linked via PO4groups to adjacent units to form

a 3-D framework The framework created contains a ber of interstitial sites where charge balancing M1and M2cations are located Within the Zr2(PO4)3 −units, there is

num-a trigonnum-al prismnum-atic site thnum-at is usunum-ally empty There num-areapproximately octahedral sites (site symmetry –3) betweenadjacent Zr2(PO4)3 −units in the same chain that are oc-cupied by Na in the parent structure In addition, thereare three more approximately octahedral sites betweenthe chains This network of full and empty sites providesthe low-energy migration pathways that give these phasestheir ionic conduction properties.1

1The description here is of the rhombohedral R ¯3c form of the

material Several NZP compositions [e.g., LiZr 2 (PO 4 ) 3 at room temperature and Na 1 +x Zr 2 (SiO 4 ) x (PO 4 ) 3 −x for 1.8<x<2.2) un-

dergo a phase transition to a monoclinic structure when cooled from elevated temperatures This phase transition is displacive and involves static rotations of the polyhedra The essential fea- tures of the structure remain unchanged, though thermal expan- sion properties are influenced.

Figure 14 Dilatometric data recorded for various

members of the ZrV 2 −x P x O 7 family (53) (data

pro-vided by the authors).

−0.35

−0.25

−0.15

−0.050.05

ZrV2O7

ZrP2O7

ZrVPO7

ZrV1.9P0.1O7ZrV1.8P0.2O7

(b)

Figure 15 The structure of NZP materials ZrO6 octahedra share

corners with PO 4 tetrahedra to form a 3-D framework Na cations

(circles) sit in interstitial sites in this framework (a) A view down

the c axis of the material (b) A view perpendicular to the c axis

em-phasizing the chains of polyhedra in the structure The “lantern”

group of two octahedra and three tetrahedra linked together is

also present in the structure of Sc (WO ) (Fig 16).

The NZP family meets many of the critera laid outpreviously The structure contains a framework of corner-sharing polyhedra that have two-coordinate bridgingoxygen atoms and a significant number of “vacant” inter-stitial sites that can accommodate vibrational modes of thepolyhedra

Interest in this family of materials was originally ulated by a paper by Boilot et al (55) who described dilato-metric data for Na1 +xZr2(SiO4)x(PO4)3 −xmaterials, where

stim-x ranges from 1 to 3, and gave the prescient statementthat “the compound x= 1 displays an important shrink-age which could allow this material to be useful when ex-pansion is undesirable.” Stemming from this original com-ment, there has been a considerable body of work on thesematerials, most notably by Roy, Agrawal, McKinstry, andtheir co-workers at Pennsylvania State University.Diffraction measurements show that the low overallthermal expansion in the simple MZr2(PO4)3phases (M=

Li, Na, K, Rb, Cs) is due to significant anisotropy in thethermal expansion of the a and c axes; the c axis typicallyshows an expansion, and the a axis shows a contraction.Lenain et al (56) related this fact to the observation thatsimilar changes in cell parameters are observed as onesubstitutes successively larger alkali metals at the M site

An elegant model was proposed to explain this anisotropybased on geometrically coupled rotations of the polyhedralunits of the structure Expansion of the M1site causes cou-pled displacements of the framework polyhedra The infer-ence from this work was that increasing temperature has

an effect comparable to increasing the size of the alkalimetal Sleight and Khosrovani (57) took a more quanti-tative approach and used a combination of the tempera-ture dependence of M–O distances from the literature andthe purely geometric distance least squares (DLS) model-ing technique to rationalize the properties of these phases.Variable temperature diffraction studies of LiGe2(PO4)3

by Alami and more recently by Lightfoot and co-workers(58–61) support these models and confirm the basic ideas.The overall change in cell parameters as a function of

Table 2 Thermal Expansion Coefficients of Selected NZP Related Materials Thermal Expansion Coefficients for These Materials Are Highly Dependent on Sample Preparation Methodsa

diffraction and dilatometry were presented in the same study.

temperature is caused by changes in the size and shape

of the M1site which causes the octahedra and tetrahedra

to undergo geometrically coupled rotations along with

mi-nor internal distortions (octahedra typically distort to a

greater extent than tetrahedra)

One of the most technologically significant features of

the NZP structure is the wide range of both iso and

alio-valent cation substitutions possible at the M1, M2, A, and

B sites The M cations can be substituted by a range of

species, including H+, Li+, K+, Rb+, Cs+, Ag+, NH+4, Mg2 +,

Ca2 +, Sr2 +, and Ba2 +; the Zr site by Na+, Mg2 +, Mn2 +, Ni2 +,

Cu2 +, Zn2 +, Al3 +, Co3 +, Fe3 +, In3 +, Sc3 +, Y3 +, Yb3 +, Nd3 +,

Ti4 +, Ge4 +, Sn4 +, Hf4 +, Th4 +, U4 +, and Nb5 +; and the P site

can be substituted by Si4 +, Ge4 +, As5 +, and S6 + This

combi-nation of doping possibilities leads to an essentially infinite

family of materials whose low thermal expansion

proper-ties are controllable (see Table 2) In terms of cation

substi-tution, NZP is one of the most flexible inorganic materials

known Another potential use of these materials is in

nu-clear waste storage Roy et al showed that NZP will

accom-modate all of the ions in a normal nuclear fuel reprocessing

scheme to form a single-phase ceramic; this has clear

en-vironmental implications

The vast majority of NZP phases actually expand at

the unit cell level, even though they show strong bulk

contraction (as measured, for example, by dilatometry)

NaTi2(PO4)3, for example, has a bulk expansion

report-edly as low as−4.55 × 10−6K−1(62), yet the intrinsic cell

volume change from the same study is +3.2 × 10−6

K−1(22–1000◦C) This discrepancy between intrinsic and

extrinsic behavior can again be related to the marked

anisotropy of expansion of these phases; the overall

prop-erties of a ceramic body depend strongly on intergrain

in-teractions The anisotropy can lead to microcracking in

bulk specimens due to stresses induced by different

expan-sions of neighbouring grains in the ceramic body and can

frequently lead to hysteresis in the thermal expansion of

composites The healing of microcracks gives an additional

component to the contraction on warming; their formation

on cooling an expansion Breval et al (63) showed that

in-troducing Si in NZP phases can lead to the formation of

glassy grain boundaries that soften at relatively low

tem-perature and lead to further discrepancy between extrinsic

and intrinsic properties

An important consequence of the possibility of

aliova-lent doping in NZP is the ability to prepare materials where

the M1 site is either fully occupied [e.g., NaZr(PO ) ],

50% occupied [e.g., Ca0.5Zr2(PO4)3], 33% occupied [e.g.,

La0.33Zr2(PO4)3], or fully vacant [e.g., NbTi(PO4)3].Because the behavior of the M1site influences polyhedraltilts in these materials to such an extent, one has a rel-atively straightforward method of controlling thermal ex-pansion properties For example, though phases such asNaZr2(PO4)3show a positive expansivity of the c and neg-ative expansivity of the a axes, this situation is reversed

in certain materials such as Ca0.5Zr2(PO4)3, with the c axiscontracts, and a expands The presence of a partially va-cant and therefore compressible M1site provides a flexiblebuffer to accommodate expansion of the full sites

Ca0.5−xSrxZr2(PO4)3is one system that has been tigated in some detail This interest was again prompted

inves-by the observation that in Ca0.5Zr2(PO4)3, the a axis tracts and c expands whereas in Sr0.5Zr2(PO4)3a expandsand c contracts.2 Workers at Penn State were successful

con-in synthesizcon-ing Ca0.25Sr0.25Zr2(PO4)3in whichαaandαcmained nearly constant when heated to 500◦C Acousticemission studies showed that microcracking in such a ma-terial when cooled can be greatly reduced (64) The overallcoefficient of thermal expansion of such a composite wasreported to be−0.1 × 10−6K−1between 298 and 873 K.Another interesting phase is NbTi(PO4)3which has fullyvacant M sites It is unusual among the NZP family be-cause it shows intrinsic negative thermal expansion; the a-axis contraction dominates the c-axis expansion Woodcock

re-et al (65) provided dre-etailed thermal expansion data forthis compound and found that α l (defined as 1/3 α V)

ranges from –3 at room temperature to around +0.5 ×

10−6 K−1 at 1000◦C The larger alkali metal (K and Rb)systems have also been reported to show intrinsic contrac-tion LiZr2(PO4)3exhibits sharp contraction [α l = −10.1 to

−22.2 × 10−6K−1) between 20 and 155◦C (66)

Agrawal and Roy also demonstrated that low sion composite bodies can be made between NZP mem-bers and other ceramics such as MgO, ZnO, Nb2O5, ZrSiO4.Another interesting application of NZP phases in produc-ing smart high-tech materials was described by Agrawaland co-workers (67) Carbon–carbon composites have anumber of properties that make them ideal for certain

expan-2 The observation that Ca 0.5Zr 2 (PO 4 ) 3 shows the behavior

typi-cal of an ordered R¯3c structure rather than the behavior of a dered R¯3 structure has been rationalized in terms of the vacancy

disor-site in Ca 0.5Zr 2 (PO 4 ) 3 which is smaller and thus less compressible than in the Sr compound.

applications in space vehicles These include a high

strength to weight ratio, good strength retention at high

temperature, and good resistance to thermal shock They

do, however, have one major drawback—they are

read-ily oxidized at high temperatures A number of

materi-als have thus been tested as potential protective

coat-ings Most, however, suffer from poor thermal mismatch

to C–C composites, which leads to their ultimate failure

Agrawal et al showed that members of the NZP

fam-ily whose coefficients of thermal expansion match those

of the composites can be successfully used as coatings

Ba0.875Zr2(SiO4)0.175(PO4)2.825, for example, can be

success-fully hot pressed around composite bars, and the resultant

body can be heated to 1200◦C without any significant

oxi-dation of carbon

The Sc 2 (WO 4 ) 3 Family

Negative thermal expansion has also been reported in the

A2(MO4)3family of materials with Sc2(WO4)3related

struc-tures These materials can be described as containing AO6

octahedra that share all six of their corners with MO4

tetrahedra (Fig 16) These materials crystallize in the

or-thorhombic space group Pnca at high temperatures, as

originally described by Abrahams and Bernstein, though

several members of the family undergo a volume-reducing

displacive phase transition to a monoclinic structure when

cooling (68,69) These materials meet many of the

crite-ria laid out previously (strong metal-to-oxygen bonds, rigid

polyhedra, no interstitial cations), and it is not

surpris-ing that they show low or negative expansion coefficients

It is also relevant to emphasize their structural

similar-ities to several of the other materials described before

Corner-sharing octahedra and tetrahedra are common to

the ZrW2O8, ZrV2O7, and NZP families of materials In

particular, the A2(MO4)3 family is closely related to the

interstial cation-free members of the NZP family such as

NbTi(PO4)3; the basic building blocks of the two structures

are identical, but their 3-D connectivity differs

x z

y

Figure 16 The structure of Sc2(WO4)3 ScO6octahedra (white)

share corners with WO tetrahedra (shaded).

1.0

Figure 17 Thermal expansion properties of selected members of

the In 2 −x Al x (WO 4 ) 3 family of materials.

This family of materials also has a high degree of ical flexibility The A3 +site can be doped by using a variety

chem-of elements ranging in size from Al3 +(r = 0.535 ˚A) to Gd3 +

(r = 0.938 ˚A) Additionally, aliovalent doping using, for

ex-ample, Zr4 + on the A site and P5 + on the M site to duce phases such as Zr2(PO4)2(WO4) is possible Thesesubstitutions again lead to controllable thermal expansionproperties (70–73) Y2(WO4)3, which contains the largest

pro-A3 + cation studied crystallographically, has the mostnegative coefficient of thermal expansion (72) The mon-oclinic to orthorhombic phase transition can also be influ-enced by the A-site electronegativity In2(WO4)3undergoes

a monoclinic to orthorhombic transition at around 600 K(Fig 17); Al2(WO4)3 is orthorhombic at all temperaturesabove 300 K By preparing solid solutions such asAlIn(WO4)3 it is possible to produce ceramic bodies dis-playing near zero net expansion (Fig 17) Expansion coef-ficients for selected materials are included in Table 3

It is worth noting that, like the NZP materials, mostmembers of the Sc2(WO4)3 family display anisotropicexpansion Intrinsic expansion (measured by diffractionmethods) can be very different in bulk properties

Sc2(WO4)3, for example, has a cell volume expansion ficientα V = −6.5 × 10−6K−1corresponding toα l = −2.2 ×

coef-10−6K−1but can show linear expansion coefficients as low

as−11 × 10−6K−1as a bulk ceramic body

Zeolites

Zeolites are another group of materials that meets many

of the criteria laid out previously for unusual sivity Zeolites can be described as frameworks built upfrom corner-sharing SiO4 and AlO4 tetrahedra and havecharge-balancing species (atomic or molecular) located invoids of the frameworks Thus, their general formula can

expan-be expressed as Mx/n[(AlO2)x(SiO2)2 −x].yH2O, where Mn +cations occupy cavity sites in the framework There arearound 200 aluminosilicate frameworks known of whicharound 40 are natural minerals In addition to the zeolites,there are a significant number of materials, often calledzeotypes, that contain elements such as P, Be, Ga, and

Zn at the tetrahedral sites Of particular interest in the

Table 3 Thermal Expansion Coefficients of Selected A 2 (MO 4)3 Materialsa

is given in parentheses.

search for negative thermal expansions are the pure silica

frameworks [i.e., x= 0 in the preceding general formula]

in which there are no interstitial cations Therefore,

they can be described as (metastable) polymorphs of

silica Cation-free frameworks are also found in the

Al/P framework materials such as AlPO4 The thermal

expansion properties of the zeolites provide an interesting

example in which computer modeling predicted an

un-usual property before it was experimentally measured In

1993, Parker and co-workers used atomistic simulations

to predict that certain zeolites should contract as

tem-perature increases (74) These properties were confirmed

by subsequent diffraction experiments In 1995, they

modeled 18 different zeolite structures and predicted that

all but two ought to contract as temperature increases

This behavior was subsequently confirmed by diffraction

measurements Negative thermal expansion has since been

found in a number of zeolites, including pure SiO2

poly-morphs of faujasite and chabazite ITQ-1, ITQ-3, ITQ-9,

SSZ-23, ZSM-S, AlPO-5, and ALPO-17 (75–80) Park

et al (79) emphasized the difference in thermal expansion

properties of synthesized and calcined zeolites Typical

coefficients of thermal expansion are included in Table 4

β-Eucryptite/Spodumene

β-Eucryptite, LiAlSiO4, can be described as a

“stuffed-quartz” structure (20) Half of the Si atoms of quartz are

replaced by Al atoms, and charge-balancing Li+cations

Table 4 Expansion Coefficients of Zeolitesa

Zeolite α (diffraction) × 10−6K−1 Temperature Range Ref.

reside in the sixfold channels visible in Fig 6 (at the cellorigin) Al/Si atoms are ordered parallel to the c axis lead-ing to a doubling of the unit cell c axis, and Li ordering

in the various tetrahedral sites of the channels leads todoubling of the a and b axes.β-Eucryptite has been well

studied due to its low or negative thermal expansion; α v

is−6 × 10−6K−1from 20 to 523 K and+0.29 × 10−6 K−1from 523 to 873 K (82) This low expansion is achieved viasignificant anisotropy; the c axis contracts, and the a axisremains approximately constant (<298 K) or expands As

a result,β-eucryptite and other structurally similar

com-pounds in the Li2O–Al2O3–SiO2(LAS) system are used tensively in low-expansion glass ceramics Products based

ex-on these systems include oven-to-table cookware and highprecision engineering components

Various workers have proposed mechanisms to explainthe low expansion ofβ-eucryptite Recently Xu et al com-

pared the expansion properties of ordered and disorderedAl/Si samples (82) They concluded that a variety of mecha-nisms determine the overall expansion properties, includ-ing tilting of the Si/Al tetrahedra, tetrahedral distortion,and disordering of the Li atoms This Li ion disorderingalso gives rise to significant ionic conductivity parallel tothe c axis of the material Several authors have discussedhow the population of different Li sites at different tem-peratures may influence thermal expansion properties Atlow temperatures, however, the principal mechanisms un-derlying anisotropic expansion are tetrahedral tilting andthe expansion of an unusually short O–O bond which arises

from edge sharing of LiO4and SiO4tetrahedra; tetrahedra

are described as lengthening parallel to c and shortening

perpendicular to this axis

A second important member of the LAS family is

β-spodumene (LiAlSi2O6) Whereasβ-eucryptite can be

de-scribed as a stuffed quartz structure, β-spodumene can

be described as derived from keatite, a high pressure

form of SiO2that has not been found in nature Keatite

is a tetragonal material built from corner-sharing SiO4

tetrahedra and contains five- seven-, and eight membered

rings.β-Spodumene is derived from this structure and

con-tains Al/Si disordered across two tetrahedral sites and Li

disordered across four sets of split tetrahedral sites

β-Spodumene again shows highly anisotropic behavior; the

c-axis expands and the a and b axes contract as

temper-ature increases This effect has again been explained by

coupled rotations of tetrahedral groups

One interesting application which has been

demon-strated forβ-eucryptite, and for which materials such as

ZrW2O8 are also being investigated, is for athermalizing

fiber Bragg gratings (FBGs) FBGs are created in

opti-cal fibers by introducing a periodic modulation into the

refractive index of a material and can be manufactured

so that they have precisely controlled optical transmission

and reflection properties As such they can be used to

con-trol the wavelength of light, combine lights of different

wavelengths, and split certain wavelengths from a fiber

These are important properties needed, for example, in

wavelength-division multiplexing (WDM) for optical

com-munication networks

One problem of FBGs is the thermal variation of the

Bragg wavelengthλB, which is influenced by both the

tem-perature dependence of the refractive index (frequently

the dominant term) and the thermal expansion of the

ma-terials used A typical temperature dependence in GeO2

-doped SiO2at 1550 nm is 0.012 nm K−1 This temperature

dependence can be elminated by actively controlling the

temperature of the unit, though this is an expensive and

complex option Passive control can be achieved by

mount-ing the gratmount-ing on a negative expansion support Early

designs relied on using two materials that have

differ-ent (positive) expansion to achieve this effect By using an

NTE mount, both the reliability and ease of manufacture

may be improved Corning, for example, reported a system

based on aβ-eucryptite glass ceramic substrate in which

the temperature dependence ofλB is reduced from 0.012

to 0.001 nmK−1 In this application, the microstructure

(determined by processing conditions) of the NTE support

is extremely important By correct processing, it is possible

to reduce the expansion coefficient ofβ-eucryptite ceramics

from their crystallographic value of∼ −0.4 × 10−6 K−1to

as low as−7 × 10−6K−1(293–393 K) Achieving this high a

negative coefficient is important because the temperature

dependence ofλBcan be approximated by

where  is the grating spacing For an average

refrac-tive index of n = 1.461 and a value of dn/dT = 11 ×

10−6K−1, a thermal expansion coefficient around−7.5 ×

10−6 K−1 is required to compensate for refractive indexchanges (83)

A second interesting technical challenge in the ment of such devices is the production of frits for bondingfibers to low or negative thermal expansion substrates Forthe device described before, Corning used SnO–ZnO–P2O5inorganic frits containing Co/Mg2P2O7 fillers (84) Thesefillers themselves undergo a martensitic transition that isaccompanied by a significant decrease in volume This vol-ume change can be used to adjust the expansion properties

develop-of the frit to match the components it must bond

Cordierite

Cordierite (Mg2Al4Si5O18) has been widely investigatedfor its low thermal expansion properties (85) The struc-ture can again be described in terms of a framework ofcorner-sharing Al/SiO4tetrahedra that form four- and six-membered rings There is an additional octahedral sitethat is occupied by Mg The Al and Si atoms in cordieriteare ordered in six-membered tetrahedral rings that have

an overall orthorhombic symmetry In many synthetic ples, however, there is no such order, and the symmetry ishexagonal; such materials have been called indialites Thestructure of cordierite is closely related to the mineral beryl(Al2Be3Si6O18) in which the tetrahedral sites are occupied

sam-by a mixture of Be and Si Like many silicates, a variety

of cations can be doped into this structural type Syntheticsolid solutions have been investigated using elements such

as Ga, Ge, and Mn; cation substitution again leads to trollable changes in thermal expansion properties Cationsubstitutions in the beryl framework are also evidenced

con-in commercially important phases such as emerald (Cr/Fesubstitution) and related minerals such as aquamarine(blue green), morganite (pink), and heliodor (yellow).These materials (similarly to β-eucryptite and β-

spodumene) achieve a low overall volume expansion cally<∼2 × 10−6 K−1 between 298 and 873 K) via signi-ficant structural anisotropy Most substituted cordieritesexpand in the a–b plane and contract along the c axis The

(typi-magnitude of c-axis contraction typically decreases as T

increases In cordierite itself, c reaches a minimum valuearound 500◦C This behavior can again be ascribed to cou-pled rotations of the constituent polyhedra Cordierite ma-terials are used extensively as low-expansion catalyst sup-ports, especially in automative catalytic converters

Selected Commercially Available Materials

Zerodur is a glass ceramic material developed by SchottGlass for various applications where controlled low ther-mal expansion is required (86) An initial glassy mate-rial is produced that contains predominantly SiO2/Al2O3/

P2O3/Li2O, though it also has smaller amounts of cationssuch as Zn, Mg, Ti, Zr, Na, and As This glass is then sub-jected to careful heat treatment to produce a compositeglass–ceramic containing approximately 70–78% crys-talline material The crystalline component has the β-

quartz structure and displays a negative coefficient ofthermal expansion which compensates for the positiveexpansion of the glassy component Correctly processed,Zerodur can have a coefficient of thermal expansion as low

as 0± 0.02 × 10−6K−1 Zerodur is routinely fabricated in

blocks weighing several tons, and still larger blocks can

be manufactured for special applications Circular mirror

blanks for high powered telescopes up to 8.6 m in

diame-ter that weigh 45 tonnes have been manufactured Zerodur

also has extremely low permeability to He which has led

to its use as building frames for laser gyroscopes which

are used for angle measurements in aircraft, helicopters,

and in spacecraft such as the Ariane rocket Other areas

of application include temperature stable distance spacers

in lasers, supports in microlithography, and in the imaging

optics for microchip manufacture

In the 1960s Corning developed “ULE”, an ultra low

expansion titanium silicate glass The average coefficient

of thermal expansion between 5 and 35◦C is certified as

0± 0.03 × 10−6K−1 The thermal expansion properties of

ULE are unchanged by thermal cycling, regardless of the

heating rates employed ULE lightweight mirror blanks

can also be produced that are reduced in weight up to

95% by using a frit bonding process to produce a

“honey-comb” structure whose cell walls are as thin as 1.3 mm A

fusion-based process leading to an 80% weight reduction in

which small pieces of ULE are welded together was used to

manufacture the 2.4-m mirror blank for the Hubble Space

3 Y Okada and Y Tokumaru, J Appl Phys 56: 314–320 (1984).

4 T.H.K Barron, J.G Collins and G.K White, Adv Phys 29:

609–730 (1980).

5 K Rottger, A Endriss, J Ihringer, S Doyle and W F., Kuhs,

Acta Crystallogr B 50: 644–648 (1994).

6 A., Schauer, Can J Phys 43: 524–531 (1965).

7 D., Taylor, Br Ceram Trans J 83: 129–134 (1984).

8 R Roy, D.K Agrawal, and H.A McKinstry, Annu Rev Mater.

Sci 19: 59–81 (1989).

9 D Taylor, Br Ceram Trans J 90: 64–69 (1991).

10 J.S.O Evans, W.I.F David, and A.W Sleight, Acta Crystallogr.

B 55: 333–340 (1999).

11 J.S.O Evans, unpublished results (1999).

12 J.S.O Evans, T.A Mary, and A.W Sleight, J Solid State Chem.

137: 148–160 (1998).

13 P.M Forster and A.W Sleight, J Inorg Mater 1: 123–127

(1999).

14 M Blackman, Proc Phys Soc B 70: 827 (1957).

15 T.H.K Barron, Ann Phys 1: 77 (1957).

16 G.K White, Contemporary Phys 34: 193–204 (1993).

17 K.S Knight, unpublished results (1999).

18 S.M Bennington, J.C Li, M.J Harris, and D.K Ross, Physica

B 263: 396–399 (1999).

19 H.D Megaw, in Crystal Structures a Working Approach W.B.

Saunders: Philadelphia, 1973.

20 P.J Heaney, In “Silica Physical Behavior, Geochemistry &

Materials Applications” Rev Miner., P.J Heaney, C.T Prewitt,

and G.V Gibbs, eds., 1994; Vol 24.

21 M.T Dove, M Gambhir, K.D Hammonds, V Heine, and A.K.A.

Pryde, Phase Transitions 58: 121–143 (1996).

22 P.R.L Welche, V Heine, and M.T Dove, Phys Chem Miner.

27 D Taylor, Br Ceram Trans J 84: 181–188 (1985).

28 W.F Schlosser, G.M Graham, and P.P.M Meincke, J Phys.

Chem Solids 32: 927–938 (1971).

29 S Chikazumi, J Magn Magn Mater 10: 113–119 (1979).

30 D Gignoux, D Givord, F Givord and R Lemaire, J Magn.

Magn Mater 10: 288–293 (1979).

31 O Sigmund and S Torquato, J Mech Phys Solids 45: 1037–

1067 (1997).

32 R.J Weiss Proc R Soc London A 82: 281–288 (1963).

33 M van Schilfgaarde, I A Abrikosov and B Johansson, Nature

400: 46–49 (1999).

34 T.A Mary, J.S.O Evans, T Vogt, and A.W Sleight, Science 272:

90–92 (1996).

35 A.K.A Pryde, K D Hammonds, M.T Dove, V Heine, J.D Gale

and M.C Warren, J Phys C 8: 10973–10982 (1996).

36 W.I.F David, J.S.O Evans, and A.W Sleight, Europhys Lett.

39 R Mittal and S.L Chaplot, Phys Rev B 60: 7234–7237 (1999).

40 J.S.O Evans, T.A Mary, T Vogt, M.A Subramanian, and A.W.

Sleight, Chem Mater 8: 2809–2823 (1996).

41 J.S.O Evans, Z Hu, J.D Jorgensen, D.N Argyriou, S Short,

and A.W Sleight, Science 275: 61–65 (1997).

42 Z Hu, J.D Jorgensen, S Teslic, S Short, D.N Argyriou,

J.S.O Evans, and A.W Sleight, Physica B 241: 370–372

(1997).

43 J.D Jorgensen, Z Hu, S Teslic, D.N Argyriou, S Short, J.S.O.

Evans, and A.W Sleight, Phys Rev B 59: 215–225 (1999).

44 C.A Perottoni and J.A.H Da Jornada, Science 280: 886–889

47 J.S.O Evans, P.A Hanson, R.M Ibberson, U Kameswari, N.

Duan, and A.W Sleight, J Am Chem Soc 122: 8694–8699

(2000).

48 U Kameswari, A.W Sleight, and J.S.O Evans, Int J Inorg.

Mater 3: 333–337 (2000).

49 R.L Withers, Y Tabira, J.S.O Evans, I.J King, and A.W.

Sleight, J Solid State Chem 157: 186–192 (2001).

50 R.L Withers, J.S.O Evans, J Hanson, and A W Sleight, J.

Solid State Chem 137: 161–167 (1998).

51 E Tillmanns, W Gebert, and W.H Baur, J Solid State Chem.

7: 69–84 (1974).

52 J Sanz, J.E Iglesias, J Soria, E.R Losilla, M.A.G Aranda,

and S Bruque, Chem Mater 9: 996–1003 (1997).

53 V Korthuis, N Khosrovani, A.W Sleight, N Roberts, R.

Dupree, and W.W Warren, Chem Mater 7: 412–417 (1995).

54 E Breval, H.A McKinstry, and D.K Agrawal, Br Ceram.

58 M Alami, R Brochu, J.L Soubeyroux, P Gravereau, G.L.F.,

and P Hagenmuller, J Solid State Chem 90: 185–193 (1991).

59 D.A Woodcock, P Lightfoot, and C Ritter, Chem Commun.

64 V Srikanth, E.C Subbarao, D.K Agrawal, C.Y Huang, R Roy

and G.V Rao, J Am Ceram Soc 74: 365–368 (1991).

65 D.A Woodcock, P Lightfoot, and R.I Smith, Mater Res Soc.

68 S.C Abrahams, J Chem Phys 45: 2745–2752 (1966).

69 A.W Sleight and L.H Brixner, J Solid State Chem 7: 172–174

73 T.A Mary and A.W Sleight, J Mater Res 912–915 (1999).

74 J.W Couves, R.H Jones, and S.C Parker, J Phys Condens.

Mater 27: L329–L332 (1993).

75 D.A Woodcock, P Lightfoot, P.A Wright, L.A Villaescusa,

M.J DiazCabanas, and M.A Camblor, J Mater Chem 9: 349–

80 P Lightfoot, D.A Woodcock, M.J Maple, L.A Villaescusa, and

P.A Wright, J Mater Chem.11: 212–216 (2001).

81 D.A Woodcock, P Lightfoot, L.A Villaescusa, M.-J

Diaz-Cabanas, M.A Camblor, and D Engberg, Chem Mater 11:

INTRODUCTION

Triboluminescence is the emission of photons caused byapplying mechanical energy to a solid The word Tribolu-minescence, coined by Wiedemann (1) in 1895, has its root

in the Greek “tribein” to rub The word has broad generalusage and covers a variety of mechanical methods of exci-tation, spectroscopic origins of luminescence, and mecha-nisms of excitation

Triboluminescence (TL) has very long history To thebest of our knowledge, TL was first reported by FrancisBacon in the early seventeenth century and is mentioned

in The Advancement of Learning (2,3) He observed that

lumps of sugar emit light when they are scraped Atpresent, it is known that nearly one-half of all inorganiccompounds and between one-quarter to one-third of all or-ganic compounds exhibit TL Spectroscopic examinationand characterization of the emitted light are useful for de-termining the excited-state origins of TL The best generalspectroscopic characteristics that can be expected from acrystal is that the TL is similar to photoluminescence (PL)

at the same temperature TL and PL spectra are expected

to be identical if the sole effect of a fracture is a population

of excited electronic states The importance of tions to the crystal and/or the emitting centers caused byfracture is evidenced by the differences between PL and TLspectra For instance, some crystals show TL at room tem-perature but do not exhibit PL at that temperature Morecommonly, a triboluminescent crystal exhibits PL, but the

perturba-TL spectrum contains features that are absent from the

PL spectrum These features may be emission bands that

do not occur in the PL spectrum or changes in the relativeintensities of bands in the TL spectrum compared to those

in the PL spectrum

A number of different mechanisms are required to

ac-count for the phenomenon of TL, although its details are

still poorly understood In this review, the mechanisms of

different types of TL along with their spectroscopic

experi-mental examples are given first Then, our experiexperi-mental

results on TL in inorganic crystals that contain rare-earth

ions are presented Finally, the potential applications of TL

are briefly described

CLASSIFICATION OF TL

The mechanisms by which TL is excited have not yet been

well clarified Thus, in addition to elucidating excited-state

origins and spectral features that are possible due to

ad-vances in TL spectroscopy, classification of the mechanisms

of TL is required to understand the phenomenon The four

mechanisms of triboexcitation that are currently under

ac-tive consideration can be broadly categorized as

electri-cally induced, thermally induced, chemielectri-cally induced and

mechanically induced Each consists of a broad range of

physical processes As will be seen, the problem is not an

absence of mechanisms that can explain the excitation, but

rather is too many variable possibilities The purpose of

this section to present briefly the mechanisms currently

under consideration

Electrically Induced TL

The electrical mechanisms that have been proposed all

require that mechanical energy generates free electrons,

which leads to the emission of photons by electron collision

with molecules, recombination luminescence of cations and

anions, or electroluminescence (EL)

Piezoelectricity in a crystal from pressure requires that

the crystal be noncentrosymmetric (4) When a

piezoelec-tric crystal is cleaved or fractured, one of its newly created

surfaces becomes positively charged and the other surface

of the crack is negatively charged (Fig 1) Generally, all

+

++

+++++

Figure 1 Schematic illustration of charge separation in the crack

of a piezoelectric crystal The crystal is subjected to a tensile stress

σ directed along a polar axis of the crystal.

Table 1 Examples of Triboluminescent Materials Induced by Piezoelectricity

piezoelectric crystals exhibit the phenomenon of TL, andthe crystals that do not show TL are nonpiezoelectric (5).Furthermore, several polymorphic crystals that belong tothe piezoelectric point group exhibit TL, whereas crystalsthat do not belong to the piezoelectric point group do notshow TL (5) Walton et al (6) calculated the voltage gen-erated by the fracture of a piezoelectric crystal under theassumption that the crystal is subjected to a tensile stressdirected along a polar axis of the crystal and estimatedwhether or not the voltage is sufficient to initiate gas dis-charge The piezoelectric constant in this direction is gen-erally of the order of 10−12C/N, and the stress near the tip

of a crack is of the order of 108N/m2 Thus, if it is assumedthat the relaxation time of the stress is less than the time

it takes for the fracture to spread across the crystal, thecharge density of the newly created surfaces is of the or-der of 10−4C/m2 The electric field between the oppositelycharged density of newly created surfaces isρ/ε0, that is,nearly 107 V/m, whereρ is the charge density and ε0 isthe permittivity constant Such a field has sufficient mag-nitude to cause dielectric breakdown of the ambient gas.Several examples of materials where triboluminescence is,induced by piezoelectricity during facture are tabulated inTable-1 (5–18) The TL of the first two materials is so in-tense that it can be seen in daylight

The TL spectrum of sucrose provided clear evidence ofemission from nitrogen gas molecules The nitrogen emis-sion observed corresponds to3u→3gfluorescence (19) Awell-defined vibronic band structure is observed, as shown

in Fig 2 (11) A minimum energy of 8.9× 104cm−1is needed

to excite nitrogen to its3uexcited state The ultravioletpart of the nitrogen gas discharge can excite the PL of sur-rounding materials; the TL of uranyl nitrate hexahydrate,for instance, has such an origin (5,12) On the other hand,

it was found that the TL spectrum of Cu, Ag or Mn-dopedZnS that is piezoelectric corresponds to the PL spectrumthat has a small shift in the energy of the peak maxima(18) All of these shifts in TL spectra were in the same di-rection as those observed upon applying hydrostatic pres-sure The spectrum of Cu-doped ZnS exhibits two bandsand matches the EL spectrum more closely than the PLspectrum

300 310 320 330 340 350 360 370 380 390 400410420

Wavelength

Sucrose

123

Figure 2 Triboluminescent spectrum of sucrose, corresponding

to emission from a nitrogen molecule [from (11)].

In several cases, the crystals and glasses whose

struc-tures possess a center of symmetry and are

nonpiezoelec-tric exhibit TL This means that processes other than

piezo-electricity are also responsible for charging newly created

surfaces In alkali halides such as LiF and NaCl without

piezoelectricity, direct evidence of positive and negative

surface charges on fresh fracture surfaces was provided

by Wollbrandt et al (20) using an electrostatic probe that

had a 100-µm spatial resolution Such electrical activity

was so strongly supported by the observation that

electro-magnetic radiation in the radio range was detected

coin-cident with the fracture of LiF, NaCl, and MgO (21,22)

One possible source for creation of charged surfaces is the

motion of charged dislocations intersected by the fracture

surface (23,24) Another source is internal electrification

at cleavage or shear planes in the crystal (23,24) If the

charge density is sufficiently high, microdischarge takes

place in the vicinity of the crack tip, where catastrophic

ion-ization results in dielectric breakdown of the ambient

at-mosphere The electrical discharge yields emission of

pho-tons, electron, and other particles that accompanies the

fracture of solids, including metals, semiconductors, and

insulators (25–27) The emission of photons, electrons, and

other particles can also excite other luminescent centers,

such as impurity ions or defect centers Even in the

ab-sence of microdischarging, local concentrations of surface

charge accelerate previously emitted electrons toward the

surface Because the energy of emitted electrons is in the

keV range (28), electron bombardment can be sufficiently

energetic to cause further emission of photons, electrons,

and other particles

In the 1965 Matsushiro earthquake swarm, where

many cracks and fissure zones were formed at the surface

(29), luminescence, as well as an anomalous change in the

earth’s potential, was observed (30) Such surface

crack-ing causes emission of charged particles such as the

exo-electron, which is responsible for exciting luminescence,

as suggested by Brady and Rowell (31) They recorded the

emission spectra of photons induced by the fracture of rocks

such as granite and basalt in argon, helium, air, water, andvacuum and found that the emission consists of line spectracorresponding to the excitation of the ambient atmosphere.Although many rock materials contain minerals, known aspiezoelectrics, the basalt was free from any strong piezo-electric minerals In addition, the spectrum of basalt inargon was identical to that produced by granite in argon.Deformation of piezoelectric minerals does occur duringfracture of rocks, but these facts imply that the piezoelec-tric mechanism is not associated with TL in these minerals

It was proposed that an exoelectron bombardment nism is responsible for exciting the ambient atmosphere.Moreover, photons, electrons, and other particles areemitted by the fracture of silica glass (32–34) Silica glasshas excellent optical properties that can be used for prac-tical applications One of the superior properties of silicaglass as an optical material is high transparency, whichmakes it possible to use silica glass as windows, lenses,prisms, waveguides, and fibers for optical telecommunica-tions Kawaguchi (34) investigated the time-resolved TL

mecha-of silica glass in vacuum and nitrogen atmosphere Twoemission bands at 1.9 and 2.7 eV were observed in thespectra The 1.9-eV band peaks around 5µs and decays

around 100µs after the instant of fracture The 2.7-eV

band rises in about 50µs, peaks around 500 µs, and decays

in several tens of milliseconds after fracture The energyposition and the time response of the two bands weresimilar to those in PL The 2.7-eV band was ascribed to theluminescence of oxygen vacancies on the fracture surface

of the silica glass, and the 1.9-eV band was related torelaxation of nonbridging oxygen hole centers A plausiblemechanism by which TL can be excited is that the defectcenters created at fracture are excited by collisions ofemitted electrons and ions Another mechanism is that thedefect centers are excited directly during defect formation

by mechanical energy supplied

Chemically Induced TL

Chemical reaction takes place between atoms and ions erated during the fracture of crystals, and subsequently,the release of energy gives rise to luminescence Moreover,molecules of the surrounding gases are absorbed or ad-sorbed on newly created surfaces when the solid is frac-tured, and the release of energy in this process causes lumi-nescence Kasemo and Walden (35) reported spontaneousemission of photons and electrons during chemisorption ofchlorine on sodium Kasemo (36) also observed lumines-cence during chemisorption of oxygen on aluminum andmagnesium surfaces

lib-Thermally Induced TL

When fracture occurs, plastic work is transformed intoheat The heat produced near the crack tips either stim-ulates the defect centers and causes luminescent excita-tion or leads to blackbody radiation or incandescent emis-sion if it is very high Although incandescent emission doesnot involve luminescence, it is included here to account fordifferent processes of light emission during deformation orfracture

Wick (37) pointed out that the TL of X- orγ -ray

irradi-ated synthetic fluorites (CaF2) that contain different

rare-earth ions occurs by deformation-induced

thermolumines-cence The TL spectra were composed of emission bands

characteristic of trivalent rare-earth ions such as Dy and

Tb Deformation-induced thermoluminescence was also

ob-served for X- orγ -ray irradiated alkali halides (6) On the

other hand, by solving the thermal conduction equation,

it was shown that temperatures of the order of 104K can

be sustained within the plastically strained regions in the

vicinity of crack tips (6) Fracture of alkali halide crystals

such as LiF, NaF, and NaCl in vacuum was accompanied by

flashes of high temperature luminosity, that is,

incandes-cent TL (6) The blackbody curves were fitted to the spectra

by assuming a temperature of 104K for LiF and NaF and of

6× 103K for NaCl Because these temperatures are higher

than the melting points of the samples, it was postulated

that the time taken by the crack to grow by a length equal

to the diameter of the local hot spot is less than the time

for the heat to be conducted away through the body of the

crystal

Chapman and Walton (38) measured TL spectra of a

va-riety of glasses and of crystalline quartz cut slowly by a

ro-tating diamond-impregnated saw blade and found that the

TL spectra resemble the emission of a blackbody radiator at

a temperature that corresponds to that of the heated

ma-terial Emission temperatures were estimated at around

1850 K for armor plate glass, 2100 K for Pyrex glass, 2400 K

for soda lime glass, 2300 K for high-density lead glass, and

2800 K for cut quartz Because the TL spectra of quartz and

glasses did not contain incandescent emission during

frac-ture, it seems that the blackbody emission produced during

cutting of these materials by an diamond-impregnated saw

blade is attributable to the heat produced by the friction

between the blades and the samples

Mechanically Induced TL

There is a possibility that thermally activated electronic

transitions occur in large strain or high stress regions This

effect is the inverse of thermal radiationless quenching of

excited states When displacements along the

configura-tion coordinate fluctuate greatly, the potential curves of

the ground and excited states cross each other (Fig 3),

and the subsequent transition of excited electrons to the

ground state gives rise to luminescence The mechanism

of radiationless transitions that lead to a population of

excited electronic states in large strain or high stress

re-gions near a crack tip was theoretically advanced by Lin

et al (39) Recently, Li and his co-workers (40,41)

inves-tigated TL during cleavage of semiconductors, and found

that when the number of atomic bonds that are broken on

an average cleaved pair of surfaces is about 1015, the

num-ber of emitted photons is nearly 108–109 The observation

that the number of emitted photons is much less than the

number of broken bonds reveals that the cleavage process

is mainly nonradiative It appears that the TL of Si, Ge,

InP, and other semiconductors is caused by radiationless

transitions via thermal excitation, where the interatomic

distance among some of the atoms of broken bonds is very

Figure 3 Schematic illustration of TL originating from an

ex-cited electronic state via radiationless transition The |g and |e represent the potential energy curves of ground and excited states, respectively.

large and therefore, the thermal movement of electronsfrom the ground state to the excited state is possible

On the other hand, Chapman and Walton (42) observed

TL for single crystals of fluorite (CaF2) doped with Tb3 +,

Dy3 +, Sm3 +, or Eu3 + The crystals were chosen becausethe possibility of piezoelectrically induced TL can be ruledout When the crystals were cut by a diamond-impregnatedcircular saw, the TL spectra show significantly more struc-ture than the room-temperature PL spectra Although theexcited-state origins of TL are rare-earth ions, it is likelythat emitting rare-earth ions are located in the vicinity

of the tip of a growing crack and/or on the faces of thevirgin crack, where the applied stress is concentrated in

a severely distorted crystal lattice Hence, the rare-earthions are subjected to a significantly distorted environmentfor time of the order of an atomic vibrational period This,

in turn, leads to changes in the Frank–Condon factors andthus to enhanced vibrational spectral components The per-turbing effect of high stress or large strain on electronicstructures unambiguously appears in TL spectra How-ever, questions remain as to whether or not the TL offluorites doped with rare-earth ions originates from thepopulation of excited electronic states by radiationlesstransitions because the microdischarge of gas moleculestakes place in nonpiezoelectric crystals, as mentioned be-fore Photons that originate from the microdischarge mayexcite the rare-earth ions, and subsequently yield TL Mea-surements of TL in fluorites without rare earths are needed

to reveal the detailed mechanisms of TL

TL OF OXIDE CRYSTALS DOPED WITH RARE EARTHS

Rare-earth-doped inorganic crystals and glasses have tical properties that are interesting and important from

op-fundamental and practical viewpoints The optical

prop-erties of rare-earth-doped inorganic materials are

appli-cable to the development of optoelectronics devices such

as lasers, optical amplifiers, optical memories, and

op-tical modulators As mentioned in the previous section,

some inorganic crystals doped with rare-earth ions

ex-hibit TL For instance, TL was observed in single

crys-talline fluorite (CaF2) doped with trivalent rare-earth ions,

and it was shown that the rare-earth ions are

responsi-ble for TL (42) As for oxide materials, our research group

found that polycrystalline hexacelsians (BaAl2Si2O8 and

SrAl2Si2O8) doped with rare-earth ions exhibit TL caused

by the 4f–4f and/or 4f–5d electronic transitions of

rare-earth ions (43–45) Intense TL was also observed in

poly-crystalline Sr3Al2O6:Eu,Dy and SrAl2O4:Eu2 +, as revealed

by Akiyama et al (46) and Xu et al (47,48), respectively In

addition, Sage et al (49) demonstrated TL in a rare-earth

complex and urged its application to sensing of structural

damage and fracture

In the next section, we describe our recent experimental

results for TL in polycrystalline oxides doped with

rare-earth ions

TL of Hexacelsians Doped with Rare Earths

The fact that polycrystalline barium hexacelsians

(BaAl2Si2O8) doped with rare-earth ions exhibit TL was

accidentally discovered by Ishihara et al (43) The

poly-crystalline samples were prepared from reagent-grade

BaCO3, Al2O3, SiO2, and rare-earth oxides After the raw

materials were mixed thoroughly, the mixture was heated

above the melting point of BaAl2Si2O8 A densified body

of polycrystalline BaAl2Si2O8 doped with rare-earth ions

was obtained by cooling from its liquid state The crystal

structure of BaAl2Si2O8 is schematically illustrated in

Fig 4 The structure consists of layers of silicate and

aluminate structural units between which Ba2 +layers are

inserted

TL was measured at room temperature while

press-ing the polycrystalline samples with a pressure device TL

spectra were obtained by using a CCD detector equipped

with a multichannel analyzer Photoluminescence (PL)

spectra were measured using a fluorescence

spectropho-tometer for comparison As an example, TL (solid circles)

and PL (solid line) spectra of BaAl2Si2O8:Dy3 +are shown

in Fig 5 The emission lines at 483, 576, and 662 nm in

both TL and PL spectra are ascribable to the4F9/2–6H15/2,

4F9/2–6H13/2, and4F9/2–6H11/2transitions of Dy3 +,

respec-tively The peak positions and the relative intensities of the

emission lines in the TL spectrum are almost identical to

those in the PL spectrum It is unambiguous that the TL is

caused by the electronic transitions of the doped Dy3 +ions

Figure 6 shows TL (solid circles) and PL (solid line)

spec-tra of BaAl2Si2O8:Tb3 + The excitation wavelength for PL

is 350 nm All of the emission lines observed in the TL

spec-trum are assigned to the5D4–7FJand5D3–7FJtransitions

of Tb3 +, as indicated in the figure The emission lines due

to the5D3–7F4and5D3–7F5transitions in the TL spectrum

do not appear in the PL spectrum, presumably because the

excitation wavelength (350 nm) is not suitable for emission

from the5D state In fact, these emission lines become

BaAl2Si2O8Hexagonal

a = 0.525 nm

c = 0.784 nmBaOSiAl

Figure 4 Schematic illustration of crystal structure of BaAl2Si2O8.

TriboluminescencePhotoluminescence

Figure 5 TL (closed circles) and PL (solid line) spectra of a

BaAl Si O :Dy 3 +polycrystal.

visible in the PL spectrum when the excitation wavelength

is changed to 250 nm Similar TL and PL spectra were

obtained for polycrystalline SrAl2Si2O8:Tb3 + Thus far, TL

characterized by the electronic transitions of doped

rare-earth ions was observed in BaAl2Si2O8 doped with Eu2 +,

Sm2 +, Sm3 +, Yb2 +, or Ce3 +and SrAl2Si2O8doped with Eu2 +

or Dy3 +, in addition to the examples previously mentioned

In our early work on the TL of hexacelsians, we argued that

the emission peak of TL is shifted to a longer wavelength

compared with PL for 4f–5d transitions (44) However, the

difference in the peak position between TL and PL spectra

is explained mainly in terms of the incomplete sensitivity

correction of the detector used for measuring TL, in

ad-dition to the fact that the S/N ratio in the TL spectra is

very low

It is also known that TL is observed in BaAl2Si2O8

with-out intentional dopants The TL spectrum of BaAl2Si2O8

without intentional dopants manifests intense emission

lines at around 315, 335, 355, and 380 nm, as shown in

Fig 7 Whether or not a crystal exhibits TL depends on its

crystal structure Table 2 shows the relationship among

crystal structure, piezoelectricity, and the presence of TL,

as summarized by Chandra (5) The hexacelsian has a

space group of symmetry D 6h (P6 /mmm) which precludes

piezoelectricity according to this table Nonetheless, TL is

observed for a hexacelsian without intentional dopants, as

shown in Fig 7 Brady and Rowell (31) proposed that the

light emission observed in earthquakes is caused by

exo-electron excitation of the ambient atmosphere Nakayama

(50) found that many charged particles are emitted from

Figure 8 X-ray diffraction pattern of a SrAl2O4:Eu 2 +,Dy3 +

poly-crystalline sample prepared by the conventional solid-state tion All of the diffraction lines but those indicated by closed and open circles, which are ascribable to R3Al5O12and RAlO3(R cor- responds to Eu 3 +and/or Dy3 +), are assigned to SrAl O .

reac-Table 2 Relationship among Crystal Structure, Piezoelectricity, and TL proposed by Chandra (5)

Point Group International Schoenflies

Piezoelectricity Triboluminescence (present: + (present: + absent: −) absent: −)

the light emission brought about by the discharge from

ni-trogen molecules The discharge presumably results from

cleavage at the Ba2 + layer of the BaAl2Si2O8 Doped

rare-earth ions are excited by photons caused by the

dis-charge from nitrogen molecules because a TL was

mea-sured in air The excited rare-earth ions lead to

lumines-cence due to 4f–5d and/or 4f–4f transitions

TL of Alkaline-Earth Aluminates Doped with Rare Earths

As described before, intense TL was observed from

poly-crystalline Sr3Al2O6:Eu,Dy and SrAl2O4:Eu2 + (46–48)

Rare-earth-doped strontium aluminate crystals are very

interesting because some of them exhibit long lasting

phos-phorescence (51) In particular, SrAl2O4doped with Eu2 +

and Dy3 +shows intense phosphorescence that lasts a very

long time The long lasting phosphorescence is a

phe-nomenon where a solid irradiated by UV or white light

beforehand continues to emit light even after the

excita-tion ceases In this secexcita-tion, we menexcita-tion our

experimen-tal results for TL of polycrysexperimen-talline SrAl2O4:Eu2 +,Dy3 +,

Dy3 +-doped (Sr,Ba)Al O , and (Sr,Ca)Al O polycrystals

An interesting application of the TL of SrAl2O4:Eu2 + forsensing stress was demonstrated by Xu et al.(47,48) and iscited in the following section

Figure 8 shows the X-ray diffraction pattern of talline SrAl2O4:Eu2 +,Dy3 + prepared by the conventionalsolid-state reaction Although very weak diffraction linesascribed to R3Al5O12 and RAlO3 (R corresponds to Eu3 +and/or Dy3 +) are observed, as indicated by solid and opencircles in the figure, almost all of the diffraction lines areattributable to SrAl2O4 In addition, although Dy3 + and

polycrys-Eu3 + form the crystalline phases previously mentioned,some of the Dy3 +and Eu2 +ions are incorporated into theSrAl2O4phase and replace the Sr2 +ions because the sam-ple exhibits long lasting phosphorescence

Figure 9 shows photographs of TL taken at various ods using a video camera after uniaxial compressive stresswas applied to the SrAl2O4:Eu2 +,Dy3 +polycrystal In thefigure, the periods calculated from the shutter velocity ofthe video camera are represented below each photograph

peri-as times of 0s to 11/30s, respectively No luminescence wperi-asfound before the fracture of the sample (at 0s) A greenemission is clearly observed in the photographs after the

0s 1/30s 2/30s

Figure 9 Photographs of the TL of a SrAl2 O 4 :Eu 2 +,Dy3 +polycrystal The photographs were taken

at various periods after a uniaxial compressive stress was applied to the sample In (48), a real-time image of stress-induced luminescence was also demonstrated for SrAl 2 O 4 :Eu 2 + Here, it should be

stressed that our TL data are based on the fracture of solids, whereas nondestructive deformation

of the crystal gives rise to TL in (48).

Wavelength (nm)

Figure 10 TL (closed circles) and PL (solid line) spectra of a

SrAl 2 O 4 :Eu 2 +,Dy3 +polycrystal.

fracture of the sample, although the stress required for

the fracture could not be estimated by using our

equip-ment Because the TL was measured after the long lasting

phosphorescence ceased, the luminescence in Fig 9 is

un-doubtedly caused by fracture of the sample In Fig 9, it is

observed that the part that exhibits green emission within

the sample varies from place to place with time This time

dependence results from the possibility that the stress is

not applied uniformly to the polycrystalline sample

TL (solid circles) and PL (solid line) spectra of

SrAl2O4:Eu2 +,Dy3 + are shown in Fig 10 The excitation

wavelength for PL is 330 nm A broad band is observed

at around 500–520 nm in both spectra The wavelength

region of this broad band coincides with the green

emis-sion A comparison of the broad band between TL and PL

spectra reveals that the peak position and the profile are

similar to, indicating that the origin of the luminescence

is the same as that of TL This broad band is ascribable to

the 4f65d–4f7transition of the doped Eu2 +ion Emission

due to the 4f–4f transitions of Dy3 +is not observed in the

TL nor the PL spectrum

In PL, light at a wavelength of 330 nm excites Eu2 +

ions, and radiative decay in Eu2 +is observed It is thought

that Dy3 + plays the role of acceptor for an electron or a

positive hole in long lasting phosphorescence (51,52) As

for TL, emission of electrons, ions, and photons takes place

when triboluminescent solids are fractured Hence, the

fol-lowing model is proposed for the mechanism of TL in a

SrAl2O4:Eu2 +,Dy3 +polycrystal Initially, photons emitted

by the fracture of SrAl2O4excite Eu2 +ions Some excited

electrons relax to the ground state of Eu2 + and radiate

Other electrons and positive holes also formed by the

ex-citation are trapped at defect sites relevant to Dy3 +, and

luminescence due to recombination of the hole and

elec-tron at the Eu2 +site takes place via long lasting

phospho-rescence We speculate that the former process is

domi-nant because the emission does not last for a long period,

as indicated in Fig 9 The intensity of green emission cays rapidly (within one second or so), as found in the pho-tographs in Fig 9 Another possible cause of TL is fracture-and/or deformation-induced thermoluminescence (53) Thefrictional heat generated by the fracture and/or deforma-tion of a crystal stimulates an electron and trapped at adefect site in advance, and the recombination of the re-leased electron and a hole trapped beforehand at the Eu2 +site (i.e., Eu3 +) yields the emission due to Eu2 + It is com-monly known that the europium ion can be in its trivalentstate (Eu3 +) in many solids

de-Note that another mechanism was proposed by Xu et al.(47) for TL based on the nondestructive deformation ofSrAl2O4:Eu2 + They argued that a positive hole trapped

at a certain localized level in the energy gap is released

by the movement of a dislocation which is caused by thedeformation of the crystal, and that the recombination

of the hole with an electron trapped at the Eu2 + site(i.e., Eu+) gives rise to the emission of photons as TL.Nonetheless, the monovalent state of europium ion (Eu+)

is unusual from the chemical viewpoint, and the existence

of Eu+is unclear, to date Thus, the mechanism of TL inthis material is controversial

To clarify the influence of crystal fracture on the localstructure and on the luminescent properties of doped rare-earth ions in alkaline earth aluminates, we have measured

TL in Dy3 +-doped (Sr,Ba)Al2O4and (Sr,Ca)Al2O4tals Dy3 +was selected because the coordination state rele-vant to Dy3 +, including the coordination symmetry around

polycrys-Dy3 +and the electronic state of the chemical bond between

Dy3 +and a ligand, can be readily deduced from the tive intensity of the emission lines assigned to the 4f–4ftransitions of Dy3 + Besides, one can alter the ligand fieldaround Dy3 +systematically by using a solid solution such

rela-as (Sr,Ba)Al2O4and (Sr,Ca)Al2O4as the host material It isknown that broad composition ranges of solid solution arepresent and the compositional dependence of the lattice pa-rameter manifests a monotonic variation in the SrAl2O4–BaAl2O4and SrAl2O4–CaAl2O4systems (54) According toX-ray diffraction analysis, the substitution of Sr by Ba inthe (Sr,Ba)Al2O4 system does not lead to any change inthe crystal structure when the amount of Ba that replaces

Sr is less than 40 mol% The structure of Sr1 −xBaxAl2O4polycrystals prepared by the conventional solid-state re-action is mainly monoclinic (α-SrAl2O4structure) at leastfor x< 0.4, although the hexagonal phase (BaAl2O4struc-ture) exists mainly when the concentration of Ba2 +is largerthan 40 mol% On the other hand, for the SrAl2O4–CaAl2O4system, a drastic change in the X-ray diffraction pattern,that is, crystal structure, is observed when the compositionchanges from Sr0.9Ca0.1Al2O4 to Sr0.8Ca0.2Al2O4, whereasthe crystal structure of SrAl2O4 is very similar to that

of Sr0.9Ca0.1Al2O4 The Sr1−xCaxAl2O4 polycrystals takepseudohexagonal and monoclinic structures for x> 0.2 and

x< 0.2, respectively.

TL spectra of (Sr,Ba)Al2O4 and Sr0.9Ca0.1Al2O4 dopedwith Dy3 + are shown in Fig 11 Three emission lines areobserved in all of the spectra These lines are assigned

to the 4f–4f transitions of Dy3 +; the emission lines ataround 480, 575, and 660 nm are attributable to the4F9/2–

6H /2,4F/2–6H /2, and4F/2–6H /2 transitions of Dy3 +,

respectively, as indicated in the figure These samples

ex-hibit PL as shown in Fig 12 These spectra were obtained

under excitation at 350 nm All of the emission lines in

the figure are attributed to the 4f–4f transitions of Dy3 +,

as indicated in the figure Noted that TL was barely

ob-served in Sr0.8Ca0.2Al2O4:Dy3 + and Sr0.6Ca0.4Al2O4:Dy3 +,

although these compounds exhibit PL under excitation

at 350 nm Presumably, this occurs because the crystal

structure of Sr0.8Ca0.2Al2O4 and Sr0.6Ca0.4Al2O4, a

pseu-dohexagonal structure, is different from those of the other

compounds; the other crystalline phases take a monoclinic

structure, as mentioned before

Comparison of Figs.11 and 12 suggests that the

inten-sity ratio of the 480-nm emission (4F9/2–6H15/2transition)

to the 575-nm emission (4F9/2–6H13/2transition) is almost

independent of the composition for the TL spectra, whereas

the relative intensity of the 480-nm emission to the 575-nm

emission varies with composition for the PL spectra

Be-sides, the intensity ratio of 480-nm emission to 575-nm

emission is smaller in the TL than in the PL spectra

Figure 12 PL spectra obtained under excitation at 350 nm

for Dy 3 +-doped (a) Sr0.6Ba0.4Al2O4, (b) Sr0.8Ba0.2Al2O4,

(c) Sr 0.9Ba 0.1Al 2 O 4 , (d) SrAl 2 O 4 , and (e) Sr 0.9Ca 0.1Al 2 O 4 The assignment of emission lines is indicated in the figure.

To visualize this relationship, the integrated intensitywas evaluated for the 480-nm and 575-nm emissions, andthe intensity ratio was plotted against the mean ionicradius of the alkaline earth in the crystal in Fig 13 Theopen and closed circles correspond to TL and PL, respec-

tively The ratio I(480nm)/I(575nm)increases as the mean ionicradius increases for PL, whereas the ratio is almost inde-pendent of the mean ionic radius for TL The intensity ra-tio is also smaller for TL than for PL Two possibilities aresuggested to explain this phenomenon One is the effect ofself-absorption (55); the photons emitted from an opticallyactive center such as rare-earth ions are reabsorbed by theother active center in TL Because TL usually occurs withinthe bulk, photons are reabsorbed until they come out of thebulk A comparison among TL, PL, and excitation spectrafor the transition that corresponds to 480 nm is shown forSrAl2O4:Dy3 +in Fig 14 The TL, PL, and excitation spec-tra are represented by closed circles, solid line, and opencircles, respectively An overlap of the PL and the excita-tion spectra is observed at around 475 nm Although it is

155 160 165 170

Mean ionic radius (pm)

00.20.40.6

Figure 13 Variation of integrated emission intensity ratio,

I(480 nm)/I(575 nm) vs the mean ionic radius of the alkaline earth

in the crystals Open and closed circles denote the TL and PL,

respectively.

difficult to compare the integrated intensity of these

emis-sion and absorption lines quantitatively, the possibility of

self-absorption cannot be ruled out

Another possibility for the difference in relative

emis-sion intensity between TL and PL spectra involves the

characteristics of the ligand field around Dy3 + It is well

known that the emission intensity for each electronic

tran-sition of a rare-earth ion can be approximately connected

to ligand fields via the Judd–Ofelt theory (56,57)

Accord-ing to this theory, the radiative transition probability for

Figure 14 TL (closed circles), PL (solid line), and excitation (open

circles) spectra at around 480 nm for SrAl O :Dy 3 +.

the electric dipole transition A is expressed as

λ3

 

n(n2+ 2)29

S(aJ :bJ)=

t =2,4,6

whereaJ U (t) bJ is the reduced matrix element that can

be calculated for each electronic transition of rare-earthions The reduced matrix element is independent of thedifference in the environment in which the rare-earth ion

is located On the other hand, because the parameter t

includes the distance between the rare-earth ion and theligand, the charge of the ligands, and so forth, t variesdepending on the kinds of ligand fields and the rare-earthions The line strength in Eq (2) is proportional to the inte-grated intensity of the lines in the optical absorption andemission spectra, so that the parameter which repre-

sents the characteristics of ligand fields is reflected by theintegrated intensity of absorption and emission lines It

is thought that 2 is relevant to the coordination metry of ligands and6 is an indicator of the covalency

sym-of the chemical bond between a rare-earth ion and a and (58,59) As for the Dy3 +, the integrated intensity ra-tio of 480-nm to 575-nm emissions correlates with6/2.This suggests that the integrated intensity ratio becomessmall when the coordination symmetry for Dy3 +is low (60).Hence, the variation of the relative integrated intensity for

lig-PL shown in Fig 13 indicates that the ligand field for Dy3 +changes as the size of alkaline earth ion changes, as ex-pected Presumably, this variation is based on the averagecoordination structure around Dy3 +because Dy3 +ions canoccupy at least two different positions at Sr2 +sites in theSrAl2O4crystal (61) On the other hand, the fact that therelative integrated emission intensity is almost indepen-dent of the mean ionic radius of the alkaline earth in TLsuggests that the average coordination symmetry for Dy3 +which gives rise to the TL does not change even though thekind of alkali-earth ion is varied Furthermore, the smaller

value of I(480 nm)/I(575 nm)in TL spectra compared with PLspectra indicates that the average coordination symmetryfor Dy3 +, which brings about the TL, is low Consequently,

it is suggested that the TL observed is the emission of tons from Dy3 +placed in a distorted site such as a fracturedsurface and/or the vicinity of a crack tip

pho-APPLICATIONS OF TL

Recent findings for rare-earth-doped materials that showintense TL have attracted attention because of their po-tential application for sensing structural damage, fracture,and deformation We describe one example of sensing de-formation demonstrated by Xu et al (47,48) using theSrAl O :Eu2 + polycrystal They mixed the SrAl O :Eu2 +

op-fundamental and practical viewpoints The optical

prop-erties of rare-earth-doped inorganic materials are

appli-cable to the development of optoelectronics... the fracture and/ or deforma-tion of a crystal stimulates an electron and trapped at adefect site in advance, and the recombination of the re-leased electron and a hole trapped beforehand at the