Motivated by the advantages of two-electrode flash sintering over normal sintering, we have investigated the effect of an external electric field on the viscosity of glass. The results show remarkable electric field-induced softening (EFIS), as application of DC field significantly lowers the softening temperature of glass. To establish the origin of EFIS, the effect is compared for single vs. mixed-alkali silicate glasses with fixed mole percentage of the alkali ions such that the mobility of alkali ions is greatly reduced while the basic network structure does not change much. The sodium silicate and lithium-sodium mixed alkali silicate glasses were tested mechanically in situ under compression in external electric field ranging from 0 to 250 V/cm in specially designed equipment. A comparison of data for different compositions indicates a complex mechanical response, which is observed as field-induced viscous flow due to a combination of Joule heating, electrolysis and dielectric breakdown.

Electric fields applied with a pair of electrodes can reduce the furnace temperature and processing time of ceramics.1–4 This effect falls into two regimes, field assisted sintering (FAST) and flash sintering. In FAST, which occurs at low fields, sintering is somewhat enhanced but progresses gradually with time.1 By comparison, at higher fields flash sintering occurs abruptly in just a few seconds, when a critical temperature is reached at a given applied field.1 It is characterized by a power surge produced by an abrupt increase in electrical conductivity.

The mechanism of flash sintering is controversial. The sudden increase in conductivity produces Joule heating, which has been the first explanation of the flash effect. A recent study5 using dynamic modeling with non-uniform temperature argues against the synergistic effect and supports Joule heating runaway to be responsible for enhancing the sintering rates. A detailed study4 suggests that Joule heating alone may not be able to explain flash sintering.

Electrical fields are often used in glass processing but the mechanism is not well understood. For example, electro-thermal poling is used to enhance biological, physical, chemical properties, and nonlinear optical susceptibility.6–15 The process involves a glass sample sandwiched between ion-blocking electrodes with a DC potential. It is then heated to some specified temperature below the glass transition temperature (Tg) to allow for significant ionic conductivity.7 The glass is then cooled down to ambient under the DC potential to freeze ionic displacements. Surface patterns can be imprinted using the DC electric fields at and below the Tg.16–18 It is possible that local softening of glass occurs during electric imprinting. A reduction in furnace temperature for bulk glass softening has potential for increasing the efficiency of glass processing.

Here, we report electric field-induced softening (EFIS) phenomenon in glass. The results show that glass softens at furnace temperatures well below Tg under the DC electric fields. The softening is accompanied by optical emission.

Three disilicate glass compositions were selected for this study: a single alkali sodium silicate (NS) and two lithium-sodium mixed alkali silicates (2L8NS and 5L5NS). Their compositions are listed in Table I. Owing to the well-known mixed alkali effect, these glasses have similar network structure but different DC electrical resistivity.19 All glasses were prepared as previously reported.20 For the EFIS measurements, rectangular blocks were cut and polished to a final cross-section of 5 mm × 5 mm and a height of about 10 mm. Carbon paste was applied to the top and bottom of the sample to ensure good electrical contacts with the graphite electrodes.

TABLE I.

Glass compositions and their Tg and Ts0 values.

Glass typeCompositionTg (°C)Ts0 (°C)
NS 0.3Na2O • 0.7SiO2 475 550 
2L8NS 0.33[0.2 Li2O • 0.8Na2O] • 0.67SiO2 427 491 
5L5NS 0.33[0.5 Li2O • 0.5Na2O] • 0.67SiO2 418 793 
Glass typeCompositionTg (°C)Ts0 (°C)
NS 0.3Na2O • 0.7SiO2 475 550 
2L8NS 0.33[0.2 Li2O • 0.8Na2O] • 0.67SiO2 427 491 
5L5NS 0.33[0.5 Li2O • 0.5Na2O] • 0.67SiO2 418 793 

Differential scanning calorimetry (DSC) using a NETZSCH 404/3F microcalorimeter measured Tg. Measurements were done from ambient temperature to 800 °C at 10 °C/min. The values for Tg are reported in Table I.

A pneumatic creep tester (Applied Test Systems model 2605) was modified to mechanically test the glass samples under the DC electrical fields. A schematic of the experimental setup is given in Fig. 1. Optical emission was recorded with Ocean Optics USB4000 UV-VIS-ES spectrometer with an optical resolution of 1.5 nm.

FIG. 1.

Experimental setup inside the modified ATS model 2605 pneumatic creep tester where the anode was located at the top of the sample and the cathode at the bottom. The system was electrically insulated from the rest of the furnace.

FIG. 1.

Experimental setup inside the modified ATS model 2605 pneumatic creep tester where the anode was located at the top of the sample and the cathode at the bottom. The system was electrically insulated from the rest of the furnace.

Close modal

EFIS experiments were carried out under compressive load of 10 MPa. The displacement was measured with an Omega LD621–15 linear variable differential transformer (LVDT) gauge. Electric field-induced softening was measured by the onset of displacement in the push rod.

A constant heating rate of 10 °C/min was used for all samples. The electric field varying from 0 to 250 V/cm was applied with a power supply (Harrison Laboratories model 890A). A resistor of 250 Ω was inserted in series with the sample to limit the current in the circuit. The voltage and current were recorded using a DATAQ Instruments model DI-149HS. Voltage was measured across the sample. Furnace temperature was determined with a thermocouple placed next to the sample.

Electrical impedance was measured following a previously reported procedure.20 Complex impedance analysis yielded DC conductivity as a function of temperature.21 

Energy-dispersive X-ray spectroscopy (EDS) was used for chemical analysis using scanning electron microscopy (SEM) (Hitachi 4300SE/N).

Representative behavior of glass softening is shown in Fig. 2 for the NS glass, in terms of the displacement (shown as negative values since the experiments were conducted in compression) as a function of temperature that was increased at a constant heating rate. Results from several samples each tested at a certain applied field are reported. The fields were varied from 0 to 250 V/cm. Thermal expansion of the loading structure is responsible for the slight displacement before the onset of viscous flow. We define the softening temperature (TsV) by the onset of large deformation under applied field, V. Note that the temperature for the onset of viscous flow decreases as the electrical field is increased.

FIG. 2.

Displacement vs. temperature of NS at various applied electric fields at a heating rate of 10 °C/min. NS Tg is indicated by arrow. Reference alumina rod displacement of loading structure is shown by light green dashes. Transition between stage I and stage II is depicted by vertical line.

FIG. 2.

Displacement vs. temperature of NS at various applied electric fields at a heating rate of 10 °C/min. NS Tg is indicated by arrow. Reference alumina rod displacement of loading structure is shown by light green dashes. Transition between stage I and stage II is depicted by vertical line.

Close modal

The curves for 0, 50, and 100 V/cm fields in Fig. 2 show compressive displacement that occurs gradually. But at fields greater than 125 V/cm viscous flow occurs abruptly. (A video demonstration is provided as supplementary material). Thus, we observe two regimes of the EFIS behavior. At low electric fields (≤100 V/cm in Fig. 2, stage I), the viscosity decreases gradually, in a manner that is similar to the reference sample at 0 V. However, at high fields (≥125 V/cm, stage II), NS samples flow quickly, within a few seconds, at furnace temperatures that are well below Tg (see Fig. 3 below for details). This behavior is similar to the observations in flash sintering experiments with ceramics.1 

FIG. 3.

Normalized EFIS effect for NS (black, squares), 2L8NS (blue, triangles), and 5L5NS (pink, inverted triangle). The Difference ΔT(V) is plotted versus applied electric field where ΔT(V) = (Ts0 − TsV). Arrows indicate approximate threshold applied electric field for each glass composition.

FIG. 3.

Normalized EFIS effect for NS (black, squares), 2L8NS (blue, triangles), and 5L5NS (pink, inverted triangle). The Difference ΔT(V) is plotted versus applied electric field where ΔT(V) = (Ts0 − TsV). Arrows indicate approximate threshold applied electric field for each glass composition.

Close modal

The effect of glass composition on the EFIS is analyzed by normalizing the data with respect to zero-field softening temperature, Ts0, as listed in Table I. The normalized EFIS effect is quantified with parameter ΔT(V) = (Ts0 − TsV). These results, shown in Fig. 3, delineate two regimes of behavior (stage I and II) for all glasses, similar to FAST and flash sintering in ceramics. At low electric fields, in stage I, ΔT(V) is small and increases slightly with field. At higher fields, typically above a threshold value, ΔT(V) begins to increase rapidly, marking a transition from stage I to stage II. For NS, 2L8NS, and 5L5NS, the threshold fields are approximately 95, 100, and 175 V/cm, respectively. Above the threshold, the value of ΔT(V) is significantly higher for the binary NS glass than for the mixed alkali glasses. This difference in the EFIS is attributed to the higher resistivity of or correspondingly lower current flow through the mixed alkali glasses than the single alkali NS glass. A higher resistivity of the former reduces Joule heating, which appears to be responsible for the lower values of ΔT(V). However, it is not clear why ΔT(V) for the NS glass increases continually with applied field, whereas it reaches a plateau at 175 V/cm for the mixed alkali glasses. Further investigation of this plateau will include testing NS at higher electric fields and testing mixed alkali glasses at lower heating rates to allow longer times for charge flow.

The relationship between the current flowing through the sample and viscous flow is shown in Fig. 4 using the example of the NS glass at 175 V/cm. Before softening, the current rises steadily with temperature as expected from increased ionic conductivity. At the onset of EFIS, the current rises sharply. The high conductivity of the sample is evident from the equally sharp drop in voltage. This behavior is similar to the observation of flash sintering where the power supply was switched from voltage to current control when the current rises abruptly.2 In the present experiments, the current was limited by the 250 Ω resistor placed in series with the sample.

FIG. 4.

Time dependence of current (blue, solid), voltage (red, solid), and displacement (black, dashed) for NS with 175 V/cm field at a heating rate of 10 °C/min. Note: Current amplifier became saturated during high current regime.

FIG. 4.

Time dependence of current (blue, solid), voltage (red, solid), and displacement (black, dashed) for NS with 175 V/cm field at a heating rate of 10 °C/min. Note: Current amplifier became saturated during high current regime.

Close modal

Measurements of optical emission from the specimen correlated with the current spikes. They are believed to be related to an electron avalanche breakdown process.22 A finite current passed through the sample prior to EFIS. It has been reported6–8,10 that this current is due to ionic transport of alkali ions, which forms a depletion layer in the glass near the anode. As the depletion layer forms, a greater amount of the applied electric field drops across this layer. It occurs due to the increasing removal of alkali ions causing the depletion layer resistance to increase. This results in an increase of the local electric field near the anode which can reach approximately to the order of 107 V/cm, which is close to the dielectric strength of silicate glasses.7 This process has recently been outlined by Zakel et al.,7 for a bioactive glass.

In industry, Joule heating is used in melting and fining of glass.23,24 This technique depends on glass resistivity and its temperature dependence. So we ask if EFIS is simply due to Joule heating from DC ionic conductivity. DC resistivity was determined from AC complex impedance analysis with a value of 3.88 × 108 Ω cm for NS, 1.19 × 1011 Ω cm for the 2L8NS, and 9.49 × 1011 Ω cm for 5L5NS at room temperature. The corresponding activation energies for NS, 2L8NS, and 5L5NS were 0.69, 0.92, and 0.97 eV, respectively, which are in good agreement with the literature values.25 

Fumes were observed arising from the anode region in all glass compositions above the threshold fields, indicating very high local temperature. They were accompanied by buildup of white powder on the push rods. Beyond this stage, the sample was dramatically modified, but data prior to this stage represented reproducible EFIS phenomenon. The EDS analysis showed sodium accumulation at the cathode, as to be expected from electrolysis. The powder from 2L8NS at 200 V/cm was analyzed by EDS as well; its composition was consistent with glass that had condensed on the surfaces during cooling. This is an unexpected observation since these glass compositions should not vaporize at these furnace temperatures. It indicates additional mechanisms of heating besides Joule heating from normal ionic conductivity of glass. Evidently, temperatures significantly higher than the furnace were reached in the depletion layer next to the electrodes,6,7 since its resistance should be much greater than the remaining bulk of the sample.

The power delivered to the sample during the high current regime was enough to vaporize glass. An estimate of sample temperature (10 W peak over 50 s) could raise the temperature by approximately 500 °C, if we assume uniform Joule heating and neglect heat loss due to thermal conduction during this brief period. An analysis of structural changes across the sample would be needed to delineate the relative contributions of the EFIS mechanisms, which may include non-uniform Joule heating, dielectric breakdown, and electrolysis. Here non-uniform Joule heating would result with localization of heat to the depletion layer due to its relatively high resistance. In this region, the actual temperature rise could be high enough to account for melting and even vaporization. Post-mortem inspection of samples showed that most of the deformation occurred around the anode side. Thus heat dissipation from the depletion layer into the bulk may be sufficient for the softening of the remaining sample. Another possible mechanism of EFIS is that the high fields within the depletion layer ionize the atoms producing an avalanche. Charge injection and consequent high current through the whole sample may ensue, causing more or less uniform softening. Furthermore, the migration of sodium towards the cathode is expected to break up the glass network by converting bridging oxygen into non-bridging oxygen, as observed in a long duration XPS experiment.26 This modification of electric field-induced structural change will reduce the softening temperature of the region away from the anode, thereby inducing premature softening relative to initial glass composition.

Just prior to and during EFIS each glass composition produced photoemission at fields greater than 50 V/cm. Fig. 5(a) (See Ref. 27 for a multimedia visual demonstration) shows the 5L5NS sample under 150 V/cm field at a furnace temperature before photoemission. When the temperature was increased above the flashing temperature (TF), the first photoemission and accompanying current spike were observed as seen in Fig. 5(b) for a furnace temperature TF < T < TS. Fig. 5(b) visually captured the photoemission near the anode. When the recorded data were synchronized with the video, it was observed that the glass samples would flash during the current spikes and sustained high current shown in Fig. 4. This marks high conductance of the material which suggests dielectric breakdown of the glass presumably across the depletion layer via electron avalanche. At the same time, there would have to be electronic conductivity throughout the sample. It has been observed that NS flashed for a longer time frame (2–5 min) before EFIS, while 2L8NS and 5L5NS flashed for about a minute or less before EFIS.

FIG. 5.

Images of 5L5NS during 150 V/cm when the furnace temperature was (a) T < TF and (b) TF < T < TS of EFIS. The anode is at the top. (See Ref. 27 for a multimedia visual demonstration).

FIG. 5.

Images of 5L5NS during 150 V/cm when the furnace temperature was (a) T < TF and (b) TF < T < TS of EFIS. The anode is at the top. (See Ref. 27 for a multimedia visual demonstration).

Close modal

The photoemission spectra for NS and 2L8NS are shown in Fig. 6. The photoemission spectra for the two compositions are similar, with a broad background in the visible and near-infrared region and several sharp characteristic peaks. All samples show an intense emission peak at 589 nm. In the presence of lithium, in 2L8NS, two additional peaks at 611 and 671 nm are observed. The peak energies match very well with the electron energy level transitions for the alkali ions, as in Grotrian diagrams provided by NIST.28 The sharp peaks suggest photoemission from gaseous species as opposed to electroluminescence reported for stage III flash sintering of yttria stabilized zirconia as exciton recombination.29 A comparison of the photoemission peaks is presented in Table SI27 for NS and 2L8NS at 200 V/cm. An impurity peak located at 767 nm is identified using EDS due to potassium that is present as an impurity similar to that has been seen in the spectra of high pressure sodium vapor lamps.30–32 The observed photoemission peaks follow a behavior similar to breakdown conduction of Al-SiO-Al, Al-SiO-Ni, and Al-MgF2-Al capacitors as well.22 

FIG. 6.

Photoemission spectra at the 200 V/cm test condition ranging from 350 to 900 nm for NS (black, bottom) and 2L8NS (blue, top). 5L5NS has the same peaks as 2L8NS. Note: Intensity of 2L8NS was offset by an arbitrary amount for comparison.

FIG. 6.

Photoemission spectra at the 200 V/cm test condition ranging from 350 to 900 nm for NS (black, bottom) and 2L8NS (blue, top). 5L5NS has the same peaks as 2L8NS. Note: Intensity of 2L8NS was offset by an arbitrary amount for comparison.

Close modal

The large broad background in the optical emission has been identified as bremsstrahlung radiation. This indicates electrons are undergoing deceleration in the Coulombic field of the atoms.33 The bremsstrahlung radiation provides a short-wavelength limit, which is the maximum energy of a single collision event within the sample. The highest energy observed can be estimated by the short-wavelength limit as shown by Goldstein.33 This limit for the spectra shown in Fig. 6 for 200 V/cm is approximately at 460 nm, which corresponds to a maximum energy value of an electron at about 2.70 eV. This value is significantly less than the energy band gap of 5.8 eV (Ref. 34) for the NS glass, ruling out direct electron-hole recombination as the cause of photoemission.

Similar to electro-thermal poling, electrons are injected into the conduction band of the glass from the cathode towards the anode for charge compensation. These electrons would then migrate towards the anode, and get accelerated upon reaching the depletion layer due to the presence of an extremely high potential drop. It is conceivable that electrons such as at the non-bridging oxygen atoms are also injected from the region next to the depletion layer. At this field strength, which can be close to the critical field strength, impact ionization may result from electron avalanche across the depletions layer. Since bremsstrahlung radiation is indicative of electron-generated radiation, the accelerated electrons are presumed to collide with alkali ions migrating towards the cathode. These interactions can also excite and relax discrete energy levels within alkali ions, creating characteristic photoemission peaks.

EFIS of glass has been identified. The application of electric field produces deformation at furnace temperature that is significantly lower than the glass transition temperature. A higher field continues to reduce the furnace temperature for the onset of softening. The EFIS effect is stronger for the single than for the mixed alkali silicate glasses, suggesting a role for ionic mobility in this phenomenon. The results suggest the following mechanism: The application of the DC fields during heating forms an alkali ion depletion layer near the anode. Joule heating within the sample becomes significant at higher temperatures as ionic conductivity increases overcoming heat dissipation. An electron avalanche occurs over the depletion layer resulting in photoemission comprising of bremsstrahlung radiation and characteristic alkali ion electron energy level peaks. These processes create a positive feedback system of Joule heating, charge injection, space charge formation, and electrolysis resulting in glass softening, melting, and ultimately vaporization under relatively moderate overall electric field across the sample.

This work has been supported by the National Science Foundation through the International Materials Institute for New Functionality in Glass (DMR 0844014) at Lehigh University. The authors thank Christie Hasbrouck for her help with the experiments. R.T. and R.R. were supported by a grant from the Basic Energy Sciences Division of the Department of Energy under Grant No. DE-FG02-07ER46403.

1.
M.
Cologna
,
B.
Rashkova
, and
R.
Raj
,
J. Am. Ceram. Soc.
93
,
3556
(
2010
).
2.
K. S.
Naik
,
V. M.
Sglavo
, and
R.
Raj
,
J. Eur. Ceram. Soc.
34
,
2435
(
2014
).
3.
M.
Cologna
,
J. S. C.
Francis
, and
R.
Raj
,
J. Eur. Ceram. Soc.
31
,
2827
(
2011
).
4.
R.
Raj
,
J. Eur. Ceram. Soc.
32
,
2293
(
2012
).
5.
R. I.
Todd
,
E.
Zapata-Solvas
,
R. S.
Bonilla
,
T.
Sneddon
, and
P. R.
Wilshaw
,
J. Eur. Ceram. Soc.
35
,
1865
(
2015
).
6.
C. R.
Mariappan
and
B.
Roling
,
J. Non-Cryst. Solids
356
,
720
(
2010
).
7.
J.
Zakel
,
M.
Balabajew
, and
B.
Roling
,
Solid State Ionics
265
,
1
(
2014
).
8.
E. C.
Ziemath
,
V. D.
Arajo
, and
C. A.
Escanhoela
,
J. Appl. Phys.
104
,
054912
(
2008
).
9.
A. L.
Moura
,
M. T.
de Arajo
,
E. A.
Gouveia
, and
M. V. D.
Vermelho
,
Opt. Express
15
,
143
(
2007
).
10.
C. M.
Lepienski
,
J. A.
Giacometti
,
G. F.
Leal Ferreira
,
F. L.
Freire
, Jr.
, and
C. A.
Achete
,
J. Non-Cryst. Solids
159
,
204
(
1993
).
11.
J.
Zakel
,
V.
Heddinga
,
S. O.
Steinmller
, and
B.
Roling
,
Solid State Ionics
237
,
46
(
2013
).
12.
C. R.
Mariappan
and
B.
Roling
,
Solid State Ionics
179
,
671
(
2008
).
13.
J.
Kruempelmann
,
C. R.
Mariappan
, and
B.
Roling
,
Phys. Rev. B
82
,
224203
(
2010
).
14.
U. K.
Krieger
and
W. A.
Lanford
,
J. Non-Cryst. Solids
102
,
50
(
1988
).
15.
A. L. R.
Brennand
, Doctoral dissertation, University of Southampton,
2002
.
16.
N.
Ikutame
,
K.
Kawaguchi
,
H.
Ikeda
,
D.
Sakai
,
K.
Harada
,
S.
Funatsu
, and
J.
Nishii
,
J. Appl. Phys.
114
,
083514
(
2013
).
17.
S.
Youn
,
H.
Takagi
,
S.
Park
,
M.
Takahashi
, and
R.
Maeda
,
J. Nanosci. Nanotechnol.
12
,
3181
(
2012
).
18.
A.
Lipovskii
,
M.
Kuittinen
,
P.
Karvinen
,
K.
Leinonen
,
V.
Melehin
,
V.
Zhurikhina
, and
Y.
Svirko
,
Nanotechnology
19
,
415304
(
2008
).
19.
D. E.
Day
,
J. Non-Cryst. Solids
21
,
343
(
1976
).
20.
C.
Saiyasombat
and
H.
Jain
,
J. Electroceram.
34
,
57
(
2015
).
21.
H.
Jain
, “
Measurement of electrical conductivity of glasses
,” in
Experimental Techniques of Glass Science
, edited by
C. J.
Simmons
and
O. H.
El-Bayoumi
(
American Ceramic Society
,
1993
), pp.
433
461
.
22.
P. P.
Budenstein
,
IEEE Trans. Electr. Insul.
EI-15
,
225
(
1980
).
23.
C. F.
De Voe
, U.S. patent 2,490,339 (Dec. 6,
1949
).
24.
J.
Stanek
and
J.
Matej
,
J. Non-Cryst. Solids
84
,
353
(
1986
).
25.
N. P.
Bansal
and
R. H.
Doremus
,
Handbook of Glass Properties
(
Academic Press, Inc.
,
Orlando
,
1986
), pp.
381
449
.
26.
A.
Sharma
,
H.
Jain
, and
A. C.
Miller
,
Surf. Interface Anal.
31
,
369
(
2001
).
27.
See supplementary material at http://dx.doi.org/10.1063/1.4934945 for Table II describing photoemission peak identifications and a multimedia visual demonstration of EFIS on NS with 150 V/cm.
28.
C. E.
Moore
and
P. W.
Merrill
, Partial Grotrian Diagrams of Astrophysical Interest, NSRDS Vol. 23 (National Bureau of Standards, Washington, D.C., 1968), available at http://www.nist.gov/data/nsrds/NSRDS-NBS-23.pdf.
29.
K.
Terauds
,
J.-M.
Lebrun
,
H. H.
Lee
,
T. Y.
Jeon
,
S. H.
Lee
,
H. J.
Je
, and
R.
Raj
,
J. Eur. Ceram. Soc.
35
,
3195
(
2015
).
30.
J. J.
de Groot
and
J. A. J. M.
van Vliet
,
The High-Pressure Sodium Lamp
, 1st ed. (
MacMillan Education LTD
,
London
,
1986
), p.
23
.
31.
M.
Raki
and
G.
Pichler
,
Opt. Commun.
284
,
2881
(
2011
).
32.
C. D.
Elvidge
,
D. M.
Keith
,
B. T.
Tuttle
, and
K. E.
Baugh
,
Sensors
10
,
3961
(
2010
).
33.
J.
Goldstein
,
D.
Newbury
,
D.
Joy
,
C.
Lyman
,
P.
Echlin
,
E.
Lifshin
,
L.
Sawyer
, and
J.
Michael
,
Scanning Electron Microscopy and X-Ray Microanalysis
, 3rd ed. (
Springer
,
New York
,
2007
), pp.
271
296
.
34.
G. H.
Sigel
,
Jr., “Optical absorption of glasses
,” in
Glass I: Interaction with Electromagnetic Radiation: Treatise on Materials Science and Technology
, edited by
M.
Tomozawa
and
R. H.
Doremus
(
Elsevier
,
2013
), Vol.
12
, pp.
5
82
.

Supplementary Material