ARTICLE IN PRESS

0022-2313/$ - se

doi:10.1016/j.jlu

�Correspondi+41 31 631 4399

E-mail addre

Journal of Luminescence 114 (2005) 53–59

www.elsevier.com/locate/jlumin

Highly efficient near-infrared to visible up-conversion processin NaYF4:Er

3þ;Yb3þ

J.F. Suyver�, J. Grimm, K.W. Kramer, H.U. Gudel

Department of Chemistry and Biochemistry, University of Bern, Freiestrasse 3, 3000 Bern 9, Switzerland

Received 3 September 2004

Available online 5 January 2005

Abstract

Up-conversion (UC) entails the addition of two (or more) photons to generate one of higher energy. This process is

very interesting, both from a fundamental point of view as well as for (future) applications. Here, dependence of the

emission spectra on excitation power and measurement temperature are reported for the very efficient near-infrared to

visible photon UC material NaYF4: 2% Er3þ; 18% Yb3þ: In the high-power limit for continuous-wave laser excitation,roughly 50% of all near-infrared photons absorbed by the material are up-converted and emitted in the visible spectral

range. The excitation spectrum shows a 39 cm�1 energy gap between the two lowest crystal-field components j0i and j1i

of the Yb3þ 2F5=2 multiplet, which sensitizes the UC emission. There is very efficient energy transfer to the lowest

energy crystal-field component of the Er3þ 4I11=2 state from 2F5=2j1i; resulting in an activation energy for all Er3þ

related up- and down-conversion emissions that scales with the number of excitation photons required multiplied by the

39 cm�1 energy gap.

r 2004 Elsevier B.V. All rights reserved.

PACS: 42.65.Ky; 42.70.Nq; 78.20.�e; 78.55.�m; 78.55.Fv

Keywords: Up-conversion; Spectroscopy; NaYF4:Er; Efficiency

1. Introduction

Fascinating and novel phenomena are attribu-ted to photon up-conversion (UC) in inorganicmaterials [1]. Applications such as lasers [2], next

e front matter r 2004 Elsevier B.V. All rights reserve

min.2004.11.012

ng author. Tel.: +4131 631 4254; fax:

.

ss: suyver@iac.unibe.ch (J.F. Suyver).

generation imaging devices [3] and near-infrared(NIR) quantum counting devices [4] have beendemonstrated. As high-power NIR lasers areaffordable and easily obtainable, efficient NIR tovisible UC is expected to have significant techno-logical impact. Besides future applications, a greatinterest also exists in the fundamental physics ofefficient UC materials due to the interestinginterplay between linear and non-linear optical

d.

ARTICLE IN PRESS

J.F. Suyver et al. / Journal of Luminescence 114 (2005) 53–5954

excitation processes, resulting in complex depen-dencies on the excitation power and non-trivialluminescence transients [5]. Studying these energytransfer processes will further support the under-standing of UC mechanisms in general and canlead to better tailored UC materials.Since the early 1970s, NaYF4:Er

3þ;Yb3þ isknown as one of the most efficient NIR to visibleUC materials [6–10], but no systematic study ofthe physical properties of this material has beenundertaken as of yet. The most important omis-sions in the literature are the total spectraldistribution (including all NIR emissions) as wellas the power dependence and the temperaturedependence of the emission bands under NIRexcitation. So far, these have not been investigatedin detail for any NIR to visible UC material. Inthis paper, a study of these properties ofNaYF4:Er

3þ;Yb3þ will be presented as well as aproof of the highly efficient UC mechanisminvolved. From the experimental data a determi-nation of the absolute efficiency as well as thefundamental (highly efficient) excitation mechan-ism for near-infrared to visible UC inNaYF4:Er

3þ;Yb3þ is presented. Our investigationproves that this material is a very efficient NIR tovisible up-converter due to the near-perfect reso-nance between a specific crystal-field level ofYb3þ and one of Er3þ: As a result, the UC cantake place through direct (resonant) energy trans-fer from Yb3þ to Er3þ and no phonon-mediatedstep is required. These results are pivotal forunderstanding the properties of this material aswell as for its application in a device. From ourdata, it is also clear that the red to green emissionratio changes with excitation power, which will beimportant when functionalizing this material in adevice. This change in red to green emission ratiois indicative of a change in excitation mechanism,which in itself is interesting from a fundamentalpoint of view.

2. Experimental

Our group recently published the optimizedsynthesis of high-quality NaYF4:Er

3þ;Yb3þ pow-der [11]. This study indicated that the material

should be in the hexagonal phase with Er3þ andYb3þ concentrations of 2% and 18%, respectively,to obtain the highest UC intensity. The routedescribed in Ref. [11] was used to fabricate thematerial used in this study.The powders were excited with a multimode,

standing wave Ti:sapphire laser (Spectra Physics3900S), pumped by an argon-ion laser in all-linesmode (Spectra Physics 204515/4S) or the secondharmonic of a Nd:YVO4 laser (Spectra PhysicsMillennia CS-FRU). The wavelength control ofthe Ti:sapphire laser was achieved by an in-chworm-driven (Burleigh PZ-501) birefringentfilter and a wavemeter (Burleigh WA2100). Samplecooling was achieved using a quartz helium gasflow tube. The sample luminescence was dispersedby a 0.85m double monochromator (Spex 1402)with gratings blazed at 500 nm (1200 grooves/mm).The signal was detected with a cooled photomul-tiplier tube (Hamamatsu R3310) and a photoncounting system (Stanford Research SR400).Device control and data acquisition were doneby a personal computer. The luminescence spectrawere corrected for the instrument response and aredisplayed as a photon flux per constant energyinterval [12]. To measure the power dependence,the beam was attenuated with a series of neutraldensity filters (Balzers). The luminescence back-ground of the (unfocussed) Ti:sapphire laser wasremoved using a double prism.

3. Results and discussion

Fig. 1 shows room temperature emission spectraof NaYF4:Er

3þ;Yb3þ for three different excitationpowers, chosen to be in the low-, intermediate- andhigh-power regimes, respectively. For excitationpowers \400mW no further change in relativeemission intensities is observed [13]. For eachspectrum, the area under emission band i corre-sponds to the number of photons emitted in thatband [12], and is denoted by Fi: As a result, f i �

Fi=P

8j Fj represents the fraction of all photonsemitted in band i. The assignment of the electronictransition responsible for each band is given inTable 1 together with the fraction of all photonsemitted in each of the six emission bands, for the

ARTICLE IN PRESSPh

oton

flu

x (c

ount

s/s)

Emission energy (103cm-1)

x 25 (a)

(b)

(c)

10 15 20 2560

40

80

120

0

20

40

30

0

2

4

6

10

21

3

45

6

Fig. 1. Emission spectra of NaYF4:Er3þ;Yb3þ excited at

10 238 cm�1 (indicated by the arrow, laser-line removed

manually) with (a) 10, (b) 100, (c) 600mW. Recorded at room

temperature and relative intensities can be compared. Relative

photon fractions and assignment of all emissions, as labeled in

Fig. 1(c), are shown in Table 1.

J.F. Suyver et al. / Journal of Luminescence 114 (2005) 53–59 55

different excitation powers used. Due to theoverlap of two emissions around 10 000 cm�1 andtwo around 18 200 cm�1; no attempt is made toresolve the separate emissions and only one valueis shown for the total band. Further emissionmeasurements have been done in the spectralregion down to 3300 and up to 35 000 cm�1; butno significant emissions were observed outside theregion shown in Fig. 1.The excitation spectrum of the 18 200 cm�1 UC

emission, shown in Fig. 2(a), was recorded at 5K

to minimize temperature-induced broadening.This spectrum is typical for Yb3þ 2F7=2 !

2F5=2

absorption. The excitation spectra of the otheremission bands are similar. At increased tempera-ture the lines broaden, but the Yb3þ character-istics remain pronounced. From Fig. 2(a) the twolowest energy crystal field components within the2F5=2 multiplet are identified: 10 199 and10 238 cm�1; denoted as 2F5=2j0i and 2F5=2j1i;respectively. The broader excitation bands athigher energy are attributed to 2F5=2j2i andphonon-assisted sidebands. Fig. 2(b) shows theexcitation spectrum of the same emission (re-corded at 12K) in a NaYF4:10% Er3þ sample.The lines correspond to electronic and vibronicexcitations of the Er3þ 4I11=2 multiplet. The lowestcrystal field level of this multiplet, denoted by4I11=2j0i; is at 10 244 cm�1: Note that this is athigher energy than the 2F5=2j0i level of Yb

3þ; butin near-perfect resonance with 2F5=2j1i: At roomtemperature the excitation lines are homoge-neously broadened and there is optimal spectraloverlap for 2F5=2j1i !

4I11=2j0i energy transfer.From comparing Figs. 2(a) (circled part) and (b) itis evident that even at 5K more than 95% of theUC emission in the co-doped sample isYb3þ sensitized.Let pi denote the number of NIR excitation

photons required to induce emission in band i

(the values of pi are indicated in Table 1) and Fi

the number of photons emitted in that band.Then T �

P8j pjFj is the (minimum) number of

NIR excitations required to induce the emissionspectrum. It was found that T is independentof temperature when corrected for the tempera-ture-dependence of the Yb3þ excitation cross-section [14]. This implies that between 5 and300K no significant non-radiative multiphononrelaxation to the Er3þ or Yb3þ ground stateoccurs, because such non-radiative processeswould depend on the sample temperature.Furthermore, the luminescence lifetimes of theYb3þ 2F5=2 and Er3þ 4I13=2 states were mea-sured under direct excitation and both werefound to be in the millisecond-range andeffectively independent of the sample tem-perature. Therefore, each photon absorbed mustcontribute to the emission of a photon and this

ARTICLE IN PRESS

Table 1

Fraction of all photons emitted in each band (denoted by f i) as well as the fraction of excitation photons absorbed that are

subsequently used to excite each of the emission bands (denoted by Ri and defined in Eq. (1)) for the six significant emission bands in

NaYF4:Er3þ;Yb3þ under 10 238 cm�1 excitation at room temperature (data shown in Fig. 1). The different excitation powers are

indicated. Also shown are the assignments of the emission bands and the (minimum) number of excitation photons needed to excite

these emission bands (denoted by pi)

i Energy of emission

band maximum ðcm�1Þ

Electronic transition

involved

pi f i Ri

10mW 100mW 600mW 10mW 100mW 600mW

1 6500 Er3þ:4I13=2 !4I15=2 1 3.0% 13% 18% 2.9% 10.8% 13.4%

2 10 000 Yb3þ: 2F5=2 !2F7=2

Er3þ: 4I11=2 !4I15=2

8<:

1 93% 68% 48% 90.3% 56.8% 35.9%

3 11 800 Er3þ:4S3=2 !4I13=2 2 0.6% 1.7% 2.1% 1.2% 2.8% 3.1%

4 15 000 Er3þ:4F9=2 !4I15=2 2 0.5% 7.9% 16% 1.0% 13.2% 24%

5 18 200

18 800

�Er3þ: 4S3=2 !

4I15=2

Er3þ: 2H11=2 !4I15=2

2 2.6% 9.5% 15% 4.9% 15.8% 22.5%

6 24 000 Er3þ:2H9=2 !4I15=2 3 0.01% 0.2% 0.5% 0.03% 0.50% 1.1%

Phot

on f

lux

(arb

.uni

ts)

Excitation energy (cm-1)

(b)

(a)

4 I 11/

2|0⟩

2 F 5/2

|0⟩

2F5/

2 |1⟩

Direct 4I11/2excitation

10200 10300 10400 10500

Fig. 2. Excitation spectrum of the 18 200 cm�1 emission in (a)

NaYF4: 2% Er3þ; 18% Yb3þ and (b) NaYF4:10% Er3þ: Thespectra were recorded at (a) 5K and (b) 12K. The vertical

scales in Figs. 2(a) and (b) are not comparable. The arrow

indicates the excitation energy used for obtaining the data

shown in Fig. 3.

J.F. Suyver et al. / Journal of Luminescence 114 (2005) 53–5956

implies that

Ri � piFi

X8j

pjFj

,(1)

will be the fraction of absorbed infra-red pho-tons emitted in band i. The values for Ri at 300Kare shown in Table 1 for different excitationpowers.Note that one important assumption has been

made in Eq. (1): it is assumed that intra-excitedstate emissions do not present a significantcontribution to the total emission spectrum.Specifically, it is assumed that the 4I13=2 state ismainly populated through a one-photon processrather than through 4S3=2 !

4I13=2 emission. Sincethe photon flux related to this intra-excited statetransition represents only a small fraction of thephoton flux related to the 4I13=2 !

4I15=2 and4S3=2 !

4I15=2 transitions (see the values of therespective f i as shown in Table 1), this assumptionis expected to be valid.Under this assumption a justified assessment

regarding the true efficiency of the NaYF4:Er3þ;

Yb3þ sample is made from the data shown inTable 1: in the high-power limit (i.e. at unfocussed

ARTICLE IN PRESS

J.F. Suyver et al. / Journal of Luminescence 114 (2005) 53–59 57

excitation powers of\400mW) nearly 50% of theNIR photons absorbed are emitted in the threeemission bands in the visible (i ¼ 4; 5; 6), clearlydemonstrating the very high UC efficiency of thismaterial. Furthermore, the data in Table 1 andFig. 1 also clearly show that the ratio of redð15 000 cm�1Þ to green ð18 200 cm�1Þ emission isdependent on the excitation power. Thus theoutput of a device based on NaYF4:Er

3þ;Yb3þ can be tuned to meet specific color-demandswhen the saturation-limit has not been reached(excitation power t400mW).Fig. 3 shows the temperature dependence of

four integrated emission bands when the sample isexcited at 10 199 cm�1: Identical results wereobtained under 10 238 cm�1 excitation. The resultsfor the 11 800 and 15 000 cm�1 emission bandsare similar to those shown in Fig. 3(c). Withthe exception of the 6500 cm�1 band, as shown in

(c)

(a)

0 50 100 150 200Temperature (K)

Phot

on f

lux

(cou

nts/

s)

p5 = 2

p1 = 2

p1 = 1105

104

103

102

101

105

104

103

102

101

Fig. 3. Temperature dependence of the integrated photon flux of the e

24 000 cm�1; respectively. Excitation was at 10 199 cm�1 (indicated by

Figs. 3(a)–(d) cannot be compared. Note the logarithmic vertical axes.

limit (� 500mW:&) are shown. The lines through the data are predic

with parameter values DE ¼ 39 cm�1 and pi as indicated.

Fig. 3(a), no effect of the excitation power wasobserved.Fig. 4 shows the relevant energy levels of

Er3þ and Yb3þ as well as some of the mostimportant energy transfer and light emissionprocesses. Indicated by the sets of two dashedarrows connected with a dotted line are the energytransfer up-conversion (ETU) processes that occurin this material: one Yb3þ ion transfers back to theground state, while simultaneously the Er3þ iontransfers to a state of higher energy. This ETUprocess can occur several times in succession,resulting in an Er3þ ion in a high excited state,such as the 2H9=2 state at 24 000 cm�1 (emissionband 6 in Fig. 1), which is reached after threeenergy transfer steps.The Yb3þ 2F5=2j0i and 2F5=2j1i states are in

thermal equilibrium immediately after the excita-tion. As the 2F5=2j0i state is at too low energy

105

(d)

(b)

105

104

103

102

101

0 50 100 150 200Temperature (K)

78

6

4

5

Phot

on f

lux

(cou

nts/

s)

p6 = 3

mission bands centered at (a) 6500, (b) 10 000, (c) 18 200 and (d)

the arrow in Fig. 2(a)) with 100mW and the vertical scales in

(a) Both the low-power limit (� 15mW:K) and the high-power

tions using Eq. (2) for Fig. 3(b) and Eq. (3) for Figs. 3(a), (c), (d)

ARTICLE IN PRESS

4F5/24F7/2

4F3/2

2F5/2

2F7/2

4F9/24I9/24I11/2

4I13/2

4I15/2

4S3/2

2H9/2

2H11/2

Er3+ Yb3+

Energy(103cm-1)

0

5

10

15

20

25

∆E

4I11/2|0⟩

10200

10250

Energy (cm-1)

2F5/2|1⟩

2F5/2|0⟩

Er3+ Yb3+

1

2

3

4

5

6

2

Fig. 4. Schematic energy-level diagram showing luminescent

processes (indicated by a solid arrow, the numbers correspond

to the six emission bands shown in Fig. 1), Yb3þ–Er3þ energy-

transfer UC processes (set of dashed arrows) and multiphonon

relaxation processes on Er3þ (dotted arrows). The inset shows

the lowest two crystal-field levels within the Yb3þ 2F5=2

multiplet as well as the lowest level within the Er3þ 4I11=2multiplet.

J.F. Suyver et al. / Journal of Luminescence 114 (2005) 53–5958

to transfer an excitation to the 4I11=2j0i state (seeFig. 2), this state will luminesce. However, the2F5=2j1i state will efficiently transfer to the

4I11=2j0istate of Er3þ: Based on this, the temperaturedependence of the emission band centered at10 000 cm�1 is determined by the Boltzmannfraction present in the 2F5=2j0i state and thetemperature dependence of the Yb3þ excitationcross-section, sYbðTÞ [14],

I2ðTÞ ¼ I2;0 1� exp �DE

kBT

� �sYbðTÞ; (2)

where DE is the energy difference between the2F5=2j0i and 2F5=2j1i states and I2;0 denotes the10 000 cm�1 intensity at 0K. A comparablemechanism has recently been suggested to explainthe temperature dependence of Tb3þ cross-relaxa-tion [15].Because efficient energy transfer is possible from

2F5=2j1i to 4I11=2j0i; the temperature dependenceof the other five emission bands will be determinedby the Boltzmann fraction present in the 2F5=2j1istate to the power pi (to account for the numberof photons required for excitation of emission

band i), multiplied by ½sYbðTÞ pi

I iðTÞ ¼ I i;1 exp �pi DE

kBT

� ½sYbðTÞ pi ðia2Þ: (3)

Here I i;1 denotes the intensity of the emissionband in the high-temperature limit and DE

remains the energy difference between the2F5=2j0i and 2F5=2j1i states. Since Er3þ emissionfrom band i requires pi excitations of Yb3þ tobegin with, Eq. (3) also depends on sYbðTÞ to thepower pi: Note that Eq. (3) implies a direct and(near) resonant energy transfer from the 2F5=2j1istate to the 4I11=2j0i; and that no phonon-mediatedstep is required. As a result, this process isinsensitive to the phonon density of states in theNaYF4 host-lattice.Eqs. (2) and (3) determine the temperature

dependence of the six emission bands, since DE

(39 cm�1; as found from Fig. 2) and the pi (seeTable 1) are known. The lines through the data inFig. 3(a)–(d) clearly show that the experimentaldata is predicted very well, with the exception ofthe high-power data (squares) of the 6500 cm�1

emission band shown in Fig. 3(a): its temperaturedependence does not correspond to a one-photonprocess but is predicted well using p1 ¼ 2; asshown by the line through the data. Thus the 4I13=2population becomes predominately a two-photonprocess at high excitation power. This effectis attributed to the importance of cross-relaxationin the high-power limit. Due to the large4S3=2 population, the j4S3=2;

4I15=2i ) j4I9=2;4I13=2i

cross-relaxation process becomes important. As aresult, two NIR excitation photons are used topopulate the 4I13=2 state in the high-power limit,while in the low-power limit the state is populatedthrough a non-radiative 4I11=2 !

4I13=2 multipho-non relaxation. Furthermore, this limits theapplicability of Eq. (1), suggesting that the high-power values for Ri (shown in Table 1) maystill be somewhat underestimated. These twocompeting mechanisms for populating the 4I13=2state have further consequences in the excitation-power dependence of the ratio of red to greenemission (as shown in Fig. 2), since 4F9=2 ispopulated from 4I13=2 through ETU. As a result,the 4F9=2 population, and thus the red emission,will become a three-photon process at high

ARTICLE IN PRESS

J.F. Suyver et al. / Journal of Luminescence 114 (2005) 53–59 59

excitation power, changing the red to greenemission ratio.

4. Conclusions

In conclusion, this study provides quantitativedata on the efficiency of near-infrared to visibleUC in NaYF4:Er

3þ;Yb3þ and demonstrates thatthe material is very efficient: in the high-powerlimit over 30% of all photons are emitted in thevisible, which corresponds to � 50% of all near-infrared photons absorbed. The huge temperaturedependence observed for all emission bands isexplained through efficient energy transfer fromthe second crystal field level of the 2F5=2 multipletin Yb3þ to the lowest level of the 4I11=2 multiplet inEr3þ and efficient UC on Er3þ: The 6500 cm�1

emission from Er3þ has a temperature dependencethat changes with excitation power: at highexcitation power, this level is significantly popu-lated through a two-photon cross-relaxation pro-cess.

Acknowledgements

Daniel Biner is gratefully acknowledged forsynthesizing the samples. This work was finan-cially supported by the Swiss National ScienceFoundation.

References

[1] F. Auzel, Chem. Rev. 104 (2004) 139.

[2] M.F. Joubert, Opt. Mater. 11 (1999) 181.

[3] E. Downing, L. Hesselink, J. Ralston, R. Macfarlane,

Science 273 (1996) 1185.

[4] J.S. Chivian, W.E. Case, D.D. Eden, Appl. Phys. Lett. 35

(1979) 124.

[5] D.R. Gamelin, H.U. Gudel, Top. Curr. Chem. 214 (2001)

1 and references therein.

[6] N. Menyuk, K. Dwight, J.W. Pierce, Appl. Phys. Lett. 21

(1972) 159.

[7] T. Kano, H. Yamamoto, Y. Otomo, J. Electrochem. Soc.

119 (1972) 1561.

[8] A. Bril, J.L. Sommerdijk, A.W. de Jager, J. Electrochem.

Soc. 122 (1975) 660.

[9] J.L. Sommerdijk, J. Lumin. 6 (1973) 61.

[10] R.H. Page, K.I. Schaffers, P.A. Waide, J.B. Tassano, S.A.

Payne, W.F. Krupke, W.K. Bischel, J. Opt. Soc. Am. B 15

(1998) 996.

[11] K.W. Kramer, D. Biner, G. Frei, H.U. Gudel, M.P.

Hehlen, S.R. Luthi, Chem. Mater. 16 (2004) 1244.

[12] E. Edjer, J. Opt. Soc. Am. 59 (1969) 223.

[13] J.F. Suyver, A. Aebischer, P. Gerner, S. Garcia-Revilla,

H.U. Gudel, Accepted by Phys. Rev. B (2005).

[14] As no measurement of sYbðTÞ could be performed due to

competing UC processes, the approximation sYbðTÞ /

expð�kBT=EqÞ was used, were Eq denotes an empirical

quenching energy and is used as a fitting parameter. This

form accurately describes the temperature dependence of a

(Lorentzian) excitation cross-section [16]. The lines shown

in Fig. 3(a)–(d) were calculated using Eq ¼ 85 cm�1:[15] T.A. Lawrence, K.A. Murra, P.S. May, J. Phys. Chem.

B 107 (2003) 4002.

[16] B. Henderson, G.F. Imbusch, Optical Spectroscopy of

Inorganic Solids, Clarendon Press, Oxford, 1989.